Sources of resistance to Pseudocercospora fijiensis, the cause of black Sigatoka in banana

Abstract Black Sigatoka, caused by Pseudocercospora fijiensis, is one of the most devastating diseases of banana. In commercial banana‐growing systems, black Sigatoka is primarily managed by fungicides. This mode of disease management is not feasible for resource‐limited smallholder farmers. Therefore, bananas resistant to P. fijiensis provide a practical solution for managing the disease, especially under smallholder farming systems. Most banana and plantain hybrids with resistance to P. fijiensis were developed using few sources of resistance, which include Calcutta 4 and Pisang Lilin. To broaden the pool of resistance sources to P. fijiensis, 95 banana accessions were evaluated under field conditions in Sendusu, Uganda. Eleven accessions were resistant to P. fijiensis. Black Sigatoka symptoms did not progress past Stage 2 (narrow brown streaks) in the diploid accessions Pahang (AA), Pisang KRA (AA), Malaccensis 0074 (AA), Long Tavoy (AA), M.A. Truncata (AA), Tani (BB), and Balbisiana (BB), a response similar to the resistant control Calcutta 4. These accessions are potential sources of P. fijiensis resistance and banana breeding programmes can use them to broaden the genetic base for resistance to P. fijiensis.

The fungus is heterothallic and produces asexual conidia and sexual ascospores throughout the year (Fouré, 1987). The disease is polycyclic and results in multiple infections in a banana cycle, leading to substantial leaf damage and yield losses of >50% (Guzmán et al., 2019). In large-scale plantations, black Sigatoka is managed by the frequent application of fungicides (Churchill, 2011). Small-scale farmers have limited access to fungicides and often cannot afford them (Alakonya et al., 2018). They therefore suffer massive losses from this disease.
In East Africa, IITA and the National Agricultural Research

Organisation in Uganda (NARO) developed 27 improved East African
Highland banana (EAHB) hybrids, known as NARITAs. The NARITAs have high yields, and some of them are resistant to black Sigatoka . One of these hybrids, NARITA 7, has been deployed to farmers in Uganda (Nowakunda et al., 2015). The Fundacion Hondureña de Investigación Agrícola (FHIA) in Honduras has also developed improved diploids and hybrids with resistance to black Sigatoka (Pillay et al., 2012;Rowe & Rosales, 2000). The hybrids include  which are now grown in many African countries, including Ghana, Kenya, Nigeria, Tanzania, and Uganda (Tenkouano & Swennen, 2004).
The success of resistance breeding is dependent on the availability of good sources of resistance (Pillay et al., 2012). Several banana varieties resistant to black Sigatoka have been identified and used in banana improvement programmes (Pillay et al., 2012;Vuylsteke et al., 1993). Among these, Calcutta 4 (M. acuminata subsp. burmannicoides) and Pisang Lilin (M. acuminata subsp. malaccensis) are the most extensively used (Pillay et al., 2012;Vuylsteke et al., 1997).
However, a vast genetic diversity does exist in bananas that may serve as potential donors of resistance (Christelová et al., 2017), but these have not been used by breeding programmes, mainly because of sterility of some of the clones and low seed set (Ortiz & Swennen, 2014).
An overreliance on a few sources of disease resistance to P. fijiensis poses a risk to the sustainability and durability of host resistance. P. fijiensis undergoes regular sexual recombination, which suggests that the fungus might overcome existing sources of resistance (McDonald & Linde, 2002). Examples of this have already been reported. Fullerton and Olsen (1995) reported that P. fijiensis isolates in Papua New Guinea and the Pacific Islands overcame resistance in young Calcutta 4 plants. In the Cook Islands, the resistant cultivars Paka and T8 (a Paka × Highgate AAAA hybrid) were reported to have become susceptible (Fullerton & Olsen, 1995). Yangambi KM5, a variety once considered highly resistant to P. fijiensis (Fouré, 1987), also became susceptible to black Sigatoka in Cameroon (Mouliom-Pefoura, 1999), Costa Rica (Escobar-Tovar et al., 2015), and Tanzania (Kimunye et al., 2019). In Cuba, the resistant FHIA-18 hybrid became susceptible to P. fijiensis (Miranda et al., 2006). All these reports point to a changing pathogen virulence profile and the risk of relying on a narrow genetic pool. The existing resistant banana gene pool therefore needs to be broadened to ensure that durable resistance to black Sigatoka is being developed by Musa breeding programmes.
The identification and introgression of new and effective P. fijiensis resistance genes into banana hybrids and cultivars has now become necessary.
Bananas and plantains have been screened for resistance to black Sigatoka before. Fouré (1994) evaluated more than 350 accessions for response to black Sigatoka in Njombe in Cameroon. However, these accessions have not been evaluated in other locations in Africa, especially in the East African highlands. Host response to infection can also depend on plantation management, including soil fertility regimes and nutrients (Kablan et al., 2012), as well as pathogen characteristics. Isolates with differing levels of aggressiveness and virulence have been reported. For example, Romero and Sutton (1997) reported higher black Sigatoka severities on Grand Naine and False Horn with isolates from Colombia, Costa Rica, and Honduras compared to those from Cameroon and Asia, while Fullerton and Olsen (1995) reported P. fijiensis strains with differential virulence from those collected in Papua New Guinea and the Pacific Islands.
Banana genotypes used as resistance sources must therefore be evaluated in different environments before being used in breeding programmes. Eight of these were selected from the first trial based on their response to P. fijiensis, five were diploid accessions previously used to generate improved diploids, six were improved diploids, and three were tetraploids used in the NARO/IITA breeding pipeline. Yangambi KM5 was included to validate the reduced resistance observed in farmers' fields (Kimunye et al., 2019), while Williams and Mbwazirume served as susceptible checks, and Calcutta 4 as resistant check.
The field experiment consisted of rows comprising seven plants per accession, of which five were used for disease ratings, planted in a randomized complete block design with three replications. The trial was established with suckers collected in Sendusu and Kawanda. Plants were planted with a spacing of 2 × 3 m. The suckers were pared before planting, and the rhizomes treated with Dursban (chlorpyrifos) for 20 min to eliminate nematodes and weevils. A P. fijiensis-susceptible Matooke variety (EAHB, cooking type), Enzirabahima, was used as a disease spreader row to ensure there was enough inoculum in the field. The accessions were evaluated every 3 months, starting at 6 months after planting, for three crop cycles (mother plant, daughter, and granddaughter), concluding evaluations in November 2018. Each of the cycles lasted 9-12 months depending on the cultivar.
Field management of the two trials was similar. At planting each hole was filled with 10 kg cow manure, after which dry grass was applied as mulch 4 months after planting. Weeding was done by hand until flowering. A herbicide (Weedall, a glyphosate-based nonselective herbicide) was thereafter used to manage weeds.
Detrashing was minimal and limited to dry leaves hanging around the pseudostem.

| Disease evaluation
Disease was scored by counting the number of standing leaves (NSL). Each leaf was visually rated for the stage of symptom development, as described by Fouré (1987): Stage 1, development of faint, minute, reddish-brown specks on the lower surface of the leaf; Stage 2, narrow reddish-brown streaks; Stage 3, streaks that change colour from reddish-brown to dark brown or black that are clearly visible at the upper surface of the leaf; Stage 4, streaks broaden and become spindle-shaped with water-soaked borders; Stage 5, lesions with dark brown or black centres that are slightly depressed with water-soaked borders; and Stage 6, grey lesions with dried out centres ( Figure 1).
According to this scale, 0 = no visible symptoms, 1 = <1%, 2 = 1%-5%, 3 = 6%-15%, 4 = 16%-33%, 5 = 34%-50%, and 6 = 51%-100% of leaf area covered with disease symptoms. At the most advanced stage of symptoms (SSD), that is, the stage at which symptom progression where X i = proportion of the host tissue damaged at ith day, t i = the time in months after appearance of the disease at ith month, and n = the total number of observations.

| Data analysis
Variation among accessions was assessed using one-way analysis of variance (ANOVA), and the means separated using the least significant difference at the 95% confidence level. For the second trial, no significant differences between the mother, daughter, and granddaughter plants were obtained, so the data were combined and subjected to an ANOVA. Pearson's correlation was used to determine the association between the different disease parameters,

| Genetic grouping
Accessions were assigned to different genomic groups and ploidy level according to the Musa Germplasm Information system (MGIS) database (https://www.crop-diver sity.org/mgis/). The accessions were then grouped into genetic clusters based on simple-sequence   Accessions grouped together based on the morphological traits and into cluster as defined using simple sequence repeats (Christelová et al., 2017;Nakato et al., 2018 Note: Accessions in bold are the improved diploids and tetraploids. Abbreviations: AUDPC, area under disease progress curve; YLst, youngest leaf with streak symptoms; YLS, youngest leaf spotted; SSD, most advanced stage of symptoms; INSL index of nonspotted leaves (%); DSI, disease severity index. Disease parameters collected at 3-month intervals and averaged over three crop cycles. a Genome group and ploidy level assignment was based on Musa Germplasm Information System. b Accessions grouped into clusters as defined using simple sequence repeats (Christelová et al., 2017;Nakato et al., 2018 (Carlier et al., 2003).
g Mbwazirume, an East African Highland banana, and Williams were used as susceptible local checks.

| Black Sigatoka symptoms
Black Sigatoka symptoms were observed on all banana accessions evaluated. The symptoms ranged from Stage 2 to the late necrotic stage (Stage 6) (Figure 1)

| Relationship between disease parameters
Significant correlations (p < 0.0001) were observed between black Sigatoka assessment parameters in both screening trials (Table 3).  (Table 3).
SSD and AUDPC had a higher coefficient of determination than DSI, YLS, YLst, and INSL for the two trials, and were therefore used in subsequent analysis (Table 3).

TA B L E 3 Pearson coatoka evaluation parameters
in several accessions did not differ significantly from Calcutta 4, and these were classified as resistant (Table 1) Truncata. The other highly resistant accessions were Tani and Balbisiana within the BB genome group (Table 1).
Hierarchical clustering revealed three groups representing resistant and susceptible accessions, while some accessions had an intermediate response. The resistant group comprised 28 accessions that clustered with Calcutta 4 (Figure 2). Some of the accessions in this group, such as K.N. Khom, Kayinja, and Pisang Lilin, had a significantly higher AUDPC than Calcutta 4, but with symptoms that did not progress beyond Stage 4 ( Table 1). The second group con-  (Table 1).

| Trial 2
The response of the banana accessions to P. fijiensis in Trial 2 varied significantly (p < 0.05) ( Table 2). The accessions also grouped into three clusters. Long Tavoy, followed by Calcutta 4, were most resistant, with AUDPC values of 39.8 and 57.0, respectively (Table 2).
In the two trials, 31 accessions were considered resistant to black Sigatoka, of which one was an improved diploid (02145/1320),

| Genetic grouping
The accessions evaluated were distributed across 22 subgroups within nine clusters (Table 4; Figure 4) (Table 4). Most of the susceptible accessions were in the subgroup AAA Lujugira/Mutika (Cluster X) and AA cv.
African (Cluster IX), with 15 and 14 accessions, respectively (Table 4). Earlier studies of banana varieties in Cameroon also reported resistant accessions in these subgroups (Fouré, 1994). More accessions from these subspecies should be screened to expand the available sources of resistance to P. fijiensis. Improved diploids are routinely used as male parents in banana breeding programmes Vuylsteke et al., 1993). The improved diploids 10969S-1 and TMB2X5265-1, for instance, were reported as good sources of black Sigatoka resistance (Batte et al., 2019). However, in the current study they were not resistant to P. fijiensis. These contrasting observations could be attributed to the high genetic diversity and emergence of new and highly virulent pathotypes arising from frequent sexual reproduction documented in P. fijiensis isolates from Uganda . Additional studies to characterize the virulence of P. fijiensis population are recommended.

| DISCUSS ION
Symptom progression in Calcutta 4 and other accessions in this study stopped at the early streak stage (Stage 2). This corresponds to the host reaction previously described by Meredith and Lawrence (1970) and Fouré (1994). The reaction in Calcutta 4 has been described as a hypersensitive response (Fouré, 1994;Guzmán et al., 2019), a type of resistance thought to be controlled by a major gene. Type 2, characterized by typical but slow symptom progression up to necrosis (Fouré, 1994). This is because they allowed the pathogen to produce asexual spores, which are normally produced at A substantial number of accessions resistant to black Sigatoka contained the B genome (BB, AB, AAB, ABB). This is contrary to the findings of Fouré (1994), who reported that cultivars with a B genome were mainly susceptible or had partial resistance. M. balbisiana has several desirable attributes including drought tolerance (Ravi et al., 2013), but their inclusion in banana breeding has been limited until now, primarily due to the banana streak virus (eBSV) that is encoded in the B genome (Bakry et al., 2009). Recent studies have shown that the recombination of M. balbisiana and M. acuminata resulted in an eBSV-free progeny (Noumbissié et al., 2016;Umber et al., 2016).
This presents the possibility of using M. balbisiana to broaden and improve resistance to black Sigatoka, as well as to introduce other desirable traits such as drought tolerance.
In this study, tetraploid hybrids derived from the cross of the P. fijiensis-susceptible EAHBs (Nante, Nfuka, and Entukura) with the P. fijiensis-resistant Calcutta 4 were susceptible to black Sigatoka. These included 376K-1 (Nante × Calcutta 4), 222K-1 (Nfuka × Calcutta 4), and 1438K-1 (Entukura × Calcutta 4). The tetraploids were derived from genetically related Matooke bananas (Němečková et al., 2018) that are highly susceptible to black Sigatoka. This deviates from earlier findings whereby susceptible plantain triploids were crossed with diploid Calcutta 4, resulting in mostly black Sigatoka-resistant tetraploid hybrids . Thus, selections for advancement in breeding need to be made based on the reaction of individual hybrids to black Sigatoka. Resistance to P. fijiensis in Musa hybrids is conferred by a major recessive gene bs1 and two modifiers genes, bsr 1 and bsr 2 , with an additive effect (Craenen & Ortiz, 1997;Ortiz & Vuylsteke, 1994). Segregation of the three loci result in progeny with a variable response to P. fijiensis (Ortiz & Vuylsteke, 1994), thus making progeny predictions based on parental phenotype unreliable. An understanding of the genetics of resistance of a parental cultivar can guide breeders to make informed decisions on the choice of parents to use in their breeding programmes, to minimize the risk of a breakdown in resistance.
The accessions Saba, IC2, and Pelipita, which were susceptible to P. fijiensis in the current study, were previously reported as moderately resistant in Cameroon (Fouré, 1994;Guzmán et al., 2019).
Pisang Ceylan was reported resistant in this study, but moderately resistant in Cameroon (Guzmán et al., 2019). These results are probably a reflection of differences in environmental factors, including different weather patterns, soil characteristics, and fertility regimes, and/or the presence of isolates differing in virulence profiles. It is therefore important that environmental factors and pathogen profiles be investigated at different locations to understand what other factors influence genotype response to infection.

F I G U R E 4
Simple-sequence repeat (SSR)-based genetic clusters (with a bold rectangle around them), into which the accessions assessed for response to infection with Pseudocercospora fijiensis were categorized. The 95 accessions grouped into Clusters I, III, VII, VIII, IX, X, XI, and XII as indicated by Christelová et al. (2017). The individual sets of the clustered accessions are indicated in Tables 1 and 2. The diagram was adopted and modified from Christelová et al. (2017) [Colour figure can be viewed at wileyonlinelibrary.com]