Recombinant production of antimicrobial proteins in bacterial expression systems: Escherichia coli vs. lactic acid bacteria Authors: Sergi Travé-Asensio, Oriol Gómez, Jose Vicente Carratalá, Carla Robles, Francesc Fàbregas, Adrià López-Cano, Xavier Rodríguez Rodríguez, Imma Ratera, Anna Arís, Elena Garcia-Fruitós Published: 2025 ABSTRACT AbstractBackgroundThe development of antimicrobial treatments as alternatives to antibiotics to combat antimicrobial-resistant (AMR) bacteria is a global priority. Antimicrobial peptides and proteins such as Host Defense Peptides (HDPs) and endolysins are one of the alternatives that are being explored. HDPs are small, cationic, and amphiphilic antimicrobial peptides derived from the innate immune system exhibiting a broad-spectrum antimicrobial activity. On the other hand, endolysins are enzymes produced by bacteriophages to hydrolyze the bacterial peptidoglycan layer, offering more specific antimicrobial activity than HDPs. While short peptides can be chemically synthesized, this approach presents several limitations, and recombinant production is also being used. Escherichia coli is the most used bacterial expression system for protein production. Alternative systems based on Generally Recognized as Safe (GRAS) microorganisms such as Lactic Acid Bacteria (LAB) have also been employed. However, so far, no comparative studies have evaluated the production and activity of antimicrobial proteins expressed in E. coli versus LAB and this study aims to address that gap.ResultsTo evaluate potential differences in the production of antimicrobial proteins using E. coli and two LAB (Lactococcus lactis and Lactiplantibacillus plantarum) hosts, various proteins were evaluated. These included two HDPs fused to a GFP, two multidomain HDP-based proteins and one endolysin. The results revealed a clear influence of the expression system on the quality of HDP-based protein, including both GFP fusions and multidomain constructs. Protein yield was higher in E. coli and all HDP-based proteins exhibited higher antimicrobial activity when expressed in E. coli compared to L. lactis and L. plantarum. In contrast, endolysin activity was comparable when produced in E. coli and L. lactis.ConclusionsThese results demonstrate that the choice of bacterial expression host significantly affects not only the yield but, more importantly, the antimicrobial activity of HDP-based proteins. For these proteins, the antimicrobial activity was consistently higher when produced in E. coli. In contrast, endolysins exhibited similar characteristics regardless of whether they were expressed in E. coli or in L. lactis.Keywords: Recombinant host defense peptides, Recombinant endolysins, Escherichia coli, Lactic acid bacteria, Protein activity Background In recent years, bacterial infections caused by antibiotic resistant pathogens have continued to rise, leading to the implementation of restrictions on antibiotic use. In this scenario, the development of new antimicrobial therapeutic alternatives has become a major challenge [1–3]. Some of these emerging alternatives include therapies based on bacteriophages [4], antimicrobial peptides and proteins [5], antibodies [6], nanoparticles [7, 8] or essential oils [9, 10]. Among antimicrobial peptides, Host Defense Peptides (HDPs) stand out as one of the most promising alternatives due to their broad-spectrum antimicrobial activity [5, 11, 12] and their ability to exert immunomodulatory responses [13, 14]. In nature, HDPs are an important part of the defense mechanism of nearly all living organisms. Their mechanism of action relies on their cationic and hydrophobic properties, which enable them to interact with and permeabilize bacterial cell membranes [15]. In recent years, many HDPs have been identified, especially in animals. In mammals, these peptides are typically classified into two main families based on their structure: defensins and cathelicidins [13]. Their extraction from natural sources is laborious and typically results in low yields of purified peptide. As an alternative, chemical synthesis -particularly solid-phase peptide synthesis- offers a method that yield peptides with high purity. However, this approach is associated with high manufacturing costs, limitations in the synthesis of longer peptides, and the use of environmental hazardous reagents, which hinders its scalability for in vivo applications [16–18]. In contrast, recombinant production represents a cost-effective option for large-scale HDP production [19]. Nevertheless, their short length and cationic nature make them highly susceptible to proteolytic degradation during the production process. This limitation has previously been addressed by fusing HDPs to carrier proteins that help stabilize the peptide. However, the removal of these carriers to obtain the final product add extra steps to the purification process, ultimately reducing the overall peptide yield [20, 21]. A recently developed alternative for efficient recombinant production involved fusing different HDPs into a single, larger polypeptide, thereby avoiding the use of non-functional carrier proteins. This strategy not only increases the molecular size but also results in a multidomain protein composed entirely of functional peptides, resulting in a functional protein with improved activity [22, 23]. Another group of antimicrobials currently under investigation are endolysins -enzymes secreted by bacteriophages (phages) during their lytic cycle [24, 25]. Endolysins hydrolyze the bacterial peptidoglycan layer to ensure the phage release. Unlike HDPs, endolysins exhibit more specific antimicrobial activity by cleaving particular bonds within the peptidoglycan structure. Although peptidoglycan composition is pretty conserved among bacteria, variations exist in the glycan chain, peptide stem or the types of cross-linking between peptides [26]. That differences explain why endolysins can discriminate between different types of bacterial cell wall. Depending on which bonds they target, endolysins are classified as glycosidases (glucosaminidases, muramidases, and transglycosylases), amidases or endopeptidases [27]. Due to their size, endolysins cannot be synthesized chemically and are instead produced recombinantly in various host systems. In general, Escherichia coli has been the most widely used host for recombinant protein production [19, 28, 29], owing to its extensive array of genetic tools, plasmids, and well-established protocols, which typically results in higher protein yields [30, 31]. However, the presence of lipopolysaccharide (LPS) in its cell wall and its ability to trigger undesired immune response, makes its removal necessary from the purified product for in vivo applications. In contrast, Lactic Acid Bacteria (LAB) are LPS-free bacteria widely used in the food industry due to their probiotic properties and their classification as Generally Recognized As Safe (GRAS) [32]. For this reason, LAB species such as Lactiplantibacillus plantarum or Lactococcus lactis are promising candidates as recombinant host systems for biopharmaceutical production [33–37]. Each host system presents its own advantages and disadvantages, which must be carefully considered when selecting the most suitable expression platform for a given protein. While some studies have previously compared the heterologous expression of proteins across different hosts [38, 39], to the best of our knowledge, no studies have directly compared the production and activity of the same antimicrobial protein expressed in both E. coli and two LAB (L. plantarum and L. lactis), to determine the most optimal strategy for their recombinant production. Methods Bacterial strains E. coli BL21 (DE3), L. lactis subsp. cremoris NZ9000 (MoBiTech, Goettingen, Germany) and L. plantarum NZ7100 (kindly provided by Dr. Pascal Hols, UCLouvain) [40] were used for recombinant protein production. Klebsiella pneumoniae carbapenemase (KPC), methicillin resistant Staphylococcus aureus (MRSA), Pseudomonas aeruginosa, Acinetobacter baumannii and methicillin-resistant Staphylococcus epidermidis (MRSE) were used to evaluate the antimicrobial activity of the HDPs studied. E. coli and P. aeruginosa strains were grown in Luria-Bertani (LB) medium, L. lactis strains in M17 broth (Sigma-Aldrich, Madrid, Spain) with 0.5% glucose, L. plantarum in Man, Rogosa and Sharpe (MRS) medium (Condalab, Madrid, Spain), and KPC, MRSA and MRSE strains in Brain-Heart Infusion (BHI) broth. Genetic construct design and cloning Mature sequences of bovine lingual antimicrobial peptide (LAP; Uniprot entry Q28880, V25-K64) and cathelicidin LL37 (Uniprot entry P49913, L134-S170) were fused to green fluorescence protein (GFP) gene through the linker sequence SGGGSGGS, as described in [23]. The same linker was used to combine bovine β-defensin 3 (bβD3; Uniprot entry P46161 Q16-W57), bovine β-defensin 1 (bβD1; Uniprot entry P46159), LAP and bactenicin 5 (Bac5; Uniprot entry P19660, R131-P173) into a single polypeptide sequence (βD3βD1LAPBac5). The same strategy was followed to fuse four domains of bβD1 in a tandem multidomain protein (βD1βD1βD1βD1). The endolysin sequence used (LysJEP8) has been previously described by [41]. All the constructs were finally fused to a C-terminal 6xHis-tag for protein purification purposes. The final constructs were chemically synthesized by GeneArt (Life technologies, Regensburg, Germany). The codon optimized sequences for E. coli were cloned into pET22b (AmpR) vector and transformed by heat shock in competent E. coli BL21 (DE3) cells. The codon optimized sequences for L. lactis were cloned into pNZ8148 (CmR; MoBiTech) and transformed by electroporation in competent L. lactis NZ9000 [42] and L. plantarum cells [43]. Determination of soluble and insoluble protein fraction E. coli strains cultures were grown in 300 mL shake flasks each containing 50 mL of LB medium supplemented with 100 µg/mL of ampicillin. Overnight (O/N) cultures at an initial OD600nm of 0.05 were grown at 37 °C and 250 rpm and they were used to inoculate the shake flasks that were grown at 37 °C and 250 rpm until reached an OD600nm = 0.4–0.6. Then, protein expression was induced by 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG, Panreac Applichem, Castellar del Vallès, Spain) and subsequently cultures were grown for 3 h, being production conditions previously established in the lab [44]. L. lactis strain was grown in M17 medium supplemented with 0.5% of glucose and 5 µg/mL of chloramphenicol. O/N cultures at an initial OD600nm of 0.05 were grown at 30 °C without agitation and they were used to inoculate 50 mL tubes with 50 mL of media that were grown under static conditions at 30 °C until reached an OD600nm = 0.4–0.6. Then, heterologous protein production was induced with 12.5 ng/mL nisin, and cultures were grown for additional 3 h, as previously described [45]. Once the production process finished, 25 mL of each culture were recovered and centrifugated (6,200 x g for 15 min at 4 °C) and the resultant pellets were stored at -80 °C. The next day, the pellets were resuspended in 0.5 mL of PBS with EDTA-free protease inhibitor cocktail (Complete EDTA-free, Switzerland, Roche) and bacteria were disrupted by sonication (3 cycles of 3 min, 0.5s on, 0.5s off at 10% amplitude) (Branson SFX550 Sonifier, Emerson, Spain). Then, samples were centrifugated (15,000 x g for 15 min at 4 °C) to separate the soluble and insoluble protein fractions. To determine the percentage produced of each fraction, they were analyzed by Western blot gels and ImageJ software. Recombinant protein production Recombinant production was done as previously mentioned but using bigger volumes. E. coli cultures were grown in 2.5 L shake flasks each containing 500 mL of LB medium supplemented with 100 µg/mL of ampicillin. Then, protein expression was induced by 1 mM IPTG and subsequently cultures were grown for 3 h. L. lactis and L. plantarum were grown in 1 L bottles with 1 L of media (M17 + 0.5% glucose for L. lactis or MRS for L. plantarum) supplemented with 5 µg/mL of chloramphenicol and that were grown under static conditions at 30 °C and 37 °C, respectively. Heterologous protein production was induced with 12.5 ng/mL nisin in L. lactis and with 50 ng/mL sakacin P inducing peptide (SppIP) (GenScript, Sequence: MAGNSSNFIHKIKQIFTHR), and cultures were grown for additional 3 h. In all the cases, cells were harvested by centrifugation (6,200 x g for 15 min at 4 °C) and the resultant pellets were stored at -80 °C until purification. Recombinant protein purification Cell pellets of 500 mL harvested during the production process were first resuspended in 30 mL binding buffer (20 mM Tris, pH 7.4, 500 mM NaCl and 20 mM imidazole) with EDTA-free protease inhibitor cocktail. Then, the samples were disrupted using a Constant Systems CF1 disruptor (Constant Systems, Daventry, United Kingdom) at 40 KPsi during 2 rounds for L. lactis and L. plantarum samples and at 20 KPsi during 1 round for E. coli samples. The disrupted samples were centrifugated for 45 min at 15,000 x g and 4 °C. After that, the supernatant was recovered and filtered (Ø 0.2 μm). The proteins were then purified by Immobilized Metal Affinity Chromatography (IMAC) in ÄKTA Start (Cytiva, Hospitalet de Llobregat, Spain) using 1mL HisTrap chelating HP columns (Cytiva, Hospitalet de Llobregat, Spain) following the manufacturer instructions. The 6xHis-tag was not removed. The eluted protein was dialyzed O/N in 0.01% acetic acid (or in 20 mM Tris-HCl pH 7.4 in the case of the endolysin LysJEP8) at 4 °C in agitation. Dialysis was performed to remove imidazole and have the proteins in the optimal buffer for activity assays. Protein concentrations were quantified by Qubit 4 (Thermo Fisher, Madrid, Spain) and their purity was determined by Western blot and Coomassie gels with further analysis with ImageJ software. Antimicrobial assay BacTiter-Glo™ Microbial Cell Viability assay (Promega, Madison, WI, USA) was used to determine the antimicrobial activity of the antimicrobial proteins. Briefly, an O/N culture of the selected pathogenic strain was grown in BHI (KPC, MRSA and MRSE) or LB (P. aeruginosa PAO1, P. aeruginosa ATCC-10145 and A. baumannii ATCC-15308). The next day, the O/N was diluted 1:100 in BHI, separated in different eppendorfs and centrifugated at 6,200 x g and 4 °C for 15 min. The resultant pellets were resuspended with 150 µL of either 0.01% acetic acid (or with 20 mM Tris-HCl pH 7.4 for LysJEP8) (negative control) or 5 µM of HDPs or 4 µM of endolysin and disposed in a sterile 96-well plate of polypropylene (Costar, Madrid, Spain). Then, the microplate was incubated for 5 h at 37 °C. After incubation, 100 µL of each well were transferred to a 96-well polystyrene microplate and then mixed with 100 µL of BacTiter-Glo™ reagent. The microplate was shaken for 30 s and then incubated for 5 min at room temperature. Finally, the luminescence was measured in a LUMIStar Omega microplate reader (BMG Labtech, Ortenberg, Germany). The bacterial survival was assessed by comparing the luminescence of the treated wells with the negative controls (previously detailed). Dynamic light scattering Particle size of the antimicrobial proteins was determined by dynamic light scattering (DLS) in a Zetasizer Pro (Malvern Instruments Ltd, Malvern, UK). Proteins were diluted to a final concentration of 1 mg/mL in 0.01% acetic acid (or with 20 mM Tris-HCl pH 7.4 for LysJEP8) and mixed with increasing concentration of SDS (0, 1, 5, 10, 15 and 20%). After 30 min incubation at RT, the samples were centrifugated at 15,000 x g 15 min at 4 °C to remove non-specific aggregates. Further, the supernatants were analysed in quintuplicate. Zetasizer explorer software was used for data recovery and representation. Sulfhydryls analysis Sulfhydryl groups were quantified using the 4,4′-dithiodipyridine (DTDP) method as previously described [46] with minor modifications. Briefly, each protein replicate was adjusted to a final volume of 1 mL and diluted with 0.01% acetic acid to a final concentration of 30 µM of cysteines present in each protein. The pH was adjusted to ≤ 7.0 by adding 200 µL of a strong buffer (100 mM NaH₂PO₄ and 0.2 mM EDTA, adjusted to pH 6.8 with NaOH). This buffer (referred to as buffer 6.8 throughout the manuscript) was used to maintain the pH stability during the experiment. Subsequently, 50 µL of 4 mM DTDP reagent was added, followed by immediate vortexing. Absorbance at 324 nm (A₃₂₄) was measured against a water blank after a 5 min incubation at room temperature. Reagent and protein blanks (A₃₂₄r and A₃₂₄p, respectively) were also measured. All measurements were performed in triplicate. The amount of free sulfhydryls (nmol) was calculated as described by Riener et al. [46]. Results and discussion The choice of recombinant expression host -and even the specific strain- can influence not only protein yield, but also protein characteristics, with biological activity being one of the most critical factors to consider [39, 44, 47]. In this context, the present study aims to investigate whether the use of different bacterial expression systems, including Gram-negative and Gram-positive species, affects the production yield and activity of different antimicrobial proteins, such as HDPs and endolysins. First, two different HDPs, a beta-defensin named LAP and cathelicidin LL37 were studied when produced as fusion proteins with GFP as a carrier. Both were produced in soluble and insoluble fractions (Figs. 1A and 2A). Notably, differences between E. coli and L. lactis were observed only for LAP-GFP, which showed greater solubility when expressed in L. lactis compared to E. coli (Fig. 1A). However, in terms of total protein yield, LAP-GFP was produced at higher levels in E. coli than in L. lactis (Fig. 1A). In contrast, LL37-GFP exhibited a different pattern, with similar production yields in both expression systems (Fig. 2A). Antimicrobial activity assays revealed that both LAP-GFP and LL37-GFP displayed significatively higher activity when produced in E. coli (Figs. 1B and 2B). The survival rate of MRSA was significantly lower using both LAP-GFP (Fig. 1B) and LL37-GFP (Fig. 2B) produced in E. coli compared to those produced in L. lactis. Similar results were observed using LL37-GFP produced in E. coli against KPC (Fig. 2B), whereas LAP-GFP performed similar but without significant differences (Fig. 1B). A Graph: Soluble (light grey) and insoluble (dark grey) yields of LAP-GFP production in E. coli and L. lactis strains. Table: Purified soluble protein yields per L of culture (mg/L) and per L of culture normalized to culture optical density (mg/L·OD). B Antimicrobial activity of LAP-GFP produced in E. coli (grey) and L. lactis (white) against MRSA and KPC. C Sulfhydryls determination of LAP-GFP obtained in E. coli (grey) and L. lactis (white) strains. D Particle size determination by DLS. Unpaired t test with Welch’s correction was used as statistical model for A, B, and C sections. Asterisks depict significant differences between samples (p < 0.05) A Graph: Soluble (light grey) and insoluble (dark grey) yields of LL37 production in E. coli and L. lactis strains. Table: Purified soluble protein yields per L of culture (mg/L) and per L of culture normalized to culture optical density (mg/L·OD). B Antimicrobial activity of LL37-GFP produced in E. coli (grey) and L. lactis (white) against MRSA and KPC. C Particle size determination by DLS. Unpaired t test with Welch’s correction was used as statistical model for A and B sections. Asterisks depict significant differences between samples (p < 0.05) The number of free sulfhydryl groups in LAP-GFP was analyzed to assess their potential influence on the protein’s final activity (Fig. 1C). This analysis was not applicable to LL37 which lacks disulfide bonds in its structure. Despite the observed differences in protein activity, no significant differences were found on the number of free sulfhydryl groups in LAP-GFP (Fig. 1C). Regarding protein structure, particle size determination was done by DLS. Proteins expressed in E. coli formed nanoparticles of approximately 20 nm, which were highly stable and only became unstructured (reaching sizes of around 1 nm) when exposed to SDS at 20% for LAP-GFP (Fig. 1D) and 5% for LL37-GFP (Fig. 2C). In contrast, when LAP-GFP was produced in L lactis, it formed aggregates (Fig. 1D), which were disrupted by low SDS concentrations (1%) resulting in particles of sizes between 1 and 2 nm (Fig. 1D). To determine whether the differences in protein activity observed for HDPs fused to GFP in the two bacterial expression systems were an isolated case or a general trend, other antimicrobial proteins were evaluated. Two multidomain antimicrobial proteins (βD3βD1LAPBac5 and βD1βD1βD1βD1) were selected. These chimeric proteins combine multiple HDPs in a single polypeptide, eliminating the need for a carrier protein such as GFP. Solubility in each expression system was protein dependent. βD3βD1LAPBac5 was fully soluble when expressed in E. coli but not in L. lactis (Fig. 3A), whereas βD1βD1βD1βD1 was completely soluble in L. lactis but showed partial aggregation when produced in E. coli (Fig. 4A). E. coli again proved to be the most suitable expression system to produce HDP-based antimicrobial proteins in terms of yield (Figs. 3A and 4A) and antimicrobial activity (Figs. 3B and 4B). βD3βD1LAPBac5 and βD1βD1βD1βD1 exhibited higher activity when produced in E. coli (Figs. 3B and 4B), particularly βD1βD1βD1βD1, which showed significant lower MRSE survival rates compared to the same protein produced in L. lactis (Fig. 4B). In these cases, the increased activity values of chimeric proteins correlated with a lower number of free sulfhydryl groups (Figs. 3C and 4C). Nanostructuration analysis revealed protein-dependent behavior. For βD3βD1LAPBac5, we observed differences in the size of the proteins when E. coli and L. lactis are compared (Fig. 3D). However, βD1βD1βD1βD1 exhibited similar characteristics and behavior regardless of the host used (Fig. 4D). A Graph: Soluble (light grey) and insoluble (dark grey) yields of βD3βD1LAPBac5 production in E. coli and L. lactis strains. Table: Purified soluble protein yields per L of culture (mg/L) and per L of culture normalized to culture optical density (mg/L·OD). B Antimicrobial activity of βD3βD1LAPBac5 produced in E. coli (grey), L. lactis (white) against MRSE. C Sulfhydryls determination of βD3βD1LAPBac5 obtained in E. coli (grey) and L. lactis (white) strains. D Particle size determination by DLS. Unpaired t test with Welch’s correction was used as statistical model for A, B and C sections. Asterisks depict significant differences between samples (p < 0.05) A Graph: Soluble (light grey) and insoluble (dark grey) yields of βD1βD1βD1βD1 production in E. coli and L. lactis strains. Table: Purified soluble protein yields per L of culture (mg/L) and per L of culture normalized to culture optical density (mg/L·OD). B Antimicrobial activity of βD1βD1βD1βD1 produced in E. coli (grey) and L. lactis (white) against MRSE. C Sulfhydryls determination of βD1βD1βD1βD1 obtained in E. coli (grey) and L. lactis (white) strains. D Particle size determination by DLS. Unpaired t test with Welch’s correction was used as statistical model for B, C and D sections. Asterisks depict significant differences between samples (p < 0.05) At this stage, a third expression system (Lactiplantibacillus plantarum) was included in the study to assess whether the reduced antimicrobial activity was specific to L. lactis or could be generalized to other LAB. The antimicrobial activity of the proteins βD3βD1LAPBac5 (Fig. 5A) and βD1βD1βD1βD1 (Fig. 5B) against P. aeruginosa PAO1 showed comparable activity values when produced in L. lactis and L. plantarum, both of which were consistently lower than those observed for the same protein produced in E. coli. Antimicrobial activity of βD3βD1LAPBac5 (A) and βD1βD1βD1βD1 (B) produced in E. coli (grey), L. lactis (white) and L. plantarum (dark grey) against P. aeruginosa PAO1. Brown-Forsythe and Welch ANOVA test was used as statistical model. Same letters depict not significant differences between groups, while different letters depict significant differences between groups (p < 0.05) Taken together, these results demonstrated a clear influence of the expression system on the yield and quality of HDP-based protein. A consistent trend was observed in terms of antimicrobial activity: all proteins exhibited higher activity when expressed in E. coli compared to LAB systems (Figs. 1B, 2B, 3B and 4B, and 5). Moreover, in general, the yield was also higher when E. coli was used (Figs. 1A, 2A and 3A, and 4A). An antimicrobial protein from a category distinct from HDPs such as an endolysin protein named LysJEP8, was also included in this study. In this case, unlike the HDP-derived proteins, the expression system had no observable impact on protein’s antimicrobial activity (Fig. 6B). Although a higher production yield was achieved in E. coli (Fig. 6A), the quality of LysJEP8 was not determined by the expression host (Fig. 6B). A Soluble (light grey) and insoluble (dark grey) yields of LysJEP8 production in E. coli and L. lactis strains. B Antimicrobial activity of LysJEP8 produced in E. coli (grey) and L. lactis (white) against A. baumannii and P. aeruginosa These results indicated that, unlike what was observed for HDP-derived proteins, endolysins did not exhibited improved performance when produced in E. coli. Specifically, LysJEP8 expressed in E. coli and L. lactis showed comparable antimicrobial activity, as reflected by similar bacterial survival ratios in the assays (Fig. 6). Several factors may account for these differences. First, HDPs and endolysins have distinct modes of action. The endolysin LysJEP8 used in this study is an endopeptidase that specifically cleaves the peptide bond between L-alanine and D-glutamic acid in the peptidoglycan structure present in Gram-negative bacteria [41]. In contrast, HDPs are broad-spectrum and non-specific antimicrobial peptides with different modes of action, including membrane disruption combined in many cases with an action against intracellular targets, being also able to modulate the immune system by various means. Therefore, the cellular context in which these proteins are expressed may differently affect their folding, stability and ultimately biological activity. Furthermore, unlike endolysins, a characteristic feature of mammalian beta defensins is the presence of six conserved cysteine residues that form three disulfide bonds (Cys1–Cys5, Cys2–Cys4, and Cys3–Cys6) [48]. The formation of these bonds may influence protein activity. In this study, the antimicrobial activity of LAP-GFP (Fig. 1B) did not correlate with the disulfide bonds formation (Fig. 1C), which is consistent with previous studies reporting that disulfide bonds in HDP are dispensable for their antimicrobial function [48–50]. However, in the case of multidomain proteins based on HDPs, we observed a lower number of free sulfhydryl groups, indicating greater disulfide bond formation, when expressed in E. coli (Figs. 3C and 4C), and this was correlated positively with their antimicrobial activity (Figs. 3B and 4B). Therefore, we did not observe a specific cause that could explain the observed pattern or predict the conditions that consistently enhance HDP activity. Moreover, contrary to the wide accepted view, but in agreement with previous findings from our group [44], our results showed that E. coli favors the disulfide bonds formation in HDPs. DLS analysis discarded nanostructuration as the unique parameter impacting on protein activity. LAP-GFP and LL37-GFP proteins produced in E. coli (the most active ones) formed more stable nanoparticles compared to those produced in L. lactis (Figs. 1D and 2C). By contrast, in the case of multidomain proteins, we observed a similar DLS profile for βD1βD1βD1βD1 (Fig. 4D) or even bigger nanoparticles in the case of L. lactis for βD3βD1LAPBac5 (Fig. 3D). Overall results suggested that the relationship between protein structure and function may be context-dependent and influenced by multiple unknown host-specific factors that can affect the final peptide processing. Multiple parameters must be considered when selecting between E. coli and L. lactis as expression hosts for antimicrobial protein production. Although a wider range of E. coli strains, plasmids and genetic tools is available compared to those for LAB used as expression systems, it remains essential to evaluate the overall outcome when selecting a specific strain for antimicrobial protein production purposes. Regarding yield, this study demonstrated that E. coli achieved the highest production yields for both HDPs and endolysins. However, when production yields were normalized to culture optical density (mg/L·OD) this trend varied depending on the protein produced (Figs. 1A, 2A, 3A, 4A and 6A). Moreover, this parameter can be optimized and scale-up under more controlled conditions, likely reducing differences between the two expression systems. Analysis of protein solubility (Figs. 1A, 2A, 3A, 4A and 6A) revealed that the solubility is also protein-dependent and therefore not a decisive criterion for system selection. In terms of protein activity E. coli outperformed L. lactis (Figs. 1B, 2B, 3B, 4B and 5), while endolysin quality remained unaffected by the expression system (Fig. 6B). Consequently, protein activity becomes the determining factor for host selection, making E. coli the only viable option for HDP production. For endolysins, both E. coli and LAB are suitable, with E. coli offering the advantage of being a well-established platform for biopharmaceutical production, whereas LAB provides a superior safety profile due to its long-standing use in the probiotic industry. Conclusions In summary, this study demonstrates that the bacterial expression host significantly affects the activity of HDP-based proteins in a multifactorial manner, but not that of endolysins. Therefore, it is crucial to evaluate each protein group individually, assessing not only protein yields but also protein activity, to select the most appropriate bacterial expression system for each specific case. Acknowledgements The authors are indebted to AGAUR for project 2021 SGR 01552 and 2021 SGR 438 and to CERCA Programme (Generalitat de Catalunya) and the European Social Fund for supporting our research. Author contributions ST-A designed the work and performed most of the experiments, interpreted the results and drafted the manuscript. OG, JVC, CR, FF, AL-C and XR-R performed experiments. AA, IR and EG-F conceived and supervised the study. All authors contributed to and approved the final version. Funding This work was financially supported by MICIU/AEI/ 10.13039/501100011033, grant number PID2022-136521OB-I00, PID2022-137332OB-I00 and CEX2023001263-S. ST-A. received a pre-doctoral fellowship from Ministerio de Universidades (Spain), and JVC was supported with a Margarita Salas grant for the training of young doctoral graduates (grant no. 722713). The authors are also grateful for the financial support received from Instituto de Salud Carlos III through and the Networking Research Center on Bioengineering, Biomaterials, and Nanomedicine (CIBER-BBN). The authors also acknowledge the ICTS NANBIOSIS for the support of the Biomaterial Processing and Nanostructuring Unit (U6) at ICMAB-CSIC (https://www.nanbiosis.es/portfolio/u6- biomaterialprocessing-and-nanostructuring-unit/) and the assistance of the ICMAB-CSIC Scientific & Technical Services: SOFT. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Anna Arís and Elena Garcia-Fruitós have contributed equally to this work. Contributor Information Anna Arís, Email: anna.aris@irta.cat. Elena Garcia-Fruitós, Email: elena.garcia@irta.cat. References Associated Data Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.