A selection of ten isolates of P. fragariae were sourced from the Atlantic Food and Horticulture Research Centre (AFHRC), Nova Scotia, Canada. An additional isolate of P. fragariae, SCRP245, alongside three P. rubi isolates, were sourced from the James Hutton Institute (JHI), Dundee, Scotland (detailed in Table 1).

All work with P. fragariae and P. rubi was conducted in a Tri-MAT Class-II microbiological safety cabinet. Isolates were routinely subcultured on kidney bean agar (KBA) produced as previously described (Maas, 1972). Isolates were grown by transferring two pieces of between 1 and 4 mm² onto fresh KBA plates, subsequently sealed with Parafilm^®. Plates were then grown at 20°C in the dark in a Panasonic MIR-254 cooled incubator for between 7 and 14 days.

Mycelia were also grown in liquid pea broth, produced as previously described (Campbell et al., 1989) with the addition of 10 g/L sucrose. These plates were inoculated with five pieces (1 – 4 mm²) of mycelium and media, subsequently sealed with Parafilm^®. These were grown at 20°C for 4 to 5 days in constant darkness.

Mother stock plants of F. × ananassa were maintained in 1 L pots in polytunnels. Runner plants were pinned down into 9 cm diameter pots filled with autoclaved 1:1 peat-based compost:sand. The clones were grown on for 3 weeks to establish their own root system and then were cut from the mother plant. Inoculations were performed as described previously (van de Weg et al., 1996). Plants were placed in a growth chamber with 16/8 h light/dark cycle at a constant 15°C. Inoculated plants stood in a shallow layer of tap water (2 – 7 mm) for the entire experiment and were watered from above twice a week. After 6 weeks, plants were dug up and the roots were rinsed to remove the soil/sand mix. Roots were then assessed for distinctive disease symptoms of “rat’s tails”, which is the dieback from the root tip and “red core”, which is the red discolouration of the internal root visible when longitudinally sliced open. Samples for which infection was unclear were visualised under a light microscope for the presence of oospores. This was performed through squash-mounting of the root tissue, where a sample of root was excised and pressed between a microscope slide and cover slip. This sample was then examined at 40x magnification under high light intensity with a Leitz Dialux 20 light microscope.

For Illumina sequencing, gDNA was extracted from 300 mg of freeze-dried mycelium using the Macherey-Nagel NucleoSpin^® Plant II Kit. The manufacturer’s protocol was modified by doubling the amount of lysis buffer PL1 used, increasing the incubation following RNase A addition by 5 min, doubling the volume of buffer PC and eluting in two steps with 35 μL of buffer PE warmed to 70°C. For PacBio and Oxford Nanopore Technologies (ONT) sequencing, gDNA was extracted using the Genomic-Tip DNA 100/G extraction kit, following the Tissue Sample method.

To create Illumina PCR-free libraries, DNA was sonicated using a Covaris M220 and size-selected using a BluePippin BDF1510 1.5% gel selecting for 550 bp. Libraries were constructed using the NEBNext enzymes: End repair module (E6050), A-Tailing module (E6053), Blunt T/A ligase (M0367) and Illumina single-indexed adapters. Sequencing was performed to generate 2 × 300 bp reads on a MiSeq^TM system using MiSeq Reagent Kit V3 600 cycle (MS-102-3003). PacBio library preparation and sequencing was performed by the Earlham Institute, United Kingdom on a PacBio RS II machine. ONT sequencing libraries were created using the SQK-LSK108 kit following the manufacturer’s protocol and sequenced using the FAH69834 FLO-MIN106 flow cell on a GridION for approximately 28 h.

Growth of isolates for inoculations were performed on fresh KBA plates for approximately 14 days as described above, before the mycelium had reached the plate edge. Plugs of mycelium growing on agar were taken using a flame sterilised 10 mm diameter cork borer and plugs were submerged in chilled stream water. Plates were incubated in constant light for 3 days at 13°C, with the water changed every 24 h. For the final 24 h, chilled Petri’s solution (Judelson et al., 1993) was used. Roots of micropropagated F. × ananassa “Hapil” plants (GenTech, Dundee, United Kingdom), maintained on Arabidopsis thaliana salts (ATS) media (Taylor et al., 2016), were submerged for 1 h before transfering back to ATS plates. These plates were kept at 15°C, with 16/8 h light/dark cycle, with a photosynthetic photon flux (PPF) of 150 μmol m^–2 s^–1 provided by fluorescent lamps (FL40SSENW37), in a Panasonic MLR-325H controlled environment chamber. Root tissue was harvested at a selection of time points post-inoculation by rinsing root tissue successively in three beakers of sterile dH₂O to remove all media. Roots were separated below the crown tissue, flash frozen in liquid nitrogen and stored at −80°C.

Extraction of total RNA from inoculated root tissue was performed similarly to the previously described 3% CTAB₃ method (Yu et al., 2012). Briefly, plant material was disrupted under liquid nitrogen in a mortar and pestle, previously decontaminated through cleaning with RNaseZap^TM solution and baking for 2 h at 230°C to deactivate RNAse enzymes. This material was transferred to a warmed extraction buffer (Yu et al., 2012) containing β-mercaptoethanol with 0.01 g of PVPP added per 0.1 g of frozen tissue. The remaining steps were performed as previously described (Yu et al., 2012), except that 60 μL DEPC-treated H₂O was used to elute RNA.

Mycelium was grown in liquid pea both and dried on a Q100 90 mm filter paper (Fisher Scientific) using a Büchner funnel and a Büchner flask attached to a vacuum pump. Total RNA was then extracted using the QIAGEN RNEasy Plant Mini kit with the RLC buffer. Extraction was carried out following the manufacturer’s instructions with an additional spin to remove residual ethanol.

RNA was checked for quality using a NanoDrop 1000 spectrophotometer, quantity with a Qubit 2.0 Fluorometer and integrity with a TapeStation 4200 before being prepared for sequencing via Reverse Transcription-Polymerase Chain Reaction (RT-PCR) and sequencing on an Illumina HiSeq^TM 4000 by Novogene, Hong Kong, Special Administrative Region, China. Timepoints for sequencing were selected through the detection of β-tubulin transcripts by RT-PCR. Reverse transcription was performed with the SuperScript^TM III Reverse Transcriptase kit with an equal amount of RNA used for each sample. The complementary DNA (cDNA) was then analysed by PCR with 200 μM dNTPs, 0.2 μM of each primer (detailed in Supplementary Table S1), 2 μL of cDNA template and 2.5 units of Taq DNA polymerase and the buffer supplied in a 20 μL reaction. Reactions were conducted in a Veriti 96-well thermocycler with an initial denaturation step at 95°C for 30 s, followed by 35 cycles of a denaturation step at 95°C for 30 s, an annealing temperature of 60°C for 30 s and an extension step of 72°C for 30 s. This was followed by a final extension step of 72°C for 5 min and held at 10°C. Products were visualised by gel electrophoresis on a 1% w/v agarose gel at 80 V for 90 min, stained with GelRed. Following this, three biological replicates of samples taken at: 24, 48, and 96 hpi for BC-16, 48 hpi for BC-1 and 72 hpi for NOV-9 were sequenced (Supplementary Figure S1). Additionally, in vitro mycelial RNA was sequenced.

Phytophthora rubi RNA-Seq reads were sequenced on a HiSeq^TM 2000 system.

Prior to assembly, Illumina reads were cleaned and sequencing adaptors were removed with fastq-mcf (Aronesty, 2013). Quality control of PacBio data was performed by the Earlham Institute, Norwich, United Kingdom. ONT reads were basecalled with Albacore version 2.2.7 and adaptors were removed with Porechop version 0.2.0 (Wick, 2018) and the trimmed reads were corrected with Canu version 1.4 (Koren et al., 2017).

Assemblies of isolates sequenced solely with Illumina data were generated with SPAdes version 3.11.0 (Bankevich et al., 2012) with Kmer sizes of 21, 33, 55, 77, 99, and 127. PacBio reads were assembled using FALCON-Unzip (Chin et al., 2016). FALCON version 0.7 + git.7a6ac0d 8e8492c64733a997d72a9359e1275bb57 was used, followed by FALCON-Unzip version 0.4.0 and Quiver (Chin et al., 2013). Error corrected ONT reads were assembled using SMARTdenovo version 1.0.0 (Ruan, 2018).

Following assembly, error correction was performed on ONT assemblies by aligning the reads to the assembly with Minimap2 version 2.8r711-dirty (Li, 2018) to inform ten iterations of Racon version 1.3.1 (Vaser et al., 2017). Following this, reads were again mapped to the assembly and errors were corrected with Nanopolish version 0.9.0 (Simpson, 2018). PacBio and ONT assemblies had Illumina reads mapped with Bowtie 2 version 2.2.6 (Langmead and Salzberg, 2012) and SAMtools version 1.5 (Li et al., 2009) to allow for error correction with ten iterations of Pilon version 1.17 (Walker et al., 2014).

Following assembly, all contigs smaller than 500 bp were discarded and assembly statistics were collected using Quast version 3.0 (Gurevich et al., 2013). BUSCO statistics were collected with BUSCO version 3.0.1 (Simão et al., 2015) using the eukaryota_odb9 database on the assemblies, as the stramenopile database was not available at the time of the analysis. Identification of repetitive sequences was performed with Repeatmasker version open-4.0.5 (Smit et al., 2015) and RepeatModeler version 1.73 (Smit and Hubley, 2015). Transposon related sequences were identified with TransposonPSI release 22nd August 2010 (Haas, 2010).

Gene and effector prediction was performed similarly to a previously described method (Armitage et al., 2018). Firstly, raw RNA-Seq reads of BC-16 from both mycelial samples and the inoculation time course were cleaned with fastq-mcf (Aronesty, 2013). Reads from the inoculation time course were first aligned to the Fragaria vesca version 1.1 genome (Shulaev et al., 2011) with STAR version 2.5.3a (Dobin et al., 2013) and unmapped reads were kept. These unmapped reads and those from in vitro mycelium were mapped to the assembled P. fragariae genomes with STAR (Dobin et al., 2013). RNA-Seq data from P. rubi were aligned to P. rubi assemblies with the same method. Further steps were performed as previously described (Armitage et al., 2018). Additionally, putative apoplastic effectors were identified through the use of ApoplastP (Sperschneider et al., 2018). Statistical significance of the differences between predicted numbers of effector genes was assessed with a Welch two sample t-test in R version 3.4.3 (R Core Team, 2017). Supporting annotation data for UK2 isolate BC-16 (Supplementary Table S2).

Orthologue identification was performed using OrthoFinder version 1.1.10 (Emms and Kelly, 2015) on predicted proteins from all sequenced P. fragariae and P. rubi isolates (Supplementary Table S3). Orthogroups were investigated for expanded and unique groups for races UK1, UK2 and UK3. Unique orthogroups were those containing proteins from only one race and expanded orthogroups were those containing more proteins from a specific race than other races. Venn diagrams were created with the VennDiagram R package version 1.6.20 (Chen and Boutros, 2011) in R version 3.2.5 (R Core Team, 2016).

Variant sites were identified through the use of the GATK version 3.6 HaplotypeCaller in diploid mode (McKenna et al., 2010) following alignment of Illumina reads for all sequenced isolates to the FALCON-Unzip assembled genome with Bowtie 2 version 2.2.6 (Langmead and Salzberg, 2012) and filtered with vcflib (Garrison, 2012) and VCFTools (Danecek et al., 2011). Additionally, structural variants were identified with SvABA (Wala et al., 2018) following the alignment of Illumina reads from all sequenced isolates to the FALCON-Unzip assembled genome with BWA-mem version 0.7.15 (Li, 2013). Population structure was assessed with fastSTRUCTURE (Raj et al., 2014) following conversion of the input file with Plink version 1.90 beta (Purcell et al., 2007). Finally, a custom Python script was used to identify variant sites private to races UK1, UK2, or UK3 (see Availability of Computer Code below).

RNA-Seq reads of BC-1, BC-16, and NOV-9 were aligned to the assembled genomes of BC-1, BC-16, and NOV-9 as described above. Expression levels and differentially expressed genes were identified with featureCounts version 1.5.2 (Liao et al., 2014) and the DESeq2 version 1.10.1 R package (Love et al., 2014). Multiple test correction was performed as part of the analysis within the DESeq2 package.

Candidate avirulence genes were identified through the analysis of uniquely expressed genes and uniquely differentially expressed genes for each of the three isolates. Uniquely expressed genes were those with a Fragments Per Kilobase of transcript per Million mapped reads (FPKM) value greater than or equal to 5 in any time point from the inoculation time course experiment for a single isolate. Uniquely differentially expressed genes were those with a minimum LFC of 3 and a p-Value threshold of 0.05. Genes were compared between isolates through the use of orthology group assignments described above and scored on a one to six scale, with five and six deemed high confidence and one and two deemed as low confidence.

To identify homologous genes in other Phytophthora spp., candidate RxLRs were processed by SignalP-5.0 (Armenteros et al., 2019) to identify the cleavage site. The signal peptide sequence was removed before being submitted for a tblastn (Altschul et al., 1997) search on GenBank; interesting top hits are reported.

The expression levels of a strong candidate PfAvr2 gene, PF003_g27513 and a candidate PfAvr3 gene, PF009_g26276, were assessed via RT-qPCR. RNA-Seq results for reference genes identified previously in P. parasitica (Yan and Liou, 2006) were examined for stability of expression levels, resulting in the selection of β-tubulin (PF003_g4288) and WS41 (PF003_g28439) as reference genes. The gDNA of Pseudomonas syringae pv. maculicola 5422 and primers for 16S (Clifford et al., 2012) were used as an inter-plate calibrator. Primers were designed using the modified Primer3 version 2.3.7 implemented in Geneious R10 (Untergasser et al., 2012). Reverse transcription was performed on three biological replicates of each sequenced timepoint, alongside: 24 hpi, 72 hpi and 96 hpi for BC-1 and 24 hpi, 48 hpi and 96 hpi for NOV-9 with the QuantiTect Reverse Transcription Kit. Quantitative PCR (qPCR) was then performed in a CFX96^TM Real-Time PCR detection system in 10 μL reactions of: 5 μL of 2× qPCRBIO SyGreen Mix Lo-Rox, 2 μL of a 1:3 dilution of the cDNA sample in dH₂O and 0.4 μL of each 10 μM primer and 2.2 μL dH₂O. The reaction was run with the following conditions: 95°C for 3 min, 39 cycles of 95°C for 10 s, 62°C for 10 s and 72°C for 30 s. This was followed by 95°C for 10 s, and a 5 s step ranging from 65°C to 95°C by 0.5°C every cycle. At least two technical replicates for each sample were performed and the melt curve results were analysed to ensure the correct product was detected. Relative gene expression was calculated using the comparative cycle threshold (C_T) method (Livak and Schmittgen, 2001). Where there was a difference of at least 1 C_T value between the minimum and maximum results for technical replicates, further reactions were conducted. Outlier technical replicates were first identified as being outside 1.5 times the interquartile range. Following this, additional outliers were identified using the Grubb’s test and excluded from the analysis. Technical replicates were averaged for each biological replicate and expression values were calculated as the mean of the three biological replicates.

The regions upstream and downstream of PF003_g27513 and PF009_g26276 in the FALCON-Unzip assembly of BC-16 and the SMART de novo assembly of NOV-9 were aligned with MAFFT in Geneious R10 (Katoh et al., 2002; Katoh and Standley, 2013). Variant sites were identified through visual inspection of the alignment alongside visualisation of aligned short reads from BC-16 and NOV-9 to both assemblies with Bowtie 2 version 2.2.6 (Langmead and Salzberg, 2012) in IGV (Thorvaldsdóttir et al., 2013).

Transcription factors were identified with a previously described HMM of transcription factors and transcriptional regulators in Stramenopiles (Buitrago-Flórez et al., 2014). These proteins were then analysed through methods described above for gene loss or gain, nearby variant sites and differential expression between isolates.

All computer code is available at: https://github.com/harri- sonlab/phytophthora_fragariae, https://github.com/harrisonlab/phytophthora_rubi, and https://github.com/harrisonlab/popgen/blob/master/snp/vcf_find_difference_pop.py.

The P. fragariae isolates A4 (race US4), BC-1 (race CA1), BC-16 (race CA3), NOV-5 (race CA1), NOV-9 (race CA2), NOV-27 (race CA2), and NOV-71 (race CA2) (Table 1) were phenotyped on a differential series of four F. × ananassa cultivars with known resistance. These were: “Allstar” (containing Rpf1, Rpf2, and Rpf3), “Cambridge Vigour” (containing Rpf2 and Rpf3), “Hapil” (containing no resistance genes), and “Redgauntlet” (containing Rpf2) (van de Weg et al., 1996; R. Harrison, unpublished). Following assessment of below ground symptoms, it was shown that race CA1 was equivalent to UK1, race CA2 was equivalent to UK3, race CA3 was equivalent to UK2 and race US4 was equivalent to UK2 (Figure 1 and Table 2).

Long read PacBio sequencing generated a highly contiguous P. fragariae isolate BC-16 (UK2) reference genome of 91 Mb in 180 contigs with an N50 value of 923.5 kb (Table 3). This was slightly larger than the closely related, well studied species from Clade 7, P. sojae, at 83 Mb (Table 3; Tyler et al., 2006). The BC-16 assembly contained 266 of 303 eukaryotic Benchmarking Universal Single-Copy Orthologs (BUSCO) genes (Table 4), compared to 270 in P. sojae (Armitage et al., 2018) and so likely represented a similar completeness of the genome as the P. sojae assembly. The P. fragariae genome was shown to be highly repeat rich, with 38% of the assembly identified as repetitive or low complexity, a larger value than the 29% shown for P. sojae (Armitage et al., 2018). A total of 37,049 genes encoding 37,346 proteins were predicted in the BC-16 assembly, consisting of 20,222 genes predicted by BRAKER1 (Hoff et al., 2016) and 17,131 additional genes added from CodingQuarry (Testa et al., 2015). From these gene models, 486 putative RxLR effectors, 82 putative crinkler effectors (CRNs) and 1,274 putative apoplastic effectors were identified (Table 5). Additionally, a total of 4,054 low confidence gene models were added from intergenic ORFs identified as putative effectors. These consisted of 566 putative RxLRs, 5 putative CRNs and 3,483 putative apoplastic effectors. This resulted in a combined total of 41,103 genes encoding 41,400 proteins with 1,052 putative RxLRs, 85 putative CRNs, and 4,757 putative apoplastic effectors (Tables 3, 5).

Ten isolates of P. fragariae and three isolates of P. rubi were additionally resequenced, de novo assembled and annotated (Tables 1, 4). These isolates showed similar assembly statistics within and between species; an average of 79 Mb in 12,804 contigs with an N50 of 19.3 kb in P. fragariae, compared to an average of 78 Mb in 13,882 contigs with an N50 of 16.8 kb in P. rubi. All assemblies showed 31% of the assembly was identified as repetitive or low complexity sequences. On average both P. fragariae and P. rubi showed high levels of completeness, with an average of 274/303 BUSCO genes identified as single copies in these assemblies. Interestingly, a total of 17 genes were consistently not identified in Illumina, PacBio and Nanopore assemblies of P. fragariae and P. rubi isolates, suggesting these genes may be absent from these species and as such may not be considered true eukaryotic BUSCO genes. An average of 67 CRNs were predicted in P. fragariae and an average of 118 CRNs in P. rubi. A significantly (p = 0.013) smaller number of RxLRs were predicted in P. rubi isolates than P. fragariae isolates alongside a significantly (p = 0.039) smaller number of apoplastic effectors predicted in P. rubi than P. fragariae (Table 5). However, it is important to note that these effectors were predicted from a subset of secreted proteins, of which there were significantly (p = 0.030) fewer predicted in P. rubi than in P. fragariae (Table 5).

An orthology analysis of all predicted proteins from the isolates of P. fragariae and P. rubi assigned 481,942 (98.7%) proteins to 38,891 orthogroups. Of these groups, 17,101 contained at least one protein from all fourteen sequenced isolates and 13,132 of these groups consisted entirely of single-copy proteins from each isolate. There were 2,345 of these groups unique to P. rubi and 1,911 of these groups unique to P. fragariae. Analysis of unique and expanded orthogroups for isolates of the UK1, UK2 and UK3 races did not lead to the identification of candidate avirulence genes, as these groups did not contain putative effector genes (Figure 2).

A total of 725,444 SNP sites and 95,478 small INDELs were identified within the P. fragariae and P. rubi isolates by the Genome Analysis ToolKit (GATK) haplotypecaller (McKenna et al., 2010) and 80,388 indels and 7,020 structural variants were identified by SvABA (Wala et al., 2018). Analysis of high quality, biallelic SNP sites allowed the identification of a distinct population consisting of the isolates of race UK1, UK2 and UK3, hereafter referred to as the UK1-2-3 population, with SCRP245, the only United Kingdom isolate, potentially forming an ancestral or hybrid isolate between the UK1-2-3 population and the population represented by BC-23 and ONT-3 (Figure 3). Additionally, clear separation between isolates of P. fragariae and P. rubi was observed. Further analysis within the UK1-2-3 population allowed for the identification of private variants, which were only present in isolates of one of the three races in this population. This resulted in the identification of twelve private variants in race UK2, shared between both A4 and BC-16; however, neither the genes they fell within or were neighbours to, were predicted to encode effectors or secreted proteins and likely did not explain the differences in pathogenicity (Supplementary Table S4).

RNA-Seq data were generated from an inoculation time course experiment for representatives of races UK1, UK2 and UK3; BC-1, BC-16 and NOV-9, respectively. Following determination of expression levels of the predicted genes in both in planta and mycelia samples, a correlation analysis showed biological replicates of each timepoint grouped together (Figure 4A). Additionally, the 48 h post-inoculation (hpi) BC-1 time point grouped with the BC-16 24 hpi time point, suggesting these may represent similar points in the infection process. However, clear separation between the BC-16 timepoints was observed (Figure 4A). A total of 13,240 transcripts from BC-16 (32%) showed expression above a FPKM value threshold of five in at least one sequenced BC-16 time point and 9,329 (23%) of these showed evidence of differential expression in at least one sequenced in planta time point compared to mycelium grown in artificial media, suggesting large scale transcriptional reprogramming. As three time points post-inoculation were sequenced for BC-16, the changes in expression over the course of the infection process was investigated. A total of 2,321 transcripts (6%) showed a Log₂ Fold Change (LFC) greater than or equal to one or less than or equal to minus one, representing a general reprofiling during infection in comparison to growth in artificial media. Of transcripts differentially expressed in planta compared to artificial media, fewer transcripts were differentially expressed at both 24 hpi and 96 hpi than between sequential timepoints (297 compared to 1,016 and 1,604 transcripts). This suggested that changes during the progress of infection were captured by this dataset (Figure 4B).

Levels of expression of effector genes were also assessed. A total of 274 (26%) RxLR effectors, 27 (31%) CRNs and 880 (18%) putative apoplastic effectors from BC-16 showed evidence of expression above the FPKM threshold of 5 in at least one sequenced time point. The majority of these effector genes also showed evidence of differential expression during in planta time points compared to in vitro mycelium. A total of 253 (24%) RxLR effectors, 19 (22%) CRNs and 888 (18%) putative apoplastic effectors showed an LFC greater than or equal to one or less than or equal to minus one, representing wide scale transcriptional reprogramming of effector genes during infection (Figures 4C–E). Ranking of the fifty highest expressed genes in planta with a LFC of ≥3 in comparison to its respective mycelium, identified four putative RxLR genes that were upregulated by all three isolates (Table 6). These genes represent putative core P. fragariae RxLR’s important for pathogenicity on strawberry. Interestingly, a BLASTP search of one of these putative core effectors, PF003_g16448 (amino acids 19-139), showed homology to P. sojae Avr1b (58% pairwise identity), GenBank accession AF449625 (Shan et al., 2004).

Comparing transcripts with the highest LFC (in planta vs. mycelium) between isolates led to the identification of the putative RxLR effector encoding transcript PF003_g27513.t1 as a potential candidate for PfAvr2 in BC-16. PF003_g27513 had a peak FPKM value in BC-16 of 9,392 compared to peaks of 0 and 33, in BC-1 and NOV-9, respectively (Supplementary Table S5). Subsequent RT-qPCR analysis of further timepoints in the in planta timecourse supported the findings of the RNA-Seq timepoints and showed the expression of putative PfAvr2 in the UK2 isolates BC-16 and A4 was significantly (p < 0.05) different to those from all samples of BC-1 (UK1) and NOV-9 (UK3; Figure 4F). A BLASTP search of PF003_g16448 (amino acids 26-137) revealed homology to P. sojae Avh6/Avr1d (37% pairwise identity), GenBank accession JN253642 (Wang et al., 2011; Na et al., 2013).

The surrounding sequence of putative PfAvr2 in BC-16, BC-1 and NOV-9 was investigated for sequence variants that may explain the expression difference. As no variants were identified near the gene from the above mentioned variant panel, an assembly of the NOV-9 isolate was created from Nanopore sequencing data, producing an assembly of 93.72 Mbp in 124 contigs with an N50 of 1,260 Kb. This resulted in the identification of a SNP from T in BC-16 to G in NOV-9 ∼14 Kb downstream of the stop codon and a 30 bp insertion in NOV-9 ∼19 Kb upstream of the start codon (Figure 5A). The upstream variant appeared to be a sequencing error following the investigation of the alignment of short reads of BC-16 and NOV-9 to each assembly and so was rejected. Expression of the genes surrounding PF003_g27513 was investigated and PF003_g27514, directly upstream of putative PfAvr2 is expressed by all three isolates (Supplementary Table S5).

Additionally, putative transcription factors and transcriptional regulators were identified in all sequenced genomes. This resulted in the identification of 269 genes in the BC-16 isolate of P. fragariae. However, analysis of gene loss or gain and an investigation of variant sites showed no race specific differences, though expression level variation was observed.

Further analysis of the RNA-Seq datasets identified the putative RxLR effector encoding transcript PF009_g26276 (an orthologue of PF003_g27386) as a potential candidate for PfAvr3 in NOV-9, with a peak FPKM value in NOV-9 of 199 compared to in planta peaks of 6 and 12 in BC-1 and BC-16, respectively (Supplementary Table S6). Similar to putative PfAvr2, no sequence differences in putative PfAvr3 were observed between the three isolates. Analysis of the surrounding region between NOV-9 and BC-16 showed no sequence differences 13,000 bp upstream and 3,761 bp downstream of PF009_g26276 (Figure 5B). The two genes upstream from putative PfAvr3 are not expressed in any isolate (Supplementary Table S6), whereas the gene directly downstream in BC-16, PF003_g27385, is expressed by all the three isolates analysed.

Subsequent RT-qPCR analysis of further timepoints in the in planta time course revealed that the putative PfAvr3 was not expressed during any timepoints by the UK1 isolate BC-1 or by the UK2 isolates BC-16 and A4 (Supplementary Figure S2). However, absolute expression of PF009_g26276 in NOV-9 remained low. Expression in NOV-9 at 72 hpi was significantly (p < 0.05) greater than those from all samples of BC-1, BC-16, and A4.

Analysis of the RNA-Seq datasets did not reveal an obvious candidate for PfAvr1. Further analysis using custom scripts identified genes, which were uniquely expressed in BC-1 and those uniquely differentially expressed. These genes were scored for confidence of race avirulence determinant; high, medium and low. No genes were identified in the high confidence class in BC-1, but four genes were scored as medium confidence, one of which was a putative apoplastic effector (Supplementary Table S7). The analysis was repeated for BC-16 and NOV-9, in case the putative RxLR candidates are not PfAvr2 and PfAvr3, respectively (Supplementary Table S7).