Field Resistance to Rose Rosette Disease as Determined by Multi-Year Evaluations in Tennessee and Delaware

Rose rosette disease (RRD) caused by the rose rosette emaravirus (RRV) and transmitted by the eriophyid mite Phyllocoptes fructiphilus (Pf), both native to North America, has caused significant damage to roses over the last several decades. As cultural and chemical control of this disease is difficult and expensive, a field trial was established to systematically screen rose germplasm for potential sources of resistance. One hundred and eight rose accessions representing the diversity of rose germplasm were planted in Tennessee and Delaware, managed to encourage disease development, and evaluated for symptom development and viral presence for three years. All major commercial rose cultivars were susceptible to this viral disease to varying levels. The rose accessions with no or few symptoms were species accessions from the sections Cinnamomeae, Carolinae, Bracteatae, and Systylae or hybrids with these. Among these, some were asymptomatic; they displayed no symptoms but were infected by the virus. Their potential depends on their ability to serve as a source of viruses. The next step is to understand the mechanism of resistance and genetic control of the various sources of resistance identified.


Introduction
Rose rosette disease, caused by the rose rosette emaravirus (Emaravirus rosae), was first reported in the western part of North America and has since moved to the eastern coast. Within the last couple of decades, it has emerged as the major disease in cultivated roses in North America [1]. Common symptoms across host cultivars include witches' broom (shoot proliferation), hyper-thorniness, enhanced red shoot coloration, strapped leaves, and distorted flowers, all leading to poor plant growth and eventual death [1][2][3][4][5]. RRV infection spreads systemically and has been detected in the roots, stems, leaves, and flowers of infected plants [6,7].
RRV has a segmented negative-sense RNA genome enclosed by a double membrane envelope [8][9][10][11]. The genetic diversity among RRV isolates from 16 states revealed little geographic clustering, likely due to the commercial movement of infected plants [9,12]. Nucleotide similarities among examined genomes were from 97.7 to 100% [12].
RRV is dispersed by wind via the eriophyid mite Phyllocoptes fructiphilus. This mite primarily resides in protected areas within floral tissues and secondarily in leaf bases, preferentially feeding on tender new growth. It overwinters under the bark, old bud scales, flowers, and other protected places on the plant. Pf populations have a short life cycle and are prevalent throughout the growing season with periodic population spikes [13][14][15]. Mites become viruliferous after 5 days of feeding on infected plants but can transmit RRV in less than one hour [6]. Susceptible roses infested by viruliferous Pf can develop symptoms after 1 to 6 months [16][17][18]. Pf is widely distributed on wild and cultivated roses [19] with southern and northern geographic limits [20].
Roses are one of the most popular plants in public gardens and commercial/residential landscapes. Ornamentals, including roses used in home, private, and public landscapes, promote human well-being, enhance air and water quality, reduce runoff and erosion, facilitate rain capture and stormwater management, reduce noise and dust pollution, and increase property values [31]. The rose industry contributed USD 777 million in direct economic impacts to the US economy in 2014 [32]. Rose growers produced~37 million garden rose bushes worth USD 203 million in 2014 but only~25 million bushes worth USD 168 million in 2019. There has also been a decrease in shrub rose producers from 1808 to 1469 over this same period [33,34]. These losses are due primarily to the RRD epidemic [1] and black spot disease.
The landscape rose market accounts for 35% of the roses sold, but this share recently decreased by 10-15% due to RRD. The estimated loss of rose sales from the RRD epidemic is USD 5-10 million annually. A recent introduction of RRD into a California production block resulted in expenses of over USD 1,000,000 to eliminate the threat and has led others to change their protocols at a high cost to prevent the disease from getting into their production blocks. Increased production costs drive up rose wholesale and retail prices while depressing rose sales in a market already being devastated by the fear of RRD. This is reflected in a 23% decrease in annual value (adjusted for inflation), a 33% decrease in the number of shrub roses produced, and 19% fewer businesses producing shrub roses from 2014 to 2019 [33,34].
Since there is no chemical control of the virus, current control approaches include exclusion via the use of clean stock material, the decrease in inoculum/vector populations by roguing of symptomatic plants in and around the garden, the use of barriers to slow the vector's spread, and chemical control of the vector [5]. An approach that is currently lacking is the use of resistant cultivars. Disease resistance is not only the easiest control option to employ but also the most desired trait among consumers and the industry [35,36]. The willingness to pay for this and other adaptation traits is in the range of USD 10 to USD 15 per plant [32]. Producers and breeders benefit from long-market-life cultivars through increased returns for product investment. The economic impact of RRD-resistant rose cultivars is estimated to be at least USD 35 million/year. These improvements will lower the cost of production, increase the ease of growing, and increase consumer satisfaction with garden roses.
The objective of this field trial was to evaluate a suite of roses that represent the wide genetic diversity of the cultivated and wild roses for their resistance to RRD.

Field Plot Design
The Tennessee (Lily loam) and the Delaware (Elsinboro silt/loam) sites are in the USDA Hardiness zones 6b and 7a, respectively. The fields were organized using a randomized block design with three replications. Each rose plant was an experimental unit. In Tennessee, Pathogens 2023, 12, 439 5 of 13 roses were planted 2.4 m apart in a staggered double rose configuration that was 1.8 m apart. Roses were mulched~8 cm deep and irrigated with drip irrigation. Irrigation was run to ensure that the plants received one inch of water per week. Fertilization was conducted according to soil test data received each year from the UT Soil and Plant Disease/Pest Diagnostic Center (Nashville, TN, USA). In Delaware, the rows were spaced 2.4 m apart. Within rows, roses were planted with 0.9 m spacing between each plant. Supplemental watering was provided to the rose, initially by hand and later via drip irrigation. The irrigation was run up to twice a week depending on the rainfall, temperature, and plant age. Roses were planted by hand and kept weeded, mulched, and pruned during the growing season to encourage the growth of soft tissue, which is required by the eriophyid mite vector for feeding and reproduction. No fungicides or insecticides were sprayed during the trials.

RRD Augmentation Procedures
At both locations, there were Rosa multiflora plants infected by the virus and infested by the vector mite (Phyllocoptes fructiphilus), which served as a natural source of the virus and mite. In Tennessee, Knock Out Roses ® infected with rose rosette emaravirus (RRV) and infested with the vector mite were dug from a commercial bed in Nashville, TN, and transplanted into the experimental block to enhance disease pressure. The plants in both locations had disease pressure augmented by the release of large numbers of viruliferous mites (600-1200 mites per plant) directly to the roses being screened. This was performed by collecting rosetted shoots from RRV-infected roses and attaching sprigs of these (5-15 cm in length) by clipping with binders or by twist-tying to actively growing shoot tips of a target rose. Augmentation was performed 2-4 times per season with at least one symptomatic shoot attached to each test rose.

RRD Evaluation in the Field
Plants were evaluated for RRD symptoms several times from July to October each year. In Tennessee, the plants were rated for the severity of symptom development using a 0 to 3 scale (0 = no symptoms, 1 = small single shoot with rosetting, 2 = 2-3 shoots with rosetting, and 3 = 4 or more shoots with rosetting), whereas in Delaware, the plants were rated either with no symptoms or with RRD symptoms. Thus, for analysis, the severity scores were transformed into a binary score for each year and replication. The average score per plant over the 3-year trial was used as the phenotypic score for RRD. Thus, if the plant had RRD symptoms all three years, it had an average score of 1.0, and if the plant was asymptomatic all three years, the average score would be zero.
In both locations, plants were tested for the presence of RRV using the RRV2 primer pair [41]. In Tennessee, all plants were tested, whereas in Delaware, only symptomatic plants were tested for the presence of the virus. Samples of leaf tissue were either processed while fresh or kept frozen at −80 • C until processing. In Delaware, approximately 100 mg of symptomatic leaf tissue from each rose was submerged in liquid nitrogen and agitated in a Mini-Beadbeater for 40 s. Total RNA was extracted using Qiagen's RNeasy Plant Mini Kit. In Delaware, RT-qPCR amplification was performed on the same day. The products were visualized on a 2% agarose gel. In Tennessee, the viral RNA from 2 gm of homogenized rose tissue was extracted using the antigen capture protocol [42]. Virus detection was performed using RT-qPCR [41]. The cycle threshold (Ct) values for the detection of RRV were between 5 and 37 cycles. Samples with Ct values less than or equal to 29 and greater than 5 were considered strong positives, while samples between 30 and 37 were considered moderate to weakly positive for the presence of RRV [7].

Data Tabulation and Analysis
A mixed model was used to analyze the mean RRD evaluations over years with the rose accession as a fixed variable and location as a random variable using JMP Pro 16 software. The calculated means were compared to the overall mean, and the rose accessions were divided into three groups according to rose rosette symptom development-low, moderate, and high. The separation of the low and high groups from the mean score was performed at the 0.10 significance level. The results of virus testing are overlaid on these scores in Tables 2-4.   These individuals had scores that were higher than the overall mean at the 0.10 level of significance. The mixed model used had genotype as fixed and the location and interaction terms as random effects.

Disease Development over Years
The spread of the disease throughout the experimental plot as indicated by symptom development was low in the first year (0.08), increased four-fold in the second year to 0.36, and peaked in the third year at 0.47. This general pattern is seen for both locations, although the Delaware location, especially in the first year, had lower levels of symptom development than the Tennessee site. Thus, even with augmentation, it takes a three-year field trial to have good confidence in the plant's ability to resist the infection by this disease.

Cultivar × Location Interaction
The interaction between the cultivar and location was significant although the variation explained by the cultivar was much higher. Upon inspection of this interaction, it was clear that the Delaware site did not differentiate among the levels of disease incidence as did the Tennessee location. This is reflected in the greater number of cultivars with low symptom development in Delaware (34 roses) as compared to Tennessee (19 roses). All disagreements among cultivars were explained by the movement of cultivars among adjacent categories (i.e., low and moderate or moderate and high). This suggests that the interaction effect is due to non-uniform disease pressure rather than variation in virus pathogenicity.

Genetic Relationships of the Roses
The modern rose is a multi-species complex that was developed initially by combining species from the Gallicanae and Indicae sections, which was followed by using species mainly from the Systylae section and, to a lesser extent, species from the Cinnamomeae and Carolinae sections [43][44][45]. The roses tested represent this diversity of the rose group.
In our genetic relationship tree (Figure 1), the A group includes species within the sections Cinnamomeae, Carolinae, Banksianae, Laevigatae, Platyrhodon, and Bracteatae and rose cultivars with introgressions from species mentioned in the first two sections. The subgroup A1 contains the species within these sections as well as recent hybrids with R. rugosa (Purple Pavement, 'Basye's Purple', 'Therese Bugnet', 'Sir Thomas Lipton', Linda Campbell, Star Delight, and Moore's Striped Rugosa) and with R. carolina and R. virginiana ('Basye's Blueberry'). The A2 subgroup contains the cultivars with more distant introgressions with these species. As expected, the Canadian cultivars ('Winnipeg Parks', 'Morden Centennial', 'Morden Fireglow', 'Morden Blush', 'John Davis', 'John Cabot', and 'Champlain') cluster together within this A2 group. These have parentage derived from R. kordesii, R. arkansana, and R. spinosissima. It is in this A group where most of the roses with no or few symptoms are found. kordesii, R. arkansana, and R. spinosissima. It is in this A group where most of the roses with no or few symptoms are found. Figure 1. Nearest joining neighbor tree using Prevosti's distance [37] showing genetic relationships among roses. Data used: Axiom WAgRhSNP array. Names followed by an asterisk (*) and in bold font mark individuals that showed little to no RRD symptoms.
The B group contains several species of the Systylae section (R. multifora, R. wichurana, and R. setigera) as well as cultivars with recent introgression from these species. Within this group are two roses that showed no or few symptoms. The predominance of floribundas in group C indicates a significant influence of the Systylae species, especially R. multiflora. There is a lessening of Systylae background in group D with primarily shrub and grandiflora types. Group E contains species from the Indicae section and related cultivars. The hybrid tea types, which consist of intercrosses primarily among species in the Gallicanae and Indicae sections, are within the F and G groups, along with a few floribundas, grandifloras, and climbing shrubs. Group H, which includes cultivars such as Knock Out ® , Miracle on the Hudson™, and Carefree Beauty™, are distinct from groups F and G, indicating some introgressions from the section Systylae and perhaps Cinnamomeae. As such, the collection evaluated does represent the wide diversity seen among commercial rose germplasm. Figure 1. Nearest joining neighbor tree using Prevosti's distance [37] showing genetic relationships among roses. Data used: Axiom WAgRhSNP array. Names followed by an asterisk (*) and in bold font mark individuals that showed little to no RRD symptoms.
The B group contains several species of the Systylae section (R. multifora, R. wichurana, and R. setigera) as well as cultivars with recent introgression from these species. Within this group are two roses that showed no or few symptoms. The predominance of floribundas in group C indicates a significant influence of the Systylae species, especially R. multiflora. There is a lessening of Systylae background in group D with primarily shrub and grandiflora types. Group E contains species from the Indicae section and related cultivars. The hybrid tea types, which consist of intercrosses primarily among species in the Gallicanae and Indicae sections, are within the F and G groups, along with a few floribundas, grandifloras, and climbing shrubs. Group H, which includes cultivars such as Knock Out ® , Miracle on the Hudson™, and Carefree Beauty™, are distinct from groups F and G, indicating some introgressions from the section Systylae and perhaps Cinnamomeae. As such, the collection evaluated does represent the wide diversity seen among commercial rose germplasm.

RRD Resistance among Roses
The combined analysis of 108 rose accessions showed a wide range of symptom development, with all of the roses showing no or few symptoms in groups A and B. This ranged from all plants at both sites being asymptomatic to all plants at both sites being symptomatic all three years. There were 21, 67, and 20 accessions that were rated as having low, moderate, and high symptom development (Tables 2-4). Most commercial cultivars included in this trial showed moderate to high symptom development.
Among those with no or low symptom development are nine species accessions. Accessions of five North American species (R. arkansana, R. carolina, R. folialosa, R. virginiana, and R. woodsii) and one Asian species (R. rugosa) belong to two closely related sections, Carolinae and Cinnamomeae [46]. Previous work has reported that accessions of several North American species from the subgenus Carolinae (R. arkansana, R. blanda, R. carolina, R. californica, and R. palustris) did not show symptom development when they were grafted with infected buds [19], indicating that there are strong sources of resistance to RRD among this group of rose species native to North America.
Unfortunately, these have been seldom used in rose breeding due to sexual incompatibility with cultivated roses, with a couple of exceptions. Dr. Robert Basye, in his breeding in Texas, used two species from this group (R. carolina and R. virginiana) to develop the rose 'Basye's Blueberry' [47], which also showed few to no symptoms in this trial. The species R. arkansana and R. spinosissima, both of which have been previously reported to be resistant to RRD [19], are in the background of the Canadian roses 'Morden Blush', 'Morden Centennial', 'Morden Fireglow', and 'Winnipeg Parks' bred by Henry Marshall. These species were incorporated by crosses with a local accession of R. arkansana as well as through his use of the roses 'Prairie Princess' and 'Assinboine'. 'Morden Blush' and 'Winnipeg Parks' showed few symptoms, whereas 'Morden Centennial' and 'Morden Fireglow' showed moderate symptom development. RRV was detected in 'Winnipeg Parks' and 'Morden Centennial' but not in 'Morden Blush' and 'Morden Fireglow'. These have good fertility with most commercial shrub germplasm and may be good sources of RRD resistance. Currently, there is an experiment with populations developed with 'Morden Blush' and 'Morden Fireglow' to identify the genes involved in their resistance to RRD.
The other two species accessions with no or low symptoms are R. bracteata (section Bracteatae) and R. wichurana (section Systylae). Rosa bracteata, or the McCartney rose, is native to China but, once introduced into the US, became established in the southeastern USA. Rosa bracteata has been reported as resistant to the mite that transmits RRD as the mite was not able to reproduce on it [19]. Unfortunately, this species has not been used much by rose breeders due to its sexual incompatibility with most commercial germplasm derived from roses from the subgenera Systylae, Indicae, and Gallicanae. Successful hybrids such as 'Mermaid' (W. Paul, 1918) and Muriel ('MORmurl') have poor fertility [45].
Resistance from RRD is found within the Systylae section, which has been used extensively in the commercial breeding of roses [43][44][45]. Here, we report that Rosa wichurana thornless ARE showed few symptoms, whereas R. wichurana poterifolia ARE and R. soulieana RM showed moderate to severe symptoms. In addition, the source of the RRD resistance on LG5 appears to be derived from Rosa wichurana 'Basye's Thornless' [28,29]. It is important to note that RRD resistance in one accession does not mean all accessions within the species are resistant. This is clear from the results of this trial as well as previous reports of Rosa wichurana [19]. Another member of this subgenus that has been reported to be resistant to RRD is Rosa setigera [19], which is a North American native within this mainly Asian group.
Another large group of accessions with low symptom development is the hybrids with R. rugosa ('John Davis', 'Fuzzy Wuzzy Red', Moore's Striped Rugosa ('MORbeauty'), Purple Pavement ('HANpur'), 'Sir Thomas Lipton', Star Delight ('MORstar90'), and 'Therese Bugnet'). These vary extensively in their sexual compatibility depending on how recently the Rosa rugosa was used in the parentage. Those that are one or two generations from the initial cross with Rosa rugosa, such as 'Fuzzy Wuzzy', 'Purple Pavement', 'Sir Thomas Lipton', 'Star Delight', and 'Therese Bugnet', have less fertility. Nevertheless, these have some breeding potential in an introgression program to incorporate RRD resistance into a wide diversity of cultivated roses. This is the first report of RRD resistance in this species, although it should be noted that some R. rugosa hybrids, such as 'Basye's Purple', 'Fru Dagmar Hastrup', and Linda Campbell ('MORredrug'), showed moderate symptom development.
The final two accessions that developed no or low symptoms of RRD over the 3-year trial were the floribunda rose Chuckles ('SIMmimi') and the miniature rose Fairy Moss (MORfairpol'). Both of these clustered in the B group, which also contained the resistant accession of R. wichurana, the TAMU RRD resistant breeding line M4-4, and the previously reported resistant species R. setigera. Thus, it is possible that these have some Systylae background from which their resistance to RRD is derived.
Of those with no or low symptoms, about half tested positive for the presence of RRV. These include one of the rugosa accessions (Rosa rugosa Bailey), R. wichurana thornless ARE, Fair Molly, four Rosa rugosa hybrids ('John Davis', Moore's Striped Rugosa, Star Delight, and 'Therese Bugnet'), and one Canadian rose ('Winnipeg Parks') with Rosa arkansana/spinosissima background. The usefulness of these materials depends on how easily the mite can acquire and transmit the virus from these to a susceptible rose. It is possible that mite populations are lower on these roses because they do not form rosettes in which the Pf populations proliferate, but this remains to be studied experimentally. Lower mite populations could reduce the spread of the vector if, as some believe, the mite is more likely to balloon off to find another host plant under high population conditions (James W. Amrine, personal communication).
All other commercial roses and species roses included in this trial showed moderate to severe symptoms of RRD, and most tested positive for RRV. Those cultivars showing moderate RRD symptoms without a positive RRV diagnosis were 'Caldwell Pink', 'Lafter', Manetti, 'Morden Fireglow', and Sorcerer ('SAVasorc'). These roses generally developed rosettes of an RRV infection but not until late in the trial. To ensure high virus titer, rosetted tissue was tested at least twice with different samples. The potential reasons for this anomaly are the following. (1.) The rose leaf tissue contains inhibitory phenolic compounds that disrupt the extraction and/or PCR reaction [41], causing these tests to be false negatives. (2.) The virus titer was below detectable amounts. (3.) The test did not pick up the specific virus variant present in the rose. (4.) The rosetting symptoms observed were caused by a different virus, pathogen, pest, or abiotic stress. At this point, we cannot distinguish between these alternatives. These roses will be replanted and further observed for symptom development and tested for virus with multiple primers and extraction techniques to properly characterize their reaction to RRV.
Previous greenhouse work conducted in Arkansas found that out of the 25 roses examined for their susceptibility using mite and graft inoculations, 24 were susceptible, with 2 of these susceptible roses (Bonica and Homerun) being difficult to infect [6]. Both Bonica and Homerun showed moderate symptom development in the 3-year field trial. As Bonica was previously reported to be resistant to the mite [19], perhaps its moderate symptom development is related to mite behavior or biology on the rose and not its intrinsic resistance to the virus. Only the climbing rose Stormy Weather resisted infection with both techniques. Surprisingly, in the field trial, Stormy Weather showed high development of symptoms and tested positive for the virus. After confirming the identity of the rose in all trials and sequencing the virus in two locations, there was no obvious reason why the results of the two studies differed. More work needs to be conducted to systematically assess the biological diversity of the pathogen/mite vector and the environmental parameters that may affect the expression of host plant resistance to RRD.

Mechanism of Resistance
Although we have identified roses that did not test positive for RRV and/or develop symptoms of RRD under field conditions, it is not known how the plant is able to resist this viral pathogen or its vector or suppress symptom development. It is possible that the lack of symptom development for Rosa bracteata and only moderate symptom development in Bonica is due to their reported mite resistance. If this translates to lower mite numbers or less feeding, this could be a mechanism of resistance protecting these accessions.
It has also been reported that the QTL for partial resistance on LG5 affects both symptom development as well as virus titer as measured by Ct values [28,29]. This implies that this resistance QTL affects virus proliferation in the host plant. Unfortunately, the mechanism of resistance is not known at this point and needs to be further investigated by comparing virus and mite reproduction and behavior on a series of resistant and susceptible cultivars.

Conclusions
As we examine the genetic background of roses with low RRD symptom development in these trials, common parentage includes species within the Cinnamomeae and Carolinae sections. This is consistent with previous work [19] and suggests that more screening needs to be performed within this group. Low symptom development was also seen with some accessions of Rosa wichurana. This species, along with Rosa setigera, was reported to have resistance to RRD, which would support future screening of the R. wichuranaand R. setigera-derived climbing and rambling roses. Many of the roses that developed moderate to high symptoms possess a breeding background containing species primarily in the sections Systylae, Indicae, and Gallicanae.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.