Engineering

The Design-Build-Test-Learn (DBTL) cycle is a core engineering framework in synthetic biology, enabling systematic iteration and optimization of biological systems. In our project, we applied DBTL across four interconnected experimental lines: chassis selection, formaldehyde detection assay, whole-cell biosensor design, and enzyme-based biosensor design. Rather than treating these components as isolated modules, we leveraged insights from the Learn phase of each to inform and enhance the others. This integrative approach fostered cross-component refinement and accelerated development. The interplay between these DBTL cycles is illustrated in Figure 1.

For both our whole-cell two-component system and methanol dehydrogenase enzyme designs, we needed to select one chassis organism. Correct chassis selection is vital when working in backgrounds with varying conditions. For a methanol biosensor that works in cocktails, it is crucial our chassis microbe can tolerate different levels of methanol, ethanol, and varying pH. Our lab commonly sees synthetic biology work conducted in E. coli and Saccharomyces cerevisiae. These two model organisms were thus subject to the DBTL cycle to assess their tolerance of our required conditions.

CYCLE 1

Design

To mimic the conditions present in a methanol-tainted alcoholic beverage, we designed a maximum tolerance test for E. coli DH5α and S. cerevisiae BY4742. The range of ethanol and methanol concentrations were designed to match those potentially found in cocktails, while the pH levels ranged from neutral pH to 2, as cocktails can vary from neutral to acidic.

Build

A set of serial dilutions were prepared in 96-well plates. Methanol and ethanol concentrations ranged from low to high, and pH values ranged from neutral to acidic. Tolerance testing conditions are listed in Table 1. A constant amount of cell culture — either LB or YPD for the respective microbes — was added to all wells.

Table 1. We chose to subject the microbial candidates to the ranges listed below. Varying conditions of Methanol, Ethanol, and pH levels.

Test

Plates were incubated overnight in a Log Phase 600 plate reader, with optical density measurements taken to assess growth. Growth curves were visualized using the Log Phase 600 App software. To quantify overall growth inhibition across the tested concentration ranges, the area under the curve (AUC) was calculated for each tolerance assay. AUC values enabled a direct comparison of tolerance profiles between E. coli and S. cerevisiae.

Learn

Both chassis options showed progressive decline as the methanol and ethanol concentrations increased and pH decreased. AUC results showed that E. coli performed better than yeast across methanol and ethanol concentrations and comparably in lower pH levels. The results guided the choice of E. coli as the optimal chassis for future biosensor research. The findings underscore the significance of empirical chassis selection in biosensor design and provide a basis for developing resilient, field-deployable synthetic biology instruments to combat methanol adulteration in alcoholic drinks.

For our biosensor's enzyme design, we planned to employ a colorimetric assay to indicate the presence of methanol. As the enzyme design involves the reaction of methanol dehydrogenase oxidising the methanol to formaldehyde, a colorimetric test for detecting formaldehyde was required.

CYCLE 1

Design

We began by reviewing literature on colorimetric assays for formaldehyde detection, prioritising methods that were simple to perform, cost-effective, used readily available reagents and equipment, and produced a rapid and distinct colour change. The assay described by Thepchuay et al. [1] met these criteria and was selected for further development. In this method, formaldehyde is oxidized to formic acid using hydrogen peroxide under alkaline conditions. Bromothymol blue (BTB) serves as the pH indicator, shifting from blue (alkaline) to yellow (acidic) as formic acid is generated.

Methanol to formaldehyde reaction
2 CH₃OH + O₂ → 2 CH₂O + 2 H₂O

Formaldehyde to formic acid reaction
2 CH₂O + H₂O₂ + 2 NaOH → 2 HCOONa + 2 H₂O + H₂

The stoichiometric conversion of one mole to one mole provides a simple proxy for methanol detection via formaldehyde.

Build

Assay reagents were prepared according to the optimized concentrations reported in the paper Thepchuay et al. [1]. Unfortunately, we found the assay could not be replicated as described in the paper and as such required further refinement (see Notebooks and Results pages).

Test

Multiple reagent combinations with varying component volumes were tested to identify the most effective setup for formaldehyde detection. A range of concentrations of formaldehyde were used to evaluate the performance of the reagent. The goal was to determine which formulation produced the clearest and fastest colour change.

Learn

Thepchuay et al. [1] reported that the colour change should occur within 5 minutes before formic acid begins to evaporate. However, in our trials, no colour change was observed within this timeframe across all tested formaldehyde concentrations.

CYCLE 2

Design

To address the lack of colour change observed in Cycle 1, we conducted a further literature review and modified the assay design by incorporating iron salts into the reagent solution. Hydrogen peroxide alone is a relatively weak oxidising agent; however, through the Fenton reaction, iron ions catalyse its activity, enhancing the oxidation of formaldehyde to formic acid [2]. This adjustment aimed to improve the sensitivity and reliability of the colorimetric response.

Build

We continued to revise the assay using a 96-well microplate format to allow for replicates and visual comparison across a gradient of formaldehyde concentrations. The plate was configured to test two concentrations of hydrogen peroxide and two types of iron salts — iron(II) sulphate and iron(III) chloride — to evaluate their catalytic efficiency.

Test

We tested a range of formaldehyde concentrations aligned with the detection range of Supelco® formaldehyde test strips, enabling direct benchmarking against a commercial standard. Additionally, the observation window was extended beyond 5 minutes to assess whether delayed colour change could occur under the modified conditions.

Learn

Our finding was that it took one week to achieve the most discerning colour range indicative of the formaldehyde concentration range tested (see Results page). We would like to further explore measures to accelerate the speed of detection since it has been reported possible within 5 minutes [1]. Despite having access to the commercially available Supelco® formaldehyde test strips, the 96-well format chemical assay represents an opportunity for high throughput assessment of both methanol dehydrogenases (MDH), and promoter–MDH combinations, in future experiments.

For our whole-cell biosensor, we chose to utilise a two-component system (TCS), as it enables bacteria to sense and respond to specific environmental stimuli. These systems offer modularity and tunability, two important attributes for synthetic biology biosensors [3]. We began with a literature review to identify a well-characterised TCS responsive to methanol. Previous studies have demonstrated successful methanol detection using an engineered EnvZ/OmpR system, which initially captured our interest [4][5]. These described a chimeric protein that had the MxaY methanol sensing domain fused to the EnvZ histidine kinase, allowing the system to respond to methanol and signal a downstream response [4]. While this work inspired our approach, the papers did not present their methods in a reproducible manner. We were also wary that since EnvZ histidine response is associated with outer membrane porins, such a design may unintentionally allow excess toxic methanol and ethanol into the cells [6].

We next looked for a more well-characterised TCS. The CusRS TCS was recently optimised for biosensor applications by adjusting signal amplification and expression levels of CusR and CusS [7] for the purpose of copper detection. We decided to move forward by designing our TCS with the CusRS system fused with the MxaY domain from Paracoccus denitrificans.

Plasmid Synthesis

CYCLE 1

Design

The CusRS two-component system detects environmental stimuli through its membrane-bound histidine kinase, CusS, which undergoes autophosphorylation upon activation. The phosphate group is then transferred to the response regulator, CusR, enabling CusR to bind DNA and initiate transcription from the CusR-regulated promoter [7]. In our design we utilise a chimeric histidine kinase that responds to methanol, while the response regulator remains unchanged.

For reporter output, we placed the gene encoding amilCP, a bright blue chromoprotein, under control of the CusR-regulated promoter. The strong visual signal of amilCP may provide an easily detectable output upon methanol exposure; amilCP is a standard iGEM biological part and common in teaching labs Part: BBa_K592009. Additionally, the bidirectional nature of the CusR promoter allows for placement of two copies of the amilCP gene, further enhancing signal output. Our design includes both the chimeric histidine kinase and response regulator expression under a constitutive promoter ensuring the biosensor components are continually available to sense and respond. To maximize expression, we selected the constitutive promoter for gapA, which includes multiple transcriptional start sites (P1, P2, and P3) and ensures strong, constitutive expression in E. coli throughout various growth phases. In designing the plasmid we used the IDT plasmid design tool for codon optimisation.

Build

Synthesis of the plasmid was initially outsourced to Twist Bioscience. Sequences were uploaded to their systems to check the complexity which came back as standard. However, the initial plasmid synthesis was unsuccessful. Twist was informative, suggesting potential problems including a hairpin formed by the bidirectional amilCP, and repetitive regions.

Re-design

We aimed to eliminate the problems in the first plasmid design that led to failed synthesis. The bidirectional arrangement of the amilCP reporter gene resulted in the formation of a large hairpin structure, which complicated DNA synthesis and stability. To resolve this, we used two different synonymous DNA sequences encoding the same amilCP protein, effectively removing the hairpin issue. To further address repetitive regions, we incorporated three different terminators (T7, T7hyb10, and rrnBT1), thus breaking up sequence repeats. IDT tools were used to monitor the sequence complexity, which initially scored in the unacceptable range due to both repeats and filler DNA. We corrected these problems by deleting unnecessary filler sections, and once the complexity score was lowered, we ordered the improved design again from Twist. Unfortunately, Twist could not synthesise this new design either. Thankfully, ANSA Biotechnology offered to sponsor the Macquarie University team and successfully synthesised our new plasmid design.

Additionally, to enable functional validation of the chromoprotein output, a simplified plasmid was designed consisting only of the CusR promoter and bidirectional amilCP gene arrangement. This was successfully synthesised by IDT. In this design, the absence of the chimeric methanol sensor allows for direct induction by copper, leveraging the natural E. coli CusRS system to drive amilCP expression. This provided an efficient platform to confirm chromoprotein production and assess promoter responsiveness independent of the synthetic methanol-sensing module. 

Re-build

Synthesis of the main chimera plasmid was successful with ANSA, a pivotal step that allowed us to progress with our project. The resulting plasmid included the codon-optimised chimeric methanol-responsive histidine kinase, the CusR response regulator, the bidirectional amilCP reporter arrangement, and kanamycin resistance. The simplified plasmid ordered from IDT was successfully synthesised and will be used to test our chromoprotein.

Test

After receiving the synthesised plasmids, they were transformed into DH5α E. coli cells. A plasmid miniprep was performed to purify the plasmid DNA from the bacterial cultures. The purified plasmid samples were then sent for sequencing. Whole plasmid sequencing was performed by Plasmidsaurus using Oxford Nanopore Technology with custom analysis and annotation, which confirmed the sequences were as designed.

Learn

This whole process proved to be an invaluable lesson in plasmid construction, highlighting critical factors that must be considered at each stage. The challenges we faced, particularly regarding repetitive regions and hairpins, underscore the importance of rigorous sequence design. These insights will guide all subsequent plasmid projects. Prioritising sequence uniqueness and minimising repetitive elements will be essential for improving synthesis reliability and expediting successful construct assembly.

Functional Validation of Methanol and Copper-Responsive Reporter Systems

CYCLE 1

Design

The plasmids designed and sequence confirmed in the plasmid design section were transformed into E. coli. Experiments were designed to test the ability of our engineered cells to respond to methanol or copper respectively by producing amilCP. Both NEB 5-alpha and NEB 10-beta strains were chosen. Inductions with copper and methanol from low to high at multiple concentrations were planned to compare the sensitivity and specificity of each system.

Build

Overnight starter cultures were grown to mid-exponential phase, providing optimal conditions for induction. For each experimental condition, negative controls without plasmid were included to establish a baseline for non-specific colour development. All technical controls and biological replicates were run in parallel to ensure reproducibility and reliability. All induction concentrations were set up at increasing concentrations from minimal detection ranges to those becoming inhibitory for growth.

Test

Induction assays were carried out across a range of copper concentrations as well as varying levels of methanol. Post-induction, cultures were incubated, pelleted and then assessed for visible colour development. Negatives were included to monitor baseline colour and confirm specificity of the response, especially since copper chloride has a blue tinge. No visible colour development was detected in the chimera-containing cells under any induction condition, indicating a lack of functional reporter expression or activation. Visible blue colour development was observed in the amilCP-only cells following copper induction, confirming that the reporter system and induction method were functioning as expected.

Learn

Given the absence of colour development in chimera-containing cells, the next steps would involve troubleshooting the functionality, such as confirming protein expression and optimising induction protocols. The successful production of colour from the simplified amilCP design led us to register the bidirectional CusR promoter as a composite part BBa_25zau49i.

However, after seeking input from academic advisors and stakeholders, the consensus is to streamline efforts by focusing on a single approach. While building the chimera has been instructive, the enzyme-based design now presents a less risky, less regulated, and more scalable pathway. Although neither strategy has yet yielded complete success, the enzyme system is considered more promising for future development.

Methanol dehydrogenases (MDH) are essential enzymes that catalyse the oxidation of methanol to formaldehyde—the first and critical step in methanol metabolism. There are three main types of MDHs, classified by their electron acceptors: NAD-dependent, pyrroloquinoline quinone (PQQ)-dependent, and O₂-dependent [8]. NAD and PQQ-dependent MDHs are typically found in methylotrophic bacteria, while O₂-dependent variants are derived from yeast [8]. Based on learnings from our Chassis Selection DBTL cycle, we opted for a bacterial system over yeast. Although NAD-dependent MDHs are often favoured in synthetic methylotrophy due to their simpler genetic requirements, they exhibit lower substrate affinity for methanol compared to PQQ-dependent MDHs. Given that high substrate specificity is vital for MethaNO’s performance, we chose to pursue PQQ-dependent MDHs despite the added engineering complexity.

Plasmid Synthesis

CYCLE 1

Design

We selected the MDH from Methylorubrum extorquens AM1 (strain ATCC 14718), as it is one of the most extensively studied PQQ-dependent MDHs. Key literature, including gene identification [9], metabolic mechanism [10], and enzyme kinetics and purification technique [11], provided foundational insights for our construct. Based on this, we identified a set of essential genes required for functional MDH expression and sourced their protein sequences from UniProt. A His-tag was added to mxaF (Table 2) to facilitate downstream protein purification.

Table 2. Four genes identified as essential for a functional MDH synthesis – mxaF and mxaI code for the enzyme subunits, whilst mxaA and mxaK code for supporting proteins to facilitate co-factor Ca²⁺ incorporation into the enzyme.

To streamline plasmid construction, we used the Integrated DNA Technologies (IDT) plasmid design tool to codon-optimise the sequences for E. coli, remove restriction sites incompatible with iGEM assembly standards, and reduce the overall complexity score of the design. The T7 promoter was chosen for its high transcriptional efficiency, enabling strong protein expression. Since MDH enzymes require an oxidising environment to function properly, signal peptides were incorporated to direct recombinant proteins from the cytoplasm to the periplasmic space. We adopted signal sequences with enhanced performance including optimised translation initiation regions (TIRs) as described in Merzadeh et al. [12].

Build

Synthesis of the initial plasmid design (Figure 3) was outsourced to Twist Bioscience. However, the synthesis attempt was unsuccessful. Twist kindly shared informative information regarding potential repetitive regions and hairpins which may have thwarted these initial synthesis attempts.

Re-design

The redesigned plasmids minimised repetitive regions by utilising a greater variety of signal peptides. The tools from IDT proved particularly helpful in identifying areas of complexity. To further favour successful synthesis the new design was ordered with genes distributed across two distinct vectors (Figure 4 and Figure 5). Care was taken to ensure these included different antibiotic resistance markers and origins of replication for co-transformation compatibility:

• The mxaFI operon was cloned into a kanamycin-resistant pET vector.
• The mxaAK operon was cloned into a chloramphenicol-resistant pTwist vector.

Additionally, both plasmids were ordered from ANSA Biotechnology, who only offered kanamycin-resistant vectors.

Re-build & Test

Synthesis was successful for both plasmids via ANSA, whereas Twist successfully synthesised only the mxaAK plasmid. We greatly appreciate ANSA’s support in providing free synthesis services for complex DNA design and it was their successful synthesis of the mxaFI plasmid that allowed us to progress with our experiments.

Test

Whole Plasmid Sequencing was performed by Plasmidsaurus using Oxford Nanopore Technology with custom analysis and annotation. This confirmed the plasmids we cloned in C3019I NEB 10-Beta competent E. coli cells and C2987I NEB 5-Alpha competent E. coli cells were the sequence as designed.

Learn

Avoiding repetitive regions is critical for successful plasmid synthesis. Future designs should prioritise sequence uniqueness to enhance synthesis reliability.

MDH Enzyme Expression

CYCLE 1

Design

Plasmid expression was induced using 0.4 mM isopropyl β-D-1-thiogalactopyranoside (IPTG), as per documentation in Notebook page. To support cofactor-dependent enzyme activity, pyrroloquinoline quinone (PQQ) was supplemented at an initial concentration of 3 µM. Although E. coli do not synthesize PQQ endogenously, they possess high-affinity PQQ transporters that function effectively in environments with elevated PQQ concentrations (0.1–3 µM) [13].

Build

The double plasmid transformants were prepared on LB agar plates containing kanamycin and chloramphenicol (KanR and CamR), then overnight cultures inoculated. Three sample types were prepared: pellet resuspension, cell lysates, and purified protein. Detailed protocols are available in the Experiment and Notebook sections.

Test

Lysate and purified protein samples were analysed via SDS-PAGE to separate proteins by molecular weight and identify MDH. The MDH, assembled by two subunits, were expected at band ~70 kDa for mxaF and band ~10 kDa for mxaI. No clear bands of these sizes were observed in the purified protein samples.

Learn

Key areas for optimisation in Cycle 2 include:

• Induction efficiency: IPTG concentration or timing may require adjustment
• Cofactor loading: PQQ concentration may be suboptimal
• Cell density: Achieving OD₆₀₀ ≥ 0.6 is critical for sufficient biomass
• Purification yield: Current protocol may result in low recovery of target protein

CYCLE 2

Design

To enhance protein expression, the IPTG induction concentration was increased to 0.8–1 mM, based on recommendations from other researchers in our lab who reported successful yields at 1 mM. The concentration of PQQ was also doubled to 6 µM to ensure sufficient cofactor availability, given the unconventional source used. Instead of research-grade PQQ from laboratory suppliers, we opted for a commercially available dietary supplement due to cost constraints. A bottle of 30 capsules (40 mg each) costs approximately US$23, whereas Sigma-Aldrich’s reagent-grade PQQ is priced at US$238 per milligram. Notably, the supplement contains PQQ in its disodium salt form, which is reported to be more stable and water-soluble than the free acid form [14].

Build

Cultures were induced with the revised concentrations of IPTG and PQQ, and growth was monitored to ensure the target OD₆₀₀ of 0.6 was achieved. Sample preparation followed the same protocol as in Cycle 1. To address previously observed low protein recovery, the purification step was modified by increasing the quantity of magnetic beads used for His-tag capture and performing additional wash steps to improve purity and reduce background protein contamination.

Test

Purified protein, crude lysate, and resuspended cell pellet samples were analysed via SDS-PAGE using the same protocol as Cycle 1, but with double the loading volume to enhance protein detectability. Despite this adjustment, no visible bands corresponding to our target protein were observed in the purified sample. In contrast, the lysate and cell pellet lanes displayed broad, high-intensity bands, likely due to overloading, which compromised resolution.

Learn

Across both cycles, we found that optimising IPTG induction, cell biomass, and purification yield did not result in successful protein expression. Further investigation, including direct communication with ANSA, revealed that the vector used to synthesise our plasmid was a single-copy backbone, which only transitions to medium-copy upon arabinose induction—a detail we had not anticipated. This likely meant our expression relied on just one plasmid copy per cell, explaining the absence of detectable protein on SDS-PAGE.

In future work (beyond our current timeline), we plan to repeat the expression protocol with arabinose induction to activate higher plasmid copy number. Additionally, we aim to reconstruct the ANSA-synthesised insert into a more conventional protein expression backbone using Gibson Assembly. If SDS-PAGE confirms successful expression in purified samples, we will proceed with periplasmic extraction to evaluate the functionality of our signal peptides.

MDH Functional Characterisation

Cycle 1

Design

Following advice from our mentor, Prof. Rob Willows, we adopted a top-down experimental strategy—starting with resting cells, then progressing to crude lysate and finally purified protein. The rationale was that if methanol oxidation could be demonstrated at the whole-cell level, downstream characterisation would be more straightforward. In this cycle, we focused on implementing the resting cell methanol assay. This assay was also appropriate as we did not successfully obtain gel bands for lysate nor purified protein expression.

Build

The resting cell assay was designed to evaluate methanol oxidation activity in MDH-expressing cells without cell lysis. Induced cell pellets were resuspended and exposed to methanol at two concentrations: 0.1% and 2.0%. Detailed protocols are available on the Experiments page.

Test

To detect formaldehyde production as a proxy for methanol oxidation, we used Supelco formaldehyde test strips. Measurements were taken at multiple time points post methanol addition, ranging from immediate exposure to 1 hour. However, no formaldehyde was detected across conditions. Full results are documented on the Results page.

Learn

Recent literature highlights the essential role of the mxaJ gene in MDH enzyme maturation [15]. Without mxaJ, the enzyme remains non-functional even if mxaF and mxaI are expressed [15]. mxaJ encodes a chaperone protein that facilitates PQQ incorporation into the catalytic site of the mxaF-encoded alpha subunit, enabling proper assembly with the mxaI-encoded beta subunit [15]. Moving forward, we plan to include mxaJ in our plasmid design and switch to research-grade PQQ to ensure cofactor purity and functional enzyme assembly.

1. Y. Thepchuay, W. Chairit, N. Saengsane, P. Porrawatkul, and R. Pimsen, “Simple and green colorimetric method for the detection of formaldehyde in vegetable samples,” Journal of Food Composition and Analysis, vol. 111, p. 104623, 2022.
2. P. Wang et al., “A pH-responsive production of hydroxyl radical in Fenton process,” Environmental Science and Ecotechnology, p. 100566, 2025.
3. G. Özer Bergman, S. Mecacci, V. A. P. Martins dos Santos, and E. Asin-Garcia, “Engineering chimeric signaling proteins for microbial whole-cell biosensors: from design to deployment,” Trends in Biotechnology, 2025, doi: 10.1016/j.tibtech.2025.08.002.
4. I. Ganesh, S. Vidhya, G. T. Eom, and S. H. Hong, “Construction of Methanol-Sensing Escherichia coli by the Introduction of a Paracoccus denitrificans MxaY-Based Chimeric Two-Component System,” Journal of Microbiology and Biotechnology, vol. 27, no. 6, pp. 1106–1111, 2017, doi: 10.4014/jmb.1611.11070.
5. V. Selvamani, M. K. Maruthamuthu, K. Arulsamy, G. T. Eom, and S. H. Hong, “Construction of methanol sensing Escherichia coli by the introduction of novel chimeric MxcQZ/OmpR two-component system from Methylobacterium organophilum XX,” The Korean Journal of Chemical Engineering, vol. 34, no. 6, pp. 1734–1739, 2017, doi: 10.1007/s11814-017-0063-8.
6. L. J. Kenney and G. S. Anand, “EnvZ/OmpR Two-Component Signaling: An Archetype System That Can Function Noncanonically,” Ecosal Plus, vol. 9, no. 1, 2020, doi: 10.1128/ecosalplus.esp-0001-2019.
7. Y. Fu, J. Li, J. Wang, E. Wang, and X. Fang, “Development of a two component system based biosensor with high sensitivity for the detection of copper ions,” Communications Biology, vol. 7, no. 1, Art. no. 1407, 2024, doi: 10.1038/s42003-024-07112-6.
8. T.-K. Le, Y.-J. Lee, G. H. Han, and S.-J. Yeom, “Methanol dehydrogenases as key biocatalysts for synthetic methylotrophy,” Frontiers in Bioengineering and Biotechnology, vol. 9, p. 787791, 2021.
9. K. Mirzadeh et al., “Increased production of periplasmic proteins in Escherichia coli by directed evolution of the translation initiation region,” Microbial Cell Factories, vol. 19, no. 1, p. 85, 2020.
10. C. J. Morris, Y. M. Kim, K. E. Perkins, and M. E. Lidstrom, “Identification and nucleotide sequences of mxaA, mxaC, mxaK, mxaL, and mxaD genes from Methylobacterium extorquens AM1,” Journal of Bacteriology, vol. 177, no. 23, pp. 6825–6831, 1995.
11. C. Anthony and P. Williams, “The structure and mechanism of methanol dehydrogenase,” Biochimica et Biophysica Acta (BBA) - Proteins and Proteomics, vol. 1647, no. 1–2, pp. 18–23, 2003.
12. T. Karaseva et al., “Isolation and characterization of homologically expressed methanol dehydrogenase from Methylorubrum extorquens AM1 for the development of bioelectrocatalytical systems,” International Journal of Molecular Sciences, vol. 23, no. 18, p. 10337, 2022.
13. F. Munder, M. Voutsinos, K. Hantke, H. Venugopal, and R. Grinter, “High-affinity PQQ import is widespread in Gram-negative bacteria,” Science Advances, vol. 11, no. 22, p. eadr2753, 2025.
14. “Is Pyrroloquinoline Quinone the Same as PQQ Disodium Salt?,” Santa Biotech. [Online]. Available: https://santabiotech.com/ (accessed Sep. 26, 2025).
15. H. Zhou et al., “Deciphering the assembly process of PQQ dependent methanol dehydrogenase,” Nature Communications, vol. 16, no. 1, p. 6672, 2025.