Background

Cryptic species can show remarkable diversity in aspects of their ecology, behaviour, and at the level of the genome. They exist across metazoan taxa [1], including globally important arthropod pest taxa, such as whiteflies [2], disease-vectoring mosquitoes [3], fruit flies [4], thrips [5, 6] and mites [7, 8], some of which are characterised by cryptic species complexes. Discovering cryptic diversity has important consequences for estimates of global biodiversity, conservation planning, and the management of pests and diseases. Morphologically similar species can vary in pest status due to differences in genotypic and/or phenotypic traits that influence their host range and specificity, geographic distribution, the ability to vector diseases, or insecticide resistance [8–10]. Therefore, recognising cryptic species and the differences in their biology and ecology are essential for effective management, with important implications for public health, agriculture and trade.

The diamondback moth, Plutella xylostella, is the major pest of brassicaceous crops worldwide, costing an estimated US$4 to US$5 billion annually in direct losses and management costs [11, 12]. Insecticide resistance is widespread in P. xylostella populations around the world, fuelling wide-ranging research to develop alternative management tactics [11, 13]. Plutella xylostella was initially recorded in Australia in the late 1800s and rapidly became a widespread pest of Brassica vegetables, and then canola following its expanded production from the 1990s [14, 15]. Recently, Landry and Hebert [16], through mtDNA barcoding, identified a cryptic lineage of Plutella in Australia not detected in previous molecular studies of P. xylostella [14, 17–21]. Although external morphology was indistinguishable from P. xylostella, deep mtDNA divergence (8.6%), differences in genital morphology and endemism in Australia led them to describe a new species, Plutella australiana Landry & Hebert. Plutella australiana was originally collected together with P. xylostella in light trap samples in eastern Australia, suggesting at least some ecological overlap [16], but its biology, ecology and pest status were unknown.

The management of P. xylostella in Australian Brassica crops has been a significant challenge for decades [15, 22], but the discovery of P. australiana has made the relative abundance and pest status of both species in these crops uncertain. With rare exception, P. xylostella and allied species feed on plants in the order Brassicales, mainly within the family Brassicaceae [16, 23, 24], implying that the host range of P. australiana may include cultivated brassicas. Widespread resistance to pyrethroid and organophosphate insecticides has been attributed to Australian populations of P. xylostella from all vegetable and canola production regions, which has led to ineffective control during outbreaks [22, 25]. Plutella xylostella is well known as a migratory insect with a high capacity for gene flow [11, 13], facilitating the rapid spread of resistance alleles. Australian P. xylostella are thought to disperse frequently, based on indirect evidence from ecological and genetic studies [14, 15, 26]. Most studies have found a lack of genetic differentiation at microsatellite loci and low sequence variation in mitochondrial DNA markers among Australian and New Zealand populations of P. xylostella, consistent with high gene flow and/or recent ancestry [14, 15, 17, 18]. While species identification was not in question in these studies, somewhat inconsistent findings in two studies from eastern Australia using allozymes or SSR markers [19, 20] might reflect the confounding presence of P. australiana samples [16]. Given these considerations, future management of Plutella in Australian crops will require thorough understanding of the ecological requirements, genetic traits and pest status of the two Plutella species. In addition, reproductive isolation between these two species is unknown but has implications for evolutionary inference and the potential for gene flow. The capacity for hybridization and introgression could lead to the exchange of insecticide resistance or other adaptive alleles [27, 28].

Although mtDNA markers are widely used in studies of species identity and population structure [29–31], mitochondrial variation within or between species can be influenced by direct and/or indirect selection, or introgressive hybridization [32, 33]. One factor that can confound mtDNA-based inference is interaction with inherited bacterial symbionts [34, 35]. Wolbachia is a widespread endosymbiont thought to infect at least half of arthropod [36] and 80% of lepidopteran [37] species. It is mainly transmitted vertically from infected females to their offspring through the egg cytoplasm, and inheritance is therefore linked with mtDNA. To facilitate its spread, Wolbachia manipulates host reproductive biology to favour the fitness of infected females by inducing host phenotypes that distort sex ratios (male-feminization, male-killing or induction of parthenogenesis) or cause sperm-egg cytoplasmic incompatibility (CI) [38, 39]. In the simple case involving a single CI-inducing strain, crosses with infected females are fertile but crosses between uninfected females and infected males fail to produce offspring. If maternal transmission is efficient and infected females have a reproductive advantage, Wolbachia infection can spread rapidly through an insect population [40], driving a selective sweep of a single haplotype and reducing mtDNA diversity [41]. Limited surveys to date have identified Wolbachia strains infecting P. xylostella at low frequency in populations from North America, Africa, Asia and Europe [18, 42, 43]. Because symbionts can contribute to reproductive isolation and influence mtDNA diversity [34, 44], assessing their role can provide important insights into host evolution and population structure [35, 45–47].

Here we investigated the biology, ecology and population genetic structure of two cryptic Plutella species by collecting Plutella from brassicaceous plants throughout Australia and screening individuals to identify mtDNA lineages and Wolbachia infections. For a subset of populations, we examined genetic diversity using thousands of nuclear SNPs from across the genome. In addition, we assessed reproductive compatibility in laboratory crosses and determined the susceptibility of each species to commercial insecticides.

Methods

Sample collection

Plutella larvae (rarely, eggs or pupae) were collected from canola crops, Brassica vegetable crops, forage brassicas and wild brassicas throughout Australia between March 2014 and December 2015 (Table 1). The wild species included wild radish, Raphanus raphanistrum, turnip weed, Rapistrum rugosum, sea rocket, Cakile maritima, Ward’s weed, Carrichtera annua, African mustard, Brassica tournefortii, and mixed stands of sand rocket, Diplotaxis tenuifolia, and wall rocket, D. muralis. At each location, at least 25 individuals were collected from randomly selected plants to achieve a representative sample. Insect samples were collected from Brassica vegetables by hand, from sea rocket by beating plants over a collection tray and from other hosts using a sweep net. Each population sample was separately reared in ventilated plastic containers on leaves of the original host material for 1–2 days and thereafter on cabbage leaves. Non-parasitised pupae or late-instar larvae were fresh frozen at −80 °C.

Table 1 Collection details showing the frequency (f) of Plutella species and Wolbachia infections among Plutella populations from Australia
Full size table

DNA isolation and COI genotyping

For each population sample, we aimed to genotype a minimum of 16 individuals where possible after removing parasitized individuals. Individual pupae (but not larvae) were sexed under a dissecting microscope, then genomic DNA was isolated by homogenising whole individuals followed by two phenol and one chloroform extractions according to Zraket et al. [48]. DNA was treated with RNase A, then precipitated and re-suspended in TE buffer. Plutella lineages were distinguished using a PCR-RFLP assay [49]. A 707 bp COI region was amplified using a combination of two primer pairs: (i) PxCOIF (5′-TCAACAAATCATAAAGATATTGG- 3′) and PxCOIR (5′-TAAACTTCAGGGTGACCAAAAAATCA- 3′), and (ii) PaCOIF (5′-TCAACAAATCATAAGGATATTGG- 3′) and PaCOIR (5 ′-TAAACCTCTGGATGGCCAAAAAATCA- 3′). Ten microliter reactions were run with 2 µL of MyTaq 5x buffer, 0.2 µL of each primer (10mM stocks), 1 µL of DNA (approx. 5 ng) and 0.05 µL of MyTaq polymerase (Bioline). Samples were amplified at 95 °C for 2 min, then 35 cycles at 95 °C for 10 s, 52 °C for 20 s, 72 °C for 30 s followed by a 5 min final extension at 72 °C. PCR products were digested at 37 °C for 1 h with 1 unit of AccI (NEB) restriction enzyme with 2 µL Cutsmart Buffer in a 20 µL reaction. Following digestion, products were separated using agarose gel electrophoresis (1.5%). Plutella xylostella products are approximately 516 bp and 191 bp and P. australiana products are 348 bp and 359 bp [49]. To examine mtDNA haplotypes, sequencing of the 707 bp COI amplicon was performed for 44 P. xylostella and 37 P. australiana individuals at the Australian Genome Research Facility (AGRF). In addition, we downloaded sequence trace files from Landry and Hebert [16] (dx.doi.org/10.5883/DS-PLUT1) and re-analysed, aligned and trimmed all sequences in GENEIOUS version 10.0.6 [50]. Haplotype networks were constructed using R package pegas version 0.9 [51].

Wolbachia screening and phylogenetics

Wolbachia infection was detected using two separate PCR assays of the 16S rRNA gene (16S-2 and 16S-6) according to Simoes et al. [52]. To identify Wolbachia strains, the Wolbachia surface protein (wsp) gene was sequenced in a subset of individuals. Amplification was performed using wsp81F and wsp691R sequence primers [53]. Amplicons were sequenced using the reverse primer and aligned in GENEIOUS version 10.0.6 [50]. We used a 493 bp alignment to construct a maximum likelihood phylogeny in RAxML version 8.2.4 [54] using a general time reversal substitution model [55] with 1000 bootstrap replicates.

RADseq library preparation and sequencing

Libraries were prepared for restriction-site-associated DNA sequencing (RADseq) according to a protocol modified from Baird et al. [56]. Genomic DNA was quantified using a Qubit 2.0 fluorometer (Invitrogen) and 200 ng digested with 10 units of high fidelity SbfI in Cutsmart Buffer (NEB) for 1 h at 37 °C, then heat inactivated at 80 °C for 20 min. One microlitre of P1 adapter (100nM) with a 6-base molecular identifier (MID) (top strand 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGxxxxxxTGCA- 3′, bottom strand 5′-[P]xxxxxxCTGTCTCTTATACACATCTGACGCTGCCGACGA- 3′, x represents sites for MIDs) were then added using 0.5 µL T4 DNA ligase (Promega), 1 nM ATP and Cutsmart buffer. Library pools were sheared using a Bioruptor sonicator (Diagenode), then DNA fragments end-repaired using a Quick Blunting Kit (NEB), adenine overhangs added then P2 adapters (top strand 5′-[P]CTGTCTCTTATACACATCTCCAGAATAG- 3′, bottom strand 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGT- 3′) ligated. DNA purification between steps was performed using a MinElute PCR purification kit (Qiagen). Libraries were amplified using KAPA HiFi Hotstart Readymix (Kapa Biosystems) and Nextera i7 and i5 indexed primers with PCR conditions: 95 °C for 3 min, two cycles of 98 °C for 20 s, 54 °C for 15 s, 72 °C for 1 min, then 15 cycles of 98 °C for 20 s, 65 °C for 15 s, 72 °C for 1 min followed by a final extension of 72 °C for 5 min. Libraries were size-selected (300-700 bp) on 1–1.5% agarose gel and purified using a minElute Gel Extraction Kit (Qiagen), then Illumina paired-end sequencing was performed using HiSeq2500 (100 bp) or NextSeq500 (75 bp) at the AGRF.

Read filtering and variant calling

Sequence reads were demultiplexed using RADtools version 1.2.4 [57] allowing one base MID mismatch, then TRIMMOMATIC version 0.32 [58] was used to remove restriction sites, adapter sequences and a thymine base from reverse reads introduced by the P2 adapter, and quality filter using the ILLUMINACLIP tool with parameters: TRAILING:10 SLIDINGWINDOW:4:15 MINLEN:40. Paired reads were aligned to the P. xylostella reference genome (accession number: GCF_000330985.1) using STAMPY version 1.0.21 [59] with --baq and --gatkcigarworkaround options and expected substitution rate set to 0.03 for P. xylostella and 0.05 for P. australiana to reflect expected levels of sequence divergence relative to the P. xylostella reference genome. Duplicate reads were removed using PICARD version 1.71 [60]. Genotypes were called using the Genome Analysis Toolkit (GATK) version 3.3-0 [61, 62] HaplotypeCaller tool. We determined that base quality score recalibration using bootstrapped SNP databases was inappropriate for this dataset as it globally reduced quality scores. For downstream comparisons between species, we joint-genotyped P. australiana and P. xylostella individuals using the GATK GenotypeGVCFs workflow. To examine finer scale population structure, we also joint-genotyped the P. australiana individuals alone. All variant callsets were hard-filtered with identical parameters using VCFtools version 0.1.12a [63]: We removed indels and retained confidently-called biallelic SNPs (GQ\(\geqslant \)30) genotyped in at least 70% of individuals with a minimum genotype depth of 5, minQ\(\geqslant \)400, average site depth of 12–100, minimum minor allele frequency of 0.05, in Hardy-Weinberg equilibrium at an alpha level of 0.05. To avoid linked sites, we used the VCFtools --thin function to retain only SNPs separated by a minimum of 2000 bp. To estimate genetic diversity, we generated a set of all confidently-called variant and invariant sites (GQ\(\geqslant \)30), and hard filtered to remove sites within repetitive regions and retain sites genotyped in at least 70% of individuals with an average site depth of 12–100. Sites from the mitochondrial genome were excluded from all datasets.

Genetic diversity and population structure

Genetic diversity was calculated for Plutella populations of both species from five locations. The R package hierfstat [64] was used to calculate observed heterozygosity, gene diversity and the inbreeding coefficient, FIS, according to Nei [65]. Population means for site depth and number of SNPs, indels and private sites were calculated using the --depth function and vcfstats module in VCFtools version 0.1.12a [63]. The number of heterozygous sites within individuals was determined from all confidently-called sites excluding indels using a custom python script parseVCF.py [66] and visualised using R [67].

To examine population structure in P. australiana, a global estimate of FST [68] with bootstrapped 99% confidence intervals (104 bootstrap replicates) was calculated in R package diveRsity [69]. Pairwise FST values for all population pairs were calculated and significance determined using exact G tests (104 mc burnins, 103 batches, and 104 iterations per batch) in GENEPOP version 4.6 [70] after Bonferroni correction for multiple comparisons. Separate analysis of population structure was performed using the Bayesian clustering program STRUCTURE version 2.3.4 [71], first for all individuals of co-occurring Plutella species, and second for P. australiana alone. For all runs, we used a burnin length of 5×105 followed by a run length of 106 MCMC iterations and performed ten independent runs for each K value from 1 to 10, where K is the number of genotypic clusters, using a different random seed for each run, assuming the locprior model with correlated allele frequencies and λ set to 1. The optimal value of K was determined using the delta K method [72] implemented in STRUCTURE HARVESTER [73] and inspection of the likelihood distribution for each model. Q-matrices were aligned across runs using CLUMPP version 1.1.2 [74] and visualised using DISTRUCT version 1.1 [75].

Laboratory cultures of Plutella species

Laboratory cultures of P. australiana and P. xylostella were established from field populations and used for crossing experiments and insecticide bioassays. Plutella adults were collected at light traps at Angle Vale and Urrbrae, South Australia, in October–November 2015. Females were isolated and allowed to lay eggs, then identified using PCR-RFLP and progeny pooled to produce separate cultures of each species. A laboratory culture of the Waite Susceptible P. xylostella strain (S) has been maintained on cabbage without insecticide exposure for approximately 24 years (≈310 generations) and was used as a bioassay reference strain. All cultures were maintained in laboratory cages at 26 ± 2.0°C and a 14:10 (L:D) hour photoperiod at the South Australian Research and Development Institute, Waite Campus, Adelaide, South Australia. The P. australiana culture was maintained on sand rocket, Diplotaxis tenuifolia, and the P. xylostella culture was maintained on cabbage, Brassica oleracea var. capitata. The purity of cultures was assessed regularly using PCR-RFLP.

Crossing experiments

Plutella australiana and P. xylostella pupae were sexed under a stereo microscope, then placed into individual 5 mL clear polystyrene tubes with fine mesh lids and gender visually confirmed after eclosion. Enclosures used for crossing experiments were 850 mL polypropylene pots (Bonson Pty Ltd) modified with lateral holes covered with voile mesh for ventilation. Crosses of single mating pairs were performed on laboratory benches at 26 ± 2.0 °C and 14:10 (L:D) photoperiod using 3-week old D. tenuifolia seedlings as the host plant. After seven days, adults were collected into a 1.5 mL tube and fresh frozen at −80 °C for species confirmation using PCR-RFLP. Seedlings were examined and eggs counted under a stereo microscope, then returned to enclosures to allow egg hatch. Larvae were provided with fresh 3–4 week old seedlings until pupation, then pupae were individually collected into 5 mL tubes. Hybrid F1xF1 crosses and back-crosses were then performed as above. The presence of egg and adult offspring was recorded for all replicates, and for the majority of replicates (> 80%), the numbers of offspring were counted and used to calculate a mean.

Insecticide bioassays

Insecticide susceptibility of field-collected Plutella strains was compared to the susceptible P. xylostella (S) reference in dose-response assays using four commercial insecticides: Dominex (100 g L−1 alpha-cypermethrin), Proclaim (44 g kg−1 emamectin benzoate), Coragen (200 g L−1 chlorantraniliprole) and Success Neo (120 g L−1 spinetoram). Bioassays were performed by placing 3rd instar larvae onto inverted leaf discs embedded in 1% agar in 90 mm Petri dishes. Cabbage leaves, Brassica oleracea. var. capitata were used for P. xylostella and canola leaves, B. napus var. ‘ATR Stingray’, were used for P. australiana. Eight concentrations and a water-only control were evaluated for each insecticide using four replicates of ten larvae. A 4 mL aliquot of test solution was applied directly to leaves using a Potter Spray Tower (Burkard Manufacturing Co. Ltd.) calibrated to deliver an aliquot of 3.52 ± 0.09 mg cm-1. After application, each dish was placed in a controlled temperature room at 25 ± 0.5 °C, then mortality was assessed after 48 h (Dominex, Success Neo) or 72 h (Proclaim, Coragen). Dose-response analysis was performed using log-logistic regression in R package drc [76] and the fitted models were used to estimate the lethal concentration predicted to cause 50% (LC50) and 99% (LC99) mortality of the test population. Resistance ratios were calculated by dividing the LC50 and LC99 estimates for field strains by the corresponding LC estimates for the P. xylostella (S) reference strain.

Results

Geographic distribution and host associations

Plutella larvae were collected from brassicaceous plants at 75 locations in Australia and 1477 individuals were genotyped at the COI locus using PCR-RFLP to identify species. Of these, 88% (n=1300) were genotyped as P. xylostella, 10% (n=147) as P. australiana and 2% (n=30) were unresolved (Table 1). Plutella australiana was identified in around one quarter (n=20/75) of collections distributed across southern Australia, while P. xylostella occurred at all locations except Cunnamulla, Queensland, in a collection from wild African mustard, Brassica tournefortii (Table 1). The sex ratio was not different from 1:1 for P. xylostella (481 females, 517 males, χ2=1.2986, p=0.2545) or P. australiana (63 females, 55 males, χ2=0.5424, p=0.4615). The relative incidence and abundance of P. australiana was >2-fold higher in the eastern state of New South Wales than in other states (Fig. 1). Plutella australiana larvae were detected in 29% (n=5/17) of collections from wild brassicas and from species including wild radish, Raphanus raphanistrum, wild turnip, Rapistrum rugosum, African mustard, B. tournefortii, and mixed stands of sand rocket, D. tenuifolia and wall rocket, D. muralis (Table 2). Among cultivated crops, P. australiana larvae occurred in 36% (n=14/39) of samples from canola, consisting of 11% of total Plutella individuals from those crops, but were not identified from commercial Brassica vegetable farms (Table 2). However, P. australiana eggs were collected from kale at one farm.

Fig. 1
figure 1

The geographic distribution of P. xylostella (light grey) and P. australiana individuals (black) in larval collections from brassicaceous plants in Australia during 2014 and 2015. Pie diagrams show the relative proportion of each species at each location. Overlapped pies represent locations with 100% P. xylostella. Green highlighted circles indicate five locations from which individuals of each species were RAD sequenced

Full size image
Table 2 Frequency of P. australiana in Plutella collections from different Brassica host types
Full size table

Wolbachia infections

Plutella individuals (n=1447) were screened for Wolbachia infection using 16S rRNA PCR assays. Only 1.5% (n=19/1300) of P. xylostella collected from eight different locations were infected (Table 1). In contrast, all 147 P. australiana individuals were infected with Wolbachia across the 20 locations where this species occurred. To identify Wolbachia strains, a Wolbachia surface protein (wsp) amplicon was sequenced from 14 P. xylostella and 30 P. australiana individuals. Each species was infected with a different strain. The wsp sequence for Australian P. xylostella showed 100% identity to a Wolbachia supergroup A isolate infecting P. xylostella from Malaysia, plutWA1 [18]. For P. australiana, the wsp sequence showed 100% identity to a Wolbachia supergroup B isolate infecting a mosquito, Culex pipiens, from Turkey and the winter moth, Operophtera brumata, from the Netherlands (Fig. 2).

Fig. 2
figure 2

Maximum likelihood phylogeny of 493 bp of Wolbachiawsp amplicons for Plutella and other arthropods. The strain infecting P. australiana (wAus) was identical to a Wolbachia supergroup B strain reported from Culex pipiens and Operophtera brumata. The strain infecting Australian P. xylostella was identical to a supergroup A strain (plutWA1) reported from Malaysian P. xylostella. Labels include the Wolbachia strain, host species and GenBank accession number. Labels in bold denote strains sequenced in this study. The scale bar shows the mean number of nucleotide substitutions per site

Full size image

Crossing experiments

Inter-species single pair mating experiments showed that hybridization between P. australiana and P. xylostella was possible, yet less successful than intra-species crosses. While most intra-species crosses produced adult offspring, the fecundity of P. xylostella was >2-fold higher than P. australiana (Table 3). Both reciprocal inter-species crosses produced F1 adult offspring, but success was asymmetric and notably higher in the pairs with P. australiana females. In this direction, there was a strong male bias in the F1 progeny: from 76 cross replicates, 16 collectively produced 9 female and 80 male adults, a ratio of 8.9. Hybrid F1xF1 crosses for both parental lines produced F2 adult offspring (Table 4). For the P. australiana maternal line, parental back-crosses using F1 hybrid males successfully produced offspring, while parental back-crosses with F1 hybrid females were sterile. For the P. xylostella maternal line, low fitness allowed only a single parental back-cross replicate, which involved a hybrid female and was sterile.

Table 3 Fecundity of intra-species and reciprocal inter-species single pair crosses of P. australiana (P.aus) and P. xylostella (P.x)
Full size table
Table 4 Fecundity of hybrid F1 crosses and back-crosses
Full size table

Mitochondrial haplotype diversity

Mitochondrial haplotype networks of Australian Plutella were constructed using a 613 bp COI alignment that included 81 sequences from this study and 108 from Landry and Hebert [16]. We found low haplotype diversity within Australian P. xylostella, consistent with previous reports [17, 18, 77]. Only five haplotypes were identified among 102 individuals, including three identified by Saw et al. [17] and three occurring in single individuals (Fig. 3a). The most common haplotype, PxCOI01, occurred at high frequency and differed by a single base mutation from other haplotypes (Fig. 3a, Additional file 1: Table S1). Nine closely related haplotypes were identified in 87 P. australiana individuals with seven occurring in single individuals (Fig. 3b). The most common haplotype, PaCOI01, occurred at high frequency and differed by 1-2 base mutations from other haplotypes (Fig. 3b, Additional file 1: Table S2).

Fig. 3
figure 3

Mitochondrial DNA haplotype network for aP. xylostella (n=102,44 from this study, 58 from [16]) and bP. australiana (n=87, 37 from this study, 50 from [16]) individuals from Australia based on a 613 bp COI sequence alignment. Haplotypes shared by more than one individual are shown in circles with a grey border with the number of individuals indicated inside the circle. Haplotypes connected by a line differ by a single mutation

Full size image

Nuclear diversity and population structure

At five collection locations, P. australiana co-occurred with P. xylostella in sufficient numbers to enable comparison of nuclear genomes, though the relative abundance of species varied between locations. To ensure representation from the south-west region of Australia, the Esperance population (n=5) was formed by including one P. australiana individual from Boyup Brook. Despite only two P. xylostella individuals at Gilgandra, this population had 17 P. australiana individuals and was included. To generate nuclear SNP markers, we performed RADseq for a total of 52 P. australiana and 47 P. xylostella individuals.

Illumina sequencing and demultiplexing using RADtools [57] yielded 276.8 million raw sequence reads. Following read quality filtering and mapping, genotypes were called for 99 individuals from the two species. Hard filtering retained 300,241 confidently-called variant and invariant nuclear sites at a mean depth >36 per individual, and a subset of 689 widely-dispersed nuclear SNP variants (to avoid linkage bias) at a mean depth >36 per individual, for comparative analyses of genetic diversity and population structure. The dataset with all confidently-called sites was used to estimate population-level genetic diversity.

Estimates of nuclear genetic diversity across 300,241 variant and invariant sites revealed a striking contrast between Plutella species, with notably higher diversity within populations of P. australiana than co-occurring populations of P. xylostella (Table 5). The mean observed heterozygosity within populations ranged from 0.013–0.016 for P. australiana and 0.009–0.010 for P. xylostella. Similarly, the average numbers of SNPs, indels and private alleles were considerably higher within P. australiana populations. As P. australiana may have fixed nucleotide differences relative to the P. xylostella reference genome that may affect population level statistics, we also removed indels from this dataset and directly compared the heterozygosity among individuals using 289,347 sites. Plutella australiana individuals had on average a 1.5-fold higher proportion of heterozygous sites than P. xylostella individuals (Fig. 4).

Fig. 4
figure 4

Boxplot showing the proportion of heterozygous sites across 289,347 confidently-called nuclear sites for individuals of P. xylostella (light grey boxes, n=47) and P. australiana (dark grey boxes, n=52) from five locations. Heterozygosity was consistently higher in P. australiana

Full size image
Table 5 Population statistics for variant and invariant sites for sympatric populations of P. australiana (P. aus) and P. xylostella (P. x) from five locations
Full size table

Genetic structure among co-occurring populations of Plutella species was investigated using 689 widely-dispersed nuclear SNPs in the program STRUCTURE. The delta K method predicted a strong optimal at K=2 genotypic clusters. Plutella australiana and P. xylostella individuals were clearly separated into distinct genotypic clusters in accordance to their species identified through mtDNA genotypes regardless of geographic location (Fig. 5, left panel). Five individuals across four locations showed greater than 1% admixture as shown by sharing of colored bars.

Fig. 5
figure 5

Proportional assignment of Plutella individuals to genotypic clusters, K, based on STRUCTURE analysis. Individuals are represented by vertical bars and genotypic clusters are represented by different colors. Left panel: Analysis at K=2 for 52 P. australiana and 47 P. xylostella individuals sorted left-to right by proportion of cluster membership. The predominantly red bars correspond to P. australiana individuals and the predominantly blue bars correspond to P. xylostella individuals identified through mtDNA genotypes. Locations are labelled for five individuals showing >1% genotypic admixture. Right panel: Analysis at K=3 for 52 P. australiana individuals sorted left-to-right by proportion of cluster membership within geographic locations, showing a high degree of genotypic admixture among individuals across locations

Full size image

Assessing population structure from datasets with multiple species can mask heirachical structure [78]. To address this, genotypes were separately called for 52 P. australiana individuals, and hard filtering retained a set of 974 widely-dispersed SNP variants at a mean depth >33 per individual for examination of finer scale structure among five populations. The delta K method predicted a weak modal signal at K=3, but the highest likelihood value occurred at K=1. Bar plots for K=3 showed a high degree of admixture among individuals across the five populations, consistent with high levels of gene flow across Australia (Fig. 5, right panel). Pairwise FST was then calculated for the five P. australiana populations using 974 SNPs. The global estimate of FST was not significantly different from zero, indicating the populations are not differentiated (FST=0.0002, 99% CI = -0.0274–0.0387). Further, pairwise FST values were low, ranging from –0.0041 to 0.0038, suggesting substantial gene flow among populations separated by distances of between 341 and 2700 kilometres (Table 6).

Table 6 Pairwise comparisonsa of Weir and Cockerham’s [68] FST (below diagonal) and geographic distance in kilometres (above diagonal) among populations of P. australiana from five locations
Full size table

Insecticide susceptibility

Bioassays revealed highly contrasting responses to insecticide exposure in P. xylostella and P. australiana field strains (Fig. 6). Plutella australiana showed extremely high susceptibility to all four insecticides evaluated: resistance ratios at the LC50 and LC99 estimates were less than 1.0 and showed that this strain was 1.5-fold to 7.4-fold more susceptible than the laboratory P. xylostella (S) reference (Additional file 1: Table S3). In contrast, resistance ratios at the LC50 for the field P. xylostella strain ranged from 2.9 for Success Neo to 41.4 for Dominex, indicating increased tolerance to all insecticides. Comparison of the LC99 estimates with commercial field doses for each insecticide implies differences in field efficacy between species. The commercial field rate of Dominex was >8-fold lower than the LC99 for P. xylostella, suggesting likely poor field control of this strain, but was >17-fold higher than the LC99 for P. australiana (Fig. 6). Control mortality was similar for the field and reference strains, averaging 3.1 to 4.4% across all bioassays.

Fig. 6
figure 6

Insecticide bioassay dose-response curves for P. australiana (dotted line) and P. xylostella (dashed line) field strains collected from Angle Vale and Urrbrae, South Australia, and a susceptible P. xylostella (S) reference strain (solid line), exposed to four commercial insecticides: Dominex, Coragen, Proclaim and Success Neo. Points are the mean observed response across 4 bioassay replicates of 10 larvae each and lines are the fitted log-logistic response curves with 95% confidence intervals shown in grey shading. The vertical red line represents the approximate commercial field dose for each insecticide and vertical black lines represent the estimated LC99 for the corresponding Plutella strain

Full size image

Discussion

Cryptic species arise when divergence does not lead to morphological change [79]. The recent discovery of a cryptic ally, P. australiana, to the diamondback moth, P. xylostella, was unexpected given the breadth of previous molecular studies of this insect. Several factors may have contributed to this discovery, including the use of light traps for specimen collection, rather than limiting sampling to Brassica vegetable farms. Landry and Hebert [16] also isolated DNA from legs, keeping most of each specimen intact and providing a morphological reference for examining unexpected genotypes. It is also possible that P. australiana was previously overlooked from nuclear DNA studies due to biases in amplification of divergent alleles. Here, we sought to determine whether P. australiana is an agricultural pest, and to understand its ecological and genetic differences from P. xylostella.

Extensive larval sampling from wild and cultivated brassicaceous plants revealed that P. australiana co-occurs widely with P. xylostella throughout southern Australia and utilizes some of the same host plants. The relative abundance of P. australiana was on average 9-fold lower than P. xylostella. We observed higher proportions of P. australiana in larval collections from the eastern state of New South Wales, similar to the light trap samples from Landry and Hebert [16], possibly reflecting habitat suitability. Although we did not detect P. australiana in limited sampling from the island state of Tasmania, the presence of brassicas in the region and evidence from light traps that wind currents can transport Plutella moths across Bass Strait (Lionel Hill, Pers. Comm.) suggest it is likely to occur there.

Our study confirms that the host range of P. australiana includes canola crops and wild brassicaceous species. In laboratory rearing, P. australiana completed development on sand rocket, D. tenuifolia, and canola, B. napus, and was also collected from several other wild species, though without rearing to confirm host status. Our sampling focused on relatively few introduced brassicaceous species common in agricultural areas, yet the Australian Brassicales is represented by 11 plant families [80], including several non-Brassicaceae on which P. xylostella and its allies have been documented feeding, such as Capparaceae [24], Cleomaceae [16] and Tropaeolaceae [23]. The Australian Brassicaceae has records for 61 genera and 205 species [80], including many introduced species but also a diversity of native genera, such as Lepidium, Blennodium, and Arabidella, that occur over vast areas of Australia. Wider sampling of native Brassicales may identify other suitable hosts for P. australiana.

Plutella australiana larvae were not identified among samples from sixteen commercial Brassica vegetable crops despite the high suitability of these crops for P. xylostella [81], however eggs were collected from kale. It is possible that extreme insecticide susceptibility prevents juvenile P. australiana populations from establishing, as commercial Brassica vegetable crops are typically sprayed multiple times per crop cycle [22]. Our data show that P. australiana is far more susceptible than P. xylostella to four commonly used insecticides. At commercial application rates, these insecticides are likely to provide high-level control of P. australiana in Australian Brassica crops, but some products may provide marginal or poor control against P. xylostella due to insecticide resistance (Fig. 6) [22, 25]. Alternatively, some vegetable cultivars may not be attractive for oviposition or suitable for larval survival in P. australiana. We noted that P. australiana cultures provided with cabbage seedlings failed to produce viable eggs over seven days, but after replacing cabbage with Diplotaxis seedlings, egg-laying then occurred within 24 h. Exposure to host plants stimulates reproductive behaviour in P. xylostella [82], but olfactory cues for host recognition or oviposition [83–85] may differ between these Plutella species. Host preference and performance studies are required to test these hypotheses.

Insecticide bioassays have been conducted routinely on Australian P. xylostella to monitor levels of insecticide resistance in field populations [22, 25]. This method appears unlikely to be affected by the presence of P. australiana under typical conditions, as a period of laboratory rearing is usually necessary to multiply individuals prior to screening. In our experience, laboratory rearing of the two Plutella species on cabbage plants selects against P. australiana individuals when competing with P. xylostella in cages, causing the complete loss of P. australiana within a few generations. The reasons for this are unknown but may include differences in host preference or development rate, or direct competition.

Crossing experiments revealed that hybridization can occur between P. australiana and P. xylostella under controlled conditions and is most likely to occur in crosses involving Wolbachia-infected P. australiana females. Hybridization occurs in around 10% of animal species, particularly in captivity [86], but asymmetric reproductive isolation is commonly observed in reciprocal crosses between taxa [87]. In our experiments, a strong male bias in the offspring of interspecific crosses and failure to back-cross hybrid females both follow Haldane’s rule [88], which predicts greater hybrid inviability or sterility in the heterogametic sex (female, in Lepidoptera). This pattern can arise from epistatic interactions between sex-linked and/or autosomal genes that result in genetic incompatibilities [89, 90]. Although the back-crosses with F1 hybrid females were sterile, the back-crosses with hybrid males (to both species) were viable, which could enable the transfer of genes between hybrid and/or parental species. However, it is unclear whether hybridization occurs in the wild.

Although P. australiana and P. xylostella show deep divergence (8.6%) in mtDNA [16], the sole use of mtDNA can be unreliable for inference of evolutionary history and should be corroborated using evidence from nuclear markers [34]. Our analysis revealed striking differences in nuclear diversity across the genome between co-existing populations of each Plutella species collected at the same locations and times, and from the same host plant species. Plutella xylostella populations from Australia and New Zealand have low levels of genetic diversity compared with populations from other continents, thought to reflect the recent introduction of this species from a small founding population [14, 17, 77]. Consistent with this view, we found a remarkable 1.5-fold reduction in heterozygosity across >300,000 sites in P. xylostella compared with sympatric P. australiana populations. However, both species showed limited mtDNA diversity with a single predominant haplotype. While outgroups from other continents were not available, comparative analysis of these closely-related Australian Plutella species suggested that patterns of mitochondrial and nuclear diversity are concordant in P. xylostella and consistent with a demographic bottleneck [17, 18], but discordant in P. australiana.

Sequence variation in mitochondrial DNA can be strongly influenced by Wolbachia infection [41]. Extensive Wolbachia screening showed that each Plutella species was infected with a different strain at contrasting frequencies, and fit a ‘most-or-few’ pattern whereby species infection rates are often very low (<10%) or very high (>90%) [91]. Infection incidence in P. xylostella was lower in Australia (1%) than previously reported across global samples (5%) [18]. Our finding of a single supergroup A strain showing 100% sequence similarity to a strain reported in P. xylostella from Malaysia, plutWA1 [18], provides some support of an Asian origin for Australian P. xylostella [17], though does not preclude this strain also occurring elsewhere.

Fixation of infection in P. australiana suggests that Wolbachia manipulates the reproductive biology of this species. We found no evidence of sex-ratio distortion, which has been associated with a Wolbachia strain, plutWB1, in P. xylostella [18]. High infection can be driven by cytoplasmic incompatibility (CI) [40]. The high frequency (87%) of a single mtDNA haplotype among P. australiana individuals implies that the spread of Wolbachia infection has driven a selective sweep of co-inherited mtDNA through the population, causing a loss of mtDNA diversity [41]. High nuclear diversity (relative to sympatric P. xylostella) supports this hypothesis, because a demographic bottleneck should reduce diversity across the entire genome [34].

Plutella australiana and P. xylostella have co-existed in Australia for at least 125 years (\(\geqslant \)1300 generations), yet have strongly divergent mitochondrial and nuclear genomes, Wolbachia infections and insecticide susceptibility phenotypes. Our observations during laboratory rearing and crossing experiments also suggested that interspecific differences in host plant use may exist. What explains such strong divergence between the two Plutella species, given sympatry and the capacity to hybridize? Endemism of P. australiana [16] implies an ancient evolutionary history in Australia, and our data provide support for existing views that Australian P. xylostella were recently introduced from a small ancestral source population, possibly from Asia [17, 18, 77]. Therefore, the two Plutella species may have diverged in allopatry and recently come into secondary contact. Maintenance of divergence suggests strong continuing reproductive isolation, which can evolve as a side-effect of allopatric divergence [44]. All 99 individuals that were RAD sequenced showed concordance in nuclear multilocus genotypes and mtDNA genotypes identified through PCR-RFLP regardless of geographic location, as shown by STRUCTURE analysis. Cryptic species in sympatry provides strong evidence of limited genetic exchange [79]. A small degree of genotypic admixture evident for a few individuals in the STRUCTURE plots might be explained by ancestral polymorphism or introgressive hybridization [28], or alternatively, could be an artefact if our dataset is not representative of the entire genetic background [33]. The level of hybridization that may be occurring between these species is unknown. Isolation may not be uniform across the genome [92, 93], and scans of larger genomic regions may be required to identify introgression and detect hybrids.

The factors leading to reproductive isolation between the two Plutella species in nature are unknown but could include a range of pre- or post-mating isolation mechanisms, such as assortive mating or hybrid fitness costs. Behavioural mating choices are often the main isolating factor in sympatric animals [86]. Does Wolbachia cause a reproductive barrier? The contrast in infection status creates the potential for cytoplasmic incompatibility between species [94]. Interspecific crosses showed a pattern of asymmetric isolation consistent with the expected effects of unidirectional CI, where 21% crosses involving infected P. australiana females produced viable offspring, while the reciprocal CI-cross direction (uninfected P. xylostella females crossed with infected P. australiana males) was nearly sterile. However, this pattern was not continued in the F1 generation: infected hybrid males (derived from the P. australiana maternal line) produced offspring at comparable rates when back-crossed to either uninfected P. xylostella or infected P. australiana female parents. The role of Wolbachia-induced postzygotic isolation between the two Plutella species requires further study, though our results suggest it could be more important in the F0 generation. Wolbachia can contribute to post-zygotic genetic isolation after speciation by complementing hybrid incompatibilities [94, 95]. Symbiont infections could also influence mating behaviour and contribute to pre-mating isolation [96].

Conclusions

The discovery of cryptic pest species introduces complexities for their management and also exciting opportunities for understanding ecological traits. We found strong genomic and phenotypic divergence in two cryptic mitochondrial Plutella lineages co-existing in nature, supporting their status as distinct species [16] despite the capacity to hybridize. Reproductive isolation is likely to have evolved during allopatric speciation, and genome-wide sequence data suggest it has been maintained following secondary contact. Variation in Wolbachia infections might be one factor reinforcing reproductive barriers.

Plutella australiana co-occurs with P. xylostella throughout agricultural regions of southern Australia, but made up only 10% of Plutella juveniles collected from cultivated and wild brassicaceous plants. A lack of population structure across neutral SNP markers suggests that P. australiana populations are linked by high levels of gene flow, and also that P. australiana is a highly mobile species, which is supported by light trap collections [16] and seasonal colonization of canola crops. Future molecular analysis of Australian Plutella should include a species identification step using a molecular diagnostic assay. For ecological studies, it may be possible to perform molecular species identification to confidently distinguish a representative sub-sample of individuals or pooled samples. Our study has shown that while P. australiana can attack canola crops, there is no evidence of pest status in commercial Brassica vegetables crops, and bioassays suggested that field populations should be easily controlled with insecticides. Though P. australiana is a potential pest of some Australian Brassica crops, it is of secondary importance to the diamondback moth, P. xylostella.