*Perspective* **Breeding Tools for Assessing and Improving Resistance and Limiting Mycotoxin Production by** *Fusarium graminearum* **in Wheat**

**Sandiswa Figlan 1,\* and Learnmore Mwadzingeni <sup>2</sup>**


**Abstract:** The recently adopted conservation and minimum tillage practices in wheat-production systems coupled with the concomitant warming of the Earth are believed to have caused the upsurges in Fusarium head blight (FHB) prevalence in major wheat-producing regions of the world. Measures to counter this effect include breeding for resistance to both initial infection of wheat and spread of the disease. Cases of mycotoxicosis caused by ingestion of wheat by-products contaminated with FHB mycotoxins have necessitated the need for resistant wheat cultivars that can limit mycotoxin production by the dominant causal pathogen, *Fusarium graminearum*. This manuscript reviews breeding tools for assessing and improving resistance as well as limiting mycotoxin contamination in wheat to reflect on the current state of affairs. Combining these aspects in wheat research and development promotes sustainable quality grain production and safeguards human and livestock health from mycotoxicosis.

**Keywords:** contamination; health; infection; molecular techniques; selection

#### **1. Introduction**

Breeding wheat for Fusarium head blight (FHB) resistance involves systematic genetic manipulation of the crop to incorporate superior biochemical and morpho-physiological traits that safeguard it against the damaging effects of the dominant causal species, *Fusarium graminearum.* Infection of crops by *F. graminearum* does not only reduce yield, but also exposes the grain to contamination by mycotoxins. Mycotoxin contamination in grain crops intended for processing food, feed and beverages often results in the accumulation of these toxic fungal metabolites in foodstuffs, causing health hazards to both human beings and livestock. *F. graminearum* species complex infects grain crops including wheat, barley and maize. Breeding for resistance against FHB aims to reduce the impact of the pathogen on crop yield as well as mycotoxin contamination in infected grain. Various strategies for breeding against Fusarium head blight have been embarked on because resistance against the disease is multigenic and is further confounded by the large influence of genotype by environment interactions [1,2]. Resistance against FHB is conferred by more than 250 quantitative trait loci (QTL) distributed across the entire chromosome cascade of the wheat genome [3–5]. To effectively compart the negative effects of the disease, strong background knowledge is needed on various aspects including the importance of FHB as a grain disease, mycotoxin contamination of infected grain, breeding strategies to reduce mycotoxin contamination in grain as well as the tools used to assess and limit mycotoxin contamination during breeding, selection and the entire wheat value chain.

Fusarium head blight, also known as 'scab', is a wheat disease that is mainly caused by the fungal complex called *F. graminearum* Schwabe (teleomorph *Gibberella zeae* Schwein.

**Citation:** Figlan, S.; Mwadzingeni, L. Breeding Tools for Assessing and Improving Resistance and Limiting Mycotoxin Production by *Fusarium graminearum* in Wheat. *Plants* **2022**, *11*, 1933. https://doi.org/10.3390/ plants11151933

Academic Editor: Alessandro Vitale

Received: 30 May 2022 Accepted: 20 July 2022 Published: 26 July 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Petch). It is one of the most common diseases affecting bread wheat (*Triticum aestivum* L.) around the world. Epidemics of FHB occur in cycles of four or five years worldwide [6] and in shorter periods under favorable conditions, particularly where no-till or minimum tillage practices, high humidity and/or high temperature coincide with early flowering to the soft dough stages of susceptible wheat cultivars. In addition to the enormous grain yield losses, *F. graminearum* infection is associated with the accumulation of mycotoxins that put the health of human beings and livestock consuming infected grain at risk [7]. Ingestion of huge amounts of mycotoxin contaminated grain may lead to mycotoxicosis, which under severe circumstances may cause death. It is important to note that there is very high genetic variability of *F. graminearum* species, which results in high resilience and complicates efforts towards breeding for FHB resistance due to genotype by isolate and isolate by environment interactions [8].

*F. graminearum* produces two groups of toxins, namely zearalenone and trichothecenes. Zearalenone (previously referred to as F-2 toxin) is one of the most prevalent estrogenic mycotoxins produced through the polyketide pathway. This mycotoxin is denoted as 6-[10 hydroxy-6-oxo-*trans*-1-undecenyl]-B-resorcyclic acid lactone. Zearalenone derives its name from *Gibberella zea* and, resorcyclic acid lactone because of the C-10 to C-2 'double bonds. The '-one' denotes its ketone group [9]. The toxicity of zearalenone is through binding to estrogen receptors ending up in estrogenicity, occasionally causing hyperestrogenism in livestock and human beings, especially women. Eventually, the toxicity of zearalenone may lead to myelofibrosis, reproductive system disorders, cancers, skeletal malformations and weakening [10], nervous disorders [11] and various other physiological malfunctions. Trichothecenes, on the other hand, are chemically tricyclic sesquiterpenes, which have double bonds at the C-9, 10 position and a C-12, 13 epoxy functional group. The most common contaminants of cereals are type-A and type-B trichothecenes [12,13]. Type-A trichothecenes are different from type-B by the absence of a carbonyl group at C-8 and hydroxylation at C-7. Type-A mycotoxins include diacetoxyscirpenol, T-2 and HT-2 toxins while type-B trichothecenes include fusarenone-X, nivalenol and deoxynivalenol. The effects of trichothecene ingestion through contaminated foodstuffs by animals and human beings include diarrhea, vomiting and death when the toxicosis is severe.

Various national and multi-national organisations have drafted guidelines on food safety to ensure that consumers are safe from the risks of eating contaminated food. The regulating bodies include the European Food Safety Authority, Codex Alimentarius and the USA Food and Drug Administration. Realizing that it is not possible to produce mycotoxinfree wheat grain, the regulating bodies have set threshold limits, which are practically attainable to reduce the incidence of mycotoxins in wheat products and other foodstuffs. The threshold regulations mainly protect the health of animals and human beings from the dangers caused by mycotoxins. Contamination of wheat with mycotoxins occurs during infection by mycotoxin, producing fungi such as *F. graminearum*, and further toxin accumulation may occur postharvest during grain storage [14–17]. Various interventions are necessary to limit mycotoxin contamination of the wheat grain by *F. graminearum* mycotoxins.

FHB can be managed using various strategies including cultural, biological and chemical control methods as well as breeding for resistance against the disease. Wheat production has thrived for ages through selection for superior traits and painstaking efforts to incorporate disease resistance. With increased efforts to incorporate FHB resistance into wheat, disease incidence and the spread of infection decrease, resulting in a subsequent reduction in mycotoxin contamination. Moreover, resistance may be specific to reduce mycotoxin production by the infecting *F. graminearum*. Various breeding strategies are being embarked upon to ensure minimal mycotoxin contamination of wheat grain. It is also important to develop laboratory tools to assist the selection of wheat varieties that suppress mycotoxin production as well as to ensure compliance with wheat grain safety standards. This review discusses these aspects beginning with various breeding strategies employed against FHB. Emphasis has been put on traditional breeding strategies, new techniques of resistance breeding and tools for monitoring mycotoxin levels in the harvested wheat grain.

#### **2. Resistance against Fusarium Head Blight in Wheat**

Resistance to FHB is categorised into various types of which the most prominent ones are type I and type II [18]. Type I refers to the resistance against initial infection and is exhibited by the ability of the cultivar to create a barrier to initial entry of the pathogen into the plant. On the other hand, Type II resistance is resistance to the spread of the pathogen after it has gained entry into the plant. The later type of resistance is more stable. Type I and II resistance can be tested under both field and artificial environments [19]. Usually, screening for resistance against FHB takes place in the advanced generations like F<sup>4</sup> onwards [20]. Select breeding lines are chosen and are artificially inoculated with the pathogen isolate(s)/races(s) to screen for resistance [21]. Assessment of resistance to FHB is done through generally visualizing discolouration of the spikes and by precisely assessing the intensity and number of affected grains. Affected grain may have a pinkish discolouration, sometimes with a chalky appearance. Assessment covers both the proportion of kernels that are diseased and the level of mycotoxins in the affected grain [22]. Resistance against mycotoxin accumulation is called type III resistance, which requires special tools for assessment, unlike type I and II which can be assessed visually. Both type I and type II resistance have indirect effects on toxin accumulation, but resistance to toxin accumulation, type III resistance, still has to be a targeted breeding objective on its own. Generally, genotypes to be used as donors of resistance in FHB breeding programmes and ultimate varieties must (1) resist initial infection (type I), (2) limit the pathogen spread in infected spikes (type II), (3) reduce mycotoxin accumulation in the grain (type III)), (4) resist kernel damage (type IV) and (5) tolerate the presence of the disease without much yield penalty (type V) [19]. Knowledge of the genetic basis underlying these observable types of resistance is slowly being demystified through advanced biotechnology and genetics.

#### **3. Breeding Focus against Fusarium Head Blight**

With the development of settlements for human beings and crop domestication, early farmers selected plants that had desirable traits and the resulting gene pool formed the basis of today's domesticated crops. Natural selection for superior agronomic traits was accelerated by the active mating and selection of offspring with desirable traits. Crops progressively improved, hence, huge monoculture practices were established to what has become modern agriculture. Wheat is one of the crops that has been extensively bred over the years leading, notably, to the Green Revolution of the 1960s. After a prolonged period of painstaking breeding efforts, Dr. Norman Borlaug, the Father of the Green Revolution, developed high yielding wheat varieties in India and Pakistan, a move that averted massive hunger. Despite this milestone, various diseases continue to threaten the crop, particularly wheat rusts and Fusarium head blight. Breeding for disease resistance continued to protect yields of high yielding varieties, among other control strategies. The wheat disease resistance breeding strategy at the International Centre for Maize and Wheat Improvement (CIMMYT) systematically grouped breeding needs of various regions in the world into mega-environments [23]. Breeding for resistance against FHB falls within the needs of mega-environment 2, which is characterized by high rainfall. China has been a significant source of resistance to FHB and hundreds of wheat lines carrying resistance have been shared with CIMMYT. Among the Chinese lines that carry FHB resistance are Sumai#3, Shanghai#5, Suzhoe#6, Yangmai#6, Wuhan#3 Ning 7840, and Chuanmai 18, which have been developed using traditional breeding methods. Genes for resistance against FHB are mostly additive, requiring a meticulous programme for resistance incorporation and selection [24].

Genetic variation for FHB resistance breeding is large. Therefore, there is a wide pool of sources of resistance. This makes it easy for resistance to be incorporated into wheat with options from exotic and native sources. However, Asian sources of resistance against FHB such as the Chinese spring wheat, Sumai#3, are prominently used worldwide. Resistance to FHB is mostly additive, being controlled by the effects of multiple genes. Quantitative trait loci controlling FHB across all 21 bread wheat chromosomes have been mapped and

identified, with just a few validated and used in breeding [4,5,25]. These QTL are prevalent in Chinese genotypes derived from Sumai#3 and they contain *Fhb1*, *Fhb2*, as well as *Qfhs.ifa-5A* [26–32]. Nevertheless, other resistance QTL do exist outside of Sumai#3. The presence of *Fhb1* (Sumai#3) and *Qfhs.nau-2DL* (breeding line CJ9306), which confer resistance to both type II and type III resistance, are of particular interest. *Fhb1* improves the detoxification of deoxynivalenol (DON) to DON-3-glucoside [33]. *Qfhs.ifa-5A* confers type III resistance by suppressing mycotoxin accumulation. Although resistance to FHB acquired from sources such as Sumai#3 has been useful, its use has been moderate and therefore new sources of resistance are desperately needed, especially resistance to curb toxin accumulation in wheat infected with *F. graminearum*. The current shortfalls in breeding for resistance against FHB therefore require radical use of new technologies. These technologies will help to improve wheat productivity to meet the needs of the growing global population.

Wheat breeding programs against FHB also aim to reduce mycotoxin production by the infecting fungus *F. graminearum*. From a food safety concern, this is an important breeding objective to ensure that harvested grain is strictly below the mycotoxin threshold level. To breed for resistance against FHB, a reliable inoculation method is needed. This allows repeatable assessment of resistance to ensure selection of resistant lines under high and uniform disease pressure. It is also important to use a cocktail of isolates/races for inoculation to ensure selection for broad-spectrum or multi-race resistance, preferably using races prevalent in the area where the resistant cultivars will be released. Isolates that produce higher levels of DON, a type-B trichothecene, are found to be more aggressive and could be useful for effective selection for type III resistance [34–39]. Resistance of wheat to DON accumulation is acquired through the ability of the plant to degrade the mycotoxin, for example, the possession of a putative deoxynivalenol-glycosyl transferase that detoxifies DON [33,40]. Newer strategies for resistance breeding have been adopted over the years and progress has been made ever since the adoption of these technologies. Breeding programs that aim to limit DON production by *F. graminearum* in wheat have greatly benefited from these new technologies.

#### **4. Traditional Crop Breeding against Fusarium Head Blight**

Conventional breeding is a systematic hybridization and selection strategy aimed to release superior genotypes. In certain instances, the trait of interest is transferred from a wild relative of the crop to be improved and this is termed wide crossing. Breeding for disease resistance often takes a different strategy from conventional breeding for complex agronomic traits such as yield. There has to be a source of resistance, which donates the resistance gene/genes to the recipient genotype containing most of the desirable agronomic traits, except for the resistant gene(s) of interest. In such a scenario, backcross breeding, which is the most prominent classical breeding technique against plant diseases, is used to recover most of the recipient genotype's genome. In certain instances, the resistance incorporated into a cultivar against FHB may be race-specific, though in most cases it is race non-specific. It is always important to adopt a clear resistance breeding strategy so that broad-spectrum and durable resistance may be incorporated into the cultivar. When using traditional breeding techniques, it is critical to select effectively in the early generations for FHB resistance; otherwise the promising gene combinations are lost irretrievably [41]. Thus, the selection efficiency increases when the breeding method can be used to select successfully in the early generations of selection [41]. Following the vast research investments that were put towards FHB resistance, backcross breeding is no longer sorely classical but is now fused with various molecular marker techniques for effective and timely selection as well as gene and QTL introgression.

#### **5. Molecular Breeding Techniques**

The use of resistant cultivars remains a valuable tool for the control of FHB. It therefore remains imperative to intensify breeding efforts and optimize breeding and selection strategies for resistance against FHB and mycotoxin production. The development and

improvement, in recent years, of molecular techniques like real-time polymerase chain reaction (PCR), marker-assisted selection, marker-assisted QTL backcrossing, next generation sequencing technologies and genetic engineering, are boosting research on FHB resistance and its associated mycotoxicosis. Screening for resistance against FHB usually takes place in advanced generations like F<sup>4</sup> onwards when select breeding lines are chosen and artificially inoculated with the pathogen to screen for resistance [42]. This task is very laborious and requires time for completion. In this case, advanced molecular techniques are required to monitor levels of inoculation, to select for resistance in genotypes to be used as parents in breeding for resistance to FHB and to introgress resistance genes into elite genotypes. These molecular tools are therefore useful in wheat pre-breeding and breeding against FHB.

#### *5.1. RNA Interference to Reduce Mycotoxin Contamination in Fusarium graminearum Infected Wheat*

The discovery of more sophisticated biotechnological approaches such as ribonucleic acid (RNA) interference (RNAi) offers new transformation opportunities to enhance resistance against *F. graminearum* and other invading wheat pathogens [43]. This is achieved through induced silencing of target virulent genes. RNA interference is an essential cellular system involved in gene regulation and protection of eukaryotes against infection by viruses [44]. It is an important systematic mechanism that can be employed to fight mycotoxigenic plant pathogenic fungi like *F. graminearum*. RNAi post-transcriptionally converts double stranded RNA molecules into short-stranded RNA duplexes of about 21 to 28 nucleotides often termed short interfering RNAs (siRNAs), which then cleaves to complimentary mRNA, effecting gene silencing or regulation [45–48]. RNA interference pathways are often triggered by the presence of viral RNAs providing gene regulated defense against specific RNA viruses. In this case, the mechanism will be termed virusinduced gene silencing (VIGS), whose success is highly dependent on designing effective vectors that will produce complementary siRNA species, efficient uptake of siRNAs by the fungus and amplification of the silencing effect within the target organism [43]. Silencing of target genes has recently been proved to be effective against plant pathogenic fungi [49] and has been demonstrated on *Puccinia* in wheat among other crop species and their respective fungal pathogens. Machado et al. [50] reviewed the recent advances in RNAi-mediated FHB control and suppression of mycotoxin contamination in a number of cereals. This involves the use of the barley stripe mosaic virus (BSMV) vector. *P. striiformis* genes were also observed to be silenced using the host-induced RNA interference mechanism [51]. In a more recent study, Cheng et al. [52] reported that wheat resistance against pathogenic fungi can be improved through RNAi sequences originating from chitin synthase (Chs) 3b gene originating from *F. graminearum.* These sequences are used for host-induced silencing of the chitin synthase gene in plant pathogenic fungi. This is one of the techniques that holds future promise for the incorporation of resistance against *F. graminearum* in wheat.

#### *5.2. Gene Transfer in General and Specifically against Fusarium Head Blight*

Gene transfer technologies that insert foreign genes in plants are another molecular breeding strategy with potential to enhance wheat resistance to FHB [53]. These technologies include particle bombardment or biolistic transformation and *Agrobacterium*-mediated genetic transformation [54]. The former bombards deoxyribonucleic acid (DNA)-coated gold or tungsten micro-projectiles into the target crop's genome using a particle gun, thereby inserting foreign genes. The later technique uses *A. tumefaciens* as a vector that copies and transfers the transfer DNA (T-DNA) molecules on a tumour-inducing (Ti) plasmid into the nucleus of target plant cells, thereby incorporating foreign DNA that is eventually inserted and becomes part of the plant genome. *Agrobacterium* transformation, however, works effectively with selected plant species, and inserts mostly three genes, including two T-DNA molecules and a selectable marker per transformation construct [55]. Biolistic transformation non-randomly targets AT-rich regions with matrix attachment region (MAR) motifs that are nuclear matrix prone eukaryotic DNA elements [56,57]. The

MARs create open chromatin, allowing the host plant genome to be accessible to transgenes. An advantage shared by both *Agrobacterium* transformation and biolistic transformation is that they can integrate two trans-genes into the target host genome [58].

The *Agrobacterium*-mediated transformation stages involve initiation, which includes identification, isolation and insertion of the gene of interest into a suitable functional construct consisting of the gene expression promoter, gene of interest, selectable marker and codon modification. This is followed by *Agrobacterium*-mediated transformation or bacterium-to-plant transfer and finally nucleus targeting [59–61]. During gene transfer within the plant cell, the transformed *Agrobacterium* facilitates the transfer of T-DNA molecules into the plant genome, then the transgene is randomly incorporated into the plant chromosome. Integration of T-DNA into the plant DNA sequence is then facilitated by non-homologous end-joinings.

Transfer of foreign genes that enhance FHB resistance into wheat is a viable alternative which has, in recent years, been used extensively to increase not only the crops' genomic variability, but also the fitness of wheat against *F. graminearum*. Among first genes to be transferred since 1992 was the *Bar* gene used as a selective marker and various others including the *TaPIMP1* gene [62], the *Yr10* gene [63] and the *TcLr19PR1* gene [64]. Various genes that encode pathogenicity related proteins (PR proteins) could be the new sources of wheat resistance against FHB. These PR proteins are defensins, which have a broad range of antifungal properties [65]. Defensin RsAFP<sup>2</sup> with growth inhibitory characteristics against *F. graminearum* was incorporated into variety Yangmai 12 using biolistic particle bombardment [66]. The success of the transformation was confirmed using PCR and Southern blot analysis. Expression of the *RsAFP<sup>2</sup>* genes in transformed wheat lines was confirmed using RT-PCR and Western blotting. Disease resistance was assessed, and the transformed lines showed resistance against *F. graminearum* compared to the untransformed control lines [66]. The low transformation efficiency using the biolistic particle bombardment, however, warrants the need for other gene transformation techniques alongside. *Agrobacterium*-mediated transformation is one such technique that has been used successfully to introduce foreign genes into the wheat plant with improved transformation efficiency.

In one effort, chitinase and #beta#-1,3-glucosanase genes were transformed into wheat to improve resistance against FHB. The transformation of chitinase and #beta#- 1,3-glucosanase genes (constructed into binary vector pCAMBIA3301) was mediated by *Agrobacterium* and the resultant transgenic lines showed resistance against FHB in the field [67]. Transformation of plant cells with exotic genes mediated with *Agrobacterium* is the initial step in introducing genes into plant cells that generate into adult plants capable of producing normal seeds. However, this process is difficult with wheat because of its complex hexaploid genome. Therefore, a more efficient protocol for wheat transformation called, 'Pure Wheat', was introduced [68]. This technique has renewed hope in accelerating transgenic wheat plants with superior traits such as FHB resistance and its associated ability to limit mycotoxin production.

#### *5.3. Genome Editing for FHB Resistance*

Major improvements in wheat will likely be brought about by genome editing, which promises to supersede the traditional random mutagenesis and conventional breeding. Genome editing technologies include the clustered regularly interspaced short palindromic repeat-associated endonucleases (CRISPR/Cas) technique, which is gaining much popularity, and other sequence-specific nucleases (SSNs) such as the transcription activator-like effector nucleases (TALENs) and zinc-finger nucleases (ZFNs). These technologies offer the benefits of gene knock-out, knock-in, replacement, activation and DNA repair [69–72]. Among these genome editing technologies, the CRISPR/Cas technology seems to hold more promise with regards to FHB resistance. The Cas nuclease system has been used with success in understanding fungal biology, with various reports in *Neurospora crassa* [73], *Aspergillus* spp. [74,75], *Penicillium chrysogenum* [76], *Alternaria alternata* [77], *Pyricularia oryzae* [78] and *Ustilago maydis* [79]. Following on these milestones, a Cas9-based genome

editing system was established in *F. graminearum* [80] and hopefully this study will generate leads to a breakthrough in *F. graminearum* control.

Several research groups have made concerted efforts to develop transgenic and mutagenic lines that confer resistance to FHB. Table 1 summarizes some of the genes that have been manipulated in wheat, barley, *Brachypodium* and *Arabidopsis* that were manipulated through advanced technologies and proved to confer reduced *F. graminearum* infection and DON accumulation. However, much effort is still needed to link the various research institutions with public and private seed companies to ensure that research and development are aimed at variety release to benefit farming communities in FHB prone areas. This effort should involve pre-commercial field-testing activities including multi-environmental trials and end-use quality analysis.

**Table 1.** Genetic transformation to enhance resistance against *Fusarium graminearum* causing Fusarium head blight in wheat and other cereals.



**Table 1.** *Cont.*

#### *5.4. Association Mapping to Find FHB Molecular Markers*

Molecular breeding and selection for FHB resistance in wheat have largely benefited from association mapping of putative QTL through associating phenotypic reactions to genotypes. Currently, high-density wheat 90 K single nucleotide polymorphism (SNP) assays are being used in genome-wide association (GWAS) studies aimed to dissect the genetic basis of resistance to Fusarium head blight in wheat breeding populations [92]. Association mapping studies have enabled the discovery of several loci associated to the resistance to FHB spread and DON accumulation. Alternative to the GWAS approach, candidate-gene association mapping can be used by targeting associations of pre-specified FHB resistance genes and the observed phenotypic reaction [93]. A recent GWAS study identified 16 significant SNPs associated with Fusarium-damaged kernels and DON levels on wheat chromosomes and suggested that FHB severity can even be reduced by smalleffect QTL [94]. Such studies form the basis of maker-assisted selection and marker-based gene and/or QTL introgression by identifying putative markers linked to genetic regions controlling particular traits. Quality phenotypic data, often with high heritability from multi-environmental trials, is required for effective association studies.

All these advanced technologies that can be employed to enhance FHB resistance have their own advantages and disadvantages when compared to traditional breeding methods. Table 2 highlights some of these pros and cons to guide future research. Generally, this indicates that the recent technologies can not completely be divorced from all aspects of traditional breeding, particularly phenotyping or field testing to account for the expression of introduced genes under real production conditions and assessing the ultimate impact on final yield.


**Table 2.** Pros and cons of using traditional breeding methods against using recent technologies.


**Table 2.** *Cont.*

#### **6. Tools to Assist Breeding for Resistance against FHB and Mycotoxin Contamination**

Laboratory analytical tools are useful to assess toxin accumulation in wheat infected with *F. graminearum.* These tools can be used in breeding programmes to assess if resistance to mycotoxin accumulation by *F. graminearum* is incorporated and in monitoring the safety of food products made from wheat grain. To incorporate Fusarium head blight resistance in wheat, various assessment methods are employed for each breeding objective. Resistance against pathogen penetration and resistance against disease spread after initial infection can be monitored visually. Monitoring resistance against mycotoxin accumulation requires specialized equipment that is able to detect even trace amounts of the mycotoxins. For the purposes of the current review, real-time PCR, chromatography and mass spectrometrybased approaches are discussed as tools to assist selection.

#### *6.1. Real-Time PCR*

Inoculation with *F. graminearum* and then determining the quantity of the inoculum is done by real-time PCR. Real-time PCR is important for diagnoses using species-specific primers to detect a suspect pathogen and for quantifying pathogen titre in infected kernels [95–98]. The technique has the potential to unpack the gene expression in response to FHB infection through monitoring transcriptome expression patterns within specific plant tissue after inoculation. Newer genomic technologies, such as genome-wide single polymorphism mapping, genome sequencing, microarrays and RNA sequencing, have been instrumental in identifying genotypes with FHB resistance. These techniques have also been useful in identifying QTL, linking resistance with other phenotypic traits as well as detecting and validating diagnostic markers.

#### *6.2. Chromatography and Mass Spectrometry-Based Approaches to Assist Selection*

Regulatory standards with threshold prescriptions for wheat products such as the Codex Alimentarius Commission 2015 require that there are monitoring procedures to quantify the DON toxin in harvested wheat grain and grain products. Chromatography and mass spectrometry-based techniques become handy in such circumstances to ensure safety of wheat products in the market. Notably, high performance liquid chromatography (HPLC) is commonly used for separation, identification, and quantification of mycotoxin levels in flour, food and feed mixtures. Other techniques include gas chromatography– mass spectrometry (GC-MS) and thin-layer chromatography (TLC), which are also effective for early detection and quantification of DON in wheat. Equally important is the use of

these quantitative techniques in screening breeding material and donor lines to be used in breeding against FHB, especially for type II resistance. Chromatography and mass spectrometry have been useful in identifying mycotoxin contaminants of wheat [96] and mycotoxin accumulation [99]. Because of their ability to detect and quantify contaminants and trace elements, chromatography and mass spectrometry-based techniques are useful in routine monitoring of grain safety to ensure compliance to prescribed standards. This could be the extension of the use of these techniques beyond research. With these state-of-the-art tools, breeding and selection of FHB resistant genotypes are becoming more efficient and reliable data are being produced on resistance to infection and mycotoxin contamination in the wheat grain.

#### **7. Conclusions**

The safety of wheat products is essential to ensure that human and animal lives are not endangered. Mycotoxins produced by the wheat-infecting *Fusarium graminearum* pathogen pose serious health risks to animals and human beings. It is therefore of the utmost importance to breed wheat varieties that are able to limit the accumulation of mycotoxins in wheat kernel that have been infected with *F. graminearum*. Traditional breeding techniques have been utilized to incorporate resistance against *F. graminearum* from resistance sources such as Sumai#3. However, the limitations of traditional plant breeding require integration of new and more sophisticated methods for cultivar improvement to fast-track *F. graminearum* resistance breeding. These techniques will also bolster resistance against mycotoxin accumulation. Clustered regularly interspaced short palindromic repeat-associated endonucleases (CRISPR/Cas) as well as RNA interference are some of the advanced tools that have revolutionized crop improvement efforts. Various molecular techniques like real-time PCR and biochemical analytical tools such as chromatography and mass spectrometry are also useful for detecting levels of infection by *F. graminearum*, and their use remains relevant for the future.

**Author Contributions:** Each author participated sufficiently in the completion of this work. Conceptualization, S.F.; writing—original draft preparation, S.F. and L.M.; writing—review and editing, S.F. and L.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research has received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors acknowledge the University of South Africa for the overall research support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Thomas Thomidis 1,\*, Konstantinos Michos <sup>2</sup> , Fotis Chatzipapadopoulos <sup>2</sup> and Amalia Tampaki <sup>2</sup>**


**Abstract:** Olive leaf spot (*Venturia oleaginea*) is a very important disease in olive trees worldwide. The introduction of predictive models for forecasting the appearance of a disease can lead to improved disease management. One of the aims of this study was to investigate the effect of temperature and leaf wetness on conidial germination of local isolates of *V. oleaginea*. The results showed that a temperature range of 5 to 25 ◦C was appropriate for conidial germination, with 20 ◦C being the optimum. It was also found that at least 12 h of leaf wetness was required to start the germination of *V. oleaginea* conidia at the optimum temperature. The second aim of this study was to validate the above generic model and a polynomial model for forecasting olive leaf spot disease under the field conditions of Potidea Chalkidiki, Northern Greece. The results showed that both models correctly predicted infection periods. However, there were differences in the severity of the infection, as demonstrated by the goodness-of-fit for the data collected on leaves of olive trees in 2016, 2017 and 2018. Specifically, the generic model predicted lower severity, which fits well with the incidence of the disease symptoms on unsprayed trees. In contrast, the polynomial model predicted high severity levels of infection, but these did not fit well with the incidence of disease symptoms.

**Keywords:** leaf wetness; temperatures; validation; *Venturia oleaginea*

## **1. Introduction**

Olive leaf spot (*Venturia oleaginea* (Castagne) Rossman & Crous, comb. nov.) is the cause of a very important disease in olive trees worldwide. According to Trapero and Blanco [1] and Viruega et al. [2], *V. oleaginea* over summers as mycelium in infected leaves that remain in the tree canopy or fallen to the soil surface, while in autumn, mycelia resume growth from the latent infections caused during the last spring or from old lesions, and new conidia are produced, which are dispersed by rain splash and run-off. The main symptoms of the disease are dark sooty spots (commonly known as peacock spots) which appear on the upper surface of leaves, mainly in the low canopy. Rarely, similar spots may also appear on the stem and fruit [3]. Heavy premature defoliation, which sometimes leads to twig death of olive (*Olea europaea* L.), can been caused by this pathogen [4] when no preventive or curative sprays are applied. According to Prota [5], in the Mediterranean region, fungicides are usually applied in the main shoot-growth seasons (spring and/or autumn).

Meteorological factors play a key role in infection by the olive leaf spot fungus. Temperature and moisture are the main climate factors influencing the development of *V. oleaginea* in olive trees. Relatively mild to low temperatures and free moisture on the leaves favor infections during the rainy periods in fall, winter, and spring [6–8]. Previous works have shown that the minimum temperature for conidia germination of the fungus is 5 ◦C, the optimum 20 ◦C, and the maximum 30 ◦C [6,9,10]; this pathogen is able to sporulate at temperatures from 5 to 25 ◦C [11]. In Greece, these temperatures occur mainly in the

**Citation:** Thomidis, T.; Michos, K.; Chatzipapadopoulos, F.; Tampaki, A. Evaluation of Two Predictive Models for Forecasting Olive Leaf Spot in Northern Greece. *Plants* **2021**, *10*, 1200. https://doi.org/10.3390/ plants10061200

Academic Editor: Alessandro Vitale

Received: 17 March 2021 Accepted: 10 June 2021 Published: 12 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

period of September to May [12]. Saad and Masri [13] also demonstrated the relationship between conidial germination and leaf wetness duration. They found that a minimum of 42 h leaf wetness was required for *V. oleaginea* conidia to germinate at 12 ◦C, while at 20 ◦C, 18 h was required. Infection occurred from 5 to 25 ◦C, and disease severity was the greatest at ~20 ◦C for wetness durations of 12 to 24 h and at ~15 ◦C for longer durations, while the optimum temperature and minimum wetness durations for infection were 15.5 ◦C and 11.9 h [6,12,14]. Obanor et al. [15] found that temperature affected olive leaf spot severity, with the lesion numbers increasing gradually from 5 ◦C to a maximum at 15 ◦C, and then declining to a minimum at 25 ◦C, while the numbers of lesions increased with increasing leaf wetness period at all temperatures tested. The minimum leaf wetness periods for infection at 5, 10, 15, 20 and 25 ◦C were 18, 12, 12, 12 and 24 h, respectively.

Several forecasting models to predict the infection for specific plant diseases have been developed. Each of the developed models has their strengths and weakness, so choosing the right one is based on many factors. A widely used generic infection model is that developed by Magarey et al. [16], which described pathosystems in which the parameters of temperature and wetness duration were supplied for each of the studied pathogens. In addition to the generic model, an infection model using regression equations, such as those based on polynomials, logistic equations can be developed by conducting combinations of multiple temperature and wetness. A forecasting model to predict olive leaf spot infection was developed by Obanor et al. [15] based on a polynomial equation with linear and quadratic terms of temperature, wetness and leaf age. However, this model has not been validated under field conditions.

The successful development of a plant disease forecasting system also requires the proper validation of a developed model. There are a large number of predictive models for the many important plant diseases in the international [17] literature. However, the accuracy of the predictions of each model must be tested under field conditions in order to reduce: (a) erroneous indications of high risk in cases where, in fact, no disease was observed, and (b) erroneous indications of no risk where, in fact, the disease was observed [18]. Although the effect of temperature and leaf wetness on the conidial germination of *V. oleaginea* has been previous studied [15], repetition to validate previous results with local isolates of the pathogen is essential to fit these parameters in the generic model developed by Magarey et al. [16] for local uses. Thus, one of the main aims of this study was to investigate the minimum, maximum and optimum temperatures and leaf wetness durations for conidia germination of local isolates of *V. oleaginea*. Because validation of prediction models for forecasting plant disease under field conditions is important prior to commercial use, a second aim of this study was to validate the above generic model and the polynomial model developed by Obanor et al. [15] under the field conditions in Potidea Chalkidiki, Northern Greece.

#### **2. Results**

#### *2.1. Effect of Temperature and Leaf Wetness on Conidial Germination*

There was no significant difference between repeated trials (*p* = 0.201), so the data from the two trials were pooled. Temperature significantly influenced (*p* < 0.001, SE = 0.687) conidial germination. Under continuous wetness, the optimum temperature for conidial germination was 20 ◦C, whereas conidial germination was inhibited at 30 and 0 ◦C. Conidial germination was significantly less at 15 and 25 ◦C than at 20 ◦C. The percentage of conidial germination was significant higher at 15 and 25 ◦C than at 10 ◦C. Conidial germination was significantly less at 5 than 10 ◦C. The estimates of the parameters from the quadratic function of temperature (R<sup>2</sup> = 0.739; Y = 7.46 + 5.22 <sup>×</sup> <sup>X</sup> <sup>−</sup> 0.16 <sup>×</sup> <sup>X</sup> 2 ) and leaf wetness (R<sup>2</sup> = 0.946; Y = 23.88 + 4.43 <sup>×</sup> <sup>X</sup> <sup>−</sup> 0.06 <sup>×</sup> <sup>X</sup> 2 ) are presented in Figure 1.

**Figure 1.** Effect of temperature on conidial germination of *Venturia oleaginea.* The parameters of minimum, maximum and optimum temperatures were fit to the generic model. **Figure 1.** Effect of temperature on conidial germination of *Venturia oleaginea.* The parameters of minimum, maximum and optimum temperatures were fit to the generic model. **Figure 1.** Effect of temperature on conidial germination of *Venturia oleaginea.* The parameters of minimum, maximum and optimum temperatures were fit to the generic model.

There was no significant difference between repeated trials (*p* = 0.186), so the data from the two trials were pooled. Leaf wetness also significantly influenced (*p* < 0.001, SE = 0.073) conidial germination. Under constant temperature at 20 °C, the conidial germination started after 12 h of continues wetness. In contrast, no conidial germination was observed after 6 h of wetness. The percentage of conidial germination after 18 h of wetness was significantly higher than 12 h, but significantly less than 24 h. No significant difference was observed in the percentage of conidial germination after 24, 36 and 48 h of leaf wetness. The estimates of the parameters from the quadratic function are presented in There was no significant difference between repeated trials (*p* = 0.186), so the data from the two trials were pooled. Leaf wetness also significantly influenced (*p* < 0.001, SE = 0.073) conidial germination. Under constant temperature at 20 ◦C, the conidial germination started after 12 h of continues wetness. In contrast, no conidial germination was observed after 6 h of wetness. The percentage of conidial germination after 18 h of wetness was significantly higher than 12 h, but significantly less than 24 h. No significant difference was observed in the percentage of conidial germination after 24, 36 and 48 h of leaf wetness. The estimates of the parameters from the quadratic function are presented in Figure 2. There was no significant difference between repeated trials (*p* = 0.186), so the data from the two trials were pooled. Leaf wetness also significantly influenced (*p* < 0.001, SE = 0.073) conidial germination. Under constant temperature at 20 °C, the conidial germination started after 12 h of continues wetness. In contrast, no conidial germination was observed after 6 h of wetness. The percentage of conidial germination after 18 h of wetness was significantly higher than 12 h, but significantly less than 24 h. No significant difference was observed in the percentage of conidial germination after 24, 36 and 48 h of leaf wetness. The estimates of the parameters from the quadratic function are presented in Figure 2.

Figure 2.

**Figure 2.** Effect of leaf wetness on the conidia germination of *Venturia oleaginea.* The parameters of minimum and maximum leaf wetness were fit to the generic model. **Figure 2.** Effect of leaf wetness on the conidia germination of *Venturia oleaginea.* The parameters of minimum and maximum leaf wetness were fit to the generic model. **Figure 2.** Effect of leaf wetness on the conidia germination of *Venturia oleaginea.* The parameters of minimum and maximum leaf wetness were fit to the generic model.

*2.2. Evaluation of Model Accuracy* 

*2.2. Evaluation of Model Accuracy* 

## *2.2. Evaluation of Model Accuracy*

*Plants* **2021**, *10*, x FOR PEER REVIEW 4 of 16

The average temperature, rainfall, and leaf wetness for the period May to December for each of these three years is presented in Figure 3. Figure 4 presents the predictions of the generic and polynomial models for the period of May to December for three consecutive years (2016, 2017, 2018). The average temperature, rainfall, and leaf wetness for the period May to December for each of these three years is presented in Figure 3. Figure 4 presents the predictions of the generic and polynomial models for the period of May to December for three consecutive years (2016, 2017, 2018).

**Figure 3.** *Cont*.

*Plants* **2021**, *10*, x FOR PEER REVIEW 5 of 16

**Figure 3.** *Cont*.

*Plants* **2021**, *10*, x FOR PEER REVIEW 6 of 16

**Figure 3.** *Cont*.

*Plants* **2021**, *10*, x FOR PEER REVIEW 7 of 16

**Figure 3.** *Cont*.

*Plants* **2021**, *10*, x FOR PEER REVIEW 8 of 16

**Figure 3.** Average temperature, rainfall, and leaf wetness for the period May to December in 2016, 2017 and 2018. **Figure 3.** Average temperature, rainfall, and leaf wetness for the period May to December in 2016, 2017 and 2018.

In 2016, the polynomial model predicted risk >29 on the 15th, 20th and 31st of May, 23th September, 10th and 15 October, 8th November and 1st December (Table 1). In contrast, no prediction of risk >29 was given from the generic model in the period April–December 2016. Very low incidence symptoms (percentage of diseased leaves < 5%) of the disease were observed in the unsprayed control trees at the same period. The mean temperature was 18.6 ◦C in May, which increased to 24.8 ◦C, 26.6 ◦C and 26.2 ◦C in June, July and August, and decreased to 22.7 ◦C, 17.8 ◦C, 13.8 ◦C in September, October and November respectively. The above temperatures were not a limiting factor for infections of the olive trees by the fungus *V. oleaginea*. In contrast, the total degree of hourly leaf wetness was very low in the period of May to December and not favorable for the development of the disease.

**Table 1.** Dates of the first seasonal infection of *Venturia oleaginea* predicted by the Generic and Polynomial models in 2016, 2017 and 2018; corresponding values of the risk index calculated by the model; times of actual disease onset; and percentage of diseased leaves.


<sup>a</sup> Date when model prediction value is higher to 29. <sup>b</sup> Risk was calculated on a scale from 0 (no risk) to 100 (maximum risk) based on weather data and tree growth stage. <sup>c</sup> Days after predicted infection–incubation period. <sup>d</sup> Model predicted risk >29 was also observed on other dates of the year without correlated with actual disease onset (see Figure 4).

*Plants* **2021**, *10*, x FOR PEER REVIEW 9 of 16

**Figure 4.** Predictions (as presented on the screen of the computer) of the generic model (orange line) and polynomial models (black line) to forecast infections from the fungus *Venturia oleaginea* on olive trees for the period May to December of three consecutive years (2016, 2017, 2018) (FO = First Observation; PDL = Percentage of the Diseased Leaves). **Figure 4.** Predictions (as presented on the screen of the computer) of the generic model (orange line) and polynomial models (black line) to forecast infections from the fungus *Venturia oleaginea* on olive trees for the period May to December of three consecutive years (2016, 2017, 2018) (FO = First Observation; PDL = Percentage of the Diseased Leaves).

In 2016, the polynomial model predicted risk > 29 on the 15th, 20th and 31st of May, 23th September, 10th and 15 October, 8th November and 1st December (Table 1). In contrast, no prediction of risk > 29 was given from the generic model in the period April– December 2016. Very low incidence symptoms (percentage of diseased leaves < 5%) of the disease were observed in the unsprayed control trees at the same period. The mean temperature was 18.6 °C in May, which increased to 24.8 °C, 26.6 °C and 26.2 °C in June, July and August, and decreased to 22.7 °C, 17.8 °C, 13.8 °C in September, October and November respectively. The above temperatures were not a limiting factor for infections of the olive trees by the fungus *V. oleaginea*. In contrast, the total degree of hourly leaf wetness In 2017, the polynomial model predicted nearly continuously risk >29 between the 5th and 31st of May. The generic model predicted risk >29 at 15th and 31st May (Table 1). The first symptoms were observed at 28th May, and the final incidence of the symptoms was moderate when recorded 15 days later (percentage of diseased leaves in unsprayed trees was 18%). The mean temperature was 19.8 ◦C in May, while the total degree of hourly leaf wetness per day was about 20 in the same period. Those climate conditions were favorable for infections of olive trees by the fungus *V. oleaginea*. The optimum temperature and leaf wetness for growth of *V. oleaginea* occurred from 15th to 28th May, justifying the short incubating period of 14 days.

was very low in the period of May to December and not favorable for the development of the disease. In 2017, the polynomial model predicted nearly continuously risk > 29 between the 5th and 31st of May. The generic model predicted risk > 29 at 15th and 31st May (Table 1). The first symptoms were observed at 28th May, and the final incidence of the symptoms was moderate when recorded 15 days later (percentage of diseased leaves in unsprayed trees was 18%). The mean temperature was 19.8 °C in May, while the total degree of hourly leaf wetness per day was about 20 in the same period. Those climate conditions In 2018, the polynomial model predicted risk >29 at 2–5th, 10th, 20th, 25th and 30th May, 8th and 30th June, 25 July, 28th September (Table 1), while the predicted risk >29 was nearly always the period October-December. In contrast, no prediction of risk >29 was given from the generic model in the period April–September 2018. The generic model predicted risk >29 at the 3rd, 6th, and 20th October, 26th October, 16th November, 22th and 28th November, 3rd, 8th and 15th December. The first symptoms of the disease were observed at the 26th October, while the incidence of the symptoms was relatively high at the 10th November (percentage of diseased leaves <36%). No other results were collected after 10 November. The mean temperature was 21.7 ◦C in May, which increased to 24.6 ◦C, 26.6 ◦C and 27.4 ◦C in June, July and August, and decreased to 23.4 ◦C, 18.8 ◦C, 14.3 ◦C, 9.2 ◦C in September, October, November, December, respectively. The above

temperatures were not a limiting factor for infections of the olive trees by the fungus *V. oleaginea*. The total degree of hourly leaf wetness was very low in the period of May to September, and not favorable for the development of the disease. In contrast, there was a high number of hourly leaf wetness in October, making the climate conditions favorable for the development of the disease.

The estimates of the parameters from the linear regression analysis to find the relationship between model predictions and level of the disease (Generic Model: R<sup>2</sup> = 0.917, Y = 4.74 + 0.47 <sup>×</sup> X, Beta Value = 0.958; Polynomial Model: R<sup>2</sup> = 0.578, Y = 7.86 + 0.34 <sup>×</sup> X, Beta Value = 0.76) are presented in Figure 5. *Plants* **2021**, *10*, x FOR PEER REVIEW 11 of 16

**Beta Value = 0.958** 

**Beta Value = 0.76** 

**Figure 5.** Relationship between model (generic and polynomial) predictions and percentage of diseased leaves. **Figure 5.** Relationship between model (generic and polynomial) predictions and percentage of diseased leaves.

#### **3. Discussion**

So far, control of olive leaf spot has been based on prognosis. This method is adequate, but possesses disadvantages including inopportune and unnecessary spray applications. It increases the cost of production, and also the risk of environmental pollution. The introduction of predictive models to forecast the appearance of a disease could improve crop management by reducing the number of spray applications and improving the effectiveness of spray applications conducted.

Magarey et al. [16] developed a generic model appropriate for predicting the appearance of a high number of plant diseases. This model requires some climate parameters such as the minimum, maximum and optimum temperatures, as wells as the minimum and maximum numbers of hours of leaf wetness. The results of this study showed that a temperature range of 5 to 25 ◦C was appropriate for the conidial germination on detached olive leaves with 20 ◦C being the optimum. Previous works showed that conidial germination of *S. oleagina* (synonym of *V. oleaginea*) on agar was at its minimum at 5 ◦C, with an optimum at 20 ◦C, and a maximum at 30 ◦C [9,19,20], while Saad and Masri [13] found that conidial germination of *S. oleagina* could be observed in temperatures ranging from 5 to 25 ◦C. The above range of temperatures at which *V. oleaginea* conidia germinate suggests that infection may occur mainly throughout the period of September–June in olive growing regions of Northern Greece. This study also showed that at least 12 h of leaf wetness was required to start the conidial germination of *V. oleaginea* at optimum temperatures. According to Obanor et al. [15], the minimum leaf wetness periods for infection of olive trees from *S. oleagina* were 18, 12, 12, 12 and 24 h at 5, 10, 15, 20 and 25 ◦C, respectively, while the minimum leaf wetness periods required for conidial germination at 5, 10, 15, 20 and 25 ◦C were 24, 12, 9, 9 and 12 h, respectively [6].

Specific factors, such as pathogen biology, host phenology, and host variety in a specific area may significantly affect the input variables for a predictive model. As the predictive model could contain assumptions about site specific conditions, each model must be validated for a specific location by testing for one or more seasons under local conditions to verify that it works with precision in this location. Obanor et al. [15] developed a regression model to predict the infections of olive trees by the fungus *S*. *oleagina.* However, this model was not evaluated and validated under field conditions. In this study, the generic model developed by Magarey et al. [16] and the polynomial model developed by Obanor et al. [15] were evaluated to predict infection of olive trees by the fungus *V. oleaginea* under the climate conditions of Potidea Chalkidiki, Northern Greece. Because the purpose of the model was to be part of a warning system for olive leaf spot management, the ability to correctly predict infection periods is crucial. The results showed that both models correctly predicted infection periods. However, there was difference in the severity of the infection, as demonstrated by the goodness-of-fit for the data collected on leaves of olive trees in 2016, 2017 and 2018. Specifically, the generic model predicted lower severity of the infection which fits very well with the incidence of the symptoms of the disease. In contrast, the model developed by Obanor et al. [15] predicted high severity of the infection, but these did not fit as well with the incidence of the symptoms. Based on the above results, the polynomial model gave false positive predictions and did not generate proper spray recommendations increasing the fungicide applications with indirect results the increase of cost production and possible environmental pollution. It is recommended that the polynomial model be calibrated and re-validated under field conditions before commercial use. In contrast, the generic model gave a correct prediction for the appearance of the disease, and it seems to fit better in computer-assisted Decision Support Systems (DSSs).

Considering that this study did not include olive cultivars with different levels of susceptibility, it was not possible to evaluate whether each of the above predictive model can be fitted better to specific olive cultivars as the same climate conditions could favor different incidence of the symptoms depending on the level of cultivar susceptibility.

#### **4. Materials and Methods**

#### *4.1. Effect of Temperature and Leaf Wetness on Conidial Germination*

#### 4.1.1. Effect of Temperature

The effect of temperature on conidial germination of *V. oleaginea* was investigated by using the methodology described by Obanor et al. [6]. Leaves (cv. Chondrolia Chalkidikis) with symptoms of olive leaf spot were collected from a commercial field established in Potidea Chalkidiki (40.1939◦ N, 23.3301◦ E) in November 2015. The leaves were agitated in distilled water and the conidial suspension filtered through a double layer of cheesecloth to remove leaf debris. Inoculum suspensions were adjusted to 6 <sup>×</sup> <sup>10</sup><sup>4</sup> conidia mL−<sup>1</sup> by using a hemocytometer. Seven temperatures (0, 5, 10, 15, 20, 25, and 30 ◦C) were tested to find the upper and lower limits of spore germination. Fully expanded leaves (4 weeks old) without symptoms of the disease were excised from olive trees grown in a commercial field by cutting at the stem end of the petiole. The leaves, before being inoculated, were disinfested by dipping them in 10% vol/vol domestic bleach solution (4.85% NaOCl) for 5 min, washed three times with sterile distilled water and left to dry at room temperature. The leaves were inoculated with two drops (10 µL) of the conidial suspension deposited on the upper leaf surface. After inoculation, the leaves were placed in 9-cm petri dishes (wet paper towel was placed onto bottom and leaves was placed on plastic sticks so that to avoid any contact) arranged randomly in the growth chamber (97–100% RH) described below. Results were collected by recording the germination of 30 conidia/leaf (10 leaves for each treatment) 24h later. A conidium was considered germinated when the germ tube was equal to the greatest diameter of the swollen conidia (1 to 1.5×) [21,22].

#### 4.1.2. Effect of Leaf Wetness

Similarly, fully expanded leaves of olive trees (cv. Chondrolia Chlakidiki) without symptoms of disease were inoculated with two drops (10 µL) of the conidia suspension as described above. After inoculation, the leaves were placed in 9-cm petri dishes (wet paper towel was placed onto the bottom, and leaves were placed on plastic sticks so as to avoid any contact) arranged randomly in the growth chamber (Emmanuel E. Chryssagis, Growth Plant Chambers—GRW 500/CMP2) (97% ± 3 Relative Humidity) under continuous wetness at the 20 ◦C (optimal temperature identified above) and incubated for 6, 12, 18, 24, 36 and 48 h. Results were collected by recording the germination of 30 conidia as described above.

Both experiments were repeated once. General linear regression analysis was performed (SPSS Grad Pack 23, SPSS Inc., Chicago, IL, USA) in order to determine the relationship between leaf wetness, temperatures and conidia germination.

#### *4.2. Model Development and Validation*

#### 4.2.1. The Models

The generic model developed by Magarey et al. [16] was used. The parameters were used to run the model based on the results produced in the above experiments: Minimum Temperature (Tmin) = 5 ◦C, Maximum Temperature (Tmax) = 25 ◦C, Optimum Temperature (Topt) = 20 ◦C, Minimum Leaf Wetness (Wmin) = 12 h, Maximum Leaf Wetness (Wmax) = 24 h. In addition to the above, the predictive model (polynomial model; Y = <sup>β</sup>0 + <sup>β</sup>1A + <sup>β</sup>2T + <sup>β</sup>3W +β4(Ax W) + <sup>β</sup>5T2 + <sup>β</sup>6W2), where Y is √ (disease severity), A is the leaf age (weeks), T is temperature (◦C), W is wetness period (h), and β0....β6 are determined parameters)to forecast the appearance of olive leaf spot developed by Obanor et al. [15] was simultaneously evaluated under the field conditions of Potidea Chalkidiki.

The leaf wetness was estimated from the hourly data: if an (i) hour is wet, it is counted as 1, or when it is dry it is designated as 0 (so the dry hours are not counted and are not taken into account). Continuous wet hours are summed to determine leaf wetness. However, if there is an interruption of fewer than or equal to 20 dry hours and low relative humidity (<70%) (based on the result published by Villalta et al. [23] for the fungus *Venturia*

*pirina*), the summation of hours is continued. In contrast, if the interruption of dry hours is longer than 20 dry hours, a new summation of hours is started. Temperature is the event average temperature during each wet period. Cultivar susceptibility and inoculum level were not considered because insufficient information was available about their effects on the occurrence of infection.

## 4.2.2. Evaluation of Model Accuracy under Field Conditions

Model accuracy in predicting the day of infection was evaluated by comparing actual and predicted times of symptom appearance. In the Potidea Chalkidiki, which is one of the most important olive production areas in Greece, a telemetric meteorological station (NEUROPUBLIC S. A., Information Technologies & Smart Farming Services, Piraeus, 18545, Attica, Greece) was established to record weather data, which were used to run the models. The model was operated hourly, starting from the 1st of May (aiming to include both periods favorable for the development of the disease (May to June) and unfavorable periods for the development of the disease (July to August)), 00.01, and ending at 31st December using hourly leaf wetness and hourly temperatures as driving variables for calculation. The date of the first observation of the symptoms (in young leaves) was used to verify the prediction of the models, while the final incidence of the symptoms was recorded 15–20 days later by calculating the percentage of diseased leaves to a sample of 100 leaves randomly selected from each of 10 trees in total. The period of possible appearance of the disease was calculated on each day when Risk (LW, T) > 30, which was considered to be the incubation period. According to Bakari´c [24], an incubation period depends on the environmental conditions, and it lasts 15 days, but can be extended from three to eight or more months. The model predictions were ranged from 0 (when Risk = 0) to 100 (when Risk = the highest possible value).It was calculated with 0 being the minimum value that could be given by the model, and 100 the maximum value. All the other values were distributed between 0 and 100. Previous preliminary work under field conditions to find the threshold for the model predictions showed that no symptoms or very lightsporadic symptoms (percentage of diseased leaves < 5%) of the disease could be observed when the model predictions were in the range 0–29 (indicating that a spray application against olive leaf spot disease would not be financially justifiable; the spray program usually includes copper-based fungicides applied before the onset of the main infection periods, which often coincide with the main shoot-growth seasons (spring and autumn)).

A commercial olive field (cv. Chondrolia Chalkidiki, 7- to 10-year-old trees), located in Potidea, Chalkidiki, was chosen to record the appearance of olive leaf spot symptoms. Trees were pruned to a vase shape by hand pruning. Five to six irrigations were provided yearly. Nitrogen was applied yearly as (NH4)2SO<sup>4</sup> at 100 N units per hectare. Selected trees did not show any symptom of the disease before starting the trial. The trees (kept unsprayed) were inspected twice per week to determine the time of symptom onset. The trees were carefully inspected for the appearance of the first symptoms, which are dark sooty spots (commonly known as peacock spots) appear on the upper surface of leaves, mainly in the low canopy. When the symptoms were unclear, the leaves were marked and observed during the following surveys. To assess the severity of the disease, 100 random leaves were observed for the symptoms of the disease per tree (results were collected from 10 trees), and the disease incidence was calculated as the percentage of leaves with leaf spot symptoms. The predicted period of disease onset was then compared with the actual one. The model was judged to have provided an accurate prediction when the observed symptom onset coincided with the time interval predicted by the model [25].

This experiment was conducted for three consecutive years (2016, 2017, and 2018). General linear regression analysis was performed (SPSSGradPack23, SPSS Inc., Chicago, IL, USA) in order to determine the relationship between model predictions and level of the disease (observations).

#### **5. Conclusions**

The effects of the air temperature and leaf wetness on *V. oleaginea* infection of olive leaves were clarified. Based on those results, disease forecasting systems were developed to find the proper timing of foliar fungicidal sprays. The generic model predicted lower severity, which fits well with the incidence of the symptoms of the disease in unsprayed trees, and this model could be used to schedule the spray applications against olive spot disease. In contrast, the polynomial model predicted high severity levels of infection, but this did not fit well with the incidence of the symptoms. This study could help pest managers and researchers predict the risk of olive leaf spot in different regions or under different crop management practices.

**Author Contributions:** Conceptualization, T.T. and A.T. methodology, F.C. software, K.M.; A.T. validation, T.T. formal analysis, T.T.; writing—original draft preparation, T.T.; K.M., F.C., A.T. writing review and editing, T.T.; supervision. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was fund by NEUROPUBLIC S. A., Information Technologies & Smart Farming Services, Piraeus, 18545, Attica, Greece.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to the agreement with the NEUROPUBLIC S. A.

**Acknowledgments:** In this section, you can acknowledge any support given which is not covered by the author contribution or funding sections. This may include administrative and technical support, or donations in kind (e.g., materials used for experiments).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

