Efficient production and characterization of the novel and highly active antifungal protein AfpB from Penicillium digitatum

Filamentous fungi encode distinct antifungal proteins (AFPs) that offer great potential to develop new antifungals. Fungi are considered immune to their own AFPs as occurs in Penicillium chrysogenum, the producer of the well-known PAF. The Penicillium digitatum genome encodes only one afp gene (afpB), and the corresponding protein (AfpB) belongs to the class B phylogenetic cluster. Previous attempts to detect AfpB were not successful. In this work, immunodetection confirmed the absence of AfpB accumulation in wild type and previous recombinant constitutive P. digitatum strains. Biotechnological production and secretion of AfpB were achieved in P. digitatum with the use of a P. chrysogenum-based expression cassette and in the yeast Pichia pastoris with the α-factor signal peptide. Both strategies allowed proper protein folding, efficient production and single-step purification of AfpB from culture supernatants. AfpB showed antifungal activity higher than the P. chrysogenum PAF against the majority of the fungi tested, especially against Penicillium species and including P. digitatum, which was highly sensitive to the self-AfpB. Spectroscopic data suggest that native folding is not required for activity. AfpB also showed notable ability to withstand protease and thermal degradation and no haemolytic activity, making AfpB a promising candidate for the control of pathogenic fungi.

Nowadays fungal infections have become a serious threat to human health and food security. Human fungal infections have severe consequences on the growing number of immune-compromised patients with high mortality rates. In addition, the control of plant diseases caused by phytopathogenic fungi represents a big challenge in agriculture. The emergence of antifungal resistant strains is in continuous growth, emphasizing the urgent need for the development of novel antifungal agents with properties and mechanisms of action different from existing ones 1 .
Antifungal proteins (AFPs) secreted by filamentous fungi have been considered promising candidates for the development of novel antifungal compounds and therapies. AFPs are small, highly stable, cationic, cysteine-rich proteins (CRPs) stabilized by up to four disulphide bridges 2 . They are usually secreted in high amounts by filamentous Ascomycetes, mainly from the genera Penicillium and Aspergillus, and show potent antifungal activity against non-self fungi at micromolar concentrations.
It is increasingly clear that fungi have a complex repertoire of AFPs that offers a great potential to obtain new antifungal agents. In a previous study, fungal genome sequences were searched to identify AFP-like sequences and conduct detailed phylogenetic studies 3 . Based on phylogenetic clustering but also on sequence alignment, cysteine pattern, intron position and Pfam domain identification, we proposed the classification of fungal AFPs into three classes: A, B and C 3 , expanding the two previously reported ones that divided the fungal AFPs into two clusters 4,5 . Recently, a fourth and distantly related group of AFPs has been described 6 . The fungal genomes that encode more than one AFP include those from Penicillium chrysogenum and Penicillium expansum, which have one AFP belonging to each of three different classes 3 . In the case of the three P. chrysogenum AFPs (class A PAF,

Results
Recombinant production of AfpB in P. digitatum and P. pastoris. We used the P. chrysogenum-based expression cassette 19 to express AfpB in P. digitatum under the regulation of the strong paf promoter and terminator sequences (Fig. 1). In our previous work, the presence of the paf SP-pro sequence warranted the secretion of PAF, PAF variants and NFAP proteins into the supernatant independently of the Penicillium species used as cell factory. Here, in order to study the functionality of the native SP-pro sequence of the AfpB protein, two different transformation vectors with two expression cassettes for the afpB gene insertion in P. digitatum were constructed. Both constructions included the paf promoter and terminator sequences. However, in the first gene construction, the full-length afpB coding sequence (afpB) was cloned, while the second approach included the in silico predicted mature afpB coding sequence (afpB*) fused to the paf SP-pro sequence (Fig. 1a). After The PAF diagram constitutes the schematic representation of the P. chrysogenum-based expression cassette; in blue: paf promoter (Ppaf), paf gene including the paf SP-pro sequence (SP), and paf terminator (Tpaf). Amino acids involved in the SP-pro peptide cleavage are coloured in blue. The AfpB diagram represents the genetic construction with the full-length AfpB coding sequence (in red) cloned under the control of the Ppaf and Tpaf sequences (in blue). Amino acids involved in the predicted SP-pro peptide cleavage are in red. The AfpB* diagram corresponds to the genetic construction with the in silico predicted AfpB coding sequence (afpB*) cloned under the control of the Ppaf, paf SP-pro sequence and Tpaf. (b) The PpAfpB* diagram represents the construction with the in silico predicted AfpB coding sequence (afpB*, in red) cloned under the control of the AOX1 promoter for methanol-induced expression and terminator sequences (in black) and yeast α-factor signal sequence (α-factor SS) (in green). evaluation of positive transformants for protein production, one clone with the highest production of the recombinant AfpB protein for each construction was selected for further characterization. The selected producer strains were PDSG2441 for AfpB and PDSG3543 for AfpB*. The genetic modification of these strains was confirmed molecularly ( Supplementary Fig. S1).
The growth of the selected transformant strains in solid medium is shown in Fig. 2. Since the P. chrysogenum PAF was already expressed in P. digitatum 19 , the PAF transformant strain (PDSG1521) was chosen as internal control in this study. AfpB transformants showed a moderate reduction of colony diameter on potato dextrose agar (PDA) plates ( Fig. 2a,b), higher for the AfpB (PDSG2441) than for the AfpB* (PDSG3543) strain. However, the growth was indistinguishable from that of the parental strain in P. digitatum minimal medium (PdMM) (Fig. 2b). The PAF transformant did not show different phenotype to the parental strain in both media. The moderate reduction of growth shown in PDA plates for the transformants AfpB and AfpB* did not result in a difference in their pathogenicity and virulence shown during orange fruit infection ( Supplementary Fig. S2).
Heterologous production of AfpB* was also addressed in P. pastoris. The genetic construct was designed for its gene inducible expression (AOX1 promoter) and secretion of the corresponding protein (PpAfpB*, for AfpB* produced in P. pastoris) to the extracellular medium (Fig. 1b). With this purpose, afpB* cDNA was fused in frame to a shortened yeast α-factor SS, in which the Ste13 protease cleavage site was removed and the Kex2 protease cleavage site maintained. This modification was introduced to avoid the Ste13 partial cleavage of the secretion signal resulting in the accumulation of an inactive protein, as previously reported for the AFP from A. giganteous 23 . The resulting construct was used to transform P. pastoris cells, and two independent transformant were selected for PpAfpB* production.
Single-step cationic exchange chromatography allowed AfpB purification from P. digitatum and P. pastoris culture supernatants. Selected clones for AfpB production in P. digitatum were grown in PdMM and, after clearing the culture broth from insoluble matter, the proteins in the supernatant were purified by one-step cation-exchange chromatography. Optimal production was achieved after 11 days of growth, and the protein amounts varied between 12 (AfpB) and 20 (AfpB*) mg/l. Both proteins eluted as a single chromatography peak at 0.25 M NaCl, and SDS-PAGE analysis revealed a single-protein band in both, having the same apparent molecular mass of approximately 6 kDa in agreement with the AfpB predicted molecular mass (6.46 kDa) (Fig. 3a, top panel). When comparing with the slightly smaller PAF (6.2 kDa), AfpB showed faster migration than expected. Anomalous migration in SDS-PAGE was also observed for NFAP 19 , PgAFP 9 and other fungal AFPs 5 and seems to be related to the extreme isoelectric points of these proteins 26 .
Two independent P. pastoris colonies were assayed to evaluate the PpAfpB* accumulation upon methanol induction at different time points. A polypeptide showing similar electrophoretic mobility to the AfpB produced in P. digitatum was observed in the supernatant of P. pastoris cells harbouring afpB* cDNA construct and grown under inducing conditions (Fig. 3b, top panel). The intensity of this band increased with time, reaching maximum levels at 48 hours of growth in methanol medium with strong aeration. This band was absent in the supernatant  Immunodetection confirmed the absence of AfpB in the parental strain. Western blot analyses of 11-days old P. digitatum culture supernatants of different fungal strains and purified AfpB proteins were performed with antibodies against P. chrysogenum PAF 27 and PAFB. PAFB is a protein from P. chrysogenum that is 88% identical to AfpB, and from which an antiserum was raised (unpublished). Culture supernatants of PDMG122 (null ΔafpB) and PDMG152 (constitutive afpB C ) were also included in the analyses as controls and to confirm previous absence of detection by Coomasie Blue staining 3 . Results indicated that both AfpB proteins and the corresponding supernatants reacted with anti-PAFB antibody (Fig. 2a, middle panel) while only purified PAF and the corresponding transformant strain PDSG1521 supernatant reacted with PAF antiserum (Fig. 2a, bottom panel). As expected, no protein band in the supernatant of the previous null transformant strain (ΔafpB) reacted with the anti-PAFB antibody. Remarkably, no immunoreaction was observed in the culture supernatant of either the parental strain (CECT 20796) or the constitutive afpB c transformant strain (PDMG152), fully demonstrating our previous conclusions on the absence of detection of any differential band or AfpB-like protein in those P. digitatum culture supernatants 3 .
Western blot analyses were also performed for the heterologously produced PpAfpB* protein with antibodies against PAFB (Fig. 3b, bottom panel). Supernatants from P. pastoris pPICZαA-AfpB* containing colonies grown under inducing conditions and purified PpAfpB* reacted with the antiserum while, as anticipated by Coomassie staining, no protein was detected in the non-induced supernatants or in cell lysates.
Mass spectrometry revealed different processing of PAF and AfpB SP-pro sequence. Molecular mass of AfpB proteins purified from P. digitatum was also determined by matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF MS) analysis. Single peaks corresponding to average masses of 6570.53 and 6458.01 Da were detected for AfpB and AfpB*, respectively (Fig. 4). The experimental mass of AfpB* is consistent with the calculated theoretical mass of the oxidized protein predicted after cleavage from the PAF SP-pro sequence (6456.23 Da), indicating the presence of three intra-molecular disulphide bonds and the absence of other post-translational modifications. Interestingly the average mass detected for AfpB revealed the presence of an extra leucine residue at the N-terminus end of the protein (calculated theoretical mass 6575.39 Da) suggesting a different cleavage of the native AfpB SP-pro sequence.
The MS corresponding to PpAfpB* revealed one main signal of 6455.20 Da ( Supplementary Fig. S3) in accordance with the calculated molecular mass of the oxidized protein form and the proper processing of the yeast α-factor SS. An additional minor signal of 7104.54 Da that did not correspond to any potential variable N-terminus was detected.
Electronic circular dichroism spectroscopy assays revealed incomplete refolding capability of the AfpB* variant. Electronic circular dichroism (ECD) spectroscopy was used to determine protein conformation and proper folding of both P. digitatum AfpB and AfpB* variants. The ECD spectra recorded at 25 °C were nearly identical for both variants ( Fig. 5 and Supplementay Fig. S4), and similar to those recorded for PAF, NFAP 19 and other disulphide bridged, β-structured proteins 28 . The spectra had two maxima at 195 and 229 nm. The maximum at 229 nm was mainly attributed to the presence of disulphide bridges while the maximum centred at 195 nm reflected contributions from both β-pleated conformation and the electronic transitions of disulphide bridges. Spectra measured at 95 °C reflected the loss of ordered structure in both AfpB variants. After cooling back to 25 °C, the native fold of AfpB was restored completely and almost immediately (Fig. 5a). In contrast, a slow structural reorganization of AfpB* took place with incomplete refolding even after 72 h (Fig. 5b). This observation may be attributed to the absence of the leucine residue at the N-terminus of this protein variant.
AfpB was highly active against Penicillium species including P. digitatum. Both AfpB protein variants, and PAF as an internal control, were tested for their antimicrobial activity towards Escherichia coli, Saccharomyces cerevisiae and a selection of filamentous fungi that include, in addition to the P. digitatum parental strain, several plant pathogens such as the citrus fruit specific Penicillium italicum, the main postharvest pathogen of pome fruit P. expansum, the polyphagous Botrytis cinerea, the rice blast fungus Magnaporthe oryzae, and the soilborne plant pathogen Fusarium oxysporum. The PAF producer P. chrysogenum strain, and a strain from Aspergillus niger which is particularly sensitive to the PAF protein were also tested. No differences in antimicrobial activity were observed among AfpB, AfpB* and PpAfpB* produced in either P. digitatum or P. pastoris ( Supplementary Fig. S5). Therefore, only results from AfpB produced in P. digitatum are shown. AfpB was inactive against E. coli and S. cerevisiae at the highest concentration tested (200 µg/ml). AfpB showed antifungal activity and inhibited the growth of all fungi tested, with the exception of M. oryzae (Fig. 6). The minimum inhibitory concentration (MIC) values varied from 1.6 µg/ml in P. italicum to 100 µg/ml in F. oxysporum. The three plant pathogenic Penicillium species tested were considerably sensitive to the protein, including its producer strain. With the exception of the PAF-sensitive A. niger, AfpB showed higher antifungal activity than PAF, with MIC values at least one order of magnitude lower.
AfpB showed high protease and thermal resistance. The sensitivity of AfpB to proteolytic digestion was tested using proteinase K and evaluating the residual antifungal activity against P. digitatum. Preincubation of AfpB and AfpB* with proteinase K did not decrease the antifungal activity of any of both protein variants ( Supplementary Fig. S6), pointing out the high protease resistance of these antifungal proteins.
The incomplete refolding of AfpB* found with the ECD analyses allowed us to test and determine the effect of heat treatment and folding on the antifungal activity. After heat treatment at 95 °C for 5 min and cooling back to 25 °C (mimicking the ECD conditions) the antifungal activity of both protein variants was practically equal and comparable to that observed without treatment ( Supplementary Fig. S7), suggesting that native AfpB refolding is not necessary for its antifungal activity.
Further experiments were conducted to compare the thermal stability of PAF and AfpB. After heat treatment at 80 °C or 95 °C for 10 and 60 min and cooling back to 25 °C, the antifungal activity of PAF and AfpB were tested against P. digitatum. Fungal conidia were exposed to twofold MIC concentrations of heat-treated proteins (100 µg/ml for PAF, Fig. 7a, and 6.5 µg/ml for AfpB, Fig. 7b). The antifungal activity of PAF against P. digitatum at that concentration resulted in a growth reduction of 70-80% after 10 min of heat treatment at 95 °C and 80 °C, respectively in comparison with the untreated control (>95%). After 60 min of heat treatment, PAF showed a P. digitatum growth reduction of 60% at 80 °C and 55% at 95 °C. In the same conditions, AfpB showed total inhibition of the fungus growth (>95%, comparable to the non-treated control), demonstrating that AfpB is more resistant to heat treatment than PAF.
AfpB showed no haemolytic activity against human red blood cells. Haemolytic activity assays are performed in order to determine the cytotoxicity of specific proteins and peptides against eukaryotic cells by their ability to lyse human red blood cells (RBCs) 29 . In this study, we determined the haemolytic activity of both AfpB variants at different concentrations from 1 to 100 μM (approximately 200 times the MIC against P. digitatum). In these experiments, PAF and melittin from honeybee 30 were included as negative and positive controls, respectively. Since the haemolytic properties of cationic peptides may show ionic strength dependence, haemolytic assays were conducted not only with a high ionic strength phosphate NaCl buffer (PBS) but also with a low ionic strength isotonic glucose phosphate buffer (PBG) 29 .
As expected, the haemolytic peptide melittin caused 100% haemolysis at the highest concentration tested and around 80% at 10 μM (Fig. 8). By contrast, none of the AFPs showed haemolytic activity at any of the concentrations tested, neither in the presence of NaCl as in PBS (Fig. 8a) nor glucose (Fig. 8b), with values close to the control where no protein was added.

Discussion
In this study, we described the biotechnological production and characterization of AfpB, the only AFP encoded in the phytopathogenic fungus P. digitatum. This protein is a new member of AFPs grouped into the class B phylogenetic cluster 3 . For the first time, AfpB was produced, detected and purified from culture supernatants of recombinant P. digitatum and P. pastoris strains. The homologous production of AfpB in P. digitatum with the paf-derived expression cassette resulted in high yields around 12-20 mg protein/l, whereas the heterologous production in P. pastoris resulted in yields 10-fold lower (1.2-1.4 mg/l). Both approaches allowed one-step purification of fully active AfpB, and the purified protein from the two systems showed the same antifungal potency. Our results demonstrate that both filamentous fungi and yeasts are good production platforms for AFPs. The homologous system appears as the most efficient system reaching high yields; whereas the heterologous system has the advantage of time reduction since optimal production was achieved after 11 days of growth in P. digitatum while only 2 days were needed with the P. pastoris system.
Previously, three other class B AFPs were characterized: Anafp from A. niger 31 , PgAFP from P. chrysogenum 32 and MAFP1 from Monascus pilosus 33 . They show an amino acid identity with P. digitatum AfpB of 79, 88, and 83%, respectively, much higher than the percentage of identity shared between AfpB and the class A PAF from P. chrysogenum, with only 33% identity 3 . The three of them were successfully isolated from the culture supernatants of the corresponding native and producer fungus. Notably, the concentration of purified PgAFP obtained from P. chrysogenum CECT 20922, which was isolated from cured meat, was up to 700 µg/ml 32 . Also, the class A Figure 7. Effect of heat treatment on the antifungal activity of PAF and AfpB. PAF (a) and AfpB (b) (100 and 6.5 µg/ml, respectively) were exposed to 80 and 95 °C for 10 and 60 min, respectively. Values represent the percentage of growth (%) of P. digitatum in the presence of non-treated and treated protein compared to P. digitatum in the absence of protein (control, C). Data show mean ± s.d. OD 600 of triplicate samples. Proteins were used at the concentrations indicated (from 1 to 100 µM). The cytolitic peptide melittin was included for comparison. The haemolytic activity is given as the mean ± s.d. of the percentage of human RBCs haemolysis (three replicates), as compared with the positive control in the presence of the detergent Triton X-100 (regarded as 100% haemolysis).
PAF and AFP are secreted in large amounts by P. chrysogenum 17 and A. giganteus 13 , respectively. However, previous attempts to detect AfpB in P. digitatum culture supernatants failed, even in P. digitatum transformants for afpB constitutive expression under the strong gpdA promoter from A. nidulans 3 . Remarkably, those transformant strains showed a drastic reduction of axenic growth, abnormal hyphal morphology and delayed conidiogenesis 3 . Here, we have confirmed with immunodetection that AfpB is not produced by any of the previous strains. By contrast, afpB expression under the regulation of the strong paf promoter and the corresponding paf terminator warranted the production and secretion of the protein at high levels, concomitant with only a moderate reduction of axenic growth. We showed that AfpB production is similar in the presence of either the paf-or the afpB-specific SP-pro sequences, indicating that the impossibility to detect AfpB in either the parental or constitutive strain cultures is not due to the native AfpB SP-pro peptide, but to the regulatory sequences. The P. chrysogenum paf promoter has been demonstrated to have greater efficiency than the A. nidulans gpdA constitutive promoter used for the expression of reporter genes in P. chrysogenum 34 , however, the approximately 2-fold difference in efficiency cannot account for the absence of AfpB accumulation when the gpdA promoter was used. In addition, differences in 5′ untranslated regions (UTRs) in the mRNA have been demonstrated to play an important role in determining the translation efficiency of proteins in filamentous fungi 35 . Nevertheless, AfpB production was also achieved in the P. pastoris expression system under the control of the AOX1 promoter using the yeast α-factor SS. Additional studies are required to elucidate the exact mechanism underlying the problematic of native afpB translation and protein accumulation in P. digitatum.
In accordance with the antimicrobial profile of native AFPs that possess antifungal but not antibacterial activity 1,17 , AfpB displayed antifungal activity against selected filamentous fungi and it was inactive against bacteria (E. coli) or yeast (S. cerevisiae). However, the homologous protein AnAFP from A. niger inhibited the proliferation of Candida albicans and S. cerevisiae 31 whereas the other two class B representatives did not show any effect on different bacteria and yeast species 32,33 . Recently a new AFP from N. fischeri named NFAP2 proved to be highly effective against targeted yeasts including clinically relevant Candida species 6 . NFAP2, which seems to be the first member of a new, phylogenetically distinct fourth group among AFPs, showed MIC values in the range of 0.2-1.5 µg/ml. Potential AfpB anti-yeast activity against non-laboratory strains and/or pathogenic isolates deserves future research.
Until this study it was assumed that AFPs are not active against the producer fungus, and it has been speculated that fungi possess innate sensing or defense systems which enable them to discriminate between AFPs from self or non-self origin 36 . Whether the function of fungal AFPs is mainly defensive or associated with fungal growth and development is still controversial. Null mutation of the afpB gene in P. digitatum did not affect spore production, growth or virulence 3 , contrarily to that described for the null mutation of the paf gene in P. chrysogenum that severely affected asexual development, spore production and associated gene regulation 37 . Recently, novel functions related to nutrient recycling during starvation, autophagy, and development for the AnAFP produced by A. niger have been proposed 36 . It is very tempting to speculate that AFPs from different classes do have distinct biological functions in the producer fungus and also distinct mechanisms of antifungal action. This seems to be the case in the AfpB/PAF couple, according to the different phenotypes of the corresponding null mutants but also to the different activity profiles of both proteins against P. digitatum, P. expansum or A. niger.
AfpB is highly active against the own producer P. digitatum strain when added exogenously to the culture media (MIC = 3.2 µg/ml; Fig. 6), which is not the case of PAF against P. chrysogenum (MIC > 200 µg/ml). However, the fungal strain engineered to express the afpB gene produces high amounts of AfpB without deleterious effect to the axenic growth or pathogenicity of P. digitatum. This fungus is also susceptible to PAF protein (MIC = 50 µg/ml) but P. digitatum engineered strains were also able to produce PAF in high quantities 19 . In contrast, the heterologous expression of the nfap gene in a NFAP-sensitive A. nidulans strain only permitted the production of low amounts of protein (approximately 1.7 mg/l), but it provoked reduced hyphal growth and delayed and abnormal germination 38 . Other fungal AFPs such as AnAFP and A. giganteous AFP have been described as only moderate-active towards the producer strains 36 whereas the class B PgAFP was not active against its producing strain P. chrysogenum CECT 20922 while it showed antifungal activity against another P. chrysogenum strain, pointing out strain-dependent activity 32 .
AfpB is unique for its very high antifungal activity against filamentous fungi that include the fungus from which the gene was identified. Our side-by-side experiments demonstrate that AfpB has higher levels of antifungal efficacy when compared to PAF (Fig. 6). With respect to other class B proteins already characterized, AnAfpB showed similar antifungal activity against F. oxysporum 31 , and PgAfpB was as well active against other P. expansum and A. niger strains 32 . The activity of AfpB against the Penicillium species tested deserves to be emphasized. All of them were highly sensitive to AfpB: MIC values in the range of 1.6-6.25 µg/ml that is equivalent to 0.25-1 µM. Remarkably, current fruit losses due to Penicillium decay are very important in agriculture since Penicillium species reproduce very rapidly and their spores are ubiquitous in the atmosphere, facilitating the dissemination on fruit surfaces. P. digitatum and P. italicum are the main causal agents of the citrus green and blue mold diseases, respectively 10 . On the other hand, P. expansum, the causal agent of the blue mold disease, is one of the most important pathogens of pome fruits, causing serious crop losses worldwide 11 . Unfortunately, there are only a few drugs available for the effective treating of fungal infections, and the development of resistance against fungicides used in agriculture is increasingly alarming. Thus, the use of AfpB as a powerful alternative in the control of pathogenic fungi could be of further interest.
In a previous work, we identified two cysteine-containing cationic peptides, PAF112 and PAF118, derived from two surface-exposed loops in AfpB, with moderate antifungal activity against P. digitatum with a MIC value of 64 µM 22 . Now our results show that AfpB is about 100 times more active than the peptides derived from its sequence suggesting that additional motifs/sequences in the protein are required to achieve full activity.
As generally described for fungal AFPs 16 , AfpB showed a remarkable ability to withstand protease and thermal degradation. It has been assumed that the compact tertiary structure stabilized by disulphide bridges of AFPs favours protein stability. Although the tertiary structure of AfpB has not yet been determined, the predicted AfpB structure by in silico modelling showed a similar tertiary structure to those reported for A. giganteous AFP 12 and P. chrysogenum PAF 16,39 . In the present study ECD spectroscopic measurements, consistent with the previous modelling, suggested the presence of disulphide bridges and a β-pleated conformation, and proved that both AfpB variants have the same structural elements as other fungal AFPs 19,24 . Our data also demonstrate that the lack of the leucine residue at the N-terminal end of AfpB* does not affect neither the antifungal activity nor the protease and thermal tolerance, although this protein variant showed incomplete refolding in the conditions tested. Moreover, the demonstration that the heat treatment which denatures AfpB*, mostly irreversibly, does not impair antifungal efficacy indicates that protein folding is not critical for the antifungal activity.
Finally, our results suggest that AfpB can be preliminary regarded as safe since it did not show haemolytic activity against human RBCs even in assays conducted at low ionic strength isotonic conditions, which are considered more sensitive for detecting the haemolytic activity of cationic peptides 29 .
In summary, this study identifies AfpB from P. digitatum as a highly active antifungal protein against filamentous fungi, including the own producer fungus. This new member of class B AFPs was produced for the first time in two eukaryotic cell factories, the fungus P. digitatum and the yeast P. pastoris. Both expression systems allowed proper protein folding, efficient production and single-step purification from culture supernatants. The remarkable stability, absence of haemolytic activity and high levels of antifungal efficacy against filamentous fungi, especially Penicillium species, suggest the potential use of AfpB as an antifungal agent.

Methods
Strains, media and culture conditions. The parental isolate P. digitatum CECT 20796 (PHI26) 10 and all of the transformants were cultured on PDA plates for 7-10 days at 25 °C. To analyse the growth on solid medium, 5 µl of conidial suspension (5 × 10 4 conidia/ml) were deposited on the centre of PDA and PdMM plates 19 , and the diameter of growth was monitored daily from 3 to 10 days. For AfpB production, strains were inoculated at a concentration of 10 6 conidia/ml in 500 ml of PdMM and incubated at 25 °C with shaking for 14 days. For the antimicrobial assays, B. cinerea CECT 2100, F. oxysporum 4287, P. expansum CMP1, P. italicum CECT 2294, P. chrysogenum Q176, and M. oryzae PR9 were incubated at 25 °C, E. coli JM109 and A. niger CBS 120.49 at 37 °C, and S. cerevisiae BY4741 at 30 °C. For the fungal transformation, vectors were propagated in E. coli JM109 grown in Luria Bertani (LB) medium supplemented with 100 µg/ml ampicillin or 75 µg/ml kanamycin. Agrobacterium tumefaciens AGL-1 strain was grown in LB supplemented with 20 µg/ml rifampicin at 28 °C. P. pastoris X-33 wildtype strain was cultured on yeast extract peptone dextrose (YPD) medium at 28 °C.
Generation of the P. digitatum AfpB producer strains. To generate the P. digitatum AfpB producer strains, two different genetic approaches were performed. The first approach included the full-length AfpB coding sequence (afpB) cloned under the control of the paf gene promoter and terminator sequences from P. chrysogenum. The second approach included the in silico predicted mature AfpB coding sequence 3 (afpB*) cloned under the control of the paf gene promoter, paf SP-pro sequence and the paf gene terminator (Fig. 1a). The specific primers used to generate the P. digitatum AfpB producer strains are described in Supplementary Table S1 and Supplementary Fig. S1a. The afpB gene was amplified from P. digitatum CECT 20796 genomic DNA, while the paf gene promoter, SP-pro and terminator sequences were generated by PCR amplification of the fragments from the vector pSK275paf 19 . PCR reactions were performed using AccuPrime High-Fidelity polymerase (Invitrogen) and the constructs verified by Sanger DNA sequencing. The two different DNA constructions were generated by Fusion PCR 40 and cloned into the pGEM-T ® Easy vector system (Promega) from where they were excised with XmaI and XbaI and inserted into the digested binary vector pBHt2 41 containing the hygromycin resistant cassette (hph) used as positive selection marker. The binary vectors were transformed into A. tumefaciens AGL-1, and the fungal transformation of P. digitatum parental strain was performed by A. tumefaciens-mediated transformation (ATMT) 41,42 . Positive transformants were confirmed by PCR amplification of genomic DNA 43 (Supplementary Fig. S1; Supplementary Table S1).
Generation of the P. pastoris AfpB producer strains. To generate the P. pastoris AfpB producer strain, the P. digitatum cDNA encoding the predicted AfpB protein sequence (afpB*) was obtained as previously described 43 and amplified by PCR using the primers MO3 and MO4 (Supplementary Table S1). The afpB* cDNA sequence was inserted into the XhoI and XbaI digested pPICZαA plasmid (Invitrogen) carrying AOX1 promoter for methanol-induced expression of afpB* in P. pastoris. The afpB* cDNA was inserted in frame with a modified version of the yeast α-factor SS, lacking the Ste13 cleavage sites, but including the Kex2 signal cleavage site (Fig. 1b). Constructs were verified by Sanger DNA sequencing. The digested plasmid pPICZαA-afpB with AvrII was used to transform P. pastoris X-33 cells by electroporation. After 5 min of incubation at 4 °C, cells were subjected to a pulse (1.5 kV, 200 Ω) in 0.2-cm cuvettes and were immediately diluted with 1 ml of ice-cold 1 M sorbitol and plated on YPD medium containing 1 M sorbitol and 100 µg/ml zeocin. Plates were incubated for 3-4 days at 28 °C. Two independent colonies were selected (#1 and #3) and used for PpAfpB* production. Transformant cells were grown in buffered minimal glycerol medium BMG (0.1 M potassium phosphate buffer, pH 6; 1.34% yeast nitrogen base; 4 × 10 -5 % biotin; 1% glycerol), and transferred to BMM (buffered minimal medium with 0.5% methanol instead of glycerol as carbon source) for afpB* induction.
Protein production and purification. AfpB variants were purified from the supernatants of P. digitatum transformant strains (AfpB and AfpB*) and P. pastoris positive transformants (PpAfpB*). Cell-free supernatants of P. digitatum grown on PdMM for 11 days or P. pastoris grown on BMM for 2 days were collected by centrifugation and dialyzed (2 K MWCO,) against 20 mM phosphate buffer pH 6.6. Dialyzed solutions were applied to an AKTA Purifier system equipped with a 6 ml RESOURCE S column (GE Healthcare) for P. digitatum supernatants Haemolytic assays. The haemolytic activity of the proteins was determined in a 96 round-bottom microtiter plate on human 1:40 diluted RBCs as previously described 29,46 with some modifications. Briefly, RBCs were harvested by centrifugation for 15 min at 100 × g and washed three times in 35 mM PBS (pH 7, 150 mM NaCl) or PBG (250 mM glucose as osmoprotectant). One hundred µl of twofold protein concentration were mixed with 100 µl of RBCs in triplicate. Plates were incubated for 1 h at 37 °C and centrifuged for 5 min at 300 × g. Supernatants (100 µl) were transferred to a new microtiter plate and the absorbance was measured at 415 nm.
No haemolysis and 100% haemolysis were determined in controls with a mixture of PBS or PBG, and 0.1% Triton X-100, respectively. The haemolytic activity was calculated as the percentage of total haemoglobin released compared with that released by incubation with 0.1% Triton X-100.
Fruit infection assays. The inoculation of P. digitatum strains on non-treated freshly harvested orange fruits (Citrus sinensis L. Osbeck cv Navelina) was conducted as previously described 47 . Briefly, three replicates of five fruits were inoculated with 5 µl of conidial suspension (5 × 10 4 conidia/ml) at four wounds around the equator. Orange fruits were stored at 20 °C and 90% relative humidity. Each wound was scored daily for infection symptoms on consecutive post inoculation days. Data availability. All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).