Through comprehensive literature review and with guidance from Prof. Lei Zhang's research group, we acquired 11 phylogenetically diverse, family-representative antifreeze proteins (AFPs) as chassis scaffolds. These proteins underwent systematic evaluation via fermentation-based expression, ice recrystallization inhibition (IRI) assays, and differential scanning calorimetry (DSC) characterization. Subsequent rational protein engineering and directed evolution enabled screening of variants with enhanced cryoprotective activity for functional applications in: Cosmeceutical formulations (e.g., skin cryoprotectants), dermatological barrier products (e.g., frost-protective hand creams), bioinspired anti-icing coatings, to mitigate low-temperature challenges in textile and fashion industries.

We first obtained the expression plasmid pET28a-AFPs, which carries the antifreeze protein gene, through commercial DNA synthesis. This plasmid was then transformed into Escherichia coli strains. After fermentation and induction of protein expression, the bacterial cells were lysed to extract the target antifreeze proteins. The antifreeze activity of the obtained proteins was further characterized using the Ice Recrystallization Inhibition (IRI) assay, and differential scanning calorimetry (DSC) characterization.

Modification and Design of AFPs

After obtaining the natural antifreeze protein, we implemented rational protein engineering of the AFPs to enhance cryoprotective efficacy. This optimization strategy was guided by in silico predictions from molecular dynamics simulations and binding energy calculations, enabling targeted modifications to improve ice-binding functionality.

The work of our Dry lab leveraged the ESM-1v protein language model to identify mutation sites, distinguishing: evolutionarily conserved core residues critical for structural stability, surface-exposed variable residues as prime targets for functional optimization.

We prioritized directional mutagenesis on surface variable regions while preserving global folding integrity. Four helical AFPs (PDB: 3ULT, 3WP9, 4NU2, 6A8K) were selected for mutational simulation. Computational predictions (see Model section) confirmed the feasibility of enhancing ice-binding activity without compromising structural integrity.

Guided by in silico results, we introduced site-directed mutations into the gene sequences of: 3ULT (named as MUT4), 3WP9 (named as MUT8), 4NU2 (named as MUT9) and 6A8K (named as MUT10).

We obtain gene sequences with the pET28a vector through gene synthesis. Then transformed it into E. coli for fermentation. The crude enzyme and crude enzyme extract were characterized for antifreeze activity using IRI experiments.

Using an inverse folding approach, we employed 3ULT, 3WP9, 4NU2, and 6A8K as initial templates. Based on a multi-objective optimization framework, we integrated deep learning with molecular simulation techniques to achieve systematic improvement of AFP performance, resulting in optimized novel sequences for AFPs.

We obtain gene sequences through gene synthesis. Expression was induced through fermentation, and antifreeze performance was characterized via IRI assays.

Design and Construction of Novel AFPs

Beyond mutational modification of existing antifreeze proteins as templates, we also designed novel antifreeze proteins in De novo way. First, computational design and prediction were performed from three perspectives: structural periodicity, ice lattice matching, and AI-generated backbones. Computational models including RFdiffusion2 and ProteinMPNN were sequentially applied to provide theoretical basis and design templates for engineering antifreeze functionality in strains. Based on computational predictions, we aim to construct novel antifreeze proteins with different functional priorities, with detailed workflows and simulation methods provided in Model.

Based on computational design results, the output sequences will be used to synthesize novel antifreeze protein genes via gene synthesis. This will enable construction of expression-ready pET28a-De novo series plasmids. Antifreeze performance will then be characterized through IRI assays.

In preliminary studies, we observed significant structural differences in antifreeze proteins from different biological sources. For example, Type I AFPs are predominantly α-helical, while Type III AFPs mainly consist of β-sheets. Existing research has demonstrated that domain fusion of glucoamylase genes can yield chimeric enzymes with novel properties and improved thermostability[12]. Inspired by this, we hypothesize that recombining structural domains from different antifreeze proteins may influence their antifreeze performance.

We conducted systematic domain fusion design to further enhance the functional synergy of the chimeric proteins. By rationally recombining structural domains from AFPs of different origins using flexible linkers (GGGGSGGG) and rigid linkers (EAAAKEAAAKEAAAK), we optimized their spatial orientation, enabling more efficient interaction with ice crystal surfaces and thereby improving antifreeze performance.

We will obtain genes encoding domain-fusion proteins through gene synthesis and construct pET28a-wpy series plasmids for surface display of these fusion proteins. These plasmids will be transformed into E.coli for fermentation-induced expression. Antifreeze performance will then be characterized via IRI assays.

Based on industry consultations with SNEFE Research Institute and Bloomage Biotechnology Co., Ltd. (detailed in Human Practices), we identified significant challenges in scale-up AFP production: extensive protein purification steps incur substantial time-cost expenditures while yielding suboptimal purity. To address these industrial constraints, we engineered an AFP chassis strain incorporating a yeast surface display system.

The EBY100 platform utilizes the Saccharomyces cerevisiae α-agglutinin anchoring mechanism. Genomically expressed Aga1p covalently binds β-glucans in the cell wall extracellular matrix. Plasmid-encoded Aga2p (expressed from pYD1 vector) forms disulfide bonds with Aga1p, enabling surface localization. Target AFP fusion constructs are thus displayed via C-terminal fusion to Aga2p[13].

We genetically engineered the Saccharomyces cerevisiae EBY100 strain by fusing 11 antifreeze protein (AFP) genes to the Aga2p coding sequence via a glycine-serine linker (GS linker), constructing AFP-Aga2p fusion cassettes. Post-translationally, the fusion proteins form disulfide bonds with cell wall-anchored Aga1p, resulting in extracellular localization of AFPs on the yeast surface. This configuration enables simultaneous intra- and extracellular cryoprotection, establishing a prototype antifreeze chassis strain with dual-phase cryopreservation functionality.

Based on the results, we selected high-performance point-mutated AFP genes and seamlessly cloned them into the pYD1 vector to construct the plasmid PR-DK, which enables surface display of the mutated AFPs. This plasmid was then transformed into the S. cerevisiae strain EBY100 for induced fermentation.

The antifreeze performance of the engineered yeast chassis was evaluated through IRI experiments and yeast freezing-thaw survival rate assays. A comparison with the original AFP-based chassis allows us to assess the effectiveness of the site-directed mutations.

During structural analysis of 11 AFPs for point mutation design, we observed that Type I antifreeze proteins contain multiple aggregated α-helical structures. Given that ice-binding surfaces are enriched with alanine[14] and abundant alanine facilitates α-helix formation, we hypothesized that enhancing the α-helical content of antifreeze proteins could improve cryoprotective capability. Based on this, we propose that either truncating the α-helical regions of antifreeze proteins or introducing α-helical fragments with superior antifreeze activity from other species may modulate their cryoprotective performance.

We selected the 4MTJ protein as the engineering template, incorporating α-helices from 4NU2 and 6A8K for substitution. Concurrently, we computationally designed de novo α-helical structures (Helix A and B) conforming to antifreeze protein characteristics. Through PCR primer-mediated mutagenesis, we introduced mutations into the 4MTJ gene sequence to achieve helical replacements, thereby integrating α-helical structures predicted to enhance ice-binding activity into the 4MTJ protein architecture.

After preliminary construction and characterization, we have initially obtained antifreeze chassis systems based on natural antifreeze proteins, as well as improved or even de novo designed chassis systems with enhanced antifreeze performance. During discussions with industry partners, concerns were raised regarding the limited antifreeze efficacy of single AFP, as well as the narrow scope of the evaluation criteria for cold resistance. In response to these concerns, under the guidance of Prof. Yi Wu, we obtained a yeast strain with high stress tolerance from his laboratory. The stress resistance of this strain is attributed to the endogenous synthesis of multiple protective metabolites. After comprehensive evaluation, we selected trehalose as a key functional component for its ability to effectively inhibit ice recrystallization and its excellent water-retention properties, which contribute to enhanced cell viability under low-temperature stress. We aim to further amplify its biosynthetic pathway and integrate it into our antifreeze chassis strain, thereby achieving a synergistic effect of antifreeze and hydration through the combined action of trehalose and AFP.

In collaboration with Professor Wu’s team, we utilized the yeast strain to perform chromosomal aneuploidy amplification of chromosome III under environmental stress conditions. Through chromosomal rearrangement and a stepwise screening strategy, we progressively narrowed down the functionally relevant genomic regions. Ultimately, we identified five key genes associated with the regulation of trehalose synthesis.

Building on this foundation, we further propose a synergistic mechanism between trehalose and antifreeze proteins, aiming to enhance the anti-freezing capacity of the chassis strain.

We obtained five key genes for endogenous trehalose synthesis via PCR amplification from yeast strains. To streamline downstream construction, we first ligated these genes into two fragments using Overlap Extension PCR: a 4902bp fragment (gene1-gene2-gene3) and a 4002bp fragment (gene4-gene5), with homologous arms incorporated via primers. Subsequently, both fragments were assembled with the pRS413 vector backbone through Gibson Assembly, yielding the recombinant plasmid pRS413-trehalose containing all five target genes. This plasmid was co-transformed separately with each of the pR_DK05, pR_DK09, pR_DK10, and pR_DK11 plasmids into Saccharomyces cerevisiae EBY100, generating engineered chassis strains capable of simultaneous endogenous trehalose synthesis and AFP expression under freezing Stress. Following fermentation-induced expression, cryoprotective effects were characterized through IRI assays of cell suspensions and freeze-thaw survival rate quantification of yeast cells.

Performance was systematically compared to baseline strains expressing only surface-displayed native AFPs without trehalose synergy, thus demonstrating the cooperative cryoprotection conferred by integrating trehalose biosynthesis with AFP expression.

Through the Cre/loxP stochastic rearrangement system, we can obtain optimal AFP combinations. However, differential expression intensity ratios among these genes may significantly impact the cryoprotective capacity of the chassis strain. Inspired by the work of Prof. Wu Yi's team[16], we propose implementing a multiplexed inversion system to screen optimal gene expression ratios, enabling precise control of target protein expression levels.

As a first step, we acquired the plasmid pRS415-GAL1-Rci from Prof. Wu Yi's laboratory, which expresses the DNA invertase Rci.

During the construction of the invertible plasmid, we first established a promoter library with varying strengths using PCR and OE-PCR techniques. The promoters included strong promoters (pGMP1, pTDH3, pTEF1), medium-strength promoters (pPGK1, pPDC1, pTEF2, pTPI1), and weak promoters (pCHO1, pADH1, pCYC1). Each promoter was flanked by sfxa101 sites at both the upstream and downstream ends, forming an invertible cassette. This cassette serves as a modular element for subsequent expression regulation and screening. These promoters regulate target gene expression to modulate intracellular protein yield.

Subsequently, we amplified AFP gene fragments by PCR, adding 40bp homology arms to both ends for seamless fusion with promoters and terminators. These fragments serve as building blocks for inversion cassettes, using pRS413 as the vector backbone.

Next, we employed OE-PCR to assemble each AFP gene with promoter and terminator modules, forming complete inversion cassettes (basic functional units).

The plasmids pTJ25-MI and pRS415-GAL1-Rci were co-transformed into the engineered yeast strain. Under the action of the Rci enzyme, the promoter library fragment in pTJ25-MI undergoes multiple inversions, with each inversion state corresponding to a distinct expression strength. Clones with superior antifreeze performance were identified and characterized through yeast freezing survival assays. This system enables reversible regulation of AFP expression under different conditions, facilitating the optimization and screening of expression levels.