A droplet digital PCR (ddPCR) assay to detect Helicoverpa armigera (Lepidoptera: Noctuidae) in bulk trap samples

Moths in the genus Helicoverpa are some of the most important agricultural pests in the world. Two species, H. armigera (Hübner) and H. zea (Boddie), cause the majority of damage to crops and millions of dollars are spent annually on control of these pests. The recent introduction of H. armigera into the New World has prompted extensive survey efforts for this species in the United States. Surveys are conducted using bucket traps baited with H. armigera pheromone, and, because the same pheromone compounds attract both species, these traps often capture large numbers of the native H. zea. Adult H. armigera and H. zea are very similar and can only be separated morphologically by minor differences in the genitalia. Thus, a time consuming genitalic dissection by a trained specialist is necessary to reliably identify either species, and every specimen must be dissected. Several molecular methods are available for differentiating and identifying H. armigera and H. zea, including two recently developed rapid protocols using real-time PCR. However, none of the published methods are capable of screening specimens in large batches. Here we detail a droplet digital PCR (ddPCR) assay that is capable of detecting a single H. armigera in a background of up to 999 H. zea. The assay has been tested using bulk extractions of 1,000 legs from actual trap samples and is effective even when using poor quality samples. This study provides an efficient, rapid, reproducible, and scalable method for processing H. armigera survey trap samples in the U.S. and demonstrates the potential for applying ddPCR technology to screen and diagnose invasive species.


Introduction
Moths in the genus Helicoverpa are some of the most important agricultural pests in the world [1,2,3]. Two species, H. armigera (Hübner) and H. zea (Boddie), cause the majority of damage to crops and millions of dollars are spent annually on control of these pests [2,4]. Prior to 2012, H. zea and H. armigera were geographically separated, with H. zea widespread in both North a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 and South America and in the Caribbean, and H. armigera prevalent throughout much of the Old World, including Europe, Asia, Africa, and Australasia [1,2]. Helicoverpa armigera was first discovered in the New World in Brazil in 2012/2013, and it has since spread to much of South America and the Caribbean, including Puerto Rico [5,6,7,8].
Although not yet present in the continental U.S., three individuals were captured in Florida in 2015 suggesting that establishment is imminent [9,10].
The extreme similarity of H. armigera to the native H. zea makes detection and identification of H. armigera difficult. It is hypothesized that H. zea split from H. armigera by a founder event from H. armigera stock colonizing the New World approximately 1.5 million years ago [11,6]. This hypothesis is supported by DNA data as well as mating compatibility between the two species and very similar morphology. Only minor differences in genitalia separate the two species morphologically [12,13], and dissection by a trained specialist is necessary to reliably identify either species.
The need to dissect every specimen for positive identification is especially problematic when conducting surveys for H. armigera adults. In the U.S., domestic monitoring for invasive pests is conducted through the USDA's Cooperative Agricultural Pest Survey (CAPS) program. The CAPS program is designed to "provide a survey profile of exotic plant pests in the United States deemed to be of regulatory significance," and it includes adult surveys for many lepidopteran pests, including H. armigera. Surveys for male H. armigera are usually performed using pheromone-baited wire-cone traps [14] or plastic bucket traps, although plastic bucket traps were economically more efficient [15,16]. The pheromone lure, a rubber septum impregnated with a mixture of Z11-16:Ald, Z9-16:Ald, and BHT [16], is placed in the bucket trap along with an insecticidal strip. The same pheromone compounds are used by H. armigera and H. zea [12], thus traps baited with H. armigera pheromone also attract male H. zea, often in large numbers; we have observed bucket traps from CAPS surveys which contain up to 500 H. zea individuals. In surveys where hundreds of non-targets are encountered per trap, it is simply not practical to use morphology as the primary means of identification.
Many molecular methods have been used to identify and separate H. armigera from its close relatives. Ming and Wang were able to differentiate between H. armigera and H. assulta using AFLP analysis [17], and Kranthi et al. [18] used RFLP analysis for separating the same taxa. Behere et al. [11] also used RFLP analysis to differentiate between H. armigera and H. zea with mitochondrial genes. Several studies have used COI sequencing and DNA barcoding to separate Helicoverpa species. Li et al. [19] used COI sequences to separate H. armigera and H. assulta. Tay et al. [6], Leite et al. [20], Mastrangelo et al. [21] (and others) used COI data to separate and characterize H. armigera and H. zea populations in the New World. Nagoshi et al. [22] used COI along with an additional marker, the Z-linked Tpi marker, to establish a method of identifying potential H. zea-H. armigera hybrids. Two studies have focused on the use of "second generation" real-time PCR to identify Helicoverpa species. Gilligan et al. [7] developed an assay based on dual-labeled hydrolysis probes capable of identifying H. armigera and H. zea using the Internal Transcribed Spacer 2 (ITS2) gene. Perera et al. [23] used real-time PCR in conjunction with dissociation curve analysis to identify H. armigera and H. zea using the Internal Transcribed Spacer 1 (ITS1) gene, and they also demonstrated that H. armigera DNA could be detected in a background of H. zea DNA at ratios of 1:24 when specimens of both species were extracted together. With the exception of Perera et al. [23], each of the aforementioned studies focus on identifying individual specimens. None of the studies listed here are capable of rapidly detecting H. armigera in a single reaction using bulk samples from pheromone traps that may contain up to 500 or more individual moths, predominantly consisting of H. zea.
Third generation PCR technology, specifically droplet digital PCR (ddPCR), has the potential to fulfill the need for a rapid, reproducible diagnostic method for detection and species determination of individual specimens from bulk samples. This relatively new technology allows for sensitive detection of very small amounts of target DNA by partitioning a single PCR mixture into approximately 20,000 individual nanoliter-sized water-in-oil droplets [24]. After thermal cycling, the droplets are read one-by-one in a specialized machine and assigned a "1" if positive and a "0" if negative (the basis for the term digital). Partitioning the PCR into thousands of individual droplets is functionally equivalent to running a separate PCR on each individual DNA fragment, leading to reliable amplification of target sequences and decreased sensitivity to PCR inhibitors. Furthermore, partitioning allows for, on average, less than one target sequence per droplet permitting the calculation of the absolute amount of target DNA in the total PCR mix [24]. Initially, detection of positive droplets in ddPCR experiments was done using hydrolysis probes, but recent advancements in ddPCR technology have allowed for a streamlined approach to identifying positive droplets with the use of EvaGreen, an intercalating DNA dye [25]. This new system allows for the direct identification of PCR products without the use of probes, reducing per-sample cost [26].
Many ddPCR studies have adapted real-time PCR methods to determine the functionality of ddPCR in previously established assays [27,28,29,30]. For example, ddPCR has been used to observe fecal contamination and the presence of invasive species in environmental DNA samples, in the detection of GMO plant specimens, and to determine the prevalence of rare mutations in circulating, cell-free DNA in patient blood samples [27,28,30,24]. These types of studies set a precedent for our experiments, as they represent varying situations in which small amounts of target DNA are identified within a large background of non-target DNA. Here we present a method for identifying low levels of H. armigera DNA in a background of H. zea DNA using EvaGreen ddPCR. The assay is tested using actual trap samples for bulk extractions consisting of legs from up to 1,000 individual specimens.

Sample collection and preparation
Specimens of H. zea used in this study were obtained from USDA CAPS pheromone traps deployed across the United States between 2012 and 2016. Those specifically used for batch extraction testing were collected using pheromone traps at the Colorado State University College of Agricultural Science Agricultural Research & Education Center (ARDEC) in September, 2016 and obtained from CAPS traps deployed in Oregon in August and September, 2015. Specimens of H. armigera used in this study were obtained from a captive laboratory colony in Germany (established using individuals from Queensland, Australia) and from specimens collected in South Africa. All specimens were identified to species by morphological dissection following Brambila [13] and/or probe-based real-time PCR following Gilligan et al. [7]. CAPS surveys were conducted on state or private land with permission of the land owner and specimens were obtained under the authority of the United States Department of Agriculture. Field collections of H. armigera in Gauteng Province, South Africa did not require a collecting permit under the Nature Conservation Ordinance of 1983 or the South African National Environmental Management: Biodiversity Act of 2004. No endangered or protected species were collected for this study. The specimens used for this study are representative of the age and condition of insect material expected to be obtained from pheromone traps. Pinned specimens were stored dry; specimens from pheromone traps were stored in envelopes at −50˚C until ready for use. Once removed from specimens, individual legs were stored in 100% ethanol at −20˚C.

Single specimen DNA extraction
Genomic DNA was extracted from individual specimens using a Qiagen DNeasy Blood and Tissue Kit (Qiagen, Valecia, California) from single legs of H. armigera or H. zea. Dry legs were crushed in a 1.5 mL microcentrifuge tube and incubated in Buffer ATL plus Proteinase K overnight at 56˚C in a GeneMate Digital dry bath (BioExpress, Kaysville, Utah). The rest of the extraction was carried out following manufacturer's instructions and eluted in 100 μL buffer AE. DNA concentration was estimated using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, Massachusetts), and is reported as the average of two readings per sample.

DNA batch extraction
DNA was extracted from H. armigera and H. zea legs using "squish buffer" as described by Perera et al. [23], modified from Gloor et al. [31]. Briefly, legs from H. armigera and H. zea were placed in 1.5 mL microcentrifuge or 7 mL Falcon tubes with a specific number of 2.3 mm zirconia/silica beads (BioSpec Products, Bartlesville, Oklahoma) and ground dry in a Mini-Beadbeater (BioSpec Products) for 55 seconds on high speed. Two beads were used for smaller amounts of tissue, with up to 30 beads for the largest samples. After the legs were reduced to a fine powder, 15 μL/leg of squish buffer (10 mM Tris-HCl, 0.5 mM EDTA, and 12.5 mM NaCl) was added to each tube (large samples were transferred to 50 mL Falcon tubes for extraction). The tubes were incubated at 80˚C in an Isotemp 102S water bath (Fisher Scientific, Waltham, Massachusetts) for 2 hours with occasional agitation for large volumes and a shaking table-top Thermomixer FP dry bath (Eppendorf AG, Hamburg, Germany) at 80˚C for 2 hours for smaller volumes. Following incubation, large samples were transferred to 1.5 mL microcentrifuge tubes and spun down at 14,000 rpm in a Centrifuge 5418 (Eppendorf AG) for 10 minutes to pellet debris. Small volume samples were transferred to or remained in 1.5 mL microcentrifuge tubes and spun down at 14,000 rpm in an Centrifuge 5418 (Eppendorf AG) for 10 minutes to pellet debris. Supernatant was transferred to clean 1.5 mL microcentrifuge tubes and stored at −20˚C until use. Batch leg extractions were carried out in ratios of (armigera:zea) of: 1:4, 1:9, 1:24, 1:49, 1:99, 1:199, 1:499, and 1:999. Additionally, a sample of 5 legs of H. armigera, 5 legs of H. zea, and a No Tissue Control (NTC) were processed in the same manner.

Real-time PCR and dissociation curve analysis
Dissociation curve differentiation of H. armigera from H. zea was carried out as described in Perera et al. [23] based on the Internal Transcribed Spacer 1 (ITS1) region of ribosomal RNA (rRNA). The conserved 18S rRNA region was used as a control. The PCR master mix included the following (primers are listed in Table 1): 0.4 μM 3374Ha_ITS1R, 0.4 μM 3373Ha_Hz_ITS1F, 0.2 μM 3377Hz_ITS1R1, 1.2 μM 3695_18S_1150F, 1.2 μM 3696_18S_1232R, 10 μL Bio-Rad SYBR Taq, and 1.2 μL sterile H 2 O; 2 μL of template DNA (from batch extractions described above) was added for a total reaction volume of 20 μL. Reactions were carried out in white Bio- The PCR mixture (20 μL total) was then added to the middle wells of disposable DG8 Cartridges for the QX100/QX200 Droplet Generator (Bio-Rad Laboratories Inc.), after which 70 μL of EvaGreen droplet generation oil (Bio-Rad Laboratories Inc.) was added to the bottom wells of the same cartridges. The cartridge was sealed with a disposable rubber DG8 Gasket (Bio-Rad Laboratories Inc.) and the cartridge was placed in the Bio-Rad QX200 Droplet Generator System (Bio-Rad Laboratories Inc.) for droplet generation. The final volume of droplets in oil was approximately 40 μL.
After droplet generation, the droplets were transferred from the top well of the DG8 cartridge to an Eppendorf semi-skirted 96-well plate (Eppendorf AG) using an Eppendorf Xplorer Plus 5-100 μL electronic pipettor (Eppendorf AG) set to the lowest draw and expel speed in an excess volume to maximize droplet recovery after transfer. The plate was sealed with a Bio-Rad Pierceable Foil Heat Seal (Bio-Rad Laboratories Inc.) in a PX1 PCR Plate Sealer (Bio-Rad Laboratories Inc.) and sealed for 5 seconds at 180˚C. The PCR reaction was carried out in a Bio-Rad C1000 Touch Thermal Cycler (Bio-Rad Laboratories Inc.) with a 96-deep well reaction module using the following program: (1) 95˚C for 5 minutes, (2) 95˚C for 30 seconds, (3) 57˚C for 1 minute, (4) steps 2 and 3 repeated for 39 cycles, (5) 4˚C for 5 minutes, (6) 95˚C for 5 minutes, and (7) an infinite hold at 4˚C. In between each step of the protocol, the ramp rate was 2˚C/second to ensure the droplet temperature changes in conjunction with the surrounding oil. Steps 5 and 6 as well as the low infinite hold temperature are specific to EvaGreen ddPCR and are required to cement the spherical shape of the droplets so they can be easily read.
After thermal cycling, the plate was placed in the block of a Bio-Rad QX200 Droplet Reader (Bio-Rad Laboratories Inc.). Droplets were read at a rate of 32 wells/hour and data were analyzed in QuantaSoft version 1.7.4 (Bio-Rad Laboratories Inc.). Single well thresholding was used to group droplets using the software's default internal algorithm. Poisson statistics were also determined by the software. Error bars represent the Poisson 95% confidence maximum and minimum based on the software's algorithm. Poisson-corrected copies/droplet (CPD) was calculated using the following equation (Bio-Rad, unpublished): To provide a better estimate of the number of positive and negative droplets and to increase reproducibility of results, the Javascript program "definetherain" [32] (http://www. definetherain.org.uk) was used to set the threshold fluorescence amplitude following each experiment. Positive controls used to calculate fluorescence thresholds consisted of runs using only H. armigera DNA. All ddPCR data reported are based on thresholds calculated by "definetherain." In each figure, the amplitude of the droplets is shown in A), positive droplets are blue while negative droplets are grey; the concentrations for each reaction are plotted in B), error bars represent the minimum and maximum Poisson 95% confidence intervals; and the table in C) shows full results from each ddPCR run. The raw data used to produce each figure is provided in S1-S5 Tables.

Initial ddPCR testing
Initial ddPCR tests were performed with purified undiluted H. armigera DNA template. Fourteen samples (plus a NTC) were selected with a range of DNA concentrations from 5.35 ng/μL to 878.5 ng/μL, as estimated by NanoDrop. Droplet digital PCR analysis followed the protocol listed above, including DNA digestion.
Initial mixed template testing was performed using DNA from two samples of H. armigera (HELICOV-121 and 146) and two samples of H. zea (HELICOV-161 and 171); DNA template concentrations ranged from 171 ng/μL to 411 ng/μL. To ensure that the number of copies of ITS1 did not exceed 100,000, which is the recommended upper limit for absolute quantification with the QX200 [33,24], DNA template was diluted 100x (based on initial DNA concentration in ng/μL). Diluted DNA template was then mixed in ratios (armigera:zea) of 1:1, 1:5, and 1:10. Diluted H. armigera DNA was used as a positive control. This experiment was repeated twice following the ddPCR protocol listed above including DNA digestion.

Bulk extraction ddPCR testing
The ddPCR protocol outlined above was performed on digested DNA from batch leg extractions in ratios (armigera:zea) of: 1:4, 1:9, 1:24, 1:49, 1:99, 1:199, 1:499, and 1:999. Each assay included DNA from H. armigera and H. zea extracted in the same manner as positive and negative controls, and a NTC. This experiment was performed twice with the same samples. Additional assays were performed with a separate set of samples for high ratios (1:199 to 1:999) and on samples with different DNA concentrations (detailed in the results).

Limit of detection
The false positive detection rate (FPR) for the assay was estimated in order to calculate the limit of detection (LoD), or the lowest number of positive droplets at which detection of H. armigera ITS1 is possible. The FPR was estimated by replicating the batch extraction protocol (including DNA digestion) on the following samples: three H. armigera; three NTCs; two batch extractions of 99 H. zea individuals; and 32 individually extracted H. zea. The FPR was calculated by tabulating the total number of positive droplets for H. zea and dividing by the total number of wells containing H. zea (34 total). The FPR was used to determine the call threshold, which was then used to determine the LoD. Threshold and LoD values were obtained from lookup tables provided by Bio-Rad based on calculations by Armbruster and Pry [34].

H. armigera was detected in batch extractions using real-time PCR
Perera et al. [23] used pooled H. armigera and H. zea legs to demonstrate that H. armigera could be detected in a background of H. zea at ratios (armigera:zea) of 1:19 and 1:24. We replicated this experiment by extracting DNA from combinations of H. armigera and H. zea legs in Helicoverpa armigera was detected using a dissociation melt curve in ratios from 1:4 to 1:49 (Fig 1A-1D). In ratios of 1:99 to 1:999, no H. armigera sequences were detected, and the results were similar to those in Fig 1F. Peak height for the 18S control is similar to that observed by Perera et al. [23] for batch extracted legs of both species, and is directly related to amplicon length and nucleotide composition. The relatively small peak for ITS1 in H. armigera is likely due inefficient amplification of this gene region in the presence of competing reverse primer for H. zea in the multiplex PCR, as discussed by Perera et al. [23].

H. armigera DNA was detected in an excess of H. zea DNA by ddPCR
Initial ddPCR experiments consisted of running H. armigera DNA extractions in a range of known concentrations to observe the relationship between ng/μL and copies/μL (Fig 2). DNA concentrations of the samples ranged from 5.35 ng/μL to 878.5 ng/μL as measured by Nano-Drop. The results showed a rough correlation between DNA concentration and copies/μL (positive droplets; Fig 2) for some samples; however, several samples had similar DNA concentrations but very different numbers of positive droplets (e.g., HELICOV-120 vs. 140).
In order to determine the utility of ddPCR for identification of H. armigera DNA mixed with H. zea DNA, we replicated the Perera et al. [23] experiment using the parameters they developed for real-time PCR, in a ddPCR experiment. Diluted DNA template from H. armigera and H. zea was mixed in ratios (armigera:zea) of 1:1, 1:5, and 1:10. Copies of H. armigera were detected in the positive control, as well as in each of the other mixed template reactions (Fig 3). DNA concentration of H. armigera did not correlate with the number of copies of ITS1 detected, although copies/μL and positive droplets/well were consistent for each specimen, regardless of the dilution factor ( Fig 3A). Sample HELICOV-121 (411 ng/μL average DNA concentration) produced an average of 2.4 copies/μL and sample HELICOV-146 (198 ng/μL average DNA concentration) produced an average of 16.3 copies/μL. Results from this experiment were used to optimize primer concentrations and PCR conditions for all following experiments.

H. armigera was detected in every ratio of batch extractions using ddPCR
The same samples used for the real-time PCR dissociation curve analysis were used to test the effectiveness of ddPCR in detecting H. armigera. The ddPCR assay was tested on batch extractions of H. armigera and H. zea legs in ratios (armigera:zea) of: 1:4, 1:9, 1:24, 1:49, 1:99, 1:199, 1:499, and 1:999. Helicoverpa armigera ITS1 sequences were detected in all ratios tested (Fig 4), and the results were consistent on a second run of the same samples. The number of positive droplets/well ranged from 9,721-9,788 for H. armigera-only to 37-40 for a 1:999 ratio of armigera:zea (Fig 4; second run not shown).
Initial bulk extractions used H. armigera legs from specimens originating from the German lab-maintained colony. To quantify the amount of DNA in these specimens, a single leg from armigera specimens were used to run initial tests of ddPCR. Helicoverpa armigera was detected in each reaction throughout the range of DNA concentrations (as measured by NanoDrop). The relationship between DNA concentration and positive droplet count is shown in (A). Concentration (copies/μL) is shown in (B); error bars represent the minimum and maximum Poisson 95% confidence intervals for copies/μL. The table in (C) shows NanoDrop readings for each sample as well as the data generated by QuantaSoft and the Poisson corrected copies/droplet. https://doi.org/10.1371/journal.pone.0178704.g002 ddPCR to detect Helicoverpa armigera each was retained and DNA was extracted individually using squish buffer. These legs yielded a relatively high concentration of DNA, approximately 300 ng/μL as measured by NanoDrop. To test the effectiveness of the assay with lower DNA concentrations, we used legs from South African H. armigera museum specimens for which DNA had been previously extracted (using a Qiagen DNeasy Blood and Tissue kit) and quantified (using NanoDrop), with the assumption that all legs from the same specimen will yield a similar amount of DNA. We selected legs from specimens with relatively low DNA concentrations (approximately 5 ng/μL), and repeated high (1:199, 1:499, and 1:999 armigera:zea) leg ratio extractions and ddPCR. Helicoverpa armigera ITS1 sequences were detected in all ratios tested, and the results were consistent on a second run of the same samples (Fig 5; second run not shown). Surprisingly, copies/μL were higher than in the previous experiment (e.g., 14.3 vs. 4.4 for 1:999), suggesting that DNA concentration estimates can be inaccurate for squish buffer extractions and that the yield of bulk extractions likely varies for each sample.

Limit of detection
False positive rates were estimated with runs of only H. zea, either singly or bulk extracted ( Fig  6). The total number of false positive droplets for all wells (49) was divided by the total number of wells containing H. zea (34) to obtain a FPR of 1.44 droplets/well (data in Table 2). Based on Bio-Rad lookup tables, the threshold to call a well positive at this FPR is 0.32%, which results in a LoD of 14 droplets/well at the 99% confidence level. Thus, the assay is able to positively detect H. armigera ITS1 locus using as few as 14 positive droplets/well, which is approximately half of the lowest number of positive droplets/well obtained from 1:999 armigera:zea batch extractions (37 droplets/well).

Discussion
Invasive pests present a serious problem with regards to crop losses and decreasing biodiversity. In the United States, H. armigera presents a serious threat to a number of economically important crops if a population becomes established. Early detection through surveys is an important step in delaying or preventing the incursion of this species. A large number of moths are captured each year in the U.S. during domestic surveys and this number increases as more and more states implement or expand surveys. Because the majority of trap catches are H. zea, and H. armigera and H. zea are externally indistinguishable, a rapid, scalable, reproducible assay to process an entire trap's contents is required. Here we detail a droplet digital PCR assay that is capable of detecting small amounts of H. armigera DNA in a background of H. zea DNA up to ratios (armigera:zea) of 1:999. The assay has been tested using bulk extractions from actual trap samples, and is effective even when using poor quality samples. The assay has a false positive rate of 1.44 droplets/well and a limit of detection of 14 droplets/well. Thus, the assays can determine, with 99% confidence, that a well contains H. armigera ITS1 DNA with as few as 14 positive droplets/well.
Our experiments represent a new, practical application for ddPCR that has been tested through all phases of sample processing (collection, extraction, assaying, and data analysis). Most previous studies have used ddPCR to amplify low concentrations of target sequences, such as when detecting endangered species or fecal contamination in water samples, but many rely upon artificially spiking predetermined amounts of purified target DNA into reactions to determine the utility of ddPCR and compare it to real-time PCR (e.g., [27,28,24]). This assay is one of the first ddPCR applications capable of detecting rare sequences in moth samples collected from field traps, and not simply in laboratory simulations.
The ddPCR assay described here does have limitations, mainly that it is not capable of identifying the total number of copies of H. armigera ITS1, and thus providing an estimate of the total number of moths. Variability in sample quality, associated differences in DNA concentration, and multiple genomic copies of ITS1 contribute to this issue. Although there is a general trend that lower DNA concentrations (ng/μL) result in fewer positive droplets, and higher DNA concentrations result in more positive droplets, the relationship is neither consistent nor linear (Fig 2). Consequently, DNA concentration cannot be reliably estimated using positive droplet count. In addition, there is an upper limit to the number of gene copies that can be identified by ddPCR due to the partitioning that occurs. The maximum number of copies/well that can be reliably quantified with ddPCR on the QX200 is 100,000, which is approximately five copies/droplet [33,24]. Other studies have been able to push that limit, with up to 131,000 copies per well quantified by Pinheiro et al. [35], but absolute quantification becomes less reliable at these upper bounds. Overall copy number quantification, however, is not a requirement for positive detection of H. armigera ITS1 DNA in a trap sample, thus there is no upper limit for the number of individual H. armigera that can be included in a single run of the assay.
We were concerned initially at how the quality of DNA from field-collected specimens would influence our ability to identify H. armigera in bulk extractions. Traps are routinely left in the field for 2-4 weeks, subjecting the specimens to non-ideal conditions (heat, moisture, etc.) that can quickly degrade DNA. Using a large data set of field collected Helicoverpa from a previous study [7], we were able to determine that the average minimum DNA concentration for an extraction from a field-collected adult was approximately 5 ng/μL. We tested the assay using specimens of H. armigera with a similar low DNA concentration and were able to detect H. armigera ITS1 sequence in ratios (armigera:zea) as high as 1:999 ( Fig 5). These data show that the ddPCR assay is effective for detection across the range of sample quality expected from field-collected specimens.
The number of copies of H. armigera ITS1 detected in batch extractions varied between ratios, but was relatively consistent for extractions at the same ratios, even when using specimens with different DNA concentrations (e.g., Fig 4 vs. Fig 5). The dilution of H. armigera DNA into a background of H. zea DNA resulted in an overall decrease in the number of positive droplets, but the decrease was not linear, likely due to the increased amount of tissue and the corresponding increased amount of extraction buffer used for higher ratios. This decrease in detectable H. armigera DNA in relation to increasing H. zea DNA signifies a practical limit to the number of moths that can be processed in a single reaction, although this limit is likely well above 1,000. We were unable to test at higher ratios (5,000 or 10,000) due to the sheer number of specimens required to replicate such an experiment. Practically, the maximum number of Helicoverpa that we have observed from a single CAPS pheromone trap is less than 500 individuals, with 50-100 the likely average. H. armigera was detected in all batch extractions using ddPCR. DNA was extracted from one leg of H. armigera pooled with an increasing number of H. zea legs in ratios (armigera:zea) of 1:4, 1:9, 1:24, 1:49, 1:99, 1:199, 1:499, and 1:999, and the resulting DNA was tested in the ddPCR assay. Helicoverpa armigera ITS1 sequences were detected in all ratios tested, and the results were consistent on a second run of the same samples. See Fig 3 legend for descriptions of A Two factors allow us to be certain that positive droplets detected in the bulk extraction experiments do, in fact, contain copies of H. armigera ITS1: (1) the amplitude of the positive droplets in comparison to negative droplets and intermediate "rain" is consistent, and (2) positive droplets/well for all experiments exceed the LoD of 14 droplets/well at the 99% confidence level. In cases where the H. zea control or NTC samples were scored as containing positive droplets, the amplitude of the "positive" droplets is far less than the 20,000 average observed for wells containing H. armigera processed using the final optimized bulk extraction protocol  (34) to obtain a FPR of 1.44 droplets/well. Based on Bio-Rad lookup tables, the threshold to call a well positive at this FPR is 0.32%, which results in a LoD of 14 droplets/well at the 99% confidence level. See Fig 3 legend for descriptions of A (Figs 4 and 5). Post-processing of droplet data using the program "definetherain" greatly improved our ability to call positive droplets and eliminated ambiguity for intermediate droplets ("rain"). Results using fluorescence thresholds determined by "definetherain" were more repeatable than simply relying on QuantaSoft's default threshold algorithm.
Data generation with ddPCR is fast, reproducible, and scalable. The rate limiting step for the experiments presented here is removing an individual leg from each moth. The workflow presented here could be improved by incorporating Bio-Rad's plate-based AutoDG Automated Droplet Generator and a more rapid sampling method, possibly by grinding whole moths rather than removing individual legs. Using the current protocol, up to 1,000 moths can be processed in a single well, allowing for screening of up to 96,000 moths on a single plate, and providing a very efficient method for screening for H. armigera in the United States.
Supporting information S1