In sentinel plantations, non-native plants are grown in a country out of their natural distribution range (e.g. native European trees planted in China) and monitored for potentially damaging agents which may provide useful data for PRA (Fig. 1). If novel pest/pathogen-host plant combinations occur, the plants are likely to develop symptoms due to a lack of coevolution with the native organism (Parker and Gilbert 2004; Vettraino et al. 2015). The assessment of symptoms and signs, along with sampling of symptomatic tissues, and the isolation of potential pest/pathogen organisms, should be prioritized. Therefore, methods and protocols used in sentinel plantations should aim to characterize damage morphotypes, followed by isolation or collection and species level identification of the causal agent(s) (Roques et al. 2017).

It is necessary to carry out HTS analysis of a representative sample of the propagation material (e.g. seeds) intended to be used before export to the country where the sentinel planting will be located. Knowledge of the plant’s endophytic community in its native range can give a baseline for interpretation of, for example, fungi contributing to disease. In sentinel plantation trials in China, absence of controls in the propagation material did not allow confirmation of the Asiatic origin of detected OTUs (Vettraino et al. 2015).

In a sentinel nursery, native plants are grown in their natural distribution range to identify potential pests or pathogens which could be spread with the international trade of these plants (Fig. 2). In this case, the results obtained will be helpful in CRA (Kenis et al. 2018). Assuming that host-parasite co-evolution of native species might not result in obvious symptom expression, a host shift to a taxonomically similar plant species in the final location of the plant may give rise to novel host-parasite interactions. Therefore, diagnostic methods that can detect endophytic or latent pathogens must be employed (Vettraino et al. 2017) in addition to standardized diagnostics for symptomatic tissue. Thus, sampling must be oriented to both symptomatic and non-symptomatic material. In this system, the use of HTS is useful for screening of the microbial communities even in the absence of symptoms. One possible way to filter large datasets arising from HTS is to group the OTUs according to their functional guild, focusing the sampling and identification on what are grouped as pathogens or opportunistic pathogens. In the case of fungi, online applications, such as FUNGuild (http://www.stbates.org/guilds/app.php), can be used for this purpose as a base for downstream analysis (Nguyen et al. 2016).

Previous fungal studies in sentinel nurseries have not provided conclusive evidence of identified risks but rather provided information that must be analyzed to arrive at a selection of taxa for further study of whether these organisms pose a threat if introduced in a naïve habitat (Vettraino et al. 2017). Information including a collection of isolates, with molecular barcoding and, eventually, taxonomic positions and a database of OTUs resulting from HTS analysis, would greatly strengthen further analyses. Large data sets can be difficult to interpret and require appropriate databases of molecular data and plant pathogens and, certainly, the scientific literature, to make full use of their potential. A limit to data interpretation is the fact that only a small percentage of global microorganism diversity is so far present in the databases. A positive aspect is that a large number of undescribed taxa are present as sequences in molecular databases, which may provide unexpected matches with OTUs from sentinel plantings and useful information on previous detection.

During arthropod studies in sentinel nurseries (Roques et al. 2015), systematic sequencing of the “morphospecies” (defined as a group of individuals that are recognized as probably belonging to a same species based on morphological characteristics) of immature stages and adults was achieved using the “barcode” COI gene to compare potentially, newly recognized species with sequence data already present in global genetic databases. However, only a limited number of the organisms found, essentially lepidopteran larvae, could be identified to the species level. Therefore, arthropod DNA barcoding does not replace the classical approach of morphology-based species identification (Hebert and Gregory 2005; Pires and Marinoni 2010). The combination of both techniques has proven successful in numerous cases (Pires and Marinoni 2010; Okiwelu and Noutcha 2014; Kirichenko et al. 2015) and should be applied also in sentinel nurseries and plantations (Roques et al. 2015).

A sentinel arboretum (Fig. 3) comprises a broad range of both native and non-native tree species from diverse regions around the world, which can allow testing of various ecological hypotheses on biological invasions, as possible host-shifts, one of the main barriers to establishment of alien plant pests and pathogens, can be examined (Kirichenko et al. 2013; Kirichenko and Kenis 2016; Morales-Rodríguez et al. 2018). Non-native species are exposed to inoculum of native, potentially pathogenic organisms harboured by native trees species growing in the same or nearby environment. An expanded assumption here is that all native and non-native tree species planted in the same area are cross-exposed to inoculum harboured by each of the tree species in a latent native-to-native interaction.

Protocols used in sentinel arboreta should aim to characterize damage morphotypes, followed by isolation or collection, and species level identification of the organisms causing these symptoms. The non-native trees might harbour endophytic microflora since the time of their introduction into arboreta as propagation material (e.g. seeds, seedlings, cuttings). HTS can be useful in detecting non-symptomatic native host endophytic species or latent infections, contributing to characterization of the donor host microbiome and to the description of a novel host-shift event. Recently, using HTS and traditional isolation methods, several novel host-interactions between Quercus species and fungal pathogens were described in the Ataturk arboretum in Turkey by Morales-Rodríguez et al. (2018). Differing from sentinel plantations, sentinel arboreta may also allow surveys of the recruitment of insects by mature trees, and especially of particular groups, such as xylophagous pests (Roques et al. 2015).

For the three cases of sentinel plantings presented above, confirmation of pathogenicity on the host plant is an essential step for determining the causal agent of disease (Koch’s postulates). Thus, collection and isolation of the organism from symptomatic plants is crucial for establishing the causative relationship between a microbe and the disease or symptoms it produces. This procedure, however, is limited to mainly non-biotrophic organisms which can be cultured onto nutrient media. Once the causal agent is known, additional inoculation trials can be designed and carried out to evaluate its potential host range. Colonizing insects observed on sentinel plants must not be incidental, but clearly capable of completing the entire life cycle on the given host, especially when non-native plants are used in sentinel plantings. This process is difficult to ascertain because rearing possibilities on non-native plants could be limited when such plants are only growing within a sentinel plot. One way to distinguish between incidental species and potential pest could be to consider the number of successive colonization events attained over a number of years by an insect species on the same non-native tree. Roques et al. (2015) considered two groups of insects, a first one (38 species) which had shown five colonization events per year, at least on European trees in China, and a second one (7 species) that has been more frequently observed (more than 15 colonization events per year) and probably more capable of switching to European trees. Hence, repeatability and reliability in the observations are critical to drawing sound conclusions on the potential risks to plant health that are needed for PRA and CRA.

The successful detection of potentially harmful pests and pathogens in sentinel plantings relies on several conceptual, methodological and organizational factors. Among these, experimental design (i.e. how sentinel plantings are organized, e.g. how many replicates of each tree species), and sampling design (i.e. how, when and what should be sampled) are critical to making sampling as efficient and reliable as possible (Eschen et al. in prep). Similar-looking symptoms might have different causes, and for this reason, the diagnostic procedure can be challenging. Although sentinel plants might be colonized and/or damaged by a broad range of organisms, some general principles about sample collection and preservation apply to all organisms (Kirichenko and Csóka 2017; Prospero et al. 2017). Among these principles, one should consider the following:

1. As different organisms can affect a single plant, the whole plant should be carefully checked for different damage morphotypes (hereinafter referred to as damage characteristic of a certain pest or pathogen) (Tables A1, A2) and the presence of damaging organisms (Moreira et al. 2017). Samples should be taken from a range of representative symptomatic organs (Nelson and Bushe 2006).

2. Before collecting symptomatic plant material, high-resolution photographs of the whole plant, of the damaged organ(s), and, if present and visible, possible damaging agent(s) should be taken. Categorization of damage morphotypes (Tables A1, A2) might give some hints about the potential causal agents.

3. Cross-contamination from sampling instruments (e.g. secateurs, pruning saw, forceps) should be avoided; this is of particular importance when sampling for pathogens.

4. The best period for sampling varies according to the affected tissues and the suspected causal agents. If possible, at least three samplings per year (spring, summer and fall) should be conducted.

5. Samples should also be taken from apparently healthy tissue to know what healthy plant tissue looks like during normal growth, to potentially detect differences in microbial community composition between healthy and symptomatic tissues, and to study latent infection or endophytes.

6. Proper labelling of sampled material is an essential step without which biological specimens lose their scientific value (Krogmann and Holstein 2010). The minimal data recorded should include locality, GPS coordinates, host plant, date of collection, collector name, and unique identifying number.

7. The stringency of sample disinfection before processing represents an additional variable, especially for biological detection of culturable microorganisms. However, the adoption or not of surface sterilization of samples also represents a conceptual decision. Specifically, in the case of sentinel nurseries, superficial contamination of plants might represent an additional pathway of introduction of alien microorganisms that deserves further attention (Vettraino et al. 2017).

Apart from these general principles, which apply to all groups of damaging agents, there are approaches for sample collection that are specific to the affected plant tissues and causal agent groups (Table A3).

Pathogens can affect all plant tissues and cause a broad range of symptoms, which could affect the whole plant (e.g. general dieback) or be more localized (e.g. wilting of individual branches). Based on the tissue affected and the type of damage induced (i.e. damage morphotype, Table A1), it may be possible to recognize which group(s) of causal agent(s) is(are) involved. The strategy for sampling symptomatic material varies according to which tissue is damaged (Table A1). It is important to collect not only the symptomatic parts, to optimize the chances of isolating and identify the causal agent(s). To optimize the chances of isolating the causal agent of the symptoms and not a secondary pathogen, samples should include the region where healthy tissue borders infected tissue (Prospero et al. 2017). Evidence of insect attack (holes in the bark, galleries under the bark, sawdust, resin flows) may also be helpful for detecting the presence of pathogens, as insects can act as vectors of other damaging organisms (Weintraub 2007; Zhao et al. 2007; Akbulut and Stamps 2012; Drenkhan et al. 2017).

Similar to pathogens, sampling of invertebrates varies depending on the affected plant tissue (Table A2) (Kirichenko and Csóka 2017). Invertebrate pests are generally sampled while feeding on plant tissue (to exclude collecting occasional agents that might be on the plant by chance) and preserved for identification. When sampled as immature stages, some arthropods, particularly insects, can be reared to adults in the laboratory as it is the preferred stage for species diagnostics (Gillott 2005). Additionally, plant material with typical arthropod damage can be collected and stored in herbarium collections and used for defining feeding guilds that have added value for identification (Roques et al. 2017). To collect pests, various tools might be used, including nets, umbrellas, collecting trays, aspirators, beating sheets, hand lenses, forceps, and sticky and pheromone traps (Gibb et al. 2006).

Classical techniques

Conventional detection of pathogens involves macroscopic and microscopic examination of symptomatic plant material and isolation of the causal agent. Often, specific isolation protocols, based on optimal requirements for types of pathogens are available, potentially increasing isolation success. However, when working with sentinel plants, there is a risk that causal agents are unknown to science. For this reason, sampled material should be analyzed using a variety of isolation methods, different culture media and temperatures.

Once isolated in pure culture, macroscopic traits, including colony shape, texture and color, and microscopic characteristics of vegetative and reproductive structures are useful criteria for characterization and identification of isolates (Beales 2012).

One problem with the identification of pathogens is the impossibility to grow some organisms on artificial/synthetic media. Obligate parasites such as rust fungi, powdery mildews, viruses and mollicutes require a living host to grow and reproduce. For these organisms vegetative and/or reproductive structure characteristics must be observed on specimens directly from the living host using optical microscopy, or electron microscopy for viruses and mollicutes. Apart from the EPPO protocols, many useful taxonomic manuals, such as Ellis and Ellis (1997), Brenner et al. (2005), Braun and Cook (2012) or Ristaino (2012) can be consulted for morphological identification of fungal, oomycete and bacterial organisms.

Serological tests

Commercially designed kits, such as enzyme-linked immunosorbent assays (ELISA) and lateral flow devices (LFDs) (Lane et al. 2007) are available for detecting and identifying common and known plant pathogens such as the bacterial pathogens Ralstonia solanacearum (Smith) Yabuuchi and R. pseudosolanacearum Safni (EPPO 2018), and viral pathogens like tomato yellow leaf curl begomovirus and tomato mottlebegomovirus (EPPO 2005). With sentinel systems, species-specific serological tests are however unlikely to prove useful, since many of the target microrganisms could be unknown. Thus, only genus-specific LFDs are useful for rapid in situ screening of samples and the selection of appropriate isolation methods for further laboratory testing. For example, for suspected Phytophthora infections, commercial LFDs can give a positive signal enabling the isolation protocol to be oriented towards the use of Phytophthora selective media in the laboratory (Lane et al. 2007).

Molecular barcoding

Molecular-based techniques using polymerase chain reaction (PCR) and Loop-mediated isothermal amplification (LAMP) assays are generally more specific and much faster than conventional techniques and can be applied to non-culturable microorganisms. Plant protection organisations routinely rely on diagnostic methods based on PCR assays, e.g. EPPO Standards (https://www.eppo.int/RESOURCES/eppo_standards). The most commonly used markers for molecular identification of fungal pathogens are the ribosomal DNA transcribed spacers, particularly the internal transcribed spacer (ITS) regions ITS1 and ITS2 (Schoch et al. 2012; Romanelli et al. 2014). Although ITS regions perform generally well as barcoding markers for many fungal taxa, this region is less useful for some genera, such as Fusarium or Penicillium, as these taxa have narrow or no barcode gaps in the ITS regions (Raja et al. 2017). Thus, additional regions must be sequenced. Commonly used regions include the two largest subunits of RNA polymerase II (RPB1, RPB2), β-tubulin regions or translation elongation factor 1 α (TEF1α), which can resolve identification of individual species within the various groups (Schoch et al. 2012). These gene regions are routinely used, depending on the organism (Romanelli et al. 2014). The 16S ribosomal RNA gene and chaperonin-60 (cpn60) are used as bacterial barcode marker genes and to study bacterial phylogeny (Chakraborty et al. 2014). Detection and identification of phytoplasma and spiroplasma are primarily based on 16S rRNA (16Sr) amplification followed by restriction fragment length polymorphism analysis (Bertaccini et al. 2019). When genetic information is available, PCR and reverse transcription PCR are used to detect plant viruses (Jeong et al. 2014).

Rapidly evolving high-throughput sequencing (HTS) technologies enable simultaneous identification of thousands of organism species from numerous and complex samples, with protocols available for viruses, bacteria, fungi, oomycetes and animal pests (Abdelfattah et al. 2018; Tedersoo et al. 2018). The available HTS platforms and details for analysis steps are outlined in Tedersoo et al. (2018). Selecting molecular markers of enough resolution, primers of high affinity to templates, negative and positive control samples and reliable reference sequence databases are the most important factors for HTS-based pest and pathogen identification (Tedersoo et al. 2018). Correct reference data are critical in the precise identification of plant pathogens and, at present, not all publically available databases are sufficiently accurate to enable accurate identification (Jayasiri et al. 2015). Thus, it is crucially important to improve and correct pest and pathogen sequences in publicly databases (Nilsson et al. 2014)

Third-generation sequencing technologies such as PacBio (www.pacificbiosciences.com) and Oxford Nanopore (www.nanoporetech.com) present the possibility to sequence long reads. These technologies have not yet been used in sentinel systems. The benefits arising from amplifying other regions (with sequences longer than ITS1 or ITS2), that could give better identification at the species level, are countered by the absence of adequate reference databases to blast the result obtained. Moreover, these sequencing technologies currently have higher error rates compared with Illumina (Weirather et al. 2017). Despite this problem, it is necessary to emphasize that the new HTS system, such as the MinION device from Oxford Nanopore has great promise as a useful tool in field applications since its portability allows for in situ (on-site) analysis and real-time data generation, thus making the workflow fully versatile.

The use of HTS platforms for biosecurity purposes such as identifying latent or potentially opportunistic pathogens in asymptomatic host tissues requires some consideration of the technological limitations, including the quality of data output (e.g. Illumina MiSeq). While bioinformatics processing can provide useful data output for biodiversity studies (e.g. metacommunity analysis), blast searching of filtered sequence data against custom or public databases generally results in a limited number of identified species, but with many OTUs assigned to higher taxonomic levels. This problem arises due to following reasons: 1) the low power of single-marker short sequences in differentiating taxa, 2) the low taxonomic coverage of databases, and 3) sequencing errors accumulated in the output reads (the sum of amplification and HTS errors). The result is a limited number of OTUs assigned at the species level which may give some value to biodiversity studies but not for biosecurity purposes.

Classical techniques

The observation and evaluation of damage on plants is the first step towards a diagnosis of damaging arthropod and nematode pests. Damage morphotypes can be effectively utilized in sentinel planting surveys as an identifier to assign phytophagous pests to certain feeding guilds, prior to species identification using morphology-based taxonomy (Roques et al. 2017). Classical taxonomy based on morphological characteristics is undoubtedly a powerful tool for arthropod and nematode identification, but some limitations exist, mainly due to the immense diversity and existing gaps in taxonomic knowledge. In most cases, keys are useful only for certain geographic regions and are often based on the identification in the adult stage (Gillot 2005). Furthermore, morphology-based taxonomy may not be helpful for discrimination of closely related species (e.g. sibling or cryptic species) (Bickford et al. 2007). Moreover, disagreements between taxonomists on defining morphological characters, redefining and synonymizing the species may complicate species identification procedures (Okiwelu and Noutcha 2014). Developments in visualizing tools (electron, fluorescent and scanning microscopy) have led to immense improvements in classical taxonomy and continue to contribute to the precision of morphological observations of arthropods and their documentation, which greatly increased the accuracy of species identification (Klaus and Schawaroch 2006; Lee et al. 2009). Some biometric parameters of arthropod body characters could provide added value for distinguishing species (Su et al. 2015). The nematode species can be identified based on the morphological features of the sexual organs of adult male nematodes (Seesao et al. 2016). Knowledge of species biology (life cycle, phenology) and ecology (range, habitat, ecological niche, host plant association) may provide important additional data when identifying taxa (Panizzi and Parra 2012).

The rapid development of computer vision technologies has led to applications in highly promising automatized arthropod identification platforms based on multivariate biometric features of the taxon. This novel approach, based fully on classical taxonomy and computer algorithms, allows species identification procedures to be performed even by non-taxonomists, with a high degree of reliability (Watson et al. 2003; Hassan et al. 2014; Yang et al. 2015; Favret and Sieracki 2016; Wang et al. 2017). Despite being highly attractive, automated species identification suffers from a number of limitations, the most significant being the limited applicability of automated platforms which have for now been created only for a few groups of insects (e.g. individual families of Lepidoptera or Diptera) (Watson et al. 2003; Yang et al. 2015; Favret and Sieracki 2016; Wang et al. 2017), whereas other large groups of important arthropod pests remain far outside the scope of these systems. The process preceding the automated species identification can be tedious, including specimen preparation for scanning and precise positioning for digitizing and recognition by the software. In addition, the computer algorithms may not always be perfect and identification accuracy may not be satisfactory. Despite these and other disadvantages, this developing technology and its possible utilization in mobile devices and other digital instruments in user-friendly mode, would be in high demand for modern forestry and agriculture (Wang et al. 2017) and could also be highly applicable to the identification of potential arthropod pests in sentinel nurseries and plantations.

Molecular barcoding

DNA barcoding is a well-known molecular approach to species identification (Hebert et al. 2003), applicable to any life stage of arthropods, including immature stages (egg, larva, pupa) most often be identified reliably to species level by morphological characteristics (Hebert and Gregory 2005). The method can be highly useful in sentinel plantings, where the pests are usually found in immature stages (Roques et al. 2015).

For arthropods, DNA barcoding uses a short genetic marker – a fragment of mitochondrial DNA (mtDNA) of the cytochrome oxidase I gene (COI; barcoding fragment 658 bp) (Hebert et al. 2003). However, this gene might not always be enough to delineate arthropod sibling species robustly and other molecular methods are required, including nuclear sequencing and/or amplified fragment length polymorphism genotyping (Dasmahapatra et al. 2010; Kirichenko et al. 2015).

As for pathogens, one of the limitations of DNA barcoding is the lack of appropriate reference databases, which would cover all formally described arthropods. To date, comprehensive databases have been accumulated mainly for certain insect taxa (e.g. Lepidoptera and Coleoptera on http://www.boldsystems.org/; Ratmasingham and Hebert 2007), whereas other groups of arthropods remain underrepresented. In the existing databases, inaccuracies may also appear which can lead to misidentification. The quality and accuracy of the sequences stored in the genetic databases might not always be satisfactory, especially considering that any user can access and add sequences (Hebert and Gregory 2005). In a recent survey of insects that colonized a sentinel plantation in China, DNA barcoding enabled to reliably identify only one quarter of sample insect species (Roques et al. 2015)

For nematodes, several genes are targeted for identification such as the mitochondrial cytochrome b locus (mtDNAcytb) (Mattiucci et al. 2003), the gene encoding the mitochondrial cytochrome oxidase 2 (COX2) (Valentini et al. 2006) and the mitochondrial cytochrome oxidase 1 (COXI) (Blouin, 2002), the ribosomal RNA of the small (ssrRNA) and large subunit (lsrRNA) (Hu et al. 2001). Other nuclear genes were also selected such as the internal transcribed spacer 1 (ITS1) of rDNA to identify Strongylidae and Anisakidae (Roeber et al. 2013). NEMBASE (http://www.nematodes.org/nembase4), a publicly available database, provides access to sequences and associated meta-data on parasitic nematode expressed sequence tags (Elsworth et al. 2011). WormBase is an international consortium of biologists and computer scientists dedicated to the research community and providing accurate, current, accessible information concerning the genetics, genomics, and biology of Caenorhabditis elegans Maupas and related nematodes (http://www.wormbase.org).