Genetic diversity of Colletotrichum lupini and its virulence on white and Andean lupin

Lupin cultivation worldwide is threatened by anthracnose, a destructive disease caused by the seed- and air-borne fungal pathogen Colletotrichum lupini. In this study we explored the intraspecific diversity of 39 C. lupini isolates collected from different lupin cultivating regions around the world, and representative isolates were screened for their pathogenicity and virulence on white and Andean lupin. Multi-locus phylogeny and morphological characterizations showed intraspecific diversity to be greater than previously shown, distinguishing a total of six genetic groups and ten distinct morphotypes. Highest diversity was found across South America, indicating it as the center of origin of C. lupini. The isolates that correspond to the current pandemic belong to a genetic and morphological uniform group, were globally widespread, and showed high virulence on tested white and Andean lupin accessions. Isolates belonging to the other five genetic groups were mostly found locally and showed distinct virulence patterns. Two highly virulent strains were shown to overcome resistance of advanced white lupin breeding material. This stresses the need to be careful with international seed transports in order to prevent spread of currently confined but potentially highly virulent strains. This study improves our understanding of the diversity, phylogeography and pathogenicity of a member of one of the world’s top 10 plant pathogen genera, providing valuable information for breeding programs and future disease management.

genes [28][29][30] , whereas in white, Andean and yellow lupin no such single gene resistance is known and the observed quantitative resistance is considered to be polygenic [31][32][33] . The increasing demand for plant-based protein is renewing the interest for lupins as a high quality protein crop [34][35][36] , the current anthracnose pandemic, however, severely hampers cultivation.
The pathogen was first described as Gloesporium lupini, followed by C. gloeosporioides and C. acutatum until it was fully described as C. lupini 14,37,38 . Currently two genetic groups (I and II) are distinguished within C. lupini based on vegetative compatibility groups (VCG) 38 , the ITS (internal transcribed spacer) region 37 and multilocus phylogeny of the ITS, GAPDH (glyceraldehyde-3-phosphate dehydrogenase), CHS-1 (chitin synthase), HIS3 (histone), ACT (actin), TUB2 (β-tubulin 2), HMG (HMG box region) and APN/MAT1 (Apn2-Mat1-2-1 intergenic) loci 39 . The TUB2 and GAPDH loci were shown to be the most informative within the C. acutatum species complex and APN/MAT1 the most informative within C. lupini, whereas classification based on the ITS region can be problematic due to low resolution within the complex 14,39 . Although only two groups within C. lupini have been distinguished, with most of the reported strains belonging to group II 39 , intraspecific diversity is thought to be greater as a high diversity was found in a Chilean C. lupini collection using random amplified polymorphic DNA (RAPD) markers 40 and a distinct lupin infecting C. acutatum group was identified in Ecuador based on the ITS region 41 . This suggests that highest intraspecific diversity is found in South America, which is believed to be the center of origin of members belonging to clade 1 of the C. acutatum species complex 10,15 .
The overall aim of this study was to assess a worldwide collection of lupin-infecting Colletotrichum isolates through (i) multi-locus phylogeny, (ii) morphology and (iii) virulence on white and Andean lupin. Insights into

Results
Colletotrichum lupini comprises of six genetic groups supported by morphology. From the 50 sequenced isolates, 39 belonged to C. lupini (Table 1). A globally representative subset of 28 C. lupini isolates was characterized based colony morphology (form, aerial mycelium, margin type and color of the reverse side) and 18 of those were further characterized for growth rate and conidial shape and size, revealing ten distinct morphotypes (A-J; Fig. 2, Table 2, Supplementary Figs. S1, S2). Despite certain variability, all observed conidia shared features typical for C. lupini (hyaline, smooth-walled, aseptate, straight and with one acute end) as described by Damm et al. 14 . Morphotype A was the most common and was observed for isolates from across the world (Europe, Australia, North-and South America), all belonging to genetic group II. Morphotypes B, C and G were observed for isolates from South Africa and morphotypes D, E, G, I and J were observed for isolates from South America.
Multi-locus phylogenetic analyses of 50 Colletotrichum isolates identified six distinct genetic groups within C. lupini (I-VI; Fig. 3, Supplementary Fig. S3). The combined sequence dataset contained 2251 characters (ITS: 1-496, GAPDH: 497-745, TUB2: 746-1200, APN/MAT1: 1201-2251) including alignment gaps. The APN/MAT1 locus showed the highest variability across the nucleotide data set, with 75.8% conserved sites for the whole data set (including out-groups) and 97.4% within C. lupini (Supplementary Table S1). The TUB2 and GAPDH loci showed 89.9% and 81.1% identical sites for the entire dataset and 97.8% and 98.4% identity within C. lupini, respectively. The ITS region showed the lowest variability with 97% identical sites across the whole dataset and 99.2% within C. lupini. As shown in Fig. 3, most C. lupini strains clustered with a high bootstrap support (BS) value of 79 and posterior probability (PP) of 1 with reference strains representing genetic group II (CBS 109221, IMI 375715 and RB221). Strains within group II showed a high identity among each other (> 99.9%) and showed morphotype A, except for Chilean strain JA15 showing morphotype D (Fig. 2). South African strain JA10 and Peruvian strain JA20, with morphotypes G and F, respectively, clustered together with a BS of 84 and PP of 1, forming a highly supported group (III). South African strains JA11 and JA12, with morphotypes C and B, respectively, clustered together with a BS of 98 and PP of 1, forming a highly supported group (IV). Ecuadorian strains JA18 and JA19 with distinct morphotypes I and J, respectively, showed 99.7% identity with reference strains of group II and clustered together with a BS of 60 ( Fig. 3, Supplementary Fig. S3) and a PP of 1 in (Fig. 3), forming a distinct group (V). The reference strains for group I (CBS 109225 with morphotype H, CBS 109226 and CBS 509.97) are clustered together with a BS of 99 and PP of 1 and show 100% identity with each other and 99.6% identity with reference strains of group II. South American strains JA21, JA22 and CBS 109216, with morphotype E, cluster together with a BS of 98 and PP of 1 (Fig. 3) and a BS of 54 ( Supplementary Fig. S3) forming a highly supported group (VI). JA21 and JA22 showed 99.8% and CBS 109216 showed 99.7% identity with reference strains of group I and 99.4% and 99.2% identity with references strains of group II, respectively. Distinct virulence patterns on white and Andean lupin. Virulence assays performed on two white lupin (L. albus L.) accessions (Feodora and Blu-25) and two Andean lupin (L. mutabilis Sweet.) accessions (LUP 17 and LUP 100) with strains representing the different morphotypes and genetic groups indicated in Fig. 3, revealed strong strain (p < 0.0001), lupin species (p < 0.0001) and strain × lupin species interaction effects (p < 0.0001). A strong accession effect was found within white lupin (p < 0.0001), whereas for Andean lupin there was no significant accession effect (p = 0.43). Strain (p < 0.0001) and strain × accession (p < 0.0001) interaction effects were found for both species. Strains belonging to genetic group II with morphotype A, caused severe disease on white lupin accession Feodora and both Andean lupin accessions ( Supplementary Fig. S4), showing standardized area under the disease progress curve (sAUDPC) means ranging from 3.95 to 5 (Fig. 4). On the more tolerant white lupin accession Blu-25, sAUDPC means for strains of group II with morphology A were more variable, with JA01 and IMI 375715 showing moderate (2.7-2.9) and Chilean strains JA16 and 17 showing high (3.8-4.1) virulence. Chilean strain JA15, also belonging to genetic group II but with a different morphology (D), caused low disease on LUP 100 and Blu-25 (1.9), showing a different virulence spectrum compared to the other tested strains of genetic group II. South African strains JA11 and JA12, belonging to genetic group IV with morphotypes C and B, respectively, showed a similar virulence spectrum on white lupin as strains of group II. JA10 and JA20, representing group III and morphotype G and F, respectively, were overall avirulent (< 2), with the exception of JA10 on Feodora, showing moderate virulence (2.95). Peruvian strain JA21, representing genetic group VI and morphotype E, caused low disease on white lupin (1.4-1.8), but severe disease on Andean lupin (4.25-5). A similar observation was found for the two Ecuadorian strains JA18 and JA19 of genetic group V and morphotypes I and J, respectively. These two strains caused low disease on white lupin and high disease on Andean lupin LUP 100. On Andean lupin LUP 17, however, a severe disease phenotype was only found for JA18 (3.6), whereas JA19 barely caused any disease symptoms (1.25). Similar to the observations for JA19, the Ukrainian strain CBS 109225 (genetic group I, morphotype H) caused severe disease on Andean lupin LUP 100 (3.36) and low disease on Andean lupin LUP 17 and white lupin (1.2-2). The C. tamarilloi and C. acutatum strains were avirulent across the lupin accessions (< 1.26).

Discussion
This study compared 39 C. lupini and 11 Colletotrichum spp. isolates collected from across the world to explore intraspecific diversity of C. lupini and to better understand the dynamics of the current lupin anthracnose pandemic and potential implications of further migrations of distinct pathogenic strains. Based on multi-locus phylogeny supported by isolate morphology, we identified four distinct genetic groups additional to previously www.nature.com/scientificreports/ described genetic groups I and II. Highest intraspecific diversity was identified among C. lupini isolates collected from across the South American Andes region. This is in line with reports of Falconí et al. 41 and Riegel et al. 40 showing high diversity in Ecuador and Chile, respectively. In those regions, Andean lupin has been cultivated for more than 2000 years 42 growing alongside numerous wild lupin species 43 . Isolates collected in South Africa showed a distinct morphology and virulence spectrum, indicating higher diversity than previously shown 44 .
Although lupins form a significant part of the local agriculture and have been researched there since at least 1897 45 , they are not native to South Africa and lupin anthracnose was not reported in South Africa until 1993 46 .
Taking into account the relatively recent reports of anthracnose in South Africa, the low diversity in Europe and Australia and the center of origin for species within clade 1 of the C. acutatum species complex being in South America 10,15 , we consider the South American Andes to be the center of origin of C. lupini. The majority of the C. lupini isolates (26 out of 39) belong to the highly virulent genetic group II, showing morphotype A, and were collected in Europe, Australia, South Africa, the USA and Chile. This result confirms previous reports classifying most C. lupini strains from across the world in the same genetic group 14,39,[47][48][49] . The low genetic diversity among strains of group II, the uniform morphology and non-observed sexual morph 14 indicates clonality as suggested by Talhinhas et al. 20 . Pathogenicity of group II strains has also been shown on blue 28 , yellow 32 and various other lupin species across the world 20 , indicating a broad host range within the genus Lupinus. Reports from South Korea and China indicate that group II strains also cause disease in those regions 50,51 , highlighting that these strains are globally widespread and are the cause of the current anthracnose pandemic in lupin. The group II strain RB221 can be used as reference, as it is now fully sequenced 52 and tested on both Andean and white lupin 53 .
The stem-wound inoculation assay used in this study was previously described to be highly reproducible and strongly correlated to field performance under natural infection pressure 26 . In the present study, virulence assays based on stem-wounding showed strong strain x accession interaction effects for white and Andean lupin, suggesting a strain-dependent host spectrum and the existence of different physiological races within C. lupini. Similar observations were described by Falconí et al. 41 , showing a C. lupini strain x Andean lupin accession interaction effect. The existence of physiological races has been observed for various Colletotrichum species, such as for C. lindemuthianum on common bean 54 , C. sublineola on sorghum 55 and C. truncatum on lentil 56 , but, in general, this is not common within the genus Colletotrichum. The similar virulence levels of isolates belonging to group II observed on Andean and white lupin accessions are in line with Alkemade et al. 26 , in which equal virulence was observed for IMI 375715 (Australia) and JA01 (Switzerland) when inoculated on six different white lupin accessions. However, an exception within group II is Chilean strain JA15, which, besides having a distinct morphology, was less virulent on Andean lupin LUP 100 and white lupin Blu-25. Further, Chilean strains JA16 and JA17 (also group II) overcame resistance of the resistant advanced breeding line Blu-25, which has been specifically bred for anthracnose resistance in Chile and was shown resistant under Swiss field conditions 26 . These results indicate that new introductions of highly virulent foreign strains can have severe consequences as seen for many other crops [57][58][59] and it should be investigated if this high virulence is also affecting other resistant (white) lupin material 26,31,60 . Although disease development after stem-wounding of seedlings correlated strongly to field disease scores of mature plants 26 , we cannot exclude the possibility that conclusions drawn on virulence level might differ for secondary infection processes (e.g. via rain splash).
This study provides first solid evidence that, based on multi-locus phylogeny and morphology, genetic diversity within C. lupini is higher than previously shown. High-resolution genome-wide sequencing and an increased sampling density from especially the South American Andes region are now necessary to increase genetic resolution and to better understand C. lupini phylogeny and phylogeography. This could provide the basis for in-depth comparative genomic studies to identify effector gene clusters within the C. lupini genome. This study confirms that the current lupin anthracnose pandemic is caused by a genetically uniform group of highly virulent strains. The identification of strains with an increased virulence on tolerant white lupin breeding material and the observation of strain-specific virulence patterns should be taken into account in lupin resistance breeding programs. Table 1. Isolation details and GenBank accessions of Colletotrichum strains used in this study. JA strains from the FiBL culture collection characterized in this study, RB personal collection of Riccardo Baroncelli described in Dubrulle et al. 39 , CBS collection of the Westerdijk Fungal Biodiversity Institute, Utrecht, The Netherlands, IMI Culture collection of CABI Europe UK Centre, Egham, UK, ITS internal transcribed spacers 1 and 2 together with 5.8S nrDNA, GAPDH glyceraldehyde-3-phosphate dehydrogenase, TUB2 β-tubulin 2, APN/ MAT1 Apn2-Mat1-2-1 intergenic. Codes in bold were used for morphology analysis in this study. Accession numbers in bold are newly determined in this study.   Table 2). Strain codes are followed by country of origin and roman numbers (I-VI) indicate genetic groups. Fungal isolation and culture conditions. Symptomatic (dried) lupin stem or pod tissue (Fig. 1 Table 2. Growth rate, conidial size and shape, and colony morphology for the different morphotypes observed within Colletotrichum lupini. L length, W width. a Mean ± SD, see also Supplementary Fig. S2. b Observed conidia were rather variable in shape and size, but all conidia were hyaline, smooth-walled, aseptate, straight, with one end round and one end acute as described for Colletotrichum lupini in Damm et al. 14

Morphology.
A globally representative subset of 28 C. lupini isolates was characterized based on colony morphology (form, aerial mycelium, margin type and color of the reverse side). From those, a subset of 18 isolates was further characterized for growth rate (mm/day), and conidial shape and size 19 . Isolates were subcultured by placing a droplet of 5 μl spore suspension in the middle of three PDA plates and grown for 14 days at 22 °C in the dark. Culture diameter was recorded every 3 days. Photographs were taken from the front and reverse sides of the PDA plates after 14 days of incubation. Conidia were collected with a sterile spreader after flooding the Petri plate with 2 ml sterile ddH 2 0, the spore suspension was filtered with sterile cheese cloth and microscopic slides were prepared with sterile ddH 2 O. Conidia morphology was observed using light microscopy (DM2000-LED, Leica Microsystems, Wetzlar, Germany) equipped with a high definition camera (Gryphax Subra, Jenoptik AG, Jena, Germany). A minimum of at least 50 measurements were performed to determine conidia length and width. DNA extraction, PCR amplification and sequencing. Mycelium from single-spore cultures was collected after 7-10 days on PDA at 22 °C with a sterile spreader after flooding the Petri dish with 2 ml sterile ddH 2 0. Genomic DNA was isolated with a CTAB extraction protocol 63 . Partial gene sequences were determined for the internal transcribed spacer (ITS) region using primers ITS5 and ITS4 64 , the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene using primers GDF1 and GDR1 65 , the β-tubulin 2 (TUB2) gene using primers   71 using a Markov Chain Monte Carlo (MCMC) algorithm using four chains and starting from a random tree topology. Substitution models for each locus were included for each partition. The analysis ran for 500,000 generations with trees sampled every 1000 generations to reach average standard deviations of split frequencies below 0.01. The first 25% of saved trees were discarded at the 'burn-in' phase and the 50% consensus trees and posterior probabilities (PP) were determined from the remaining trees. Bootstrap Anthracnose severity is expressed in standardized area under the disease progress curve (sAUDPC) and estimated means are shown. Strain codes are followed by abbreviated country of origin and morphotype (A-J). Different capital letters above bars indicate significant differences between strains (Tuckey-HSD, p < 0.05). Error bars indicate the standard error of the estimated mean.
Scientific Reports | (2021) 11:13547 | https://doi.org/10.1038/s41598-021-92953-y www.nature.com/scientificreports/ support values (BS) from the ML analysis were plotted on the Bayesian phylogeny. Further phylogenetic analyses were performed with the unweighted pair group method with arithmetic mean (UPGMA) with 10,000 replicates in Mega X. All generated sequences were deposited in GenBank (Table 1) and alignments and trees in TreeBASE.
Virulence. Virulence tests were performed on white and Andean lupin with representative C. lupini strains (see Fig. 3), C. tamarilloi strain CBS 129814 and C. acutatum strain CBS 369.73 through stem-wound inoculation as described by Alkemade et al. 26 , which was shown to highly correspond to field performance in Switzerland (r = 0.95). Disease scores ranging from 1 (non-pathogenic), 2 (low virulence) to 9 (highly virulent) were taken 4, 7 and 10 days post inoculation (dpi) and the standardized area under the disease progress curve was calculated (sAUDPC) 26 . All inoculations were performed in a growth chamber (25 ± 2 °C, 16 h light and ~ 70% relative humidity) in a completely randomized block design with a minimum of six replicates per experiment.
Statistical analysis. Statistical analyses were performed with R 4.0.3 using the packages lme4 72 , lmerTest 73 and emmeans 74 , following a mixed model with factors of interest (i.e. strain, lupin species, lupin accession) as fixed and replicated block nested in experiment as random factor. Datasets that did not follow assumptions of normality of residuals and homogeneity of variance were log10 transformed. Data are presented as estimated least-squares means using the aforementioned mixed model. A Tukey-HSD test (p ≤ 0.05) was applied for pairwise mean comparison of the different Colletotrichum strains within each lupin accession.

Data availability
The data that support the findings of this study are shown in this manuscript or, in the case of new sequences data, are openly available in Genbank at https:// www. ncbi. nlm. nih. gov/ genba nk/ (for reference numbers see Table 1) and in Treebase at http:// purl. org/ phylo/ treeb ase/ phylo ws/ study/ TB2: S27356? x-access-code= 26013 6f8e6 416a0 614b9 3528d dbfe0 ef& format= html.