This is the submitted version of the article: Carratalá, José Vicente; Cano Garrido, Olivia; Sánchez, Julieta María; [et al.]. «Aggregation-prone peptides modulate activity of bovine interferon gamma released from naturally occurring protein nanoparticles». New Biotechnolgy, Vol. 57 (July 2020), p. 11-19. DOI 10.1016/j.nbt.2020.02.001 This version is avaible at https://ddd.uab.cat/record/218927 under the terms of the license 1 Title. 2 Aggregation‐prone peptides modulate activity of bovine interferon gamma released from 3 naturally occurring protein nanoparticles 4 Author names and affiliations. 5 José Vicente Carrataláa,b, Olivia Cano‐Garridoa,b,c,1, Julieta Sáncheza,2, Cristina 6 Membradoa,b, Eudald Péreza,b, Oscar Conchillo‐Soléa, Xavier Dauraa,d, Alejandro Sánchez‐ 7 Chardie, Antonio Villaverdea,b,c, Anna Arísf, Elena Garcia‐Fruitósf, Neus Ferrer‐Mirallesa,b,c# 8 a 9 Institut de Biotecnologia i de Biomedicina, Universitat Autònoma de Barcelona, 10 Bellaterra, Barcelona, Spain b 11 Departament de Genètica i de Microbiologia, Universitat Autònoma de Barcelona, 12 Bellaterra, Barcelona, Spain c 13 CIBER de Bioingeniería, Biomateriales y Nanomedicina (CIBER‐BBN), Bellaterra, Barcelona, 14 Spain d 15 Catalan Institution for Research and Advanced Studies, Barcelona, Spain e 16 Departament de Biologia Evolutiva, Ecologia i CiènciesAmbientals, Facultat de Biologia, 17 Universitat de Barcelona, Av. Diagonal 643, 08028 Barcelona, Spain. f 18 Department of Ruminant Production, Institut de Recerca i Tecnologia Agroalimentàries 19 (IRTA), Caldes de Montbui, Barcelona, Spain 1 20 #Address correspondence to Neus Ferrer‐Miralles, neus.ferrer@uab.cat. 1 21 Present address: Nanoligent SL. Edifici Eureka. Av. de Can Doménech s/n. Campus de la 22 UAB. Bellaterra, Barcelona, Spain 2 23 Permanent address: Universidad Nacional de Córdoba, Facultad de Ciencias Exactas, 24 Físicas y Naturales, ICTA and Departamento de Química, Cátedra de Química Biológica, 25 Córdoba, Argentina, CONICET, Instituto de Investigaciones Biológicas y Tecnológicas 26 (IIBYT), Córdoba, Argentina. 27 28 29 30 Abstract 31 Efficient protocols for the production of recombinant proteins are indispensable for the 32 development of the biopharmaceutical sector. Accumulation of recombinant proteins in 33 naturally‐occurring protein aggregates is detrimental to biopharmaceutical development. 34 In recent years, the view of protein aggregates has completely changed with the 35 recognition that these aggregates are a valuable source of functional recombinant 36 proteins. In this study, bovine interferon‐gamma (rBoIFN‐) was engineered to enhance 37 the formation of protein aggregates (also known as protein nanoparticles (NPs)) by the 38 addition of aggregation‐prone peptides (APPs) in the generally recognized as safe (GRAS) 39 bacterial Lactococcus lactis expression system. The L6K2, HALRU and CYOB peptides were 2 40 selected to assess their intrinsic aggregation capability to nucleate protein aggregation. 41 These APPs enhanced the tendency to aggregate of the resulting protein at the expense of 42 the total protein yield. However, fine physico‐chemical characterization of the resulting 43 intracellular protein NPs, the protein released from these protein NPs, and the protein 44 purified from the soluble cell fraction indicated that the compactability of protein 45 conformations is directly related to the biological activity of variants of IFN‐, which is 46 used here as a model protein with therapeutic potential. APPs enhance aggregation 47 tendency of fused rBoIFN‐ while increasing compactability of protein species. Biological 48 activity rBoIFN‐ is favored in more compacted conformations. Naturally‐occurring protein 49 aggregates can be produced in GRAS microorganisms as protein depots of releasable 50 active protein. The addition of APPs to enhance aggregation tendency has a positive 51 impact in overall compactability and functionality of resulting protein conformers. 52 53 Keywords 54 Interferon‐gamma, protein nanoparticles, protein aggregation, Lactococcus lactis, GRAS, 55 conformational compactability 56 57 58 1. Introduction 3 59 The efficient production and purification of recombinant proteins in a wide range of 60 expression hosts has driven the launch of a large number of biopharmaceutical products. 61 One of the most‐studied expression systems is Escherichia coli (E. coli), which is also one of 62 the most‐used gene expression systems for biopharmaceutical products [1,2]. However, 63 pro‐inflammatory contaminant lipopolysaccharide (LPS) components of the outer leaflet 64 of the outer membrane of E. coli need to be removed from the purified protein to ensure 65 the safety of the final product, increasing the final cost [3,4]. Prokaryotic endotoxin‐free 66 expression systems are being explored to overcome this limitation, including E. coli LPS 67 mutant strains [5,6] and generally recognized as safe (GRAS) microorganisms, such as 68 Lactococcus lactis (L. lactis) [7,8]. These strains are envisioned as sound alternatives that 69 avoid the safety concern of LPS contamination retaining the advantages of culturing 70 prokaryotic hosts [9,10]. 71 During recombinant gene expression, the great stress posed to the protein quality control 72 machinery leads, in most cases, to the accumulation of the recombinant protein in 73 aggregates that form intracellular nanoparticles (NPs), known as inclusion bodies (IBs) [11‐ 74 13]. In fact, the aggregation of a heterologous is one of the evaluated key parameters 75 when stablishing the production and purification processes [14‐16]. Intracellular protein 76 aggregates are dynamic and complex nanostructures with a variable content of 77 recombinant protein [17‐19]. The trapped protein was formerly thought to be biologically 78 inert due to, aberrant protein conformations or inactive partially folded species 79 incompatible with the presence of biological activity. Thus, the recombinant protein could 80 be often recovered, with low efficiency, from the insoluble cell fraction by in vitro 4 81 denaturing/refolding processes [20]. This scenario has been replicated in biotechnological 82 research, and the main goal of recombinant protein production is to minimize protein 83 aggregation (and, in consequence, maximize protein solubility) during the production 84 process. 85 However, in recent decades, the view of naturally occurring protein aggregates as inert 86 material has changed completely since the detection of biologically active protein 87 embedded in these aggregates [21‐23]. The classic view of protein aggregates as mere 88 inactive folding intermediates has been transformed into the idea of heterogeneous 89 porous multimeric structures stabilized by a scaffold of cross beta‐sheet structures that 90 contain conformers of the recombinant protein in which a spectrum of recombinant 91 protein species containing native‐like conformations are incorporated [12]. In fact, it has 92 been proven that biologically active protein species can be extracted from IBs by applying 93 mild solubilization protocols, indicating the biologically active nature of proteins forming 94 these aggregates [24]. In addition, further ground‐breaking studies have suggested 95 applications of IBs in nanomedicine. In these studies, IBs are envisioned as recombinant 96 protein depots capable of slow release of active recombinant protein to replace specific 97 biological activities in defective cell lines, to recover cell viability under stress conditions in 98 cell culture [25], or even to target specific cell types in tumors when subcutaneously 99 implanted in animal models [26,27]. Furthermore, IBs have been proposed as a novel 100 biomaterial for use in tissue engineering due to the stimulation of mechanical and physical 101 signals induced in surrounding cells, even in the absence of cell growth‐promoting protein 102 factors in their formulation [28,29]. Therefore, IBs are envisioned as non‐toxic, 5 103 biocompatible and mechanically stable materials from which biologically active molecules 104 of the recombinant protein can be released under mild solubilization and physiological 105 conditions. 106 In this scenario, interest in the possibility of controlling the aggregation of recombinant 107 proteins in these types of nanostructures is increasing, and several aggregation‐prone 108 peptides (APPs) have been identified for fusion with recombinant proteins to enhance the 109 aggregation process in the producing cell [30]. In this study, we selected interferon‐ 110 gamma (IFN‐) as a model protein to study the effect of the addition of APPs in naturally 111 occurring protein aggregates due to interest in this activity in biomedicine and its 112 potential use in animal health. 113 IFN‐ is the sole type II IFN. This cytokine is produced and secreted by different cell types 114 and is involved in immunostimulatory processes through the activation of specific cellular 115 pathways at the transcriptional level [31]. IFN‐ secretion, by natural killer (NK) cells and 116 antigen‐presenting cells (APCs), enhances the innate immune response against detected 117 pathogens, while T‐lymphocytes are involved in the secretion of IFN‐ in the adaptive 118 immune response [32,33]. IFN‐ secretion is mainly regulated by other cytokines produced 119 by APCs, including IL‐12 and IL‐18. The activity of IFN‐ depends on its interaction, as a 120 dimer, with the IFN‐ receptor (IFNGR). IFNGR is a tetrameric complex formed by two 121 ligand‐binding proteins (IFNGR1) and two signal‐transducing proteins (IFNGR2). 122 IFN‐ is one of the biopharmaceuticals approved by the FDA under the trade name 123 ACTIMMUNE® (Horizon Pharma, Inc., USA). An Iranian biosimilar (‐IMMUNEX, Exir 6 124 Pharmaceutical Company, Iran) is distributed in Asia. IFN‐ is used to prevent infectious 125 diseases in patients suffering from chronic granulomatous disease and is also indicated to 126 slow osteopetrosis [34]. In recent years, IFN‐ has been investigated in approximately 80 127 clinical trials for a number of indications, mainly related to immune system disorders, 128 infectious diseases and cancer (https://www.clinicaltrials.gov). Recombinant IFN‐ is 129 administered either as a unique drug in these clinical trials [35‐41] or in combination with 130 other products [42‐44]. Therefore, the central key role of IFN‐ in immunostimulatory and 131 immunomodulatory effects will lead to an increase in the use of this cytokine in the 132 coming years in human health. The immunostimulatory properties of this cytokine have 133 also been evaluated as alternative therapeutics in animal health [45]. In fact, the 134 administration of IFN‐ by intramammary infusion in productive dairy cows for mastitis 135 treatment has been proposed as a putative strategy to reduce the spread of antibiotic 136 resistance among zoonotic bacteria [46]. Furthermore, the wide use of antibiotics in the 137 prevention of animal diseases and growth promotion appears to be a source of resistant 138 bacteria, and WHO and the United Nations have deployed global action against this threat 139 to health security. Therefore, the rational use of antibiotics in addition to the 140 development of immunostimulatory alternatives in the treatment and prevention of 141 animal diseases may play a role in the control of antibiotic resistance. 142 Interestingly, recombinant products for animal health must not only be stable and 143 safely and effectively delivered on a large scale and under standard conditions but also 144 present obvious advantages over existing products to prove its commercial viability [47]. 145 In this context, approved recombinant human IFN‐ can be obtained from the E. coli 7 146 expression system, but novel protein formulations need to be developed. In fact, in most 147 reported studies of the expression and purification of IFN‐, the recombinant protein is 148 recovered from the purified IBs through extensive denaturation‐refolding processes [48‐ 149 50]. 150 Given the importance of expanding the catalogue of prokaryotic expression strains to 151 improve biophamaceutical safety, to explore the possibility of formulating recombinant 152 proteins in alternative cost‐effective formats and developing novel treatments to reduce 153 the reliance on antibiotics, in this work, we produced recombinant bovine IFN‐ (rBoIFN‐) 154 in GRAS lactic acid bacteria (L. lactis) in the form of protein NPs. We analyzed the ability of 155 APPs fused to rBoIFN‐ to enhance the aggregation propensity of the recombinant 156 cytokine. Furthermore, we assessed the link between the biological activities contained in 157 protein NPs of IFN‐ variants and their physico‐chemical characteristics. We determined 158 that the activity of the IBs is related to the specific biological activity of the recombinant 159 protein they contain, whereas the proportion of released protein is not the main factor. 160 The data presented here illustrate the great potential of endotoxin‐free protein NPs as 161 active biomaterials to formulate, at the nanoscale level, releasable proteins of biomedical 162 interest. 163 2. Materials and methods 164 2.1. Bacterial strains and plasmids 165 E. coli MC4100 (StrepR) [51] was used for cloning genes for protein production in L. 166 lactis. E. coli DH5 was used for cloning genes for protein production in E. coli. L. lactis 8 167 cremoris NZ9000 (Boca Scientific), and ClearColi® BL21(DE3) (Lucigen) were used in 168 expression experiments for each expression system. For L. lactis expression vectors, IFN‐ 169 of bovine (Bos taurus) origin was cloned at the NcoI/XbaI restriction sites of the CmR 170 pNZ8148 plasmid (MoBiTech). The digestion products were ligated into the expression 171 plasmid pNZ8148, and ligation product was used to transform L. lactis NZ9000 competent 172 cells by electroporation [52]. Electroporation was performed using a Gene Pulser from 173 Bio‐Rad with settings of 2500 V, 200 Ω and 25 μF in a pre‐cooled 2‐cm electroporation 174 cuvette. The electroporated cells were then supplemented with 900 μl of M17 broth with 175 0.5 % glucose and incubated for 2 h at 30 °C. The electroporation mix was centrifuged for 176 10 min at 10,000 x g at 4 °C, and the pellet was resuspended in 100–200 μl of M17 media 177 and plated. In addition, fusions of rBpIFN‐ with APPs were constructed (rBOIFN‐_L6K2, 178 rBoIFN‐_HALRU and rBoIFN‐_CYOB). All genes were C‐terminally fused to a His‐tag for 179 detection and quantification by western blot analysis. A Lys residue was included at the N‐ 180 terminus of the tag for putative elimination of the tag by exopeptidases. Gene sequences 181 were codon optimized for the L. lactis expression host as indicated (Geneart). For the E. 182 coli expression vector pETDuet (Novagen), the L. lactis codon‐optimized IFN‐ gene of 183 bovine (Bos taurus) origin was cloned at the NcoI/HindIII restriction sites of the pETDuet 184 plasmid. The expression vector was transformed into the ClearColi® BL21(DE3) strain. 185 Electroporation was performed using a Gene Pulser from Bio‐Rad with settings at 2400 V, 186 750 Ω and 25 μF in a pre‐cooled 2‐cm electroporation cuvette. The electroporated cells 187 were then supplemented with 900 μl of lysogeny broth (LB) and incubated for 1 h at 37 °C. 188 The cells were then plated on LB‐agar plates containing ampicillin (100 μg/ml) and 9 189 incubated at 37 °C overnight. For all clones, in the sequence design we added an NcoI 190 restriction site at the 5′ end followed by the nucleotides CA to restore the reading frame. 191 This cloning strategy adds an Ala to the N‐terminus of the protein. Therefore, the 192 recombinant proteins were produced as the mature form of the IFN‐γ (from Gln24 to 193 Thr166 NP_776511.1) with an additional Ala at the N‐terminus to restore the frameshift 194 introduced by the NcoI restriction site (Fig. 1a). 195 2.2. Selection of aggregation‐prone peptides (APPs) 196 APPs were selected by scanning the Disprot v6.02 database [53] with AGGRESCAN 197 software [54]. We selected two unstructured regions from two different proteins that 198 displayed a higher HSA. This selection was based on the assumption that APPs in solvent‐ 199 exposed regions were the best candidates for the purposes of this study. Additionally, 200 L6K2 was selected based on previous experimental results [55] after analysis with 201 AGGRESCAN showed that this peptide had a high NHSA and high a4vAHS. 202 203 2.3. Production and purification of rBoIFN‐ protein from the soluble cell fraction 204 Cultures of ClearColi® BL21 (DE3) cells transformed with the plasmid containing rBoIFN‐ 205  gene (pETDuet‐rBoIFN‐) were incubated in a shake flask at 37 °C and 250 rpm in LB 206 medium supplemented with 100 μg/ml ampicillin. When the cultures reached an 207 OD 550 of approximately 0.5 – 0.7, protein expression was induced by adding 1 mM IPTG 208 (isopropyl‐β‐D‐thiogalactopyranoside). The cultures were then incubated at 20 °C and 209 250 rpm overnight (for protein production). During the purification process, cells were 10 210 collected by centrifugation (15 min, 6,000 x g, 4 °C), and proteins were released by 211 sonication. Briefly, cells were lysed by sonication with 5 rounds of 3‐min cycles and pulses 212 of 0.5 s at 15 % amplitude (Lab Sonic ultrasonicator). L. lactis cremoris cells transformed 213 with plasmids containing the rBoIFN‐ gene (pNZ8148‐rBoIFN‐, pNZ8148‐rBoIFN‐ 214 L6K2pNZ8148‐rBoIFN‐HALRU and pNZ8148‐rBoIFN‐_CYOB) were incubated in a 215 shake flask at 30 °C without shaking in M17 broth + 0.5 % glucose supplemented with 5 216 µg/ml chloramphenicol and 2.5 µg/ml erythromycin. When the cultures reached an 217 OD 550 of approximately 0.4‐0.6, protein expression was induced by adding 12.5 ng/ml 218 nisin. Then, the cultures were incubated at 30 °C without shaking for the indicated time. 219 The cells were collected by centrifugation (15 min, 6,000 x g, 4 °C), and proteins were 220 released by a French press (3 rounds at 15000 PSI/machine pressure). 221 From this point, the same protocol was followed for both proteins. The soluble and 222 insoluble cell fractions were separated by centrifugation (40 min, 15,000 x g, 4 °C), and the 223 soluble cell fraction was filtered using a pore diameter of 0.2 μm. The recombinant protein 224 in the soluble cell fraction was purified by immobilized metal affinity chromatography 225 (IMAC) using a HiTrap Chelating HP 1‐ml column (GE Healthcare) with an ÄKTA purifier 226 FPLC system (GE Healthcare). The eluted proteins were then dialyzed against phosphate‐ 227 buffered saline (PBS) buffer. The control protein rBoIFN‐_Std (produced in E. coli) was 228 obtained from R&D Systems (2300‐BG‐025, R&D Systems) and corresponded to a mixture 229 of bovine IFN‐ Gln24‐Thr166 and Gln24‐Arg162, both with an N‐terminal. 230 11 231 2.4. Production and purification of rBoIFN‐ protein nanoparticles. 232 L. lactis cells transformed with expression plasmids derived from pNZ8148 were grown 233 in M17 medium enriched with 0.5 % glucose at 30 °C without shaking. E. coli was grown in 234 LB rich medium at 37 °C and 250 rpm. NP production was induced by adding 12.5 ng/ml 235 nisin (Sigma‐Aldrich) to L. lactis or 1 mM IPTG to E. coli cultures. After induction, the 236 cultures were grown for 5 h. Antibiotics were used for plasmid maintenance at the 237 following concentrations: chloramphenicol (5 μg/ml) and erythromycin (2.5 μg/ml) for L. 238 lactis and ampicillin (100 μg/ml) and streptomycin (30 μg/ml) for ClearColi. 239 Once produced, the protein NPs were purified using the purification protocol described 240 previously [9], including, at the beginning of the process, a mechanical disruption step by 241 French press. The protocol was performed under sterile conditions and all incubations 242 were carried out under agitation. The purified protein NPs were diluted 1:10 in PBS and 243 resuspended. 244 245 2.5. Quantitative protein analysis 246 The amounts of recombinant proteins produced by the expressing cells or present in 247 NPs were quantified by denaturing SDS‐PAGE as described previously [56]. Bands were 248 identified using a commercial polyclonal serum against the histidine tag (#A00186‐100 249 Genscript) and an anti‐mouse secondary antibody (#170‐6516, Bio‐Rad). The recombinant 250 protein yield was estimated by comparison with a standard curve of known amounts of a 12 251 purified GFP‐H6 protein quantified by the Bradford assay. Quantification was performed 252 with Quantity One software (Bio‐Rad). 253 254 2.6. Ultrastructural characterization 255 To characterize the morphometry (size and shape) of the NPs, microdrops of protein 256 aggregate suspensions were deposited for 2 min on silicon wafers (Ted Pella Inc.), air‐ 257 dried and observed in a nearly native state under a field emission scanning electron 258 microscope (FESEM) Zeiss Merlin (Zeiss) operating at 1 kV. Micrographs of the NPs were 259 acquired with a high‐resolution in‐lens secondary electron (SE) detector. Images were 260 taken at magnifications ranging from 20,000x to 80,000x. 261 262 2.7. Z potential analysis 263 Z potential (ZP) characterization of each kind of protein NPs was carried out using DLS 264 equipment (Malvern Nanosizer). To prevent the electrodes from burning, the samples 265 were prepared in deionized (MilliQ) water, a low ionic strength medium. Each sample was 266 analyzed in triplicate. 267 268 2.8. Determination of rBoIFN‐γ biological activity in bovine cells 269 The different rBoIFN‐ formulations described here were analyzed by a modified 270 kynurenine bioassay [57]. Bovine fibroblast‐like cells (EBTr cells), (87090202 Sigma‐ 271 Aldrich) were cultured in Dulbecco’s modified Eagle’s medium (Gibco) with 10 % fetal 272 bovine serum (FBS). Depending on cell growth, cultures were split every 2 days using 13 273 ratios of 1:2 to 1:3. Before an experiment, a lower ratio was selected to obtain the 274 maximum cell density. For activity analysis, the cells were seeded in 96‐well flat‐bottom 275 microtiter plates (5000 cells per well) in Dulbecco’s modified Eagle’s medium (Gibco) 276 supplemented with 100 µg/ml L‐ Trp. Serial dilutions of both, the soluble and NP forms of 277 rBoIFNy at quantities ranging from 6 nmol/L to 0.024 nmol/L were incubated with cells for 278 96 h at 37 °C. Aliquots of 160 µl of the supernatants were then mixed with 10 µl of 30 % 279 trichloroacetic acid (T6399 Sigma‐Aldrich) and incubated at 50 °C for 30 min. After a 280 centrifugation step (10 min, 600 x g), aliquots of 100 µl of the supernatants were mixed 281 with an equal volume of 4 % w/v Ehrlich´s reagent 4‐(dimethylamino) benzaldehyde 282 (156477, Sigma‐Aldrich) in glacial acetic acid (Fisher Chemical A/0360/PB15). After 10 min, 283 the absorbance at 490 nm was measured in a conventional luminometer and VICTOR3V 284 1420 multilabel reader (PerkinElmer). 285 The absorbance vs IFN‐ quantity curves were adjusted to Eq. 1. Abs490 is the 286 absorbance at 490 nm, which represents an indirect measurement of IFN‐ binding to the 287 receptor, Absmax is the maximal binding of IFN‐ to the receptor, and KD is the equilibrium 288 dissociation constant. A low value of KD indicates high IFN‐ affinity to the receptor. 289 Abs𝑚𝑎𝑥 x IFN 290 Abs490 (1) IFN 𝐊𝐝 291 292 2.9. Assay of protein solubilization from protein nanoparticles 293 The rBoIFN‐ protein NPs (rBoIFN‐_L, rBoIFN‐_L6K2, rBoIFN‐_CYOB and rBoIFN‐ 294 _HALRU) were solubilized in different volumes of PBS depending on their initial protein 14 295 amounts. In all cases, the concentration was adjusted to 20 µg/ml. After manual agitation, 296 every sample was incubated at 37 °C for 96 h to reproduce the conditions used during the 297 biological activity determination. The soluble and insoluble fractions were then isolated by 298 centrifugation (15 min, 15,000 x g). The protein amounts in the soluble fractions were 299 quantified, and the concentrations were adjusted. The biological activity of every sample 300 was determined at a single concentration (3 ng/µl) as described previously. The protein 301 released from the protein NPs was resuspended in Laemmli buffer (1x) (from Laemmli 302 buffer (4x): 1.28 g of Tris base, 8 ml of glycerol, 1.6 g of sodium dodecylsulfate (SDS), 4 ml 303 of ‐mercaptoethanol and 9.6 g of urea in a final volume of 100 ml), boiled at 98 °C for 304 45 min, and loaded onto SDS–polyacrylamide gel electrophoresis (10 % acrylamide) 305 denaturing gels. Protein bands were detected by Western blot using a commercial 6xHis 306 monoclonal antibody (631212, Clontech) and a goat anti‐mouse IgG (H + L)‐HRP conjugate 307 (1706516, Bio‐Rad) as the secondary antibody. Images of the membranes were obtained 308 using the ChemiDoc Imaging System (Bio‐Rad), and bands were quantified with Image Lab 309 Software (Bio‐Rad) using known concentrations of commercial rBoIFN‐ (20, 15, 10, 5, 3 310 and 1 ng; 2300‐BG‐025, R&D Systems). 311 312 2.10. Interferon size determination 313 The volume size distribution of interferon  was determined by Dynamic Light 314 Scattering (DLS). A 60‐µl aliquot (stored at ‐80 °C) was thawed, and the volume size 315 distribution of each protein format was immediately determined at 633 nm (Zetasizer 316 Nano ZS, Malvern Instruments Ltd.). 15 317 318 2.11. Analysis of protein conformation by intrinsic tryptophan fluorescence 319 Fluorescence spectra were recorded on a Cary Eclipse spectrofluorometer (Agilent 320 Technologies). A quartz cell with a 10‐mm path length and a thermostated holder was 321 used. The excitation and emission slits were set at 5 nm. The excitation wavelength (ex) 322 was set at 295 nm. Emission spectra were acquired within a range from 310 to 550 nm. 323 The protein concentration was 0.3 mg/ml in PBS. To evaluate conformational differences 324 between the proteins, we applied the CSM. CSM is the weighted average of the 325 fluorescence spectrum peak. 326 The CSM was calculated for each of the fluorescence emission spectra [58] according to 327 Eq.2, where Ii is the fluorescence intensity measured at wavelength i. 328 329    .I i i (2) I i 330 331 2.12. Statistical analysis 332 Prior to the use of parametric tests, normality and homogeneity of variances were 333 tested using the Shapiro‐Wilk test for all quantitative data or the Levene test for raw or 334 transformed data. First, divergences between groups were tested with one‐way ANOVA, 335 and pairwise comparisons were made with Student’s t tests. The results were expressed 336 as the arithmetic mean for non‐transformed data ± the standard error of the mean 337 (x̅ ± SEM), except otherwise stated. 16 338 The least squares method was applied to fit functions through a regression analysis to 339 determine the Kd values according to Eq. 1. Significance was accepted at p < 0.05, and 340 Bonferroni correction was applied for sequential comparisons. All statistical analyses were 341 performed with SPSS v. 18 for Windows. 342 343 3. Results and discussion 344 3.1. Production of rBoIFN‐ in L. lactis 345 The IFN‐ gene of mammalian species encodes a preproprotein of 155‐166 amino acids, 346 including a signal peptide of 22‐23 amino acids and a propeptide in the last few residues, 347 rendering a mature protein of 15.6 to 17 kDa. Heterogeneity of the C‐terminus has been 348 described, giving rise to variants of human IFN‐ending at residues 150, 160 or 161 [59]. 349 Human IFN‐ is usually produced in the E. coli expression system and is purified from IBs 350 by using denaturing/refolding methodologies [48‐50]. The same strategy has been used 351 for mouse IFN‐ [60]. In other approaches, recombinant proteins of bovine and ovine 352 origin are obtained from the soluble cell fraction of E. coli and Corynebacterium 353 glutamicum [61,62]. In the present work, the mature form of bovine IFN‐ (rBoIFN‐ 354 protein (UniProtKB P07353, residues 24 to 166) was used as a model protein and it was 355 produced in L. lactis with the aim of opening up new opportunities for novel protein 356 production platforms. Specifically, different APPs were fused to rboIFN‐to evaluate the 357 ability of the peptides to increase protein aggregation and to analyze the biological activity 358 retained in the naturally occurring protein aggregates. 17 359 To improve gene expression, the DNA sequence of the recombinant gene was codon 360 optimized for the L. lactis expression system. Three peptides, CYOB, HALRU and L6K2, 361 were selected based on their predicted aggregation propensity (Table 1 and Fig. 1a). 362 AGGRESCAN was used to identify aggregation‐prone segments in proteins deposited in the 363 Disprot protein database version v6.02 [53]. CYOB was selected as the peptide displaying 364 the highest hot spot area (HSA). HALRU showed a high normalized hot spot area (NHSA) 365 and average aggregation‐propensity hot spot (a4vAHS) while maintaining a significantly 366 high HSA value relative to the other identified peptides. Finally, L6K2 was previously 367 identified as a surfactant‐like peptide with the ability to enhance the aggregation 368 propensity of several proteins [55]. In the analysis, this peptide exhibited high NHSA and 369 a4vAHS values despite having shorter sequence. A linker with a predicted random coil 370 conformation was positioned between the IFN‐ and APP as previously described [55]. 371 In L. lactis, most of the protein was detected in the soluble cell fraction in the absence 372 of any of the APP (Fig. 1b). This observation is in agreement with previous results for the 373 expression of the natural DNA sequence of the bovine IFN‐ gene in E. coli in which His‐ 374 tagged rBoIFN‐ was purified by affinity chromatography from the soluble cell fraction 375 [62]. The presence of the APPs in the recombinant protein caused a noticeable shift of the 376 final products toward the insoluble cell fraction, as expected (Fig. 1b, bottom). The purity 377 of the protein aggregates ranged between 50‐60 % in all constructs (data not shown). The 378 APP resulting in the highest aggregation tendency was the L6K2 peptide. In addition, the 379 presence of an APP tag also had a negative effect on the total recombinant protein 380 produced in the cell (Fig, 1a, top). This negative effect was maximal at 3 h, when protein 18 381 levels of 13.82 ± 2.01 g/ml, 11.38 ± 0.36 g/ml, and 10.36 ± 0.45 g/ml were observed 382 for the IFN‐ variants fused with the L6K2, HALRU and CYOB peptides, respectively, 383 compared with 211.99 ± 51.46 g/ml for wild type IFN‐. Therefore, the best APP in terms 384 of aggregation propensity and protein yield in the insoluble cell fraction, corresponded to 385 the IFN‐ L6K2 formulation. Surprisingly, the performance of this surfactant‐like peptide 386 exceeded the predicted aggregation‐prone capabilities of CYOB and HALRU peptides 387 (Table 1). 388 389 FIG 1 (a) IFN‐constructs produced in E. coli and L. lactis. Residues of the IFN‐ protein are 390 depicted in the corresponding light grey rectangles. APPs are indicated as wavy pattern 391 boxes. The amino acid sequences of the APPs and the linker (black rectangles) between 19 392 bovine IFN‐ and the APPs are shown below the drawings. The H6‐tag fused to the C‐ 393 terminus of all constructs is shown in dark grey. (b) Quantification of the production of 394 IFN‐ in IB‐like nanoparticles in L. lactis (top) and solubility (bottom) of IFN‐ in L. lactis. 395 Significant results are shown as * p ≤ 0.05 and ** p ≤ 0.005. 396 397 398 3.2. Nanoarchitectonic characterization of protein nanoparticles 399 The morphometry of purified protein NPs of the rBoIFN‐ variants was examined by 400 field emission scanning electron microscopy (FESEM; Fig. 2a). The images revealed the 401 presence of multimeric complexes comprising discrete NPs in addition to isolated protein 402 NPs (inset Fig. 2a). First, the NPs were similar to rBoIFN‐ protein NPs obtained previously 403 in this expression system [9]. Z potential (ZP) measurements showed that all of the NPs 404 presented negatively charged surfaces with negative values ranging from ‐38 to ‐28 mV 405 (Fig. 2b), indicating the stability of the NP suspension. The higher values of ZP obtained for 406 the IFN‐ variants provide information about particle stability, as NPs displaying higher ZP 407 values (higher than +30 mV or lower than ‐30 mV) exhibit increased stability due to 408 greater electrostatic repulsion between particles [63]. 20 409 410 FIG 2 (a) Ultrastructural characterization by FESEM of protein aggregates and purified 411 protein nanoparticles of rBoIFN_, rBoIFN‐_CYOB, rBoIFN‐_HARLU and rBoIFN‐_L6K2. 412 Scale bars correspond to 200 nm. (b) Z potential of purified protein nanoparticles. 413 414 3.3. Biological activity of soluble IFN‐ and nanoparticles of IFN‐ 415 The activity of IFN‐ is usually determined by an antiviral assay [64]. This assay must be 416 performed in facilities with an appropriate biosafety level, and viral stocks have to be 417 maintained over time. Therefore, alternative assays have been developed to simplify the 418 procedure. One approach to evaluate IFN‐ activity mediated by IFN‐‐receptor binding is 419 the detection of kynurenine. The antiproliferative activity of IFN‐ in this assay is related to 420 the induction of the expression of the indoleamine 2,3‐dioxygenase 1 (IDO1) gene, which 421 is the first and rate‐limiting enzyme in tryptophan catabolism. IDO1 catalyzes oxidative 422 cleavage of tryptophan to N‐formylkynurenine. Following a hydrolysis step, the latter is 423 transformed into kynurenine by Ehrlich’s reagent, giving a yellow‐colored compound 21 424 absorbing at 490 nm [65]. The activity of IFN‐is highly species‐specific and the detection 425 of tryptophan hydrolysis by IDO1 has been then stablished for different cell lines, including 426 bovine cells [57]. Thus, a specific assay for the bovine IFN‐ was developed and validated 427 in this study. For the validation process of the developed procedure, the activity of three 428 soluble rBoIFN‐ proteins was tested (Fig. 3a). rBoIFN‐_Std exhibited the lowest 429 dissociation constant (KD) among the proteins purified from the soluble cell fraction (Fig. 430 3a). The difference in this parameter with in‐house IFN‐ produced in Clearcoli (rBoIFN‐ 431 _E. coli) may be related to the absence of C‐terminal variants in this sample or the effect 432 of the fused His‐tag to the C‐term although may also be attributable to other variables 433 [66]. The protein obtained from the L. lactis expression system displayed less activity, 434 which may be to differences in the production process among prokaryotic expression 435 systems [67,68]. Once the activity assay was validated, the biological activity contained in 436 the IFN‐ protein NPs produced in L. lactis was determined. For that, bovine cells were 437 incubated with increasing amounts of rBoIFN‐NPs, that is, protein NPs purified from the 438 insoluble cell fraction (Fig. 3b). The results showed that all cells were able to elicit 439 responses to the presence of the protein NPs, and the IFN‐_L6K2 formulation displayed 440 the highest initial rate and kynurenine production. The addition of HALRU and CYOB APP 441 to IFN‐ had a moderate effect on the cell response. As shown in Fig. 3a and Fig. 3b, the 442 experiments were performed at the same time with the same stock of cells and 443 conditions. We wondered why the sample corresponding to protein NPs of IFN‐_L6K2 444 had the highest activity and initial rate, even compared with commercial IFN‐. Consistent 445 with this observation, a previous analysis of the activity of recombinant ‐galactosidase 22 446 produced in E. coli in the form of protein NPs revealed higher specific activity than the 447 corresponding soluble version of the protein [21]. However, these protein NPs obtained 448 from E. coli have not been characterized in detail. The activity displayed by IBs produced in 449 E. coli has been attributed to the release of a spectrum of conformers of the recombinant 450 protein, which leaves a scaffold that is resistant to proteolysis and has an extensive cross‐ 451 beta‐pleated sheet conformation [69,70]. In fact, in the case of protein NPs of rBoIFN‐ 452 produced in L. lactis, 30‐40 % of the material is resistant to proteolysis, indicating that the 453 protein NPs obtained in this expression system follow the same principles as the E. coli 454 system [9]. Therefore, the activities displayed by the protein NPs are probably due to the 455 partial release of the IFN‐ that forms part of the macromolecular complex [27]. To better 456 evaluate the ability of the protein NPs to release protein, they were incubated in PBS for 457 96 h to emulate the protein release conditions established during the biological activity 458 assay of the protein NPs (see the experimental design used to obtain the different protein 459 samples in Fig. 3c). Release of 52.67 %, 5.30 %, 0.42 % and 0.46 % was observed for IFN‐, 460 IFN‐_L6K2, IFN‐_HALRU and IFN‐_CYOB NPs, respectively. Aiming to analyze the specific 461 activity of the proteins released from the protein NPs, an activity assay was performed 462 and the results were compared with soluble proteins obtained directly from the soluble 463 cell fraction (Fig. 3d). The results showed that the maximal specific activity corresponded 464 to the IFN‐_L6K2 protein released from NPs. In addition, the comparison of the specific 465 activity of the rBoIFN‐ protein produced in L. lactis and purified from the soluble cell 466 fraction with that of the corresponding protein released from NPs showed that the 23 467 released protein elicited better conformational performance (compare the second and 468 last bars in Fig. 3d). 469 470 FIG 3 Kynurenine levels measured by absorbance at 490 nm after treatment of EBTr cells 471 for 96 h with increasing amounts of rBoIFN‐ from different origins. (a) Soluble rBoIFN‐ 472 produced in the indicated expression system. (b) Protein nanoparticles of rBoIFN‐ 473 produced in L. lactis. The Kd values are indicated in the plot. (c) Schematic representation 474 of the protein samples used in the activity assays: soluble protein obtained from the 475 soluble cell fraction, protein NPs purified from the insoluble cell fraction, soluble protein 476 obtained from the protein NPs, and the NP core after a resolubilization procedure. (d) 477 Comparison of the activity between rBoIFN‐ protein obtained from solubilization of 478 protein NPs and purified rBoIFN‐ from the soluble cell fraction as indicated at 0.72 24 479 nmol/L. Different letters depict differences between proteins (p < 0.001) except rBoIFN‐ 480 from protein NPs and rBoIFN‐_E (p = 0.024). 481 In this context, the addition of APPs to the rBoIFN‐protein improved the aggregation 482 profile of the produced protein (Fig. 1b). However, the presence of this type of peptides 483 had a negative effect on the overall production of the protein and, in the case of HALRU 484 and CYOB, a huge impact on biological activity (Fig. 3). Therefore, AGGRESCAN software is 485 able to predict the propensity to aggregate of the resulting APP‐containing recombinant 486 IFN‐ and is a reliable tool for analyzing solubility performance in the design of 487 recombinant genes [54]. 488 489 3.4. Physico‐chemical characterization of soluble IFN‐ and nanoparticles of IFN‐ 490 The precise physico‐chemical analysis of recombinant proteins is important due to 491 safety concerns, and specific regulatory guidelines were redesigned for protein‐based 492 pharmaceuticals after the TGN1412 clinical trial [71,72]. Therefore, the physico‐chemical 493 characterization of recombinant proteins needs to be further developed, and technical 494 approaches to obtain reliable data on the quality of the final product are needed. 495 To further analyze the protein in different formats, DLS measurements were performed 496 (Fig. 4a‐4d). The commercial bovine IFN‐ exhibited a peak with a maximum at 7.6 nm, 497 was quite similar to the peak at 6.13 nm for the in‐house version of IFN‐ produced in L. 498 lactis. This configuration might correspond to the dimeric form of the cytokine. However, 25 499 the IFN‐ obtained in‐house in E. coli showed a tendency towards a larger size. Therefore, 500 the specific activity of the different rBoIFN‐formats is not directly linked to the dimeric 501 configuration, which is the functional conformation when binding to the cell receptor, and 502 some other variables might be involved. When analyzing the size of the purified NPs, a 503 peak above 1,000 nm was detected, which is above the upper sensitivity limit of the 504 equipment (Fig. 4b). The NPs were clustered in higher‐order complexes from monomeric 505 versions of 200 nm (Fig. 2a). All samples exhibited the same profile. After solubilization of 506 the protein embedded in the NPs, the size of the remaining material remained above 507 1,000 nm since the scaffold of the NPs retained the overall structure after the protein was 508 released (Fig. 4d). The released protein showed a narrow dispersion ranging from the 509 dimeric size of the protein identified in the samples obtained from the commercial IFN‐ 510 or the soluble version purified from L. lactis detected in the upper panel of Fig. 4a (Fig. 4c). 511 In addition, the polydispersity index (PI) of these samples was higher than that of the 512 soluble IFN‐ versions. The PI corresponds to an estimate of the width of the distribution, 513 and the higher values of PI are in accordance with the data showing the presence of a pool 514 of conformers in the folding of recombinant proteins when the proteins are produced in 515 the cell. By contrast, in the protein versions purified from the soluble cell fraction, the 516 downstream process selects only a narrow collection of conformers, indicating that the 517 protein obtained during solubilization assays from protein NPs is more representative of 518 the diversity in conformations of a single protein that are produced in the expression 519 system. 26 a b 25 25 20 rBoIFN‐_L. lactis (7.6±1.89; 0.55) 20 no TAG(1230; 0.3) rBoIFN‐_E. coli (19.6±9.9; 0.24) L6K2 (1158; 0.16) t rBoIFN‐_Std (6.13±1.86; 0.08) CYOB(1168; 0.05) 15 15 HALRU(1136; 0.08) %volume 10 10 5 5 0 0 c d 25 25 20 20 no TAG(7.7±7.22; 0.74) no TAG(1358; 0.264) L6K2 (10±3.24; 1) L6K2 (1632; 0.2) %volume CYOB(11.3±3.99; 0.66) 15 CYOB(1595; 0.45) 15 HALRU(7.9±5.97; 0.94) HALRU(1348; 0.18) 10 10 5 5 0 0 1 10 100 1000 10000 1 10 100 1000 10000 diameter (nm) diameter (nm) 520 521 FIG 4 Recombinant IFN‐ sizes in different supramolecular arrangements (purified soluble 522 IFN‐ and INF‐ IBs). (a) Soluble rBoIFN‐ from different origins: commercial rBoIFN‐_Std, 523 in‐house rBoIFN‐ from E. coli and L. lactis. (b) rBoIFN‐ IBs produced in L. lactis. (c) PBS 524 solubilized rBoIFN‐ from IBs after interferon release. (d) Scaffold of rBoIFN‐ IBs 525 incubated for 96 hours at 37 °C. The mean size and polydispersity index are indicated in 526 brackets. The average size data of the soluble proteins were analyzed by one‐way ANOVA 527 (t corresponds to P  0.07). 528 To further analyze the link between the physico‐chemical properties and the specific 529 activity of the proteins, the fluorescence emission of Trp was recorded. Each fluorescence 530 emission spectrum was transformed into a center of spectral mass (CSM) value. This 531 parameter is related to the relative exposure of the Trp to the protein environment. The 27 532 maximum red‐shift in the CSM of the Trp spectrum is compatible with large solvent 533 accessibility [73‐75]. By contrast, the blue shift in the CSM corresponds to a Trp hidden in 534 a more hydrophobic milieu [76]. The mature form of BoIFN‐ has a unique Trp. This Trp 535 residue is partially buried in the 3D structure of the protein (PDB 1D9C) [77] and is not 536 involved in either monomer or in cytokine‐receptor interactions, as shown in the 3D 537 structure of the human tetrameric complex of the cytokine dimer with the receptor (PDB 538 1FG9) [78]. A remarkable aspect of the intrinsic fluorescence analysis is that all of the 539 rBoIFN‐ variants within the NPs or after solubilization from the protein NPs exhibited 540 lower CSM values than the samples obtained from the soluble fraction (Table 2). These 541 results suggest that the protein forming part of the NPs and the protein solubilized from 542 the aggregates have a more compact conformation than the soluble version. The 543 differences between the soluble IFN‐ versions are due to differences in protein origin, 544 sequence and size. The most efficient IFN‐ soluble version corresponded to the 545 commercial IFN‐, which had the lowest CMS due to its highly compacted structure. The 546 proteins obtained from the soluble fraction of E. coli and L. lactis exhibited higher CMS 547 values than the commercial protein. These differences might be related to the distinct 548 sizes detected (Fig. 4a). The in‐house rBoIFN‐_E. coli was approximately three times 549 larger than the same NPs produced in L lactis, indicating that the Trp residue sensed a 550 more polar environment compared with the L lactis form (Table 2). Interestingly, for the 551 protein originating from the particulate form, a blue shift was observed compared with 552 the soluble versions, and the CSM increased as it was resolubilized (lines 2 and 4 of Table 553 2), at least in versions with a percentage of release above 5 %. Interestingly, the CSM 28 554 value of the PBS solubilized rBoIFN‐_L. lactis protein sample did not reach that of the 555 soluble counterpart (lines 3 and 4 of Table 2). When the APPs were incorporated in the 556 engineered protein constructs, the solubilized proteins showed a decrease in the CSM 557 values compared with the protein NPs samples (lines 5 and 7). This behavior suggests a 558 possible self‐arrangement of the tag within the protein that could replace water molecules 559 and increase the hydrophobicity of the Trp environment. The CYOB construct (line 6 of 560 Table 2) required a specific analysis as this tag contributes five additional Trp residues to 561 the whole protein structure. In this case, the solubilized protein spectrum exhibited a 562 modest red shift (higher CMS value) compared with the particulate form, indicating that 563 the solubilization process exposed some of the Trp residues to a hydrophilic environment. 564 The CSM values of the CYOB and HALRU protein NPs remained unaltered after 565 solubilization (Table 2, lines 6 and 7). These data are in accordance with the higher 566 stability of the particulate forms, which exhibited low levels of protein release. 567 The NP form of IFN‐ also favored the specific activity (insets, Fig. 3, L. lactis and non‐ 568 tagged rBoIFN). This phenomenon is not only due to more active conformation of the 569 protein (Fig. 4 and Table 2, line 3 vs line 4) [79] but also to the heterogeneous distribution 570 of the protein and the ability of the protein NPs to increase the effective concentration of 571 protein in the proximity of the receptor. Moreover, the formulation containing L6K2 was 572 the most efficient, even compared with the commercial protein. Solubilization clearly 573 conferred the most active and altered conformation of the protein without the tag. 574 Although a low percentage of protein was released from the NPs containing L6K2, at least 29 575 in the physiological buffer PBS, this released protein seems to be sufficient to surpass the 576 activity of the released protein without a tag (Fig.3). 577 Another interesting aspect is the effect of the size of the tag on the structure‐function 578 of the protein. The incorporation of a tag larger than 17 aa beyond the linker (Fig. 1a) 579 could generate steric problems preventing the interaction of tagged IFN‐ with the 580 receptor. The larger the tag, the more negative the effect on the activity is detected. As 581 shown in Fig. 1, L6K2 is only 8 aa, compared with 17 aa for HALRU and 38 aa for CYOB. The 582 short size of the L6K2 tag might reduce the difficulty of the interaction between L6K2‐IFN‐ 583  and the receptor compared with the longer IFN‐ tags. 584 The size of the APP is an important parameter, and as a general rule, the longer the 585 aggregation peptide, the higher the HSA. However, small peptides with a discrete HSA but 586 high normalized and average HSA have also been reported to enhance the aggregation of 587 the accompanying proteins, as in the case of L6K2 [55]. Surprisingly, the L6K2 peptide also 588 leads to the production of highly active protein in the protein NPs. The detailed physico‐ 589 chemical characterization of the protein samples revealed higher compactability of the 590 protein conformations in the protein samples with the highest biological activity. The 591 protein peaks of the soluble protein samples indicated that the purification strategy of a 592 protein selects for a narrower distribution of protein conformations compared with 593 protein samples recovered from naturally occurring protein NPs (compare the PDI of the 594 samples obtained from the soluble cell fraction with the protein samples obtained after 595 solubilization of protein NPs in Fig. 4). 30 596 In the recombinant protein production platform, the general consensus for improving 597 protein yield is to improve the solubility of the protein. Consequently, the major 598 established strategies for producing recombinant protein involve improving solubility by 599 including folding modulators or reducing the growth rate. However, solubility and 600 conformational quality are not necessarily coincident parameters [80]. The functionalities 601 of the protein obtained from the soluble cell fraction or the protein NPs of rBoIFN‐_L. 602 lactis in the present work were in accordance with these previous findings, as the protein 603 obtained from the soluble cell fraction was less active than that recovered from the 604 protein NPs. The compactabilities of the conformations of these proteins were in 605 agreement with their dissimilar biological activity. The results obtained in this study are in 606 accordance with previous analyses indicating that the compactability of protein 607 conformations is a key parameter related to stability and function [81,82]. 608 Consistent with these observations, the addition of the L6K2 APP to rBoIFN‐_L. lactis 609 improved the biological performance of the resulting conformational species retained in 610 the corresponding protein NPs. Interestingly, this APP showed higher capability as a 611 compactability factor compared with the other APPs tested with IFN‐. In fact, the 612 addition of the other two APPs to rBoIFN‐ (HALRU and CYOB) surprisingly had negative 613 effects on the functionality of the resulting proteins, despite levels of compactability 614 similar to that of the protein sample lacking the APPs. These sequences might have 615 negative effects by preventing correct binding of the dimer to the cell receptor. In 616 conclusion, the detailed physico‐chemical characterization of the purified recombinant 31 617 proteins needs to be further developed to select a more appropriate product to optimize 618 the quality of the recombinant product. 619 620 4. Conclusions 621 The demand for recombinant proteins in the pharmaceutical industry is steadily 622 increasing. Emerging novel protein formulations, including naturally occurring protein NPs, 623 might be an alternative to soluble variants for fine analysis at the biophysical level. Such 624 analyses are important to address safety about biological molecules. 625 This study analyzes the effect of aggregation‐prone peptides (APPs) on the improvement 626 of the production of naturally occurring protein nanoparticles (NPs) of interferon gamma 627 (IFN‐) in the generally recognized as safe (GRAS) Lactococcus lactis expression system. In 628 addition, the fine physico‐chemical characterization of the resulting proteins, either 629 obtained from the soluble or insoluble cell fractions, indicates that the selected 630 engineered proteins embedded in the protein NPs show higher compactability than their 631 soluble protein counterparts. Conformational compactability is directly related to the 632 biological performance of the recombinant IFN‐. This data prompt us to further explore 633 the relationship between compactability of protein conformations and protein quality in 634 naturally occurring protein NPs, envisioned as protein depots for the controlled release of 635 therapeutic proteins for biomedical applications. 636 32 637 Acknowledgments 638 This work was supported by grants from INIA, MINECO, Spain to N.F.M. and E.G.F. 639 (RTA2015‐00064‐C02‐01 and RTA2015‐00064‐C02‐02). The authors acknowledge financial 640 support granted to A.V. from AGAUR (2017 SGR‐229) and from the Centro de Investigación 641 Biomédica en Red (CIBER) de Bioingeniería, Biomateriales y Nanomedicina financed by the 642 Instituto de Salud Carlos III with assistance from the European Regional Development. We 643 are also indebted to the CERCA Programme (Generalitat de Catalunya) and European 644 Social Fund for supporting our research. J.V.C. received a pre‐doctoral fellowship from 645 UAB, O.C.G. received a PhD fellowship from MECD (FPU), and E.G.F. received a post‐ 646 doctoral fellowship from INIA (DOC‐INIA). AV has been distinguished with an ICREA 647 ACADEMIA Award. The authors also acknowledge ICTS ‘‘NANBIOSIS”, more specifically the 648 Protein Production Platform of CIBER in Bioengineering, Biomaterials & Nanomedicine 649 (CIBER‐ BBN)/IBB, at the UAB sePBioEs scientific‐technical service 650 (http://www.nanbiosis.es/unit/u1‐protein‐production‐platform‐ppp/) and the UAB 651 scientific‐technical services LLEB, SM and SCAC 652 (https://www.uab.cat/web/research/scientific‐technical‐services/all‐scientific‐technical‐ 653 services‐‐1345667278676.html). The authors would like to thank Milena Tileva for her 654 helpful advice on technical issues related to the experimental adjustment of the IFN‐ 655 activity bioassay. Special thanks to Sandra Párraga‐Ferrer for the design of Fig. 3c. E. 656 Garcia‐Fruitós and N. Ferrer‐Miralles designed and supervised the experiments. J.V. 657 Carratalà, O. Cano‐Garrido, J. Sánchez, C. Membrado, E. Pérez, O. Conchillo‐Solé and A. 658 Sánchez‐Chardi performed the experiments. J. V. Carratalà, O. Cano‐Garrido, J. Sánchez 33 659 and N. Ferrer‐Miralles analyzed the data. All authors edited the manuscript. N. Ferrer‐ 660 Miralles wrote the paper. 661 662 References 663 664 [1] M.N. Baeshen, A.M. Al‐Hejin, R.S. Bora, M.M. Ahmed, H.A. Ramadan, K.S. Saini, 665 N.A. Baeshen, E.M. Redwan, Production of Biopharmaceuticals in E. coli: Current 666 Scenario and Future Perspectives, J Microbiol Biotechnol 25 (2015) pp. 953‐962. 667 https://doi.org/10.4014/jmb.1412.12079. 668 [2] L. Sanchez‐Garcia, L. Martin, R. Mangues, N. Ferrer‐Miralles, E. Vazquez, A. 669 Villaverde, Recombinant pharmaceuticals from microbial cells: a 2015 update, 670 Microb Cell Fact 15 (2016) p.33. https://doi.org/10.1186/s12934‐016‐0437‐3. 671 [3] P.O. Magalhaes, A.M. Lopes, P.G. Mazzola, C. Rangel‐Yagui, T.C. Penna, A. Pessoa, 672 Jr., Methods of endotoxin removal from biological preparations: a review, J 673 Pharm. Pharm. Sci 10 (2007) pp. 388‐404. 674 [4] D. Petsch and F.B. Anspach, Endotoxin removal from protein solutions, J 675 Biotechnol 76 (2000) pp. 97‐119. 676 [5] F. Rueda, M.V. Cespedes, A. Sanchez‐Chardi, J. Seras‐Franzoso, M. Pesarrodona, N. 677 Ferrer‐Miralles, E. Vazquez, U. Rinas, U. Unzueta, U. Mamat, R. Mangues, E. Garcia‐ 678 Fruitos, A. Villaverde, Structural and functional features of self‐assembling protein 679 nanoparticles produced in endotoxin‐free Escherichia coli, Microb Cell Fact 15 680 (2016) p.59. 10.1186/s12934‐016‐0457‐z [doi];10.1186/s12934‐016‐0457‐z [pii]. 681 [6] U. Mamat, K. Wilke, D. Bramhill, A.B. Schromm, B. Lindner, T.A. Kohl, J.L. Corchero, 682 A. Villaverde, L. Schaffer, S.R. Head, C. Souvignier, T.C. Meredith, R.W. Woodard, 683 Detoxifying Escherichia coli for endotoxin‐free production of recombinant proteins, 684 Microb Cell Fact 14 (2015) p.57. 10.1186/s12934‐015‐0241‐5 685 [doi];10.1186/s12934‐015‐0241‐5 [pii]. 686 [7] A.A. Song, L.L.A. In, S.H.E. Lim, R.A. Rahim, A review on Lactococcus lactis: from 687 food to factory, Microb Cell Fact 16 (2017) p.55. 10.1186/s12934‐017‐0669‐x 688 [doi];10.1186/s12934‐017‐0669‐x [pii]. 34 689 [8] E. Garcia‐Fruitos, Lactic Acid Bacteria: a promising alternative for recombinant 690 protein production, Microb Cell Fact 11 (2012) p.157. 1475‐2859‐11‐157 691 [pii];10.1186/1475‐2859‐11‐157 [doi]. 692 [9] O. Cano‐Garrido, A. Sanchez‐Chardi, S. Pares, I. Giro, W.I. Tatkiewicz, N. Ferrer‐ 693 Miralles, I. Ratera, A. Natalello, R. Cubarsi, J. Veciana, A. Bach, A. Villaverde, A. Aris, 694 E. Garcia‐Fruitos, Functional protein‐based nanomaterial produced in 695 microorganisms recognized as safe: A new platform for biotechnology, Acta 696 Biomater. 43 (2016) pp. 230‐239. https://doi.org/10.1016/j.actbio.2016.07.038. 697 [10] N. Ferrer‐Miralles and A. Villaverde, Bacterial cell factories for recombinant 698 protein production; expanding the catalogue, Microb Cell Fact 12 (2013) p.113. 699 https://doi.org/10.1186/1475‐2859‐12‐113. 700 [11] R.R. Kopito, Aggresomes, inclusion bodies and protein aggregation, Trends Cell 701 Biol 10 (2000) pp. 524‐530. 702 [12] U. Rinas, E. Garcia‐Fruitos, J.L. Corchero, E. Vazquez, J. Seras‐Franzoso, A. 703 Villaverde, Bacterial Inclusion Bodies: Discovering Their Better Half, Trends 704 Biochem Sci 42 (2017) pp. 726‐737. https://doi.org/10.1016/j.tibs.2017.01.005. 705 [13] E. Rodriguez‐Carmona, R. Mendoza, E. Ruiz‐Canovas, N. Ferrer‐Miralles, I. Abasolo, 706 Schwartz S Jr, A. Villaverde, J.L. Corchero, A novel bio‐functional material based on 707 mammalian cell aggresomes, Appl. Microbiol Biotechnol 99 (2015) pp. 7079‐7088. 708 https://doi.org/10.1007/s00253‐015‐6684‐0. 709 [14] G.L. Rosano and E.A. Ceccarelli, Recombinant protein expression in Escherichia 710 coli: advances and challenges, Front Microbiol 5 (2014) p.172. 711 https://doi.org/10.3389/fmicb.2014.00172. 712 [15] H.P. Sorensen and K.K. Mortensen, Soluble expression of recombinant proteins in 713 the cytoplasm of Escherichia coli, Microb Cell Fact 4 (2005) p.1. 714 https://doi.org/10.1186/1475‐2859‐4‐1. 715 [16] S. Peternel and R. Komel, Active protein aggregates produced in Escherichia coli, 716 Int J Mol Sci 12 (2011) pp. 8275‐8287. https://doi.org/10.3390/ijms12118275. 717 [17] E. Garcia‐Fruitos, Inclusion bodies: a new concept, Microb Cell Fact 9 (2010) p.80. 718 https://doi.org/10.1186/1475‐2859‐9‐80. 719 [18] B. Jurgen, A. Breitenstein, V. Urlacher, K. Buttner, H. Lin, M. Hecker, T. Schweder, 720 P. Neubauer, Quality control of inclusion bodies in Escherichia coli, Microb Cell 721 Fact 9 (2010) p.41. https://doi.org/10.1186/1475‐2859‐9‐41. 35 722 [19] A. Singh, V. Upadhyay, A.K. Upadhyay, S.M. Singh, A.K. Panda, Protein recovery 723 from inclusion bodies of Escherichia coli using mild solubilization process, Microb 724 Cell Fact 14 (2015) p.41. https://doi.org/10.1186/s12934‐015‐0222‐8. 725 [20] H. Yamaguchi and M. Miyazaki, Refolding techniques for recovering biologically 726 active recombinant proteins from inclusion bodies, Biomolecules. 4 (2014) pp. 235‐ 727 251. https://doi.org/10.3390/biom4010235. 728 [21] E. Garcia‐Fruitos, N. Gonzalez‐Montalban, M. Morell, A. Vera, R.M. Ferraz, A. Aris, 729 S. Ventura, A. Villaverde, Aggregation as bacterial inclusion bodies does not imply 730 inactivation of enzymes and fluorescent proteins, Microb Cell Fact 4 (2005) p.27. 731 https://doi.org/10.1186/1475‐2859‐4‐27. 732 [22] S. Jevsevar, V. Gaberc‐Porekar, I. Fonda, B. Podobnik, J. Grdadolnik, V. Menart, 733 Production of nonclassical inclusion bodies from which correctly folded protein can 734 be extracted, Biotechnol Prog. 21 (2005) pp. 632‐639. 735 https://doi.org/10.1021/bp0497839. 736 [23] J. Nahalka and B. Nidetzky, Fusion to a pull‐down domain: a novel approach of 737 producing Trigonopsis variabilisD‐amino acid oxidase as insoluble enzyme 738 aggregates, Biotechnol Bioeng. 97 (2007) pp. 454‐461. 739 https://doi.org/10.1002/bit.21244. 740 [24] L. Gifre‐Renom, O. Cano‐Garrido, F. Fabregas, R. Roca‐Pinilla, J. Seras‐Franzoso, N. 741 Ferrer‐Miralles, A. Villaverde, A. Bach, M. Devant, A. Aris, E. Garcia‐Fruitos, A new 742 approach to obtain pure and active proteins from Lactococcus lactis protein 743 aggregates, Sci Rep. 8 (2018) p.13917. 10.1038/s41598‐018‐32213‐8 744 [doi];10.1038/s41598‐018‐32213‐8 [pii]. 745 [25] E. Vazquez, J.L. Corchero, J.F. Burgueno, J. Seras‐Franzoso, A. Kosoy, R. Bosser, R. 746 Mendoza, J.M. Martinez‐Lainez, U. Rinas, E. Fernandez, L. Ruiz‐Avila, E. Garcia‐ 747 Fruitos, A. Villaverde, Functional inclusion bodies produced in bacteria as naturally 748 occurring nanopills for advanced cell therapies, Adv Mater. 24 (2012) pp. 1742‐ 749 1747. 10.1002/adma.201104330 [doi]. 750 [26] M.V. Cespedes, Y. Fernandez, U. Unzueta, R. Mendoza, J. Seras‐Franzoso, A. 751 Sanchez‐Chardi, P. Alamo, V. Toledo‐Rubio, N. Ferrer‐Miralles, E. Vazquez, S. 752 Schwartz, I. Abasolo, J.L. Corchero, R. Mangues, A. Villaverde, Bacterial mimetics 753 of endocrine secretory granules as immobilized in vivo depots for functional 754 protein drugs, Sci Rep. 6 (2016) p.35765. https://doi.org/10.1038/srep35765. 755 [27] U. Unzueta, M.V. Cespedes, R. Sala, P. Alamo, A. Sanchez‐Chardi, M. Pesarrodona, 756 L. Sanchez‐Garcia, O. Cano‐Garrido, A. Villaverde, E. Vazquez, R. Mangues, J. Seras‐ 757 Franzoso, Release of targeted protein nanoparticles from functional bacterial 36 758 amyloids: A death star‐like approach, J Control Release 279 (2018) pp. 29‐39. 759 https://doi.org/10.1016/j.jconrel.2018.04.004. 760 [28] J. Seras‐Franzoso, C. Steurer, M. Roldan, M. Vendrell, C. Vidaurre‐Agut, A. 761 Tarruella, L. Saldana, N. Vilaboa, M. Parera, E. Elizondo, I. Ratera, N. Ventosa, J. 762 Veciana, A.J. Campillo‐Fernandez, E. Garcia‐Fruitos, E. Vazquez, A. Villaverde, 763 Functionalization of 3D scaffolds with protein‐releasing biomaterials for 764 intracellular delivery, J Control Release 171 (2013) pp. 63‐72. 765 https://doi.org/10.1016/j.jconrel.2013.06.034. 766 [29] J. Seras‐Franzoso, K. Peebo, E. Garcia‐Fruitos, E. Vazquez, U. Rinas, A. Villaverde, 767 Improving protein delivery of fibroblast growth factor‐2 from bacterial inclusion 768 bodies used as cell culture substrates, Acta Biomater. 10 (2014) pp. 1354‐1359. 769 https://doi.org/10.1016/j.actbio.2013.12.021. 770 [30] U. Krauss, V.D. Jager, M. Diener, M. Pohl, K.E. Jaeger, Catalytically‐active inclusion 771 bodies‐Carrier‐free protein immobilizates for application in biotechnology and 772 biomedicine, J Biotechnol 258 (2017) pp. 136‐147. 773 https://doi.org/10.1016/j.jbiotec.2017.04.033. 774 [31] K. Schroder, P.J. Hertzog, T. Ravasi, D.A. Hume, Interferon‐gamma: an overview of 775 signals, mechanisms and functions, J Leukoc. Biol 75 (2004) pp. 163‐189. 776 https://doi.org/10.1189/jlb.0603252. 777 [32] P.C. Le, P. Genin, M.G. Baines, J. Hiscott, Interferon activation and innate 778 immunity, Rev. Immunogenet. 2 (2000) pp. 374‐386. 779 [33] J.R. Schoenborn and C.B. Wilson, Regulation of interferon‐gamma during innate 780 and adaptive immune responses, Adv Immunol. 96 (2007) pp. 41‐101. 781 https://doi.org/10.1016/S0065‐2776(07)96002‐2. 782 [34] A. Razaghi, L. Owens, K. Heimann, Review of the recombinant human interferon 783 gamma as an immunotherapeutic: Impacts of production platforms and 784 glycosylation, J Biotechnol 240 (2016) pp. 48‐60. 785 https://doi.org/10.1016/j.jbiotec.2016.10.022. 786 [35] K. Jiang, S. Cao, J.Z. Cui, J.A. Matsubara, Immuno‐modulatory Effect of IFN‐gamma 787 in AMD and its Role as a Possible Target for Therapy, J Clin Exp. Ophthalmol. Suppl 788 2 (2013) pp. 0071‐0076. https://doi.org/10.4172/2155‐9570‐S2‐007. 789 [36] C. Kosmidis, K. Sapalidis, T. Koletsa, M. Kosmidou, C. Efthimiadis, G. Anthimidis, N. 790 Varsamis, N. Michalopoulos, C. Koulouris, S. Atmatzidis, L. Liavas, T.M. Strati, G. 791 Koimtzis, A. Tsakalidis, S. Mantalovas, K. Zarampouka, M. Florou, D.E. Giannakidis, 792 E. Georgakoudi, S. Baka, P. Zarogoulidis, Y.G. Man, I. Kesisoglou, Interferon‐ 793 gamma and Colorectal Cancer: an up‐to date, J Cancer 9 (2018) pp. 232‐238. 794 https://doi.org/10.7150/jca.22962. 37 795 [37] B.E. Marciano, R. Wesley, E.S. De Carlo, V.L. Anderson, L.A. Barnhart, D. Darnell, 796 H.L. Malech, J.I. Gallin, S.M. Holland, Long‐term interferon‐gamma therapy for 797 patients with chronic granulomatous disease, Clin Infect. Dis. 39 (2004) pp. 692‐ 798 699. https://doi.org/10.1086/422993. 799 [38] S.M. Rowe, D.S. Borowitz, J.L. Burns, J.P. Clancy, S.H. Donaldson, G. Retsch‐Bogart, 800 S.D. Sagel, B.W. Ramsey, Progress in cystic fibrosis and the CF Therapeutics 801 Development Network, Thorax 67 (2012) pp. 882‐890. 802 https://doi.org/10.1136/thoraxjnl‐2012‐202550. 803 [39] L. Seyer, N. Greeley, D. Foerster, C. Strawser, S. Gelbard, Y. Dong, K. Schadt, M.G. 804 Cotticelli, A. Brocht, J. Farmer, R.B. Wilson, D.R. Lynch, Open‐label pilot study of 805 interferon gamma‐1b in Friedreich ataxia, Acta Neurol. Scand. 132 (2015) pp. 7‐15. 806 https://doi.org/10.1111/ane.12337. 807 [40] G.C. Smaldone, Repurposing of gamma interferon via inhalation delivery, Adv Drug 808 Deliv. Rev. (2018). https://doi.org/10.1016/j.addr.2018.06.004. 809 [41] C.C. Wu, M.J. Econs, L.A. DiMeglio, K.L. Insogna, M.A. Levine, P.J. Orchard, W.P. 810 Miller, A. Petryk, E.T. Rush, D.M. Shoback, L.M. Ward, L.E. Polgreen, Diagnosis and 811 Management of Osteopetrosis: Consensus Guidelines From the Osteopetrosis 812 Working Group, J Clin Endocrinol. Metab 102 (2017) pp. 3111‐3123. 813 https://doi.org/10.1210/jc.2017‐01127. 814 [42] D.S. Green, A.T. Nunes, V. David‐Ocampo, I.B. Ekwede, N.D. Houston, S.L. Highfill, 815 H. Khuu, D.F. Stroncek, S.M. Steinberg, K.C. Zoon, C.M. Annunziata, A Phase 1 trial 816 of autologous monocytes stimulated ex vivo with Sylatron((R)) (Peginterferon alfa‐ 817 2b) and Actimmune((R)) (Interferon gamma‐1b) for intra‐peritoneal administration 818 in recurrent ovarian cancer, J Transl. Med 16 (2018) p.196. 819 https://doi.org/10.1186/s12967‐018‐1569‐5. 820 [43] D. Grimaldi, O. Pradier, R.S. Hotchkiss, J.L. Vincent, Nivolumab plus interferon‐ 821 gamma in the treatment of intractable mucormycosis, Lancet Infect. Dis. 17 (2017) 822 p.18. https://doi.org/10.1016/S1473‐3099(16)30541‐2. 823 [44] S.J. Harris, J. Brown, J. Lopez, T.A. Yap, Immuno‐oncology combinations: raising 824 the tail of the survival curve, Cancer Biol Med 13 (2016) pp. 171‐193. 825 https://doi.org/10.20892/j.issn.2095‐3941.2016.0015. 826 [45] V. Janardhana, M.E. Ford, M.P. Bruce, M.M. Broadway, T.E. O'Neil, A.J. Karpala, M. 827 Asif, G.F. Browning, K.A. Tivendale, A.H. Noormohammadi, J.W. Lowenthal, A.G. 828 Bean, IFN‐gamma enhances immune responses to E. coli infection in the chicken, J 829 Interferon Cytokine Res 27 (2007) pp. 937‐946. 830 https://doi.org/10.1089/jir.2007.0020. 38 831 [46] L.K. Fox, H.D. Liggit, T. Yilma, L.B. Corbeil, The effect of interferon‐gamma 832 intramammary administration on mammary phagocyte function, Zentralbl. 833 Veterinarmed. B 37 (1990) pp. 28‐30. 834 [47] L. Gifre, A. Aris, A. Bach, E. Garcia‐Fruitos, Trends in recombinant protein use in 835 animal production, Microb Cell Fact 16 (2017) p.40. 10.1186/s12934‐017‐0654‐4 836 [doi];10.1186/s12934‐017‐0654‐4 [pii]. 837 [48] R. Khalilzadeh, S.A. Shojaosadati, N. Maghsoudi, J. Mohammadian‐Mosaabadi, 838 M.R. Mohammadi, A. Bahrami, N. Maleksabet, M.A. Nassiri‐Khalilli, M. Ebrahimi, H. 839 Naderimanesh, Process development for production of recombinant human 840 interferon‐gamma expressed in Escherichia coli, J Ind Microbiol Biotechnol 31 841 (2004) pp. 63‐69. https://doi.org/10.1007/s10295‐004‐0117‐x. 842 [49] L. Perez, J. Vega, C. Chuay, A. Menendez, R. Ubieta, M. Montero, G. Padron, A. 843 Silva, C. Santizo, V. Besada, ., Production and characterization of human gamma 844 interferon from Escherichia coli, Appl. Microbiol Biotechnol 33 (1990) pp. 429‐434. 845 [50] S.T. Vaiphei, G. Pandey, K.J. Mukherjee, Kinetic studies of recombinant human 846 interferon‐gamma expression in continuous cultures of E. coli, J Ind Microbiol 847 Biotechnol 36 (2009) pp. 1453‐1458. https://doi.org/10.1007/s10295‐009‐0632‐x. 848 [51] J.G. Thomas and F. Baneyx, Roles of the Escherichia coli small heat shock proteins 849 IbpA and IbpB in thermal stress management: comparison with ClpA, ClpB, and 850 HtpG In vivo, J Bacteriol 180 (1998) pp. 5165‐5172. 851 [52] C. Labarre, C. Divies, J. Guzzo, Genetic organization of the mle locus and 852 identification of a mleR‐like gene from Leuconostoc oenos, Appl. Environ. 853 Microbiol 62 (1996) pp. 4493‐4498. 854 [53] D. Piovesan, F. Tabaro, I. Micetic, M. Necci, F. Quaglia, C.J. Oldfield, M.C. 855 Aspromonte, N.E. Davey, R. Davidovic, Z. Dosztanyi, A. Elofsson, A. Gasparini, A. 856 Hatos, A.V. Kajava, L. Kalmar, E. Leonardi, T. Lazar, S. Macedo‐Ribeiro, M. 857 Macossay‐Castillo, A. Meszaros, G. Minervini, N. Murvai, J. Pujols, D.B. Roche, E. 858 Salladini, E. Schad, A. Schramm, B. Szabo, A. Tantos, F. Tonello, K.D. Tsirigos, N. 859 Veljkovic, S. Ventura, W. Vranken, P. Warholm, V.N. Uversky, A.K. Dunker, S. 860 Longhi, P. Tompa, S.C. Tosatto, DisProt 7.0: a major update of the database of 861 disordered proteins, Nucleic Acids Res 45 (2017) p.D219‐D227. 862 https://doi.org/10.1093/nar/gkw1056. 863 [54] O. Conchillo‐Sole, N.S. de Groot, F.X. Aviles, J. Vendrell, X. Daura, S. Ventura, 864 AGGRESCAN: a server for the prediction and evaluation of "hot spots" of 865 aggregation in polypeptides, BMC. Bioinformatics. 8 (2007) p.65. 866 https://doi.org/10.1186/1471‐2105‐8‐65. 39 867 [55] B. Zhou, L. Xing, W. Wu, X.E. Zhang, Z. Lin, Small surfactant‐like peptides can drive 868 soluble proteins into active aggregates, Microb Cell Fact 11 (2012) p.10. 869 https://doi.org/10.1186/1475‐2859‐11‐10. 870 [56] O. Cano‐Garrido, F.L. Rueda, L. Sanchez‐Garcia, L. Ruiz‐Avila, R. Bosser, A. 871 Villaverde, E. Garcia‐Fruitos, Expanding the recombinant protein quality in 872 Lactococcus lactis, Microb Cell Fact 13 (2014) p.167. 873 https://doi.org/10.1186/s12934‐014‐0167‐3. 874 [57] K. Spekker, M. Czesla, V. Ince, K. Heseler, S.K. Schmidt, G. Schares, W. Daubener, 875 Indoleamine 2,3‐dioxygenase is involved in defense against Neospora caninum in 876 human and bovine cells, Infect. Immun. 77 (2009) pp. 4496‐4501. 877 https://doi.org/10.1128/IAI.00310‐09. 878 [58] J.R. Lakowicz, J. Kusba, W. Wiczk, I. Gryczynski, H. Szmacinski, M.L. Johnson, 879 Resolution of the conformational distribution and dynamics of a flexible molecule 880 using frequency‐domain fluorometry, Biophys. Chem 39 (1991) pp. 79‐84. 881 [59] Y.C. Pan, A.S. Stern, P.C. Familletti, F.R. Khan, R. Chizzonite, Structural 882 characterization of human interferon gamma. Heterogeneity of the carboxyl 883 terminus, Eur. J Biochem 166 (1987) pp. 145‐149. 884 [60] M. Kumar, M. Singh, S.B. Singh, Optimization of conditions for expression of 885 recombinant interferon‐gamma in E.coli, Mol Biol Rep. 41 (2014) pp. 6537‐6543. 886 https://doi.org/10.1007/s11033‐014‐3537‐3. 887 [61] H. Billman‐Jacobe, A.L. Hodgson, M. Lightowlers, P.R. Wood, A.J. Radford, 888 Expression of ovine gamma interferon in Escherichia coli and Corynebacterium 889 glutamicum, Appl. Environ. Microbiol 60 (1994) pp. 1641‐1645. 890 [62] G.Y. Li, Z.Z. Xiao, H.P. Lu, Y.Y. Li, X.H. Zhou, X. Tan, X.Y. Zhang, X.L. Xia, H.C. Sun, A 891 simple method for recombinant protein purification using self‐assembling peptide‐ 892 tagged tobacco etch virus protease, Protein Expr. Purif. 128 (2016) pp. 86‐92. 893 https://doi.org/10.1016/j.pep.2016.08.013. 894 [63] S. Bhattacharjee, DLS and zeta potential ‐ What they are and what they are not?, J 895 Control Release 235 (2016) pp. 337‐351. 896 https://doi.org/10.1016/j.jconrel.2016.06.017. 897 [64] A. Meager, Biological assays for interferons, J Immunol. Methods 261 (2002) pp. 898 21‐36. 899 [65] M. Boyanova, R. Tsanev, I. Ivanov, A modified kynurenine bioassay for 900 quantitative determination of human interferon‐gamma, Anal. Biochem 308 (2002) 901 pp. 178‐181. 40 902 [66] A.K. Hess, P. Saffert, K. Liebeton, Z. Ignatova, Optimization of translation profiles 903 enhances protein expression and solubility, PLoS One 10 (2015) p.e0127039. 904 https://doi.org/10.1371/journal.pone.0127039. 905 [67] M. Boumaiza, A. Colarusso, E. Parrilli, E. Garcia‐Fruitos, A. Casillo, A. Aris, M.M. 906 Corsaro, D. Picone, S. Leone, M.L. Tutino, Getting value from the waste: 907 recombinant production of a sweet protein by Lactococcus lactis grown on cheese 908 whey, Microb Cell Fact 17 (2018) p.126. https://doi.org/10.1186/s12934‐018‐ 909 0974‐z. 910 [68] G. Marini, M.D. Luchese, A.P. Argondizzo, A.C. de Goes, R. Galler, T.L. Alves, M.A. 911 Medeiros, A.L. Larentis, Experimental design approach in recombinant protein 912 expression: determining medium composition and induction conditions for 913 expression of pneumolysin from Streptococcus pneumoniae in Escherichia coli and 914 preliminary purification process, BMC. Biotechnol 14 (2014) p.1. 915 https://doi.org/10.1186/1472‐6750‐14‐1. 916 [69] E. Garcia‐Fruitos, A. Aris, A. Villaverde, Localization of functional polypeptides in 917 bacterial inclusion bodies, Appl. Environ. Microbiol 73 (2007) pp. 289‐294. 918 https://doi.org/10.1128/AEM.01952‐06. 919 [70] A.K. Upadhyay, A. Murmu, A. Singh, A.K. Panda, Kinetics of inclusion body 920 formation and its correlation with the characteristics of protein aggregates in 921 Escherichia coli, PLoS One 7 (2012) p.e33951. 10.1371/journal.pone.0033951 922 [doi];PONE‐D‐11‐18189 [pii]. 923 [71] T. Hunig, The storm has cleared: lessons from the CD28 superagonist TGN1412 924 trial, Nat Rev. Immunol. 12 (2012) pp. 317‐318. https://doi.org/10.1038/nri3192. 925 [72] G. Suntharalingam, M.R. Perry, S. Ward, S.J. Brett, A. Castello‐Cortes, M.D. 926 Brunner, N. Panoskaltsis, Cytokine storm in a phase 1 trial of the anti‐CD28 927 monoclonal antibody TGN1412, N. Engl. J Med 355 (2006) pp. 1018‐1028. 928 https://doi.org/10.1056/NEJMoa063842. 929 [73] T.M. Li, J.W. Hook, III, H.G. Drickamer, G. Weber, Plurality of pressure‐denatured 930 forms in chymotrypsinogen and lysozyme, Biochemistry 15 (1976) pp. 5571‐5580. 931 [74] R. Mohana‐Borges, J.L. Silva, J. Ruiz‐Sanz, G. de Prat‐Gay, Folding of a pressure‐ 932 denatured model protein, Proc Natl Acad Sci U S A 96 (1999) pp. 7888‐7893. 933 [75] K. Ruan and G. Weber, Hysteresis and conformational drift of pressure‐dissociated 934 glyceraldehydephosphate dehydrogenase, Biochemistry 28 (1989) pp. 2144‐2153. 935 [76] J.M. Sanchez, V. Nolan, M.A. Perillo, beta‐galactosidase at the membrane‐water 936 interface: a case of an active enzyme with non‐native conformation, Colloids Surf. 41 937 B Biointerfaces. 108 (2013) pp. 1‐7. 938 https://doi.org/10.1016/j.colsurfb.2013.02.019. 939 [77] M. Randal and A.A. Kossiakoff, The 2.0 A structure of bovine interferon‐gamma; 940 assessment of the structural differences between species, Acta Crystallogr. D. Biol 941 Crystallogr. 56 (2000) pp. 14‐24. 942 [78] D.J. Thiel, M.H. le Du, R.L. Walter, A. D'Arcy, C. Chene, M. Fountoulakis, G. Garotta, 943 F.K. Winkler, S.E. Ealick, Observation of an unexpected third receptor molecule in 944 the crystal structure of human interferon‐gamma receptor complex, Structure 8 945 (2000) pp. 927‐936. 946 [79] S.S. Flores, V. Nolan, M.A. Perillo, J.M. Sanchez, Superactive beta‐galactosidase 947 inclusion bodies, Colloids Surf. B Biointerfaces. 173 (2019) pp. 769‐775. 948 https://doi.org/10.1016/j.colsurfb.2018.10.049. 949 [80] M. Martinez‐Alonso, E. Garcia‐Fruitos, A. Villaverde, Yield, solubility and 950 conformational quality of soluble proteins are not simultaneously favored in 951 recombinant Escherichia coli, Biotechnol Bioeng. 101 (2008) pp. 1353‐1358. 952 https://doi.org/10.1002/bit.21996. 953 [81] S. Ghobadi, M.R. Ashrafi‐Kooshk, H. Mahdiuni, R. Khodarahmi, Enhancement of 954 intrinsic fluorescence of human carbonic anhydrase II upon topiramate binding: 955 Some evidence for drug‐induced molecular contraction of the protein, Int J Biol 956 Macromol. 108 (2018) pp. 240‐249. 957 https://doi.org/10.1016/j.ijbiomac.2017.12.011. 958 [82] L. Wang, Q. Dong, Q. Zhu, N. Tang, S. Jia, C. Xi, H. Zhao, S. Han, Y. Wang, 959 Conformational Characteristics of Rice Hexokinase OsHXK7 as a Moonlighting 960 Protein involved in Sugar Signalling and Metabolism, Protein J 36 (2017) pp. 249‐ 961 256. https://doi.org/10.1007/s10930‐017‐9718‐x. 962 963 42 Name UNIPROT HS HS Sequence HSA NHSA a4vAHS Ref region size CYOB P0ABI8 591‐629 39 AGIVIAAFSTIFGFAMIWHI 30.696 0.787 0.767 This study WWLAIVGFAGMIITWIVKS HALRU Q9BP37 1‐17 17 MTYMCSILICLVLILCA 15.842 0.932 0.904 This study L6K2 NA 1‐6 6 LLLLLLKK 6.211 1.035 0.949 [55] 964 Table 1 Selection of APPs from predictions of “hot spots (HS)” of aggregation in 965 polypeptides by AGGRESCAN [54]. CYOB: Cytochrome bo3 ubiquinol oxidase subunit 1 966 from E. coli, HALRU: Aragonite protein AP7. NA: Not applicable. HS: hot spot. HSA: hot 967 spot area. NHSA: normalized HSA. a4vAHS: average aggregation‐propensity in each HS. 968 969 43 Soluble rBoIFN‐ rBoIFN‐ NPs Soluble rBoIFN‐ from NPs CSM (nm) Soluble Protein NPs NP core Resolubilized 1 rBoINF‐_Std 356.5 2 rBoIFN‐_ E. coli 358.1 3 rBoIFN‐_ L. lactis 357.4 4 rBoIFN‐_L. lactis 352.4 353.6 354.2 5 rBoIFN‐_L6K2_L. lactis 352.7 353.6 351.7 6 rBoIFN‐_CYOB_L. lactis 353.1 353.1 354 7 rBoIFN_HALRU_L. lactis 354.2 354.2 352 970 971 Table 2 Center of spectral mass (CSM) of IFN‐ protein preparations in soluble formats or 972 in protein NPs analyzed before and after the resolubilization protocol. 44