A problem well stated is a problem half solved. -- Antoine de Saint-Exupéry
Fig. 1 | Synthetic biology adaptation of design-build-test-learn cycle.
Fig. 2 | Overview of the engineering cycle, illustrating the problem input, intermediate solution steps, and the target output.
Our project aims to construct Saccharomyces cerevisiae strains capable of efficiently expressing the small antimicrobial peptide LL-37 as an alternative to the extensive use of traditional antibiotics. In this section, we elucidate the fundamental principles of our design and decision-making, describe how we engineered S. cerevisiae to enhance production yields, and how we synthesized and tested LL-37 variants with enhanced antimicrobial properties.
1.Goals:
Our project aims to construct genetically modified S. cerevisiae strains capable of efficiently expressing the antimicrobial peptide LL-37. The cationic and hydrophobic properties of LL-37 enable it to interact with bacterial lipopolysaccharides, form transmembrane pores, cause intracellular content leakage, and induce bacterial death. Through efficient expression of LL-37, we hope to build a S. cerevisiae cell factory capable of expressing various small antimicrobial peptides, thereby providing a solution to the global antibiotic abuse problem.
2.Problems:
2.1 How to select suitable chassis strains? Can we design S. cerevisiae strains that express the small antimicrobial peptide LL-37?
2.2 How to successfully secrete the small antimicrobial peptide LL-37 extracellularly? Since the cations of LL-37 cause it to adsorb to negatively charged cell membranes. How to purify LL-37?
2.3 How to construct a S. cerevisiae factory that efficiently expresses the small antimicrobial peptide LL-37? We employ multiple methods to increase small antimicrobial peptide production, such as promoter engineering, protease gene knockout...
2.4 How to predict LL-37 variants with better antimicrobial effects? And how to synthesize and test the designed LL-37 variants?
Design:
To achieve LL-37 expression in S. cerevisiae, we constructed the plasmid pESC-URA3-PGAL1-LL37 and introduced it into four different S. cerevisiae chassis strains: engineering strain CEN.PK113-7D, laboratory strain S288C, and MATa haploids (URA3 deletion type) of wild-type strains PGY7 and PGY34, designated as QY1.1, QY1.2, QY1.3, and QY1.4 respectively.
Fermentation was performed in SC-Uracil medium with an initial glucose concentration of 4 %; after glucose depletion, raffinose and galactose were added to final concentrations of 2% each for induction. Samples were taken at 24, 48, and 96 h to measure optical density (OD600), product titer, and perform Western blot (WB) analysis.
Build:
We used one-step cloning to link the LL-37 coding sequence with the GAL1 promoter and CYC1 terminator to construct the recombinant plasmid. The verified plasmid was transformed into the four S. cerevisiae chassis strains mentioned above for fermentation to express the target protein.
Test:
OD measurement, titer detection, and WB analysis were performed on fermentation samples.
Fig. 3 | Four chassis strains were tested for 72 h fermentation sampling, with hollow circles representing OD600 and bar charts representing LL-37 titers.
The OD600 values of engineering and laboratory strains in fermentation medium were significantly higher than wild-type, reaching a maximum of 18.11 (±0.74), while wild-type reached only 12.08 (±0.63). Meanwhile, the engineering strain CEN.PK113-7D showed the highest production at 48 hours, reaching 386.81 (±16.07) μg/L, which was 7.67% higher (p=0.093) than the second-highest value of 359.25 (±14.62) μg/L achieved by wild-type strain QY1.3 at 72 hours, and significantly higher than all other strains at all time points. Therefore, further modifications will be based on this strain.
We also noted that after the engineering strain reached its peak at 48 hours, the titer began to decline, indicating that intracellular accumulation of LL-37 is prone to degradation, requiring development of secretion pathways.
Additionally, while WB bands were very bright, the yield detected by liquid chromatography was low, leading us to suspect that LL-37 might still be attached to cell debris after cell disruption. Next, we will conduct intracellular fluorescence localization experiments of LL-37 to explore bottlenecks in yield detection.
Learn：
• The engineering strain CEN.PK113-7D performed best in terms of yield, reaching peak at 48 h, demonstrating good potential as a chassis cell for cell factory
• Since intracellular accumulation of LL-37 is prone to degradation, secretion pathways need to be developed to enhance stability
• We hypothesize that LL-37 may attach to cell debris after cell disruption, resulting in lower HPLC yields than WB. To verify this, we will conduct intracellular fluorescence localization experiments of LL-37 to establish more reliable yield determination methods for subsequent cell factory construction.
Design:
To investigate the intracellular localization of LL-37 and extraction limiting factors, we designed the pESC-PGAL1-LL37-eGFP fusion expression plasmid, transformed it into CEN.PK113-7D strain, performed fermentation induction expression, and took samples at 48 hours for subcellular localization observation by fluorescence microscopy.
Build:
Successfully constructed pESC-PGAL1-LL37-eGFP plasmid and transformed it into CEN.PK113-7D strain.
Test:
Samples were observed under fluorescence microscopy, and the results are shown in Fig. 4 after 48 h of fermentation.
Fig. 4 | Fluorescence microscopy results of protein localization after 48 h fermentation.
Learn:
• LL-37-eGFP fusion protein localized to the cell membrane region, possibly due to electrostatic adsorption between the positive charges carried by LL-37 and the negative charges on the S. cerevisiae cell membrane surface
• Subsequently, we will attempt to elute LL-37 by charge shielding under high salt and strong acid conditions to obtain an appropriate yield determination method
Design:
Based on fluorescence localization results, we hypothesized that LL-37 adsorption on the cell membrane is caused by charge interactions. Therefore, we attempted to use high concentration NaCl solution and low pH conditions for electrostatic shielding of LL-37, thereby eluting it from the cell membrane.
Build:
We prepared an extraction solution with the following composition: 20 % acetonitrile, pH adjusted to approximately 2 with 1 ‰ TFA, 1 M NaCl, and 1 ‰ Triton X-100. The organic solvent prevents LL-37 precipitation due to hydrophilic aggregation, low pH conditions protonate the cell membrane and LL-37 to generate electrostatic repulsion, NaCl provides high ionic strength to shield electrostatic adsorption between LL-37 and cell membrane, and Triton X-100 as a non-ionic surfactant is used to solubilize membrane components.
Test:
Fermentation samples were centrifuged and supernatant discarded, 1 mL of prepared extraction solution was added for cell disruption and extraction. After 1 h of treatment with a disruptor, the extraction was completed and centrifuged again, with supernatant and cell disruption precipitate taken separately for Western blot (WB) detection.
Fig. 5 | WB measurement of fermentation with different carbon sources , where M represents the marker.
WB results showed clear bands in the extraction supernatant, indicating that the extraction solution formula has certain effectiveness. However, although some LL-37 can be extracted from the disrupted precipitate, it still cannot completely elute LL-37 adsorbed on cell debris.
Learn:
• High salt strong acid extraction solution can extract LL-37 from cell disruption precipitate
• This method still cannot completely elute LL-37 adsorbed on cell debris
Design:
Given the limited efficiency of high salt low pH extraction solution, we plan to use surfactants to further enhance extraction effectiveness.
Build:
Referring to protein purification strategies, surfactants play an important role in membrane protein extraction. We prepared various acetonitrile salt water extraction solutions containing surfactants (concentration 1 %), with specific types and characteristics shown in the table below:
Table 1 | Properties of different surfactants
| Surfactant | Property Description |
|---|---|
| SDS | Anionic detergent, used to solubilize and denature proteins |
| LDS | Anionic detergent, superior solubility and denaturing effect compared to SDS |
| CHAPS | Zwitterionic detergent, suitable for membrane protein extraction under non-denaturing conditions and disrupting protein interactions, can be removed by dialysis |
| Tween-20 | Common non-ionic detergent, used for pre-extraction of cell membranes to remove peripheral proteins |
| Triton X-100 | Widely used non-ionic surfactant, can be used for membrane protein recovery under mild non-denaturing conditions |
Test:
Fermentation samples were centrifuged and supernatant discarded, 1 mL of different surfactant extraction solutions were added for mechanical disruption and 1 hour extraction. After extraction and centrifugation, supernatant and disruption precipitate were taken separately for WB detection.
Fig. 6 | WB measurement of extraction efficiency of different surfactants, where M represents the marker.
Learn:
• Most surfactants showed better extraction effects than high salt low pH extraction solution
• However, surfactants easily clog C18 chromatography columns, hindering LL-37 purification, so we ultimately chose high salt low pH extraction solution for experiments
The GAL promoter is a key element for synthesizing the target product LL-37. Through constructing reporter plasmids, random mutagenesis, and replacing core promoters, we search for efficient promoters to generate LL-37, thereby increasing cell factory production.
Design:
Initially, we had a plasmid with an engineered GAL promoter, typically consisting of a UAS region controlling both GAL1 and GAL10 promoters simultaneously. Without modification, we could only judge the strength of the GAL promoter by testing LL-37 production levels, but this process is time-consuming. Therefore, we planned to connect different fluorescent protein genes downstream of GAL1and GAL10 respectively to construct reporter plasmids, allowing subsequent assessment of promoter strength through bacterial fluorescence intensity.
Build:
We selected two common fluorescent protein genes with distinct colors: green eGFP and blue mTagBFP2. Through homologous recombination, we first independently connected green eGFP downstream of GAL1 and blue mTagBFP2 downstream of GAL10, constructing pESC-PGAL1-eGFP and pESC-PGAL10-mTagBFP2 plasmids respectively. After transforming these two plasmids into yeast strain CEN.PK113-7d ura3△, normal expression of respective fluorescent proteins was detected by fluorescence microscopy after induction. We then further combined the two, constructing a co-expression plasmid pESC-PGAL1-eGFP-PGAL10-mTagBFP2 controlled by the GAL promoter for green and blue fluorescent proteins (Fig. 7).
Fig. 7 | Plasmid construction for fluorescent protein expression.
Two plasmids were designed using homologous recombination. In pESC-PGAL1-eGFP, the green fluorescent protein gene eGFP was inserted downstream of the GAL1 promoter. In pESC-PGAL10-mTagBFP2, the blue fluorescent protein gene mTagBFP2 was inserted downstream of the GAL10 promoter.
Test:
We transformed the co-expression plasmid into yeast cells for testing. Fluorescence microscopy observations are shown in the figures (Fig. 8,Fig. 9). Although the plasmid still showed uneven expression with obvious cell-to-cell variation, it should be usable for analyzing overall fluorescence expression intensity.
Fig. 8 | Fluorescent protein expression results.
Fig. 9 | Co-expression of blue and green fluorescent proteins in Saccharomyces cerevisiae under galactose induction.
Learn:
• Construction of reporter plasmid for promoter modification is complete, allowing for subsequent promoter engineering.
Design:
A common method to enhance promoters is to change bases in the promoter sequence. We hoped to perform large-scale single-point mutations on the promoter to efficiently screen for positive mutations resulting in stronger promoters, so we chose to use error-prone PCR for random mutagenesis of the promoter.
· Error-prone PCR : Usually refers to PCR amplification using error-prone thermostable DNA polymerase, thereby introducing random mutations into target gene sequences.
Build:
First, we used a gene random mutagenesis kit and RandomMut DNA polymerase to perform error-prone PCR directly on the promoter, obtaining numerous mutants. Second, for the mutants, to ensure the modified promoter could normally initiate in both directions, we used high-throughput screening technology to simultaneously monitor eGFP driven by GAL1 and mTagBFP2 driven by GAL10 for 96 picked colonies. We then extracted plasmids from strains corresponding to wells with the highest relative fluorescence intensity and sent them for sequencing.
Test:
High-throughput measurements showed that most strains showed no observable fluorescence, while some strains showed only one type of fluorescence. The 4 wells with the highest relative fluorescence intensity for both types: A11, B10, H1, H12 were picked for plasmid extraction and sequencing.
Fluorescence kinetics results showed these four promoters basically maintained simultaneous protein expression in both directions, with intensity reaching relatively high levels. However, after sequencing, we found these four promoter sequences were identical to the original, with no mutations occurring. Even though we subsequently reconstructed the promoter mutation library, final sequencing results showed no positive mutant promoters were screened.
Fig. 10 | Second-round screening: Engineering modification of bidirectional GAL promoter and dual-fluorescence kinetics analysis with OD600 measurement
Learn:
• Single-point random mutagenesis of promoters showed limited effectiveness
• Perhaps overall replacement of promoter elements could achieve our goals
Design:
Through literature review, another common strategy to enhance promoters is replacing core promoters to construct hybrid promoters. Known commonly used strong promoters include TEF1 and *TDH3, and many hybrid promoters constructed using them show good performance. Therefore, we hoped to retain the UAS sequence of the GAL promoter while replacing GAL1 and GAL10 with TEF1 and TDH3* respectively.
Build:
Genscript, who sponsored us with 20kb gene synthesis, directly synthesized TEF1 and TDH3 promoters. We PCR amplified the GAP promoter, TEF1 fragment, and backbone with UAS_GAL, and connected them to obtain strong promoters with UAS sequences connected to core promoters on one side - (1) PGAL1-TDH3 and (2) PGAL1-TEF1. The specific construction process is shown in the figure (Fig. 11).
Fig. 11 | Mechanistic illustration of promoter engineering.
Test:
We transformed the two constructed plasmids into yeast, plated them, picked single colonies for induction and fluorescence intensity measurement. Ideally, colonies on plates would not fluoresce and would only emit light after induction, but in reality, promoter leakage occurred before induction, with visible fluorescence to the naked eye (Fig. 12). After induction, fluorescence intensity improved, but it affected yeast growth, slowing growth rate (Fig. 13)
Fig. 12 | Fluorescence of multiple yeast colonies on different plates under various promoter conditions.
Fig. 13 | Characterization of five heterozygous promoters.
Learn:
• The method of constructing hybrid promoters by replacing core promoters is effective, but we need to consider how to compensate for the inhibitory effect on growth rate
• Subsequently, we will construct TEF1 and TDH3 on both ends of UAS simultaneously and observe experimentally
Design:
In the earlier attempts, replacing the core promoter alone yielded limited improvement in transcriptional activity. According to literature, inserting additional UAS elements or tandemly repeating UAS sequences can significantly enhance promoter strength. In particular, the upstream activating sequence (UAS) from the Saccharomyces cerevisiae cell cycle gene CLB2 is known for its strong transcriptional activation capacity, especially during specific stages of the cell cycle, where it drives high levels of gene expression. Therefore, we attempted to insert varying copy numbers of UAS CLB2 elements into the original promoter to test the activity of the newly constructed hybrid promoters.
Build:
Based on the principle of homologous recombination, the UAS CLB2 fragment was PCR-amplified from the S. cerevisiae genome and ligated into the reporter plasmid backbone. This resulted in hybrid promoters containing one, two, or three tandem copies of UAS CLB2 (i.e., 3×UAS CLB2–PGAL). The schematic diagram of the construction strategy is shown in Fig. 14.
Test:
The three hybrid promoters were transformed into S. cerevisiae and plated. Single colonies were picked into 96-well plates for activation and induced fermentation. Fluorescence kinetics were monitored by measuring blue and green fluorescence intensities and OD₆₀₀ at 0 h (before induction), 24 h, 36 h, and 48 h post-induction. The results showed that all hybrid promoters exhibited a certain degree of basal (leaky) expression prior to induction. Nevertheless, it is noteworthy that the triple-tandem hybrid promoter displayed a substantial increase in fluorescence intensity compared to the original promoter, accompanied by a slight improvement in cell density.
Fig. 14 | Construction and functional evaluation of hybrid promoters containing tandem UAS CLB2 elements. (a) Schematic representation of hybrid promoter construction, showing the insertion of one, two, or three tandem copies of the UAS CLB2 upstream of the PGAL promoter. Homologous recombination was used to assemble the hybrid promoters in a reporter plasmid. (b) Time-course measurements of fluorescence intensity (green and blue) and cell growth (OD₆₀₀) for strains harboring 1×, 2×, and 3× UAS CLB2–PGAL promoters at 0, 24, 36, and 48 h post-induction. Data indicate that tandem repetition of UAS CLB2 enhances promoter strength, with the triple-tandem promoter showing the highest fluorescence and slightly improved cell density.
Learn:
Previously, using S. cerevisiae engineering strain CEN.PK113-7D as chassis cells, we successfully expressed LL-37 intracellularly, but found that LL-37 could not be secreted extracellularly. This increases the difficulty of subsequent experiments for improving production and performance testing. We hope to help LL-37 secretion through signal peptides and fusion molecular chaperone.
Design:
Through literature review and previous laboratory experience, we selected leader signal peptides from five genes to construct plasmids that could potentially guide LL-37 extracellularly without affecting LL-37 synthesis.
Build:
We constructed 6 high-efficiency expression plasmids with five leader signal peptide genes (no sp, SUC2 sp, INU1 sp, PHO5 sp, SCW11 sp, HECH sp). Besides CEN.PK113-7D, we also included CPK04 and BY4741 strains as chassis strains, each chemically transformed with the six plasmids for induction.
Test:
We performed WB detection on supernatant and precipitate from 24h fermentation samples but did not detect LL-37 in the supernatant (Fig. 15). Possible reasons we analyzed include:
① All products were degraded in the secretion pathway (mainly possibly during the process from endoplasmic reticulum to Golgi apparatus).
② Degradation by extracellular proteases.
③ Small peptides were recognized as misfolded by the endoplasmic reticulum and retained.
Fig. 15 | Western blot analysis of target protein expression with different signal peptides in various yeast strains
Learn:
• Possibly due to LL-37's small molecular weight, these five signal peptides cannot help LL-37 secretion
• Besides these five common signal peptides, perhaps other signal peptides might work, and knocking out related protease genes might also be a good strategy
Design:
The small molecular weight characteristic of LL-37 might affect its secretion, while the complete α-factor secretion signal sequence is as long as 89 AA, which can guide target proteins through the endoplasmic reticulum-Golgi secretion pathway. Since this complete signal peptide is large, it might escape yeast's quality control system. Kex2 protease cleavage occurs after the binary KR sequence, then dipeptidyl aminopeptidase A (Dpp) removes the EA dipeptide to obtain secreted LL-37.
Build:
Fig. 16 | Nucleotide sequence of the α-factor signal peptide.
We constructed complete α-factor secretion signal-LL37 (Fig. 16), transformed it into E. coli for verification, and after amplification, transformed it into yeast strains for induction and WB detection.
Test:
Although we still couldn't detect corresponding bands in the supernatant, whole cell lysate showed two distinct bands: 14.6 kDa for the unprocessed original translated peptide, and the significantly deeper 5.4 kDa band for the mature cleaved peptide (Fig. 17), indicating that the α-factor signal peptide successfully guided LL-37 to the Golgi apparatus, where it was cleaved by the membrane-bound serine protease Kex2 localized in the Golgi during late transport.
Fig. 17 | Western blot analysis of LL-37 with or without α-factor signal peptide.
Combined with previous experimental results, we hypothesize that LL-37 is very likely secreted in later stages and after being expelled from cells, re-adsorbs to the yeast cell membrane.
Learn:
• Subsequently, we will construct α factor-LL37-eGFP to observe whether fluorescence is mostly localized to the cell membrane in late induction, with small amounts appearing as aggregates inside cells, distributed in vesicle-like patterns on the Golgi apparatus
• Use low pH high salt solution to elute LL-37 from cell debris, detect LL-37 content in washing solution, and use in later sample preparation to enhance yield
Design:
The small ubiquitin-like modifier (SUMO, 11 kDa) fusion tag has been successfully expressed in E. coli for a variety of proteins, where it was shown to enhance protein solubility and increase yield. In addition, the tag can be specifically cleaved by the SUMO protease (ULP1, encoded by S. cerevisiae, which cannot be knocked out). We attempted to fuse LL-37 with a SUMO tag to shield its strong cationic property, aiming to facilitate its secretion into the extracellular medium.
Build:
We amplified DNA fragments by PCR and assembled them with the vector backbone to construct the plasmid MFα-SUMO-LL37-His. After verification in E. coli, the plasmid was extracted and transformed into S. cerevisiae. Single colonies were collected and subjected to activation and induction fermentation.
Test:
Western blotting was performed twice. LL-37 was not detected in the supernatant, cell lysates, or total cell extracts. We suspected either plasmid transformation failure or premature cleavage by the SUMO protease.
Learn:
• SUMO protease exhibits high catalytic efficiency.
• The SUMO fusion strategy may still be effective, and repositioning the SUMO tag within the plasmid construct may improve the outcome.
Design:
To improve LL-37 secretion and activity, we hypothesized that intracellular proteases might degrade antimicrobial peptide LL-37, thereby reducing its yield. Through literature mining, we screened multiple candidate protease genes that might participate in LL-37 degradation. We sequentially knocked out each candidate gene in pre-screened chassis strains. Although CRISPR-Cas9 technology is commonly used for gene knockout, its off-target effects are high and might cause DNA damage. Therefore, we adopted the Cre-loxP method - achieving precise operations through pre-designed loxP sites and Cre plasmids.
Build:
We first constructed a loxP-URA3-loxP sequence containing homology arms targeting the open reading frame of each candidate gene for replacement (Fig. 18). The constructed plasmid (e.g. BBa_25NP2VZ8,BBa_253NFL7O) was transformed into E. coli DH5α for amplification and verified by colony PCR. Verified plasmids were extracted and transformed into S. cerevisiae, with colonies screened on YPD medium. To recycle the URA tag, colonies were co-cultured with Cre plasmid carrying KanMX selection marker. Recombinant strains were successfully screened based on ability to grow in G418 medium but not in SC-URA medium. Finally, the Cre plasmid was cleared from yeast strains.
Fig. 18 | A loxP-URA3-loxP sequence used to knock out genes.
Test:Knockout strains (including 17 protease gene deletion strains3, 4, GAL80 knockout strain, and background strain BY4741) were transformed with pESC expression plasmid (BBa_25K5U7FH) respectively, and OD600 values were measured after 48 hours of culture. Growth of each strain was detected by spectrophotometer, and LL-37 yield was quantified using Western blotting (Fig. 19) and High Performance Liquid Chromatography (HPLC, Fig. 20). This method effectively distinguished whether changes in LL-37 titer were due to differences in expression efficiency or cell density.
Fig. 19 | WB results of LL-37 production.
Fig. 20 | OD600 and HPLC results
Learn:
• From HPLC results, LL-37 production in PEP4 knockout and PRB1 knockout strains showed significant improvement compared to the control group
• These candidate proteases indeed play important degradation roles in LL-37 synthesis or secretion. Based on these findings, we plan to continue screening more genes with similar functions and construct multi-gene knockout strains to evaluate whether synergistic effects exist in LL-37 secretion
Design:
Through literature analysis and preliminary experiments, we identified four genes with the most significant impact on LL-37 production (Table 2). To optimize the expression system, we decided to adopt a combinatorial knockout strategy to explore whether multiple gene deletions could synergistically enhance LL-37 secretion efficiency. In this iteration, we focused on constructing double-gene knockout strains as the first step of the combinatorial knockout strategy.
Table 2 | four candidate gene and their function
| Gene name | Gene function |
|---|---|
| PEP4 | Lactate aspartate protease A |
| PRB1 | Lubricinase B（H3 N-terminus endopeptidase） |
| YSP3 | Serine proteinase |
| YPS1 | Yapsin protease family; Plasma aspartate aminotransferase |
Build:
Based on single-gene knockout strains, we transformed PCR products containing corresponding knockout cassettes into these strains and cultured on SC-URA medium plates. After colony formation, about 20 colonies were selected for clone PCR screening. Compared to wild-type control, correctly transformed colonies showed larger PCR amplification products, indicating successful integration. We then followed the same process described in Cycle 1 for URA3 marker recycling and Cre plasmid repair.
Test:
We collected the 10 constructed gene knockout strains after 24h and 48h shake flask fermentation and detected results by HPLC.
Learn:
After confirming gene expression by WB detection, we detected 48h fermentation yield by HPLC and found that double-knockout strains showed tremendous yield improvement, increasing from about 300 μg/L to around 2000 μg/L in a 25mL shake flask system.
Design:
The yeast knockout6 library we used contains approximately 5000 strains, making it very difficult to test LL-37 production in each strain individually. Therefore, we adopted a semi-rational screening strategy: randomly selecting partial strains from the mixed library for fermentation experiments, screening strains with higher LL-37 production. Subsequently, sequencing analysis of these high-producing strains was performed to determine their corresponding gene deletion sites, thereby achieving large-scale random screening of the yeast genome.
Build:
We transferred strains selected from 96-well plates to freshly prepared medium to ensure their viability. Subsequently, strains were inoculated in batches into small-scale fermentation medium (SC-URA) under identical conditions to minimize inter-group differences. Each batch of cultures was cultivated for 48 hours before sample collection for LL-37 production analysis.
Test:
High-performance liquid chromatography (HPLC) was used to analyze LL-37 production from obtained strains (Fig. 21).
Fig. 21 | Distribution of Z-scores for all 384 strains. The histogram depicts the frequency distribution using a bin width of 0.2. The red bars denote strains that exceed the significance threshold (Z > 2)
The data distribution approximately follows a normal distribution, with wild-type strain LL-37 production as the mean. Strains to the right of zero show higher LL-37 production, while those on the left show lower production. We selected single-gene deletion strains with Z-values greater than 2, which showed significantly increased LL-37 expression levels.
Learn:
Based on this, we plan to:
• Sequence high-producing strains to confirm their specific genotypes
• Conduct attribution analysis to elucidate mechanisms of production improvement
• Combine these advantageous gene deletions with targets obtained from previous rational screening (such as key protease deletions) to further enhance LL-37 production
Design:
The S. cerevisiae genome is rich in Ty1 retrotransposons surrounded by δ sequences, which can serve as homologous recombination sites for random multi-copy integration5. We utilized this natural genomic feature to design a strategy: enhancing LL-37 expression by integrating multiple LL-37 expression cassettes into these δ sites. In this process, we used URA3 as a selection marker to successfully construct stable multi-copy expression strains.
Build:
Using δ sequences on both sides of retrotransposon TY1 as homology arms（Ty1-LTR-UP，Ty1-LTR-down）, with URA3 as selection marker, we achieved multi-copy integration in the genome, constructing the multi-copy integration plasmid TOPO-HOMO-TY1-URA3-pGAL1-LL37.
Test:
Fermentation was performed in 3 mL induction medium, with samples taken at 24 h and 48 h for analysis, verified by Western Blot (WB) (Fig. 22).
Fig. 22 | WB results of Construction strategies of expression systems；P: 2μ origin-based episomal plasmid with URA3 selection marker; G: Genomic integration via δ retrotransposon sequences at TY1 locus.
Learn:
• Successful implementation of multi-copy integration, showing advantages in eliminating plasmid-related burden (such as selection pressure)
• Expression levels from free plasmids were significantly higher than multi-copy integration strains, possibly because plasmid copy numbers still exceed genomic integration copy numbers
• Subsequent experiments can be further conducted in rich nutrient medium
Design & Build:
Based on URA3-marked multi-copy integration technology, we further attempted to increase LL-37 copy number and expression levels. Previous studies have shown that high concentrations of G418 can enrich yeast strains carrying more integration copies2. Therefore, we modified the plasmid by replacing the URA3 marker with KanMX, constructing the multi-copy integration plasmid TOPO-HOMO-TY1-KanMX-pGAL1-LL37-pGAL10-LL37.
Test:
WB detection confirmed correct LL-37 protein expression with clear bands (Fig. 23). In subsequent studies, we will consider combining this high-copy integration strategy with protease knockout strains for further optimization.
Fig. 23 | Multi copy site integration of LL37 based on KanMX screening markers
Learn:
• Replacing URA3 marker with KanMX can increase copy number, thereby increasing LL-37 production
• Plan to subsequently integrate high-copy with protease-deficient strains to further enhance LL-37 production
Design:
To further enhance LL-37 production and verify its actual performance in fermenters, we combined previously constructed multi-gene knockout strains with multi-copy integration strategies to obtain recombinant strains with stronger expression capability, and conducted production testing at fermenter scale.
Build:
Based on existing quadruple-knockout strains, we additionally knocked out the GAL80 gene and achieved multi-copy integration at yeast Ty1 repeat sequence sites to enhance target gene expression levels. The final strain was yps1△;prb1△;pep4△;ysp3△;gal80△::LoxP-URA3; ty1::LoxP-KanMX-GAL1-LL37-His-GAL10-LL37-His.
Test:
We first conducted fermentation pre-experiments in shake flasks, detecting and comparing production by WB and HPLC to screen the best-performing candidate strains. Subsequently, the selected strain was inoculated into a 5 L fermenter (working volume 2.5 L), with fermentation conditions set as:
• Basic medium: YPD + 18 g/L potassium dihydrogen phosphate, 10 g/L magnesium sulfate heptahydrate, 7 g/L potassium sulfate • Feed: 50% glucose solution
• Dissolved oxygen: maintained at 30%
• pH: stably controlled at 5.5
During fermentation, samples were taken every 3 hours for the first 48 hours, with sampling frequency appropriately reduced later, for a total fermentation duration of approximately 96 hours. Up to 48 hours, OD600, residual sugar, and WB detection every 6 hours are shown in （Fig. 24）.
Fig. 24 | Data from a 5L fermentation tank and WB testing of samples taken within the first 48 hours
Learn:
• Glucose consumption during fermentation was too rapid, causing premature LL37-His production
• Since LL-37 cannot be expelled and there are still many endogenous proteases, as cell growth progresses, production and degradation basically reach equilibrium, making it difficult to accumulate in large quantities. Finally, OD600 stabilized around 156, with production difficult to reach 100 mg/L level, but peak accumulation should be at 30-40 mg/L level
To achieve industrial production of LL-37, our chosen carbon source must balance cost, yield, and expression stability. When we expressed LL-37 in Cycle 1, we used the GAL system requiring galactose induction, meaning that in actual fermentation, we need to distinguish between growth and induction carbon sources, and maintain LL-37 expression through fed-batch feeding. Although this method is feasible under laboratory conditions, it increases fermentation process complexity and overall production costs in industrial scale-up production6.
Therefore, in this cycle, we removed glucose repression effects by knocking out key regulatory factors, and conducted fermentation testing with 8 carbon sources in the modified derepressed chassis, ultimately determining carbon source substrates suitable for derepressed chassis fermentation.
Design:
Fig. 25 | Mechanism of GAL promoter regulation and glucose repression in Saccharomyces cerevisiae.
In wild yeast, glucose signals are transmitted through the Snf1/Mig1 pathway to regulate gene expression. Without glucose, Mig1 is phosphorylated by Snf1 and retained in the cytoplasm. If galactose is present, Gal3 binds to Gal80, relieving Gal80's inhibition of Gal4, allowing Gal4 to fully exert transcriptional activation. After adding glucose, Snf1 undergoes dephosphorylation and inactivation, causing Mig1 dephosphorylation and nuclear entry, binding to promoters and forming the Cyc8/Ssn6-Tup1 complex to inhibit transcription7,8. As glucose is depleted, Snf1 is reactivated, phosphorylating Mig1, and repressed genes are re-expressed9.
After consultation with our PI Anqi Chen, to avoid metabolic disorders caused by Snf1 knockout, we decided to use the Cre-loxP system to knock out GAL80, completely relieving Gal80's inhibition of Gal4, thereby eliminating galactose-only induction10, 11, 12.
Build:
We selected partial upstream and downstream sequences of GAL80 from the SGD database, combined with loxP sequences and URA3 tags to construct GAL80 and MIG1 knockout components (BBa_25KID167, BBa_2555F5WC). After verifying successful knockout by nucleic acid gel, we used Cre components to recycle the URA3 tag, obtaining the gal80Δ derepressed chassis.
Test:
To verify actual derepression effects, we inoculated the derepressed chassis into 25mL fermentation medium containing 2% glucose for shake flask fermentation, sampling every 3 hours to detect residual glucose and OD600 in samples, while monitoring LL-37 production changes over fermentation time by WB.
Fig. 26 | Time-course analysis of LL-37 production during shake-flask fermentation in glucose medium
A derepressed yeast chassis was inoculated into 25 mL medium containing 2% glucose and cultured under shake-flask conditions. Samples were taken every 3 hours to measure residual glucose concentration and OD₆₀₀, while LL-37 production was monitored by Western blot. The results show the dynamic relationship between substrate consumption, cell growth, and peptide secretion during fermentation.
From the final results (Fig. 26), it can be seen that GAL80 knockout indeed remove the restriction of galactose-only induction. The GAL system in the derepressed chassis can begin expressing LL-37 when glucose is nearly depleted without galactose induction, with expression gradually increasing over time. This indicates that our constructed derepressed chassis has potential to simplify fermentation processes, providing a feasible chassis strain for future industrial production of LL-37.
Since fermentation carbon source selection directly affects industrial production costs, we still need to test the expression effects of the derepressed chassis under various carbon source conditions and comprehensively consider production costs to screen for optimal fermentation carbon sources.
Learn:
• GAL80 knockout indeed removes the restriction of galactose-only induction
• Subsequently, we will further screen for optimal fermentation carbon sources
Design:
Considering differences in cost, metabolic pathways, and effects on LL-37 expression among different carbon sources, we designed comparative experiments using 8 different carbon sources as fermentation substrates, including glucose, fructose, maltose, lactose, galactose, raffinose, mannose, and sucrose13, 14, 15, 16. By comparing their utilization efficiency in the derepressed chassis and final LL-37 expression levels, we hope to screen for fermentation carbon sources that ensure both high yield and economic feasibility.
Build:
Using gal80Δ strain as the production strain, we inoculated it into 25mL shake flask media containing single carbon sources, maintaining identical initial inoculation amounts and fermentation conditions. Carbon source concentration in each medium was unified at 2% (for more specific preparation methods, please refer to our Protocol section).
Test:
We sampled during fermentation to detect OD600 to record chassis cell growth status, and used the extraction method obtained from Cycle 1 combined with WB to detect LL-37 expression levels, finally obtaining quantitative LL-37 results by HPLC.
Fig. 27 | Growth curves of yeast under different carbon sources
OD600 data showed that chassis cells grew poorly in medium with lactose as the sole carbon source, with curves barely rising, indicating that chassis cells basically cannot utilize lactose, which is consistent with our research findings^1^. Additionally, in medium with raffinose as carbon source, chassis cells showed significantly delayed entry into log phase with lower final OD600, galactose medium final OD600 was about 1.6, but log phase appearance was slightly delayed. Besides these, glucose, sucrose, mannose, maltose, and fructose curves almost overlapped, with final OD600 all stabilizing between 1.6-2.0, indicating these carbon sources support efficient chassis cell growth.
We also processed 36 h fermentation broth according to Cycle 1 methods for WB and HPLC detection. Results (Fig. 27, Fig. 28) show that chassis cells showed high LL-37 expression in media with glucose, fructose, sucrose, and maltose as sole carbon sources.
Due to glucose's advantages of low price, wide availability, and convenient procurement, we ultimately selected glucose as the fermentation carbon source.
Fig. 28 | HPLC analysis of LL-37 production after 36 h fermentation with different carbon sources.
Learn:
• Glucose is the most cost-effective fermentation carbon source
• Glucose and maltose as fermentation carbon sources yielded the highest LL-37 production, but glucose is more cost-effective, so we ultimately chose glucose as the fermentation carbon source
• Chassis cells basically cannot utilize lactose as a fermentation carbon source
Obtaining high-purity LL-37 is a prerequisite for subsequent protein function verification. However, unpurified fermentation broth contains large amounts of host proteins, metabolites, and medium components. These impurities not only interfere with detection accuracy but may also affect LL-37 activity assessment. Therefore, we designed a purification process for LL-37 to ensure more reliable subsequent experimental data and accumulate experience for future scale-up production processes. Based on this goal, we attempted both nickel column purification and magnetic bead purification strategies17, 18, 19, 20.
Design:
We wanted to use Ni-NTA nickel affinity chromatography columns to purify LL-37 with 6×His tags. Considering this method is widely used in laboratories with high specificity and separation efficiency, theoretically LL-37 can be separated from fermentation broth through His tag binding to Ni2+. We expected this strategy to obtain high-purity LL-37.
Build:
We used the multi-copy integration strain CEN.PK113-7D ura3△,ty1△::Loxp-KanMX-GAl1-LL37(6 x His) constructed in Cycle 2.5 for fermentation in YP+2% Lactose+Raffinose, collected fermentation broth at 24h for disruption and Ni column purification, attempting gradient washing and elution . For more specific procedures, please refer to the Protocol.
Fig. 29 | The plasmid of TOPO-HOMO-Ty1-kanMX-pGAL1-LL37
Test:
We performed WB detection on original samples, flow-through, and gradient concentration elution solutions (Fig. 30). The WB and CBS results of flow-through and original samples were basically identical, with no bands in all subsequent washing and elution steps, indicating that LL-37 could hardly bind to the nickel column.
Fig. 30 | SDS-PAGE and Western blot analysis of LL-37 purified with different imidazole concentrations using Ni-NTA chromatography.
We suspected that aging or repeated use of the nickel column packing might prevent LL-37 from binding normally to the nickel column. We subsequently attempted nickel column purification again with a newly purchased Ni column, with WB results shown in (Fig. 30):
We found that even with a new Ni column for purification, LL-37 still bound poorly to the column. Possible reasons include: LL37-6×His is too small and easily flows through gaps, while low concentration 50mM imidazole might cause excessive elution.
This result indicates that Ni-NTA columns are not suitable for LL-37 separation and purification, requiring alternative methods with stronger binding capacity and more flexible operation.
Learn:
• LL-37 cannot bind to nickel columns, presumably because LL37-6xHis is too small to bind to columns
• Ni-NTA columns are not suitable for LL-37 separation and purification, requiring replacement with alternative methods with stronger binding capacity
Design:
After the failed nickel column attempt, we redesigned our purification approach. Magnetic bead purification methods have higher binding sensitivity and more convenient operation, especially suitable for small molecules or target proteins with low expression levels. We expected this method to solve the problem of LL-37 not binding in column chromatography.
Build:
We used Ni-NTA magnetic beads as affinity medium, directly mixing expressed LL-37 samples with magnetic beads for incubation, using magnetic separation for non-specific heteroprotein removal, and obtaining target protein through stepwise elution with low and high concentration imidazole.
Test:
Western blot (WB) detection showed clear LL-37 bands in the magnetic bead elution group, verifying that magnetic beads could effectively bind and recover target protein. Subsequent HPLC detection further confirmed the correctness and relative purity of purified products.
Learn:
• LL-37 can be purified through magnetic bead purification methods
Design:
In this section, we hope to enhance the antimicrobial effect of antimicrobial peptide LL-37. In initial antimicrobial tests, LL-37 showed moderate antimicrobial effects. To improve its antimicrobial performance, our dry lab group designed the MADS model (see Model for details) to generate LL-37 variant sequences with stronger antimicrobial activity（Variant-1、Variant-2、Variant-3、Variant-4） (Fig. 31). However, chemical synthesis of these variant sequences is too expensive and time-consuming, and biosynthesis yields are insufficient for antimicrobial testing. Therefore, we planned to use the cell-free system ProteinFactory kit to synthesize these sequences and planned to use minimum inhibitory concentration method to test their antimicrobial performance. Before using ProteinFactory to synthesize our LL-37, we planned to first synthesize eGFP and His-tagged LL-37, determining ProteinFactory's synthesis effectiveness through eGFP fluorescence intensity and WB detection of LL-37(6xHis), how long it takes for LL-37 production to reach peak levels.
Fig. 31 | Predicted structural model of different LL-37 variants based on computational modeling.
Build:
We first constructed pD2P plasmids with eGFP and LL-37 (6xHis) using restriction enzyme ligation to insert LL-37 sequence into the MCS sequence on pD2P-1.08t-8His-eGFP vector, constructing recombinant vector, transforming into cloning host Trelief 5α, using Ampr resistance to screen monoclones, and finally confirming by sequencing to obtain positive monoclones. These were sequentially added to DNA Amplifier and ProteinFactory Rxn reaction systems, with eGFP fluorescence intensity detected using a microplate reader (see Protocol for more details) .
Test:
Fig. 32 | Fluorescence kinetics of the D2P reaction. Fluorescence spectra were acquired every 10 minutes using an excitation wavelength of 488 nm and an emission wavelength of 512 nm.
According to microplate reader detection results and visual observation, fluorescence intensity reached peak and stabilized at approximately 3-4 hours, with reaction solution in wells visibly turning green. WB detection of LL-37 (6xHis) reached detection limit at approximately one and a half hours.
Learn:
• Cell-free factory ProteinFactory can be used to synthesize eGFP and LL-37, with slower synthesis of eGFP and faster synthesis of LL-37.
Design:
After confirming that the cell-free system ProteinFactory can be used to synthesize LL-37 (6xHis) and determining the time required for production to reach peak levels, we planned to use ProteinFactory to synthesize LL-37 and its variants and eGFP, and attempt to use the reaction solution for antimicrobial experiments against E. coli.
Build:
The steps for synthesizing LL-37 using ProteinFactory were the same as the construction method in Iteration 1 and using the pD2P plasmid expressing LL-37 variant 1 as a template, we constructed different plasmids expressing LL-37 variants through circular PCR. These were then added to 96-well plates containing E. coli cultures, testing growth curves to observe LL-37's antimicrobial effects.
Test:
OD600 of wells with added eGFP reaction solution was higher than those without, while wells with added LL37 reaction solution (multiple variants, including negative control wells without bacterial inoculation) showed significantly different abnormal OD600. Visual observation revealed small round particulate flocculent materials, with results shown in Figures 33 and 34 (Fig. 33, Fig. 34).
Fig. 33 | Antibacterial testing of LB medium and mixed reaction solution.
Fig. 34 | Photograph of a 96-well microtiter plate containing antimicrobial assay solutions
Learn:
• Addition of eGFP control reaction solution clearly promoted E. coli growth
• The ProteinFactory reaction system contains large amounts of energy and substrates needed for transcription and translation processes, being nutritionally rich, requiring subsequent centrifugation to remove reaction solution
• Microscopic observation might reveal the cause of abnormal OD600 in LL-37 reaction solution
Design:
We planned to use inverted fluorescence microscopy to observe products from Iteration 2, hoping to find the cause of abnormal OD600 in LL-37 reaction solution through microscopic examination results.
Build:
We prepared slides of products from Iteration 2 - LB+E. coli, LB+E. coli+eGFP, LB+E. coli+LL-37, LB+LL-37, and observed them under inverted fluorescence microscopy.
Test:
Fig. 35 | Microscopic antibacterial experiment culture medium.
Results are shown in Figure 35, where A is LB medium, B is LB+eGFP, C is LB+LL37 and variants, D is LB+LL37+no bacterial inoculation. B showed significantly higher E. coli density than A, few E. coli were observed in C, small round particulate flocculent materials appeared in both D and C. Multiple experiments found these slowly aggregated over time, with aggregation speed varying greatly at different concentration additions, and contamination causes were excluded (Fig. 35).
Learn:
• We hypothesize that due to excessive concentration of LL-37 variants synthesized by ProteinFactory, protein crystallization precipitated and interfered with OD600 measurement
• Using plate assays might help determine antimicrobial effects of LL-37 variants
Design:
After observing round particulate flocculent materials at the bottom of wells with added LL-37 reaction solution, we switched to plate assays to determine antimicrobial effects of LL-37 and its variants.
Build:
We still used ProteinFactory kit to synthesize LL-37 and its variants and eGFP, with specific steps the same as in Iteration 1. LL-37 and variants, eGFP, and water were mixed with E. coli, then samples were taken at 0h and 3h for plating and counting, with eGFP and water as controls, calculating survival rates.
Test:
After adding LL-37 variants, it was clearly observed that E. coli colony numbers at T=3h were fewer than at T=0h, proving that LL-37 has certain antimicrobial effects against E. coli (Fig. 36, Fig. 37).
Fig. 36 | Survival rates of bacterial cells after treatment with LL-37 and its variants, determined by plate count assay
Fig. 37 | Survival rates of bacterial cells after treatment with LL-37 and its variants, determined by plate count assay
Learn:
• Switching to agar diffusion inhibition zone tests helps determine antimicrobial effects of LL-37 variants
• The antimicrobial effect of LL-37 variant 1 may be superior to that of the original LL-37; however, further analysis could not be conducted due to time limitations.
Wang, J., Atolia, E., Hua, B., Savir, Y., Escalante-Chong, R., & Springer, M. (2015). Natural variation in preparation for nutrient depletion reveals a cost-benefit tradeoff. PLoS Biol, 13(1), e1002041. https://doi.org/10.1371/journal.pbio.1002041
Kim, H., Kim, M., Im, S.-K., & Fang, S. (2018). Mouse Cre-LoxP system: General principles to determine tissue-specific roles of target genes. Laboratory Animal Research, 34**(4), 147–159. https://doi.org/10.5625/lar.2018.34.4.147
Sauer, B., & Henderson, N. (1988). Site-specific DNA recombination in mammalian cells by the Cre recombinase of bacteriophage P1. Proceedings of the National Academy of Sciences of the United States of America, 85(14), 5166–5170. https://doi.org/10.1073/pnas.85.14.5166
Rose, M. (1983). Structure and function of the yeast URA3 gene. Journal of Molecular Biology, 167(3), 391–412. https://doi.org/10.1016/S0022-2836(83)80193-4
Gancedo J. M. (1998). Yeast carbon catabolite repression. Microbiology and molecular biology reviews : MMBR, 62(2), 334–361. https://doi.org/10.1128/MMBR.62.2.334-361.1998
