Tianjin iGEM 2025

Overview

Our engineering workflow is divided into two parts: Part 1 focuses on antifreeze protein (AFP) engineering, and Part 2 on the development of an antifreeze chassis system. In general, Part 2 is an iteration of Part 1. It is important to note that, due to experimental timelines, some tasks across the two parts were conducted concurrently.

In Part 1, we began by characterizing the activity of eleven natural AFPs. We then performed rational modifications to these native proteins through point mutations and domain fusions, guided by in dry-lab design and simulations. Furthermore, we employed inverse folding and de novo design strategies to create a series of artificial AFPs. Finally, the safety of AFPs were confirmed through cytotoxicity assays.

In Part 2, we utilized a yeast surface display system in the EBY100 strain. We constructed chassis systems expressing the eleven natural AFPs and subsequently expanded the system to accommodate non-natural AFPs, including those generated by strategies such as point mutations and domain engineering. To further enhance the overall performance of the system, we introduced a fortified trehalose metabolic pathway. This resulted in a dual-mode antifreeze chassis, leveraging the combined protective effects of both trehalose and the displayed antifreeze proteins.

Cycle 1 The Construction of Antifreeze Proteins in Escherichia coli

Design

In order to obtain antifreeze proteins with excellent antifreeze properties, based on systematic literature research and in-depth communication with Professor Zhang Lei, this study finally selected 11 types of antifreeze proteins representative of different species, which are of family significance, including 4KE2 [1], 2ZIB [2], 1UCS [3], 3ULT [4], 1M8N [5], 2PNE [6], 3P4G [7], 3WP9 [8], 4NU2 [9], 6A8K [10], and 4MTJ [11]. To facilitate subsequent experimental analysis and comparison of results, these proteins are respectively named AFP1 to AFP11. During the construction of the expression system, a lactose-inducible regulatory module was incorporated to enable precise and tunable expression of the target AFP genes (Fig. 1).

Figure 1. Genetic circuits constructed on pET-28a-Target

Build

We synthesized the vector pET-28a-Target through the company's commercial process. It harbors a lactose operator system and the target gene expression module.

Test

To obtain the target protein, the primary seed culture was incubated overnight and subsequently diluted at a 1:100 inoculum ratio into fresh liquid medium. After 4 hours of growth, protein expression was induced by the addition of IPTG. The cells were harvested by centrifugation after 16 hours of induction, resuspended in phosphate-buffered saline (PBS), and lysed via ultrasonication to obtain the crude enzyme extract.

Considering that the protein purification process is time-consuming, prone to failure and results in yield loss, we adjusted our characterization strategy and adopted a more efficient identification method. As an alternative, we lysed the cells to obtain the crude enzyme solution containing antifreeze proteins, and then conducted subsequent characterization using the crude enzyme solution containing the antifreeze protein.

The ice recrystallization inhibition (IRI) activity of the crude enzyme solutions was quantified. ImageJ software was used to analyze the Mean Grain Area (MGA) of the ice crystals. Specifically, eight random areas (each area 176 μm × 176 μm) were selected from the 30-minute image, and the number and average size of ice crystals in each area were counted. Each sample was tested in triplicate, and the final results were averaged. Bar graphs with overlaid data points were plotted using Origin software.

Since the antifreeze proteins we selected have already been characterized in previous studies regarding their performance in IRI, in order to avoid redundant experiments and improve the research efficiency, we randomly selected some of the antifreeze proteins for verification.

Figure 2 IRI activity of crude enzyme (AFP4/8/9/10)

Figure 2. Analysis of IRI activity in crude enzyme solutions of selected AFPs. (a) Ice crystal morphology following incubation at -6 °C for 30 minutes. (b) Measurement of the average ice crystal area(μm²)

The results show that the antifreeze proteins we selected have demonstrated excellent antifreeze properties after being characterized and verified.(Fig. 2).

Learn

The experimental results confirmed that the selected AFPs exhibited the expected antifreeze capability. Based on the previous research data, we concluded that the 11 antifreeze proteins we chose could all demonstrate excellent antifreeze activity. To further optimize the performance of the antifreeze proteins, we will subsequently select some antifreeze proteins with significant helical characteristics and introduce site-directed mutations.

Cycle 2 Site-directed Mutagenesis Protein Design

Design

To enhance AFPs performance, we employed the ESM-1v model to analyze evolutionary conservation and identify potential mutation sites. This analysis revealed two distinct types of key residues: highly conserved residues critical for maintaining the structural core, and more variable surface residues suitable for functional optimization (see Dry Lab section for details). Based on this finding, mutations in the variable surface regions were prioritized to explore functional enhancement while preserving the native protein fold. Four candidates (AFP4, AFP8, AFP9, and AFP10), selected for their well-defined structures and stable helices, were used as model proteins for subsequent site-directed mutations (Fig. 3). The resulting mutants were designated MUT4, MUT8, MUT9, and MUT10, respectively.

Figure 3 Sequence alignments before/after mutation

Figure 3. Sequence alignment of wild-type and mutant antifreeze protein genes. (a) Alignment of AFP4 and MUT4, depicting mutation sites H65N, H92N, and T94N. (b) Alignment of AFP8 and MUT8, depicting mutation sites C66G, M132A, and M140F. (c) Alignment of AFP9 and MUT9, depicting mutation sites D62I and C86G. (d) Alignment of AFP10 and MUT10, depicting mutation sites R133P, S189Q, and I215G.

Build

We synthesized the vector pET-28a-MUT4, pET-28a-MUT8, pET-28a-MUT9, and pET-28a-MUT10 through the company's commercial process, each containing the complete target gene expression module.

Test

Comparative analysis was performed to evaluate the functional enhancement of the engineered proteins compared with their wild-type counterparts under identical conditions.

Figure 4 IRI activity of AFP vs MUT series

Figure 4. Comparison of IRI activity between wild-type AFPs and their mutants. (a) Ice crystal morphology after freezing at -6 °C for 30 min. (b) Measurement of the average ice crystal area (µm²).

This modified protein still retains some anti-freezing activity, but its ability to inhibit the recrystallization of ice crystals is lower than that of the original protein (Fig. 4).

Learn

Although the designed mutants did not show an improvement in freezing resistance, it is worth noting that their performance was always superior to that of the wild-type Escherichia coli. This observation led us to propose the hypothesis that we have limited sites for point mutations, but in practical engineering strategies, such protein evolution requires multiple rounds of iteration and expansion of the mutation range. However, due to experimental cycle constraints, it's difficult to complete such a large workload. Based on this hypothesis, in the subsequent experiments, we adopted the domain fusion strategy to combine different domains of these proteins with a distinct helical structure, in order to create a fusion protein with stronger functionality.

Cycle 3 Protein Domain Fusion

Design

Based on the previous research results, we combined the AFP domains from different biological sources (such as fish, beetles, and Antarctic bacteria) with different structural types (including α-helices, β-helices, and β-sheet structures) to rationally design fusion proteins, aiming to systematically explore the potential synergistic effects on antifreeze ability [12]. To optimize the spatial arrangement of the domains and enhance the ice-binding ability, we added specific-length flexible linkers between different functional modules. Experimental calculations and simulations have shown that all the designed fusion proteins had favorable theoretical properties: their folding probability exceeded 70%, the average predicted confidence was higher than 65%, and they usually exhibited high solubility scale - indicating good solubility and potential antifreeze activity (Fig. 5, Table1). Moreover, the relatively low folding free energy further indicated that these fusion proteins could theoretically form stable structures.

Figure 5. The prediction model of domain fusion protein in pymol. (a) wpy-1: 1WFB[13] (α-helix) joined by a flexible linker to 1J5B[14] (α-helix). (b) wpy-2: 1WFB (α-helix) joined by a flexible linker to 3ULT (β-helix). (c) wpy-3: 1WFB (α-helix) linked to 1EZG [15](β-helix) and then to 5IRB[16] (β-sheet) via flexible linkers. (d) wpy-4: 1WFB (α-helix) connected by a flexible linker to 1M8N (β-helix) and by a rigid linker to 5IRB (β-sheet).

Table 1. Performance evaluation of domain-fused antifreeze proteins

Build

We synthesized the vector pET-28a-wpy1 to pET-28a-wpy4 through the company's commercial process, with each vector containing the respective target gene expression module.

Test

To assess the antifreeze properties of the designed fusion proteins, we employed the same experimental procedure as outlined earlier.
We evaluated the IRI activity of these proteins.

Figure 6. IRI activity of crude enzyme solutions containing the fusion proteins wpy-1, wpy-2, wpy-3, and wpy-4. (a) Ice crystal morphology observed after incubation at -6 °C for 30 minutes. (b) Measurement of the average ice crystal area (µm²).

The domain-rearranged proteins showed no enhancement in ice recrystallization inhibition compared to controls (Fig. 6), leading us to pursue alternative design strategies.

Learn

The comprehensive results indicate that in this experiment, the domain fusion strategy failed to produce an AFP with actual functionality. We will change our approach in the subsequent iterations. In the subsequent design phase, we will adopt the inverse folding and de novo strategies to conduct a new round of protein optimization.

Cycle 4 De novo design and inverse folding of AFPs

Design

After evaluating various methods for protein structure optimization, we discussed with advisor Xiaoyan Yue and determined an advanced strategy. We employed the inverse protein folding method for structure optimization (detailed design can be found in the Dry Lab). Therefore, we selected four well-characterized antifreeze proteins - AFP4, AFP8, AFP9, and AFP10 - as the initial templates, and designed four new sequences using the inverse folding model (Fig. 7). Subsequently, we implemented a comprehensive de novo strategy, based on different conceptual frameworks, to design three new types of AFPs: helical periodic parameters, ice surface lattice matching, and protein scaffolds. By systematically exploring these three factors, a total of six design schemes were generated (as detailed in the Dry Lab), and their characterization and validation were conducted (Fig. 8).

Figure 7. The structure of invertedly folded protein is predicted through Alpha Fold. (a) The 3ULT-IF structure. (b) The 3WP9-IF structure. (c) The 4NU2-IF structure. (d) The 6A8K-IF structure.

Figure 8 Design of de novo antifreeze proteins

Figure 8. Alpha Fold for the structural prediction of de novo designs based on inverse folding (a-b) De novo 1 and De novo 2, designed using helical periodicity parameters. (c-d) De novo 3 and De novo 4, engineered for ice lattice matching. (e-f) De novo 5 and De novo 6, created through novel protein backbone design.

Build

We constructed the following expression vectors via gene synthesis: pET-28a-3WP9-IF, pET-28a-4NU2-IF, and pET-28a-6A8K-IF (inverse-folded designs), along with pET-28a-De novo1 to pET-28a-De novo6 (de novo designs).

Test

We first conducted IRI characterization on the inverse folding protein.

Figure 9. IRI activity of crude enzyme solutions containing the engineered proteins 6A8K-IF, 4NU2-IF, and 3WP9-IF. (a) Representative ice crystal images after incubation at -6 °C for 30 minutes. (b) Measurement of the average ice crystal area (µm²).

Significant antifreeze activity was observed in specific inverse-folded protein variants (Fig. 9).

In the subsequent stage, the same IRI characterization scheme will be applied to the 6 newly designed proteins to verify the antifreeze properties of the newly de novo proteins.

Figure 10. IRI activity characterization of crude enzyme solutions containing de novo proteins based on inverse folding. (a) Ice crystal morphology observed after freezing at -6 °C for 30 minutes. (b) Measurement of the average ice crystal area (µm²).

The experimental results demonstrate that some of the anti-freezing proteins we designed exhibit high IRI performance, which positively validates our experimental design strategy. However, there are also some proteins with poor anti-freezing properties (Fig. 10).

Learn

The research results indicate that both the proteins with inverse folding design and de novo design possess measurable antifreeze activity, but their performance still lags behind that of some natural antifreeze proteins. Our design of protein optimization is still at a rudimentary stage and has not been deeply explored. At the same time, the use of our protein optimization models has not shown superiority. Future research will focus on systematically optimizing the design algorithms and deeply studying the effective antifreeze proteins we have designed to develop more superior artificial antifreeze proteins.

Cycle 5 Biocompatibility testing

Design

Based on the feedback from HP and to assess the applicability of our antifreeze protein in subsequent material applications, we conducted a series of biocompatibility tests to evaluate its potential cytotoxicity.

Build

To evaluate the biocompatibility of our engineered antifreeze proteins, we selected four AFP from earlier characterization studies for cytotoxicity testing.

Test

Four antifreeze proteins (AFP4, AFP8, AFP9, MUT4) demonstrating superior performance in previous assays were selected for overexpression. Following the established protein production protocol from Cycle 1, We obtained crude protein extracts and then carried out protein purification of the crude enzyme solution using the AKTA purification instrument (Fig. 11).

Figure 16 IRI activity of AFP-TPS and AFP-TPP

Figure 11. SDS-PAGE analysis of purified antifreeze proteins. (a) Purified AFP4. (b) Purified AFP8. (c) Purified AFP9. (d) Purified MUT4.

To evaluate the potential cytotoxicity of our antifreeze proteins, L929 mouse fibroblast cells were treated with the four selected protein samples. The assay was performed using cells seeded at densities of 2,500 and 5,000 cells per well to assess dose-dependent effects.

Figure 12. Cytotoxicity assessment of antifreeze proteins in L929 mouse fibroblast cells. (a)Biocompatibility Experiment Record Sheet. (b) Cell viability results from wells initially seeded at 2,500 cells/well. (c) Cell viability results from wells initially seeded at 5,000 cells/well.

Our research results show that these antifreeze proteins possess excellent antifreeze properties and have high biocompatibility. The cell survival rate remains around 100% , indicating that they have almost no cytotoxicity (Fig. 12).

Learn

The biocompatibility test confirmed the excellent safety of our antifreeze protein. This provides strong support for integrating it into subsequent products to achieve practical transformative development. We will introduce the antifreeze protein into downstream products to exert its effect, such as in hand cream.

References

[1] Sun T, et al. An antifreeze protein folds with an interior network of more than 400 semi-clathrate waters. Science. 2014, 343(6172): 795-798.
[2] Nishimiya Y, et al. Crystal structure and mutational analysis of Ca2+-independent type II antifreeze protein from longsnout poacher, Brachyopsis rostratus. Journal of Molecular Biology. 2008, 382(3): 734-746.
[3] Ko T P, et al. The refined crystal structure of an eel pout type III antifreeze protein RD1 at 0.62-Å resolution reveals structural microheterogeneity of protein and solvation. Biophysical Journal. 2003, 84(2): 1228-1237.
[4] Middleton A J, et al. Antifreeze protein from freeze-tolerant grass has a beta-roll fold with an irregularly structured ice-binding site. Journal of Molecular Biology. 2012, 416(5): 713-724.
[5] Leinala E K, et al. A β-helical antifreeze protein isoform with increased activity: structural and functional insights. Journal of Biological Chemistry. 2002, 277(36): 33349-33352.
[6] Pentelute B L, et al. X-ray structure of snow flea antifreeze protein determined by racemic crystallization of synthetic protein enantiomers. Journal of the American Chemical Society. 2008, 130(30): 9695-9701.
[7] Garnham C P, et al. Anchored clathrate waters bind antifreeze proteins to ice. Proceedings of the National Academy of Sciences of the United States of America. 2011, 108(18): 7363-7367.
[8] Hanada Y, et al. Hyperactive antifreeze protein from an Antarctic sea ice bacterium Colwellia sp. has a compound ice-binding site without repetitive sequences. FEBS Journal. 2014, 281(16): 3576-3590.
[9] Do H, et al. Structure-based characterization and antifreeze properties of a hyperactive ice-binding protein from the Antarctic bacterium Flavobacterium frigoris PS1. Acta Crystallographica Section D: Biological Crystallography. 2014, D70(4): 1016-1025.
[10] Kondo H, et al. Multiple binding modes of a moderate ice-binding protein from a polar microalga. Physical Chemistry Chemical Physics. 2018, 20(38): 25295-25303.
[11] O’Brien J R, et al. Insight into the mechanism of the B12-independent glycerol dehydratase from Clostridium butyricum: preliminary biochemical and structural characterization. Biochemistry. 2004, 43(16): 4635-4645.
[12] Mueller G M, et al. Inhibition of recrystallization in ice by chimeric proteins containing antifreeze domains. Journal of Biological Chemistry. 1991, 266(12): 7339-7344.
[13]F.Sicheri et al., Ice-binding structure and mechanism of an antifreeze protein from winter flounder[J], Nature, 1995, 375, 427-431
[14]Edvards Liepinsh et al.,Solution structure of a hydrophobic analogue of the winter flounder antifreeze protein[J],European Journal of Biochemistry, 2002, 269(4), 1259-1266
[15]Yih-Cherng Liou et al.,Mimicry of ice structure by surface hydroxyls and water of a β-helix antifreeze protein[J], Nature, 2000, 406, 322-324
[16]Shuaiqi Guo et al.,Structure of a 1.5-MDa adhesin that binds its Antarctic bacterium to diatoms and ice[J], ScienceAdvances, 2017, 3(8), e1701440

Overview

Through communication with the enterprises, we identified key bottlenecks in their production process of functional proteins: complex, time-consuming, and costly protein purification steps, coupled with low yield. To address this challenge, we proposed the concept of "Antifreeze Chassis" . This involves using engineered strains capable of directly synthesizing antifreeze proteins as the final product, thereby by passing traditional purification processes.

Considering the sensitivity to odor in our target fields such as cosmetics and fashion, we selected the odor-friendly Saccharomyces cerevisiae as the chassis microorganism. We introduced the EBY100 yeast surface display system to enhance the antifreeze ability of the chassis. Meanwhile, by integrating the trehalose synthetic metabolic pathway, the chassis is endowed with dual antifreeze capabilities.

C1. Introduce GFP to characterize the Yeast Surface Display System Usability

Design

The EBY100 yeast surface display system consists of key proteins: the agglutinin protein Aga1 and Aga2. Aga1p is anchored to the surface of the EBY100 yeast strain. We designed a fusion protein where Aga2p is linked to the AFP. After translation, Aga2p-AFP would form disulfide bonds with Aga1p on the cell wall, thereby anchoring itself on the cell surface[1]. Simultaneously, The galactose-inducible system regulates AFP production, reducing metabolic burden on the chassis.

First, to verify the usability of the EBY100 yeast surface display system, we designed plasmid pR_DK01. Green Fluorescent Protein (GFP) , which was constructed as a fusion protein with Aga2p, was used to replace AFP for characterization. The fusion protein expressed under the control of the GAL1 promoter (Fig. 1).

Figure 1. Genetic circuit schematic of plasmid pR_DK01

Build

We obtained the GFP gene via PCR amplification from a laboratory stock plasmid and linearized the pYD1 vector. The GFP gene was ligated with the linearized pYD1 vector using Gibson Assembly and transformed into Escherichia coli. After verification by colony PCR, the correct plasmid was extracted and sequenced. The confirmed correct plasmid was then transformed into Saccharomyces cerevisiae EBY100. Additionally, we transformed the empty pYD1 plasmid into EBY100 as a control group.

Test

We validated the transformants containing the correct plasmid through PCR (Fig. 2).

Figure 2. GFP verification gel electrophoresis image. From left to right are the upper and lower interfaces of plasmid PR_DK01

The successfully constructed yeast strains and their protoplasts were observed under a fluorescence microscope to examine GFP surface display after induction. The intact yeast cells were imaged with an exposure time of 200 ms, while the protoplasts were imaged with an exposure time of 1 s. As shown in the results, the fluorescence intensity of the intact yeast cells was notably strong, with a distinct ring of fluorescent protein visible around the cell periphery. In contrast, the protoplasts exhibited much weaker fluorescence even under longer exposure. These findings confirm the effective function of our EBY100 yeast surface display system. (Fig. 3).

Figure 3. Fluorescence micrographs of GFP yeast surface display. (a) shows the surface display of EBY100 strain containing pR_DK01 after induced expression, captured with a 200ms exposure under fluorescence. (b) shows the surface display of protoplasts prepared from this strain, imaged with a 1s exposure under fluorescence microscopy

Learn

Fluorescence microscopy results revealed that our yeast surface display system can successfully display GFP on the yeast surface. Therefore, in the next cycle, we will introduce 11 natural AFPs into the system.

C2. Yeast Surface Display of 11 Natural AFPs

Design

Following the same strategy, we designed fusion proteins where 11 different AFPs were linked to the agglutinin protein Aga2p. We designed plasmids pR_DK02 to pR_DK12, with AFP expression under the control of the GAL1 promoter (Fig. 4).

Figure 4. Genetic circuit schematic of plasmid from pR_DK02 to pR_DK12

Furthermore, to demonstrate that the surface display system significantly enhances the chassis's antifreeze capability, we designed two endogenous expression plasmids, pR_DK13 and pR_DK14. These plasmids had the Aga2p sequence deleted from the pYD1 backbone, preventing AFP secretion via the agglutinin pathway and retaining the AFP within the EBY100 strain. AFP expression is controlled by the GAL1 promoter. (Fig. 5).

Figure 5. Genetic circuit schematic of Plasmid from pR_DK13 to pR_DK14

Build

We obtained the 11 AFPs genes via PCR amplification from commercially synthesized plasmids and linearized the pYD1 vector. Subsequently, Gibson Assembly was used to ligate each of the 11 AFP genes individually with the linearized pYD1 vector, and the products were transformed into E.coli for cloning and amplification.

Simultaneously, via inverse PCR amplification, we obtained the linearized vector backbone lacking the Aga2p gene. Similarly, two selected AFP genes were individually cloned into this linearized vector (lacking Aga2p) using Gibson Assembly and transformed into E.coli.

After screening by colony PCR and verification by plasmid extraction and sequencing, the correct recombinant plasmids were obtained. These were then transformed into S.cerevisiae EBY100, ultimately achieving surface display and endogenous expression of the antifreeze proteins.

Test

We validated the transformants containing the correct plasmid through PCR (Fig. 6, Fig. 7).

Figure 6. Colony PCR gel electrophoresis image. Lanes from left to right show junctions for plasmids pR_DK02 to pR_DK07, upper/lower junctions for pR_DK08, upper/lower junctions for pR_DK09, upper/lower junctions for pR_DK10, upper/lower junctions for pR_DK11, upper/lower junctions for pR_DK12

Schematic of the domain-swapped helical construct in AFP11

Figure 7. Colony PCR gel electrophoresis image. Lanes from left to right show junctions for plasmids pR_DK13, pR_DK14

After obtaining the successfully constructed EBY100 yeast strains, we tested the performance of the antifreeze chassis.

T1. Ice Recrystallization Inhibition (IRI) Activity Characterization

We first induced protein expression in the yeast strains. A certain amount of yeast strain from the plate was inoculated into SC-Trp medium and cultured in a shaker at 30°C for 36h. Subsequently, the culture was diluted to an OD of 0.6-0.8 using SC-Trp medium without glucose. After dilution, raffinose (working concentration 10g/L) and galactose (working concentration 20g/L) were added for induction, followed by shaking at 30°C for 36h.

The IRI assay was then performed. The induced culture was centrifuged at 3000×g for 5 minutes, and the cell pellet was collected. 0.08–0.085 g of the pellet was weighed and placed in a 1 mL centrifuge tube. Then, 500 μL of a 10% glycerol solution (diluted in PBS) was added to fully resuspend the cells. 12 μL of the cell suspension was dropped from a height of 0.7 meters onto a quartz crucible placed on a thin aluminum sheet pre-cooled in liquid nitrogen, forming a uniform thin ice layer. The sample was transferred to a cooling stage at -60°C, then warmed at a rate of 20°C/min to -6°C (yeast samples were warmed to -12°C), and held at this temperature for 30 minutes. Images were captured at 0, 5, 10, 15, 20, 25, and 30 minutes after the start of the holding period. ImageJ software was used to analyze the Mean Grain Area (MGA) of the ice crystals. Specifically, eight random areas (each area 176 μm × 176 μm) were selected from the 30-minute image, and the number and average size of ice crystals in each area were counted. Each sample was tested in triplicate, and the final results were averaged. Bar graphs with overlaid data points were plotted using Origin software (The 'pYD1' in the graph refers to the control group transformed with the empty pYD1 plasmid).

The bar graph shows that the chassis strains containing the empty pYD1 vector and displaying GFP exhibited the worst IRI performance. In contrast, chassis strains displaying AFP1-9 and AFP11 demonstrated excellent IRI performance, confirming that our AFPs successfully endowed the chassis with excellent IRI properties. Our antifreeze chassis shows great potential as an antifreeze agent (Fig. 8).

Figure 8. Performance characterization diagram of AFP1-11, pYD1, and GFP yeast strains in IRI. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

Meanwhile, the analysis of IRI activity in the two endogenously expressed strains revealed a significant reduction compared to the chassis strain equipped with the surface display system. This outcome further confirms that the surface display system plays an essential role in enhancing the freeze resistance of the chassis strain, thereby validating the rationality and correctness of our selection of the surface display system (Fig. 9).

Figure 9. Performance characterization diagram of pYD1, GFP, AFP9, AFP10, AFP9△Aga2p, and AFP10△Aga2p in IRI. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

T2. Freezing Survival Rate Characterization of Antifreeze Chassis

The same method was used to induce protein expression in the yeast strains, except the induction time was shortened from 36 hours to 24 hours to avoid adverse effects on the strains from prolonged induction. Before freezing, all cultures were diluted to an OD of approximately 0.8 using SC-Trp medium without glucose to minimize errors in survival rate due to differences in cell density. We selected the freezing temperatures: -80°C. The specific protocols were as follows:

After freezing treatment, the cell suspensions were diluted according to the corresponding dilution factors, and 100 µL was spread onto SC-Trp plates. The plates were incubated at 30°C. Once colonies grew to a suitable size and number, the plates were photographed and the colonies were counted. The survival rate of the strains was calculated by comparing the number of colonies formed under freezing conditions with the number of colonies in the non-frozen control group (Fig.10, Fig.11).

Figure 10. Survival rate assay of yeast strains AFP1-11, pYD1, and GFP after freezing at -80°C. (a) Growth of the strains after culture. (b) Calculated survival rates of the strains. The survival rate was determined by comparing the number of colonies formed under freezing conditions with that of the non-frozen control group

Performance characterization diagram of NEW11-9、NEW11-10

Figure 11. Survival rate assay of yeast strains pYD1, AFP9, AFP10, AFP9△Aga2p, and AFP10△Aga2p after freezing at -80°C. (a) Growth status of the strains after culture. (b) Survival rates of the strains calculated from colony counts. The survival rate was determined by comparing the number of colonies formed under freezing conditions with that of the non-frozen control group

Based on the characterization results from the above two parts, it can be observed that in the freezing survival rate assay, the 11 AFPs we constructed significantly enhanced the freeze-survival capacity of the chassis EBY100, demonstrating a clear advantage over the control group, which indicates the rationality and success of our system design. Furthermore, after the removal of the surface display system, the freezing survival capacity of the two endogenously expressed strains decreased significantly, further confirming the necessity of the surface display system in improving freeze-resistant performance.

Learn

In this iteration, we successfully constructed 11 yeast surface display chassis based on natural AFPs and simultaneously prepared two endogenous expression controls without the display system. Evaluation of the system's antifreeze capability confirmed that: 1) The introduction of AFPs is the decisive factor for improving the chassis's antifreeze properties, with AFP-displaying chassis performing significantly better than the empty vector control; 2) The two chassis lacking the display system showed weaker antifreeze capability, confirming the importance of the display system in the antifreeze chassis. To further improve the chassis performance where there is still room for optimization, subsequent research will focus on introducing a series of novel non-natural AFPs to expand the upper limit of their antifreeze capability.

C3. Yeast Surface Display of Unnatural Antifreeze Proteins

In the previous iteration, we successfully constructed yeast display systems for 11 natural AFPs and corresponding antifreeze chassis, verifying their excellent antifreeze performance. In this iteration, we introduced several novel unnatural AFPs into the system, including the aforementioned point-mutated protein MUT-4, and a new class of proteins designed by replacing protein helical domains. We further constructed these unnatural AFPs in the EBY100 yeast surface display system to prepare the corresponding antifreeze chassis.

C3.1 Surface Display System for Point-Mutated Protein

Design

Based on the results from the antifreeze protein part, we constructed several point-mutant proteins, among which MUT-4 demonstrated relatively high IRI activity. Therefore, in this iteration, we selected MUT-4 for initial introduction into the yeast surface display system to construct the antifreeze chassis. Following the same strategy, we designed a fusion protein where the point-mutant protein MUT-4 is linked to the agglutinin protein Aga2p. We designed plasmid pR_DK15, with MUT-4 protein expression under the control of the GAL1 promoter (Fig.12).

Figure 12. pR_DK15 plasmid genetic circuit schematic

Build

We obtained the MUT-4 sequence via PCR amplification from the commercially synthesized pET28a-MUT4 plasmid and ligated it with the linearized vector using Gibson Assembly. After transforming the ligation product into E.coli, PCR identification confirmed the correct recombinant plasmid, which was then extracted. Subsequently, this plasmid was transformed into yeast EBY100 to construct the antifreeze chassis.

Test

We validated the transformants containing the correct plasmid through PCR(Fig.13).

Figure 13. Colony PCR gel electrophoresis image. Verification band for pR_DK15 plasmid

Subsequently, we performed a series of antifreeze characterizations on the MUT-4 antifreeze chassis.

T1. IRI Activity Characterization

First, IRI performance was characterized. Strain induction and IRI assay methods were the same as described previously. The results show that the antifreeze performance of the MUT-4 chassis is significantly better than that of the AFP4 chassis, indicating that our mutation construction was very successful, leading to further optimization of the antifreeze chassis performance (Fig.14).

Figure 14. Performance characterization diagram of MUT-4. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

T2. Freezing Survival Rate Characterization of Antifreeze Chassis

The yeast strains were subjected to induced expression and freeze-survival characterization assays using the same methodology. Results indicate that the MUT-4 chassis exhibited improved freeze-resistant performance compared to AFP4, along with a clear advantage over the pYD1 control group. These findings demonstrate the correctness of our point mutation design and its success in optimizing the freeze-resistance of the original AFP4 (Fig.15).

Figure 15. Survival rate assay of yeast strains MUT-4 after freezing at -80°C. (a) Growth status of the strains after culture. (b) Survival rates of the strains calculated from colony counts. The survival rate was determined by comparing the number of colonies formed under freezing conditions with that of the non-frozen control group

Learn

It’s obvious that the MUT4 antifreeze chassis performance has clear advantages compared to AFP4. However, overall, its antifreeze capability still has significant room for improvement. Considering that in this iteration, we only made changes to a few specific sites of the protein, the modification scope was relatively small. In the next iteration, we will modify the protein's structural domains, expecting to substantially enhance the performance of the antifreeze chassis.

C3.2 Yeast surface display system for domain-swapped helical proteins

Design

Among the 11 natural AFPs we studied, AFP11 attracted our attention. Firstly, existing research suggests that when the molecular weight of an AFP is too large, it might act as a nucleation site for ice crystal formation, potentially promoting ice growth instead of inhibiting it; AFP11 is the only one among these 11 proteins with a gene sequence length exceeding 2000 bp. Secondly, AFP11 contains multiple typical helical structures, and such helices are considered key structural units for AFP antifreeze function. Based on the above reasons, we planned to modify the structure of AFP11, specifically replacing its largest internal helix with smaller helical structures from other proteins, aiming to optimize its antifreeze performance.

In simulations, we selected Root Mean Square Deviation (RMSD) and Root Mean Square Fluctuation (RMSF) as evaluation metrics. We replaced the "target" sequence within AFP11 with a small helical sequence from AFP9, creating a new protein, which served as the simulation object (see Dry Lab section for details). The results showed that compared to the original AFP11, the helix-replaced protein has a more stable structure, but its residue flexibility did not increase. We speculate that this higher structural rigidity might help the protein bind more stably to the ice surface, thereby enhancing its antifreeze capability (Fig.16, Fig.17).

Figure 16. Schematic of the domain-swapped helical construct in AFP11. (a) AFP11 sequence schematic. The orange "target" part is the sequence to be replaced. (b) Replaced AFP11 sequence schematic. The yellow part is the small helical sequence from AFP9 replacing the "target"

Analysis of RMSD and RMSF simulation results

Figure 17. Analysis of RMSD and RSMF simulation results. (a) RMSD simulation result graph for helix-replaced AFP11 chassis. (b) RMSD simulation result graph for AFP11 chassis. (X-axis: Time, Y-axis: RMSD value (nm)). (c) RMSF simulation result graph for helix-replaced AFP11 chassis. (d) RMSF simulation result graph for AFP11 chassis. (X-axis: Residue number, Y-axis: RMSF value (nm))

In summary, dry lab simulation results indicated that replacing the helix in AFP11 holds promise for improving its antifreeze performance.

We designed plasmids pR_DK16 and pR_DK17, which replace the "target" sequence in AFP11 with a helical segment from AFP9 and AFP10, respectively (Fig.18, Fig.19).

Figure 18. Schematic of the domain-swapped helical construct in AFP11. (a) Replaced AFP11 sequence schematic. Yellow part is AFP9 small helix replacing "target". (b) Replaced AFP11 sequence schematic. Yellow part is AFP10 small helix replacing "target".

Figure 19. pR_DK16, pR_DK17 plasmid genetic circuit schematic

Build

Our construction was based on plasmid pR_DK12. We linearized it via PCR amplification into two fragments excluding the "target" sequence. The sequences of the unnatural helices intended for replacement were incorporated into the primers used for amplification. After obtaining these two target fragments, Gibson Assembly was used to seamlessly join them into a complete plasmid, which was then transformed into E.coli competent cells. After screening by colony PCR to verify positive clones, the plasmids were extracted and further transformed into yeast strain EBY100, ultimately successfully obtaining the desired antifreeze chassis strains.

Test

We validated the transformants containing the correct plasmid through PCR (Fig.20).

Figure 20. Colony PCR gel electrophoresis image. (a) Plasmid verification gel image for plasmid pR_DK16. (b) Plasmid verification gel image for plasmid pR_DK17

Subsequently, we performed a series of antifreeze characterizations on the two helix-replacement antifreeze chassis.

T1. Ice Recrystallization Inhibition (IRI) Activity Characterization

First, the characterization of IRI performance was conducted using the same method as described earlier. The results indicate that the two chassis strains with substituted helices exhibit enhanced IRI performance compared to the original AFP11 chassis strain. This demonstrates the rationality of our helix substitution strategy and validates the accuracy of the in silico simulations(Fig.21).

Figure 21. Performance characterization diagram of NEW11-9、NEW11-10. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

T2. Freezing Survival Rate Characterization of Antifreeze Chassis

The same methodology was applied to conduct induced expression and freeze survival assays in yeast strains. As can be observed from the results, the NEW11-9 and NEW11-10 chassis strains exhibit a substantial improvement in freezing tolerance compared to AFP11, along with a clear advantage over the pYD1 control group. These results validate the rationality behind our helical domain substitution strategy, demonstrate strong agreement between the wet-lab experiments and in dry-lab simulations, and confirm a significant optimization of the original AFP11 performance (Fig.22).

Figure 22. Survival rate assay of yeast strains NEW11-9, NEW11-10 after freezing at -80°C. (a) Growth status of the strains after culture. (b) Survival rates of the strains calculated from colony counts. The survival rate was determined by comparing the number of colonies formed under freezing conditions with that of the non-frozen control group

Learn

The characterization results from this iteration showed that the performance of both helix-replacement antifreeze chassis was significantly better than the original AFP11 chassis, fully validating the rationality and effectiveness of our strategy to replace the helical domain. Given that the helices used in this cycle were natural, in the next iteration, we plan to use dry lab simulations to design unnatural synthetic helical structures and employ the same strategy to embed them into AFP11, to further explore the performance optimization limits.

C3.3 Domain substitution was performed on the AFP11 protein using an artificially designed helical domain.

Design

Based on the excellent results from the previous iteration, we decided to continue optimizing AFP11 by replacing its helical domain. In this iteration, we employed dry lab methods to assist in designing two unnatural helical structures. In the design, considering that Type I AFP structures are primarily composed of alanine-rich α-helices, and alanine is a strong helix-forming residue[2], we constructed standard α-helices using alanine as the backbone. Threonine residues were regularly introduced at an 11-residue interval. This interval precisely corresponds to three turns of the α-helix, ensuring that all antifreeze active sites are precisely arranged on the same side of the helix, forming a continuous ice-binding surface. This design process was visually verified using PyMOL (Fig.23).

Figure 23. Visualization of the rationally designed helix concept. Shows all antifreeze active sites located on the same side of the helix

Based on the above strategy, we designed plasmids pR_DK18 and pR_DK19, containing two designed unnatural helices A and B, with gene expression under the control of the GAL1 promoter (Fig.24).

Figure 24. pR_DK18, pR_DK19 plasmid genetic circuit schematic

Build

The construction was based on plasmid pR_DK12. We linearized it via PCR amplification into two fragments excluding the "target" sequence. The sequences of unnatural helices intended for replacement were incorporated into the primers. After obtaining these two target fragments, Gibson Assembly was used to seamlessly join them into a complete plasmid, which was then transformed into E.coli competent cells. After screening by colony PCR to verify positive clones, the plasmids were extracted and further transformed into yeast strain EBY100, ultimately successfully obtaining the desired antifreeze chassis strains.

Test

We validated the transformants containing the correct plasmid through PCR (Fig.25).

Figure 25. Plasmid construction verification gel image for pR_DK18, pR_DK19

For testing, we performed a series of antifreeze characterizations on the two antifreeze chassis with replaced unnatural helices.

T1. IRI Activity Characterization

First, IRI activity characterization was performed on the samples using the method described previously. The results showed that both antifreeze proteins with unnatural helix A or B replacement exhibited significantly better IRI performance than the pYD1 chassis and the AFP11 chassis. This result further validated the effectiveness of the dry lab-based design of unnatural helices and their application in domain replacement strategies, indicating that our constructed antifreeze system has achieved a further improvement in ice crystal inhibition capability (Fig.26).

Engineering Performance Characterization Diagram

Figure 26. Performance characterization diagram of NEW11-A、NEW11-B. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

T2. Freezing Survival Rate Characterization of Antifreeze Chassis

The yeast strains were subjected to induced expression and freeze-survival assays using the same methodology. Results indicate that the NEW11-A and NEW11-B chassis strains exhibit significantly enhanced cryotolerance compared to the AFP11 strain, while also demonstrating a clear advantage over the pYD1 control group. These findings confirm the validity of our helix substitution strategy and underscore the efficacy of the in silico-designed helical domain, which substantially improves the performance of the original AFP11 protein (Fig.27).

Figure 27. Survival rate assay of yeast strains NEW11-A, NEW11-B after freezing at -80°C. (a) Growth status of the strains after culture. (b) Survival rates of the strains calculated from colony counts. The survival rate was determined by comparing the number of colonies formed under freezing conditions with that of the non-frozen control group

Learn

This iteration was an absolute success, decisively validating our dry lab-guided design and replacement strategy for unnatural helices. By successfully introducing rationally designed helical domains, we directly enhanced the system's antifreeze capability, marking our contribution to the rational design and optimization of antifreeze proteins.

C4. Introduction of Trehalose

Design

To further optimize our antifreeze chassis and system, we introduced the trehalose metabolic pathway. Trehalose is a natural cryoprotectant that can specifically bind to biological macromolecules like biomembranes and proteins via hydrogen bonds, replacing water molecules to maintain their hydration shell, preventing cell damage caused by ice crystal formation and osmotic pressure changes. It also lowers the solution freezing point and inhibits ice crystal growth and recrystallization.

We collaborated with Professor Yi Wu's team to screen and construct the relevant plasmid pRS413-trehalose containing a strengthened trehalose metabolic pathway using a narrow-down strategy (Design). In this iteration, we first constructed the trehalose-producing antifreeze chassis, characterized its ability to produce trehalose, and tested the chassis's antifreeze performance.

Build

We transformed the pRS413-trehalose plasmid into the EBY100 yeast strain via yeast transformation. Meanwhile, the empty pRS413 plasmid was transformed into the EBY100 strain to serve as the control.

Test

We validated the transformants containing the correct plasmid through PCR. (Fig.28).

Figure 28. Plasmid transformation verification gel image for EBY100-pRS413, EBY100-pRS413-trehalose

Next, we characterized the production of trehalose in EBY100-pRS413-trehalose.

Figure 29. LC‑MS analysis of trehalose in different yeast strains. (a) LC‑MS chromatograms of the trehalose standard. (b) LC‑MS chromatograms of the engineered EBY100‑pRS413 strain.(c) LC‑MS chromatograms of the engineered EBY100‑pRS413‑trehalose strain. (d) Quantitative comparison of trehalose content between EBY100‑pRS413 and EBY100‑pRS413‑trehalose strains

Mass spectrometry analysis showed that both the original strain EBY100‑pRS413 and the plasmid-carrying strain EBY100-pRS413-trehalose exhibited peaks at the same retention time as the standard, indicating that both strains possess the capability to produce trehalose. However, the trehalose yield of the EBY100-pRS413-trehalose strain was significantly higher than that of the EBY100‑pRS413 strain, confirming the correctness and efficiency of the engineered strain construction (Fig.29).

Subsequently, we conducted a series of freeze resistance characterizations on the EBY100-pRS413-trehalose strain.

T1. IRI Activity Characterization

We inoculated the successfully constructed strain from the plate into SC-His liquid medium and cultured it in a shaker at 30 °C for 12 hours, with the OD value maintained between 0.5 and 0.8. The IRI index was then measured using the same method as described previously. As shown in the results, the IRI performance of the EBY100-pRS413-trehalose strain was significantly higher than that of the control group, indicating that the introduction of trehalose substantially enhanced the freeze-resistance capability of the original strain. The incorporation of trehalose shows great potential for optimizing the freeze-resistance of the system (Fig.30).

Figure 30. Performance characterization diagram of EBY100-pRS413, EBY100-pRS413-trehalose. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

T2. Freezing viability assay

The yeast strains were cultured using the same method as for the IRI characterization. In the freezing viability assay, the selected freezing conditions included exposure to -24 °C for 1 h, 1.5 h, and 2 h. After freezing, 100 μL of the bacterial suspension was plated on Sc-His agar plates, and subsequent characterization steps followed the previously described experimental procedure. As shown in the results, EBY100-pRS413-trehalose exhibited significantly higher cell viability than the original strain under all three freezing durations, demonstrating that trehalose enhances the freeze resistance of the chassis and validating the rationality behind our iterative approach (Fig.31).

Figure 31. Viability assay of EBY100-pRS413 and EBY100-pRS413-trehalose under -24°C cold stress

Learn

Based on the results of this iteration, the introduced trehalose metabolic pathway plasmid has successfully enhanced the trehalose expression level in the host strain and effectively improved its overall freeze-resistance performance. In the next iteration, we plan to integrate both the trehalose module and the antifreeze protein module into the target host EBY100, thereby constructing a reinforced anti-freeze system with a dual freeze-resistance mechanism.

C5. Coupling Trehalose with Antifreeze Proteins

Design

We coupled the trehalose module with the antifreeze protein module to provide our antifreeze system with dual antifreeze capabilities. We designed the transformation of the trehalose metabolic pathway plasmid pRS413-trehalose into EBY100 strains containing antifreeze proteins.

Build

We transformed the pRS413-trehalose plasmid into EBY100 strains containing AFP4, AFP8, AFP9, and AFP10, respectively, constructing four antifreeze chassis with dual antifreeze properties.

Test

We validated the transformants containing the correct plasmid through PCR. We successfully transformed the target plasmid into the four antifreeze chassis.

Figure 32. Plasmid transformation verification gel image

Subsequently, we performed a series of antifreeze capability characterizations on the four dual-capability strains.

T1. Characterization of Ice Recrystallization Inhibition (IRI) Activity

The successfully constructed strains were inoculated into liquid medium and cultured in a shaker at 30 °C for 36 hours. The cultures were then diluted to an OD value between 0.6 and 0.8 using glucose-free SC-His-Trp medium. After dilution, raffinose (final concentration 10 g/L) and galactose (final concentration 20 g/L) were added for induction, followed by another 36 hours of shake-flask cultivation at 30 °C. The IRI activity of the four dual-antifreeze strains was subsequently measured using the same characterization method as described previously.

Figure 33. Performance characterization diagram of four dual-mode antifreeze chassis. (a) IRI activity test: ice crystal images of yeast suspensions in 10% glycerol PBS after freezing at -12°C for 30 minutes. (b) Bar graph of ice crystal average particle size analysis

Learn

As evidenced by the results, the strain incorporating the enhanced trehalose metabolic pathway exhibited a significantly improved IRI performance compared to the original chassis. This validates the efficacy of our trehalose integration strategy and represents a further successful optimization of our antifreeze system.

References

[1] Boder ET, Wittrup KD. Yeast surface display for screening combinatorial polypeptide libraries. Nat Biotechnol. 1997;15(6):553-557. doi:10.1038/nbt0697-553.
[2] Davies PL, Baardsnes J, et al. Structure and function of antifreeze proteins. Philos Trans R Soc Lond B Biol Sci. 2002;357(1423):927-935. doi:10.1098/rstb.2002.1081.