Initial regression design delivers significant parameters for constructing the model

The study assessed the impact of eleven carefully selected factors, identified through a literature review, on bacterial growth using Design-Expert software (version 7.0.0, Stat-Ease Inc.). A total of 16 experimental runs were conducted using these variables (S1 Table) to assess the growth kinetics of nMPI-producing cells in real-time. For this purpose, the Design-Expert platform was integrated with the BioLector microbioreactor system (Beckman Coulter, Indianapolis, USA), enabling high-resolution, traceable measurements of cell growth dynamics. The PBD, a widely used statistical approach for initial screening experiments, was employed to identify the most effective components for following optimization [28]. The 16 experimental conditions derived from PBD were applied to cultures grown in a multi-well microbioreactor, with each condition performed in triplicate. The final volume in each well was adjusted to 1000 μL. Cells were incubated at 37 °C with constant shaking (800 rpm) for one hour under conditions defined in S2 Table. Throughout the incubation, scattered light intensity was continuously monitored to track cell growth in real time (S3 Fig). Among the factors tested, the concentration of IPTG was included to optimize induction efficiency and minimize production costs. Once cultures reached an optical density (OD₆₀₀) of 0.7–0.8, induction was performed using three IPTG concentrations (0.2 mM, 0.3 mM, and 0.4 mM) by the PBD matrix. After induction, cells were further incubated at 37 °C for one hour. To estimate mini-proinsulin concentration, UV/Vis spectrophotometry at 280 nm was used as a rapid and straightforward method, validated with standard lysozyme solutions. To ensure the reliability and accuracy of the spectrophotometric measurements, a validation test was first performed using manually prepared lysozyme standards at concentrations of 15 mg/mL, 10 mg/mL, 5 mg/mL, and 2.5 mg/mL. Each sample was dissolved in 1 mL of deionized water and directly measured using a Nanodrop spectrophotometer. The absorbance readings precisely matched the calculated concentrations based on the weighed amounts, confirming the instrument’s accuracy.Following this validation step, samples from both nMPI-expressing and non-expressing (control) cultures were analyzed using the same method, with confidence in the device’s performance. Final product concentrations were expressed in grams per liter and confirmed via 20% SDS-PAGE analysis (S3B Fig). Product yield was defined as the response variable and recorded for each experimental condition in S1 Table’s response section.

ANOVA results for the selected PBD factorial model demonstrated statistical significance, with a p-value of 0.05 and a coefficient of determination (R²) of 85.37% (Table 1, S3 Table). The model identified MgSO₄, glycerol, glucose, and yeast extract as significant contributors, indicating these variables had the most substantial influence on mini-proinsulin production yield (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. ANOVA Table Analysis for Selected Factorial Model of Plackett-Burman Design. The analysis of variance (ANOVA) table was obtained for the selected factorial model.

https://doi.org/10.1371/journal.pone.0329319.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Regression model analysis over residuals, half-normal plot, standardized effects, and factor interactions.

(A) Externally studentized residual shows the run number of 12 was out of the limits. (B) The half-normal plot shows that the parameters C, D, E, and H were significant for constructing the model. We can also observe the p-value of parameters in Table 1. (C) Pareto chart shows the Bonferroni and t-value limits based on selected parameters. The limit values were 3.899 and 2.306 of Bonferroni and t-Value, respectively. MgSO₄ was the only parameter that passed the Bonferroni limit. (D) The perturbation chart also shows that MgSO₄, glucose, and glycerol have a negative effect on nMPI production; however, yeast has a positive effect (see S3 Table).

https://doi.org/10.1371/journal.pone.0329319.g002

Analysis of externally studentized residuals identified an outlier in run number 12, which exceeded acceptable limits (Fig 2A). Despite this anomaly, the regression model remained statistically robust, with a coefficient of determination (R²) greater than 0.85 (Table 1), a threshold commonly considered acceptable in biological studies [29]. The half-normal plot is a diagnostic tool used to assess the normality assumption of a model by evaluating the distribution of residuals [30]. In Fig 2B, the linear alignment of residuals suggests that they follow a normal distribution. Deviations from this linear trend would have indicated violations of the normality assumption. Furthermore, the half-normal plot effectively visualized the standardized effects of independent variables, highlighting MgSO₄, glycerol, glucose, and yeast extract as the most significant factors influencing the response. To further confirm variable significance, a Pareto chart was created to display the standardized effects of each factor [31]. Threshold values for significance were set at 3.899 for the Bonferroni limit and 2.306 for the t-value. Among all tested parameters, only MgSO₄ exceeded the Bonferroni threshold, indicating a particularly strong effect (Fig 2C). However, glycerol, glucose, and yeast also showed significant impacts, supporting their inclusion in the model as significant contributors. A perturbation plot was used to compare the relative effects of individual variables around a central point in the design space [32]. This analysis confirmed that MgSO₄, glucose, and glycerol negatively influenced nMPI production, suggesting that lower concentrations of these components may enhance yield. In contrast, yeast extract exhibited a positive effect, indicating that increasing its concentration could improve nMPI output (Fig 2D and S3 Table).

Quadratic modeling identifies glycerol as a key modulator in multivariate culture

To further optimize the statistically significant factors identified in the initial screening—MgSO₄, glycerol, glucose, and yeast extract—a CCD was applied, generating 30 experimental conditions (S4 Table). These conditions were tested using the microbioreactor platform. Throughout all experiments, baseline medium components were kept constant, including 10 mM KCl, 15 mM MgCl₂, 15 mM KH₂PO₄, and 10 mM thiamine. When cultures reached an OD₆₀₀ of 0.7–0.8, induction was carried out with 0.25 mM IPTG for each condition, followed by a 1-hour incubation at 37 °C (S4A Fig). Final nMPI concentrations were quantified in grams per liter using UV/Vis spectrophotometry at 280 nm and further validated by 20% SDS-PAGE analysis (S4B Fig). nMPI yield served as the response variable and was recorded for each run in the CCD matrix (S4 Table, response section). After testing, we created a quadratic model to show how the variables relate to the response. This model helps us predict and confirm the best conditions. The model demonstrated strong statistical significance, particularly for the effect of glycerol, with a p-value of 0.0037 (Table 2), supporting its reliability. The model did not fit the data well, as shown by a statistical value of p = 0.4479. This means that the errors in predictions were small and did not significantly affect the model’s accuracy (Table 2). Although the R² value of the model was moderate (0.6866), the residual deviation (3.3%) remained well within the acceptable range of 5–10% for biological regression models [29]. This level of accuracy was consistent with our experimental validation, as reflected in the outcomes summarized in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The analysis of the variance table is used to perform ANOVA for a Response Surface Reduced Quadratic Model.

https://doi.org/10.1371/journal.pone.0329319.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. The forecasted optimum conditions for achieving the maximum production of nMPI.

https://doi.org/10.1371/journal.pone.0329319.t003

We looked at studentized residuals to check if the selected factor concentrations were suitable (Fig 3A). Most data points closely followed the probability line, which shows that the chosen concentrations were right for the optimization process. In the Predicted vs. Actual plot (Fig 3B), nMPI concentration values should ideally be close to the diagonal reference line, which shows high predictive accuracy. The model works well overall, but there are some small deviations at lower nMPI concentrations, indicating slight inaccuracies in these predictions. On the other hand, medium to high nMPI concentrations show a more consistent distribution around the reference line, proving that the model performs strongly and reliably in this range. These results show that the model reliably predicts moderate to high nMPI yields. While it performs slightly worse at low concentrations, this does not significantly affect its overall ability to make accurate predictions. Additionally, a perturbation plot was used to examine the model’s sensitivity to individual factors (Fig 3C). This analysis revealed that glycerol (factor B) had a particularly strong influence on the response variable, further highlighting its critical role in nMPI production. Instead, glucose, yeast, and MgSO₄ (factors A, C, and D, respectively) have much less impact compared to factor B, which is important for boosting nMPI production (Fig 3C). The data in Table 2 and S5 Table support this finding. Testing offers that factors A, C, and D have little effect on the concentrations we tested. We looked closely at how these factors interact, which we show in Figs 3D, 3E, and 3F. The results indicate that, at the tested concentrations, glucose does not interact with glycerol (Fig 3D). However, glycerol does interact significantly with other components. It interacts with yeast at higher concentrations (Fig 3E) and with MgSO₄ at lower concentrations (Fig 3F). Understanding these interactions is important for optimizing the medium composition. It suggests that we should adjust the amounts of yeast and MgSO₄ based on the level of glycerol to create the best conditions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Evaluation of model sensitivity and parameter interactions in nMPI production.

(A) Internally studentized residuals show the correction of chosen parameter concentration. All runs were close to the probability line, which means the selected concentrations were correct for optimization. (B) The nMPI ranges in the Predicted vs. Actual chart. Low concentrations of nMPI shifted a bit from the actual data. The other concentration ranges distribute homogeneously around the line. (C) Perturbation plot. Glycerol significantly affects nMPI production, while glucose, yeast, and MgSO₄ have no significant impact on these concentration ranges. (D, E, F) The interactions of parameters. (D) Glucose and glycerol had no interactions in these concentration ranges. However, (E) glycerol interacted with yeast at high concentrations, while at low concentrations (F), there was an interaction with MgSO₄.

https://doi.org/10.1371/journal.pone.0329319.g003

The adapted hybrid approach made the finest recipe for mini-proinsulin production

System calculated factorial contour plots (Fig 4) and response surface analysis (Fig 5A) to better understand how different factors affect nMPI yield. The two-dimensional contour plots (Fig 4) and the three-dimensional response surface plot (Fig 5A) show how the independent variables relate to nMPI yield. We changed these independent variables within a set range to see how they work together to influence nMPI production. The contour plots (Fig 4) give useful insights into how different factors interact. For example, higher levels of glycerol and glucose, along with lower amounts of MgSO₄ and yeast, reduce nMPI yield from 12.7172 mg/ml to 9.5663 mg/ml (Fig 4, top left corner). Also, increasing glycerol and yeast at higher levels of MgSO₄ and glucose decreased nMPI yield from 11.124 mg/ml to 6.84772 mg/ml (Fig 4, bottom right corner).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Factorial contour plots of Central Composite Design.

A representative plot is an effective visualization tool, capturing the intricate relationship between the independent variables and the resulting nMPI yield. The intentional manipulation of these independent variables within a predetermined range allowed for systematically exploring their collective impact on nMPI production. Elevating glycerol (B) levels adversely affects nMPI efficacy, whereas increasing glucose (A) levels positively influences nMPI yield. Maintaining yeast (C) at lower levels contributes positively to nMPI yield. Conversely, increasing MgSO4 (D) concentrations negatively impacts nMPI efficacy. The best factorial composition is characterized by elevated glucose (A) and lower levels of yeast (C) and MgSO4 (D), giving the highly yielded nMPI concentration of around 13 mg/ml (top left corner).

https://doi.org/10.1371/journal.pone.0329319.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Illustrating the interactive effects of glucose, glycerol, MgSO₄, and yeast concentrations on the efficiency of nMPI production while keeping temperature and pH constant.

(A) The 3D response surface plot indicates that 1% glycerol and 15 mM glucose positively affect nMPI production in the presence of 2.5 g/L yeast and 5 mM MgSO4, providing around 13 g/L nMPI efficacy. (B) 2-factorial contour plots validate that a lower level of glycerol and a higher level of glucose provides a high-yield product.

https://doi.org/10.1371/journal.pone.0329319.g005

We calculated a 3D response surface plot (Fig 5) to clarify how different variables interact with each other. This plot clearly showed how glucose, glycerol, MgSO₄, and yeast concentrations affect nMPI production in 3D rendering (Fig 5A). We found that using a higher glucose concentration in the presence of 2.5 g/L of yeast and 5 mM of MgSO₄ led to increased nMPI yield (Fig 5A). This combination produced more nMPI than any other variable we tested.

Four distinct scenarios were statistically derived and subjected to experimental validation based on these findings. The experimentation was performed using the micro bioreactor (1 mL), with each scenario being replicated three times to ensure the reliability and accuracy of the results. All components of the baseline medium were kept constant during the experiments, including 10 mM KCl, 15 mM MgCl₂, 15 mM KH₂PO₄, and 10 mM thiamine. Despite the absence of substantial discrepancies among the scenarios, noteworthy that scenario III—characterized by a glucose concentration of 8.78 mM, glycerol concentration of 10 g/L, yeast at 2.5 g/L, and MgSO₄ concentration of 15 mM—exhibits the slightest deviation, and the highest nMPI production compared to the further scenarios (Table 3). The final nMPI concentrations cultivated by microbioreactor (Fig 6) were quantified using UV/Vis spectrophotometry at 280 nm (Table 3) and confirmed by 20% SDS-PAGE analysis (Fig 6)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Production efficacy of nMPI and analytical characterization.

Growth kinetics of 4 scenarios obtained by RSM using the multi-well micro bioreactor. Each parameter was triplicated. 20% SDS-PAGE analysis for determining nMPI efficacy. Each band indicates each scenario’s protein yield, with no amount of equalization performed to assess direct protein efficacy. The original gel image is in the supplementary file (S4A Fig). M: marker.

https://doi.org/10.1371/journal.pone.0329319.g006

To further evaluate the scalability and practical applicability of the optimized conditions identified in the micro-scale experiments, all four statistically derived scenarios were independently tested in 3-liter bench-top bioreactors. Each condition was scaled up under controlled parameters to validate whether the nMPI yields observed in the multi-well microbioreactor could be reliably reproduced in larger volumes. In addition, a control experiment was performed using a standard LB medium formulation (30 g/L LB broth) under identical fermentation and induction (0.25 mM IPTG) conditions to benchmark the performance of the optimized media. For all the scenarios, all the ingredients in the basic mixture stayed the same during the experiments. This included 10 mM of KCl, 15 mM of MgCl₂, 15 mM of KH₂PO₄, and 10 mM of thiamine. The final nMPI concentrations were quantified using UV/Vis spectrophotometry at 280 nm and confirmed by 20% SDS-PAGE analysis (Fig 7 and S5 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. SDS-PAGE analysis of nMPI production under large-scale (3L) cultivation conditions across four optimized media scenarios compared to standard LB medium.

Panels A–D compare the elution fractions [1–9] of purified nMPI expressed in 3-L bioreactors using LB broth (30 g/L) and each of the four optimized media formulations (Scenarios I–IV), as defined in Table 3. (A) Scenario I yielded 11.87 mg/mL in microbioreactor assays but only ~1 mg/mL in 3-L scale. The protein bands in all fractions appear faint, indicating poor scalability. (B) Scenario II showed a predicted yield of 11.54 mg/mL in microscale, but only ~1.5 mg/mL was obtained in the 3-L bioreactor. Fractions remained faint and inconsistently distributed. (C) Scenario III demonstrated strong scalability: although the predicted yield was 13.00 mg/mL, the actual production in the 3-L bioreactor reached 11.5 mg/mL. Bands are clearly visible and evenly distributed across all fractions, indicating robust expression and purification efficiency. (D) Scenario IV showed moderate scalability, with an observed yield of ~5 mg/mL compared to 12.33 mg/mL in the microscale. Bands are moderately intense and visible across mid-fractions. In all panels, the top gels show production from LB broth, which served as the baseline control. The lower gels in each panel correspond to the scenario-specific media. Arrowheads indicate the expected molecular weight region for nMPI. The original gel images are in the supplementary file (S4B-F Fig). M: marker (~12 kDa).

https://doi.org/10.1371/journal.pone.0329319.g007

SDS-PAGE analysis of elution fractions from large-scale cultures revealed clear differences in nMPI expression levels across the conditions (Fig 7). Scenario I, despite an observed yield of 11.87 mg/mL in the microbioreactor, resulted in only ~1 mg/mL in the 3-L bioreactor, with faint and diffuse bands observed across fractions (Fig 7A). Similarly, Scenario II showed limited scalability, producing ~1.5 mg/mL in the bioreactor versus the observed 11.54 mg/mL in the microbioreactor, again with weak and inconsistently distributed bands (Fig 7B). In contrast, Scenario III demonstrated prominent scalability: although the observed yield was 13.00 mg/mL in microbioreactor, the actual production reached 11.50 mg/mL in large-scale format, with strong and well-resolved bands across all fractions (Fig 7C). Scenario IV showed intermediate performance, yielding ~5 mg/mL compared to an observed 12.33 mg/mL in microbioreacot, with moderate band intensity (Fig 7D). Across all scenarios, LB broth (30 g/L) produced ~3 mg/mL of nMPI and served as a baseline reference. These results confirm that Scenario III not only offers the best predictive-to-observed yield correlation in micro- and large-scale production but also exhibits superior protein expression consistency.

Microorganisms, culture media, chemicals, BioLector and Software

We transformed the E. coli BL21 Rosetta 2 strain using the pET28a(+) vector, which contains a nMPI construct with a removable 6xHis-tag at the start. We sourced all necessary chemicals from trusted suppliers, including Merck and Sigma. Our team provided the KUYBIIGM-Ladder protein weight marker. For designing and analyzing our optimization experiments, we used the Design-Expert 7.0.0 software from Stat-Ease, Inc. in Minneapolis, MN, USA. We conducted screening experiments with the BioLector XT Microbioreactor in Indianapolis, focusing on statistically determined parameters.

Media optimization

We started by screening different factors using the Plackett-Burman Design (PBD) method to find the most important variables. After that, we used the Central Composite Design (CCD) method from Response Surface Methodology (RSM) to optimize the levels of these key variables. All experiments took place in 2 mL volume wells in the BioLector XT Microbioreactor. Each well contained 1 mL of the specific culture medium, which we prepared based on certain design points and inoculated with an overnight seed culture. We grew the inoculated cells for 4 hours in each well, stirring continuously at 800 rpm and maintaining a constant temperature of 37°C. We used 48-well FlowerPlates from mp2-labs in Baesweiler, Germany. We monitored growth by measuring scattered light intensities in real time. Following bacterial growth, the cells harboring the nMPI plasmid were harvested and pelleted by centrifugation (3500 rpm x 10 mins). The resulting pellets were solubilized overnight at room temperature on a shaker using 250 μL of reducing solubilization buffer containing 25 mM Tris-HCl (pH 9.0), 5 mM 2-mercaptoethanol, and 8 M urea. Complete dissolution of the pellets was observed under these conditions. To determine the final nMPI concentration, we used a UV/Vis spectrophotometer set to a wavelength of 280 nm. The absorbance of the non-induced sample (lacking nMPI expression; before induction) was subtracted from that of the induced sample (overexpressing nMPI; after induction) to account for background signal and isolate the contribution of the recombinant protein. We analyzed the results further using 20% SDS-PAGE (Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis). Raw versions of all gel images are presented in S1 File_raw_image.

Screening of factors using the Plackett-Burman factorial design approach

We studied eleven factors that could affect our experiment. These factors included different sources of nitrogen and carbon, pH levels, various media concentrations, and the presence of salts and metal ions. We designed the experiment using a Plackett-Burman factorial design, which involved sixteen tests conducted with DoE software. After performing the BioLector experiments, we analyzed the results statistically. We presented our findings, including model validation and the importance of each variable, in tables that showed ANOVA and fit statistics. We identified significant variables (with p-values less than 0.05) by looking at the ANOVA table and tools like the Half-normal plot, Pareto chart, and Perturbation chart. These methods helped us determine which factors had a major impact on the results of our experiments.

Optimizing factors using response surface methodology

Based on the PBD results, we selected four important factors for further optimization using CCD, which required thirty experimental runs. We prepared the culture media for each run using the specific composition of the selected factors, following the design points in BioLector. We also added constant values for other media components that were not part of the model, based on nMPI results measured in mg/ml from the PBD design and SDS-PAGE results. We supplied the concentration corresponding to the highest response for less important variables, along with the −1 level of variable factors that were insignificant. After conducting the experiments in BioLector, we analyzed the responses using different models. We chose the best model based on validation parameters shown in the ANOVA table, fit statistics, and diagnostic analysis. The DoE software created a perturbation chart, and we illustrated the effects of each significant independent and dependent variable on the response using contour and 3D plots. Finally, the software generated four scenarios for optimal points based on the regression equation. We tested the predicted design scenarios with the highest desirability in triplicate using the BioLector XT microbioreactor. The optimal medium identified was Scenario-3, which yielded 13 mg/ml with a 3.3% deviation, and is ready for scale-up optimization. Raw versions of all gel images are presented in S1 File_raw_image.

Large-scale production

Initially, stock cells containing nMPI genes were cultured overnight in 150 mL of LB medium (30 g/L), supplemented with 50 µg/mL kanamycin and 35 µg/mL chloramphenicol, at a temperature of 37°C and a shaking speed of 110 rpm using a New Brunswick Innova 4430R incubator. Following this, a 150 mL cell culture was transferred to a 6 L bioreactor (Braun Biotech Sartorius 6L Jacketed Glass Vessel Bioreactor) containing the same antibiotics and 4 mL of antifoam. The main cultivation process was conducted at 37°C with an initial culture volume of 3 L. The air flow rate was set at 15 L/min, and the stirrer speed was maintained at 800 min ⁻ ¹. Thermal mass flow meters (Bronkhorst High-Tech B.V., Ruurlo, The Netherlands) were used to mix air and oxygen. The concentration of dissolved oxygen was monitored using two polarographic electrodes (Ingold, Germany) and was kept at 30% of air saturation by adjusting the stirrer speed and aeration rate. Pure oxygen was added to the inlet air as needed. The pH was maintained at 6.8 by adding aqueous ammonia (25% w/w), with all controls managed by the bioreactor’s process control system. When the optical density reached 0.7, protein production was induced by adding 0.25 mM IPTG to the cultured cells, followed by a three-hour incubation at 37°C and a shaking speed of 250 rpm. After the incubation period, cell pellets were harvested by centrifugation at 3500 rpm for 45 minutes using a Beckman Allegra 15R centrifuge, and the collected pellets were stored in a −80°C freezer.

Solubilization, purification, sulfitolysis and refolding

The downstream process of large-scaled nMPI was performed following Ayan et al. study (Ayan et al., 2025). Briefly, to partially purify and completely solubilize inclusion bodies, 1 gram of cell pellets containing nMPI proteins were resuspended in 10 mL of lysis buffer composed of 25 mM Tris-HCl (pH 8.0) and 5 mM Ethylenediaminetetraacetic acid (EDTA). The bacterial cells were lysed using sonication, and the resulting cell debris was subjected to low-speed centrifugation at 6000 × g for 7 minutes at 4°C. The obtained pellets were then resuspended in wash buffer A, which contained 25 mM Tris-HCl (pH 8.0), 5 mM EDTA, 0.1% (v/v) Triton X-100, and 1 M urea. This resuspension was sonicated for 30 seconds in an ice bath, followed by centrifugation at 8000 × g for 20 minutes at 4°C. To further isolate the inclusion bodies, the debris was resuspended at a concentration of 0.1 g/mL in a binding buffer consisting of 25 mM Tris-HCl (pH 8.4), 5 mM 2-mercaptoethanol (BME), and 8 M urea. Centrifugation at 10,000 × g for 30 minutes at 4°C enabled the recovery of inclusion bodies containing the fusion proteins. The resulting supernatants were filtered through a 0.2-micron Millipore filter. Later, sulfitolysis of nMPI was performed at 25°C for 4 hours using 200 mM sodium sulfite (Na₂SO₃) and 20 mM sodium tetrathionate (Na₂S₄O₆). Manual gravity Ni-NTA affinity chromatography was employed next, which involved column equilibration with one column volume of a buffer consisting of 25 mM Tris-HCl (pH 8.0), 5 mM BME, and 8 M urea. Elution was carried out using a buffer of 25 mM Tris-HCl (pH 8.0), 5 mM BME, 8 M urea, and 250 mM imidazole. The purity of the nMPI sample was assessed at each stage through 20% SDS-PAGE gel analysis and Coomassie staining. High-purity fractions were collected and filtered through a 0.2-micron Millipore filter. The purified nMPI (at a concentration of 0.5 mg/mL) was then dialyzed against 25 mM Tris-HCl (pH 9.00) containing 8 M urea and 0.5 mM EDTA at 4°C. After centrifugation at 10,000 × g for 30 minutes at 4°C to remove potential aggregates, the resulting mixture was dialyzed in a refolding solution of 0.1 M Gly/NaOH (pH 10.5), 0.5 M urea, 0.5 mM EDTA, and 5% glycerol, stirring at 15°C overnight. Finally, the refolded nMPI was dialyzed against 25 mM Tris-HCl (pH 8.0) containing 5% glycerol for 18 hours.

Proliferative effects of insulin monomers in hiPSC culture systems

Human-induced pluripotent stem cells (hiPSCs) were generated from peripheral blood mononuclear cells (PBMCs) through a non-integrative Sendai viral vector system that delivers Yamanaka reprogramming factors. After successful reprogramming, cell colonies were expanded on Matrigel-coated plates utilizing mTeSR1 medium (StemCell Technologies) and were regularly subcultured with 0.5 mM EDTA. Before experimental procedures, all hiPSC lines were cryopreserved and later re-adapted to the TeSRE8 culture medium for at least two weeks to guarantee stability. The nMPI, created as detailed in the Methods section, was used for comparative evaluations. For treatment assays, hiPSCs were cultured in a defined E8 medium formulation supplemented with essential additives, including 100 U/ml penicillin, 100 µg/ml streptomycin, 20 mg/ml transferrin, 3 mM sodium selenite, 100 mg/ml ascorbic acid, 1 mg/ml FGF2-G3, and 0.5 µg/ml TGF-β3. Either 20 mg/ml of the new nMPI analog or corresponding concentrations of commercial recombinant insulin were added to the medium. Cultures were kept at 37°C in a humidified environment with 5% CO₂. Cell proliferation patterns were evaluated at designated time points—Days 3 and 7—using the Mateo TL Digital Transmitted Light Microscope. Quantitative findings indicated that nMPI had a proliferative influence on hiPSCs similar to that of human insulin, underscoring its potential as a functionally similar analog.