Rapid development of a robust bioprocess for subunit rotavirus vaccine production in Escherichia coli with the Quality by Design approach

doi:10.21203/rs.3.rs-7125264/v1

Download PDF

Research Article

Rapid development of a robust bioprocess for subunit rotavirus vaccine production in Escherichia coli with the Quality by Design approach

https://doi.org/10.21203/rs.3.rs-7125264/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Rotavirus are leading cause of severe diarrhea and mortality in children less than 5 years old. Vaccination is considered to be the most effective strategy for preventing rotavirus infection. Two live attenuated oral vaccines are licensed in many countries worldwide, with significant reductions in rotavirus-associated mortality and hospitalizations. However, the effectiveness of those vaccines is lower in low- and middle-income countries partly due to the high cost of current vaccines. The spike protein VP4 of rotavirus is the main antigen for inducing neutralizing antibodies and emerges as a promising candidate for cost-effective subunit vaccine against rotaviruses. In this study, we developed a robust bioprocess for VP4 protein production in Escherichia coli with yield higher than 620 mg/L and compliant with the Quality by Design approach. First, the process parameters with potential significant effect on VP4 protein yield were identify based on our experience in virus-like particle vaccine production and screened with a Fractional Factorial Design approach in 1-L parallel bioreactor system. Then, the robust setpoint and design space of the time of induction (TOI), induction temperature (ITmp), the final concentration of IPTG (Con), and speed of feed addition (SOFA) were explored based on a the Central Composite Design approach and criteria of VP4 protein yield > 500 mg/L and probability of failure < 3%. With process parameters set at the robust setpoint, the VP4 protein yield of 685 mg/L was obtained in 1-L bioreactor. Furthermore, the VP4 protein yields with process parameters at the robust setpoint and design space vertexs were higher than 620 mg/L and within the interval of model prediction. This study may serve as a reference for development of a robust and cost-effective subunit rotavirus vaccine production process in Escherichia coli.

Rotavirus

Subunit vaccine

Escherichia coli

Recombinant protein expression

Bioprocess development

Rotaviruses are non-enveloped double-stranded RNA (dsRNA) viruses belonging to the family Reoviridae, with genome encoding 6 structural viral proteins (VP1-4, VP6 and VP7) and 6 non-structural proteins (NSP1-6) [1]. Rotavirus are frequent cause of severe diarrhea among children under five years of age across the globe [2, 3]. Rotavirus induces diarrhea through altering the function of the small intestinal epithelium, including destruction of absorptive enterocytes, intestinal secretion and enteric nervous system activation caused by rotavirus non-structural protein 4 (NSP4) [1]. The morbidity and mortality from diarrhea among children in high- and middle-income countries have greatly decreased, which is attributed to introduction of rotavirus vaccine into their national immunization programs [4]. There are two live attenuated oral vaccines, pentavalent rotavirus vaccine RotaTeq® from Merck & Co., Inc. and mono-valent rotavirus vaccine Rotarix™ from GlaxoSmithKline Biologicals, licensed for use in infants in many countries worldwide. However, it is estimated that rotavirus infections still result in more than one hundred thousand deaths annually in low-income countries [1, 5]. Thus, development of a safe, effective, and cheaper rotavirus vaccine and widespread vaccination are imminent in developing countries, where the burden of disease is highest [4].

Compared with live attenuated vaccine, subunit vaccines have several merits including propensity for minimal or negligible adverse effects, favorable safety profile for immunocompromised individuals [6]. Thus, various approaches to the development of subunit rotavirus vaccines have been proposed. Although the highly immunogenic rotavirus capsid protein VP6, forms the middle of triple layered capsids, is considered as a promising component for subunit rotavirus vaccine [7, 8]. The spike protein VP4, responsible for virus attachment and penetration, is the main antigen for inducing neutralizing antibodies and could be a potential subunit vaccine candidate [9, 10]. Production processes of heterologous recombinant rotavirus capsid proteins expression in microorganism cells are straightforward, rendering them relatively safer and more cost effective than live attenuated vaccine [6]. Escherichia coli is considered as high-yielding and economic platform for protein or vaccine production, owing to its own advantages over the Corynebacterium glutamicum and Pichia pastoris, including the ability to grow in an inexpensive medium, rapid growth rate, and the ease of molecular manipulation. Furthermore, overexpression of rotavirus capsid protein in E. coli still has an advantage in both volumetric and specific productivity [11]. However, low or unstable expression of heterologous proteins in E. coli could be attributed to the process parameters, such as cultivation temperature, inducer concentrations, cell density at the time of induction and culture medium [12, 13]. Thus, a well-characterized and robust manufacturing process for VP4 protein is required to fulfill the increasing demand for subunit rotavirus vaccine in the coming years.

Usually, the upstream process optimization was performed for cell growth and protein yield in a fed-batch fermentation of E. coli where the recombinant protein was being expressed from a plasmid harboring of the protein gene under control of the T7 promoter. Bioproces optimization based on classical one-factor-at-a-time (OFAT) methods with adjusting one parameter at a time is easy to implement but cannot study the interactions between factors and consume the time [14]. Recently, design of experiment (DoE) approach started gaining attention as a efficient tool for optimizing biopharmaceutical production processes by determining the optimal values and design space of the involved factors, which is compliant with quality by design (QbD) approach from the International Conference on Harmonization (ICH) guidance documents [15]. The shake flasks and microtiter plates are widely used as the scale-down model of stirred bioreactor for process development [16]. The throughput screening and optimization are easy achieved by increasing the number of shake flask and microtiter plate used. However, miniaturization of culture scale in shake flasks and microtiter plates impeded access to critical process parameters (e.g., agitation, aeration, pH, etc.) and may resulted in bioprocess deviations during scale-up. Compared to shake flask cultures, cultivation in parallel bioreactor system enables mimic the performance of the large-scale bioreactor and easy scale-up, which is contributing to improving efficiency of bioprocess development [17].

In this study, we developed a robust bioprocess for subunit rotavirus vaccine production in E. coli compliant with the Quality by Design approach. In detail, based on our experience in virus-like particle vaccine production in E. coli, two rounds of DoE approach were conducted to screening and optimization VP4 protein production process parameters in 1-L parallel bioreactor system. Then, the robust setpoint and design space of parameters were explored based on the predictive model and the acceptable criteria. The robust property of optimized bioprocess was verified through comparing experimental and predictive VP4 protein yield under the condition of process parameters set at the robust setpoint and design space vertexs.

2.1. Strain and medium

The recombinant Escherichia coli ER2566, containing plasmid harboring gene encoding the viral protein VP4 of rotavirus genotype P[6], was constructed previously in our laboratory. The LB medium was used for seed culture, which consisted of (g/L): tryptone 10, yeast extract 5, NaCl 5. A modified TB medium was used for VP4 protein production in bioreactor, which contains (g/L): tryptone 20, yeast extract 30, KH₂PO₄ 2.31, K₂HPO₄ 12.55, glycerol 25. The feed medium contained (g/L): yeast extract 50, glycerol 714, (NH4)₂SO₄ 25. The 100 g/L isopropyl-β-d-thiogalactoside (IPTG) solution was used for induction.

2.2. Seed expansion in shake flask

To obtain seeds for fermentation in bioreactor, 45 µL of frozen working cell bank was inoculated into 1-L shake flask containing 300 mL LB medium. Then, the flasks were placed on a rotary shaker at 30°C and 220 rpm for 9 h.

2.3. Bioreactor fermentation condition

Fed-batch cultivations for VP4 protein production were performed in 1-L DASGIP® parallel bioreactor system (Eppendorf, Germany). Forty-two milliliters seed culture was inoculated in bioreactor with 500 mL working volume. The whole cultivation process consisted of two phases, cell growth phase (phase I) and VP4 protein production phase (phase II). The cell growth phase was performed at 42°C with airflow of 0.5 L/min (1 vvm). The pH was set to 7.00 ± 0.10 and controlled by automatic addition of 25% (v/v) ammonia solution or 85% (w/v) phosphoric acid solution. The dissolved oxygen (DO) was maintained by adjusting agitation speed at 400–1200 rpm and adjusting pure oxygen proportion in aeration. The feed medium was added into bioreactor with a peristaltic pump. When a certain biomass concentration was reached, the temperature was decrease. Then the process switched into VP4 protein production phase after IPTG solution was added into bioreactor for induction. When a total cultivation time of 24 h was reached, the bioprocess was stopped and culture broth was collected.

2.4. Establishment of robust VP4 protein production process with the DoE approach

To obtain a robust process for VP4 protein production in E. coli, the DoE approach was performed by using the MODDE® software (Sartorius, Germany). The experimental design consisted of two rounds of DoE: screening and optimization. The parameters and their multilevel values for investigation were described in Table 1. First, a Fractional Factorial Design was used to screen parameters with significant impact on VP4 protein from 8 factors, including cultivation time of seed (CTOS) for seed preparation in shake flask and cultivation temperature (CTmp), DO, start time of feed (STOF), speed of feed addition (SOFA), time of temperature decrease (TOTD), final concentration of IPTG (Con), and upper limit of aeration (ULOV) for bioprocess in bioreactor. Then, a Central Composite Design was used for further optimization, robust setpoint exploration, and design space exploration. As we equipped with three DASGIP® Bioblock modules, up to 12 bioprocesses could be run at same time. Thus, experiment runs more than 12 will be divided into blocks according to the number of experiments required in DoE approach. The VP4 protein yield at harvest was used as the responses.

Table 1
Factors and responses investigated in the DoE approach based on experience
Operation Unit	Parameter	Abbreviations	Unit	Level	experimental setting
Seed preparation (shake flask)	Medium	--	--	LB medium	fixed
	Cultivation time of seed	CTOS	h	6, 9, 12	screening
	Temperature	--	^oC	37 ± 0.2	fixed
	pH	--	--	7.00 ± 0.10	fixed
	Rotation speed	--	rpm	220 ± 10	fixed
Bioprocess (Bioreactor)	Inoculum density	--	OD₆₀₀	0.40	fixed
	Yeast extract lot number	--	--	230220-09	fixed
	Cultivation temperature	CTmp	^oC	32, 37, 42	screening
	pH	pH	--	7.00 ± 0.10	fixed
	Dissolved oxygen	DO	%	10, 35, 60	screening
	Start time of feed	STOF	h	4.5, 6.0, 7.5	screening
	Speed of feed addition	SOFA	mL/h	0.875, 1.750, 2.625	screening
	Speed of feed addition	SOFA	mL/h	0.875, 1.750, 2.625, 3.500	optimization
	Time of temperature decrease	TOTD	OD₆₀₀	7.5, 12.5, 17.5	screening
	Time of induction	TOI	OD₆₀₀	7.5, 12.5, 20.0, 27.5, 35.0	optimization
	Induction temperature	ITmp	^oC	15.0, 17.5, 20.0, 22.5, 25.0	optimization
	Final concentration of IPTG	Con	µM	25, 50, 75	screening
	Final concentration of IPTG	Con	µM	12.5, 25.0, 37.5, 50	optimization
	Upper limit of aeration	ULOA	VVM	1.0, 1.5, 2.0	screening
	Cultivation duration	--	h	24	fixed

2.5. Measurements of biomass and metabolites

The cell growth was monitored by measuring the optical density of culture broth at a wavelength of 600 nm (OD₆₀₀) using a Biomate™ 160 spectrophotometer (Thermo, USA). To estimate the WCW, 1 mL of culture broth was centrifuged at 17,000 g at 4°C for 5 min. After centrifugation the supernatant was removed and the WCW was measured by calculating the difference in weight between the tube before and after sample addition.

Samples of culture broth were taken from bioreactor and analyzed for glycerol, phosphoric acid, acetic acid, and NH4⁺ by using Cedex Bio analyzer (Roche Diagnostics, Switzerland).

2.6. Assay of VP4 protein concentration

The culture broth was centrifuged at 17,000 g for 15 min using a Velocity 18R Pro centrifug (Dynamica, UK). The pellets were harvested and resuspended in cell lysis solution. Then, the cell suspensions were lysed using a Multi-Channel Ultrasonic Homogenizer (Scientz, China). Lastly, the resulting solutions were centrifuged at 17,000 g for 15 min and the supernatant were harvested for VP4 protein concentration analysis. VP4 protein concentration was measured through ELISA according to the experiment protocol established in our laboratory.

3.1. Preliminary characterization of VP4 protein production bioprocess

The characterization of E. coli-based VP4 protein production bioprocess is necessary to identify and understand the potential implications of process parameters on VP4 protein yield. More than 10 years of experience in extensive application of E. coli expression system for virus-like particle vaccine production helped us to preliminary identify and classify factors affected VP4 protein production. As shown in Table 1, the small-scale VP4 protein production bioprocess consist of seed preparation in shake flask and E. coli fermentation in bioreactor. For seed preparation, the E. coli was cultured in an LB medium with pH of 7.00 at 37^oC and 220 rpm. The cultivation time of seed, which affects bacterial metabolism and growth rate, was chosen for further investigation using screening experiment. Based on our history VP4 protein production data record, the inoculum density at range of 0.3–0.5 OD₆₀₀ and pH at range of 6.5–7.5 in bioreactor have no significant effect on VP4 protein yield in E. coli. However, the lot-to-lot variations in yeast extract could result in up to a 40% difference in VP4 protein yield (data not shown), thus yeast extract lot number was fixed throughout the entire experiment of this study. In order to efficiently maintain our current process turnover, the total cultivation duration was kept at 24 h, even if further extending the cultivation duration may increase VP4 protein yield. Therefore, for E. coli fermentation in bioreactor, the bioreactor containing TB medium with pH of 7.00 was inoculated with a density of 0.40 OD₆₀₀. The remaining 8 parameters in Table 1 were selected for further investigation using screening experiment.

3.2. Determination of critical bioprocess parameters of VP4 protein production

To efficiently identify significant parameters showing a high probability of impacting VP4 protein yield from 8 potential variables, a Fractional Factorial Design was used as screening approach. The VP4 protein yield at 24 h of each run were collected (Table 2) and used to establish a mathematical model between parameters and VP4 protein yield. Model analyses showed the model well explained and predicted the response variations with R² = 0.893 (adjusted R² = 0.818) and Q² = 0.556, respectively. The model, with a p-value of 0.319, had no lack of fit. The coefficients and analysis of variance (ANOVA) results of the model are shown in Table 3. The p-value represents the probability of model coefficient has an insignificant effect on the response. Thus, the start time of feed (STOF), final concentration of IPTG (Con), and speed of feed addition (SOFA) with p value lower than 0.05 are significant parameter. The start time of feed (STOF) and final concentration of IPTG (Con) had negative effect on VP4 protein yield within their level range. The speed of feed addition (SOFA) had positive effect on adenovirus VP4 protein yield within their level range. Furthermore, the interaction of cultivation time of seed (CTOS) and upper limit of aeration (ULOV) often have significant effect on the response, although CTOS and ULOV are not significant for the model.

Table 2
Screening design matrix and data of the Fractional Factorial Design
No.	CTOS (h)	CTmp (^oC)	DO (%)	STOF (h)	TOTD (OD₆₀₀)	Con (µM)	SOFA (mL/h)	ULOV (VVM)	Block	Yield (mg/L)
1	9	37	35	6.0	12.5	50	1.750	1.5	1	374.5
2	6	32	60	4.5	17.5	75	0.875	2.0	1	352.0
3	6	32	60	7.5	17.5	25	2.625	1.0	1	435.0
4	6	32	10	7.5	7.5	75	2.625	2.0	1	335.5
5	12	42	10	7.5	7.5	25	2.625	1.0	1	468.5
6	12	42	10	4.5	7.5	75	0.875	2.0	1	298.5
7	9	37	35	6.0	12.5	50	1.750	1.5	1	348.5
8	12	42	60	4.5	17.5	25	0.875	1.0	1	407.5
9	12	42	60	7.5	17.5	75	2.625	2.0	1	317.0
10	9	37	35	6.0	12.5	50	1.750	1.5	1	317.5
11	6	32	10	4.5	7.5	25	0.875	1.0	1	363.0
12	6	42	60	4.5	7.5	75	2.625	1.0	2	389.0
13	6	42	10	7.5	17.5	75	0.875	1.0	2	240.0
14	9	37	35	6.0	12.5	50	1.750	1.5	2	351.5
15	12	32	10	4.5	17.5	75	2.625	1.0	2	424.0
16	9	37	35	6.0	12.5	50	1.750	1.5	2	337.5
17	12	32	60	7.5	7.5	75	0.875	1.0	2	288.0
18	6	42	60	7.5	7.5	25	0.875	2.0	2	341.0
19	9	37	35	6.0	12.5	50	1.750	1.5	2	352.5
20	6	42	10	4.5	17.5	25	2.625	2.0	2	567.5
21	12	32	10	7.5	17.5	25	0.875	2.0	2	326.5
22	12	32	60	4.5	7.5	25	2.625	2.0	2	421.5
23	9	37	35	6.0	12.5	50	1.750	1.5	2	389.0

Table 3
Coefficients and ANOVA results for Fractional Factorial Design model
Terms	Coefficient (scaled and centered)
Constant	367.20	< 0.0001
CTOS	-4.47	0.5472
CTmp	5.22	0.4832
DO	-4.53	0.5417
STOF	-29.47	0.0013
TOTD	10.28	0.1786
Con	-42.91	< 0.0001
SOFA	46.34	< 0.0001
ULOV	-3.47	0.6394
CTOS * ULOV	-24.59	0.0047

3.3. Fitting of mathematical model for VP4 protein production

To further optimize the bioprocess performance, the final concentration of IPTG (Con), speed of feed addition (SOFA), the time of induction (TOI), and induction temperature (ITmp) were investigated using a the Central Composite Design. Based on the data in Table 4, the quantitative relationships between VP4 protein yield and Con, SOFA, TOI, and ITmp were built by multiple linear regression:

Table 4
Experimental matrix and data of the Central Composite Design
No.	TOI (OD₆₀₀)	ITmp (^oC)	Con (µM)	SOFA (mL/h)	Block	Yield (mg/L)
1	20.0	20.0	25.0	1.750	1	568.3
2	27.5	17.5	37.5	0.875	1	293.8
3	12.5	22.5	37.5	0.875	1	491.3
4	12.5	17.5	37.5	2.625	1	407.1
5	20.0	20.0	25.0	1.750	1	469.2
6	12.5	22.5	12.5	2.625	1	539.6
7	20.0	20.0	25.0	1.750	1	506.3
8	27.5	17.5	12.5	2.625	1	97.5
9	27.5	22.5	12.5	0.875	1	325.4
10	27.5	22.5	37.5	2.625	1	448.3
11	20.0	20.0	25.0	1.750	1	538.3
12	12.5	17.5	12.5	0.875	1	384.6
13	12.5	17.5	37.5	0.875	2	314.6
14	20.0	20.0	25.0	1.750	2	442.9
15	27.5	22.5	37.5	0.875	2	272.1
16	12.5	22.5	37.5	2.625	2	550.0
17	12.5	17.5	12.5	2.625	2	159.2
18	27.5	17.5	12.5	0.875	2	256.7
19	12.5	22.5	12.5	0.875	2	384.6
20	20.0	20.0	25.0	1.750	2	486.7
21	27.5	22.5	12.5	2.625	2	433.8
22	27.5	17.5	37.5	2.625	2	276.3
23	20.0	20.0	25.0	1.750	2	440.0
24	20.0	20.0	50.0	1.750	3	432.1
25	20.0	20.0	25.0	1.750	3	542.1
26	20.0	20.0	25.0	3.500	3	505.8
27	7.5	20.0	25.0	1.750	3	538.3
28	20.0	20.0	25.0	1.750	3	440.4
29	35.0	20.0	25.0	1.750	3	390.8
30	20.0	20.0	25.0	1.750	3	512.5
31	20.0	15.0	25.0	1.750	3	135.0
32	20.0	25.0	25.0	1.750	3	432.5

$$\:\text{Y}\text{i}\text{e}\text{l}\text{d}=-3429.7-6.5\bullet\:\text{T}\text{O}\text{I}+378.2\bullet\:\text{I}\text{T}\text{m}\text{p}+27.3\bullet\:\text{C}\text{o}\text{n}-509.0\bullet\:\text{S}\text{O}\text{F}\text{A}-9.3\bullet\:{\text{I}\text{T}\text{m}\text{p}}^{2}-0.3\bullet\:{\text{C}\text{o}\text{n}}^{2}-0.6\bullet\:\text{I}\text{T}\text{m}\text{p}\bullet\:\text{C}\text{o}\text{n}+23.1\bullet\:\text{I}\text{T}\text{m}\text{p}\bullet\:\text{S}\text{O}\text{F}\text{A}+2.5\bullet\:\text{C}\text{o}\text{n}\bullet\:\text{S}\text{O}\text{F}\text{A}$$

Model analyses are shown in Fig. 1 and Table 5. The model well explained the response variations with R² = 0.866 (adjusted R² = 0.811) and Q² = 0.598 (Fig. 1a), respectively. The model with a p-value of 0.175 indicated no lack of fit was found. The experiments were shown in the replicate plot (Fig. 1b) colored by block: green, block 1; yellow, block 2; red, block 3. The replicate plot indicated the variability of the replicates (square symbols connected by a line) in each block was small. The residual plot shows points are inside +/- 4 standard deviations (4 SD) and closed to a straight line, indicating the residual are normally distributed. The coefficients and their ANOVA results of the model are shown in Table 5. The time of induction (TOI) and induction temperature (ITmp) are significant model term. There’s a trend that decreasing the time of induction within the range of 7.5–35.0 OD₆₀₀ and increasing induction temperature within the range of 15.0–25.0°C will increase VP4 protein yield. Furthermore, the quadratic effect of final concentration of IPTG (Con) and the interaction effect between induction temperature (ITmp) and speed of feed addition (SOFA) are significant model term as well, which indicates that those two parameters act dependently upon the VP4 protein yield.

Table 5
Coefficients and ANOVA results for Central Composite Design model
Terms	Coefficient (scaled and centered)	P value
Constant	490.53	< 0.0001
TOI	-88.71	0.0003
ITmp	185.00	< 0.0001
Con	3.84	0.8783
SOFA	38.43	0.0822
ITmp * ITmp	-232.44	< 0.0001
Con * Con	-112.17	0.0011
ITmp * Con	-59.06	0.1635
ITmp * SOFA	151.31	0.0012
Con * SOFA	60.75	0.0607

3.4. Determination of the robust setpoint and design space

To establish a robust bioprocess for VP4 protein production in E. coli compliant with quality by design (QbD) conception, the robust setpoint and design space of parameters were explored based on the predictive model described above and the criteria of VP4 protein yield > 500 mg/L and probability of failure < 3%. Monte Carlo simulations of 65,536 runs were performed to find robust setpoint. The design space with a robust setpoint for VP4 protein production is shown in Fig. 1d. The robust setpoint of parameters were determined to be time of induction (TOI) of 11.2 OD₆₀₀, induction temperature (ITmp) of 22.5°C, the final concentration of IPTG (Con) of 30.6 µM, and speed of feed addition (SOFA) of 3.11 mL/h. The design space hypercube of parameters was: time of induction (TOI) of 7.5–13.0 OD₆₀₀, induction temperature (ITmp) of 21.7–23.0°C, the final concentration of IPTG (Con) of 27.5–35.0 µM, and speed of feed addition (SOFA) of 2.98–3.50 mL/h.

3.5. Verification of the robust of optimized bioprocess

To verify the robust property of optimized bioprocess, the predicted robust setpoint and design space vertexs of VP4 protein production parameters were tested in bioreactors with 500 mL working volume, respectively (Fig. 2 and Table S1). The DASware® software suite enables precise monitoring and control of temperature, pH, DO, and agitation speed. The bioreactor parameter profiles and metabolite profiles during the VP4 protein production with process parameters at the robust setpoint are shown in Fig. 2. The bioprocess started at fixed agitation speed of 400 rpm and the DO was well controlled at 30% through automatic adjustment of agitation speed within 400–1200 rpm and oxygen supplementation. The initial cultivation temperature was set at 42.0°C and changed to 22.5°C when the OD₆₀₀ of 7.5 was reached at 2.65 h. When the OD₆₀₀ reached 11.2 at 3.75 h, IPTG solution was added to introduce protein expression and the bioprocess shifted to VP4 protein production phase (phase II). Despite addition of feed medium started at 4.5 h with a speed of 3.11 mL/h, the glycerol concentration decreased with the culture time prolongs and the depletion of glycerol was observed at the time of harvest. Furthermore, the concentration of NH₄⁺, phosphoric acid, and acetic acid decreased with the increasing of cell density during VP4 protein production phase. The cell biomass concentration reached 120.2 OD₆₀₀ (176.9 g/L WCW) at the time of harvest. With process parameters set at the robust setpoint, the VP4 protein yield of 685 mg/L was obtained, which was 1.64-fold higher than that of control condition (the initial process). Furthermore, the VP4 protein yield in bioreactor with process parameters at the robust setpoint and design space vertexs were higher than 620 mg/L and within the interval of model prediction (Fig. 3 and Table S1). Thus, a robust VP4 protein production bioprocess was developed successfully with yield increased.

Rotavirus are frequent cause of severe diarrhea and mortality in children less than 5 years old. The live oral rotavirus vaccines, developed from strains pathogenic attenuated by serial passage in cell culture, remain the most efficacious and widely utilized strategy for preventing rotavirus infection [18]. However, low protective (approximately 60%) and coverage (approximately 70%) of live oral rotavirus vaccines in low-income countries resulted in high under 5 years mortality [19]. Thus, the inadequate efficacy of live oral vaccines in those low-income countries underscores the need for developing new vaccines that avoid the oral route. New parenteral rotavirus vaccines in development including non-replicating rotavirus vaccine (NRRV), inactivated RV vaccine, and subunit rotavirus vaccine [7, 18]. Among them, the subunit vaccine exhibits a property of minimal or negligible adverse effects, favorable safety profile for immunocompromised individuals, and straightforward production processes [21]. Although the rotavirus capsid protein VP6, which is highly immunogenic and contains conserved cross-reacting epitopes, is considered as an attractive candidate for subunit rotavirus vaccine [7, 8] and could induce protection against rotavirus infection in a murine model [20]. However, the rotavirus VP4 protein is one of the pivotal factors in orchestrating protective immunity, making it a more promising candidate for subunit rotavirus vaccine [6, 9, 10].

To date, E. coli is still the most widely used host for industrial production of heterologous proteins. Previous reports indicated that expression of VP6 proteins in E. coli is considered to be the most convenient and cost-effective approach for subunit rotavirus vaccine production [8, 11]. However, despite the outstanding advantages of E. coli platform, VP6 protein produced by E. coli was usually insoluble [8]. Thus, optimization approaches at expression level and process level should be implemented to improve recombinant protein expression yield in soluble form [14]. Several technological advances at the level of gene expression have widely improved recombinant protein expression in E. coli, including optimization of host strain, expression vector, promoter, codon usage, mRNA structure, fusion tags, and secretion peptides [21, 22]. The recombinant E. coli ER2566 contained gene of viral protein VP4 of rotavirus genotype P[6] was constructed previously in our laboratory. Thus, in this study, we mainly focus on optimization of bioprocess conditions to achieve higher PV4 protein yield in E. coli, including cultivation temperature, IPTG concentration, cell density at the time of induction, and feeding strategy.

Bioprocesses are complex nonlinear biochemical processes that occur in living cells and development of recombinant protein production bioprocess is generally labor-intensive and time-consuming as various process parameters (medium, temperature, pH, agitation, aeration, etc.) influence the accumulation of the protein. To develop an efficient and economic bioprocess, the optimal condition and interaction effect of critical parameter should be evaluated in parallel. In order to implement multiple parameters optimization and achieve efficient VP4 protein production process, high-throughput platforms and experimental approaches are essential. The merit of easy to handle, cost-effective, and high parallelization capability makes shake flasks and microtiter plates more suitable for the screening of culture medium, bacterial strains, and optimization of preliminary process parameters for E. coli [16, 23]. Furthermore, microtiter plate integrated with fluorescence-based pH and DO sensor sports features online measurements of pH, DO, and biomass concentration when cultured in commercial BioLector® microbioreactor (m2p-labs GmbH, Germany) [24, 25]. However, differences in geometric structure, gas mass transfer, and shear force resulted in the flask and microtiter plate cultivations are not completely comparable to bioreactor scales. Moreover, insufficient oxygen supply when cell cultured at high cell density in flask or microtiter plate may lead to altered metabolism and reduce recombinant protein yield. The commercial miniaturised and automated parallel bioreactor systems with several stirred-tank bioreactors on a mL-scale or L-scale, such as Ambr® multi-parallel bioreactors from Sartorius AG and DASGIP® parallel bioreactor system from Eppendorf Corporate, is contributing to reduction of bioprocess development time and costs, in particular due to their intrinsic capability for high throughput and property of mimic the performance of the large-scale bioreactor.

E. coli is the most applied organism for the production of recombinant proteins and their protocols for obtaining recombinant proteins are pretty straightforward [26]. After cloning gene of interest into expression vector under the promoter and transforming it into the E. coli host, the cultivation process of such a engineered strain could be divided into two phases, cell growth phase and recombinant protein production phase. As the efficient expression of recombinant protein in E. coli is a combination of optimal cell growth and transcription condition [14, 27], thus parameters from operation unit of seed preparation and E. coli fermentation, which had potential significant effect on VP4 protein yield, were evaluated in this study. In order to maximize the information between process parameters and VP4 protein yield while minimizing the number of experiment, combining parallel bioreactor system with statistical DoE method is a powerful and reasonable approach, which has the advantage of reducing bioprocess development time and costs. The DoE method is a frequently utilized statistical technique for bioprocess optimization of recombinant protein expression in E. coli, which have been found to be more effective and reasonable than the traditional OFAT method [14, 28, 29]. Although we have established a technical platform for the production of VP4 protein in E. coli based on our experience and preliminary process optimization, in order to further improve VP4 protein yield and robustness of such a bioprocess, we implemented a DoE driven QbD approach to facilitate understanding of the relationship between process parameters and VP4 protein yield.

Since VP4 protein gene under control of the T7 promoter, the inducible fed-batch fermentation process of E. coli was conducted to achieve a high biomass density and fast production of the highest possible amount of VP4 protein. With the help of parallel bioreactor systems, the effect of 7 parameters on VP4 protein yield were evaluated simultaneously at bioreactor level in this study. Although the implementation of DoE approach resulted in more than 20 experiments, the parallel bioreactor system enabled the fermentation process control more precise and the data acquisition more efficient. The DoE model analysis showed that the start time of feed (STOF), speed of feed addition (SOFA), and final concentration of IPTG (Con) are significant parameter on VP4 protein yield with p value lower than 0.05. Feeding strategy is an important factor for recombinant protein expression in E. coli, as the start time and speed of feeding have an impact on nutrients supplementation, cell growth rate, and protein biosynthesis [30, 31]. In this study, the model analysis results showed the start time of feed (STOF) and the speed of feed addition (SOFA) had negative and positive effect on VP4 protein yield, respectively, indicating that increasing nutrients supplementation resulted in VP4 protein yield increased. After inoculation and further growth, the IPTG should be added into the culture at a proper time and dosage to induce protein transcription and translation. The model analysis results showed the time of induction (TOI) and induction temperature (ITmp) are significant model terms and exhibited a tendency that the VP4 protein yield increased with the increasing the induction temperature and decreasing the time of induction. Overall, by implementing DoE approach, the feeding and induction condition were optimized to balance cell growth and protein expression, increasing the VP4 protein yield in E. coli. The VP4 protein yield of 685 mg/L was obtained in 1-L bioreactor with the optimized process, which was 1.64-fold higher than the initial process.

Most reports use DoE approach for optimizing the production process of recombinant proteins in E. coli to determine optimal parameter settings and increase protein yield, but considerations for process robustness are often ignored. The QbD concept, which emphasizes enhancing process understanding and designing a robust manufacturing process to consistently deliver the desired product quality, has received significant attention in biopharmaceutical industry due to its potential to improve the efficacy and quality of pharmaceutical products [32, 33]. ICH guidance Q8 Pharmaceutical Development defines the design space as “the multidimensional combination and interaction of input variables and process parameters that have been demonstrated to provide assurance of quality”. The process parameters working within the design space is not considered as a change and do not trigger a regulatory post approval procedure. In this study, the establishment of a robust process for subunit rotavirus vaccine with productivity increased was be performed based on the QbD concept. In detail, the correlations between process parameters and VP4 protein yield were evaluated using DoE methods and design space of process parameters was further explored based on the DoE result and mathematical model. Finally, the robust property of optimized bioprocess was verified by confirming VP4 protein yield with process parameters at the design space vertexs. Due to the advantage of rapid growth rate of E. coli, the establishment of design space of process parameters makes the VP4 protein production process more flexible.

In this study, a robust bioprocess for VP4 protein production in E. coli compliant with the Quality by Design approach was developed in 1-L parallel bioreactor system. Based on our experience in virus-like particle vaccine production, we first screened the process parameters which has potential significant effect on VP4 protein yield with a Fractional Factorial Design approach. Then, the time of induction (TOI), induction temperature (ITmp), the final concentration of IPTG (Con), and speed of feed addition (SOFA) were optimized based on a Central Composite Design approach and the robust setpoint and design space of those parameters were explored was well. Furthermore, the robust property of optimized bioprocess was verified by confirming VP4 protein yield with process parameters at robust setpoint and the design space vertexs in 1-L bioreactors, yielding VP4 protein yield higher than 620 mg/L and within the interval of model prediction. This study presents the efficient development of a well-characterized and robust subunit rotavirus vaccine manufacturing process in E. coli with the implementation of DoE driven QbD approach.

Acknowledgments

This work was supported by National Natural Science Foundation of China (No. 22478157), Research and Development of Innovative Solutions for Social Development in Jiangsu Province (BE2022694), the national first-class discipline program of Light Industry Technology and Engineering, China, (LITE2018-24).

Conflict of interests

The authors declare that there is no conflict of interest.

Author Contributions

Daning Wang: methodology, investigation, software, writing-original draft. Minming Chen: conceptualization, investigation, validation, visualization, data curation. Junyi Lin: methodology, data curation, visualization, software. Guoxing Luo: validation, visualization, data curation. Jianqi Nie: methodology, conceptualization, writing-original draft. Zhonghu Bai: conceptualization, writing-review and editing, project administration, supervision.

Data availability

The data that support the findings of this study are available on request from the corresponding author upon reasonable request..

Crawford, S.E., et al. (2017) Rotavirus infection. Nature Reviews Disease Primers. 3: 17083. DOI: doi: 10.1038/nrdp.2017.83.
Cai, L., et al. (2025) Disease burden of rotavirus related diarrhea in children under 5 years in China: a meta-analysis. Scientific Reports. 15(1): 15973. DOI: 10.1038/s41598-025-00778-w.
Black, R.E., et al. (2024) Estimated global and regional causes of deaths from diarrhoea in children younger than 5 years during 2000–21: a systematic review and Bayesian multinomial analysis. The Lancet Global Health. 12(6): e919-e928. DOI: 10.1016/S2214-109X(24)00078-0.
Tate, J.E., et al. (2012) Remaining issues and challenges for rotavirus vaccine in preventing global childhood diarrheal morbidity andmortality. Expert Review of Vaccines. 11(2): 211-220. DOI: 10.1586/erv.11.184.
Arakaki, L., et al. (2021) Prevalence of rotavirus among older children and adults with diarrhea: A systematic review and meta-analysis. Vaccine. 39(33): 4577-4590. DOI: 10.1016/j.vaccine.2021.06.073.
Kuri, P.R. and P. Goswami (2024) Reverse vaccinology-based multi-epitope vaccine design against Indian group A rotavirus targeting VP7, VP4, and VP6 proteins. Microbial Pathogenesis. 193: 106775. DOI: 10.1016/j.micpath.2024.106775.
Kirkwood, C.D., et al. (2019) The rotavirus vaccine development pipeline. Vaccine. 37(50): 7328-7335. DOI: 10.1016/j.vaccine.2017.03.076.
Kuri, P.R. and P. Goswami (2024) Unravelling aggregation propensity of rotavirus A VP6 expressed as E. coli inclusion bodies through in silico prediction. Scientific Reports. 14(1): 21464. DOI: 10.1038/s41598-024-69896-1.
Huang, Y., et al. (2025) A single residue switch mediates the broad neutralization of Rotaviruses. Nature Communications. 16: 838. DOI: 10.1038/s41467-025-56114-3.
Zeng, Y., et al. (2022) Generation and characterization of mouse monoclonal antibodies against the VP4 protein of group A human rotaviruses. Antiviral Research. 207: 105407. DOI: 10.1016/j.antiviral.2022.105407.
Bredell, H., et al. (2016) Expression of rotavirus VP6 protein: a comparison amongstEscherichia coli, Pichia pastorisandHansenula polymorpha. FEMS Yeast Research. 16(2): fow001. DOI: 10.1093/femsyr/fow001.
García-Fraga, B., et al. (2015) Optimized expression conditions for enhancing production of two recombinant chitinolytic enzymes from different prokaryote domains. Bioprocess and Biosystems Engineering. 38(12): 2477-2486. DOI: 10.1007/s00449-015-1485-5.
Kaur, J., A. Kumar, and J. Kaur (2018) Strategies for optimization of heterologous protein expression in E. coli: Roadblocks and reinforcements. International Journal of Biological Macromolecules. 106: 803-822. DOI: 10.1016/j.ijbiomac.2017.08.080.
Packiam, K.A.R., et al. (2020) Stepwise optimization of recombinant protein production in Escherichia coli utilizing computational and experimental approaches. Applied Microbiology and Biotechnology. 104(8): 3253-3266. DOI: 10.1007/s00253-020-10454-w.
Anurag S Rathore, H.W. (2009) Quality by design for biopharmaceuticals. Nature Biotechnology. 27: 26-34. DOI: 10.1038/nbt0109-26.
Mühlmann, M., et al. (2017) Optimizing recombinant protein expression via automated induction profiling in microtiter plates at different temperatures. Microbial Cell Factories. 16(1): 220. DOI: 10.1186/s12934-017-0832-4.
Teworte, S., et al. (2022) Recent advances in fed-batch microscale bioreactor design. Biotechnology Advances. 55: 107888. DOI: 10.1016/j.biotechadv.2021.107888.
Glass, R.I., et al. (2021) The Rotavirus Vaccine Story: From Discovery to the Eventual Control of Rotavirus Disease. The Journal of Infectious Diseases. 224(Supplement_4): S331-S342. DOI: 10.1093/infdis/jiaa598.
Glass, R.I., B. Jiang, and U. Parashar (2018) The future control of rotavirus disease: Can live oral vaccines alone solve the rotavirus problem? Vaccine. 36(17): 2233-2236. DOI: 10.1016/j.vaccine.2018.03.008.
Ho, P.L., et al. (2018) Rotavirus VP6 protein mucosally delivered by cell wall-derived particles from Lactococcus lactis induces protection against infection in a murine model. Plos One. 13(9): e0203700. DOI: 10.1371/journal.pone.0203700.
Jiang, R., et al. (2024) Strategies to overcome the challenges of low or no expression of heterologous proteins in Escherichia coli. Biotechnology Advances. 75: 108417. DOI: 10.1016/j.biotechadv.2024.108417.
Arcovito, A., et al. (2014) Synthesis and characterization of different immunogenic viral nanoconstructs from rotavirus VP6 inner capsid protein. International Journal of Nanomedicine: 2727. DOI: 10.2147/IJN.S60014.
Diederichs, S., et al. (2014) Phenotyping the quality of complex medium components by simple online-monitored shake flask experiments. Microbial Cell Factories. 13(1): 149. DOI: 10.1186/s12934-014-0149-5.
Mühlmann, M.J., et al. (2018) Prediction of recombinant protein production by Escherichia coli derived online from indicators of metabolic burden. Biotechnology Progress. 34(6): 1543-1552. DOI: 10.1002/btpr.2704.
Kensy, F., C. Engelbrecht, and J. Büchs (2009) Scale-up from microtiter plate to laboratory fermenter: evaluation by online monitoring techniques of growth and protein expression in Escherichia coli and Hansenula polymorpha fermentations. Microbial Cell Factories. 8(1): 68. DOI: 10.1186/1475-2859-8-68.
Rosano, G.n.L. and E.A. Ceccarelli (2014) Recombinant protein expression in Escherichia coli: advances and challenges. Frontiers in Microbiology. 5: 172. DOI: 10.3389/fmicb.2014.00172.
Lozano Terol, G., et al. (2021) Impact of the Expression System on Recombinant Protein Production in Escherichia coli BL21. Frontiers in Microbiology. 12: 682001. DOI: 10.3389/fmicb.2021.682001.
Hajihassan, Z., A. Yaseri, and M. Yazdi (2025) Optimization of recombinant neurturin expression in Escherichia coli using response surface methodology. Biotechnology Letters. 47: 36. DOI: 10.1007/s10529-025-03575-7.
Agbogbo, F.K., et al. (2020) Upstream development ofEscherichia colifermentation process withPhoApromoter using design of experiments (DoE). Journal of Industrial Microbiology and Biotechnology. 47: 789-799. DOI: 10.1007/s10295-020-02302-7.
Mahmoodi, M. and E. Nassireslami (2021) Control algorithms and strategies of feeding for fed-batch fermentation ofEscherichia coli: a review of 40 years of experience. Preparative Biochemistry & Biotechnology. 52(7): 823-834. DOI: 10.1080/10826068.2021.1998112.
Babaeipour, V., et al. (2010) A proposed feeding strategy for the overproduction of recombinant proteins in Escherichia coli. Biotechnology and Applied Biochemistry. 49(2): 141-147. DOI: 10.1042/BA20070089.
Yu, L.X., et al. (2014) Understanding Pharmaceutical Quality by Design. The AAPS Journal. 16(4): 771-783. DOI: 10.1208/s12248-014-9598-3.
Duarte, J.G., et al. (2025) Rethinking Pharmaceutical Industry with Quality by Design: Application in Research, Development, Manufacturing, and Quality Assurance. The AAPS Journal. 27(4): 96. DOI: 10.1208/s12248-025-01079-w.

No competing interests reported.

SupplementarymaterialTableS1.docx

Download PDF

Editorial decision: Revision requested
08 Aug, 2025
Reviews received at journal
08 Aug, 2025
Reviews received at journal
04 Aug, 2025
Reviewers agreed at journal
27 Jul, 2025
Reviewers agreed at journal
27 Jul, 2025
Reviewers invited by journal
24 Jul, 2025
Editor assigned by journal
15 Jul, 2025
Submission checks completed at journal
15 Jul, 2025
First submitted to journal
14 Jul, 2025

You are reading this latest preprint version

Rapid development of a robust bioprocess for subunit rotavirus vaccine production in Escherichia coli with the Quality by Design approach

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials and methods

2.1. Strain and medium

2.2. Seed expansion in shake flask

2.3. Bioreactor fermentation condition

2.4. Establishment of robust VP4 protein production process with the DoE approach

2.5. Measurements of biomass and metabolites

2.6. Assay of VP4 protein concentration

3. Results

3.1. Preliminary characterization of VP4 protein production bioprocess

3.2. Determination of critical bioprocess parameters of VP4 protein production

3.3. Fitting of mathematical model for VP4 protein production

3.4. Determination of the robust setpoint and design space

3.5. Verification of the robust of optimized bioprocess

4. Discussion

5. Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1