Detection of blueberry stunt phytoplasma in Eastern Canada using cpn60-based molecular diagnostic assays

Blueberry stunt phytoplasma (BBSP; ‘Candidatus Phytoplasma asteris’) is an insect-vectored plant pathogen that causes severe yield losses in blueberry (Vaccinium corymbosum), which is the most valuable fruit crop in Canada. Rapid, field-based diagnostic assays are desirable tools for the control of BBSP, as part of an integrated, proactive approach to production management termed biovigilance. We designed and validated a chaperonin-60 (cpn60)-targeted LAMP assay for detection of BBSP, providing a rapid, low cost, field-deployable diagnostic option. Our validation demonstrates that the assay is reproducible, with high analytical specificity and improved sensitivity when compared with 16S rRNA nested PCR. We applied the validated LAMP assay to nearly 2000 blueberry samples from Québec and Nova Scotia over three growing seasons (2016–2018). Our surveys revealed that BBSP is present in most sites across both provinces, though detection of the pathogen in individual plants varied in different tissues across sampling dates and across years, and evidence of spread between plants was limited. To quantify pathogen load in select plants, we designed additional qPCR and ddPCR assays, also based on cpn60. We found that pathogen load fluctuates in individual plants, both within and between growing seasons. Finally, we designed an interactive map to visualize the results of our surveys. These results provide a validated diagnostic assay that can be used as part of a biovigilance strategy for detecting and controlling infections caused by BBSP.


Results
Quantitative PCR assay based on BBSP cpn60. The previously reported BBSP cpn60 UT sequence (GenBank KU523402) 27 was used to design PCR primers and a hydrolysis probe for BBSP detection and quantification in blueberry tissues. To examine the PCR efficiency of this primer/probe set, a serial tenfold dilution series of BBSP cpn60 UT-containing plasmid DNA was quantified using real-time PCR and the quantification cycle (C q ) plotted against input template amount ( Supplementary Fig. S1). The PCR efficiency (E) was determined to be > 1.99, indicating a highly efficient PCR. No amplification was observed from asymptomatic blueberry plant tissue.
The hydrolysis probe-based qPCR assay was adapted to the ddPCR format, including an internal standard based on the sequence of the V. corymbosum rubisco gene. To determine the linearity and accuracy of the ddPCR assay in this duplex format, a series of mock-infected standards were prepared consisting of DNA from uninfected blueberry leaves into which was added known, variable amounts of BBSP cpn60 plasmid DNA. Determination of the fractional abundance (FA) in each of these standards using ddPCR showed that the assay was highly linear over several orders of magnitude, although very high levels of BBSP were quantified somewhat less accurately (Fig. 1).
LAMP assay targeting 'Candidatus Phytoplasma asteris' cpn60. To provide a detection assay that was suitable for screening large numbers of samples, a LAMP assay was developed targeting the cpn60 UT of BBSP. To determine its suitability for the detection of BBSP in field-collected samples, the following assay parameters were measured: Analytical specificity. The ability of the LAMP and qPCR assays to detect various phytoplasmas was assessed using either plasmids containing cloned cpn60 UT PCR products, or total genomic DNA from infected plant tissues from a wide variety of phytoplasmas (Table 1). Positive results were observed only for samples Analytical sensitivity. The LAMP assay was capable of detecting low levels of BBSP DNA in a matrix of blueberry DNA in artificially inoculated samples (spiked), with as few as 10 copies of the BBSP cpn60 UT plasmid generating a positive result (Fig. 2). However, the linearity of the assay was low (Pearson r 2 = 0.49). The limit of detection (LOD) of the LAMP assay, defined as the C 95 , or the concentration of analyte required to give a positive result in 95% of assays 29 , was determined in terms of FA using naturally infected blueberry www.nature.com/scientificreports/ tissue samples spanning the range of observed BBSP levels. Probit analysis indicated that the C 95 of the LAMP assay corresponded to a FA of 0.463 in naturally infected samples. Similar results were obtained for the LAMP assay using an optimized curve-fitting model 30 , with a single-replicate 95% LOD of FA = 0.466 in naturally infected samples (Supplementary Table S3). This corresponded to a ddPCR-determined copy number of 60.2 (Supplementary Table S3). By contrast, the LOD for the qPCR assay, determined using plasmid templates, was 7.5 copies (Supplementary Table S3).
Precision. The repeatability of the LAMP assay was determined by examining the results of replicates of naturally infected samples at various ddPCR-determined levels of BBSP ( Table 2). As expected, the assay results measured in time to positive (T p ) were more reproducible (lower coefficient of variation) at higher levels of BBSP and less reproducible at lower levels. Similarly, the qPCR assay targeting BBSP demonstrated decreased repeatability at lower template levels (3.03% CV at 5 copies) compared to higher levels (1.10% CV at 10,000 Table 1. List of samples and/or cpn60 UT clones used to measure the analytical specificity of the LAMP assay targeting BBSP. Plasmid DNA (10 7 copies, containing the cpn60 UT) amplified and cloned from plants infected with each of the specified groups and subgroups was used as template as described in the text. When available, total nucleic acids extracted from infected plant tissue was also tested, as indicated by (a). Complete descriptions of each of these strains can be found at http:// cpncl assip hyr. ca. a These samples were also tested using DNA extracted from a plant infected with the same phytoplasma, which yielded the same result. b ND not detected.  www.nature.com/scientificreports/ copies- Supplementary Table S4). Overall, the repeatability of the C q values of the qPCR assay was higher compared to the T p values generated by the LAMP assay (Table 2; Supplementary Table S4).
Linearity. As expected, the LAMP assay displayed an inverse relationship between T p and FA, with higher FA samples featuring a lower T p than lower FA samples (Fig. 3). However, this relationship was not very strong (r 2 = 0.8), which decreases the utility of the LAMP assay for quantifying BBSP in tissue samples. However, the Spearman correlation coefficient was highly significant (ρ = − 0.772; p < 0.0001), indicating a statistically significant relationship between BBSP amount and LAMP T p . Nevertheless, in light of these results the LAMP   www.nature.com/scientificreports/ assay was ultimately implemented as a binomial, providing positive or negative results, rather than a quantitative assay. The qPCR assay (real-time format) displayed a very high correlation between BBSP cpn60 copies and observed C q (r 2 = 0.9959- Supplementary Fig. S1), as is typically observed using qPCR. The same assay in ddPCR format also showed a high correlation between BBSP cpn60 copies added to the assay and the calculated copy number (r 2 = 0.998; Fig. 1).
Comparison of methods. The commonly accepted "gold standard" assay for phytoplasma detection is nested PCR targeting the F2nR2 fragment into the 16S rRNA-encoding gene 12,[31][32][33][34] . To determine how the LAMP assay compared to this gold standard, we examined 256 blueberry tissue samples from Québec and Nova Scotia using both methods and compared the binomial results obtained by both assays for each sample. The large majority of samples that were positive by nested PCR were also positive using LAMP, providing a test sensitivity of 92% (Table 3). Conversely, the LAMP assay appeared to provide a relatively high rate of type I errors (false positives), with 30 blueberry tissue samples testing positive by LAMP but negative by nested PCR (  Table S1). These results indicate that the higher rate of positives obtained using the LAMP assay reflected not type I errors (false positives) by the LAMP assay but type II errors (false negatives) by the nested PCR assay targeting the 16S rRNA-encoding gene. These nested PCR negative but LAMP positive samples tended to have a high C q in the qPCR assay (mean C q = 30), suggesting that this discrepancy may arise from a higher analytical sensitivity of the LAMP assay. Sequences were determined for samples that tested positive by 16S nested PCR but negative by LAMP, and all were shown to contain phytoplasma DNA (data not shown). Furthermore, the 168 samples examined that tested negative by both assays included all tissue types analyzed (leaves, fruits, stems, and roots), along with 28 that were taken from tissue (fruits or leaves) that was assessed as "healthy". These results demonstrate that neither the LAMP assay nor the nested PCR assay generated spurious signals from uninfected blueberry tissues. We also examined the LAMP assay in comparison to another nested PCR targeting the rp operon of 16SrI (Aster Yellows) phytoplasmas 21 , which provided an assay sensitivity of 83% and specificity of 86% (Supplementary Table S5). The kappa statistic comparing the two assays (0.7) indicated a substantial level of agreement. For the qPCR assay, an assay sensitivity of 87% and specificity of 86% was determined compared to the Aster Yellows rp nested PCR assay (Supplementary Table S6), with a kappa statistic (0.7) that also indicated substantial agreement. In Quebec, samples were taken on highbush blueberry plants belonging mainly to the cultivars Patriot, Blueray, Bluecrop and Duke or in lowbush blueberry plants (Supplementary File S1). In Nova Scotia, all samples were collected from highbush blueberry plants belonging to Liberty, Blue Gold, Duke and Burlington (Supplementary File S1). The results of the LAMP assays for BBSP on these samples are summarized in Tables 4  and 5, and the complete results are provided as a supplemental file (Supplementary File S2). In addition, the complete results of the BBSP surveys are available as an interactive map (R code and dataframe provided as supplemental file Distribution.zip).

BBSP surveys in Eastern
BBSP was found in most of the fields surveyed in both provinces in 2016, with only 2 of 8 sites determined to lack the pathogen in all samples collected (Table 4). In some fields, the majority (up to 100%) of the plants were positive in at least one sample taken, but not all samples taken from BBSP positive plants were positive at all times (Supplementary File S2). In total, 21 plants were identified in all fields examined in 2016 that were positive for BBSP. Table 3. Sensitivity and specificity of the LAMP assay, using nested PCR targeting the 16S rRNA-encoding gene as a gold standard. 95% CI 95% confidence interval, calculated as described by Banoo et al. 25 . www.nature.com/scientificreports/ In 2017, more sites in Québec and fewer sites in Nova Scotia were surveyed. The expanded survey area in Québec identified 5 more sites that contained BBSP positive plants, and another 7 sites where the pathogen was not detected (Table 5)    In Québec, two known BBSP-infected plants and four neighboring plants each were sampled throughout the growing season at two separate sites (Supplementary Fig. S2). At one site, the infected plant (Qc8P8) remained BBSP positive in both leaves and fruit throughout the growing season, while the four neighboring plants sampled demonstrated intermittent BBSP presence (Fig. 6). At the other site, the BBSP-infected plant (Qc4 PR6-1) also remained positive over several months, while three of the neighboring plants remained BBSP-free. Late in the season, one neighboring plant tested positive for BBSP (Fig. 6).
Determination of pathogen load in infected plants. BBSP was quantified by ddPCR in a series of tissue samples taken from symptomatic blueberry plants in Québec and Nova Scotia over two or three growing seasons. Two plants were examined in Québec, which were at differing stages of blueberry stunt disease. One of these plants, Qc1P1L3, was highly symptomatic and provided only a single time point for BBSP quantification in fruit, because at all subsequent time points no fruit was produced. The FA of BBSP in the single fruit sample that was available was the highest observed in any of the leaf or stem samples, but not dramatically higher (Fig. 7A). BBSP levels in the leaf and stem samples taken subsequently remained rather high throughout 2016 and 2017 (leaf only) until the plant was removed and destroyed by the producer (Fig. 7A). In contrast, plant Qc4PR6-1 had BBSP levels that were consistently much higher in fruits compared to leaves in 2016, but the levels were closer to those in fruits and leaves in 2017, mainly due to an increase in BBSP in leaf tissue (Fig. 7B). In 2018, the levels of BBSP decreased in leaf tissues but remained high in the three fruit samples analyzed from early July to mid-August (Fig. 7B). Similarly, the two plants analyzed from Nova Scotia also displayed far higher levels of BBSP in fruits compared to leaves throughout the sampling seasons 2016-2018 (Fig. 7C,D).
The BBSP levels in the leaves were not only typically much lower than those found in the fruits but also tended to fluctuate across the sampling seasons, and sometimes strongly. For example, BBSP in the leaves of plant Qc4PR6-1 was relatively low in 2016, much higher in 2017, and low again in 2018 (Fig. 7B). These fluctuations were also observed within a sampling season in the leaf tissue-plant NS1P4 had a BBSP FA of ~ 0.2-1.3 in early July to mid-August of 2018, but the sample taken on August 29 had an approximately tenfold decrease in BBSP (FA ~ 0.01), then increased again to the levels seen earlier in the summer in the September 12 sample (Fig. 7D).

Discussion
The concept of biovigilance for crop disease management encompasses several interlinked aspects along a continuum of research foci, which begin with disease awareness and detection/identification 11 . In the case of BBSP, it is important for producers to become aware of the presence of the disease in the area, since the symptoms are variable and unreliable to make a decision, and the disease is spread by phytophagous insects. It is therefore important to have in place tools that facilitate the detection of the disease in infected plant and insect tissue. To be effective, disease detection should be as rapid and accurate as possible, and users should have confidence that detection assays have been thoroughly validated and produce results that are reliable and easily interpreted. In order to meet this need, we have designed and validated a LAMP-based diagnostic for BBSP and applied this assay to field-collected blueberry samples from two Canadian provinces over three years. The validation was carried out according to performance specifications set out in Burd's validation of laboratory-developed molecular assays for www.nature.com/scientificreports/ infectious diseases 29 , including the establishment of analytical specificity, analytical sensitivity, precision, linearity and accuracy. The LAMP diagnostic specifically detected the 16SrI 'Candidatus Phytoplasma asteris' group that includes BBSP, and these results indicate the potential utility of the assay to detect other subgroups within the geographically widely distributed 16SrI (Aster Yellows) group. The precision of this assay was demonstrated by the repeatability of multiple replicates of naturally infected samples at various levels of BBSP. The LAMP assay exhibited an inverse, albeit weak, relationship between time to positive and fractional abundance. To establish accuracy compared to established methods, a comparison of methods was carried out, comparing the LAMP www.nature.com/scientificreports/ diagnostic with nested PCR targeting the 16S rRNA gene, a traditional gold standard diagnostic for this disease. In addition to its vastly increased speed and ability to target 16SrI phytoplasmas, the LAMP assay was found to have higher sensitivity compared to the 16S nested PCR, as samples that had tested positive by LAMP but negative by 16S nested PCR were demonstrated to be true positives using PCR assays targeting other genes. The performance specifications determined by the BBSP LAMP validation compare well with other recently described LAMP assays. The improved sensitivity of LAMP over PCR is typical, as LAMP is less prone to inhibition than PCR. This has also been recently shown with regards to the development of a bacterial meningitis LAMP assay that was found to be between 10 and 10,000 times more sensitive than PCR 35 and another LAMP assay targeting Puccinia triticina causing leaf rust of wheat 500-fold greater sensitivity than PCR 36 . The analytical specificity of LAMP targeting the potato pathogen Dickeya dianthicola also showed high specificity, with 16 different D. dianthicola strains being amplified while 56 other bacterial strains were not 37 . The time to positive found by Castro et al. 38 investigating a rapid diagnosis of Zika virus was found to be in the same 13-15 min range as our work described here. An inverse relationship between time to positive and input copy number that was linear over several orders of magnitude was also demonstrated for a Strawberry Green Petal phytoplasma (16SrXIII) LAMP assay 15 .
The fact that LAMP does not require thermal cycling makes it a simple, fast and cost-effective method compared with PCR. Moreover, because LAMP is isothermal, it is possible to deploy the assay to remote, potentially resource-limited settings for nearly immediate, on-site diagnostics. Such an approach would greatly facilitate the rapid and accurate detection of BBSP in infected plant tissue. While plant tissue presented a relatively abundant source of infected material for the development and validation of this assay described here, a true biovigilance-based approach, which seeks to detect and mitigate BBSP infections before they become a problem, may require the examination of known leafhopper vectors collected in or near blueberry fields. This would facilitate the detection of potential carriers before they transmit disease. On-site detection of BBSP would require a suitable means of extracting DNA from plant or insect tissue. To facilitate this, FTA matrix cards, which are designed for DNA extraction from plant tissue, have been adapted to detect phytoplasma infections in insect tissue 39 . The suitability of LAMP coupled with rapid and inexpensive nucleic acid purification and colorimetric detection for rapid and accurate detection of SARS-CoV2 has also been recently described 40 , demonstrating the utility of this approach for pathogen detection in the context of human disease.
Despite the many advantages of this technique, there are limitations to the use of LAMP as a diagnostic tool. The LAMP amplicon is stable, and the method features a high analytical sensitivity, and so can easily crosscontaminate, leading to false positive results. Additionally, primer design can be challenging as the use of multiple primers can necessitate manual primer design and can also increase the chance of primer-primer hybridizations. Hence, it is often necessary to examine multiple LAMP primers to find a set that produces suitable results, and thorough validation of assay performance is essential. While direct sequencing of LAMP products for disease detection has been developed in the context of covid (LamPORE) 41 , the technique does not generally lend itself www.nature.com/scientificreports/ to this approach for amplicon detection, nor does the timing of the positive signal reliably indicate quantity. Therefore, other tests must be undertaken to confirm sequence identity or quantify pathogen load. Recently described, alternative isothermal amplification methods such as Recombinase Polymerase Amplification have been shown to generate results with comparable speed and sensitivity, but simpler primer design than LAMP 42 .
Having developed and validated a suitable rapid, low-cost assay, we applied it to determine the geographic distribution of BBSP in the Eastern Canadian provinces Québec and Nova Scotia. Surveys of nearly 2000 samples indicate that BBSP was widespread throughout the sampling regions in both provinces throughout the course of multiple growing seasons. Only the most severely stunted and unproductive plants warranted destruction by the producer.
To quantify BBSP levels in infected plants across time, we developed and applied a ddPCR assay. Application of this assay to select survey samples showed that BBSP fractional abundance changes over time. In some blueberry plants, BBSP levels in leaves were observed to drop below the detection limit only to re-appear at later time points in the season. Other phytoplasmas have been shown to be continually in flux within plants and over time, such as Chrysanthemum yellows phytoplasma in Chrysanthemum carinatum 43 and 'Candidatus Phytoplasma aurantifolia' in lime 44 . In this work, blueberry leaf tissue was consistently found to have lower BBSP fractional abundance compared to fruit tissue. This is consistent with some phytoplasmas preferentially being found in reproductive tissues 45 . Although higher levels of BBSP are found in fruit, leaves are more continuously available, and a more convenient substrate for DNA extraction; therefore, we recommend that in Eastern Canada early August would be the optimal time for sampling blueberry leaves for BBSP. This time of the growing season is also suitable for screening blueberry plants for viruses known to co-infect with BBSP, such as Blueberry Red Ringspot Virus and Tobacco Ringspot Virus 8 .
Only the most severely symptomatic blueberry plants were found to have very high fractional abundance of BBSP. Although very low levels of BBSP may not be associated with economic losses, and may therefore not yet warrant management interventions, knowledge about the presence of BBSP remains useful for tracking distribution, spread, disease development or evaluating efficacy of IPM programs, especially where insect vectors are present. To address whether or not BBSP is spread between neighboring plants, we examined four plants immediately adjacent to a single plant known to have a high fractional abundance of BBSP. Some evidence www.nature.com/scientificreports/ of possible spread between neighboring plants was detected in one site in Québec (Fig. 6), although further investigation will need to be carried out to answer this question definitively. In the field examined in Nova Scotia, no evidence of spread to neighboring plants within a radius of approximately 150 m was observed. These results suggest that the spread of the phytoplasma infection to neighboring plants was neither rapid nor widespread, which may be the result of insect control measures undertaken by producers. It is possible that the phytoplasma we detected in the production fields was present in the plant material upon importation and was undetected when the plants were first established. However, the origin of the disease in eastern Canada is still unknown. We have designed and validated the first cpn60-targeted LAMP assay for aster yellows phytoplasma and applied it to BBSP detection in blueberry tissues, which provides the fastest and least costly in-field option for detection that has been described to date. All molecular diagnostic tests have strengths and weaknesses, and assay choice will necessarily involve trade-offs. The high analytical specificity of the LAMP method will provide answers to very specific diagnostic questions (presence of 16SrI phytoplasmas), and therefore its use as a broad screening tool for diseases of unknown etiology is limited. In situations where assay mobility is required, but cost is less important, sequence-based methods such as Nanopore can provide more open-ended diagnostic information in cases where disease etiology is unknown or mixed infections may be suspected 46 . Moreover, Nanopore can provide sequence information outside of the targeted region, thereby providing additional taxonomic and genotypic data that can inform disease management decisions. Future work will deploy this technology to Eastern Canada to diagnose co-infection of blueberries and other crops with phytoplasmas and various viral or fungal pathogens as part of a biovigilance-based approach to disease management.

Materials and methods
Study areas and sample collection. In Nova Scotia, commercial highbush blueberry fields, located within the Annapolis Valley, were selected based on symptoms. Sampling was conducted monthly throughout the growing seasons. Locations of individual plants were recorded using a hand-held GPS unit and marked with flagging tape. Plants were imaged to capture symptoms. At each location, 5 plants were 1 young leaf near the bottom, 1 young leaf near the top, 1 old leaf near the bottom and one old leaf near the top. Additional samples included flowers, berries and stems. Plant and insect samples were placed in Styrofoam boxes with freezer packs and shipped to AAFC Saskatoon, Saskatchewan on the day of sampling or within 24 h of sampling. In Québec, plants were collected from 23 localities, and were selected and sampled in similar fashion to those in Nova Scotia (Supplementary File S1).
Sample collection complied with Canadian federal guidelines and legislation. Moreover, field access was granted and permission to collect samples was obtained from participating blueberry growers in all the fields studied.
DNA extraction. Blueberry leaves, stems, roots and fruits were collected, lyophilized for 48 h, and DNA was extracted using the Qiagen DNeasy 96 Plant kit (Hilden, Germany) according to manufacturer's instructions. At least two wells per plate were processed without a sample, serving as extraction blanks to monitor possible cross-contamination. qPCR assay development. The known sequence of the BBSP cpn60 UT (GenBank KU523402) 21 was used to design PCR primers and a hydrolysis probe for BBSP detection and quantification with Beacon Designer v7 (Premier Biosoft, Palo Alto, CA). The sequences of the amplification primers and hydrolysis probe are provided in Supplementary Table S2. All primers and hydrolysis probes were synthesized by Integrated DNA Technologies (Coralville, IA). Thermal cycling conditions for real-time qPCR were optimized using gradient PCR on a C1000 thermal cycler base with a CFX96 qPCR module (Bio-Rad) (Supplementary Table S2). The qPCR assay used 300 nM of each primer, 200 nM of hydrolysis probe, and 1 × SSoAdvanced Universal Probes Supermix (Bio-Rad). The efficiency of the qPCR assay was determined using a dilution series of the BBSP cpn60 plasmid corresponding to known copy numbers and determining the C q of each template amount, then calculating PCR efficiency (E) as described by Pfaffl 34 .
The hydrolysis probe-based assay for BBSP was adapted to the droplet digital PCR format using primer (900 nM) and probe (250 nM) concentrations suggested in the BioRad ddPCR applications guide (2014 version). In addition, to facilitate the determination of results in terms of fractional abundances (FA), a second hydrolysis probe assay was developed based on the V. corymbosum Rubisco gene (GenBank KJ773964.1) (Supplementary Table S2). Amplifications used 1 × ddPCR Supermix for Probes (Bio-Rad) and a template volume of 2 µl in a total volume of 20 µl. Template DNA was digested prior to assay using EcoRI (genomic DNA samples) or PstI (BBSP cpn60 plasmid samples) at 37 °C for 60 min. Amplification conditions for ddPCR were optimized using gradient PCR on a BioRad C1000 Touch thermocycler (Supplementary Table S2). Droplets were generated using a QX100 droplet generator (Bio-Rad), and amplifications used a C1000 Touch thermocycler (Bio-Rad), Droplets were analyzed post-amplification using a QX100 droplet reader (Bio-Rad), and results were calculated using QuantaSoft v.1.6.6 (Bio-Rad).

LAMP assay.
To provide a rapid and inexpensive assay for screening large numbers of samples, an aster yellows (AY) phytoplasma cpn60 sequence (GenBank KJ940013) was used to design primers with LAMP Designer software v. 1 www.nature.com/scientificreports/ fluorometric mastermix (New England Biolabs, Whitby, ON). Alternatively, LAMP reactions were performed using 1 × Isothermal Mastermix (Prolab Diagnostics, Richmond Hill, ON). LAMP products were generated using a C1000 Touch thermocycler (Bio-Rad) and detected visually in the case of the colorimetric mastermix or using a Genie II/III instrument (Prolab Diagnostics) or CFX96 qPCR module on a C1000 thermocycler base (Bio-Rad) for fluorometric reactions. LAMP reactions proceeded for 40 min (fluorescent detection) or 60 min (colorimetric detection) at 63 °C and fluorescent reactions were followed by an annealing protocol to determine the annealing temperature (T a ) of the amplified product (Supplementary Table S2).
Validation of the LAMP assay. The performance parameters of the LAMP assay that was developed for the detection of BBSP in plant tissues were evaluated according to recommended standards for evaluating molecular diagnostic assays 24 . Specifically, the following parameters were assessed: Analytical specificity. The ability of the LAMP assay to discern AY phytoplasmas, including BBSP, from other phytoplasmas was assessed using as template genomic DNA isolated from infected plants, and/or plasmid DNA corresponding to the cloned cpn60 amplified from characterized phytoplasma strains.
Analytical sensitivity. To examine the ability of the AY LAMP assay to detect low levels of BBSP DNA, a series of mock-infected samples was prepared in which BBSP plasmid DNA was spiked into a matrix of uninfected blueberry DNA. LAMP assays were performed on each sample in both the fluorimetric and colorimetric formats. The LAMP assay limit of detection (LOD) was assessed using a series of naturally infected blueberry plant tissues in which the fractional abundances of BBSP had been measured using ddPCR. Selected samples that spanned the full range of FA observed in 41 naturally infected samples (FA = 0.13-30.1) were assayed by LAMP 6-12 times each (total of 72 replicate assays) using fluorescent detection on a Genie instrument, and the proportion of positive results at each FA was determined. The LOD was specified as the C 95 , or the concentration of BBSP that resulted in a positive result in 95% of assays 24 , and was calculated using probit analysis (SPSS v.24). LOD was also measured using a series of curve-fitting models that report the LOD as the concentration of analyte resulting in a 95% single-replicate detection rate using the best-fitting model 30 . Results were determined on naturally infected samples for LAMP (reported as FA and as the ddPCR-determined cpn60 copy number in the sample) and on plasmid templates for qPCR.
Repeatability. Assay precision was determined by examining the time to positive (T p ) generated from all of the replicates of the naturally infected samples used for LOD determination. Repeatability was determined by calculating the coefficient of variation of the T p for each of four levels of BBSP. The repeatability of the qPCR assay was determined by measuring the C q values obtained at various levels of cpn60 plasmid template DNA (5 copies, 100 copies, 1000 copies) at 12 replicates each, then determining the coefficient of variation as for the LAMP assay.
Linearity. The relationship between the BBSP FA and the T p was examined by polynomial regression analysis of the T p determined at each level of BBSP in naturally infected samples. For qPCR, the linearity was determined in the same manner, but using plasmid templates.
Comparison-of-methods. The clinical sensitivity and specificity 25 of the LAMP assay was evaluated by comparison to an acknowledged gold standard for phytoplasma detection, nested PCR targeting the 16S rRNAencoding gene 7 . To amplify this target from phytoplasmas, primers P1 35 and P7 36 were used in an initial round of PCR to generate a product of ~ 1.8 kb. The product of this PCR was diluted 1:30 and 2 µl used as a template in a second round of PCR with primers R16F2n and R16R2 7,9 . Both rounds of PCR used 2.5 mM MgCl 2 , 500 nM of each dNTP, 400 nM of each primer and 1 U Taq DNA polymerase (Invitrogen). Thermal cycling conditions for both rounds were 95 °C, 10 min (1×); 95 °C, 1:00/55 °C, 1:00/72 °C, 1:45 (35×); 72 °C, 10:00 (1×). Samples were analyzed on a 1.5% agarose gel and were characterized as positive or negative based on the appearance of a PCR product of the expected size (1.2 kb). The same samples were analyzed using the BBSP-targeted LAMP assay as described below. Results were analyzed using a 2 × 2 table and sensitivity and specificity along with 95% confidence intervals calculated as described 25 . Samples that were discordant (LAMP positive/nested PCR negative) were re-analyzed using BBSP-specific real-time qPCR as described here (Supplementary Table S2) as well as using nested PCR targeting the ribosomal protein (rp) operon 15 . For rp operon amplification, primers rpF1/rpR1 37 were used in the first round, then the PCR product was diluted 1:30 and 2 µl used in a second round with primers rp(I)FIA/rp(I)RIA 38 . PCR and thermal cycling conditions for rp (both rounds) were as reported by Lee et al. 38 , except 1U Platinum Taq polymerase (Invitrogen) was used per reaction. In some cases the secondround rp PCR products were directly sequenced using the amplification primers (Eurofins Genomics, Toronto, ON). LAMP and qPCR assay results were compared to the AY-targeted rp assay by setting the rp assay results as the gold standard and calculating the sensitivity and specificity in the same manner.
LAMP assays on field-collected samples. LAMP reactions were performed on DNA extracted from field-collected blueberry plant tissue using 2 µl of template in a 25 µl reaction volume. Initial screens used colorimetric detection and samples that tested negative (remained pink) were not examined further. Samples that yielded a positive result (yellow) or an intermediate result (orange) were subsequently examined by fluorimetric LAMP. Samples were considered positive if they provided a positive result in the secondary screen, with a T a of the amplified product within 0.8 °C of that of the positive control (product generated from BBSP cpn60 plasmid www.nature.com/scientificreports/ DNA added at 10 5 -10 6 copies per reaction). All assays included "no template" controls with water in place of template DNA. www.nature.com/scientificreports/