The Antifungal Protein AfpB Induces Regulated Cell Death in Its Parental Fungus Penicillium digitatum

Disease-causing fungi pose a serious threat to human health and food safety and security. The limited number of licensed antifungals, together with the emergence of pathogenic fungi with multiple resistance to available antifungals, represents a serious challenge for medicine and agriculture. Therefore, there is an urgent need for new compounds with high fungal specificity and novel antifungal mechanisms. Antifungal proteins in general, and AfpB from Penicillium digitatum in particular, are promising molecules for the development of novel antifungals. This study on AfpB’s mode of action demonstrates its potent, specific fungicidal activity through the interaction with multiple targets, presumably reducing the risk of evolving fungal resistance, and through a regulated cell death process, uncovering this protein as an excellent candidate for a novel biofungicide. The in-depth knowledge on AfpB mechanistic function presented in this work is important to guide its possible future clinical and agricultural applications.

F ungal infections threaten human health worldwide, causing the death of about 1.5 to 2 million people every year (1). Invasive fungi have fatal consequences in vulnerable patients, such as those immunodepressed by anti-cancer chemotherapies, corticosteroid treatments, HIV/AIDS, or organ transplantation. Fungi also have a significant impact on global food security, damaging more than 20% of crop production, spoiling an estimated 10% of harvested crops, and causing disease in domesticated animals (2). In addition, food safety is challenged by mycotoxigenic fungi that contaminate food and feed with toxins detrimental for health. Consequently, there is a huge demand for fungicides, accounting for a market of $18.7 billion in 2019 that is expected to grow with the emergence of hypervirulent strains and the geographical expansion of pathogens due to global warming and globalization. Only a few classes of fungicides are available today due to the complexity of targeting these eukaryotic microorganisms without affecting plant, animal, or human hosts. These fungicides target the main differences between fungal and host cells, including the plasma membrane and the cell wall composition, ergosterol biosynthesis, or disrupting DNA or RNA synthesis in a fungus-specific manner (3). However, the agricultural application of fungicides has led to the development of resistance to the few available fungicides not only in plant pathogens but also in those causing human diseases (2). The scarceness of licensed fungicides and the emergence of multifungicide resistance make the development of novel antifungal compounds with new modes of action crucial to fight human-and plant-pathogenic fungi.
Detailed knowledge of the mechanistic function of AFPs is an important requisite for their development and application. Notably, variations exist in the spectrum of fungi targeted by different AFPs, demonstrating that they act in a species-specific manner (5,6,16,21). Although AFPs are similar in structure, they differ in amino acid composition and sequence and, presumably, in their mode of action. To date, more than 50 members have been identified in the AFP family, AFP from Aspergillus giganteus and PAF from Penicillium chrysogenum being the first identified and best studied (4,7,22). These two proteins are quite similar in structure and antifungal spectrum, but their mechanisms of action are significantly different. Whereas AFP inhibits chitin synthesis and disturbs plasma membrane integrity, PAF is actively internalized in target cells, where it triggers apoptosis (21,(23)(24)(25). Divergence in the mode of action makes it necessary to characterize each AFP individually for their efficient biotechnological application as well as for a better understanding of their biological functions.
This work reports a detailed study on the mechanistic function of the AfpB antifungal protein compared to the well-studied PAF. AfpB is a recently identified AFP in the genome of the citrus postharvest pathogen Penicillium digitatum (26). The protein could not be detected in natural P. digitatum cultures but has been biotechnologically produced in large amounts in the same fungus or in a plant-based system (5,20). This antifungal protein is very active against major phytopathogenic fungi, especially Penicillium species, including the parental P. digitatum (5). Moreover, AfpB has been shown to be effective in plant protection treatments against fungal infections, such as Botrytis cinerea, the causal agent of gray mold disease in more than 200 vegetable species (6,20). We show here that AfpB cytotoxicity is determined by interaction with the cell wall of the target fungus and by its active internalization into fungal cells, where it induces a regulated cell death program. This knowledge is crucial for the future development of AfpB as a novel biofungicide. entered the hyphae when not exposed to the inhibitor. Hyphae treated for 2 h with AfpB in the presence of NaN 3 appeared less damaged than those not exposed to the inhibitor (Fig. 2). These observations suggest that AfpB fungicidal activity is associated with its active entry into the fungal cells.
Cell wall interaction is important for AfpB antifungal activity and is blocked in a protein mannosylation mutant. As a cell-penetrating protein, AfpB must pass through the fungal cell wall to reach the plasma membrane on its way to the cytoplasm. To examine the role of the cell wall in the activity of AfpB, we stained fungal hyphae with calcofluor white (CFW), a fluorescent dye that binds to the chitin polymers of fungal cell walls. CFW prevented AfpB from getting into the intracellular space when hyphae were stained prior to AfpB treatment, whereas CFW staining after treatment did not interfere with AfpB lytic activity (Fig. 4A). These observations indicate that the fungal cell wall has a key role in facilitating AfpB interaction with target cells and, therefore, its activity.
In addition, we characterized the response to AfpB of the Pdpmt2 deletion mutant strain, which is known to have cell wall defects and greater tolerance to antifungal peptides, such as PAF26 (30). This is a P. digitatum null mutant on the O-mannosyltransferase 2 (Δpmt2) that affects protein mannosylation, including the mannoproteins on the outer cell wall. Surprisingly, we did not observe fluorescence on Δpmt2 germlings treated with 1.5 M BODIPY-AfpB, either at the cell wall or intracellularly, at any time point (Fig. 4B). This observation suggests that cell wall mannoproteins are required for the interaction with AfpB. Moreover, growth inhibition assays showed that the Δpmt2 mutant was more resistant to AfpB than the wild-type (WT) strain (Fig. 4C). Altogether, these results suggest that mannoproteins determine AfpB interaction with the cell wall and, thus, its antifungal activity. AfpB induces rapid ROS production prior to cell death in fungal hyphae. The generation of reactive oxygen species (ROS) is associated with cell death, acting either as signaling molecules in regulated cell death or as toxic compounds causing oxidative damage to lipids, proteins, and nucleic acids and uncontrolled cell destruction. Several studies have previously reported the production of ROS in fungal cells triggered by AFPs (29,(31)(32)(33)(34). ROS production was monitored in these studies after long and high-concentration treatments, with deleterious effects observed; it was unclear  whether ROS was signaling or a consequence of cell death. To study the oxidative status of hyphae treated with AfpB, we used the permeable fluorescent CM-H 2 CFDA probe as an indicator of ROS cell content (35). Fluorescent hyphae were visible immediately after the addition of AfpB, whereas fluorescence was barely detected in PAF-treated hyphae at the same concentration (1.5 M) (Fig. 5A). These observations show that AfpB induces the rapid generation of ROS in target fungal cells. However, staining with the cell death marker propidium iodide (PI), which fluoresces upon interaction with nucleic acids and penetrates only cells with damaged plasma membranes, showed red fluorescent hyphae only visible at late times of exposure (Fig. 5A). These permeabilized hyphae could be detected from 30 min, although most frequently they were seen after 1 h of AfpB treatment. Similarly, SYTOX green, which fluoresces upon binding to nucleic acids of hyphae with compromised plasma membrane, was only visible from 30 to 40 min after AfpB treatment (Fig. S3). In contrast, no dead hyphae were observed after 2 h of PAF treatment at the same concentration (Fig. 5A).
These results indicate that cell death mediated by AfpB is delayed with respect to ROS production, suggesting that ROS acts on cell death signaling rather than as a consequence of uncontrolled cell death.
We quantified the ROS production triggered by AfpB using a fluorimeter (Fig. 5B). ROS accumulation was detected as early as 15 min after AfpB treatment at different concentrations. ROS levels increased progressively over time, reaching values similar to those of hydrogen peroxide (H 2 O 2 ) at 5 mM, used as a control. Interestingly, ROS was also detected at low concentrations of AfpB (0.15 M), concentrations that are not fungal inhibitory, suggesting once again that ROS generation triggered by AfpB has a signaling role instead of a toxic effect.
AfpB induces regulated death in P. digitatum cells. The cleavage of linker regions between nucleosomes by nucleases is a hallmark of regulated cell death (RCD). To further investigate the cell death induced by AfpB in P. digitatum, we assessed DNA fragmentation in AfpB-treated hyphae compared to hyphae treated with H 2 O 2 , an inducer of the apoptosis phenotype in yeast and filamentous fungi when applied at minimum, but toxic, concentrations (36,37). The terminal deoxynucleotidyl transferasemediated dUTP nick-end labeling (TUNEL) assay is one of the most reliable methods for detecting DNA breaks. Fungal hyphae treated with AfpB showed positive TUNEL staining (Fig. 6), with hyphae treated with H 2 O 2 , used as a positive control of DNA fragmentation, giving a similar pattern of staining. The TUNEL-stained regions colocalized with the nuclei visualized by 4=,6-diamidino-2-phenylindole (DAPI) staining. These observations showed that AfpB induces DNA fragmentation in nuclei, which is indicative of RCD.
Further insights into the intracellular molecular events underlying AfpB-mediated fungal cell death were obtained by monitoring the expression of RCD-associated genes.
In these experiments, we used an inhibitory concentration of AfpB (0.3 M) that was slightly lower than the MIC (0.5 M), since the AfpB cytotoxic effects were too strong at the MIC, making it difficult to monitor gene expression. Among the genes associated with RCD, caspase-like genes are good candidates, as caspase-like activities are involved in cell death execution. Although fungi do not encode true caspases, fungal metacaspases with different contributions to cell death have been identified (38)(39)(40). We identified two metacaspases encoded in the genome of P. digitatum, CasA and CasB. Our gene expression analysis revealed that both metacaspases were induced in hyphae in response to AfpB treatment, even at a subinhibitory AfpB concentration (0.15 M), although at lower levels than those induced by H 2 O 2 treatment, used as an RCD control ( Fig. 7A and B). Similarly, we analyzed the expression of the fadA gene encoding the ␣-subunit of a heterotrimeric G-protein that mediates RCD in response to PAF, Neosartorya fischeri NAFP, or the antifungal protein osmotin from plant (23,(41)(42)(43). The expression of fadA was induced by AfpB treatment, even at higher levels than those caused by H 2 O 2 treatment (Fig. 7C). Analysis of the nma gene, the orthologue of the human HtrA2 proapoptotic gene, which also induces apoptotic markers in filamentous fungi (44), revealed low but constant induction by AfpB at the different time points at levels similar to those of our positive control (Fig. 7D). The apoptosis-inducing factor gene (aif1), as well as the aif-homologous mitochondrion-associated inducers of death gene (amid2) (45), components of the caspase-independent pathway of RCD in fungi, were also induced in response to AfpB treatment. However, the expression level of the genes was lower than that induced by H 2 O 2 and was statistically significant only at certain time points (Fig. 7E and F). Altogether, our gene expression results indicate that AfpB regulates cell death in P. digitatum through a complex intracellular process, and membrane permeabilization is the consequence, not the cause, of its antifungal activity.

DISCUSSION
In this study, we demonstrate the strong, rapid antifungal activity of AfpB against P. digitatum at concentrations lower than the micromolar level through multiple targets, including components of the cell wall, the plasma membrane, and intracellular ele- ments, which would hinder the development of fungal resistance. Interestingly, we found that the antifungal effect of AfpB occurs inside fungal cells, where it activates a transcriptional reprogramming to trigger cell death. This mechanism of regulated cell killing is less aggressive to hosts than the membranolytic actions of other antifungals, and this, together with its potency and specificity toward fungal cells, makes this protein an excellent candidate to be developed as an antimycotic in clinical therapies as well as a biofungicide for crop and postharvest protection.
The side-by-side study of the mode of action of AfpB compared with the wellcharacterized and less active PAF protein against the phytopathogen P. digitatum indicates that both proteins have a pathway of action that fits our previously proposed model for the antifungal action of short synthetic peptides (46,47). Following this model, AfpB acts through a three-stage process to lead to fungal cell death: first, the interaction with the cell wall, then the energy-dependent cell penetration, and, finally, a series of intracellular regulated actions that end with cell collapse (Fig. 8). This mechanism differs from the one elucidated for AFP from A. giganteus (4,21,24) and other antifungal proteins, such as plant defensins (48)(49)(50)(51), which act by disrupting plasma membrane integrity, causing cell disintegration and death. However, other AFPs have a cell-penetrating mode of action similar to that of AfpB, including PAF, and P. chrysogenum AfpB orthologs, PAFB, and PgAFP and NAFP from N. fischeri (16,23,43,52).
Interestingly, we found that AfpB interaction with the cell wall is important for its activity by facilitating access to the plasma membrane and subsequent entry into the fungal cell (Fig. 8). The fungal cell wall is composed of an outermost highly glycosylated layer of mannoproteins, a thick intermediate layer of ␤-glucans, and a thin layer of chitin next to the plasma membrane. Staining P. digitatum cell walls with CFW, which binds chitin, interfered with the internalization of AfpB, indicating that a fortified cell wall can act as a barrier impeding AfpB diffusion from reaching the fungal plasma membrane, or that AfpB requires chitin interaction, blocked by CFW. Varied susceptibilities to AfpB have been observed in P. digitatum chitin synthase mutants (53), which suggests that both alternatives are feasible. Similarly, the A. giganteus AFP binds chitin, and reinforcement of the cell wall confers tolerance to AFP (54). Additionally, we show that the Δpmt2 mutant deficient in protein mannosylation had increased tolerance to AfpB, in correlation with the absence of protein interaction and internalization. Therefore, the mannoproteins surrounding the fungal cell could be mediating the interaction of AfpB with the cell wall, concentrating and bringing it near the plasma membrane for its internalization and cytotoxic activity. These results agree with previous observations with other antifungal peptides and proteins, such as synthetic PAF26 or plant osmotin, where cell wall mannoproteins were required for their activity against fungal cells (30,32,55). Moreover, our results support the idea that cell wall integrity is necessary for the activity of AfpB, as previously reported for the plant defensin NaD1 and osmotin (51,56). Further study on the implication of cell wall mannoproteins in AfpB activity would help determine its spectrum of activity and the design of improved antifungal molecules.
AfpB and PAF interacted similarly and efficiently with the cell wall of P. digitatum, but AfpB is immediately internalized and exerts its cytotoxic effects inside the cell, whereas PAF remains attached to the extracellular envelope. These observations lead us to conclude that the capability of AfpB for rapid internalization is the determinant for its potent antifungal activity. We also show that AfpB internalization inside P. digitatum cells is an active process that requires energy, presumably via an endocytosislike mechanism (Fig. 8). The active uptake of PAF and other antifungal peptides and proteins into susceptible cells has been previously observed (28,29,57). In the same way, the orthologous protein of AfpB in P. chrysogenum, PAFB, is actively taken up by the parental fungus, leading to autolysis, unlike PAF, to which the producer fungus is resistant (16). Therefore, the cell toxicity of this group of proteins correlates with active internalization in target fungi, and higher fungicidal activities are associated with proteins with efficient cell-penetrating capabilities. Future research to identify the determinants on the AFPs for efficient uptake into fungal cells will help to design potent antifungal proteins.
Our studies showed that AfpB causes a series of intracellular effects that lead to fungal cell death (Fig. 8). One of the first events was the generation of ROS, which was probably activated even earlier than AfpB cellular uptake, because ROS was quickly detected. ROS production occurred prior to cell death and not as a consequence of cell damage, suggesting that it has a signaling role. ROS accumulation seems to initiate signaling pathways that activated a transcriptional reprogramming to trigger a cell death program in the fungal cell. In addition to ROS, AfpB activates other signaling messengers, including mitogen-activated protein kinase cascades previously reported by our group (58), and G-protein signaling pathways, as indicated by the transcriptional activation of the ␣-subunit of the G-protein FadA detected in the present study. These results agree with a general role of G-protein signaling in the cytotoxicity of cellpenetrating AFPs, with previous research showing that fadA mutants from Aspergillus nidulans had reduced susceptibility to PAF (23) or NAFP (43) and reduced accumulation of G-proteins by PgAFP in Aspergillus flavus (52). This group of AFPs is reported to induce some morphological changes associated with programmed cell death, such as DNA fragmentation and phosphatidylserine externalization (23,52). Similarly, AfpBinduced cell death was accompanied by a TUNEL-positive response, revealing DNA fragmentation associated with RCD. We went on to demonstrate that several regulated cell death markers are induced in response to AfpB in the fungal cells, including cell death execution methacaspases (casA and casB), proapoptotic genes (nma), and death inductive factors (aif1 and amid2). We detected their transcriptional gene activation in response to AfpB in a reproducible quantitative manner, indicating that AfpB triggers an RCD program in its target fungi. Eventually, AfpB causes loss of plasma membrane integrity in fungal cells as a consequence of cell damage, but this is not its primary target. This information is of fundamental value in the biotechnological application of AfpB for guiding combination therapies and for evaluating secondary effects.
More intriguing is the biological function of AfpB in P. digitatum, a protein that is highly active against the parental fungus. AFPs have commonly been recognized as specific against fungi but innocuous or less active against the producer fungus (15). However, the discovery of AfpB (5) and, later, of the orthologous protein from P. chrysogenum PAFB (16) and AfpA from P. expansum (6), is bringing to light that self-activity of AFPs could be a more frequent phenomenon than thought. The finding that AfpB elicits a genetically encoded program of self-elimination makes us think that its biological role goes beyond being a weapon to fight competitors to being a weapon to ensure the survival of the whole species by regulating fungal populations. In this sense, based on a transcriptomic meta-analysis, it has been proposed recently that the AFPs from Aspergillus niger and A. giganteus are cannibal toxins in fungi with a primary function of killing genetically identical siblings for the benefit of the fungal community (22). A better understanding of the biological role of AFPs and how the producing organisms protect themselves from autolysis will broaden the knowledge of these valued antifungal molecules with relevant therapeutic applications.

MATERIALS AND METHODS
Strains, media, and culture conditions. The wild-type P. digitatum PHI26 (CECT 20796) and the genetically engineered Pdpmt2 disruptant (Δpmt2) strains were used in this study (30,59). They were cultured on potato dextrose agar (PDA) plates at 25°C for conidium collection. After dilution to the appropriate concentration, conidia were statically incubated in 1/10 potato dextrose broth (PDB) medium supplemented with 0.03% chloramphenicol during the indicated periods for antifungal assays and confocal microscopy studies.
Protein production and purification. Fully active AfpB and PAF proteins were produced in P. digitatum and purified to homogeneity as previously described (6,19).
Fluorescent labeling of AfpB and PAF. Proteins were labeled via carboxyl groups with the fluorophore BODIPY-FL-EDA (Life Technologies), as previously described (51,60). Unbound BODIPY and salts were removed through dialysis using benzoylated cellulose tubes (2,000 molecular weight cutoff; Sigma) against deionized water. Protein concentration was determined by UV spectrometry.
Antifungal activity assays. Growth inhibition assays were performed in 96-well microtiter plates as previously described (5). In these assays, antifungal proteins were normally incubated with conidia for 72 h to determine their growth-inhibitory activity. When indicated, the conidia were pregerminated for 24 h prior to incubation with the antifungal proteins. Fungicidal activity was assessed on conidial suspensions (10 4 conidia/ml) in water by incubation with different protein concentrations per triplicate for 24 h at 25°C and low shaking. After treatment, samples were serially diluted and grown on PDA plates for

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. We thank Montse Amenòs for technical assistance with confocal microscopy.