Methods

Field sites

Field work was conducted in the Torres Strait and the Northern Peninsula Area of Cape York Peninsula, Queensland (Australia) between 2017 and 2019. Trial work took place across three sites on Thursday Island, including the Frog Gully community garden (referred to as “FGG”), a roadside near Thursday Island Hospital (“TIH”) and Green Hill (“GHF”), and one site on the Australian mainland in the town of Injinoo (“INJ”) where L. sativae has never been recorded. Active populations of L. sativae were present at the FGG and TIH sites, while activity was uncommon at the GHF site. The INJ site falls outside the current known range of L. sativae on the mainland. In addition to the trial work, the methods were tested on samples that had been collected as part of regular surveillance activities throughout the Torres Strait, including on Zuna Island, Horn Island and other regions on Thursday Island.

Leaf mine preservation

After collection of plant material, leaf mines were photographed and preserved into either 100% ethanol or onto a Whatman® FTA card, both confirmed as suitable preservation techniques via a pilot study (see Supplementary Figure S2). For mines stored into ethanol, extra leaf material was cut away from the mine, and the mines were placed into a 2 mL Axygen® tube with enough ethanol to submerge the mine (~ 0.75 – 1 mL). Leaf mines were typically up to 2.5 mm wide, and between 20 to 100 mm in length. For mines preserved onto FTA cards, leaves were rubbed, mine side down for about 30 seconds, onto the surface of the card. FTA cards were stored at 4 °C and ethanol samples at -20 °C until analysis.

Experimental groups

‘Positive control’ samples refer to mines that were preserved with the larvae still present within the leaf. ‘Zero day’ samples refer to mines that were collected before larvae had naturally emerged, but the larva was then carefully removed by excision, before the mine was preserved. ‘Unmined’ samples refer to leaves that, after having been isolated in a mesh bag while still on living plants for at least four days, showed no signs of mines, and were therefore taken as absent of larvae. However, while these leaves had no visible signs of mining, L. sativae may still have been present in the area and may have had opportunity to deposit DNA on the leaf in the form of saliva and/or eggs that failed to hatch.

Experiment 1 – Testing of unmined leaves in the field

Trials were undertaken in the Torres Strait and Cape York Peninsula of Australia to investigate the potential for false positives from unmined leaves by the L. sativae eDNA method using siratro weed (Macroptilium atropurpureum ), a favoured host of L. sativae in the Torres Strait (Blacket et al. 2015). In 2018, ten random leaves of M. atropurpureum that showed no visible signs of leaf mining were selected at FGG. Each leaf was individually enclosed in a small mesh bag for the duration of the trial. The mesh bags were designed to prevent adult flies accessing the leaf surface, and thus prevent egg lay. After 11 days, each leaf was removed from their respective bag, visibly inspected for the presence of leaf mining, and placed into sealed plastic bags. In 2019, additional field trials were undertaken. A similar approach to the one described above was used except 15 random M. atropurpureum leaves were selected, and the trial repeated at three locations. These were FGG, GHF and INJ. At FGG and GHF, leaves were collected four days after the mesh bags were first installed, and immediately placed into sealed plastic bags. At the INJ site, a revisit was not possible after four days. The likelihoodof L. sativae presence at this site was extremely low, so the unmined leaves were immediately collected and placed into sealed plastic bags. Plant samples from the 2018 and 2019 trials were transported back to the laboratory and stored at -20 °C prior to molecular testing.

Experiment 2 - eDNA persistence trial

To determine the appropriate timeframe to examine the persistence of leafminer eDNA in the field, a pilot trial was conducted on a related and widespread species, L. brassicae (Riley), from which it was found that eDNA remained in leaf mines for at least 28 days under laboratory conditions (see Supplementary Figure S3). A field trial was then conducted at FGG on Thursday Island between July - August 2018 involving L. sativae . Seventy-three active leaf mines were identified on M. atropurpureum . These mines were randomly allocated to experimental groups ranging from 0 to 28 days. Within 24 h, as a result of the rapid lifecycle of L. sativae in this tropical location, some of the larvae in selected leaf mines had already exited the mine. For those that did not emerge within 24 h, the larvae were carefully excised manually, using a thin pair of tweezers, ensuring that the emergence hole created was no larger than for natural emergences. In this way, all larvae emerged, either naturally or artificially, on the same day. A photograph was taken of each mine at this point, for later reference.
Each leaf was then enclosed in a small mesh bag to ensure no further egg lay, and thus no additional leaf mining. Between 9 and 14 leaves were collected on days 0, 1, 3, 7, 14 and 28, and these were placed into separate sealed plastic bags and stored at -20 °C prior to molecular testing.
In the first four days of the trial, all leaves were monitored closely to ensure only one leaf mine developed per leaf. If additional mines were observed forming due to eggs already present in the leaf before the addition of the mesh bag hatching, we immediately excised these additional larvae to prevent further development of the unwanted mines and replaced the mesh bags. To ensure the correct mine was ultimately collected, the photograph taken on day 0 of the desired mine was referenced upon collection. Any leaves for which the original mine was intersected by the formation of a new mine were discarded from the experiment.
A temperature logger (iButton® Maxim Integrated) was placed inside a mesh bag and positioned in the shade among M. atropurpureumleaves. The logger recorded temperature and humidity every 10 minutes for the duration of the trial.

Experiment 3 - eDNA sensitivity under field conditions

The field-based sensitivity of the L. sativae eDNA method was explored on Thursday Island. Field-based sensitivity here refers to the proportion of leaves, all of which are known to have at some point been exposed to L. sativae DNA as a result of leaf mining, but for which the age, concentration, and level of degradation of the DNA is unknown, which yield positive detections via the eDNA test. Thus, the goal of this experiment was not to determine the actual threshold concentration of DNA which could be detected by the eDNA test (this was determined in the laboratory, see below), but rather to determine a realistic field measure of sensitivity, as the parameters of DNA age, concentration and degradation will almost always be unknown from field collected leaf mine samples. In May 2019, 288 mined leaves of M. atropurpureum were randomly selected from FGG and TIH (144 leaves at each location). Ninety-two leaf mines from each site were excised and placed into 2 mL Axygen® tubes with 100% ethanol. The remaining 52 leaves from each site were preserved onto FTA cards, following the methods described above. Prior to preservation, each mine was scored by its appearance as either fresh, medium or old (since mine age was unknown) and checked under the microscope for any remains of a larva (see Table 1 for specific scoring criteria). The length of each leaf mine was also estimated, and categorised as short (< 20 mm), medium (20-50mm) or long (> 50mm).
Unmined leaf samples were the same as those used during the 2019 trials in Experiment 1.

Experiment 4 - Field applications to delimit geographic range and host range

To explore the utility of the L. sativae eDNA method to host plants beyond M. atropurpureum , we applied the test to a range of host plants, selected from field collections between 2018-2019 in Torres Strait where L. sativae is known to be present. In July 2018, leaf mines that looked similar in appearance to L. sativae mining were collected from chilli (Capsicum sp.), passionfruit (Passiflora edulis ) and basil (Ocimum basilicum ) from FGG. A single leaf mine found on snakeweed (Stachytarphetajamaicensis ) was collected from Horn Island. In May 2019, five mines each from snakebean (Vigna unguiculata ssp.sesquipedalis ), tomato (Solanum lycopersicum ), wild passionfruit (Passiflora foetida ), and yellow alder weed (Turnera ulmifolia ) were collected from FGG and stored in 2 mL Axygen® tubes with 100% ethanol. Additionally, ten leaf mines fromS. jamaicensis (from a single patch of leaf mines discovered on Thursday Island), were collected and stored in 100% ethanol. These samples were transported back to the laboratory and stored at -20 °C prior to molecular testing.
To further test the application of the eDNA methodology during delimiting survey activities, leaf mines were collected in July 2018 from seven leaf mines found in M. atropurpureum on Zuna Island, Queensland, a sparsely habited island where L. sativae had not been recorded previously and leaf mining activity was known to be very low. The leafmining damage discovered in M. atropurpureumappeared to be old, and none of the seven mines contained any active larvae that could be reared or preserved for identification. In June 2019, an empty leaf mine was also collected in Cairns, Queensland, from an eggplant (Solanum melongena ), a known host of L. sativae , but also a host of other leafminer species present in Australia.
In all instances, the empty leaf mines were closely inspected under the microscope prior to preservation, and some samples were found to contain visible remains of dead fly larvae. Sections of the mine that contained these remains were preserved, and analysed, separately from the rest of the empty mines to improve amplification of DNA.

DNA extraction

Total genomic DNA was extracted using a modified Chelex extraction protocol (Walsh et al. 1991). Individual leaf mines or 5 mm2 sections of FTA cards were placed into 1.5 ml tubes along with a 3 mm glass bead (Retsch GmbH, Haan, Germany), 5 µL of proteinase K and 200 µL of 5% Chelex solution. Each tube was then shaken in a Mixer Mill (MM300, Retsch GmbH, Haan, Germany) at 30 oscillations /s for 1 min. Samples were subsequently digested at 55 °C for 60 min, followed by a final incubation at 95 °C for 15 min with periodic vortexing. Extractions were stored at -20 °C until required. Prior to real time polymerase chain reaction (qPCR) amplification, extractions were spun at 10,000 g for 2 min. Aliquots from the bottom half of the supernatant immediately above the Chelex resin was used for qPCR amplification.

Molecular assays

Species-specific PrimeTime qPCR assays (Integrated DNA Technologies) were used to target a 109 base pair (bp) fragment of the mitochondrial CO1 gene of L. sativae with sequences as described in Sooda et al. (2017) with the addition of two degenerate bases in the probe to accommodate sequence variants found within the Torres Strait Islands: NCBI accession KR476580 Haplotype S.28 (Blacket et al. 2015) and KR476573 Haplotype S.7 (Blacket et al. 2015). Forward primer SAT-F ACCCCCTGCTTTAACTCTTTT, reverse primer SAT-R AGCACCACCATGTGCAATAA and reporter probe SAT-P CAGTATAGTAGAAAATGGRGCTGGRA with a 6-FAM/ZEN/IBFQ modification.
We also developed a species specific PrimeTime qPCR assay to target a 63 bp fragment of the mitochondrial CO1 gene of L. brassicae for a pilot trial (see Supplementary Figure S2): NCBI accession KR476570 (Blacket et al. 2015). Forward primer GCCGGAACAGGATGAACAGTTTAT, reverse primer AGATGCCCCACCGTGAG, and reporter probe CCCCTCTCTTCTATTATTG with a 6-FAM/ZEN/IBFQ modification. Primer specificity was checked using a BLAST (Basic Local Alignment Search Tool) search against the National Institutes of Health NCBI (National Center for Biotechnology Information) nucleotide database (https://blast.ncbi.nlm.nih.gov/Blast.cgi), with no close matches found outside of L. brassicae.
The qPCR assays were tested for specificity against a panel of target and off-target Liriomyza genomic DNA diluted to 10 picograms. This included L. sativae , L. brassicae , L. trifoliiand L. huidobrensis , as well as DNA from four other species, L. yasumatsui , L. katoi , L. chenopodii andL. chinensis . Amplification was only evident in the respective target species, confirming species specificity of the assays developed for both L. sativae and L. brassicae .
PrimeTime qPCR assays were conducted using a Roche LightCycler 480 system (Roche Diagnostics Australia, Castle Hill, Australia) in a 384-well format. DNA extraction and qPCR assays were performed in separate isolated rooms. Reaction volumes were 10 μL, containing 5 μL of KAPA probe force master mix (KAPA biosystems), 0.5 μL PrimeTime qPCR assay (final primer and reporter probe concentration of 500 nmol/L and 250 nmol/L, respectively), 2.5 μL ddH2O, and 2 μL of DNA. Each reaction was prepared in triplicate. Included in each 384-well assay plate were control reactions containing genomic DNA of L. sativae that was serially diluted by a factor of 10 (10 ng to 0.01 pg) and a negative control with no DNA template. Quantitative PCR amplification conditions were 3 min at 98 °C, followed by 50 cycles of 10 seconds at 95 °C and 20 seconds at 60 °C. The absolute quantification module of the LightCycler 480 software package was used to calculate the assay efficiency and total amount of DNA in unknown samples based on the genomic DNA standard. The efficiency of all qPCRs was always >90%. All extractions and qPCR analyses were undertaken in a room that is dedicated to low-quantity DNA sources, with qPCR setup undertaken in a laminar flow-hood. Positive controls and standards were added immediately prior to placing samples in a Roche LightCycler 480. Negative controls were also included at all stages (DNA extraction, qPCR) so that laboratory contamination could be identified if present.

Assay efficiency and sensitivity

A tenfold dilution series ranging from 100 to 0.001 picogram of tissue derived gDNA, measured with a Qubit 2.0 fluorometer (Invitrogen, Carlsbad, CA, USA) was prepared in elution buffer AE (Qiagen). Ten replicates of each dilution were run and the reported Cq (cycle quantification) values were used to determine amplification efficiency, the coefficient of determination (R 2) value as well as the limit of detection (LOD) and limit of quantification (LOQ) following the protocol and curve fitting method described in Klymuset al. (2019).

Data analysis

To estimate the required sampling protocol to achieve a predetermined level of diagnostic confidence under different conditions a hierarchical probabilistic model was used (Lugg et al. 2018). This model captured the nested effect of eDNA presence inside a leaf, and subsequent detectability through technical replicates of the molecular assay. Positive diagnosis of leafminer DNA was modelled as a nested Bernoulli variable: eDNA presence in the leaf mine (\(Y_{i,j}=1\)) for each site\(i\) and leafmine \(j\) depends on the probability \(\theta_{i,j}\)and; DNA presence in the extraction (\(X_{i,j,k}\)) for each extraction\(k\) depends on the parameter \(p\) and whether there was extractable eDNA in the mine (\(Y_{i,j}\)). Thus,
\begin{equation} X_{i,j,k}\ \sim\ Bern\left(\text{p\ }Y_{i,j}\right)\nonumber \\ \end{equation}\begin{equation} Y_{i,j}\ \sim\ {Bern(\theta}_{i,j})\nonumber \\ \end{equation}
The probability of viable eDNA in a leaf mine is assumed to covary with site and leaf conditions which are modelled through a logit link function:
\begin{equation} \text{logit}\left(\theta_{i,j}\right)=a+B\mathbf{x}_{\mathbf{i,j}}\nonumber \\ \end{equation}
The model was fitted to observations using JAGS . Coefficients with 95% credible intervals were reported model parameters, including the binary covariates for preservation method (FTA versus ethanol), visible presence of larval remains (present versus absent), mine age (fresh versus not fresh), and mine length (long versus short/medium).
The probability of leafminer diagnosis d can be estimated using the posterior distribution of model parameters as:
\begin{equation} d\ =\ 1\ -\ \left(\left(1\ -\ \ \theta_{i,j}\right)\left(1\ -\ \left(1\ -\ p\right)^{N_{r}}\ \right)\right)^{N_{l}}\text{\ \ \ }\nonumber \\ \end{equation}
where Nr is the number of technical replicates and Nl is the number of mined leaves tested.