The gene encoding the HpaA protein was derived from H. pylori 26,695 (GenBank accession number: AE000511.1) and optimized for E. coli codon usage. The hpaA gene, synthesized without a signal peptide by GeneCreate Biotechnology Company (Wuhan, China) was inserted into the NcoI and XhoI sites of the pET28a(+) vector, yielding the recombinant plasmid. The plasmid was transformed into E. coli BL21 (DE3) for protein expression. For long-term preservation, the recombinant bacterial strains were stored at −80°C in a solution containing 8% (v/v) glycerol.

Luria Bertani (LB) medium, composed of 5 g/L yeast extract, 10 g/L tryptone, and 10 g/L NaCl was employed for both strain revival and liquid seed culture preparation. The glycerol stock of the recombinant E. coli BL21 (DE3) harboring the rhpaA/pET28a(+) plasmid was revived on LB agar plates at 37 °C for 16 h. A single colony was selected, inoculated into 10 mL of LB broth containing 100 μg/mL ampicillin, and incubated at 37°C and 220 rpm on a rotary shaker (Mingquan, Shanghai, China) overnight. Subsequently, 1 mL of the seed culture was diluted into 100 mL of induction media at a 1% (v/v) ratio and incubated under the same conditions until the optical density at 600 nm (OD₆₀₀) reached 0.6–0.8. The expression of the rHpaA protein was induced by the addition of isopropyl-β-D-thiogalactoside (IPTG) to a final concentration of 0.5 mmol/L. After a 4-h induction period, the cells were harvested by centrifugation at 12,000 g for 5 min.

The collected induced cells were suspended in phosphate-buffered saline (PBS) and sonicated. The total protein content of the lysate was quantified using the Lowry method (Lowry et al., 1951). Equal amounts of 10 μg protein were loaded for analysis by 12% sodium dodecyl sulfate‒polyacrylamide gel electrophoresis (SDS‒PAGE). After electrophoresis, the gel was stained with 0.1% Coomassie Brilliant Blue R-250 and visualized with a ChemiDoc™ MP Imaging System (Bio-Rad Laboratories, California, USA). The band areas in the SDS‒PAGE gel were quantified with ImageLab 6.1 software (Bio-Rad Laboratories, California, USA). The rHpaA yield was determined by multiplying the total protein concentration of the sample by the percentage signal intensity of rHpaA.

The scale of the rHpaA/pET28a(+)/E. coli BL21 (DE3) cultures was increased to 1 L under the optimized culture conditions. The cells were harvested and suspended at a ratio of wet cell weight (g) to buffer volume (mL) of 1:20 in ice-cold 50 mmol/L phosphate buffer (PB, pH 7.0) supplemented with 500 mmol/L NaCl, 10 mmol/L imidazole, 100 U of benzonase and 200 mmol/L MgCl₂. After high-pressure homogenization, the lysate was centrifuged at 12,000 g for 30 min. The supernatant was collected and applied to a pre-equilibrated Ni-Sepharose affinity column (Cytiva, Marlborough, MA, USA). Proteins were eluted with 50 mmol/L PB, 500 mmol/L NaCl and 100 mmol/L imidazole. The eluate was subsequently diluted 50-fold and subjected to anion-exchange chromatography using a Q-Sepharose column (Cytiva, Marlborough, MA, USA). The bound proteins were eluted with a solution of 20 mmol/L PB containing 150 mmol/L NaCl at pH 7.4 and stored at −80°C.

The purity of the purified rHpaA was assessed via size exclusion chromatography–high-performance liquid chromatography (SEC‒HPLC) on an Agilent Infinity 1260 II system (Santa Clara, CA, USA) with an Xtimate SEC-300A column (300 mm × 7.8 mm; Welch Materials Inc., Shanghai, China). The mobile phase comprised 150 mmol/L PB and 10% acetonitrile, with the flow rate of 1.0 mL/min. The column temperature was maintained at 20°C, and the injection volume was 10 µL. The protein was detected by an UV detector at 280 nm.

Six-week-old, specific-pathogen-free (SPF) female BALB/c mice were randomly divided into a control group and an experimental group, with 20 mice per group. The experimental group received immunization with 50 μg of rHpaA mixed with 1 mg of aluminum adjuvant (Thermo Fisher Scientific, Waltham, MA, USA) in an equal volume via intramuscular injection on days 0, 14, and 21. The control group received an equivalent treatment with PBS. Serum samples were collected from the tail veins of mice 1 week after the final immunization and isolated by centrifugation at 3,000 g for 5 min.

The experimental animals were purchased from Dossy Experimental Animals Co., Ltd. (Chengdu, China) and housed in the experimental animal center of West China Hospital, Sichuan University. All experiments were approved by the institutional ethics committee (number: 20220113002).

The antigenicity of rHpaA was assessed via Western blotting. The purified protein was subjected a 12% SDS-polyacrylamide gel and electrophoretically transferred onto a 0.22 μm pore-sized polyvinylidene fluoride (PVDF) membrane. The membrane was blocked with 5% (w/v) nonfat dry milk in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 h. After three washes with TBST, the membrane was incubated with either mouse anti-H. pylori lysate serum or anti-rHpaA serum, which was prepared in our laboratory, as the primary antibody for 2 h. Following three additional washes with TBST, the membrane was incubated with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (Sangon Biotech, Shanghai, China) as the secondary antibody for 1 h. The SuperFemto ECL Chemiluminescence Kit (Vazyme, Nanjing, China) was used to visualize the immunoreactive bands with a ChemiDoc™ MP Imaging System.

The immunogenicity of rHpaA was evaluated using an enzyme-linked immunosorbent assay (ELISA). The wells of a 96-well ELISA plate were coated with 100 μL purified rHpaA at a concentration of 2 μg/mL and then blocked with 1% (w/v) bovine serum albumin (BSA) for 1 h at 37°C. Serial dilutions of serum samples from immunized mice were added to the wells and incubated for 2 h. Following three washes with PBST, HRP-conjugated goat anti-mouse IgG was added and incubated for 30 min at 37°C. The plate was then washed again, and 100 μL of TMB chromogenic substrate solution was added to each well for 15-min incubation at 37°C. The enzymatic reaction was stopped by adding 50 μL of 2 mol/L sulfuric acid (H₂SO₄) per well, and the absorbance at 450 nm was measured via a microplate reader (BioTek Instruments, Winooski, VT, USA).

To enhance the yield of rHpaA, one-factor-at-a-time experiments were performed to assess the composition of the basic medium. Understanding the optimal carbon source is crucial, as it serves as the primary energy substrate for microorganisms. The experimental data (Figure 1A) showed that maltose and glucose significantly enhanced rHpaA expression, achieving levels of 44.20% and 44.33%, respectively. These levels were 10.39% and 10.71% higher than the control group. Further evaluation revealed that when glucose was used as the carbon source, resulted in the highest cell density, with an OD₆₀₀ value of 4.72, which was 1.29-fold higher than that of the control at 3.65. The maximum rHpaA yield reached 0.45 g/L, marking a 1.41-fold increase compared to the control group.

Upon identifying the optimal carbon source, the focus shifted to the nitrogen source. The experiments (Figure 1B) revealed that the addition of inorganic nitrogen sources effectively increased rHpaA expression, but ammonium citrate inhibited cell growth. The combination of NH₄Cl with yeast extract and yeast peptone resulted in the highest OD₆₀₀ at 3.87 and the protein expression of 47.25%. This mix yielded 0.40 g/L of rHpaA, which was 1.27 times greater than that of the control.

Metal ions are essential for microorganisms, often serving as enzyme cofactors that influence cellular physiological processes. The next step was to investigate these elements for their potential to enhance rHpaA production. Compared to the control, the addition of Mg²⁺, Zn²⁺, Cu²⁺, Ni²⁺ and Ca²⁺ boosted rHpaA expression. Notably, only Ca²⁺ increased biomass simultaneously, resulting in an rHpaA yield of 0.39 g/L, which was 1.31 times greater than that of the control (Figure 1C).

The types of phosphate were also investigated, as they are necessary for electrochemical gradients and cellular signaling. The results indicated that the simultaneous addition of sodium phosphate and potassium phosphate to the culture medium resulted in a higher yield of rHpaA compared to the use of a single phosphate (Figure 1C). In the culture medium containing mixed phosphates, the yield of rHpaA reached 0.38 g/L, which was 1.27 times that of the control group, with an expression level of 46.06% and an OD₆₀₀ of 3.83.

The Plackett–Burman experimental design is a factorial screening method used to identify key variables from a comprehensive multivariate analysis (Plackett and Burman, 1946; Quinlan and Lin, 2015). On the basis of preliminary one-factor-at-a-time experiments, six medium components were selected as variables, labeled A, B, C, D, E, and F representing glucose, NH₄Cl, CaCl₂, yeast extract, yeast peptone and mixed phosphate solution, respectively. The response values in the design matrix for the 12 sets of experimental conditions varied from 0.3568 to 0.4575 g/L (see Supplementary Material). Analysis with Design Expert 13 revealed that the model was statistically significant at p < 0.05. Among the factors, glucose (p = 0.0016), NH₄Cl (p = 0.0022) and yeast extract (p = 0.0102) had the most significant effects on rHpaA yield, with contributions of 34.3664%, 30.1742% and 14.6369%, respectively (Table 1).

The steepest ascent path is an optimization strategy for multifactorial experiments, designed to incrementally adjust factor levels towards the optimal design center. The coefficients from the Plackett–Burman experiment (Table 1) indicated that NH₄Cl positively affecting the production of rHpaA, whereas glucose and yeast extract had negative effects. This implied that to reach the maximum response, the concentration of NH₄Cl should be progressively increased, whereas the concentrations of glucose and yeast extract should be decreased. During the second step of the five-step ascent, the yield of rHpaA peaked at 0.585 g/L, with the concentrations of the medium components approaching the optimal solution, at which point the concentrations of glucose, NH₄Cl, and yeast paste were 7 g/L, 5 g/L, and 16 g/L, respectively (Table 2).

The current study systematically investigated the effects of various culture medium components on rHpaA production by recombinant E. coli for the first time, and the optimal component concentration of the medium was determined by RSM and ANN model optimization. It also explored the purification methods and the fundamental immunological properties of rHpaA.

Different culture medium compositions may exert varying effects on microbial growth and product metabolism. The results of one-factor-at-a-time experiments indicated that utilizing glucose as the carbon source significantly enhanced the expression efficiency of the recombinant E. coli, leading to a marked increase in rHpaA production. The preference for glucose may be attributed to its rapid uptake and metabolism by bacteria, which facilitated more efficient energy conversion (Jeckelmann and Erni, 2020). Furthermore, this study demonstrated that the combined use of inorganic and organic nitrogen sources is more advantageous for rHpaA production compared to the use of a single nitrogen source. Notably, the addition of NH₄Cl resulted in a significant increase in rHpaA production. Inorganic nitrogen sources provide more quick-acting nitrogen and reduce the lag phase, thereby promoting rapid strain growth and metabolism (Kuypers et al., 2018). Replacing complex organic nitrogen sources like yeast extract and peptone with inorganic nitrogen sources also helps control production costs for economic feasibility (Terol et al., 2019). The addition of Ca²⁺ and the use of a mixed phosphate significantly enhanced the production of rHpaA. Ca²⁺ can affect ribosome function and protein folding (Ohkuri et al., 2002). The phosphate mixture can supply both K⁺ and Na⁺, providing greater stability in maintaining ion balance and osmotic pressure across the cell membrane, thereby better supporting the transport of nutrients into cells (Stautz et al., 2021; Szatmári et al., 2020).

Microbial metabolite synthesis critically depends on the nutritional environment, with different culture media causing significant changes in protein expression, primarily due to variations in resource allocation (Goormans et al., 2020). Considering the interaction and complex effects of multiple factors in the engineered E. coli culture medium, multi-factor optimization methods were used to avoid deviation from the results of one-dimensional studies. Plackett–Burman experimental analysis identified glucose, yeast extract, and NH₄Cl as the most significant variables, which are essential carbon and nitrogen sources that greatly influence rHpaA as direct energy substrates. The steepest ascent design further determined the value range of the central composite design, promoting the transition from preliminary variable screening to a more detailed optimization process, and prepared for subsequent precise adjustments (Sztendur and Diamond, 2002). The RSM model established a quadratic regression equation relating actual variables to rHpaA yield, with a three-dimensional plot highlighting significant interactions between glucose and yeast extract, as well as glucose and NH₄Cl (Jankovic et al., 2021). Utilizing RSM data, the ANN model further dissected the implicit nonlinear relationships of input variables on rHpaA yield. With GA optimization, a definitive formula for the culture medium that significantly enhances rHpaA production was derived. Concurrently, the ANN-GA model exhibited superior regression accuracy in both the prediction and optimization process (Prabhu et al., 2017). The limitation of BBD-RSM is its capacity to construct only polynomial regression models, which restricts its generalization ability when addressing complex data sets (Hamza et al., 2024). In contrast, artificial neural networks possess a nonlinear activation function that effectively captures intricate nonlinear relationships within the data, thereby providing enhanced data processing and generalization capabilities (Liu et al., 2022). This makes neural networks more suitable for guiding the optimization process of biologics manufacturing.

The structure and immunogenicity of rHpaA were critical factors influencing the protective efficacy of the vaccine. SEC-HPLC results indicated that rHpaA, following two-step purification, exhibited high purity and homogeneity. The recombinant antigen was recognized by mouse anti-H.pylori lysate serum, demonstrating that the rHpaA possessed antigenicity comparable to that of the natural HpaA from H. pylori (Lundström et al., 2003). The quantitative analyses of specific IgG in the serum of immunized mice revealed that the rHpaA antigen displayed high immunogenicity and effectively induced specific IgG responses, thereby confirming its potential for vaccine application (Banga Ndzouboukou et al., 2021).

In the development of H. pylori vaccines, HpaA is widely regarded as one of the most promising antigens. While most previous studies have concentrated on the immune effects and mechanisms of HpaA, there has been less focus on its engineering applications. To achieve large-scale application in the vaccine field, it is crucial to systematically enhance the antigen production yield. The optimization of fermentation conditions is a key strategy to significantly improve both product yield and quality. This study successfully increased rHpaA production through a series of experiments and model design methodologies, establishing a foundation for the future commercialization of H. pylori vaccines composed of rHpaA. Despite these positive findings, it is important to acknowledge the limitations of this study, which primarily focused on optimizing the composition of the culture medium. Future research must tackle the challenge of refining a broader range of culture conditions during scale-up fermentation to ensure improved scalability and reproducibility of the production method.

This study applied a systematic optimization strategy for the culture medium, in conjunction with RSM and ANN, leading to a significant enhancement in the production of rHpaA. The obtained rHpaA exhibited robust immunogenicity in animal models, providing a viable solution for its large-scale production in H. pylori vaccines. The ANN-GA model demonstrated superior accuracy in prediction and optimization, underscoring its broad utility in biotechnology.