To identify the presence and distribution of Alternaria spp., a 3-year survey was conducted from 2016 to 2018. Potato plants with early blight symptoms were collected from large commercial fields and from a small farmer’s garden. Samples were collected in nine different districts of Serbia: North Bačka, South Bačka, Pomoravlje, Moravica, Raška, Jablanica, Zlatibor, Rasina, and City of Belgrade district. Sixty-one fields in 20 different localities were surveyed (Table 1). Infected plants were sampled at the production plots in two diagonal transects, and two leaves per plant were collected from 10 plants at equal distances across a diagonal transect of the field during early blight disease epidemics between July and September. Samples with dark and circular lesions with typical concentric rings were collected. For Alternaria isolation, necrotic leaf cuts from a margin of the lesion were surface disinfected in 1% sodium hypochlorite, rinsed with sterile distilled water, transferred to Petri dishes containing potato dextrose agar (PDA, Difco, Detroit, MI), and incubated at 23°C. The morphological characteristics of the isolates were studied by taking a 5 mm mycelial plug from the growing colony margin of a 5-day-old culture and subculturing on V8 medium (200 ml V8 juice Campbell; 3 g CaCO₃; 15 g agar; and 800 ml of sterile distilled water). The cultures were placed under cool white fluorescent light with 12 h/12 h periods of light/dark and incubated on 23°C for 5 days (Simmons, 2007).

The growing cultures of all isolates were studied with a protocol described by Simmons (2007). Conidial morphology was examined using a microscope (Olympus BX51TF, Japan) at 400 × magnification. Dimensions are based on the observation of 100 conidia and 50 conidiophores per isolate. The conidia length and sporulation pattern data were analyzed using the analysis of variance (ANOVA; p < 0.05) in the statistical software package SPSS version 20.0 (SPSS Inc., Chicago, IL, United States).

For molecular identification, 30 mg of dry weight mycelium was collected from PDA plates and used for DNA extraction according to the manufacturer’s instructions of the DNeasy Plant Mini Kit (Qiagen, Valencia, CA, United States). Polymerase chain reaction (PCR) was used with the specific primers OAsF7/OAsR6 which amplified a 164 bp fragment from the Alt a1 gene to separate the species A. solani and related species (A. grandis and A. protenta) from A. linariae, and with the primer pair OAtF4/OAtR2, which amplified a 483 bp fragment from the calmodulin-encoding gene of A. linariae (Gannibal et al., 2014; Supplementary Table 1). PCR was performed in a Mastercycler Nexus GSX1 (Eppendorf, Hamburg, Germany) with the amplification program as described in Gannibal et al. (2014). The PCR amplicons were separated by electrophoresis on 1% agarose gels run in 1 × TBE buffer at 90 V constant voltages. The gels were stained with ethidium bromide and visualized under UV light. To differentiate between A. grandis and A. solani (including A. protenta), PCR identifications of isolates that amplified a 164 bp fragment with the primer pair OasF7/OasR6 were complemented by RFLP of the calmodulin gene that was amplified using the primer pair CALDF1/CALDR1 (Lawrence et al., 2013). The resulting PCR products were cut using double restriction with enzymes HaeII and RsaI (New England Biolabs) according to Ayad et al., (2017a). RFLP patterns were discriminated by the size of the larger fragment, 420 bp for A. solani and 292 bp for A. grandis. To distinguish between A. solani and A. protenta isolates, sequencing of the Rpb2 locus was performed.

Multilocus sequence analysis was performed for all 230 isolates with amplification of three housekeeping genes by primers RPB2DF/RPB2DR (Lawrence et al., 2013) for the (Rpb2) gene, primer pair CALDF1/CALDR1 for the amplification of Calmodulin gene (Lawrence et al., 2013), and by using GPD1/GPD2 primers for the amplification of (GAPDH; Berbee et al., 1999; Supplementary Table 1). The PCR amplification conditions for RPB2 and calmodulin genes were described in Lawrence et al., (2013), whereas GAPDH gene amplification was conducted using the PCR conditions described in Berbee et al., (1999). The PCRs were conducted in a Mastercycler Nexus GSX1 (Eppendorf, Hamburg, Germany) in a 25 μl reaction mixture using the following final concentrations or total amounts: 5 ng DNA, 1 × PCR buffer (20 mM Tris/HCl pH 8.4 and 50 mM KCl), 1 μM of each primer, 2.5 mM MgCl₂, 0.25 mM of each dNTP, and 1 unit of Taq polymerase (Fermentas, Vilnus, Lithuania). The amplified PCR products were separated by electrophoresis and purified (QIAquick PCR Purification Kit, Qiagen, Valencia, CA, United States) for Sanger sequencing in both directions (Macrogene, Seoul, Korea). For final identification, the DNA sequences of all isolates were compared with reference sequences of Alternaria species using the BLAST program of NCBI. From DNA sequences of 230 isolates, 37 were selected based on the species and geographic location and as representative strains these sequences were deposited into NCBI GenBank (Supplementary Table 2).

Partial gene sequences of GAPDH (569 bp), Rpb2 (772 bp), and calmodulin (778 bp) were concatenated (2,119 bp) to conduct a multilocus phylogenetic tree. Analysis is done after the gene sequences are concatenated head to tail to form a gene alignment which is then analyzed to generate the species tree. Manual corrections, alignments, and comparisons of the aligned database were performed using CLUSTALW integrated into MEGA version 11.0 software (Tamura et al., 2021). The gamma distributed Tamura-Nei model determined by ModelTest implemented in MEGA11 was used as the best fitting model of nucleotide substitution. The reliability of the obtained trees was evaluated using 1,000 bootstrap replicates and bootstrap confidence values <50% were omitted. Phylogenetic trees were constructed using the maximum likelihood (ML) method implemented in the same software.

The phylogenetic relationships among Alternaria spp. from potato were confirmed by Bayesian phylogeny. For concatenated data, the Bayesian information criterion (BIC) was conducted using the best fitting substitution model according to jModeltest 2.1.7 (Darriba et al., 2012). The Bayesian analysis was performed in MrBayes 3.1.2 (Huelsenbeck and Ronquist, 2003) for 1 million generations, sampling every 100 trees, using simultaneous Markov chain Monte Carlo (MCMC) runs. The standard deviation of the split frequencies was checked until it reached a value below 0.01 (default burnin = 25%). The convergence of the MCMC chains and their stationarity were confirmed using Tracer 1.5 (Rambaut and Drummond, 2009). The phylogenetic tree was observed and printed using FigTree 1.4 (Rambaut and Drummond, 2012).

Pathogenicity tests were conducted with all investigated isolates by inoculating leaves of potato plants (Carrera variety) in pot experiments. Potato seed tubers were sown in pots filled with sterilized soil in a glass house at 25°C with an 8 h/16 h photoperiod. One month old plants were used for inoculations. The conidial suspensions prepared from cultures grown on V8 agar at 23°C for 5 days were adjusted to 1 × 10⁶ conidia/ml, and a 20 μl drop was applied to the leaf surface. Control plants were treated at the same time using sterile distilled water. After inoculation, plants were placed in a growing chamber with relative humidity between 95 and 100% for one week at 23°C. The test was repeated three times. Fragments of lesion areas were cut and surface sterilized before being placed into V8 agar. The morphological characteristics of reisolated pathogens were compared with original isolates to confirm Koch’s postulates. Inoculated leaves were cut from the tested potato plants, and the necrotic leaf area was measured using the program ImageJ (Schneider et al., 2012). In the pathogenicity test, the Kruskal–Wallis test was used in the statistical software package SPSS version 20.0 (SPSS Inc., Chicago, IL, United States).

Throughout a three-year survey (2016–2018), the presence of early blight was confirmed in all surveyed localities and four large-spored Alternaria pathogens: A. solani, A. protenta, A. grandis, and A. linariae were detected on potatoes in Serbia (Figures 1, 2). In total, 230 Alternaria spp. isolates were recovered from symptomatic potato leaves collected in 40 different potato fields (Table 1). Most of the isolated pathogens belonged to A. solani, accounting for 60% of the overall isolates. A. solani was not recorded in five localities: Borča (district City of Belgrade), Velika Drenova (district Rasina), Bzovik (district Raška) Navalin, and Pržine (district Jablanica). A. protenta was detected in 33% of the isolates in all major potato production areas, except in the localities Velika Drenova (district Rasina), Rudno (district Raška), Pečenjevce (district Jablanica), Krivača, Zablaće, Katići (district Moravica), Drmanovići, and Blato (district Zlatibor). A. grandis and A. linariae were identified in 4 and 3%, respectively. A. grandis was found in Krivača and Zablaće (district Moravica), while A. linariae was detected in Velika Drenova (district Rasina), Dragocvet (district Pomoravlje), and Borča (district City of Belgrade).

Morphological preliminary identification of fungi was performed 7 days after incubation on V-8 medium, and large-spored Alternaria isolates revealed slight variability in colony morphology. According to Simmons (2007), isolated species were identified by their spore size, and the isolates could be separated into four species: A. solani, A. protenta, A. grandis, and A. linariae.

A. solani isolates produced colonies with conidia dimensions as follows: 190 ± 24 μm in length and 16 ± 2 μm in width with 5–13 transverse and up to three longitudinal septa and branching conidiophores 187.7 ± 4.8 μm with numerous conidiogenous sites and spores. Isolates of A. protenta developed conidia 225 ± 23 μm in length and 17 ± 2 μm in width with 8–13 transverse septa and one longitudinal septum in a few transverse segments. Conidiophores were solitary, reaching 83 μm in primary growth up to 210 μm in the stage of secondary branching (137.7 ± 6.9 μm). A. linariae isolates had solitary conidiophores up to 142 μm long (133.7 ± 7.5 μm), rising straight from the surface with large conidia. The size of the spores was 255 ± 33 μm in length and 18 ± 2 μm in width with 1–12 transverse septa and most commonly only one longitudinal septum in many transverse segments. Isolates identified as A. grandis produced conidia 262 ± 36 μm in length and 16 ± 3 μm in width with 7–12 transverse septa and one longitudinal septum in a few transverse segments. Conidiophores arose as singly, up to 200 μm in length (157.2 ± 9.3 μm).

Assumptions for conducting statistical analyses were inspected using Shapiro–Wilk’s normality test, Box’s test of equality of covariance matrices and Levene’s test of equality of error variances. Data showed adequate values for conducting parameter statistical test one-way ANOVA for conidiophores and conidial length but not for conidial width so this category was excluded from the further statistical analyses. A significant difference among the identified groups was found only for conidial length p = 0.002. Parameters for conidial length indicated again that A. solani and A. protenta did not show significant differences in conidial size (p = 0.122).

To confirm morphological identification of Alternaria isolates, the primer pair OAsF7/OAsR6 amplified the DNA fragment at position 164 bp for 222 isolates (Supplementary Figure 1A). The second primer pair OAtF4/OAtR2 was used to confirm that A. tomatophila amplified a 483 bp fragment for eight isolates (Supplementary Figure 1B). For species differentiation between A. solani/A.protenta and A. grandis isolates, RFLP was performed with HaeII and RsaI restriction enzyme digestion of PCR products of the calmodulin gene. The restriction pattern of the A. solani/A.protenta isolates exhibited a larger fragment at 420 bp; in contrast, the restriction pattern of the A. grandis isolates exhibited a larger fragment at 292 bp (Supplementary Figure 1C). Additionally, sequencing of the Rpb2 gene was used to differentiate A. solani and A. protenta isolates. BLAST analyses of the Rpb2 gene separated A. protenta from A. solani isolates and showed that A. protenta isolates had 100% nucleotide sequence identity with the corresponding gene regions of A. protenta strain CBS 116696. Molecular identification resulted in 138 A. solani isolates, 74 A. protenta isolates, 10 A. grandis isolates, and 8 A. linariae isolates collected from potato in Serbia. After sequences analysis of all 230 isolates, DNA sequences of the GAPDH, calmodulin, and Rpb2 genes were identical for the same species and we selected 37 isolates based on species and geographic location to be representative for phylogenetic analysis.

Multilocus sequence analysis of each individual sequence of the GAPDH, calmodulin, and Rpb2 genes and comparing all investigated isolates with reference GenBank sequences (A. linariae CBS 109156, A. protenta CBS 116696, A. solani CBS 109157, and A. grandis CBS 109158), identified four different species distributed in four clusters (Figure 3; Supplementary Figures 2–4). Phylogenetic trees constructed for GAPDH and calmodulin genes were similar and grouped isolates into three diverse clusters that identified isolates as A. solani/A. protenta, A. grandis, and A. linariae, while phylogenetic trees constructed for the Rpb2 gene followed the results of previous identification and identified four different species.

PCR products of the calmodulin regions revealed two clades for A. grandis and A. linariae. The first cluster, A. linariae, separated as a monophyletic lineage with a bootstrap value of 96%, segregated four isolates into two subclusters along with the reference strains. One clade grouped isolates IZB 136A and IZB 36c-5 with reference strains CBS 109156, MF-P138061, and DA 100 obtained from the NCBI database, and the second clade grouped isolates IZB 24–1 and IZB 24–11 with reference strain DA 119. The cluster that included A. grandis isolates was divided into two subclusters, one with two isolates, IZB 101A and IZB 54–11, and the reference strains NB 251 and DA 009. The second subcluster consisted of isolates IZB 107A, IZB 107-2A, and IZB 107-3A and reference strains DA 052 and CBS 109158.

Concatenated gene sequences of the GAPDH, calmodulin, and Rpb2 genes were used to substantiate the results of separate gene analysis. Phylogenetic analyses by the Bayesian information criterion (BIC) employed in jModelTest confessed a general time reversible model with gamma distribution (TIM3 + IG) as the most suitable substitution model. The distance from the end to the end of the concatenated sequences used in this testing was 2,135 bp. Concatenated sequences of three analyzed gene regions, GAPDH, calmodulin, and Rpb2, in supplementary common phylogeny analysis were recognized as a single multilocus dataset for all isolated species. The parameters for the Bayesian analyses were Lset base = (0.2378 0.3131 0.2471) nst = 2 tratio = 1.3022 rates = equal pinvar = 0. For second phylogenetic analyses of GAPDH, calmodulin, and Rpb2 concatenated sequences, maximum likelihood analysis was employed and the T92 model was selected as the best-fit model.

After seven days of incubation, typical early blight disease symptoms were observed on inoculated potato plants for all tested isolates with differences in symptom appearance (Figure 4). To confirm Koch’s postulates, the pathogen was reisolated, and based on colonies and the morphology of conidia, recovered isolates were identical as to the original isolates.

During the first week after inoculation, the development of lesions was slow, and spots were small (≤ to 5 mm in diameter). In the second week after inoculation, symptoms appeared as brown to dark spots on leaves similar to those observed in the potato fields (Figure 4). Although inoculations with A. solani, A. protenta, A. grandis, and A. linariae isolates resulted in disease symptoms visible one week after inoculation, leaf spots differentiated in color and diameter. Concentric rings with dark color and small diameter were characteristic of isolates of A. grandis and A. linariae. A. solani was the most aggressive, with the largest spots and dark color, whereas A. protenta produced leaf spots light brown to dark in color. No symptoms were observed on plants inoculated with sterile distilled water.

A Kruskal–Wallis test was conducted to determine if there were differences in aggressiveness between isolated Alternaria species on potato leaves. The distributions of necrotic leaf area (%) were not similar for all groups, as assessed by visual inspection of a boxplot. The necrotic areas of potato leaves were significantly different between the different species, χ²(3) = 39.142, p = 0.000. Subsequently, pairwise comparisons were performed using Dunn’s (1964) procedure with a Bonferroni correction for multiple comparisons. This post-hoc analysis revealed statistically significant differences in aggressiveness among A. solani, the most pathogenic species (32.1 ± 6.5), and A. linariae, the weakest pathogen (16.8 ± 4.8) compared to other species; however, A. protenta and A. grandis revealed moderate pathogenicity (25.2 ± 6.2 and 23.9 ± 5.3, respectively) with no statistically significant difference in pathogenicity on potato (p = 0.812; Figure 5).