Botrytis is an ascomycete fungal genus of plant pathogens. Most members of the genus are specialized species infecting a narrow range of monocotyledonous host plants. Typically, they are aggressive, necrotrophic pathogens (Staats et al., 2005). Some have an extended quiescent phase following infection (Botrytis allii). An exception to the rule of narrow host range is the clade including the species Botrytis cinerea sensu lato. This clade is by far the most economically damaging group within the genus; B. cinerea s.l. has a recorded host range including over 1400 (mostly dicotyledonous) hosts (Elad et al., 2015). The typical symptoms leading to economic loss are the occurrence of spreading, fast-growing necrotic lesions bearing abundant pigmented, hydrophobic, conidia.

Botrytis conidia are dispersed in windy conditions. Rainfall aids dispersal through the sharp motions of infected tissues resulting from raindrop impact. Conidia require high humidity to infect; infection tests in laboratory are typically performed with high concentrations of spore suspensions (10⁶/ml). The likelihood that conidia will produce a lesion is greatly increased if a nutrient source is available due to host cell damage, insect honeydew or exogenously supplied sugars (van den Heuvel, 1981; Dik, 1992).

Typically B. cinerea causes loss in crops by damage to the harvestable part of the crop, flowers, fruits, or leaves, or by girdling stems. The tools to manage B. cinerea in crops include (partial) host resistance, avoidance of damage allowing saprophytic infections to occur and then spread, environmental modification to reduce the probability of conditions suitable for spore germination and dispersal, and the use of fungicide. The latter can be effective, but typically there are no consistent critical periods. Although in a single experiment, one or a few fungicide application times may stand out as effective, these application times may not be reproducible in repeat experiments or easy to relate to environmental conditions (e.g., McQuilken and Thomson, 2008). Spores are present in low concentrations in the air over most of the year (Hausbeck and Moorman, 1996; Kerssies et al., 1997; Boff et al., 2001; Boulard et al., 2008). A common strategy for application of fungicides is to maintain continuous cover. This is undesirable on environmental and economic grounds and leads to rapid development of resistance to new fungicides.

A number of publications have suggested that a range of plant species may harbor infections by B. cinerea s.l. and other species, including Botrytis deweyae, which cause no visible symptoms on the plant at the initial time of infection, are long-lived, can be isolated from newly grown host tissues, but cause necrotic lesions as the plant moves into a reproductive phase. Cultivated hosts in which this form of infection has been studied include Hemerocallis (Grant-Downton et al., 2014), hybrid commercial Primula (Barnes and Shaw, 2002, 2003), and lettuce (Sowley et al., 2010). In wild-growing plants, Rajaguru and Shaw (2010) found widespread infection in leaves of Taraxacum and, to a lesser extent, wild Primula vulgaris. A sampling for endophytes in Centaurea stoebe revealed the frequent occurrence of Botrytis spp., including isolates not derived from known species (Shipunov et al., 2008).

It is becoming increasingly clear that this situation is quite common. A simple distinction between pathogens and non-pathogens may not be possible for many fungal species, as several pathogens have extended cryptic phases (Stergiopoulos and Gordon, 2014). The most obvious examples, and most similar to the Botrytis case, are seedborne smut fungi (Schafer et al., 2010), with a cryptic phase from seed until flower maturation. In a more complex example, isolates of Fusarium oxysporum can cause devastating disease, with individual fungal genotypes often displaying a very narrow host range or be a benign root endophyte (Gordon and Martyn, 1997; Demers et al., 2015), wherein both endophytic and pathogenic behavior appear to be polyphyletic (Gordon and Martyn, 1997).

The discovery of distributed, symptomless infection by Botrytis sp. in a range of hosts suggests that a proportion of inoculum might arise from these sources, and that the fungus may exist for an important part as an endophyte or intimate phyllosphere inhabitant, with conidia only being produced when the host approaches the end of its life-cycle. Understanding this cryptic infection is important for controlling the disease in hosts in which it happens, and for managing disease in other hosts where the cryptic form may serve as an unexpected source of inoculum. In this paper we draw together data on plant species that act as hosts to symptomless distributed infections of Botrytis; how the infection varies over time and location; whether particular fungal clonal lineages are adapted to particular hosts; the fungal species involved; the effect of the infection on the host. The aim of experiments collated here was to explore an unexpected mode of infection by a fungus often considered to be a “model necrotroph.”

Specialist Botrytis species such as B. aclada and B. allii have been understood to have a life-cycle involving extended periods of symptomless growth (Maude and Presly, 1977; du Toit et al., 2004). Seed infection by species in the B. cinerea complex leading to seedling infection has been well-known for some time in species such as linseed (Harold et al., 1997) because of its contribution to seedling death. It has been less appreciated that long-lived cryptic infection with generalist Botrytis is possible. This has previously been shown in cultivated primula (Barnes and Shaw, 2003) and in lettuce (Sowley et al., 2010). Quiescent infections of a few host cells which give rise to spreading necrosis at fruit maturity are characteristic of many fruit diseases (e.g., Prusky and Lichter, 2007; Puhl and Treutter, 2008). The mode of growth discussed here is different, as it is a form of infection in which the fungus is able to grow with the host and spread to new plant parts. The results above show that many host species growing in natural settings and showing no disease symptoms frequently have disseminated infections of Botrytis species.

This is of both scientific and practical interest. We become increasingly aware that plants harbor many endophytic species (e.g., Shipunov et al., 2008) and that these may have profound effects on their susceptibility to stress and to pathogens (Rodriguez et al., 2009). Understanding that a pathogen like Botrytis, which is investigated mostly in the context of rapidly spreading necrotic infections may also exist as a widespread symptomless infection with minimal effects on the host should seriously alter our understanding of pathogenesis. There are quantitative differences between isolates in their pathogenicity on different species, but we did not obtain any evidence for isolates taken from a particular host to be particularly adapted to that host. Furthermore, both B. cinerea, B. pseudocinerea, and other lineages or cryptic species of Botrytis appeared to occur in this asymptomatic form. From a practical point of view, management of Botrytis in species where cryptic infection is common needs to focus on environmental conditions; it may be ineffective to try to prevent low levels of infection, since permanent cover by systemic fungicide would be needed. In particular, results in lettuce suggest a balanced system in which host defenses are activated sufficiently to prevent infection increasing above a certain density, because plants grown from “clean” seed acquired the same level of infection as plants grown from the infection-carrying seed infection. The question whether cryptic infection is harmful or beneficial to a plant is difficult to address because of the difficulty of growing plants free from Botrytis infection but also in a natural setting.

Because the findings reported may be deemed somewhat surprising, it is important to consider possible artifacts. The most obvious is that these isolations could represent independent, spatially restricted early stages or quiescent infections from environmental conidia. There are several arguments against an important contribution of these artifacts to the observations summarized here. (1) Barnes and Shaw (2003) showed that the sterilization procedure was highly effective in killing adherent spores on seed and other tests with inoculated material have shown the procedure to be effective on leaf tissue. (2) T. officinale and B. perennis grow in the same habitat and often in close proximity but have very different prevalence rates, showing that we are dealing with at least an established association between fungus and host. (3) In greenhouse work, Barnes and Shaw (2003) showed that the same isolate was repeatedly recovered from single hosts, but that adjacent plants rarely hosted identical isolates. This is weak evidence by itself, but supports the hypothesis that this type of infection is at least partially systemic, because it predicts that recovery of isolates or genotypes will be clustered on particular plants. However, clustering can also be ascribed to variation in susceptibility among plants.

Phylogenetic analysis of a subset of isolates sampled from dandelion in Netherlands showed that B. cinerea and B. pseudocinerea were the predominant species. The latter is a recently described species which is morphologically very similar but phylogenetically distinct from B. cinerea (Walker et al., 2011). Based on the morphology of the isolates sampled from a spectrum of plants in the UK, it may be assumed that these also mostly represent B. cinerea and B. pseudocinerea, although this would need extensive sequence analysis to be confirmed. The phylogenetic analysis also identified two isolates of distinct Botrytis species belonging to clade 2 (Staats et al., 2005), one of which (DAN39) is presumably from B. mali, a postharvest pathogen of apple with an as yet poorly explored distribution (O'Gorman et al., 2008). It is perhaps relevant to note that this isolate was sampled from a dandelion plant growing in a grass patch adjacent to an apple and pear orchard of an organic farm. The second isolate from Botrytis clade 2, DAN5, was related but not identical to B. caroliniana and B. fabiopsis, and might represent a novel Botrytis species. An earlier study by Shipunov et al. (2008) on endophytic fungi in Centaurea stoebe identified, besides B. cinerea, six distinct phylogenetic groups of Botrytis isolates. Four of these groups were closely related to B. cinerea and one of them might represent B. pseudocinerea (which had not yet been described at the time of the study by Shipunov et al., 2008). The isolates bot079 and bot378 reported in Shipunov et al. (2008), however, were very distant from the clade containing B. cinerea and grouped with B. paeoniae, alike the B. mali isolate DAN39. Unfortunately the isolates from C. stoebe are no longer available for further phylogenetic comparison (A. Shipunov, personal communication). The above results and those of Shipunov et al. (2008) demonstrate that several new Botrytis species can be retrieved as endophytes from a single host in a limited sampling. A more extensive phylogenetic analysis of Botrytis isolates from multiple hosts and locations would likely reveal many more novel species. If more extensive sampling and further phylogenetic analysis would indeed confirm a greater diversity of Botrytis species than previously appreciated (Staats et al., 2005; Hyde et al., 2014), it raises the question as to how many times the cryptic life-style has arisen.

As with all endophytic organisms, the question arises as to how the host defense system and the pathogen are interacting (Schulz and Boyle, 2005; Newton et al., 2010). There are a number of possibilities. First, the fungus may evade the host defense system by not producing signals which activate defenses strongly enough to destroy the fungus. This is consistent with the increased infection frequency in unshaded lettuce plants in experiment 6: rapidly photosynthesizing plants are likely to have more nutrients in the apoplast, and thus allow more pronounced growth if they do not also have better defense. It is likely that Botrytis produces little or no phytotoxic metabolites and proteins, or plant cell wall degrading enzymes, while it displays symptomless internal infection behavior. If it did produce such compounds, as Botrytis “normally” does (reviewed by van Kan, 2006), this would result in irreversible damage to plant cells and culminate in disease symptoms. It remains elusive why Botrytis does not produce such damaging compounds during cryptic infection. Is it by lack of a (plant-derived) signal to activate the expression of the genes, or is the plant actively suppressing the expression of virulence genes?

An alternative view of the interaction is that host defense continually destroys Botrytis locally, but establishment in new areas is frequent enough to maintain the fungus associated with the plant. In experiment 6, spray inoculation with B. cinerea spores incurred a 20% reduction in growth, without altering infection density. This reduction in growth might reflect the cost of effective defense against a forceful attack; this defense evidently does not alter the balance between systemic infection and host defense. The similarity in frequency of cryptic infection in plants which did and did not have seed-borne infection removed by fungicide in experiment 5 again points to a balanced system in which too much local growth can be eliminated by the host. The hypothesis of a balanced growth/elimination system implies that previously infested plant tissues might become uninfested, as well as the reverse. We have no data on this. Experiments to address this question would be complex because of the destructive nature of current methods to detect infection.

The spore density at the point of first encounter is likely a crucial determinant in whether the interaction becomes asymptomatic or necrotic. At high spore density, and in the presence of nutrients, the fungus can rapidly and efficiently germinate (van den Heuvel, 1981) and produce a spectrum of phytotoxic metabolites and proteins at sufficient concentrations to cause death of multiple plant cells (van Kan, 2006) and set necrotic development in motion. Causing host cell death is in essence a physiologically irreversible decision, although there are cases where primary necrotic lesions fail to develop into an expanding lesion, possibly due to an effective restriction of the fungus by the host (Benito et al., 1998; Coertze and Holz, 2002). In order to achieve symptomless Botrytis infection in plants in our experiments, it was important to start the infection with a low inoculum dosage. The low number of spores and concomitant low density of germlings on the plant surface might not release sufficient phytotoxic metabolites and proteins to activate the plant cell death machinery, and thereby fail to kill cells that would act as an entry point for the fungus. Failure to induce cell death would force the fungus to enter the plant through natural openings and grow in the plant interior where it can retrieve nutrients. The growth of the fungus inside the plant would need to remain restricted and be synchronized with that of the host, possibly by mechanisms similar to those reported for Epichloë endophytes(Tanaka et al., 2012).

A final possible view of the interaction in this system is that the infection is in some circumstances beneficial to the host plant, so that specific signals from Botrytis lead to over-riding of normal defense signals. A hint that this might be so is given by the positive relation between lettuce growth and the frequency with which Botrytis could be recovered from the plants in experiment 7, presumably representing the density of infection. The nature of such signaling processes can only be conjectured at present. Recent studies by Weiberg et al. (2013) described how small RNAs of Botrytis cinerea can actively silence host genes that are involved in immunity, and thereby promote virulence of the pathogen. Conversely, it is conceivable that small RNAs from a host plant could silence Botrytis genes that are required for the activation of virulence mechanisms, and so prevent the production of damaging fungal proteins and metabolites. In the fungal grass endophyte Epichloë, the symbiotic interaction depends on an intact reactive oxygen signaling and MAP kinase signaling. The disruption of components of either of these signaling processes results in breakdown of symbiosis and switch to disease development (Tanaka et al., 2012).

The economic importance of this form of infection could be considerable in a susceptible crop such as lettuce, cyclamen or Primula × polyantha in which persistent infections last for the lifetime of the crop, causing disease and loss at maturity or after harvest. The importance of the demonstration of a large reservoir of infested vegetation in the wider environment depends mainly on how much inoculum it supplies. If this is sufficiently low, there could be no economic implications. On the other hand, given the number of plant species shown to harbor infection, infections in natural vegetation could be an important source of conidia into otherwise clean protected environments. It is less clear what could be done about it, but it suggests that effective integrated management of Botrytis should assume that inoculum is present not only in the air but probably latent in the crops. The best management strategies will focus on the physiological state of the plant and the intrinsic resistance or tolerance of the host to necrotic disease development.

If we can understand the mechanism that determines whether the interaction becomes symptomless or necrotic, it might be possible to steer the interaction in one or other direction, by modifying the environmental conditions, or by applying chemicals that trigger a switch in lifestyle. This might result in active suppression of necrotic development of Botrytis, or in the forced interruption of the symptomless infection.

MS and JvK coordinated the sampling, designed and supervised the experiments, and wrote the manuscript. CE, DE, RT, AS, DT, and ME performed the experiments. CE prepared Figure 1, RT prepared Figure 2, DE prepared Figure 4.