A synthetic gene encoding the modular protein T22-GFP-H6 was designed in house, produced by Geneart (Thermo Fisher Scientific, Waltham, MA, United States) and subcloned into the expression vector pET22b (Novagen, Merck KGaA, Darmstadt, Germany). T22 is a chemokine receptor (CXCR4) antagonist derived from the polyphemousin II that has an antiparallel β sheet structure stabilized by two disulfide bonds (Murakami et al., 1997). E. coli Origami B (Novagen) was used as expression system for the T22-GFP-H6 formats as shown in Table 1 (Protein NPs and IBs). In this E. coli strain, the thioredoxin reductase (trxB) and glutathione reductase (gor) genes are deleted, facilitating disulfide bond formation (Xu et al., 2008). A synthetic gene encoding IFN-γ-H6 was designed in house and produced by Geneart (Thermo Fisher Scientific) as a codon optimized version for Lactococcus lactis host and subcloned into pETDuet (Novagen) for E. coli expression system and pNZ8148 plasmid (Cm®, MoBiTec GmbH, Göttingen, Germany) for L. lactis expression system. ClearColi® BL21 (DE3) strain (Lucigen, Middleton, WI, United States) was used as expression system for IFN-γ-H6 protein formats (unassembled protein format and IBs) in an E. coli endotoxin-free strain (Table 1). IFN-γ-H6 was also produced in Lactococcus lactis expression strain NZ9000 (NICE®, Boca Scientific, Inc., Dedham, MA, United States) (Mireau 2005 AMB) in the same protein formats as in ClearColi® (Table 1).

pET22b-T22-GFP-H6 plasmid was transformed into E. coli Origami B (Novagen) by heat shock at 42°C and maintained in lysogenic broth (LB) supplemented with 100 μg/ml ampicillin. Recombinant protein was produced in the same medium O/N at 20°C upon induction with 0.1 mmol/L isopropyl- β-d-1-thiogalactopyranoside (IPTG; Sigma–Aldrich, Sant Louis, MO, United States). Cells were then harvested by centrifugation (10 min at 5,000 g). T22-GFP-H6 was purified from the soluble cell fraction by affinity chromatography (Supplementary Method). Protein purity was determined by SDS-PAGE gel electrophoresis and subsequent protein immunodetection by Western blot using an anti-His monoclonal antibody (Santa Cruz Biotechnology, Dallas, TX, United States; ref: Sc-57598). Protein integrity was then verified by MALDI-TOF mass spectrophotometry in a linear mode (20–80 kDa). For that, 5 µl of protein sample was first dialyzed against milli-Q H2O in a Millipore membrane for 20min and loaded then onto a MALDI plate with Siapinic Acid in sandwich mode. Final protein concentration was determined by Bradford assay. Formation of protein NPs (generated by the self-assembling of monomeric building blocks of recombinant protein) was determined by dynamic light scanning (DLS) analysis in a Zetasizer Nano ZS (Malvern panalytical, Malvern, United Kingdom). The factors involved in the formation of protein NPs from soluble recombinant proteins have been previously described and seem to be related to the cationic nature of the peptide located at the N-terminal end of the protein and the C-terminal His-tag (Unzueta et al., 2012; Serna et al., 2016).

Cultures of ClearColi® BL21 (DE3) cells transformed with the plasmid pETDuet-IFN-γ-H6 were incubated in a shake flask at 37°C and 250 rpm in LB medium supplemented with 100 μg/ml ampicillin. Protein expression was induced in the same medium by adding 1 mmol/L isopropyl-β-d-thiogalactopyranoside (IPTG). The cultures were then incubated at 20°C and 250 rpm O/N for protein production. Cells were harvested by centrifugation (10 min at 5,000 g). IFN-γ-H6 was purified from the soluble cell fraction by affinity chromatography (Supplementary Method). The final protein concentration was determined by Bradford assay, yielding 0.80 mg/ml.

pNZ8148 plasmid (Cm^R, NICE) was transformed into competent L. lactis NZ9000 bacteria, as described elsewhere (Cano-Garrido et al., 2016). Transformed L. lactis cells with pNZ8148-IFN-γ-H6 were grown in M17 medium supplemented with 0.5% glucose at 30°C without agitation. Antibiotics were used for plasmid selection such as chloramphenicol (5 μg/ml) and erythromycin (2.5 μg/ml). Protein production was induced in the same medium by adding 500 ng/ml nisin (Sigma–Aldrich). After induction, cultures were grown for 3 h. Bacteria were harvested at 5,000 g for 5 min at 4°C and washed twice with PBS. IFN-γ-H6 was purified from the soluble cell fraction by affinity chromatography (Supplementary Method). Protein purity was determined by SDS-PAGE gel electrophoresis and final protein concentration was determined by Bradford assay.

Different protein formats of T22-GFP-H6 and IFN-γ-H6 obtained from the soluble cells fractions of the expression hosts are represented in Supplementary Figure S1.

pET22b-T22-GFP-H6 plasmid was transformed into E. coli Origami B (Novagen) by heat shock at 42°C and maintained in LB supplememted with 100 μg/ml ampicillin. Recombinant protein was produced in the same medium for 3 h at 37°C upon induction with 1 mmol/L isopropyl- β-d-1-thiogalactopyranoside (IPTG). Cells from 20 ml of culture were then harvested by centrifugation (5 min at 5,000 g). Purification of IBs of T22-GFP-H6 was performed following protocol described in Supplementary Method. Protein purity was determined by SDS-PAGE gel electrophoresis and protein concentration was determined by western blot using an anti-His monoclonal antibody (Genescript # A00186) with a calibration curve of quantified soluble recombinant GFP-H6.

pETDuet-IFN-γ-H6 plasmid was transformed into ClearColi® BL21 (DE3) cells and grown as described for the soluble protein version. IB production was induced by adding 1 mmol/L isopropyl-β-D thiogalactopyranoside (IPTG) to ClearColi® BL21 (DE3) cultures grown in LB supplemented with 100 μg/ml ampicillin. After induction, the cultures were grown for 5 h at 37°C. Cells were harvested by centrifugation (10 min at 5,000 g). Purification of IBs of IFN-γ-H6 was performed following protocol described in Supplementary Method . The recombinant protein yield was estimated by comparison with a standard curve of known amounts of purified rBoIFN-γ protein quantified by the Bradford assay. Quantification was performed with Image Lab software (Bio-Rad).

pNZ8148-IFN-γ-H6 was transformed into L. lactis cells and cells were grown as described in the production of soluble IFN-γ-H6 protein version in this prokaryotic host. IB production was induced by adding 500 ng/ml nisin (Sigma–Aldrich). After induction, cultures were grown for 5 h. Final OD_{550 nm} was 3.02. Bacteria were harvested at 5,000 g for 5min at 4°C and washed twice with PBS. Purification of IBs of IFN-γ-H6 was performed following protocol described in Supplementary Method. Protein purity was determined by SDS-PAGE gel electrophoresis, and protein concentration was determined by western blot using an anti-His monoclonal antibody (Genescript # A00186) with a calibration curve of quantified soluble recombinant GFP-H6.

Schematic representation of purification of IBs of T22-GFP-H6 and IFN-γ-H6 obtained from the insoluble cells fractions of the expression hosts are showed in Supplementary Figure S1.

Caco-2 cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) free of sodium pyruvate (Biowest, Nuaillé, France), supplemented with 10% fetal bovine serum, 1% non-essential amino-acids (Biowest) and 2.5 μg/ml Plasmocin (InvivoGen, San Diego, CA, United States) in a humidified atmosphere of 5% CO2 and 95% air at 37°C. This cell line was used as it provided low or absent expression of CXCR4 to which T22 peptide acted as an antagonist (Tamamura et al., 2006; Rubie et al., 2011; Stuckel et al., 2020).

Cell viability was determined by the Beckman counter method using a ZTM Series Coulter-Counter (Beckman Coulter Inc. Bea, CA, United States). Briefly, 150,000 cells per well were seeded on 12-well plates. The next day, the cells were exposed to increasing concentrations of the different protein formats of T22-GFP-H6 (ranging from 4 × 10³ to 70 × 10³ ng/mL, or what is the same, 0,14 to 2.3 μmol/L) and IFN-γ-H6 (ranging from 18 to 180 ng/ml, or the same, 1–10 μmol/L). After 24 h, cells were counted, and survival curves were represented.

The proportion of live, apoptotic, and necrotic cells was assessed with Annexin V FLUOS staining kit (Sigma–Aldrich) following the manufacturer’s protocol. Thus, cells were seeded at a density of 150,000 cells per well in 12-well plates and the following day, they were treated with two different concentrations of each protein: 18 × 10³ ng/ml and 70 × 10³ ng/ml T22-GFP-H6, and 72 and 180 ng/ml IFN-γ-H6. After 24 h of exposure, the supernatant was collected in a 15 ml tube. Adherent cells were trypsinized and collected in the same tube as the supernatant for each condition. The mix was centrifuged and washed once with PSB 1X. The pellet was resuspended in 100 µl of incubation buffer with 2 µl of Annexin V and 2 µl of propidium iodide (PI). Unstained cells and resuspended cell mixture were used as compensation controls to improve the gating strategy. The cells were incubated for 15 min at room temperature and in the dark for staining before adding 100 µl more of incubation buffer per sample. Then, samples were analyzed by BD FACSCanto Flow Cytometer (BD Biosciences, Franklin Lakes, NJ, United States). More than 10,000 events per sample were analyzed. Among these events, the PI-negative and Annexin V-negative cells represent the living population, PI-negative and Annexin V-positive cells are considered apoptotic and cells positive for both PI and Annexin V are considered necrotic. Cells exposed for 30 min to 100 mM H₂O₂ were used as positive controls.

The alkaline single cell gel electrophoresis (Comet) assay is a powerful technique to detect directly both oxidative stress and its effects inducing DNA strand breaks at the level of individual cells. Here it was used to evaluate the genotoxic and oxidative DNA damage for caco-2 cells exposed to the different protein forms, using formamidopyrimidine DNA glycosylase (FPG) as previously described (Bach et al., 2014). Gelbond® films (GF) were used for this assay. Cells were exposed for 24 h to high concentrations of the different recombinant proteins and forms: 4, 8, 18, 36, and 70 μg/ml. The next day, the cells were trypsinized, collected, centrifuged, and resuspended in cold PBS 1X at a concentration of 1 × 10⁶ cells/mL. In the following step, cells were mixed with 0.75% low-melting-point agarose at 37°C and three drops of 7 μl each were placed in the GF. Two GF with identical samples were processed simultaneously in each experiment. Then, a lysis step was performed overnight by immersing the GF in ice-cold lysis buffer at 4°C (2.5 mol/L NaCl, 0.1 mol/L Na₂EDTA, 0.1 mol/L Tris base, 1% Triton X-100, 1% lauroyl sarcosinate, 10% DMSO), at pH 10. The following day, the GF replicates were gently washed twice (1 × 5 min, 1 × 50 min) in enzyme buffer at 4°C and pH 8 (10 mmol/L HEPES, 0.1 mol/L KCl, 0.5 mm l/L EDTA, 0.2 mg/ml BSA) and then, they were incubated for 30 min at 37°C in enzyme buffer (negative control) or FPG-containing enzyme buffer. Later, the GF were washed with electrophoresis buffer and placed in a horizontal electrophoresis tank for 35 min to allow the DNA unwinding 0.3 ml/L NaOH and 1 mmol/L Na₂EDTA pH 13.2 before the electrophoresis, which was carried out for 20 min at 0.8 V/cm and 300 mA at 4°C. Then, GF were rinsed with cold PBS for 15 min and fixed in absolute ethanol for 2 h before air-drying it overnight at room temperature. A staining step was performed by incubating the GF for 20 min with SYBR Gold 1/10,000 in TE buffer (10 mmol/L Tris, 1 mmol/L EDTA pH 7.5). Finally, gels were mounted, visualized for comets using an epifluorescent microscope at ×20 magnification, and analyzed with the Komet 5.5 Image analysis system (Kinetic Imaging Ltd., Liverpool, United Kingdom). The levels of DNA damage were evaluated according to the percentage of DNA in the tail of the cells. One hundred randomly selected comet images were analyzed per sample. Cells exposed to the DNA damaging agent methyl methanesulfonate (MMS,Sigma–Aldrich) at 200 μmol/L for 1 h at 37°C were used as controls.

Analysis of variance followed by Dunnett’s multiple comparison test was performed appropriately to compare the effect of the exposure to the different recombinant proteins. A two-sided p < 0.05 was considered statistically significant in all cases.

Gel electrophoresis of T22-GFP-H6 and IFN-γ-H6 showed the expected protein bands (Supplementary Figure S2). In the soluble cell fraction (Supplementary Figure S2A) the main protein band corresponded to the recombinant protein showing a purity >90%. Some other accompanying protein bands of higher molecular weight corresponded to a small proportion of quaternary structures. In fact, dimerization of GFP (Wiedenmann et al., 2000) and IFN-γ (Fensterl and Sen, 2009) has been widely described and observed even under denaturing conditions of the SDS-PAGE electrophoresis. The presence of protein NPs in the T22-GFP-H6 sample was determined by DLS (Supplementary Figure S4) as previously described (Céspedes et al., 2016). Protein identification in the insoluble cell fraction was performed by immunodetection (Supplementary Figure S2B). In each sample, the protein band compatible with the corresponding expected molecular weight could be detected although analysis on polyacrylamide gels showed variable purity (Supplementary Figure S3). In fact, the relative presence of the recombinant protein in such samples depends on the protein itself and the experimental conditions set during gene expression with a wide range of purity between experiments for different proteins (Peternel et al., 2008; Castellanos-Mendoza et al., 2014; Carratalá et al., 2020b). The presence of associated proteins as chaperones and other cellular components in IBs have been widely described (Ventura and Villaverde, 2006).

The analysis of diverse endpoints linked to the toxicity potential in vitro showed that neither T22-GFP-H6 nor IFN-γ-H6 purified from ClearColi® or L. lactis were toxic in any of the protein forms assessed (natural IBs or soluble protein formats). Caco-2 cells had a cell survival rate generally over 80% when exposed to concentrations within the standard range, and no significant differences were found between exposed cells and non-exposed controls (Figures 1A,C,E). This was further confirmed by the results from the PI and Annexin V staining followed by flow cytometry analysis, where living cells were again found to be above 80% for all concentrations tested (Figures 1B,D,F), and in any case, these percentages were significantly different from the non-exposed controls. The proportion of apoptotic and necrotic cells in the culture was characterized for an intermediate and a high concentration of exposure (72 and 180 ng/ml for IFN-γ-H6; and 18 × 10³ and 70 × 10³ ng/mL of T22-GFP-H6), confirming the results of the cell viability assays (Figures 1B,D,F). Approximately 2% of the cell population were apoptotic cells in those exposed to IFN-γ-H6, while around 4% of the cells were apoptotic after exposure to T22-GFP-H6. Necrotic cells represented 1% of the caco-2 cells exposed to IFN-γ-H6 from ClearColi®, 6% of the cells exposed to IFN-γ-H6 from L. Lactis, and 10% of the cells exposed to T22-GFP-H6. Nonetheless, none of these percentages were significantly different from the proportions observed in non-exposed cells. In line with the above, the analysis of oxidized DNA bases by the comet assay with the inclusion of FPG enzyme also points out toward the absence of induced oxidative stress, as oxidative DNA damage did not exceed 10% for any of the conditions tested (Figure 2). Importantly, the analysis of DNA damage in individual cells by the Comet assay revealed no genotoxic effects of the recombinant proteins despite the increment in the dose of exposure. Genotoxic DNA damage levels did not differ from those found in the control cells, with values no higher than 12% (Figure 2). There were no significant differences in the levels of DNA damage when comparing T22-GFP-H6, IFN-γ-H6 from ClearColi®, and IFN-γ-H6 from L. lactis, nor between IBs or the soluble form of the different proteins.