Molecular detection of small hive beetle Aethina tumida Murray (Coleoptera: Nitidulidae): DNA barcoding and development of a real-time PCR assay

Small hive beetle (SHB), Aethina tumida can feed on honey, pollen and brood in honey bee colonies. It was endemic to Africa, but since 1996 has been detected in a number of countries worldwide, including Australia, Brazil, Canada, Italy, Mexico, South Korea, Philippines and the USA where it has had economic effects on local apiculture. To improve SHB identification, we obtained the first reference sequences from the DNA barcoding 5′ COI gene region for SHB and some species of the family Nitidulidae associated with beehives. Phylogenetic analysis of SHB COI sequences (3′ COI) revealed two divergent lineages, with those from Australia and USA being genetically different from the recent detection in Italy. Many countries, including New Zealand, are currently free from SHB, and require a rapid detection method for biosecurity. Here we present the development and validation of a real-time PCR assay for detection of SHB. The assay showed high specificity and sensitivity for detecting SHB, with no cross-reaction observed with closely related species, such as A. concolor. The real-time PCR is sensitive, detecting the target sequences up to 100 copies/µL. This assay should prove a useful biosecurity tool for rapid detection of SHB worldwide.


Species (Family)
Reported Pest Status

BOLD barcode* (public) Present in Australia Present in NZ
Aethina tumida Murray (Nitidulidae) Breeds in hives, damaging beehive pest 16 Table 2) was obtained from beehive via New Zealand general surveillance program. Another A. concolor specimen (T17_00549 in Table 2) was collected from and Calystegia sepium flowers. The survey of Buchholz et al. 65 showed that A. concolor individuals were observed on blossoms which were regularly visited by honey bees.  Table 2. DNA sequences of 5′fragment of COI or the Folmer region obtained in this study showed over 99% identities between the SHB samples from Australia and USA, but they shared around 94-96% identities with the SHB from Italy. Those SHB sequences were distantly related (>17% differences) to those of A. concolor (5 sequences: T16_01444, T17_00549, AQ5, GU217509 & GU217510) and A. suturalis (9 sequences: KJ480785-KJ480793), with around 81-83% identity. The sequences of those SHB is also quite divergent from the other nitudulid beetles collected from bee surveillance survey (Table 2). Of the existing 79 SHB COI sequences (for the 3′ fragments region) deposited in GenBank, however, there was little overlap with the Folmer DNA barcoding region (Fig. 1).

Phylogenetic relationship and morphological comparison of SHB from different locations.
Nearly full COI gene sequences for the SHB intercepted in New Zealand (AQ3 and AQ4), obtained from Italy (T17_01853A and T17_01853B) and A. concolor collected from Auckland were also obtained ( Table 2). Phylogenetic analysis of AQ3, AQ4, T17_01853A, T17_01853B and the 79 SHB (Accession numbers shown on Fig. 2) downloaded from GenBank was constructed by using A. concolor as outgroup. Due to the short overlap of the sequences, the specimens with only 5′COI sequences obtained in this study were not included (Fig. 2). The phylogenetic tree showed that SHB formed two clusters (Fig. 2, Group I and II), the SHB sequences obtained from Australia, USA and South Africa formed a separate clade (Group I) from the recent detection of Italian SHB sequences (Group II). The one sequence (HM056044) of SHB from Cameroon, Central Africa also belonged to the same clade as the Italian samples (Group II). Between the two clades, there are around 4-6% differences in the COI sequences (~900 bp). Despite the genetic differences outlined here between SHB Groups I and II (Fig. 2), the recent Italian SHB was morphologically identified as A. tumida 11 and no obvious morphological differences were observed (Fig. 3).
Following further examination in this study of SHB specimens from Italy and Australia, most characteristics were found to match with the description of A. tumida as described in Lee et al. 22 . Minor morphological differences were observed between the SHB specimens examined from Italy and Victoria (Australia), including difference in body colouration. Body colour was reddish/dark brown (particularly ventrally) in the Italian specimens ( Fig. 3a,b) but black/dark brown in the Australian specimens (Fig. 3f,g). The shape of the prosternal process also differed between specimens, with lateral margins appearing parallel sided and more slender in the Italian specimens ( Fig. 3c) but slightly more concave laterally in some Victorian (Australian) specimens (Fig. 3h). Finally, the ovipositor was narrower in the Italian specimens (Fig. 3e) compared with the Australian specimens (Fig. 3j). Some of the differences are subtle and may represent natural morphological variation, or be consequences of the size or life stage of individuals (eg. teneral specimens are often paler in colour) or be due to preservation conditions. Real-time PCR design. A total of 2, 329 COI sequences of SHB and closely related Nitidulidae species were downloaded from the GenBank and BOLD databases, the alignment showed that over 1,000 sequences included mainly the 3′ fragment of COI gene region. Amongst the sequences, 79 sequences were from SHB (3′ COI region), thus this region was used for real-time PCR assay design. The COI sequences of SHB were highly AT rich, with ~70% AT content. Due to the high AT richness, a BHQplus probe was designed, in which propyne modified bases (pdU and pdC) were incorporated and higher Tm was obtained using shorter bases 50 . The primers/probe have high matches to all the SHB COI sequence online and those we obtained in this study. The details about the real-time PCR primers/probe were: forward primer (Atum-3F: 5′-CCCATTTCCATTATGTWYTATCTATAGG -3′), reverse primer (Atum-3R: 5′-CTATTTAAAGTYAATCCTGTAATTAATGG-3′) and probe (Atum-3P: 5′-pdTApdTpdTpdTGCpdTApdTpdTApdTAGpdCpdCGGApdTpdTpdTGpdT-[BHQplus]-3′). The position in Real-time PCR optimization. The real-time PCR assay was initially tested with different annealing/extension temperatures using DNA extracted from the Australian (adult) and the USA (larva) specimens. The Cq of ~24 and ~17 cycles for the two DNA extracts were obtained, respectively from the temperatures of 56° to 60   The real-time assay was demonstrated to perform consistently well using different PCR mastermixes, with similar Cq values of around 17-18 cycles obtained for the DNA extracted from the USA specimen although slight differences in intensity were observed with the addition of MgCl 2 . There was negligible difference in the Cq values (~17-18 cycles for the USA specimen) when the assay was run with different concentrations of primers (125-300 nM) or probe (125-250 nM). Therefore the real-time PCR condition for SHB were chosen as: primer and probe concentration of 250 nM and 200 nM, respectively.

Real-time PCR sensitivities.
The linear dynamic range for the assay was tested on plasmids containing the COI inserts and extended from 10 7 -10 0 copies of plasmid DNA. The 95% confidence limits of the linear dynamic range were plotted in Fig. 4, with a strong correlation coefficients (r 2 = 0.993 and 0.998). The limit of detection (LOD) for the assay was estimated to be 100 copies/µL of target DNA. A template concentration of 10 copies/µL was sporadically detected in the assay with an average Cq value around 35 cycles. The calibration curves shown in Fig. 4 was able to detect 100% of the samples (and replicates) at the 100 copies/µL. This assay could also reliably detect up to 1:10 4 and 1:10 5 dilution of the DNA extracted from an adult leg (VAITC5360) and a small pieces of larva (AQ4) of the SHB (data not shown) specimen, respectively.
Real-time PCR specificity and robustness. All small hive beetle samples were successfully amplified by the real-time PCR assay, while no amplifications were observed for the non-target species ( Table 2). The assay was able to accurately identify small hive beetle samples from Australia, Italy and USA. No cross-reaction was observed with the DNA extracted from honey bee samples, varroa mites (Varroa destructor), and Acarapis mites (Acarapis woodi, A. dorsalis and A. externus).

Figure 2.
Bayesian phylogenetic analysis of 3′ fragment of COI sequences from 89 SHB using A. concolor as outgroup, with posterior probabilities of branches at each node. Bold letters for the taxa indicate sequences obtained in the current study (further details in Table 2). The accession number for each sequence is listed in the tree, except that the clades with similar sequences were collapsed for clear view of the tree. All sixteen samples tested in the blind panel test were correctly identified as either positive or negative for SHB ( Table 3). All the SHB samples were tested positive with Cq values less than 30. No amplification was observed in all the non-target species (Table 3).
Comparison with the SHB-specific PCR assay. DNA samples which tested positive in the real-time PCR assay developed in this study were used as templates in the SHB-specific PCR described in Idrissou et al. 51 . The expected size of ~193 bp PCR products were detected in the DNA extractions from Australia and the USA specimens, but very weak amplification was observed in the DNA samples from Italian specimens under the same PCR compositions and cycling conditions with and without additional MgSO 4 added. Further tests showed that detectable PCR bands were observed in all the samples tested when the PCR cycles increased to 40 or the annealing temperatures reduced to 52 °C (data not shown).
The sensitivity of the SHB-specific PCR 51 was tested against DNA extracted from the Australian (VAITC5360, adult) and the USA (AQ4, larva) specimens and compared to that of our real-time PCR assay. The SHB-specific PCR successful amplified DNA samples up to 1:10 2 and 1:10 3 dilutions whereas our real-time PCR assay successfully amplified the DNA in samples with up to 1:10 4 and 1:10 5 dilutions (see section on Real-time PCR sensitivities) from the Australian and the USA specimens, respectively. Therefore the real-time PCR developed in this study demonstrated over 100 times the sensitivity of the PCR assay by Idrissou et al. 51 . . Sensitivity analysis of real-time PCR assay for the identification of small hive beetle using plasmid DNA containing COI inserts. Plasmid containing COI insert of SHB were series diluted to create calibration curves for sensitivity calculations. The standard curve built from Cq values against the log copy number (range = 10 7 -10 0 copies) of COI insert (n = 3). The 95% confidence intervals of the slopes were plotted with a blue line for VAITC5359 (Australia sample) and a red line for AQ4 (USA sample). The R 2 = 0.993 for VAITC5359 and 0.998 for AQ4 were obtained for the assay.

Discussion
This study generated the first DNA barcoding sequences (5′ COI) of SHB (Table 2), which will be useful not only for SHB identification but also for the future phylogenetic studies of other Aethina species with the sequences in the same region. DNA barcoding of the 5′ fragment of COI region of the SHB sample from Australia and USA were highly similar with more than 99% identity (Table 2), but they are different from sequences of the Italy SHB samples (Table 2) by around 4-6% in the Folmer's region. Phylogenetic analyses of DNA sequences of the 3′ fragment of the COI gene (~900 bp) revealed divergences among SHB (Fig. 2). COI sequences were highly similar for the SHB detected in Australia, Canada, Madagascar, South Africa, USA and Zimbabwe 20,47 , but different from those detected in Italy and one from Cameroon, Central Africa, HM056044 (Fig. 2 Group II). Similarly, Lounsberry et al. 20 pointed out that the SHB COI sequence (HM056044) was 4.6% divergent from the rest samples they analysed. Our study demonstrated that this single Central African COI sequences is similar to the recent detection of SHB in Italy and formed in one clade (Fig. 2). In the Italy SHB samples, there are mainly three sub-clades formed, the central African sequence (HM056044) was close to the subclade B and C than those of sub-clade A. The COI sequences (~900 bp) for the sequences in the sub-clades B and C are less than 2% difference while they are different from the sub-clade A by around 3.3%. Therefore it suggested that the Italy detection is more related to the SHB from Cameroon, not the SHB from the other locations studied 20,47 . This is a novel finding that two genetically divergent groups are present within SHB populations. Questions surrounding the divergence of these two groups requires further investigation. An intensive morphological study of SHB from Italy, Australia and from around the world, including the original African source populations, would help to better define morphological variability within this species or, given the large sequence difference observed between the two SHB groups (Fig. 2), potentially reveal two closely related species. However, to date the two groups have been identified morphologically as SHB and both appear to cause serious damage to beehives 11 .
For a reliable assay, the primers/probe need to match COI variation from all the SHB specimens, but possess sufficient mismatches to the closely related species. Our assay was designed by aligning all the COI sequences from Nitidulidae, the sequences analysis showed that the majorities of COI sequences for SHB were toward to the 3′ fragment of the COI gene. To be able to capture all A. tumida sequences, our real-time PCR protocol was designed by targeting 3′ fragment of the COI gene (including SHB Groups I and II, Fig. 2). The primers were designed with degeneracy in order to capture all the haplotypes of the A. tumida sequences. In silico tests showed that the primers (Atum_3F and Atum_3R) and probe (Atum_3P) have 5, 5 and 2 mismatches with A. concolor, there was no cross-reaction observed in the real-time PCR assay either ( Table 2).
A real-time PCR protocol was developed in 2007, in which the probe, SHB245T is 39 bp in length 48 . The design for the real-time PCR was more likely due to the high AT rich sequences, thus longer probe was needed to be provide a higher Tm. As probe length increases, the distance between the fluorophore reporter and quencher at the probe end increases, resulting in reduced quenching and thus, poor signal to background discrimination will be observed. Therefore, the probe longer than 30 bases are unlikely to have efficient quenching as an end-labelled probe. Longer probes are also more likely to bind non-specifically, and therefore exhibit reduced specificity. Prior to develop our real-time PCR assay, attempts were made to synthesise the primers/probe from Ward et al. 48 , but most companies suggested that an internally quenched probe is needed for probe over 30 bases. In addition, there were only 15 COI sequences available when this assay was designed 48 , thus it was not necessary to order the primers/probe by Ward et al. 48 and validate the real-time PCR assay. On this other hand, the real-time PCR assay designed in this study based on 79 SHB sequences, containing new divergent haplotypes of SHB sequences (Fig. 2). In in silico analysis, mismatches in the binding sites for primers and probe in Ward et al. 48 were also observed in 31 (23 from GenBank and 8 from this study) sequences (Fig. 5). Therefore it was necessary to develop a new real-time PCR assay for the rapid detection of SHB.
The real-time PCR assay developed in this study covers the known COI diversity (i.e. including Groups I and II, Fig. 2) of haplotypes of SHB available (89 sequences) using the BHQplus probe technique (http://eu.biosearchtech.com/products/bhqplus-probes). The incorporation propyne modified bases of T and C 50 could achieve a higher Tm with a shorted sequence in the AT rich A. tumida COI sequence. Using this technique, a BHQplus probe of 25 bases was able to reach a sufficiently high Tm and allow the assay to effectively amplify the SHB samples and increase its sensitivity. The real-time PCR assay has also demonstrated its sensitivity by detecting the SHB samples in less than 30 cycles of Cq values detested in this study.
Comparison with the recent published PCR assay developed by Idrissou et al. 51 indicated that the real-time PCR assay developed in this study is more robustness in amplifying different haplotypes of SHB samples. When testing with the SHB-specific PCR assay, successful amplification using the Australian and the USA specimens, but very weak and not obvious amplifications were observed using the DNA extracted from the Italy specimens. Further in silico testing showed that the two primers (AT420F and AT623R) designed by Idrissou et al. 51 were based on one sequence obtained from A. tumida isolate BRL-Maryland unplaced whole genome genomic scaffold (NW_017855158.1), which represents one of the COI haplotypes. As there are several different haplotypes detected in the SHB specimens (Fig. 2), there are mismatches observed in the two primers: 2-3 mismatches in the AT630R primer for the available SHB COI sequence (SHB Group I and II in Fig. 2); for the AT420F, however, there are 2 mismatches in the 5′-end for the Group I while 1 in the 5′-end, 1-2 in the 3′-end mismatches for the Group II SHB specimens (Fig. 2). Therefore the SHB-specific assay 51 might provide a false negative results, for example, no detectable PCR products were observed in the PCR reactions using the DNA from some of the Italy specimens tested in this study under the condition suggested in Idrissou et al. 51 . The reasons might due to (1) the mismatches in the 3′-end (critical position) of the AT430F for the Italy SHB reduced the amplification efficiently; (2) different PCR mastermixes were used in the reaction, affecting the assay robustness. Obvious detectable PCR products were observed when increasing the PCR cycles or reducing the annealing temperatures further ScIEntIfIc REPORtS | (2018) 8:9623 | DOI:10.1038/s41598-018-27603-x indicated that the amplification efficiency are low. If this real-time PCR will be applied for diagnostic, further validation are needed.
In comparison to the SHB-specific PCR assay, the real-time assay demonstrated its sensitive by detecting 100 times less concentrated DNA samples than the PCR assay 51 . Furthermore, it is a closed-tube system which no post-PCR manipulations was required, thus reducing the risk of cross contamination between samples. Therefore the real-time PCR assay developed in this study will enable the quick, reliable and highly sensitive detection of SHB smaples.
Application of the real-time assay in the New Zealand quarantine framework provides an alternative to morphological identification, especially useful when immature stages of the beetles and the damaged specimens are intercepted. The real-time PCR assay targeting SHB developed in this study is conformant with the MIQE guidelines for qualitative assays 49 . High specificity and sensitivity are much desired in border diagnostics, the assay showed high efficiency and sensitivity in detecting the target species. In the specificity tests, all the SHB samples from Australia, Italy and USA were correctly identified, no cross-reactions with the closely related species were observed. The assay has tested on A. concolor from New Zealand, but no other Aethina species were available to validate the real-time PCR assay. However, in silico analysis of the Aethnia species with the COI sequences showed that SHB is over 17% different compared with A. suturalis (KJ480791). Therefore, the possibility of cross-reaction with A. suturalis are extremely low. The beetle samples associated with bee hives were also tested in the real-time PCR assay, no amplification observed (Table 2), further indicating the specificity of the real-time PCR assay. This real-time PCR assay could assist in rapid identification of SHB, including the immature stages of SHB which are difficult to be identified by morphological features 52 .
The real-time PCR assay developed here could detect SHB from Australia, Italy and USA of different life stages, no cross reactions were observed with honeybees, bee mites (Acarapis woodi, A. externus and A. dorsalis) and varroa mites. The assay was also tested against beetles of genera Brachypeplus, Carpophilus, Dactylosternum, Dermestes, Epuraea and Urophorus (Table 2), and no-cross reactions were observed. This assay tested against the nitidulid beetle samples intercepted and related species detected in beehives at New Zealand borders and post border, no amplification was observed, therefore it could be applied to the routine diagnostic for SHB at the New Zealand borders.
There are several exotic beetle species which are associated with bee hives ( Table 1) that have not been tested in our real-time PCR assay due to the difficulties in obtaining those specimens. However, in silico analysis showed that these species possess a divergent COI region to that of the SHB. For example, there are more than 15% differences between SHB and the species of Carpophilus lugubris, Cychramus luteus and Glischrochilus fasciatus when comparing their COI sequences. Cryptophagus hexagonalis (Coloptera: Crytophagidae), the only one not in the family Nitidulidae (Table 1), was reported in the beehive 53 , no DNA sequences available for this species, but there are sequences for this genus. Sequence comparison of SHB with Cryptophagus spp. showed that the similarities with SHB are 77-80%, thus indicating Cryptophagus spp. are more distant from SHB. Therefore we are confident that this real-time PCR assay for SHB is not likely to be able to amplify those beetle species. In addition, no other Aethina species could be obtained for testing the real-time PCR assay besides A. concolor. There are limited studies on the genus, DNA barcoding reference sequences (on BOLD) are currently available only for A. timuida, A. concolor and A. suturalis. The real-time PCR assay was designed based on the current available sequences, tested and validated with the species we could obtain, therefore further tests and analysis will be conducted if more related species obtained and DNA sequences available.
Here, we report the DNA barcoding identification of SHB, A. tumida and other related beetle specimens, including A. concolor, Brachypeplus sp., Epuraea signata, Epuraea sp. and Urophorus humeralis (BOLD project ANZN), which enrich the reference sequences for the nitidulid beetles in BOLD and GenBank databases. The real-time PCR assay developed for SHB in this study is suitable for routine application by diagnostic and research agencies, for facilitating exports and imports. It will assist in border security agencies worldwide to limit and monitor the spread of this pest. This assay provides a rapid, accurate and specific alternative to morphology, or more time intensive DNA barcoding methods, for identification for SHB. This assay has been fully optimised for

Material and Methods
Sampling. Small hive beetle samples used in this study were collected from sticky mats used for bee hive pest surveillance in Australia, intercepted at the New Zealand borders and obtained from Italy (Table 2). Additional nitidulid beetles for DNA extraction include pinned specimens in the Victorian Agricultural Insect Collection (VAIC, Melbourne, Australia), frozen and ethanol (75% and 96%) preserved specimens from the Plant Health and Environment Laboratory collection (PHEL, Auckland, New Zealand) ( Table 2). All the specimens were initially identified morphologically and some were also confirmed through DNA barcoding, using COI sequences (see section 4.2 below).
Four adult specimens of SHB collected from Italy (ANZN026-17 to ANZN029-17) and four specimens from Victoria, Australia (VAIC 63240, 63244, 74457 and 78224) were closely examined and compared to the morphological description and keys in Lee et al. 22 . Three females (one Italian and two Australian) were dissected for examination of the genitalia. Specimens were first soaked in 10% Potassium Hydroxide (KOH) overnight before removing the ovipositor from the abdomen using fine micro-tools.
DNA extraction, PCR amplification and sequencing. Total DNA was extracted using the DNeasy Blood and Tissue kit (Qiagen, Valencia, CA, USA) as per the manufacturer's instructions. A single leg of adult or segment of larva was used for each extraction and physical disruption was performed by micro-pestles, with the final DNA eluted in 100 µL of AE buffer (Qiagen, Valencia, CA, USA). For DNA barcoding of the beetle species, LCO1490 and HCO2198 primers 46 were used to amplify the PCR products of 5′fragment of COI gene from most specimens. The amplification of A. concolor and some of the specimens for 5′ fragment of COI gene using LCO1490 and HCO2198 was not successful, thus several pairs of primers were tested, including mtD-7.2 F/mtD-9.2R 54 , CI-J-1751/CI-N-2191 and CI J-1718F/COIREVA 55 . Additionally, older pinned VAIC specimens were amplified using two new "mini-barcode" internal primers, NitidulidaeCOI-F 5′-AAAGAGGAGCWGGMACWGG-3′ and NitidulidaeCOI-R 5′-GCGATATTGGATGAVAGDGG-3′. These primers were designed using Primer 3 56 from COI sequences of Carpophilus davidsoni Dobson, Urophorous humeralis, Fabricius, Brachypeplus sp., Aethina concolor (Macleay, 1872), obtained from Brown et al. 57 , for amplification of the 5′ COI region in two approximately 350-400 bp (55 bp overlapping) amplicons (Fig. 1). To obtain the 3′ fragment of the COI gene sequences, selected samples were used for PCR with primers: A1859 and AT3014 58 (Fig. 1).
For gel stained with SYBR ® safe (Life Technologies ™ ), and visualised using a Gel Doc Software system (BioRad, Hercules, CA, USA). Amplified products were sequenced bi-directionally using the amplification primers for 5′ fragment (as above) and AT2380S and AT2519A 47 for the 3′ fragment of COI gene by EcoGene ® (Auckland, New Zealand) or Macrogen (Seoul, South Korea). The obtained DNA sequences were edited and aligned using Geneious Pro 9.1.5 (http://www.geneious.com, Kearse et al. 59 ) and BLAST searched against the GenBank database 60 or BOLD database 45 to confirm morphological identifications.
The COI sequence data obtained in this study were deposited in BOLD and GenBank databases ( Table 2). BOLD accessions include full collection details along with diagnostic images of VAIC specimens obtained using a Leica M205C microscope and DFC450 camera, those of PHEL photographed using a Zeiss AxioCam HRc camera attached to a Zeiss interference-phase contrast microscope (Axio Imager 1), and can be found in the BOLD project Australian and New Zealand Nitidulidae -ANZN.
Phylogenetic relationship of SHB samples. Sequence data of COI gene from 79 SHB samples were downloaded from the GenBank database. Ten nearly full length (AQ3, AQ4, T17_01583A_1 to T17_01583A_4 & T17_01583B_1 to T17_01583B_4) and nine SHB 5′ fragment of COI sequences were obtained in this study ( Table 2). Multiple sequence alignment was performed using the Geneious aligner and re-aligned with Clustal W in Geneious version 9.1.5 (http://www.geneious.com, Kearse et al. 59 ). The alignment of SHB samples (89 samples) with mainly 3′ fragment of COI gene were used, but due to the small region of sequence overlap, some of the Australian samples with only the 5′ fragment of sequences were not used in phylogenetic analyses. Phylogenetic trees were constructed using Neighbour-Joining (NJ), Maximum-Likelihood (PHYML) and Bayesian (MrBayes) methods in Geneious Pro 9.1.5 under the default settings 61 . Multiple runs were performed using the model GTR and rate variation invgamma. For Bayesian tree estimation, the resulting perimeter files were inspected for chain convergence and mixing in Tracer 1.4 62 . The trees were rooted using A. concolor COI sequences (T6_01444 & T17_00549) obtained in this study as outgroup.
Real-time assay design. All currently available COI sequences of Nitidulidae, including A. tumida and closely related species were downloaded from GenBank and BOLD databases (up to June 2016). The sequences were aligned in Geneious version 9.1.5 (http://www.geneious.com, Kearse et al. 59 ). Suitable regions for probe and primer design were identified manually. The probe was designed using Oligo Evaluator from Biosearch (https:// ScIEntIfIc REPORtS | (2018) 8:9623 | DOI:10.1038/s41598-018-27603-x rtd.biosearchtech.com/design/OligoEvaluator.aspx), and primers were designed manually. Melting temperature (Tm) and secondary structures of primers and probe sequences were evaluated using OligoAnalyzer 3.1 from IDT Integrated DNA technologies (https://sg.idtdna.com/calc/analyzer). The primers and probe were BLAST searched to test their specificity in GenBank and checked their matches to SHB sequences in Geneious 9.1.5 (http://www.geneious.com, Kearse et al. 59 ).

Development and validation of the real-time PCR assay.
Optimization. The real-time PCR protocol was tested on a CFX96 ™ Touch Real-time platform (BioRad). Gradients of temperatures (56-66 °C), primer concentrations (200, 250 and 300 nM), probe gradients (125, 200 and 250 nM), with and without the additional MgCl 2 (1.5 mM) were used to optimise the PCR conditions. To select an appropriate real-time mastermix for the assay, preliminary comparisons were trialled with commercially available mastermixes: PerfeCTa ® qPCR ToughMix ® (Quanta Bioscience), SsoAdvanced ™ Probes Supermix (BioRad). The quality of each DNA sample was also tested with conventional PCR using COI (LCO1490 and HCO2198 primers) 46 or 18S ribosomal RNA (rRNA) gene internal control real-time PCR (Applied Biosystems, CA, USA).
Sensitivity. To evaluate the analytical sensitivity of the developed assay, a 1,100 bp template COI PCR product, amplified as described in section 4.2 was used to prepare plasmid standards of known copy number. The amplicon was cloned using the TOPO ® TA vector cloning kit (Invitrogen, Carlsbad, CA, USA) as per the manufacturer's instructions. Cloning was performed from the PCR products amplified from DNA extracted of two small hive beetles, one from USA and the other one from Australia. The two clones containing the correct insert were selected for preparing standards. Plasmid DNA were extracted using the Wizard ® Plus SV Miniprep (Promega, Madison, WI, USA). The plasmid DNA was quantified using a µDrop plate in MultiSkan GO DNA quantification system (Thermo Scientific, USA) and normalised to a concentration of 10 7 copies/µL. A dilution series of the plasmid from 10 7 -10 0 copies was created using sterile TE buffer (Sigma). Analytical sensitivity of the assay was determined using the dilution series with each concentration in triplicate per reaction. Linear regression was performed between the quantitative cycle (Cq) and the log 10 of the copy number, measuring the fit as R 2 . Amplification efficiencies for individual reactions were calculated using the formula, E = 10 |1/slope| and converted to E % by (E − 1) × 100 in the R environment version 3.1.1 63 .
Specificity and robustness. Closely related (i.e. Nitidulidae) species and other species collected from bee hives were used for the specificity test of the assay ( Table 2). DNA samples extracted from honey bee (Apis cerana cerana), tracheal mites and varroa mites were also used to test whether there were cross-reactions. In addition, DNA extracted from other beetles were tested in the real-time PCR assay ( Sixteen samples were prepared for the blind panel test (Table 3), with the operator having no knowledge of the sample identities. The samples were tested in the real-time PCR protocol for small hive beetle, each samples were tested in duplicate, including the positive and water controls.
Comparison with the SHB-specific PCR assay. The reliability and sensitivity of the real-time PCR assay developed in this study was compared with the published SHB-specific conventional PCR assay by Idrissou et al. 51 . The PCR primer set, AT420F and AT623R were ordered and used for PCR with the DNA extraction from SHB of Australia, the USA and Italy, under the same PCR compositions and cycling conditions as per Idrissou et al. 51 using GoTaq master mix (Promega, Madison, WI) with and without additional 1.5 mM MgSO 4 . Further testing using the PCR compositions with additional 1.5 mM MgSO 4 were conducted under two PCR cycling conditions, (1) PCR cycles changed to 40 instead of 35 cycles, and (2) the annealing temperature decreased to 52 °C instead of 56 °C for 35 cycles. The resulted PCR products were gel electrophoresis with 2% agrose in 1 × TAE buffer, The PCR sensitivity was also compared with series dilutions of DNA extraction from SHB samples.