1. Round 1 DBTL: Chassis and Pathway Construction

In the first engineering cycle, our goal was to build a metabolic pathway in E. coli capable of synthesizing citronellal. We chose to reconstruct the three-enzyme cascade “geranyl diphosphate → geraniol → citronellal” in the E. coli BL21(DE3) chassis, involving three heterologous enzymes: geranyl diphosphate synthase (GPS), geraniol synthase (CsTPS1), and geraniol dehydrogenase (GeDH). To enable flexible and efficient expression, we adopted a multi-plasmid co-expression system: GPS was cloned into pET28a (T7 promoter, IPTG inducible, kanamycin resistance), CsTPS1 into pET21a (T7 promoter, IPTG inducible, ampicillin resistance), and GeDH into pBAD33 (araBAD promoter, L-arabinose inducible, chloramphenicol resistance) (Fig. 1).

Fig.1 Recombinant plasmid designs for pET28a-GPS, pET21a-CsTPS1 and pBAD23-GeDH

This plasmid combination ensured compatibility and allowed independent regulation of upstream and terminal enzymes via IPTG and L-ara dual induction, providing convenience for subsequent flux balancing and condition optimization. During construction, the target genes were codon-optimized for E. coli and synthesized as fragments (Fig. 2).

Fig.2 M: Marker; 1:GeDH; 2:CsTPS1; 3:GPS; 4: pET28a; 5: pET21a; 6: pBAD23

Each fragment was cloned into the corresponding linearized vector by Gibson Assembly, and a triple-plasmid engineered strain was obtained after two rounds of transformation (Fig. 3).

Fig.3 Second transformation results

The recombinant colonies were screened and verified using colony PCR. Results showed that GPS, CsTPS1, and GeDH PCR products matched the expected sizes without nonspecific amplification (Fig. 4), indicating correct gene insertion.

Fig.4 M: Marker; 1,2:GeDH; 3,4:CsTPS1; 5,6:GPS

Sequencing further confirmed correct sequences, reading frames, and orientations, proving successful construction of the three-enzyme cascade pathway. This meant we had obtained a strain “theoretically capable of producing citronellal,” laying the foundation for the next round of experiments.

2. Round 2 DBTL: Verification of the Performance of Engineered Strains

With the engineered strain constructed, the second cycle focused on validating growth performance and confirming proper transcription and translation of the heterologous pathway. We designed three levels of experiments:

Measuring growth curves of engineered and control strains to assess effects of the multi-plasmid dual-induction system.
Using qPCR to quantify mRNA fold changes of heterologous genes before and after induction, confirming transcriptional activation.
Detecting protein expression of target enzymes by SDS-PAGE/Western Blot to confirm translation.Without induction, the engineered strain's growth curve nearly overlapped with wild-type and empty-vector controls, with similar lag phases, exponential growth rates, and final OD₆₀₀ values (Fig. 5). This indicated that carrying three plasmids did not significantly burden host growth.

Fig.5 Group 1: Engineered bacteria + LB + antibiotics; Group 2: Wild-type bacteria + LB; Group 3: Empty vector bacteria + LB + antibiotics

After induction, relative qPCR (normalized to 16S rRNA) showed significant mRNA upregulation: GPS ~16.1-fold, CsTPS1 ~18.0-fold, and GeDH ~12.6-fold (Fig. 6). This confirmed successful transcriptional activation of the entire pathway under IPTG + L-ara co-induction.

Fig.6 Relative expression level of GPS, CsTPS1 and GeDH

Notably, GeDH had slightly lower induction than the other enzymes, likely reflecting promoter strength differences or suboptimal induction conditions, suggesting directions for optimization.

At the protein level, Western Blot successfully detected the three enzymes with clear bands at their expected molecular weights: GPS ~46.9 kDa, CsTPS1 ~63.7 kDa, GeDH ~41.6 kDa (Fig. 7). This confirmed translation and expression of all key enzymes.

Fig.7 The Western blot results of GPS, CsTPS1, and GeDH.

Together, growth curve, qPCR, and Western Blot results formed a complete evidence chain from “chassis health” to “transcriptional activation” to “protein expression,” confirming that the pathway was active inside cells and setting the stage for product detection.

3. Round 3 DBTL: Product Fermentation and Quantification

The third cycle focused on verifying citronellal production and establishing reliable quantification. Engineered strains were cultured in TB medium, induced at mid-log phase (OD₆₀₀≈0.6) with IPTG + L-ara to initiate citronellal biosynthesis.

High-performance liquid chromatography (HPLC) was used as the primary analysis method. A standard curve was prepared using citronellal standards (0.4-1.6 g/L), achieving excellent linearity (R²≈0.999) (Fig. 8).

Fig.8 The standard curve of the citronellal sample.

During HPLC analysis, the citronellal peak was detected and quantified at approximately 14 minutes using a UV detector. Simultaneously, we conducted parallel UV-Vis spectrophotometry (UV-Vis) measurements as cross-validation. Furthermore, to evaluate the distribution of the product both inside and outside the cells, we separately measured the citronellal content in the fermentation supernatant and cell lysate to determine whether subsequent analyses could focus solely on the supernatant. The HPLC analysis successfully detected the citronellal product. The fermentation supernatant sample exhibited a target peak at a retention time of around 14 minutes, with the calculated citronellal concentration being approximately 0.89 g/L (Fig.9)

Fig.9: HPLC of TB fermentation supernatant

Correspondingly, citronellal was also detected in the cell lysate sample, with a concentration of approximately 0.91 g/L (Fig.10)

Fig.10: HPLC of cell lysate

The product concentrations of the two samples were very close, and statistical significance testing confirmed that the difference was not significant (Fig.11). This indicates that the distribution of citral between the culture supernatant and intracellular fraction is roughly equivalent. Therefore, the workflow can be simplified by directly measuring the fermentation supernatant to represent the total yield.

Fig.11 Significance Analysis of Concentration Differences between Cell Lysate and Fermentation Supernatant.

In terms of quantification accuracy, concentration results obtained from UV-Vis parallel tests were consistent with those from HPLC, further enhancing the reliability of the data. Through this round of fermentation experiments, we clearly demonstrated that the engineered strain is indeed capable of synthesizing citral, reaching nearly gram-per-liter levels under shake-flask conditions. This not only marks a key milestone in product generation for the project but also provides a solid quantitative foundation for subsequent optimization.

The third round of results provided three important insights and directions for improvement: 1. The engineered strain successfully “lit up” the synthetic pathway at the product level, achieving citral titers close to g/L in shake flasks, thereby demonstrating the feasibility of biomanufacturing. 2. No significant difference was observed in the intra- and extracellular distribution of citral, so future detection can be simplified by analyzing only the supernatant, improving throughput for strain screening. 3. The consistency between HPLC quantification and UV-Vis detection established a robust “dual-channel” quantification system, offering a reliable benchmark for process optimization and performance evaluation.

Based on these findings, we decided to carry out more in-depth engineering optimization in the next round to further improve production: on one hand, by investigating the effects of induction temperature and inducer concentration on yield in fermentation process optimization; and on the other, by applying enzyme engineering strategies to modify the rate-limiting enzyme, with the aim of further enhancing citral synthesis efficiency.

4. Round 4 DBTL: Process Optimization and Enzyme Engineering

In the final engineering cycle, our focus shifted from simply achieving production to attaining high production of citral. To this end, we adopted a dual strategy: process optimization and rational enzyme engineering.

For process optimization, Response Surface Methodology (RSM) was introduced to systematically optimize induction conditions. Specifically, a Box-Behnken design was used to examine the combined effects of three factors—induction temperature (A), IPTG concentration (B), and L-ara concentration (C)—each at three levels, on citral yield. A total of 15 shake-flask experiments were performed to fit a quadratic polynomial model, enabling evaluation of main effects, interactions, and quadratic effects.

In parallel, in silico enzyme engineering of the rate-limiting enzyme GeDH was conducted. First, the protein structure was predicted, and Funclib was used to identify potential beneficial mutation sites. Next, Rosetta-based molecular docking was applied to calculate binding conformations, followed by GROMACS molecular dynamics simulations to assess the stability of mutant enzyme-substrate complexes. Through multiple rounds of scoring, four key mutation sites—Q60R, L125P, F147Y, and A300D—were identified and combined into a quadruple mutant, expected to enhance the terminal oxidation step from geraniol to citral.Under the optimized RSM conditions, citral yield reached 1.01 g/L (Fig.12), representing an improvement over the previous round.

Fig.12 Optimal fermentation conditions predicted by response surface HPLC

Meanwhile, the GeDH quadruple mutant strain achieved an even higher yield under the same conditions, with HPLC quantification showing 1.36 g/L—approximately a 53% increase compared to the unmodified strain (Fig.13). This significant improvement demonstrates the practical effectiveness of rationally designed mutations in enhancing flux through the terminal step.

Fig.13 HPLC fermentation under optimal conditions for the mutant strain

By combining process parameter optimization with enzyme engineering, we successfully elevated citral titers to a new level. The fourth round experiments deepened our understanding of the engineered system and provided clear guidance for future work:1. Induction temperature is a critical lever for yield improvement—low-temperature induction markedly enhanced citral synthesis efficiency and should be adopted as the standard condition.2. Dual-inducer optimization is essential—IPTG exhibited an optimal effective concentration, beyond which no further yield increase was observed. Optimal results required coordinated tuning with L-ara levels.3. Enzyme engineering is effective—rational mutations of the rate-limiting enzyme significantly boosted pathway flux, and when combined with process optimization, the gains exceeded those of either strategy alone.

Accordingly, we established 20 °C, IPTG 0.46 mM, and L-ara 0.20% as the current optimal induction conditions, and confirmed the use of “direct HPLC quantification of the fermentation supernatant, with UV-Vis as auxiliary” as our strategy for yield evaluation. Moving forward, we plan to extend the successful GeDH engineering approach to upstream enzymes, applying saturation or combinatorial mutagenesis at key sites. For scale-up studies, additional factors such as dissolved oxygen, pH, and induction timing will be incorporated into the next round of RSM optimization, in order to further enhance volumetric productivity and ensure batch-to-batch consistency.

Through iterative optimization in each cycle, we aim to ultimately achieve efficient and stable citral production.