Taking into account the serious global concern of antimicrobial resistance and the complexity of scenarios on battlegrounds and disaster-striken areas, we aimed to engineer resistance-proof antimicrobial fusion proteins capable of binding a wide range of first aid materials and being released upon contact with wounds to achieve antimicrobial action. Thus, our design consisted of 3 main modules: the binding domain, which specifically binds to the respective type of first aid material; the antimicrobial functional module, 4 human defensins; and the small ubiquitin-like modifier (SUMO) proteolytic cleavage tag, selectively cleaved by the SUMO protease for the release of the functional module.

However, from our prototype, our design went through multiple iterations of the design-build-test-learn (DBTL) cycle, resulting in fusion proteins with a widened range of application, optimized antimicrobial strength, and increased production.

Binding domains selectively bind their respective type of first aid material. Thus, a fusion protein containing binding domains and defensins could form a layer of antimicrobial coating on first aid materials. We selected various binding domains that could bind materials suitable for first-aid applications and attached chromoproteins with different colors to them to visualize their binding efficiency in a straightforward way (Fig. 1).

Design I

Binding Domain Selection

In order to attach our defensins to various types of first aid materials, we selected the collagen binding domains α1 and α2; cellulose binding domains CBM2 and CBM3; chitosan binding domain CBM5; and alginate binding domain CBMxx.

α1 and α2 are collagen-binding subunits of the collagen integrins α1β1 and α2β1, respectively. They interact with special GXX'GER motifs (with X and X' being any amino acid) in the triple-helical structure of collagen. They could bind collagen wound dressings that decrease bleeding at wound sites (Heino & Siljamäki, 2023).

Carbohydrate-binding modules (CBMs) are generally non-catalytic modules appended to carbohydrate-active enzymes to participate in the degradation of carbohydrates (Armenta et al., 2017). They possess specificity toward different types of carbohydrates.

CBM2 and CBM3 are 2 families of CBMs that selectively bind different types of cellulose, which could be made into gauze to stop bleeding from extensive wounds.

CBM5 is another family of CBM that binds chitosan (Han et al., 2017). Because chitosan already possesses stand-alone antimicrobial property, it could be made into a wound dressing with enhanced antimicrobial strength to combat especially severe instances of infection.

CBMxx is yet another family of CBM that binds alginate (Mei et al., 2024). Alginate possesses high biocompatibility, low toxicity, and could provide a moist wound environment and promote wound healing when delivered in a hydrogel wound dressing formation (Lee & Mooney, 2012).

Build-Test-Learn I

Compared to negative controls, first-aid materials dyed with chromoproteins without binding domains, first-aid materials dyed with chromoproteins attached to binding domains showed significantly greater color retention. This verified that our design of creating a layer of antimicrobial coating on first-aid materials through the incorporation of specific binding domains into our fusion proteins is feasible.

Design II

Why Defensins?

We decided early on that the functional module, or active ingredient, of our antimicrobial first aid materials would be defensins.

Defensins are a cationic, hydrophobic family of antimicrobial peptides that possess broad range antimicrobial properties (Fu et al., 2023). They are generally composed of three beta-pleated sheets stabilized by three cystine disulfide bonds, sometimes with the addition of an alpha-helix structure. As they are naturally released by human neutrophils, intestinal Paneth cells, and other human epithelial cells, they have the advantage of high biocompatibility with the human body compared to other antimicrobial peptides (Fu et al., 2023).

Toward gram-negative bacteria such as E. coli with thinner cell walls, their bactericidal mechanism of action rely on perforating bacterial cell membranes through electrostatic interactions between their cationic amino acid residues and the anionic bacterial cell membrane (Kagan et al., 1990). Furthermore, their hydrophobic amino acid residues play a crucial role in their dimerization, which is functionally important, and allow for better insertion into the hydrophobic inner cell membrane, enhancing membrane disruption (Rajabi et al., 2012; Pazgier et al., 2012; Tai et al., 2014). Toward gram-positive bacteria such as S. aureus with thicker cell walls that make perforation difficult, they target the cell wall precursor Lipid II to inhibit cell wall synthesis (de Leeuw et al., 2010). Due to their ability of acting on cell membranes, a wider and more essential target, in a quicker and more complete manner, they are less likely to induce antimicrobial resistance compared to traditional antibiotics, which often rely on specific cellular targets that are prone to modification (Fjell et al., 2011; Reygaert, 2018). Thus, they make an effective, resistant-proof alternative to traditional antibiotics.

Aside from their direct, individual bactericidal effects, defensins have been shown to enhance the body’s adaptive immune response by recruiting dendritic and T cells to wound sites and inducing the expression of chemokines and other cytokines to further stimulate the immune system (Yang et al., 1999; Gao et al., 2021). Moreover, they have been shown to promote the migration and proliferation of epithelial cells such as keratinocytes and bone-forming cells such as osteoblasts, which can accelerate wound closure and healing (Lai & Gallo, 2009).

Why Biosynthesis?

Although some defensins are naturally synthesized by neutrophils in human blood, the extraction of defensins from them is extremely inefficient and faces ethical concerns (Xie et al., 2022).

Traditionally, defensins are synthesized in vitro through the chemical method of solid-phase peptide synthesis. However, this method is not commercially viable. Generally, this method is costly as commercially synthesizing around 5 mg of a custom peptide with 98% purity could cost $18 per amino acid residue according to the companies ABI Scientific Inc. and AAPPTec (ABI Scientific Inc., n.d.; AAPPTec, n.d). Specifically, hydrophobic and β-branched amino acid residues, which are common in defensins, are difficult to synthesize using this method (Tickler & Wade, 2007).

Microbial biosynthesis could increase the commercial viability of our product by reducing the costs and increasing the throughput of defensin synthesis through large-scale fermentation and improving the quality of synthesized peptides through adopting chassis organisms that are effective at synthesizing peptides with complex high-dimensional structures.

3D structure and bioactivity of α- and β- defensin. Some structures are dimers or tetramers.

Build-Test-Learn II

To determine whether defensins could be released by the cleavage of the SUMO tag by Ulp1 and whether the released defensins possessed antimicrobial activity, we performed deSUMOylation on samples of produced fusion proteins and performed qualitative agar dilution and quantitative broth dilution assays on common infection-causing bacteria. The results showed that our fusion proteins were indeed released and possessed antimicrobial activity, and we quantified the minimally inhibitory concentrations of our released defensins. The exact results can be found in our engineering success page.After transforming our constructed plasmids (pET28a-HNP1, pET28a-HNP4, pET28a-HD5, pET28a-HBD3) into SHuffle E. coli, our chassis organisms all showed visible growth. This did not match our hypothesis that when individually biosynthesized, defensins would be toxic to the host. However, we speculated that the synthesized defensins were either ineffective due to the disruption of their 3D structure because of the addition of extra amino acids on the N-terminus, which constituted the 6xHis-tag used for protein purification, or degraded by the host as it was synthesized due to the nature of ultra-light (less than 5 kDa) peptides being prone to the loss of function due to hydrolysis by proteases.

To avoid the direct impact of extra amino acids on the N-terminus of the defensin and increase the total length of the synthesized protein, we came up with Design III by combining binding domains with the SUMO tag and natural AMP sequences. After the SUMO tag is cleaved by the Ulp 1 enzyme on the wound site, the mature AMP sequence without extra amino acid residues on the N-terminus would be released. The addition of the binding domain and the SUMO tag also increased the molecular weight of our protein from less than 5 kDa to around 25 kDa, therefore effectively reducing the likelihood of proteolytic degradation.

Design III

When directly biosynthesized, individual defensins would inevitably be toxic toward the chassis, making this approach infeasible. Our approach of synthesizing defensins as components of fusion proteins eliminates the possibilty of host toxicity because the fusion protein would not be bioactive until the functional module, defensins, is released after the cleavage of the SUMO tag by the Ulp 1 enzyme during the process of deSUMOylation.

Why SUMO?

During our research, we found multiple fusion partners for defensins, including small ubiquitine-like modifier (SUMO), thioredoxin (Tx), glutathione-S-transferase (GST), Mxe Gyra intein, and calmodulin (CaM) (Zhu et al., 2021). However, all candidates except SUMO and Mxe Gyra intein would leave extra amino acid residues on the N-terminus of the defensin, which we speculated could impact defensin function based on our results from DBTL cycle II. Moreover, Mxe Gyra intein requires dithiothreitol (DTT) as a buffer, which could disrupt the functionally important disulfide bonds in defensins. Therefore, we selected SUMO, a tag covalently attached to proteins during a mechanism of post-translational modification called SUMOylation. During deSUMOylation, SUMO proteases, in our case being Ubiquitin-like-specific protease 1 (Ulp1), cleave out the SUMO protein, releasing the target protein in its mature form (Hickey et al., 2012). This process does not leave any extra amino acid residues on the N-terminus of the defensin and does not involve DTT, so it maximally preserves defensin function.

Initially, when we did not decide on using the SUMO system yet, we directly paired the collagen binding domain α2 with defensins. However, the fusion protein α2-defensin displayed low solubility, causing its purification to be extremly difficult. Therefore, we switched to the cellulose binding domain CBM3, which has higher solubility. In addition, the SUMO tag itself is a solubility tag and thus could increase the solubility of the entire fusion protein (Marblestone et al., 2006).

We paired CBM3 with each of the 4 defensins to determine whether the SUMO tag could be successfully cleaved by Ulp1 and whether the released defensins possessed antimicrobial activity.

Build-Test-Learn III

To determine whether defensins could be released by the cleavage of the SUMO tag by Ulp1 and whether the released defensins possessed antimicrobial activity, we performed deSUMOylation on samples of produced fusion proteins and performed qualitative agar dilution and quantitative broth dilution assays on common infection-causing bacteria. The results showed that our fusion proteins were indeed released and possessed antimicrobial activity, and we quantified the minimally inhibitory concentrations of our released defensins. The exact results can be found in our engineering page.

Design IV

After verifying the feasibility of our deSUMOylation release system using one binding domain, CBM3, we aimed to characterize other FBSDs. Due to time constraints and unexpected bacterial contamination, we only paired the collagen binding domain α2 and chitosan binding domain CBM5, both of which we verified the binding effectiveness of in DBTL Cycle I, with the defensin HNP1. We also only conducted qualitative agar dilution assays to verify the release of the defensins.

Build-Test-Learn IV

After deSUMOylation, the results showed that, like our results in DBTL cycle III, our fusion proteins were released and possessed antimicrobial activity. The exact results can be found in our results page. (link to results)

The agar dilution results verified the applicability of our fusion protein design to various binding domains. Combined with our results in DBTL cycle III verifying the activity of all 4 types of released defensins, this completed our construction of an arsenal of antimicrobial fusion proteins, each with their specific advantage in terms of bioactivity toward specific strains of bacteria and each capable of forming coatings on every type of first aid material that we selected.

Design V

The Problem: Expensive Ulp1

The core of our release mechanism of defensins is the Ulp1 protease, which cleaves out the defensin from the SUMO tag and the binding domain for its release into the wound site. However, the Ulp1 that we purchased from Beyotime, a Chinese biotechnology company, for testing has extremely low cost-effectiveness due to the complexities within enzyme synthesis, purification, and transportation causing costs to be high. Thus, from a business perspective, complete reliance on commercial Ulp1 would cause our first-aid kits to become commercially infeasible and increase our dependence on unpredictable external manufacturers.

Our Solution: Biosynthesis of rtUlp1

We decided to biosynthesize Ulp1 ourselves to eliminate costs related to transportation and the profit margin of external companies. To further reduce the cost, we synthesized a codon-optimized, truncated version of Ulp1 that we found in the literature (Wang et al., 2016). This bioactive enzyme, rtUlp1, is much shorter in length and therefore required a much simpler and cheaper synthesis process (Wang et al., 2016).

In our initial design, users need to carefully mix rtUlp1 stored in powder form with water and spray the water onto the wound dressing with defensins when using our product. However, when we presented our product to Mr. Dawa in our interview, he alerted us that our product is too complicated to use especially in emergency situation. Therefore, we decided to simplify the whole process for our users by also affixing rtUlp1 to the wound dressing by fusing it with a binding domain, so that our users only need to saturate the wound dressing with water before using it.

Build-Test-Learn V

We tested the optimized enzymes on the fusion protein CBM3-SUMO-HNP1. An agar dilution assay showed that the fusion protein with rtUlp1 added possessed an inhibitory effect. Thus, our effort of synthesizing rtUlp1 ourselves succeeded, and the synthesized protease was able to cleave out the SUMO tag from our fusion proteins. However, due to time constraints, we were not able to synthesize and affix rtUlp1 fused with a binding domain onto first aid materials.

We proved that our synthesis of the optimized rtUlp1 with a level of bioactivity comparable to commercial Ulp1 is a feasible approach to increasing the commercial viability of our product. We also realized the importance of simplifying procedures of application for our product under the complex circumstances of war and natural disasters.

As mentioned in Design II, cationicity and hydrophobicity are both key contributors to the antimicrobial activity of defensins. In addition, they are also key determinants in the dimerization rate of defensins (Rajabi et al., 2012). A stable dimer structure is also key to defensin binding and function (Pazgier et al., 2012).

Based on the principle of increasing dimerization rates, cationicity, and hydrophobicity, we attempted to improve the stability and antimicrobial activity of one α-defensin, HNP1, by rationally optimizing its amino acid sequence through individual mutagenesis. We successfully purified 3 of the 4 optimized defensins. However, all showed lower antimicrobial activity compared to the original HNP1. We speculated that our modifications negatively impacted the ability of HNP1 to inhibit Lipid II, which is a proposed mechanism of action for defensins to act on Gram-positive bacteria like S. aureus, which we conducted the broth dilution assays on.

More details can be found in our main page for modelling.

Pichia pastoris Expression System

A common issue we encountered while synthesizing FBSDs through E. coli is that the target proteins were often contained in inclusion bodies within the cell that could not be easily purified out of the cell. This significantly reduced the actual yields of FBSDs we were able to achieve and impacted the commercial viability of our product.

To improve yields and thus our commercial viability, we desgined an expression system for HNP1 through the yeast Pichia pastoris, which, as described in the literature, could directly secrete HNP1 out of the cell instead of containing it in inclusion bodies (Zhang et. al, 2018). However, due to the slow growth of P. pastoris making experimental cycles especially long and the complex culture conditions it required, we were not able to assess the antimicrobial activity of the produced HNP1.

More details can be found in the corresponding section in our main page for outlook.

Addition of Solubility Tags

Another issue that we faced during FBSD synthesis was the low solubility of the target protein complicating the purification process. As high hydrophobicity is key to one of the mechanisms of action of defensins, FBSDs containing defensin sequences naturally suffered low solubility that outweighed the additional solubility provided by the SUMO tag. This also caused a significant decrease in the amount of FBSDs that could be purified out of the cell.

To improve the solubility of our FBSDs, we added 2 solubility tags to the N-terminus of CBM3-SUMO-HNP1. However, due to time constraints, we were not able to assess the solubility of the improved FBSDs.

More details can be found in the corresponding section in our main page for outlook.