A complex resistance locus in Solanum americanum recognizes a conserved Phytophthora effector.

Late blight caused by Phytophthora infestans greatly constrains potato production. Many Resistance (R) genes were cloned from wild Solanum species and/or introduced into potato cultivars by breeding. However, individual R genes have been overcome by P. infestans evolution; durable resistance remains elusive. We positionally cloned a new R gene, Rpi-amr1, from Solanum americanum, that encodes an NRC helper-dependent CC-NLR protein. Rpi-amr1 confers resistance in potato to all 19 P. infestans isolates tested. Using association genomics and long-read RenSeq, we defined eight additional Rpi-amr1 alleles from different S. americanum and related species. Despite only ~90% identity between Rpi-amr1 proteins, all confer late blight resistance but differentially recognize Avramr1 orthologues and paralogues. We propose that Rpi-amr1 gene family diversity assists detection of diverse paralogues and alleles of the recognized effector, facilitating durable resistance against P. infestans.

Introduction

Potatois the third most important directly-consumed food crop world-wide[1]. Phytophthora infestans, an oomycete pathogen, causes late blight disease in potato, and can result in complete crop failure. Disease management is primarily based on repeated fungicide applications (10-25 times per season in Europe). However,fungicide-resistant races have emerged[2].

To elevate late blight resistance, Resistance to Phytophthora infestans (Rpi) genes were identifiedin wild relatives of potato and used for resistance breeding[3]. More than 20 Rpi genes have been mapped and cloned from different Solanum species, e.g. R2 (Rpi-blb3), R3a, R8, Rpi-blb1, Rpi-blb2 and Rpi-vnt1 [4-10]. All encode coiled-coil (CC), nucleotide binding (NB), leucine-rich repeat (LRR) (NLR) proteins[11] and some require helper NLR proteins of the NRC family[12]. However, most cloned Rpi genes can be overcome by at least one strain of P. infestans [13]. Provision of durable late blight resistance for potato remains a major challenge.

NLR-mediated immunity upon effector recognition activates “effector-triggered immunity” (ETI)[14]. In oomycetes, all identified recognized effectors, or avirulence (AVR)proteins, carry a signal peptide and an RXLR motif[15]. 563 RxLR effectors were predicted from the P. infestans genome, enabling identification of the recognized effectors[16,17]. Many P. infestans effectors show signatures of selection to evade recognition by corresponding NLR proteins[18]. NLR genes also show extensive allelic and presence/absence variation in wild plant populations[19,20] and known Resistance (R) gene loci like Mla, L, Pi9, RPP1 and RPP13 from barley, flax, rice and Arabidopsis show substantial allelic polymorphism[21-24]. Remarkably, different barley Mla and flax L gene alleles can recognize sequence-unrelated effectors[25,26].

Technical advances like RenSeq (Resistance gene enrichment and Sequencing)and PenSeq (Pathogen enrichment Sequencing) enable rapid definition of allelic variation and mapping of plant NLRs, or discovery of variation in pathogen effectors[27-29].

Combined with single-molecule real-time (SMRT) sequencing, SMRT RenSeq enabled cloning of Rpi-amr3 from Solanum americanum [30]. Similarly, long read and cDNA PenSeq enabled us to identify Avramr1 from P. infestans [31].

In this study,we further explored the genetic diversity of S. americanum, and by applying sequence capture technologies,we fine-mapped and cloned Rpi-amr1 from S. americanum,(usually) located on the short arm of chromosome 11. Multiple Rpi-amr1 homologs were found in different S. americanum accessions and in relatives, including Solanum nigrescens and Solanum nigrum. Functional alleles show extensive allelic variation and confer strong resistance to all 19 tested diverse P. infestans isolates. Although differential recognition was found between different Rpi-amr1 and Avramr1 homologs, all Rpi-amr1 alleles recognize the Avramr1 homologs from Phytophthora parasitica and Phytophthora cactorum. Our study reveals unique properties of genetic variation of R genes from “non-host” species.

Results

Rpi-amr1 maps to the short arm of chromosome 11

We previously investigated S. americanum and isolated Rpi-amr3 from an accession 944750095 (SP1102)[30]. To discover new Rpi-amr genes, we characterized additional 14 lines of P. infestans-resistant S. americanum and close relatives S. nigrescens and Solanum nodiflorum by crossing them to a susceptible (S) S. americanum line 954750186 (hereafter SP2271) (Table 1, Fig. S1). To avoid self-pollination, a resistant parent was always used as a pollen donor. All the corresponding F1 plants (6-10 per cross)were resistant in a detached leaf assay (DLA) (Table 1). Around 60-100 F2 progeny derived from each self-pollinated F1 plant were phenotyped by DLA using P. infestans isolate 88069[32]. The F2 progenies that derived from the resistant parents with working numbers SP1032, SP1034, SP1123, SP2272, SP2273, SP2360, SP3399, SP3400, SP3406, SP3408 and SP3409 segregated in a ratio suggesting the presence of a single (semi-) dominant resistance gene (fitting 3:1 or 2:1 [likely due to segregation distortion], R:S - resistant to susceptible - ratio). Two crosses showed a 15:1 segregation (resistant parent SP2300 and SP2307), suggesting the presence of two unlinked resistance genes. Investigating resistance in SP1101 required two backcrosses to SP2271 prior to selfing of resistant progeny to reveal a 3:1 R:S segregation.

Table 1

S. americanum, S. nodiflorum and S. nigrescens accessions used in this study and the corresponding Rpi-amr1 homologs

Accession	Working name	Species	Reported origin	Source	Late blight resistance	Rpi-amr1 homolog	Similarity	Cloning method
954750186	SP2271	S. americanum	Brazil	RU	Susceptible
954750184	SP2273	S. americanum var. patulum	unknown	RU	Resistant	Rpi-amr1-2273	100%	Map-based cloning
sn27	SP1032	S. americanum sensu lato	China	BGS	Resistant	Rpi-amr1-1032	92.8%	Association genomics
Veg422	SP1034	S. americanum sensu lato	unknown	NN	Resistant	Rpi-arm1-2273	100%	Association genomics
A54750014	SP1101	S. americanum sensu lato	unknown	RU	Resistant	Rpi-amr1-1101	89.4%	SMRT RenSeq
A14750006	SP1123	S. americanum sensu lato	unknown	RU	Resistant	Rpi-amr1-1123	91.8%	Association genomics
954750174	SP2272	S. americanum	unknown	RU	Resistant	Rpi-amr1-2272	89.4%	Association genomics
SOLA 226	SP2300	S. americanum	Cuba	IPK	Resistant	Rpi-amr1-2300	90.4%	SMRT RenSeq
SOLA 425	SP2307	S. americanum	America	IPK	Resistant	Rpi-amr1-2307	91.7%	Association genomics
Wang 2059	SP2360	S. americanum	China	NH M	Resistant	Rpi-arm1-2273	100%	Association genomics
A14750138	SP3399	S. americanum	unknown	RU	Resistant	Rpi-amr1-2272	89.4%	Association genomics
A14750130	SP3400	S. nodiflorum	unknown	RU	Resistant	Rpi-amr1-2273	100%	Association genomics
944750261	SP3406	S. nigrescens	Bolivia	RU	Resistant	Rpi-amr1-3406	92.5%	Association genomics
954750172	SP3408	S. nigrescens	Bolivia	RU	Resistant	Rpi-amr1-3408	92.6%	Association genomics
A14750423	SP3409	S. nigrescens	Mauritius	RU	Resistant	Rpi-amr1-3409	89.5%	SMRT RenSeq

RU - Radboud University, Nijmegen, The Netherlands

IPK - IPK Gatersleben, Germany

NHM - Natural History Museum, London, United Kingdom

BGS - Shanghai Botanical Garden, Shanghai, China

NN - Nicky's Nursery Ltd, Kent, United Kingdom

To identify Rpi genes from these resistant S. americanum accessions, we prioritized an F2 population derived from resistant parent SP2273 and named the corresponding gene Rpi-amr1. Using markers from RenSeq, genotyping by sequencing (RAD markers) and Whole Genome Shotgun sequencing (WGS), the Rpi-amr1 gene was mapped in a small population (n=188 gametes) to the short arm of chromosome 11, between markers RAD_3 and WGS_1 (Fig. 1a, Table S1, S2). We expanded the mapping population and developed a PCR marker WGS_2 that co-segregated with resistance in 3,586 gametes (Fig. 1b, Table S2). To generate the physical map of the target interval from SP2273, a BAC library was generated. Two BAC clones (12H and 5G) covering the target interval were identified by a PCR screen with the above linked marker, sequenced on the PacBio RSII platform, and assembled into a single contig of 204,128 bp (Fig. 1c). We predicted 11 potential coding sequences on the assembled contig, nine of which encode NLR genes (Fig. 1c). These NLR genes belong to the CNL class and have 80-96% between-paralog identity.

Fig. 1

Map-based cloning of Rpi-amr1 and its resistance to P. infestans.

Mapping of Rpi-amr1 in a small F2 population (n=188 gametes); the names of the markers and genetic distances are shown above or below the bar.

Fine mapping of Rpi-amr1 in the F2 population of 3,586 gametes. The names of the markers and the number of recombinants are shown above or below the bar.

Physical map of the target Rpi-amr1 interval based on the assembled BAC contig. The markers present on the BAC are shown. The predicted NLR genes are depicted as black arrows (expressed NLRs) or empty arrows (pseudogenized NLRs). Rpi-amr1 (formerly Rpi-amr1e) is indicated by a red arrow.

Four Rpi-amr1 transcripts detected by 3’ RACE PCR.

Leaves of N. benthamiana plants were infiltrated with the binary vector pICSLUS0003∷ 35S overexpressing either the late blight resistance gene Rpi-amr3 (positive control), one of seven Rpi-amr1 candidates, or the non-functional Rpi-amr3-S (negative control). Leaves were inoculated with P. infestans strain 88069 24 h after infiltration. Only leaves infiltrated with Rpi-amr3 and Rpi-amr1e (pictured) showed reduced pathogen growth, whereas P. infestans grew well in the presence of the remaining Rpi-amr1 candidates. Only Rpi-amr1c is shown as the phenotype of all other non-functional candidate genes was indistinguishable. Photographs were taken 9 dpi.

Transgenic potato cv. Maris Piper which expresses Rpi-amr1 under the native regulatory elements is resistant to P. infestans isolate 88069 (top), displaying no symptoms at the spot of inoculation. Each leaflet was inoculated with a droplet containing approximately 1,000 zoospores; photographs were taken 9 dpi.

The control plants carrying the non-functional candidate Rpi-amr1a show large necrotic lesions and sporulation. Each leaflet was inoculated with a droplet containing approximately 1,000 zoospores; photographs were taken 9 dpi.

To define which of these NLR genes are expressed, cDNA RenSeq data of the resistant parent SP2273 were generated and mapped to the BAC_5G sequence. Seven out of nine NLR genes were expressed. These genes - Rpi-amr1a, b, c, d, e, g and h - were tested as candidate genes for Rpi-amr1 (Fig. 1c).

Rpi-amr1e confers resistance in Nicotiana benthamiana and cultivated potato

To test the function of the seven candidate genes, we cloned their open reading frames from genomic DNA inclusive of introns into a binary expression vector under control of the 35S promoter. Rpi-amr3 was used as a positive control and the non-functional Rpi-amr3-S was used as a negative control. The constructs carrying each of the seven candidate genes were transiently expressed after Agrobacterium infiltration into N. benthamiana leaves, which were subsequently inoculated with the P. infestans isolate 88069 as described previously[30]. P. infestans growth was observed six days post inoculation (dpi). Only 35S::Rpi-amr1e -infiltrated leaves showed reduced pathogen growth at 6 dpi compared to other candidate genes like Rpi-amr1c, or negative control Rpi-amr3-S. (Fig. 1e). Hence, we conclude that Rpi-amr1e is the functional Rpi-amr1 (hereafter) gene from S. americanum SP2273.

To test if Rpi-amr1 confers late blight resistance in potato, we cloned it with its native promoter and terminator, and generated transgenic potato cultivar Maris Piper plants carrying Rpi-amr1. A non-functional paralog Rpi-amr1a was also transformed into Maris Piper as a negative control. As in the transient assay, stably transformed Rpi-amr1 lines resisted P. infestans 88069 in potato (Fig. 1f), but Rpi-amr1a -transformed plants did not (Fig. 1g).

Rpi-amr1 is a four exon CC-NLR

To characterize the structure of Rpi-amr1, we mapped the cDNA RenSeq data to the full length Rpi-amr1 gene and found four alternatively spliced forms of Rpi-amr1. The most abundant form, supported by >80% of reads, comprises four exons encoding a protein of 1,013 amino acids. The remaining three forms had shifts in reading frames, leading to premature stop codons or absence of some exons. This was confirmed with 3’ RACE PCR (Fig. 1d). The Rpi-amr1 is a typical CC-NB-LRR resistance protein, with a coiled-coil domain (CC; amino acids 2-146), nucleotide binding domain (NB-ARC; amino acids 179-457) and leucine-rich repeats (LRR; located between amino acids 504-900) which are all positioned in the first exon (1-918 aa, Fig. 2a). The remaining three short exons (amino acids 919-943, 944-1002 and 1,003-1,013) lack homology to any known domains. No integrated domains[33] were found in the Rpi-amr1 protein.

Fig. 2

Schematic representation of amino acid sequence alignment of Rpi-amr1 homologs (a) and P. infestans resistance in transient assay (b).

The exons and the conserved NLR domains are highlighted at the top of the alignment (exons, grey; CC, orange; NB-ARC, blue; LRR, green). Black bars in the alleles indicate the polymorphic amino acids and indels as compared with Rpi-amr1-2273. The numbers next to the alleles refer to the accession working numbers (Table 1). Figure drawn to the scale.

Nine Rpi-amr1 homologs provide resistance to P. infestans in transient complementation assay. Rpi-amr1 genes with native regulatory elements were infiltrated into N. benthamiana leaves. At 1 dpi, leaves were cut off and drop inoculated with 10 μl of zoospore suspension (50,000 zoospores/mL) from P. infestans isolate 88069. The non-functional Rpi-amr1-2271 homolog from susceptible accession SP2271 was used as negative control. Photographs were taken 8 dpi.

Functional Rpi-amr1 homologs are present in were identified from multiple resistant S.americanum and relatives

Previously, we found at least 14 S. americanum accessions and related species that resist late blight (Table 1). To test if Rpi-amr1 contributes to late blight resistance in other resistant S. americanum accessions, we genotyped 10-50 susceptible F2 plants of the populations derived from resistant accessions, with a marker positioned in Rpi-amr1 gene (56766, Fig. 1 and Table S2). We found that the marker is absent in all tested susceptible descendants of accessions SP1032, SP1034, SP1123, SP2272, SP2307, SP2360, SP3399, SP3400, SP3406 and SP3408,suggesting that the resistance is linked to the Rpi-amr1 locus. To test if in these accessions the resistance is conferred by functional Rpi-amr1 homologs, we performed SMRT RenSeq-based de novo assembly of each resistant accession, and looked for homologs with the greatest identity to Rpi-amr1. For accessions SP2307, SP3399 and SP3406, we also used cDNA RenSeq to monitor their expression. We mapped de novo contigs to the coding sequence of Rpi-amr1 allowing for 15% mismatches and gaps, and selected the closest homolog as a candidate Rpi-amr1 ortholog (Table S3). In three resistant parents, namely SP1034, SP2360 and SP3400,the functional alleles showed 100% identity at the amino acid level to Rpi-amr1, while amino acid sequences from the remaining accessions had as little as 89% identity to the functional Rpi-amr1 (Table S3). As described previously, we transiently expressed the closest related candidate Rpi-amr1 homologs in N. benthamiana leaves followed by DLA with P. infestans isolate 88069, and verified their functionality. The unique homologs of Rpi-amr1-2273 were named as Rpi-amr1-1032, Rpi-amr1-1123, Rpi-amr1-2272, Rpi-amr1-2307 and Rpi-amr1-3408.

For some accessions, like SP1101 and SP2300, the Rpi-amr1-linked markers gave ambiguous results, so we directly performed bulked segregant analysis (BSA) and RenSeq. Additional Rpi-amr1 co-segregating paralogs, Rpi-amr1-1101 and Rpi-amr1-2300,were identified and verified in transient assays as above (Fig. 2b).

Similarly, we inspected an F2 population derived from S. nigrescens accession SP3409 (Table 1). We applied BSA RenSeq and SMRT RenSeq to the resistant parents and F2 segregating population, and we found five candidate NLRs belonging to the same Rpi-amr1 clade, all of which are expressed. The five candidates were cloned, and transient assays verified one of them as a functional Rpi-amr1 homolog, Rpi-amr1-3409. However, Rpi-amr1-3409 does not co-segregate with Rpi-amr1- linked markers.We used GenSeq sequence capture-based genotyping (Chen et al. 2018), and found that Rpi-amr1-3409 locates on chromosome 1, based on the potato DM reference genome[34]. This result suggests that a fragment of DNA that locates on distal end of the short arm of chromosome 11 in other resistant accessions was translocated to the distal end of the long arm of chromosome 1 in SP3409.

When the full-length amino acid sequences of nine Rpi-amr1 homologs were aligned, the polymorphisms between different functional alleles were found to be distributed through all domains including the LRR region (Fig. 2a and Fig.S2).

Taken together, by using BSA RenSeq, SMRT RenSeq, cDNA RenSeq, association genomics and GenSeq, we cloned eight additional functional Rpi-amr1 homologs from different resistant accessions, of which all confer resistance to P. infestans 88069 in transient assays. The closest Rpi-amr1 homolog from susceptible parent SP2271 does not confer resistance (Fig. 2b).

Rpi-amr1 confers broad-spectrum late blight resistance in cultivated potato

To test the scope of late blight resistance conferred by Rpi-amr1 and its homologs, we generated stably transformed transgenic potato cv Maris Piper plants carrying Rpi-amr1-2272 and Rpi-amr1-2273, the most diverged of the homologs (Table S3), and inoculated them by DLA with 19 P. infestans isolates from UK, the Netherlands, Belgium, USA, Ecuador, Mexico and Korea (Table 2). Many of the tested P. infestans isolates can defeat multiple Rpi genes (Table 2). Our DLAs show that Maris Piper carrying Rpi-amr1-2272 or Rpi-amr1-2273 resist all 19 tested P. infestans isolates, while the wild-type Maris Piper control is susceptible to all of them. This indicates that Rpi-amr1 confers broad-spectrum resistance against diverse P. infestans races.

Table 2

Phenotypes of potato plants stably transformed with Rpi-amr1-2272 and Rpi-amr1-2273 after inoculation with multiple isolates of P. infestans.

Isolate	Rpi-amr1-2272	Rpi-amr1-2273	Maris Piper	Origin	Race[e]
NL00228	R	R	S	The Netherlands	1.2.4.7
US23	R	R	S	USA	n.a.
3928A[a]	R	R	S	UK	1.2.3.4.5.6.7.10.11[f]
EC3626[b]	R	R	S	Ecuador	n.a.
NL14538[c]	R	R	S	The Netherlands	n.a.
NR47UH[d]	R	R	S	UK	1.3.4.7.10.11[f]
T30-4	R	R	S	The Netherlands	n.a.
USA618	R	R	S	USA	1.2.3.6.7.10.11
KPI15-10	R	R	S	Korea	n.a.
IPO-C	R	R	S	Belgium	1.2.3.4.5.6.7.10.11
PIC99189	R	R	S	Mexico	1.2.5.7.10.11
UK7824	R	R	S	UK	n.a.
PIC99177	R	R	S	Mexico	1.2.3.4.7.9.11
VK98014	R	R	S	The Netherlands	1.2.4.11
NL08645	R	R	S	The Netherlands	n.a.
PIC99183	R	R	S	Mexico	1.2.3.4.5.7.8.10.11
NL11179	R	R	S	The Netherlands	n.a.
EC1[b]	R	R	S	Ecuador	1.3.4.7.10.11
NL01096	R	R	S	The Netherlands	1.3.4.7.8.10.11

Clonal lineage EU_13_A2, or “Blue13”

Overcomes Rpi-vnt1

Overcomes Rpi-vnt1 and partially Rpi-blb1, Rpi-blb2

Clonal lineage EU_6_A1, commonly known as “Pink6”

Summarized in[49]

See[50]

Differential recognition by Rpi-amr1 alleles of Avramr1 homologs

Avramr1 (PITG_07569) was identified in P. infestans race T30-4 by long-read and cDNA PenSeq, and multiple Avramr1 homologs were identified in four P. infestans isolates and classified into four subclades[31]. To investigate if all nine cloned Rpi-amr1 homologs could recognize diverse Avramr1 homologs from different P. infestans isolates, in addition to Avramr1 from race T30-4 that corresponds to clade A, we synthesized three Avramr1 homologs Avramr1-13B1, Avramr1-13C2 and Avramr1-13D1 from isolate 3928A (EU_13_A2, commonly known as “Blue 13”), corresponding to clades B, C and D, respectively (Fig. 3).We also synthesized the Avramr1 homologs from P. parasitica and P. cactorum [31]. These six Avramr1 homologs were co-expressed in N. benthamiana by agro-infiltration in all possible combinations with nine functional Rpi-amr1 homologs and the non-functional Rpi-amr1-2271 as a negative control (Fig. 3).

Fig. 3

Differential recognition of Rpi-amr1 and Avramr1 homologs.

Four Avramr1 homologs representing clades A-D, and P. parasitica and P. cactorum homologs were co-infiltrated with ten Rpi-amr1 homologs, including a non-functional homolog Rpi-amr1-2271, into N. benthamiana leaves. Colours from green to brown represent the strength of HR scored from 0 to 2 (see bottom panel). N=3. The representative HR phenotype and scoring are shown in Fig. S3.

Left: phylogenetic tree of nine functional Rpi-amr1 homologs and non-functional homolog Rpi-amr1-2271. Top: phylogenetic tree of Avramr1 homologs from four isolates of P. infestans.

* Stable Rpi-amr1-2307 N. benthamiana transformants show HR upon transient expression of Avramr1 and Avramr1-13B 1.

We found that different combinations of Rpi-amr1 alleles and Avramr1 homologs led either to strong, weak or no HR phenotype in transient assay, but the non-functional Rpi-amr1-2271 allelefailed to recognize any Avramr1 homologs (Fig. 3) The representative HR phenotype and the scoring of HR indices are shown in Fig. S3. Rpi-amr1-2300 and Rpi-amr1-2307 recognized one Avramr1 homolog each, but others detected Avramr1 homologs from more than one clade. Clade C, represented here by Avramr1-13C2,is usually not expressed[31], and when expressed from 35S promoter, this effector was not recognized by most Rpi-amr1 homologs, though a weak HR was observed upon co-expression with Rpi-amr1-2272. Avramr1-13D1 belongs to Clade D, which is absent in T30-4 but present in four other sequenced isolates[31], and was recognized by all but one (Rpi-amr1-2300) homologs in the transient assay. Surprisingly, two Avramr1 homologs from P. parasitica and P. cactorum are strongly recognized by all functional Rpi-amr1 homologs, apart from Rpi-amr1-2272 which showed a weaker HR (Fig. 3).

Collectively, our data shows that Rpi-amr1/Avramr1 homolog pairs provoke quantitatively and qualitatively different HRs, but all functional Rpi-amr1 homologs detect at least one Avramr1 homolog from P. infestans isolate 3928A.

Both Rpi-amr1-mediated resistance and effector recognition are NRC2 or NRC3 dependent

We generated a phylogenetic tree for representative Solanaceae NLR proteins. Rpi-amr1 is grouped with clade CNL-3, from which no functional resistance genes were previously cloned (Fig. 4a). The closest related cloned functional gene is Rpi-amr3 (31.2% identity on aa level) belonging to clade CNL-13 and located on chromosome 4. The phylogenetic affiliation suggested that Rpi-amr1 is likely to depend on the helper NRC clade because CNL-3 is among the large super-clade of NRC-dependent sensors (Fig. 4a)[12].

Fig. 4

Rpi-amr1 is NRC2 or NRC3 dependent.

Phylogenetic analysis of Rpi-amr1 protein and other functional Solanaceae NLR proteins. The NLR clades shown here are as described previously[30], the NRC-dependent sensor clades are marked by blue box.

Transient expression of Rpi-amr1-2273 in NRC2 / NRC3 double knockout N. benthamiana, followed by zoospore inoculation of P. infestans isolate 88069, results in large necrotic lesions indicating lack of resistance.

Transient expression of Rpi-amr1-2273 in NRC2 / NRC3 / NRC4 triple knockout N. benthamiana, followed by zoospore inoculation of P. infestans isolate 88069, results in large necrotic lesions indicating lack of the resistance.

Transient expression of Rpi-amr1-2273 in NRC4 knockout N. benthamiana, followed by zoospore inoculation of P. infestans isolate 88069results in small necrotic lesions indicating resistance.

To test this hypothesis, we transiently expressed Rpi-amr1-2273 together with PpAvramr1 in NRC4, NRC2/3 or NRC2/3/4 knock out N. benthamiana leaves[35,36](Fig. S4). The HR phenotype was abolished in NRC2/3 and NRC2/3/4 knockout plants (Fig.S5c and b), but not in NRC4 knock-out or wild-type plants (Fig.S5d and a). The HR was recovered when NRC2 or NRC3 was co-expressed in the NRC2/3/4 or NRC2/3 knock out plants, but co-expression of NRC4 did not complement the loss of HR phenotype in NRC2/3/4 knockout plants. (Fig.S5b and c). We further showed that also Rpi-amr1 mediated resistance is dependent on NRC2 or NRC3 but not NRC4, as transient expression of Rpi-amr1-2273 followed by P. infestans infection restricted pathogen growth only in NRC4 knockout N. benthamiana plants (Fig. 4b-d). These data indicate that both the effector recognition and resistance conferred by Rpi-amr1 are NRC2 or NRC3 dependent.

Functional Rpi-amr1 homologs are present in hexaploid S.nigrum accessions

Most S. nigrum accessions are highly resistant to P. infestans and S. nigrum has been reported to be a “non-host” to P. infestans [37], even though rare accessions are susceptible[38]. S. americanum may be the diploid ancestor of hexaploid S. nigrum [39]. To test if Rpi-amr1 also contributes to late blight resistance in S. nigrum, we designed nested PCR primers based on the Rpi-amr1-2273 sequence, and amplified and sequenced the coding sequence of Rpi-amr1 homologs from two resistant and one reported susceptible S. nigrum accessions[38]. From two resistant accessions (SP1088 and SP1097; Table S4), we amplified sequences with >99% nucleotide identity to S. americanum Rpi-amr1-2273, namely Rpi-nig1-1088 and Rpi-nig1-1097. The protein sequences of Rpi-nig1-1088 and Rpi-nig1-1097 are identical, with only one amino-acid (225 R->Q) change compared to Rpi-amr1-2273(Fig. S6a). The primers used for allele mining did not amplify a product of the expected size for Rpi-amr1 from the susceptible line SP999. To test their function, we performed transient assay for HR and disease resistance on N. benthamiana. We found both Rpi-nig1-1088 and Rpi-nig1-1097 show strong HR when co-expressed with PpAvramr1 and PcAvramr1.However, they activate a weaker HR to Avramr1 and Avramr1-13B1 compared to Rpi-amr1-2273 (Fig. S6b).Like Rpi-amr1-2273, but not the negative control Rpi-amr1-2271,transiently expressed Rpi-nig1-1088 and Rpi-nig1-1097 confer resistance to P. infestans 88069 (Fig. S6c).It is the first report of functional Rpi genes from S. nigrum, and our finding suggests the strong late blight resistance of S. nigrum is determined or partially determined by the Rpi-amr1 homologs,that were most likely inherited from S. americanum.

High allelic diversity at Rpi-amr1 was generated through inter-paralog and ortholog sequence exchange

Rpi-amr1 alleles show relatively high nucleotide diversity (π=0.04), which could be an indication of balancing or diversifying selection (Table S5). In addition, Rpi-amr1 alleles differ in their recognition of the Avramr1 homologs (Fig. 3) which is also consistent with selection in a host-parasite co-evolutionary arms race. To test the hypothesis that allelic polymorphism at Rpi-amr1 results from diversifying selection,we calculated diversity statistics and performed a McDonald-Kreitman test on both Rpi-amr1 alleles and Avramr1 homologs. As expected, Avramr1 homologs show a signature consistent with balancing selection (Tajima’s D = 2.27) (Table S5). Remarkably, despite the high nucleotide diversity, no clear signals of balancing or diversifying selection were detected for Rpi-amr1 (Tajima’s D = 0.09083) (Table S5). Aligning the Rpi-amr1 alleles against the reference and scrutinizing the sequences in more detail provided further insights. The nucleotide similarity of alleles varies markedly across the Rpi-amr1 homologs (Fig. 2a and Table S3); this pattern is consistent with occasional recombination between highly diverged alleles or paralogs.

To test whether recombination could explain the observed polymorphisms in Rpi-amr1 alleles, we predicted the possible recombination events using 3SEQ. Several recombination events were detected between Rpi-amr1 orthologs from different S. americanum accessions, and Rpi-amr1 paralogs from SP2273 (Table S6). Some sequence exchanges were visualized using HybridCheck (Fig. S7)[40], and these data suggest that sequence exchange occurred between functional Rpi-amr1 alleles and paralogs. To confirm these findings, we mapped all cloned Rpi-amr1 CDS back to the BAC_5G sequence from accession SP2273 (Fig.S8). As expected, some Rpi-amr1 homologs (e.g. SP2300 and SP2272) show a perfect match with the fourth NLR,and show a distribution of high identity that reflects the intron-exon structure. For some homologs (e.g. 2271), 5’ end sequences match different NLR sequences on the BAC_5G and for others (e.g. 2275) part of the sequence is highly diverged from BAC_5G. Taken together, our results indicate that the polymorphism of Rpi-amr1 alleles appears to have arisen partly due to sequence exchange between highly diverged alleles and paralogs, and not just through mutation accumulation.

Discussion

Achieving complete and durable resistance is the ultimate goal of resistance breeding. Here, we report significant progress towards durable resistance against potato late blight. Most cloned late blight resistance genes derive from wild tuber-bearing species of genus Solanum, and many have been overcome by one or more P. infestans strains[41]. Conceivably, resistance to P. infestans in nearly all S. americanum and S. nigrum accessions is due to multiple NLR genes, as zoospores from P. infestans can germinate on S. nigrum leaves but penetration is stopped by strong HR[37,42]. Rpi genes from plant species that only rarely support pathogen growth have likely not participated, or are no longer participating, in an evolutionary arms race with P. infestans,and hence, the pathogen’s effectors have not (yet) evolved to evade detection by these Rpi genes. Under this scenario, a pre-existing standing variation in the pathogen for overcoming such Rpi genes is either absent or extremely rare. This makes such genes promising candidates for provision of broad-spectrum and durable late blight resistance, provided they are not deployed alone which facilitates one-step genetic changes in the pathogen to evade them, but rather in combination with other genes, as in the source plant[43].

We report here a novel,broad-spectrum S. americanum resistance gene, Rpi-amr1. We also identified eight additional Rpi-amr1 alleles from different S. americanum accessions and relatives, including one Rpi-amr1 allele that translocated to the long arm of chromosome 1.Homology-based cloning also revealed the presence of functional Rpi-amr1 homologs in S. nigrum. All nine cloned Rpi-amr1 alleles confer late blight resistance in transient assays in N. benthamiana, and both Rpi-amr1- 2272 and Rpi-amr1-2273 in potato cv Maris Piper background confer resistance to all 19 tested P. infestans isolates from different countries, many of which overcome other Rpi genes. Thus, Rpi-amr1 is widely distributed in germplasm of S. americanum, its relatives and S. nigrum, and may contribute to the resistance of nearly all accessions to P. infestans.

Many plant R genes and their corresponding Avr genes evolved differential recognition specificities with extensive allelic series for both R gene and Avr genes. Examples include ATR1 and RPP1 or ATR13 and RPP13 from Hyaloperonospora arabidopsidis and Arabidopsis [15], Avr567 and L genes from the rust Melampsora lini and flax[44], and multiple and diverse recognized effectors from barley powdery mildew and Mla from barley. Similarly, Avramr1 and its homologs from several P. infestans races[31] were found to be differentially recognized by alleles of the Rpi-amr1 gene.Remarkably though, Rpi-amr1 nucleotide diversity of the R gene did not show any of the hallmarks of diversifying or balancing selection.

Rather than through mutation accumulation, the high allelic variation observed at Rpi-amr1 appears to have been generated partly by recombination between significantly diverged alleles and paralogs. The recombination events are likely to be rare relative to the mutation rate, given that the alleles carry many polymorphisms. This evolutionary scenario can explain the observed mosaic-like structure of high and low sequence similarities when the Rpi-amr1 alleles were mapped against the contig based on two overlapping BAC clones. The deep coalescence of alleles that is implicit in this scenario can be generated by balancing selection, but we did not find evidence of such selection when analysing the nucleotide substitution patterns. Recombination between Rpi-amr1 alleles could have eroded this signature of selection, as has been observed also in Rp1 resistance genes in grasses[45] and in the vertebrate immune genes of the major histocompatibility complex (MHC)[46,47]. Nucleotide sequence diversity across the Rpi-amr1 alleles is correlated with only slight differences in Avramr1 recognition specificity. Rpi-amr1 alleles can even recognize multiple Avramr1 paralogs from a single P. infestans strain, a scenario that might elevate durability of resistance. Since the S. americanum population recognizes multiple Avramr1 alleles and paralogs, small mutational changes in Avramr1 gene are unlikely to suffice to escape detection, which makes resistance-breaking less likely, thus promoting evolutionary durability of Rpi-amr1. Remarkably, Avramr1 (PITG_07569) was recently reported to regulate plant alternative splicing and promote the colonization of P. infestans [48], indicating Avramr1 contributes an important function for the virulence of P. infestans. We hypothesise that this enhanced recognition capacity could be key to the evolution of “non-host” resistance, offering an escape for the plant from the coevolutionary arms race. Conceivably, stacking Rpi-amr1 alleles in cis could extend the recognition specificities, which could potentially lead to even more durable late blight resistance.

Intriguingly, two Avramr1 homologs from P. parasitica and P. cactorum are recognized by all Rpi-amr1 homologs. Presumably, these genes have been under even less selection pressure to evade Rpi-amr1 recognition. This result indicates that Rpi-amr1 has the potential to provide “non-host”type resistance in S. americanum against multiple oomycete pathogens like P. parasitica and P. cactorum, which can infect a wide range of hosts. As both the resistance and effector recognition of Rpi-amr1 are NRC2 or NRC3 dependent, co-expression of NRC2 or NRC3 with Rpi-amr1 might enable it to confer resistance to other Phytophthora species outside the Solanaceae.

In summary, we cloned Rpi-amr1,a broad-spectrum Rpi gene that contributes to the strong late blightresistance of nearly all S. americanum accessions to late blight.The apparent redundancy across the Rpi-amr1 gene family may serve an evolutionary function by broadening the scope for recognizing multiple Avramr1 alleles and paralogs, and potentially reducing the probability of evolution of resistance-breaking strains.Stacking this type of Rpi gene with additional Rpi genes might help to turn host plants such as potato into non-hosts for late blight, enabling broad-spectrum and durable resistance.

Methods

Methods and associated references are in supplementary information.