The production of antimicrobial peptides or proteins, such as killer toxins, was first reported in Saccharomyces cerevisiae (Bevan and Makover, 1963), and has been widely characterised since. Three main killer toxins have been reported in S. cerevisiae: K1, K2, and K28 (Novotna et al., 2004; Orentaite et al., 2016). Yeast killer toxin producers exhibit antimicrobial properties against other S. cerevisiae sensitive strains and are immune to their own killer toxin family (K1, K2, or K28), but could be sensitive to another killer family (Orentaite et al., 2016).

A killer phenotype has been reported in NS yeasts found in grape must and other environments (Santos and Marquina, 2004; Liu et al., 2012) (Table 2). Amongst these yeasts, some of them were reported to be killer toxin producers with antimicrobial activity on the main wine spoilage yeasts, Brettanomyces and Hanseniaspora genera (Mannazzu et al., 2019).

Recently, the production of a killer toxin by a M. pulcherrima strain (TB26) isolated from grape vine was demonstrated (Büyüksırıt-Bedir and Kuleaşan, 2022). The application of the purified killer toxin was tested in ready to cook meatballs in order to extended their shelf life (Büyüksırıt Bedir and Kuleaşan, 2021). The same authors purified, characterised and tested different growth conditions for killer activity (Büyüksırıt-Bedir and Kuleaşan, 2022). Incubation at 20°C at a pH value of 7 showed the highest inhibition diameter on agar plates. An analysis by MALDI-TOF mass spectrometry gave a molecular weight of 10.3 kDa and provided the amino acid sequence of this toxin. The sequence comparison underlined that part of the sequence obtained (amino acids 31 to 50) showed a 100% correspondence score with the KHR killer toxin of S. cerevisiae characterised by Goto et al. (1990). Molecular weights of killer toxins are extremely variable. The size range reported in the literature can vary from 1.8 to 300 kDa (Kagiyama et al., 1988). The M. pulcherrima TB26 killer toxin seems to be in the low molecular weight range of the killer toxins characterised until now. These results were supported by Hicks et al. (2021) who tested the killer phenotypes of 11 M. pulcherrima strains. The authors showed that some strains exhibit a killer activity against Escherichia coli, Salmonella enterica, and Staphylococcus aureus on agar, an effect also achieved by adding the culture supernatant. They also analysed the protein extract of Metschnikowia culture and found a considerable amount of proteins with a size around 10 kDa, which corresponds to the killer toxin size identified by Büyüksırıt-Bedir and Kuleaşan (2022).

This approach remains to be explored with more strains of M. pulcherrima used for bio-protection in oenology. Farris et al. (1991) and Lopes and Sangorrín (2010) shed light on the killer phenotype (i.e., antagonist effect on agar plate by visualising a halo of inhibition) of M. pulcherrima strains isolated from grape, must, and the wine matrix. However, the killer phenomenon observed in those studies were not proved to be linked to the production of a peptidic killer toxin.

Microorganisms can produce non-peptidic compounds with antimicrobial properties. In the particular case of M. pulcherrima, the non-peptidic antimicrobial compound composed of four genes: PUL1, PUL2, PUL3 is pulcherriminic acid. The first reports in the literature on pulcherriminic acid production date from the last century and concern yeasts and Bacillus bacteria (Kluyver et al., 1953; MacDonald, 1965; Uffen and Canale-Parola, 1972). The metabolic pathway involved in pulcherriminic acid synthesis has been well characterised since, especially in Bacillus subtilis and Bacillus licheniformis. Pulcherriminic acid is produced from two leucyl-tRNA that are cyclised by cyclodileucine synthase, encoded by the yvmC gene, which leads to the production of cyclo-(L-Leu-L-Leu). The latter is then oxidised in pulcherriminic acids by pulcherriminic acid synthase, encoded by the cypX gene (coding for a P450 cytochrome family protein). After its production in the cell, pulcherriminic acid is excreted into the extracellular medium by transporters encoded by the yvmA gene, where it can chelate iron ion (Fe³⁺) by a non-enzymatic reaction to form the red pigment named pulcherrimin (Wang et al., 2018; Yuan et al., 2020). The production of pulcherriminic acid is down-regulated by three main genes in Bacillus: yvnA, yvmB and abrB. Moreover, Wang et al. (2018) showed that the iron concentration of the medium impacts pulcherriminic acid production. These authors have shown that an iron-limited medium inhibited the yvmC-cypX cluster and thus inhibited pulcherriminic acid production.

In M. pulcherrima, as well as in other pulcherrimin-producer Metschnikowia species, the metabolic pathway of pulcherriminic acid production is not yet as well-known as in Bacillus. Having prior knowledge of the genes involved in the synthesis of this pigment in Bacillus, Piombo et al. (2018) searched for homologous genes by sequence alignment in two newly sequenced M. fructicola strains. But the results were negative. Genes that may be involved in the production of this antimicrobial pigment in Metschnikowia were identified by Sipiczki (2020). The PUL (for PULcherrimin) gene cluster was identified as being involved in the production of pulcherriminic acid in Metschnikowia, Kluyveromyces, and Zygotorulaspora genera (Krause et al., 2018). This cluster is supposedly composed of four genes: PUL1, PUL2, PUL3 and PUL4 (Figure 3). Krause et al. (2018) were able to assign potential functions to each of these genes. The PUL1 gene is found to be potentially responsible for the production of cyclo-(L-Leu-L-Leu) from two leucyl-tRNA (corresponding to yvnB in Bacillus), followed by PUL2 acting to oxidise the cyclodipeptide into pulcherriminic acid (corresponding to cpyX in Bacillus). The PUL4 gene encodes a transcriptional regulator of PUL1, PUL2 and PUL3 genes. The PUL3 gene appears to be involved in the transport of iron in pulcherrimin, from outside to inside the cell for its reuse in cell metabolism. Studies conducted on Bacillus seem to indicate that iron once chelated by pulcherriminic acid becomes unavailable for the other microorganisms as well as for the bacteria producing pulcherrimin, unlike for yeasts producing this pigment which seem to possess a mechanism allowing them to reuse the iron in the medium even after complexation with pulcherriminic acid. The work of Krause et al. (2018) also showed the presence of genes homologous to PUL3 and PUL4 in other genera and species, such as S. cerevisiae, Lachancea thermotolerans, and M. bicuspidata. Some of them have both genes, and others have only PUL4. Species with only PUL4 are not able to reuse iron once chelated by pulcherriminic acid, unlike strains that have both genes (PUL3 and PUL4). Amongst them we find S. cerevisiae which has PUL3 and PUL4 and seems to be able to reuse iron despite the presence of pulcherriminic acid in the medium. It is hypothesised that PUL3 encodes the transporter that allows the internalisation of pulcherrimin and that PUL4 regulates the transcription of PUL3 by binding DNA (Krause et al., 2018).

The genes of the PUL cluster do not appear to be the only ones involved in the production of pulcherriminic acid. Using the PacBio method, Gore-Lloyd et al. (2019) sequenced a new strain of Metschnikowia, identified as a Metschnikowia aff. pulcherrima strain, as well as three spontaneous pigmentless mutants of this strain. Genome comparison between the wild-type strain and the pigmentless mutants showed a single mutation in the gene homologous to the SNF2 gene in S. cerevisiae. This mutation introduces a premature stop codon in the sequence, leading to the production of a truncated and therefore non-functional protein. In S. cerevisiae, SNF2 codes for a transcriptional regulator by chromatin remodelling (Snf2) (Hirschhorn et al., 1992). This mutation seems to impact the transcription of many genes including genes of the PUL cluster. Indeed, strains mutated on the SNF2 gene show a strong decrease of PUL cluster gene transcription. Since Snf2 is involved in chromatin remodelling, it is possible that the non-functional Snf2 prevents the remodelling of chromatin at the PUL cluster loci, and thus decreases the transcription of these genes (Gore-Lloyd et al., 2019).

Iron chelation by pulcherriminic acid makes this resource unavailable for microorganisms in the environment, and thus negatively impacts the growth of microorganisms needing Fe³⁺. The secretion of pulcherriminic acid into the medium by pulcherrimin producing yeasts, such as M. pulcherrima, could partially explain the antagonistic effect of this species on some microorganisms. The antagonistic effect of pulcherrimin was therefore studied with different species of yeasts producing this pigment. These studies were conducted in vitro (mostly with agar plates) and in vivo. Different species of Metschnikowia (citrensis, fructicola, pulcherrima) have shown an inhibitory effect on microorganisms that can alter the fruit or the resulting food products (Türkel and Ener, 2009; Hicks et al., 2021; Wang et al., 2021; Zhang et al., 2023). It was shown that the intensity of the pigmented halo around Metschnikowia on agar increases with the iron concentration in the medium, contrary to the inhibition zone which decreases when increasing the iron concentration. Indeed, in a medium with excess iron, Fe³⁺ chelation by pulcherriminic acid will not be sufficient to cause a deficiency in the environment, and thus not be deleterious for the surrounding microorganisms (Sipiczki, 2006; Saravanakumar et al., 2008; Türkel and Ener, 2009; Wang et al., 2021). Furthermore, Kregiel et al. (2022) showed that pulcherrimin has no antimicrobial effect. Their results support the hypothesis that it is iron chelation and not the pigment itself that has an inhibitory effect. In the oenological context, Oro et al. (2014) showed that the production of pulcherriminic acid by M. pulcherrima had an inhibitory effect on non-Saccharomyces yeasts found in oenology, including spoilage yeasts such as apiculate yeasts (Hanseniaspora guillermondii) and Brettanomyces bruxellensis.

Under the conditions tested, the production of pulcherriminic acid seems to explain most of the antagonistic effect of Metschnikowia observed. However, the production of this compound does not seem to be the only explanation for its antimicrobial effect. Indeed, in studies that sought to confirm the involvement of iron chelation in inhibitory effects, the authors supplemented the medium with iron or induced the knock-out or down-regulation of genes involved in pulcherriminic acid production. In most cases, the inhibitory effect is strongly reduced (by 80%) (Gore-Lloyd et al., 2019; Wang et al., 2021), but not always completely eliminated (Gore-Lloyd et al., 2019). This suggests that iron chelation is not the only negative interaction that M. pulcherrima establishes in playing its bioprotective role.

Quorum Sensing (QS) is intercellular communication mediated by the excretion of a density-dependent signal molecule. Once the target cell density is reached, the signal molecule concentration also reaches a threshold value, inducing signal transduction pathways that coordinate a response at the population level rather than at the cell level. This communication mechanism was first discovered in Vibrio fischeri for bioluminescence production (Nealson et al., 1970; Dunlap, 1999). In bacteria, QS has been found to be involved in different cellular mechanisms such as biofilm formation and the production of enzymes and other compounds required for pathogenicity or growth regulation (Miller and Bassler, 2001).

Bacteria are not the only microorganisms able to regulate their gene expression at the population scale to adapt to their environment. QS is also an interaction phenomenon investigated in yeasts, and more extensively in the pathogenic yeast Candida albicans, as it controls various cellular transformations such as biofilms, transitions between cellular growth and stationary phases, hyphal production, and many others (Kügler et al., 2000; Chen and Fink, 2006; Hogan, 2006; Sprague and Winans, 2006). In yeasts, it seems that QS is induced by aromatic alcohols in response to environmental factors such as a low nitrogen level in the medium. In the model yeast S. cerevisiae, the Quorum Sensing Molecules (QSMs) found are 2-phenylethanol and tryptophol. Tyrosol has been found in C. albicans but its role as a QSM in S. cerevisiae is still controversial (Chen and Fink, 2006; Wuster and Babu, 2009; Zupan et al., 2013; Avbelj et al., 2016; Padder et al., 2018; Mehmood et al., 2019; Jagtap et al., 2020). These molecules are produced from phenylalanine, tryptophan and tyrosine respectively, through the Ehrlich pathways. In order to produce these QSMs, amino acids must undergo three phases of transformation: transamination, decarboxylation and oxidation, involving the genes ARO8/ARO9, ADH, and ARO10, respectively (Sprague and Winans, 2006; Hazelwood et al., 2008; Wuster and Babu, 2009; Avbelj et al., 2015; Jagtap et al., 2020). Saccharomyces cerevisiae and C. albicans are not alone in being assumed to be capable of communicating through QS. Pu et al. (2014) showed that the bioprotective effect of H. uvarum on lemon through biofilm formation in fruit wounds was induced by phenylethanol secretion and thus under the potential control of QS. Furthermore, other yeasts and fungi are suspected of being able to use QS for their intercellular communication, mainly through aromatic alcohol production (Kügler et al., 2000; Sprague and Winans, 2006; Gori et al., 2011; Padder et al., 2018).

Except for C. albicans, this type of interaction remains less studied and understood in yeasts. The existence of QS as well as the role of aromatic alcohols as signalling molecules remains controversial and requires further research (Winters et al., 2019, 2022). Nevertheless, it has been shown that some non-Saccharomyces wine yeasts, including M. pulcherrima, are able to produce aromatic alcohols suspected of playing the role of QSMs in S. cerevisiae (González et al., 2018; Petitgonnet et al., 2019). It could therefore be interesting to extend research in M. pulcherrima to the involvement of these molecules as potential QSMs and to their control of gene expression amongst non-Saccharomyces yeasts mainly present on a grape must.

Amongst the NS yeasts, Metschnikowia pulcherrima stands out due to its substantial production of a very diverse range of enzymatic activities, such as amylase, lichenase, cellulase, lipase, and glucanase (Charoenchai et al., 1997; Strauss et al., 2001; Oztekin and Karbancioglu-Guler, 2021). Saravanakumar et al. (2008) showed that a M. pulcherrima strain isolated from an apple carposphere is able to secrete chitinase in the medium. This enzymatic production was found to be involved in its biocontrol activity on Botrytis cinerea. Other studies have also reported chitinase (Banani et al., 2015; Pretscher et al., 2018; Freimoser et al., 2019; Morata et al., 2019) and β-1,3-glucanase productions (Oztekin and Karbancioglu-Guler, 2021) by yeasts belonging to the Metschnikowia genus, justifying their potential use as biocontrol agents on vegetables and fruits, as well as grapevine. To our knowledge, enzyme production by M. pulcherrima and potential enzymatic activity levels have never been tested in relation to the acidity associated with oenological conditions, in order to verify the possible implication of extracellular enzymes synthesised by the yeast in a bio-protection context.

Oxygen plays a key role in the metabolism of NS yeasts. At the pre-fermentative stage, the level of dissolved oxygen (DO) in must is about 8 mg/L (at 20°C) and decreases during alcoholic fermentation. Depending on oenological practices, the concentration of DO added to the must can vary and practices such as punching down and pumping over to the vat incorporate significative quantities of oxygen into the medium (Moenne et al., 2014).

Many NS yeasts are high oxygen consumers (Visser et al., 1990). Quirós et al. (2014) studied the respiratory quotient (RQ) of S. cerevisiae and many NS yeasts. Amongst the yeasts studied, they showed that at the beginning of alcoholic fermentation when oxygen is available in the must, M. pulcherrima has an RQ of 1. This RQ means that all the sugar metabolised by the yeast is respired and not fermented, in contrast to S. cerevisiae which has an RQ of 4 (meaning that only about 10% of the sugar consumed is respired). The authors also showed that the RQ of M. pulcherrima could reach a maximum fermentative capacity of 2.6, which indicates that the yeast may consume at least 17% of the sugar by respiration.

Several studies have investigated the role of oxygen in the persistence and survival of yeasts in co-cultures with S. cerevisiae. Most of the time, NS yeasts do not persist in must, especially after inoculation with a S. cerevisiae strain. In the literature, the addition of oxygen has been found to improve the persistence of NS yeasts in the medium (Holm Hansen et al., 2001; Shekhawat et al., 2016, 2018; Englezos et al., 2018; Yan et al., 2020). Morales et al. (2015) and Shekhawat et al. (2016) showed that the decline of M. pulcherrima in co-culture with S. cerevisiae is delayed by increasing O₂ supply in white grape juice and synthetic must. Although these studies focused on oxygen requirements during co-cultures combining S. cerevisiae/non-Saccharomyces yeast, they nonetheless highlighted the importance of oxygen needs for NS yeasts. In order to better understand and control the bioprotective effect of M. pulcherrima, it appears essential to quantify its oxygen requirement, as well as those of the indigenous flora, such as Hanseniaspora yeasts, with the aim of optimising the implantation of the bioprotective strain in relation to its efficiency. High oxygen consumption exhibited by M. pulcherrima yeasts could induce competition between bioprotective and indigenous yeasts, leading to the inhibition of potential spoilage yeasts.

Oxygen is not the only resource consumed by yeast, which could lead to competitive phenomena. Amongst these nutrients, requirements for nitrogenous resources (ammonium and amino acids) are those studied most, playing a central role in yeast metabolism synthesising protein and nitrogenous bases, and in the production of aromatic compounds by the Ehrlich pathway (Hazelwood et al., 2008). As for oxygen, competition for nitrogen compounds could be involved in the negative interactions triggered by the bioprotective yeast.

Nitrogen sources are classified as “preferential” and “non-preferential” resources. Preferential resources are consumed in priority by the yeasts and support their growth more efficiently, contrary to non-preferential resources which are consumed once the preferred resources have been entirely consumed or have become limited in quantity in the medium (Crépin et al., 2012). The regulation of nitrogen consumption has been extensively studied in S. cerevisiae and appears to be regulated by two main systems: the Nitrogen Catabolic System (NCR) and regulation involving the Ssy1-Ptr3-Ssy5 (SPS) sensor of the plasma membrane (Crépin et al., 2012).

In NS yeasts, the regulatory pathways involved remain poorly investigated. Moreover, nitrogen requirements were found to be highly variable between yeast species but also strain-dependent (Kemsawasd et al., 2015a; Gobert et al., 2017; Prior et al., 2019; Roca-Mesa et al., 2020; Seguinot et al., 2020). The data heterogeneity can be explained by the great diversity of experimental conditions: variable temperatures, different strains analysed as well as different growth media (grape juice, synthetic must or synthetic medium). According to Gobert et al. (2017), the preferential sources of M. pulcherrima in grape juice at 28°C are Ile, Leu, Lys, Met, Glu and Cys, whilst at 20°C the preferential resources are Ala, Cys, Glu, His, Lys and Thr. In the synthetic medium YNBMAF at 25°C, the resources consumed are mainly Ala and Asn (Kemsawasd et al., 2015b). In synthetic must at 22°C Lys is found mainly consumed with Glu, Gln, His and Val in medium with amino acids and ammonium, or with Phe in medium with only amino acids as nitrogen sources (Roca-Mesa et al., 2020). Furthermore, Seguinot et al. (2020) showed that consumption also varies according to the initial nitrogen concentration, with an increase of the consumption of resources with the concentration of nitrogen in the environment.

To the best of our knowledge, there are still no studies that have attempted to demonstrate competition for nitrogen resources in a bio-protection strategy. Better understanding of the nitrogen requirements of M. pulcherrima in relation with the environmental conditions of growth is essential in order to investigate whether potential competition for nitrogen compounds may be involved in the negative effect on the growth of indigenous yeasts.

In must, other compounds could lead to competition, such as lipids or vitamins. Lipids are necessary for the maintenance of cell membrane integrity and they improve resistance to the stresses induced by ethanol. Moreover a medium limited in lipids can lead to languishing fermentation (Tesnière, 2019; Mbuyane et al., 2022). Vitamins are also an important factor for yeast growth, and a medium limited in vitamins could also lead to sluggish fermentation. Also, vitamin consumption appears to be species- and strain-dependent (Evers et al., 2021). The lipid and vitamin requirements of NS yeasts are still poorly explored and deserve further investigation to determine if they are involved in the bioprotective phenomenon.