Introduction

Soybean (Glycine max L.) is a major source of protein and oil that is produced on approximately 130 million hectares globally each year1. As the global population increases from 7.9 billion to an estimated 9.8 billion by 2050, we will need to produce more food on less land to ensure global food security2. The mission to increase the seed yield potential of soybeans has been at the forefront of soybean breeders for decades, and consequently, present soybean varieties produce nearly twice the seed yield they did 60 years ago3. However, efforts to achieve yield gains through genetics are often at the expense of decreased levels of disease resistance4.

Phytophthora stem and root rot (PRR) of soybean, caused by the soilborne oomycete Phytophthora sojae (Kaufmann & Gerdemann), is responsible for 1–2 billion U.S. dollars in soybean yield loss worldwide5. Outbreaks of PRR are more prevalent after heavy rains in fields where soil moisture remains saturated and facilitates the production of motile zoospores of P. sojae, which move chemotactically toward soybean root exudates and infect soybean roots. During infection, the sexual reproductive and resting structure, the oospore, is produced within infected soybean root and stem tissue6. P. sojae is homothallic, or self-fertile, and can produce viable oospores without the need for a second genotype. Oospores of P. sojae can lie dormant within plant debris in the soil for years until environmental conditions are conducive to infection (i.e., flooding) and susceptible host plants are present6,7.

Current climate models suggest that periods of intense rain may become more frequent due to anthropogenically driven climate change8. The increasing frequency of heavy rains could provide more favorable environmental conditions for PRR disease throughout the global soybean-producing regions, emphasizing that effective and economically viable management of this destructive soybean disease will be increasingly important9.

Currently, soybean varieties with single-dominant P. sojae resistance genes (termed “Rps” genes, “resistance to Phytophthora sojae”) and quantitative resistance are commercially available to mitigate losses to PRR specifically. The first Rps gene available for PRR management was Rps1a, released in 1964, followed by Rps1c in 1980, and lastly, Rps1k and Rps3a, released in 198510. The genes Rps1c, Rps1k, and Rps3a were made commercially available to manage PRR after Rps1a-mediated resistance in Ohio became no longer effective in 197210. Given the adaptation of the P. sojae population to evade Rps1a-mediated resistance over 8 years, Schmitthenner hypothesized that Rps genes would only be effective for management for 6 to 15 years before needing to be replaced by more efficacious Rps genes11. This observation raises current concerns for the efficacy and longevity of Rps1c, Rps1k, and Rps3a for PRR management on a global scale, as they have been deployed for 35–40 years10. Currently, more than 40 Rps genes have been identified in soybean germplasm; however, only a few are widely effective and released for PRR management6,12. In North and South America, Rps1c and Rps1k are the most widely available Rps genes to farmers for PRR management, while the availability of Rps1a, Rps3a, Rps6, and Rps8 depends on country, locale, and seed company6,13.

Soybean Rps proteins initiate strong qualitative defense responses against P. sojae isolates expressing specific Avr proteins (effectors). This gene-for-gene relationship has been well characterized and is the primary means of disease management for PRR in agricultural production worldwide6,13,14,15,16. Soybean Rps genes are thought to encode for NBS-LRR (Nucleotide Binding Site–Leucine Rich Repeat) receptor proteins which recognize Avr proteins or their activity from P. sojae. When P. sojae produces Avr proteins detected by the complementary soybean Rps protein(s), genetic signals are produced that lead to a hyper-sensitive response that confers complete resistance to P. sojae17. Alternatively, P. sojae isolates that do not express effectors recognized by the cognate soybean Rps protein evade detection, continue infection, and cause PRR. A standard set of soybean germplasm, which differs in the Rps gene present in the genome, is used to characterize the gene-for-gene interaction phenotypically and identify effective Rps genes in the sampled P. sojae population18. Recently, new molecular tools have been developed to rapidly identify Avr genes present in P. sojae isolates and thus assess the pathotype profile without going through the phenotypic procedures19,20. Through surveying populations of P. sojae and describing their pathogenicity phenotype, effective resistance genes for the management of the sampled population can be identified and recommended for use by commercial producers.

Regional or state sampling of P. sojae populations has occurred in North America since the first pathotype of P. sojae was described in 196511,14,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41. While P. sojae was identified and known to be prevalent in South America since the 1970s, surveys have only been conducted since the late 1980s15,42,43,44,45,46. Surveys in Asia, primarily China, where P. sojae is an invasive pest, have been performed since 1991, when the pathogen was first identified in Heilongjiang Province, China16,47,48,49,50,51,52,53,54. Surveys in other soybean-growing countries such as Japan, Iran, and Australia have been infrequently conducted55,56,57,58. These surveys typically were performed on a state or province scale and occasionally aggregated to offer insights into Rps gene efficacy on a regional scale at a single time point13,14,38,49. In addition to identifying effective Rps genes, these studies have also been used to describe and track the pathotype complexity, the number of Rps genes which isolates can cause disease against, as well as the diversity of pathotypes to better understand the durability of Rps-mediated resistance. As the pathotype complexity of the P. sojae population increases, there will be fewer individual Rps genes that can effectively manage the known population and stacking, or having multiple Rps genes present within the genotype, will be needed for effective disease management. Likewise, observations of increasingly diverse pathotypes mean it is highly unlikely these populations would be effectively managed through the use of a single Rps gene. Pathotype complexity and diversity metrics can be used to assess the potential durability of employed Rps genes in a given population. Pathotype evaluation surveys at multiple timeframes in the same regions have increased concerns about increasing pathotype complexity and diversity, and the potential loss of effectiveness of the Rps1c and Rps1k genes for PRR management in North and South America have emerged14,36,37,38,39,44,45. However, no comprehensive worldwide analysis has been performed to evaluate the durability of Rps genes over time and across continents, which would allow for more informed planning and decision-making in global commercial trait-introduction programs to adequately manage PRR in the field.

Identifying how regional P. sojae pathotypes have evolved, as well as the durability of Rps genes for PRR management, will help guide soybean breeding and disease management recommendations on a global scale. Here, we use a novel approach of collating data from the past thirty years of P. sojae pathotype surveys from four of the largest soybean-producing countries (Argentina, Canada, China, and the United States) across three continents to show the first global-scale Rps gene fluctuation. Specifically, we aimed to: (1) identify countrywide-temporal changes in pathotype complexity, (2) evaluate Rps gene efficacy on a country-temporal scale, and (3) determine how the pathotype diversity of sampled P. sojae populations have changed over time by country. We determined that: (1) the pathotype complexity of P. sojae populations in Argentina, China, and the United States has significantly increased over time, with sampled populations now able to overcome 1–3 more of the tested Rps genes than in previously sampled time frames, (2) Rps1c and Rps1k are no longer effective in the United States, Argentina, and Canada; however there is little change in Rps1c and Rps1k efficacy observed in China between the time frames studied, and (3) the diversity of pathotypes has significantly changed over time in each country, leading to inadequate PRR disease management using currently employed Rps genes. Together, these results broadly build upon foundational research on the management of PRR. Moreover, they significantly improve our current understanding of Rps gene efficacy over time and provide a rationale for deploying new Rps genes and quantitative resistance to ensure adequate management of PRR globally.

Results

Phytophthora sojae pathotype composition within populations is thought to evolve and change over time due in part to the selective pressure of always planting soybean varieties containing the same Rps gene each rotation10. A P. sojae population adapts due to this constant selective pressure from the host through mechanisms such as outcrossing with other P. sojae genotypes59, as well as mutations60,61 or gene-silencing62 of the detected P. sojae Avr-gene product. If the soybean Rps gene is unable to detect infection by P. sojae, no defense response is activated; the plant is therefore susceptible to infection, and disease ensues. In total, 41 P. sojae pathotype studies were identified through our search of the literature. Of these, 29 were used for this systematic review (Table 1; Supplementary Fig. 6). The remaining 12 studies were not included as they were either conducted outside of the time frames studied, did not test the standard soybean Rps genes used in pathotyping, or the country in which the study took place did not have sufficient additional surveys for temporal analyses22,23,24,25,26,35,43,46,55,56,57,58. In this study, we focused on temporal changes in the efficacy of eight Rps genes (Rps1a, Rps1b, Rps1c, Rps1d, Rps1k, Rps3a, Rps6, Rps7) that have been used to describe isolate pathotypes in P. sojae pathotype surveys for decades22. In soybean production, Rps1c and Rps1k are currently the most commonly deployed genes available in cultivars for the management of PRR, while Rps1a, Rps3a, and Rps6 are available to a lesser extent6,13. We did not include Rps2, Rps3b, Rps3c, Rps4 and Rps5 in this analysis, as not all studies used these resistance genes to characterize their P. sojae pathotype composition of field populations, and none of these Rps genes are used for management of PRR in commercial soybean production in these nations.

Table 1 Studies used in this systematic review
Full size table

Pathotype complexity has significantly increased over time on a national scale

Increases in pathotype complexity within populations reflect the total number of Rps genes an individual P. sojae isolate within the population can cause disease against. Thus, these genes are ineffective for PRR management purposes. The published pathotype data from Argentina, Canada, China, and the United States were grouped into two or three distinct timeframes (generally: 1990–1999, 2000–2012, and 2013–2019) for all temporal analyses, depending on the availability of data from each country (see “Methods”; Table 1). The pathotype complexity of each tested P. sojae isolate was determined and summarized for each timeframe and country (Fig. 1). Differences in pathotype complexity between the 1990s and 2000–2013 timeframes within each country were not significant at p = 0.05, with observed increases in the mean complexity of less than 1 Rps gene (Fig. 1). Significant increases in pathotype complexity over time were observed in Argentina (t = −7.9205, df = 94.767, 95% CI = −2.367:−1.418, p = 4.485 × 10−12), Canada (t = −2.562, df = 536.52, 95% CI = −0.462:−0.061, p = 0.01067), and the United States (t = −24.954, df = 1894.5, 95% CI = −1.89:−1.62, p < 2.2 × 10−16) between the 2000–2012/2013 and 2013–2019 timeframes (Fig. 1). The most notable shift was detected between the 2000–2012/2013 and 2013–2019 timeframes in Argentina and the United States, with a mean increase in population complexity of approximately 2 Rps genes (Fig. 1). In China, a small insignificant (t = −1.545, df = 119.9, 95% CI = −0.678:0.083, p = 0.1249) increase in pathotype complexity was observed between the two timeframes, with a mean increase of 0.317 Rps genes (Fig. 1).

Fig. 1: Pathotype complexity of all isolates by timepoint for each country.
figure 1

Panel A Argentina, 1989–1999 n = 174 isolates, 2000–2012 n = 65 isolates, 2013–2019 n = 210 isolates. 1989–1999 and 2000–2012 t-test (t = −1.775, df = 97.46, 95% CI = −0.906:0.051, p = 0.079), 2000–2013 and 2013–2019 t-test (t = −7.9205, df = 94.767, 95% CI = −2.367:−1.418, p = 4.485 × 10−12), 1989–1999 and 2013–2019 t-test (t = −15.614, df = 374.61, 95% CI = −2.613:−2.029, p = <2.2 × 10−16). Panel B Canada, 2000–2013 n = 253 isolates, 2014–2019 n = 394 isolates. 2000–2013 and 2014–2019 t-test (t = −2.562, df = 536.52, 95% CI = −0.462:−0.061, p = 0.01067). Panel C China, 1990–1999 n = 97 isolates, 2000–2013 n = 790 isolates. 1990–1999 and 2000–2013 t-test (t = −1.545, df = 119.9, 95% CI = −0.678:0.083, p = 0.1249). Panel D United States, 1990–1999 n = 1115 isolates, 2000–2013 n = 956 isolates, 2013–2019 n = 1067 isolates. 1990–1999 and 2000–2013 t-test (t = −1.278, df = 1998.1, 95% CI = −0.237:0.05, p = 0.2014), 2000–2013 and 2013–2019 t-test (t = −24.954, df = 1894.5, 95% CI = −1.89:−1.62, p < 2.2 × 10−16), 1990–1999 and 2013–2019 t-test (t = −28.021, df = 2168.7, 95% CI = −1.984:−1.725, p < 2.2 × 10−16). Blue coloring denotes the studies performed in the 1990s (1989–1999, or 1990–1999), gray denotes the studies performed between 2000 and 2013, and red coloring denotes studies performed between 2013 and 2019 for each respective country. Dots indicate individual isolates pathotype complexity. Median pathotype complexity is depicted by the black bar within the box. Whiskers depict the first and third quartiles of data. Mean pathotype complexity is shown as a black circle within the boxplot. Outliers are defined as a black dot outside of the first and third quartiles. Asterisks indicate statistically significant differences between the means of groups at α = 0.05. Source data are provided as a Source Data file.

Full size image

Rps1c and Rps1k are no longer effective for PRR management in Argentina, Canada, and the United States

Repeated use of an individual resistance gene to manage a phytopathogenic population over time can lead to a loss of management efficacy of that resistance gene63,64. In this study, an Rps gene is considered ineffective at managing a P. sojae population if ≥40% of the sampled population can evade detection and cause disease. This provides a conservative estimate of when new Rps genes should begin to be introduced to manage P. sojae. Here we determine if the efficacy of the primary Rps genes used for the management of PRR has been lost over time.

In Argentina, only Rps7 was ineffective in managing the P. sojae population until the 2013–2019 time frame. Nearly 100% of isolates recovered in the first two timeframes were only pathogenic on Rps7 (Fig. 2). However, isolates recovered from 2013–2019 were pathogenic on Rps1a (84.7%), Rps1c (66.6%), and Rps1k (78.1%). More than 75% of the isolates tested were pathogenic on Rps1a and Rps1k (Fig. 2). Similarly, in the United States, Rps1a and Rps7 were the only ineffective genes on a national scale until the 2013–2019 sampling timeframe. A greater percentage of isolates recovered during this third timeframe were pathogenic on Rps1b (83.5%), Rps1c (74.5%), and Rps1k (85.5%). Likewise, increases in pathogenicity between the 2000–2013 and 2013–2019 timeframes on soybean genotypes containing Rps1a (from 60.9 to 89.1%) and Rps1b (from 44.1 to 83.5%) were also observed. Interestingly, the efficacy of Rps3a and Rps6 genes on the United States P. sojae national population appears to have remained consistent throughout the past thirty years (Fig. 2; Supplementary Fig. 1).

Fig. 2: Resistance gene efficacy for each Rps gene and timepoint interaction by country.
figure 2

Facets denote the Rps genes tested; the Y-axis is the percent of isolates that are pathogenic on a given gene at a specific time frame from each study. Panel A Argentina, 1989–1999 n = 174 isolates, 2000–2012 n = 65 isolates, 2013–2019 n = 210 isolates. Panel B Canada, 2000–2013 n = 253 isolates, 2014–2019 n = 394 isolates. Panel C China, 1990–1999 n = 97 isolates, 2000–2013 n = 790 isolates. Panel D United States, 1990–1999 n = 1115 isolates, 2000–2013 n = 956 isolates, 2013–2019 n = 1067 isolates. Blue coloring denotes the studies performed in the 1990s (1989–1999, or 1990–1999), gray denotes the studies performed between 2000 and 2013, and red coloring denotes studies performed between 2013 and 2019 for each respective country. Dots within violin plots represent percent efficacy for each Rps gene by reported years isolates were recovered from included studies within each time frame. The black bar represents the mean percent pathogenic for each time frame and Rps gene within a county. The red dashed line is at 40%, indicating when the management efficacy of a gene to the population is reduced. Source data are provided as a Source Data file.

Full size image

Small differences in the P. sojae populations’ pathogenicity on the Rps1k gene in Canada were observed between the 2000–2013 and 2014–2019 sampling timeframes, from 39.9 to 43.9% (Fig. 2; Supplementary Fig. 1). Phytophthora sojae population pathogenicity on the Rps1c gene in Canada increased substantially between sampling timeframes, with 54.1% of isolates being pathogenic on Rps1c in 2000–2013 compared to 83.7% of isolates recovered from 2014–2019 (Fig. 2; Supplementary Fig. 1). Similarly, there was an increase in the proportion of the population that was pathogenic on the Rps1d gene between 2000–2013 (13.4%) and 2014–2019 (52.5%), however, population pathogenicity on Rps6 decreased from 55.7 to 11.9%. Rps1d is not widely deployed in Canadian soybean production13; it is curious why the sampled population is adapting to this resistance gene on a national scale. Of the Rps genes tested with the population from China, only the efficacy of Rps6 changed between 1990–1999 (26.8%) and 2000–2013 (58.9%). The efficacy of the other tested Rps genes remained consistent throughout the timeframes used for analyses (Fig. 2; Supplementary Fig. 1). This could be because P. sojae is an invasive species to China and has only been present in that country since the early 1990s47. Thus, the P. sojae population in China may not be as genetically diverse or may have had less exposure to the tested resistance genes deployed compared with native North American populations.

Phytophthora sojae pathotype diversity increases over time

As P. sojae populations adapt to resistance genes, differences in the pathogenic diversity among populations can be measured65,66,67. Here we investigate differences in pathotype diversity within and among temporally sampled international populations using the Gleason diversity index (see “Methods”) and Principal Coordinates Analysis (PCoA), in tandem with testing beta-dispersion and Permutational Multivariate Analysis of Variance (PERMANOVA) to determine significant shifts in these populations over time. The Gleason diversity index exhibited an increase in pathotype diversity between the first and last timeframes for the P. sojae populations in the United States, Canada, and China (Table 2). In Argentina, diversity appeared to decrease between 1989–1999 and 2000–2012 before increasing again in 2013–2019, with an overall trend of increased pathotype diversity over time (Table 2). PCoA plots for each country revealed that the national population at each sampling timeframe, while interspersed, had their own pathotype composition (Fig. 3). A test of beta-dispersion and PERMANOVA were performed on each country individually by sampling timeframe to determine if the differences observed in pathotype composition via the PCoA plots were significant. Significant changes in beta-dispersion were determined between all time frames within each country. In the United States, pathotypes became more dispersed between the 1990–1999 and 2000–2013 time frames (beta-dispersion mean difference = −0.145, p-value < 0.001; PERMANOVA R2 = 0.025, p-value < 0.001), and began to coalesce during the 2013–2019 time frame as compared to the 2000–2013 time frame (beta-dispersion mean difference = 0.045, p-value < 0.001; PERMANOVA R2 = 0.119, p-value < 0.001). Similar findings were obtained in Argentina between the 1989–1999 and 2000–2012 time frames (beta-dispersion mean difference = −0.144, p-values < 0.001; PERMANOVA R2 = 0.234, p-value < 0.001), as well as the 2000–2012 and 2013–2019 time frames (beta-dispersion mean difference = 0.195, p-value < 0.001; PERMANOVA R2 = 0.059, p-value < 0.001). Due to the availability of data, the change in pathotype beta-diversity could only be compared between 2000–2013 and 2014–2019 in Canada (beta-dispersion mean difference = 0.091, p-value < 0.001; PERMANOVA R2 = 0.086, p-value < 0.001), and between the 1990–1999 and 2000–2013 time frame in China, of this study (beta-dispersion mean difference = −0.053, p-value < 0.001; PERMANOVA R2 = 0.018, p-value < 0.001). However, the results from Canada and China corroborate what was found in the United States and Argentina during their respective time frame. Reductions in beta-dispersion over time indicate that more isolates are becoming similarly virulent on a number of the Rps genes tested. Beta-dispersion and PERMANOVA tests corroborated alpha-diversity test results; there was a significant difference in each country’s pathotype composition between sampling timeframes, with pathotypes becoming less dispersed over time (Tables 3 and 4).

Table 2 Isolate number, unique pathotypes, mean pathotype complexity and Gleason diversity index for each population by timepoint
Full size table
Fig. 3: Principal Coordinates Analysis (PCoA) of the sampled virulence phenotype in P. sojae populations for each country colored by timepoint.
figure 3

Panel A Argentina, 1989–1999 n = 174 isolates, 2000–2012 n = 65 isolates, 2013–2019 n = 210 isolates. Panel B Canada, 2000–2013 n = 253 isolates, 2014–2019 n = 394 isolates. Panel C China, 1990–1999 n = 97 isolates, 2000–2013 n = 790 isolates. Panel D United States, 1990–1999 n = 1115 isolates, 2000–2013 n = 956 isolates, 2013–2019 n = 1067 isolates Blue coloring denotes the studies performed in the 1990s (1989–1999 or 1990–1999), gray denotes the studies performed between 2000 and 2013, red coloring denotes studies performed between 2013 and 2019 for each respective country. Dots represent the Jaccard distance matrices values for each isolate within each country and time frame, respectively. 95% data ellipses are shown for each timepoint. Source data are provided as a Source Data file.

Full size image
Table 3 Tukey HSD results on Beta-dispersion comparisons
Full size table
Table 4 Pairwise PERMANOVA comparisons of country time frames
Full size table

Discussion

These analyses on a national scale support previous observations that pathotype compositions of P. sojae populations have changed over time, and pathotype complexity has increased14,37,38,39. This study demonstrates that Rps genes, such as Rps1c and Rps1k, have lost efficacy over time across major soybean-producing areas in the world, possibly because of the adaptive traits of P. sojae populations and the imposed selective pressure of frequently planting soybean varieties with the same Rps gene. Most importantly, this study reports that Rps1c and Rps1k, the most widely utilized Rps genes within the United States, Argentina, and Canada, are no longer effective for the management of PRR in these countries.

The data used in this systematic review was limited to the geographic areas sampled in P. sojae pathotype surveys conducted across the world over the past 30 years. As with any survey-based work, there are physical and financial limitations to the total area that could be sampled within each region and variations present within the methods used. Likewise, some countries did not perform these surveys in regular intervals, as was found with the absence of recent data from China, to be able to view the current efficacy of Rps genes. Similarly, we could not identify P. sojae pathotype studies conducted in Canada between 1990 and 1999.

The main potential biases in P. sojae pathotype surveys come from the method used to obtain isolates and inoculation conditions used for pathotype characterization. Isolations from field plants present an inherent bias on isolates recovered based on the plant genotype; only isolates that can cause disease on the present genotype are recovered. To account for this potential bias, field soil is collected, and a pathogen baiting technique is performed, which incites disease in a soybean genotype containing no known Rps genes. Subsequent isolation of P. sojae from those bait plants onto an oomycete selective media has become more regularly performed68 (Table 1). Inoculation conditions, such as high temperature (>25 °C), can cause erroneous phenotypic results68. Of the 29 studies used in this systematic review, three used solely plant isolations, fifteen used solely soil baiting, and eleven used a combination of plant isolations and soil baiting to recover isolates of P. sojae (Table 1). This systematic review is also limited by the fact that plant isolation and soil baiting surveys were not separated for the analysis. All studies reported adequate inoculation conditions11,13,14,15,16,27,28,29,30,31,32,33,34,36,37,38,39,40,41,42,44,45,48,49,50,51,52,53,54.

The uniform and significant increase in pathotype complexity and diversity over time is likely due to the selective pressure imposed by the frequent planting of soybean cultivars with single qualitative resistance genes over a large area. The ‘Boom and Bust’ cycle of resistance genes to phytopathogens is well documented in potatoes and the pathogen Phytophthora infestans64. Resistance genes for management of late-blight of potato, caused by P. infestans, are typically effective for only a few years before the pathogen population adapts, rendering them ineffective64. In our study, we identified that Rps1c and Rps1k were effective for PRR management for 20–30 years, which is longer than expected for R-gene-mediated management11,69,70. This may be due to biological characteristics specific to P. sojae as it,. S is homothallic, thus primarily self-fertilizing with rare occurrences of outcrossing with other genotypes; or also due to biases in the data used for the systematic review59. Additionally, P. sojae is a soilborne pathogen and thus, does not aerially disperse over long distances. These factors may contribute to the low genomic diversity observed within sampled P. sojae populations38,71. Even with limited genomic diversity, P. sojae pathotypes continue to evolve to be more diverse and pathogenic on an increasing number of soybean Rps genes. As the diversity and complexity of pathotypes within a P. sojae population increase, the likelihood that a single host resistance gene will be able to effectively manage that population decreases. Incorporating multiple qualitative resistance genes within soybean varieties will be necessary to broadly manage P. sojae populations via Rps gene-mediated resistance.

P. sojae is thought to be native to the United States as a pathogen of the indigenous legume genus Lupinus59,71,72 and was first described shortly after soybean production started in the United States. P. sojae genotypes have been shown to be largely clonal, with most of the genetic variation within the species encompassed within four distinct genotypes59,73. Subsequent work in the United States and Argentina substantiated this work, identifying low to moderate genetic variation and evidence of potential sub-populations within the United States and therefore providing more evidence that the United States is the center of origin for P. sojae38. However, no correlation between genetic diversity and pathotype diversity within these populations has thus far been identified45,59,74. P. sojae Avr genes have been shown to be concentrated within transposon-rich regions of the genome, indicating that the adaptive evolution of Avr gene sequences is more rapid than the transposon-sparse regions of the genome75. Due to this genomic arrangement, Avr gene sequences containing single nucleotide polymorphisms, insertions, and deletions have been identified within the coding region of the P. sojae Avr1a, Avr1b, Avr1c, Avr1d, Avr1k, Avr3a, and Avr6 genes subsequently conferring virulence on the corresponding soybean Rps gene19. Thirty-one typical isolates from the 2000–2013 survey in Canada were fully sequenced and were found to contain all haplotypes reported for seven Avr genes of P. sojae throughout the world19. In addition, the sequences of those genes were compared with ca. 300 isolates from the 2014–2019 survey and found a great increase in pathotype complexity from the earlier survey that could be linked to selection pressure from Rps usage in soybean fields over the years13.

To our knowledge, Rps1d has never been deployed commercially anywhere in the world, and the gene has never been identified precisely or cloned. Surprisingly, we observed an increase in isolates pathogenic against the Rps1d gene. This may indicate that Rps1d has been deployed unwittingly in certain commercial soybean varieties, as has been previously reported for Rps676,77. Additionally, previous studies have found a significant percentage of isolates with virulence on Rps1d14,74, thus supporting the unknown presence of Rps1d in some soybean varieties. Other studies in Canada have found nearly no isolates virulent against Rps1d41, suggesting that issues existed with differentials carrying Rps1d13, a distinct possibility considering the elusive nature of the gene. The rise in virulence in the 2014–2019 survey would probably be explained by the fortuitous presence of Rps1d in many commercial lines, although this will only be answered by the eventual identification of Rps1d.

Adaptation of the P. sojae population to Rps1c and Rps1k took longer than anticipated, more than double the time originally predicted by Schmitthenner11. Research into the soybean–P. sojae molecular interactions has shown that Rps1k is able to detect and mount a defense response to both Avr1k and Avr1b78. A similar study showed Rps1c detected Avr1c and Avr1a79. This partially explains the longevity of Rps1c and Rps1k as effective PRR management tools, as both Rps1c and Rps1k would confer resistance to not only those P. sojae isolates expressing the Avr1c and Avr1k genes but also the Avr1a and Avr1b genes, respectively. In this study, the efficacy of Rps1b was shown to mirror that of Rps1k in Canada and the United States, even though Rps1b has never been knowingly deployed for PRR management in these countries (Fig. 2; Supplementary Fig. 1). This could be due to the Rps1k gene providing selection pressure against the Avr1b and Avr1k genes over time. Rps1a was already ineffective for management at the earliest time points used in this study for Canada and the United States. However, in Argentina and Canada, Rps1a and Rps1c had similar efficacy trajectories over time (Supplementary Fig. 1). Rps6 was found to be ineffective in Canada during the 200–2013 sampling time frame and then effective during 2014–2019 samplings. This could be due to the increase in geographic sampling performed, and therefore an increase in the geographically dispersed P. sojae isolates during the 2014–2019 timepoint to include Manitoba and Quebec, and therefore more representative of the Canadian P. sojae population (Fig. 2).

While the loss of the Rps1c and Rps1k genes for management was gradual in the United States over the sampling timeframes investigated, the loss of efficacy of these genes occurred more quickly in Argentina between 2000–2013 and 2014–2019 (Fig. 2; Supplementary Fig. 1). Grijalba and Gally15 hypothesized that the increase in pathotype complexity in Argentina between the 1990s and the early 2000s was due to a combination of Rps gene and conservation tillage adoption, along with the presence of the indigenous population of P. sojae present in soils which were never planted with soybean80. The first Rps genes introgressed into Argentina soybean varieties were Rps1a and Rps1c in 1984, followed by Rps1k in 199181. By 1998 these Rps genes, along with conservation tillage practices, were widely adopted in Argentina (P. Grijalba, personal observation). Following the hypothesis made by Grijalba and Gally15, pathotype complexity and diversity continued to increase through the 2000s while effective resistance genes in the region decreased. Nevertheless, the loss of Rps1c and Rps1k for management of PRR in North and South American commercial soybean production presents an immediate concern for global food security.

Single-dominant qualitative resistance genes are the most economically viable option for disease management, as plants are completely resistant to the pathogen population. However, this imposes significant pressure on the phytopathogenic community, selecting for those individuals that can mitigate, manipulate, or evade the plant’s ability to detect the pathogen. As we have shown, the pathotypes of P. sojae are diverse within populations, and single Rps gene-mediated management will not be an effective long-term management strategy going forward unless new and widely effective Rps genes are deployed into commercial soybean varieties on a regular basis. In Canada, it was found that 85% of soybean growers were using an ineffective Rps gene within their fields, showing the need for novel and effective Rps genes13. Reports from the United States estimate that effective Rps genes such as Rps3a and Rps6 were present in less than 3% of soybean varieties from 2010 to 202037. The efficacy of Rps3a and Rps6 remaining relatively stable during the past thirty years in the United States is likely due in part to their very limited availability to producers and therefore minimal selective pressure on the P. sojae populations studied here (Fig. 2; Supplementary Fig. 1). There have been over 40 Rps genes discovered in soybean germplasm, however, novel Rps genes are rarely screened during traditional pathotype survey studies12. A promising new Rps gene, Rps11, was recently identified and conferred Rps-mediated resistance to 80% of P. sojae isolates tested within their study, including even highly diverse pathotypes82. The Rps11 gene sequence and associated markers for introgression into commercial soybean varieties have been patented by Corteva Agrosciences™, which is a promising first step towards the Rps11 gene being available in commercial varieties in the near future82. However, even with the added efficiency of marker-assisted breeding, it will likely take a few years before varieties with the Rps11 gene are available. Increasing the availability of already positioned and effective Rps genes, such as Rps3a and Rps6, would benefit PRR management in the interim. A concerted, discipline-wide effort will be needed to amend the soybean Rps genes tested in pathotype surveys moving forward so that new effective Rps genes can be identified for introgression into elite soybean varieties for future deployment.

Quantitative resistance (also called “partial resistance”, “horizontal resistance”, or “field tolerance”) to P. sojae is available in commercial soybean varieties6,83,84,85. Unlike qualitative resistance, quantitative resistance is governed by multiple genes acting in concert to provide partial resistance to the pathogen by allowing infection but limiting the growth and spread of the pathogen within the plant and limiting disease development. As there are many genes acting together, there is no single focal point of selective pressure for the P. sojae population to adapt and cause disease86. Currently, more than 130 quantitative trait loci (QTL) have been identified in soybean for resistance to P. sojae, along with their associated markers for marker-assisted breeding, which are available in Lin et al.12.

A conclusion of this systematic review is the predominant Rps genes used for PRR management in major soybean-producing countries of North and South America, Rps1c and Rps1k, are no longer effective, and alternative forms of resistance will be needed to manage PRR going forward. Concentrating breeding efforts on introgressing quantitative resistance into commercial soybean varieties, in conjunction with stacking and deploying novel Rps genes, may reduce the impact of future epidemics of PRR on global food security in these countries.

Methods

This review follows reporting criteria in accordance with the Preferred Reporting Items for Systematic Review and Meta-analysis87. This systematic review has not been registered in the International Prospective Register of Systemic Reviews (PROSPERO), and the protocol has not been implemented or published before. The PRISMA checklist for this manuscript can be found in the supplement (Supplementary Note 1).

Pathotype survey data

Phytophthora sojae pathotype studies were identified using Google Scholar and Web of Science on the 21st of September 2021 and searched again for studies published after the original search date on the 6th of June 2022. Data from all identified P. sojae pathotype surveys were transcribed manually into Microsoft Excel® or supplied by authors from published manuscripts (Table 1). Each survey dataset was then validated using the ‘hagis’ R package for microbial phenotypic pathogenicity data to ensure the data collected accurately reflected what was reported in the published manuscripts88. Studies that were not regional survey-based P. sojae pathotype studies on Rps gene efficacy were not used within this temporal analysis. A curated global P. sojae virulence phenotype database was established from all identified studies as of June 6th, 2022, and published online89.

Data preparation and eligibility criteria

Using the global pathotype dataset available, Argentina, Canada, China, and the United States were identified to have pathotype survey studies conducted regularly or intermittently over time for this temporal analysis11,13,14,15,16,27,28,29,30,31,32,33,34,36,37,38,39,40,41,42,44,45,48,49,50,51,52,53,54. Studies from countries that performed minimal pathotype surveys over time, or were outside of the time frames studied, were not used in analysis22,23,24,25,26,35,43,46,55,56,57,58. Phenotypic interactions of each P. sojae isolate were first reduced to a subset of Rps genes used in this study (rps (susceptible control), Rps1a, Rps1b, Rps1c, Rps1d, Rps1k, Rps3a, Rps6, Rps7) to ensure homogeneity of phenotypic observations for each isolate. The Rps7 data from Canada was excluded as the most recent survey used a molecular tool to detect P. sojae Avr genes, and Rps7 testing was not performed13. All other pathotype surveys used the soybean hypocotyl inoculation technique for pathotype characterization, as described in Dorrance et al.68. P. sojae isolates data were then grouped by country and timeframes for analysis by the year each sample was reportedly isolated from soil or from symptomatic plants. Data from studies that were outside the discrete timeframes examined were not used in the analysis, and no isolate data was used in more than one timeframe (Table 1). Timeframes for each country and the number of isolates used in each time point for analyses were as follows: United States: 1990–1999 (n = 1115), 2000–2013 (n = 956), 2013–2019 (n = 1067) (Supplementary Fig. 2). Argentina: 1989–1999 (n = 174), 2000–2012 (n = 65), 2013–2019 (n = 210) (Supplementary Fig. 3). Canada: 2000–2013 (n = 253), 2014–2019 (n = 394) (Supplementary Fig. 4). China: 1990–1999 (n = 97), 2000–2013 (n = 790) (Supplementary Fig. 5).

Temporal analysis of pathotype complexity

Using the ‘hagis’ R package, pathotype complexity (sum of tested Rps genes an isolate can cause disease against) was determined for each individual P. sojae isolate by country before calculating the mean pathotype complexity and standard deviation of a sampled P. sojae population at each timeframe and country. A Welch’s two-sample, two-sided t-test was used to determine the significance between pathotype complexity at each timeframe by country.

Temporal comparison of Rps gene efficacy

Pathotype data for each country was grouped by timeframe, and the number and percentage of isolates pathogenic on each Rps gene was calculated as follows: ‘number of isolates pathogenic’ = (sum of isolates pathogenic on each Rps gene); percent pathogenic = ((‘number of isolates pathogenic’)/(total isolate N))*100. Percent isolate pathogenic results were plotted using ‘ggplot2’90 (Version 3.3.5) with a dashed red bar at the 40% management efficacy threshold to show effective and ineffective Rps genes for each of the sampled population timeframes. Forty percent was used as a conservative estimate for management efficacy, as it is before the majority of the population would be able to cause disease against a given Rps gene.

Temporal differences in pathotype diversity

Alpha-diversity

The Gleason diversity index, and number of unique pathotypes, were calculated for country timeframes using the ‘hagis’ R package88. The Gleason diversity index is calculated as: Gleason = ((N unique pathotypes) − 1)/log(number isolates)91. The ‘hagis’ R package automatically removes non-pathogenic phenotypic data from the dataset before calculating diversity indices so that only differences in pathogenicity are described.

Beta-diversity

To make statistical comparisons between pathotype diversity for each timeframe, the data from each country was further reduced to exclude P. sojae isolates which were non-pathogenic on all of the 8 Rps genes, and the susceptible (rps) control phenotype was removed. Jaccard distances were calculated for each country individually using the ‘vegan’ R package92 (version 2.5–7). Principal Coordinates Analysis was performed using the Jaccard distance matrix for each country and visualized with 95% data ellipses from the ‘ellipses’ R package93 (Version 0.4.2) using ‘ggplot2’90 (Version 3.3.5). Beta-dispersion tests were conducted between timeframes using the function “betadisper” from the ‘vegan’ package92. An Analysis of Variance (ANOVA) was used to identify significant differences in Beta-dispersion between timeframes, and the ad-hoc Tukey’s HSD test was used to determine significant pairwise differences between groups. A Permutational Multivariate Analysis of Variance (PERMANOVA) was performed using “adonis”, with 999 permutations, from the ‘vegan’ R package to test for significant differences in the pathotype composition of the sampled populations within each country between timeframes. PERMANOVA significance was further investigated using a pairwise PERMANOVA from the ‘RVAideMemoire’ R package94 (Version 0.9-81-2) with the function “pairwise.perm.manova” with 999 permutations.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.