The 20th century philosopher Ludwig Wittgenstein said, “The limits of my language mean the limits of my world” (Wittgenstein, 1922). By learning the communication systems of other organisms, humans may be able to expand their world view and increase their understanding of other species. Chemical communication language is the syntax of a ubiquitous biological language, and highly diffusible low molecular weight volatile compounds constitute the alphabet of this language (Mithofer and Boland, 2016). Plants and microorganisms release blends of volatile compounds to their surrounding environment. The constituents of volatile blends vary with taxon as well as with organism age and exposure to biotic and abiotic stresses (Jiang et al., 2016; Niinemets and Monson, 2013; Sharifi et al., 2018). Accordingly, volatile blends reflect the presence of emitter organisms and their physiological status, whether healthy or stressed (Niinemets et al., 2013; Sharifi et al., 2018). It is not yet clear whether emitter organisms release volatiles for deliberate communication with neighbors, but receiver organisms have evolved processes for understanding information imparted by volatiles (Heil, 2014; Sharifi and Ryu, 2018a, 2018b).

Biogenic volatiles can thus be considered as infochemicals that act as intra-species, inter-species, and inter-kingdom signal communications. For example, bacterial volatile compounds (BVCs) influence several steps in plant physiological development, from seed germination to harvesting time (Sharifi and Ryu, 2018a). Biogenic volatiles therefore have potential as novel fertilizers or plant stimulant in modern agriculture. Volatiles can also induce systemic defenses against biotic stress (Chung et al., 2016): and can act as inter- and intra-species signals in plants (Yi et al., 2009), and therefore have potential as novel pesticides. In recent years, biogenic volatiles have been exploited to modulate plant physiology (Sharifi and Ryu, 2016; Song et al., 2015), to detect diseased or stressed plants (Jansen et al., 2011; Niinemets et al., 2013), and to discriminate between bacterial species (Halbfeld et al., 2018; Lough et al., 2017; Spinelli et al., 2012; Wilson et al., 2004).

Biogenic volatiles thus constitute a rich source of information that could be highly valuable in modern precision agriculture. For example, real-time analysis of microbe-induced plant volatiles (MIPVs) could allow rapid, noninvasive detection of plant diseases. Bacterial volatiles also have great potential for application in the field as fertilizers or as inducers of natural plant defenses (Ryu, 2015; Song and Ryu, 2013). However, practical use of biogenic volatiles in the field remains challenging due to their high vapor pressure. Research is lacking into the development of effective application formulations for BVCs. Here, we discuss how slow-release technologies such as microencapsulation and sol-gel could facilitate preparation of BVC formulations amenable to commercial application. The use of BVCs in drip irrigation systems, storage, nurseries, and open fields is also discussed.

Plant volatiles function as infochemicals for neighboring plants and for other organisms in the phytobiome. Infected or injured plants release blends of volatile compounds that differ from those released by healthy plants (Sharifi et al., 2018). Plant volatiles could be used to detect infected plants during early stages of disease development, earlier than detection was possible using other methods (Aksenov et al., 2014; Cellini et al., 2016; Martinelli et al., 2015; Wilson et al., 2004). Here, we review the most recent developments in non-invasive detection of MIPVs. Gas chromatography-mass spectrometry (GC-MS) is a basic analytical method for volatile detection that can be used to measure the relative amount of each compound in volatile emissions. Several methods have been developed for sampling and preconcentration of volatiles from plant samples either in lab or field conditions. In passive (static headspace) sampling methods, volatile traps by absorbent material in headspace of plant sample. The adsorbed compounds will release in thermal desorption system of GC. Head space (HS)-solid phase microextraction (SPME) is a common passive sample preparation method for plant volatiles (Deasy et al., 2016; Dettmer and Engewald, 2002; Kallenbach et al., 2014; Tholl et al., 2006). However, because low amounts of absorbent fiber (polydimethylsiloxane (PDMS)) are employed in SPME (less than 0.5 μl), stir bar sorptive extraction (SBSE) was subsequently developed, which employs a thicker volume of coating fiber on a stir bar to increase the capability and sensitivity of volatile sampling (Kallenbach et al., 2014). Kallenbach et al. (2015) proposed “silicone tubing” as a cheap, fast, high-throughput method for headspace sampling, and this method performed well in field experiments. In passive sampling methods, plant part enclose in a plastic or glass bag. This chamber alter humidity, temperature, and light spectrum around plant organs and subsequently change volatile emission. To address this limitation, direct-contact sorptive extraction were developed in which a magnetic PDMS coated stir bar held on leaf by a magnet (Kfoury et al., 2017). This method do not need any chamber and can be used in field condition. Passive volatile sampling methods could be adapt for root system. Volatiles can be sampled from soil headspace by SPME method. However, it is better to collect volatiles from the rhizosphere. Eilers et al. (2015) exploited two perforated polytetrafluoroethylene (PTFE) discs to make chamber in bottom of pot. Plant roots passed through the top disc entering the chamber and emitted volatiles from roots were trapped by PDMS tubes. Later, Deasy et al. (2016) insert some air permeable PTFE tubes around root to collect rhizosphere volatiles. SPME apparatuses were mounted on tubes allowing temporal root volatile sampling in field condition. In dynamic headspace sampling methods, air continuously pumped over plant headspace and volatiles traps and enriches in adsorbent materials such as active charcoal, Tenax, Carbotrap X and Carboxen. Dynamic sampling prevent headspace accumulation of volatiles that may have negative impact on plant volatile emission (Gallego et al., 2010; Giacomuzzi et al., 2016; Quintana-Rodriguez et al., 2015). In compared to air flow in open systems, closed-loop-stripping-analysis is a dynamic sampling method employing a closed system which minimizes air contaminants and increase its ability to traps low amount of volatiles. However, this method need solvent extraction step and is unable to detect compounds that their mass is below solvent mass such as methanol, carbon dioxide, acetaldehyde and ethanol (Giacomuzzi et al., 2017). Some of dynamic sampling methods use adsorbent suitable for thermal desorption in GC. As different adsorbents have different affinity for volatile compounds it is better to use multiple layer of adsorbents in a multi-sorbent tube (Gallego et al., 2010).

There are two main disadvantages to the current GC-MS method: 1) GC-MS takes precise snapshots of volatile profiles, but volatile organic compound (VOC) emission occurs dynamically in nature; and 2) GC-MS requires the use of heavy laboratory equipment which cannot be used in the field. To overcome the pitfalls of GC-MS and employ the technique under in situ conditions, Contreras et al. (2008) invented a portable GC-TMS, which was later improved by adding a dynamic needle trap system for online detection of plant volatiles (Beck et al., 2015). Beck et al. (2015) showed that this method could differentiate between damaged and healthy Centaurea solstitialis in the natural environment. Portable GC-MS could efficiently discriminate between stem rot and anthracnose diseases of mango using 1-pentanol and ethyl boronate as marker volatiles for stem rot and thujol as a marker of anthracnose (Moalemiyan et al., 2006). Aksenov et al. (2014) coupled differential mobility spectrometry (DMS) with GC to detect citrus greening disease in plants at early stages of disease development. The portable GC/DMS device proved more effective than qPCR in detecting the unculturable phloem-limiting bacterium Candidatus Liberibacter solanacearum in apples under field conditions. Other powerful real-time instruments such as selected ion flow tube (SIFT)-MS and proton transfer reaction (PTR)-MS are also available for analyzing plant volatiles. PTR-TOF-MS is more sensitive than SIFT-MS for real-time VOC detection (Biasioli et al., 2011; Materic et al., 2015). PTR-MS allows partial identification of volatiles and is a powerful tool for the rapid monitoring of changes in volatile emissions over time (Fig. 1). PTR-MS is more suitable for the detection of low molecular mass volatiles such as ethanol, methanol, and propanol than GC-MS (Portillo-Estrada et al., 2015); however, the lack of a capillary column in PTR-MS makes the technique unsuitable for separating isomeric analytes of the same molecular mass. Ruzsanyi and colleagues (2013) installed a 20 cm long multi-capillary column at the inlet of the PTR-TOFMS system, and this allowed the efficient separation and quantification of aldehydes with the same fragmentation pattern and separation of the isomeric ketones 3-heptanone and 2-methyl-3-hexanone. This method holds promise for the precise, non-invasive detection of MIPVs in near real time (Fig. 1).

Non-destructive detection of diseased plants through analysis of volatile profiles is likely to become widespread in modern agriculture. However, the heavy bench-top instruments that have been traditionally employed are not conducive to field, commercial greenhouse, or storage applications. In recent years, portable user-friendly devices such as E-nose, portable GC-MS, and GC/DMS have been developed to address this challenge (Fig. 1). The E-nose instrument contains an array of electronic chemical sensors, such as metal-oxide and conductive polymers, which change in their electrical conductivity upon interaction with volatiles. These conductivity changes can then be assessed by multivariate analysis techniques such as principal component analysis and cluster analysis to differentiate between samples. The E-nose technology is thus able to differentiate between sample volatile profiles but is not designed to detect individual volatiles. Biondi et al. (2014) used E-nose for detection of bacterial brown and ring rot in commercial potato storage conditions, with efficient detection of diseased tubers achieved in 90% of samples. E-nose was potentially able to detect a single infected asymptomatic tuber within a 10 kg box of healthy tubers (de Lacy Costello et al., 2000). E-nose can also be used in complex situations. Not only could the instrument discriminate plant species, but it was also able to distinguish between wounded, pest-infested, and pathogen-infected plants in experimental conditions (Laothawornkitkul et al., 2008). Cellini et al. (2016) assessed GC-MS, PTR-MS, and E-nose for early detection of fire blight on apple plants. E-nose was identified as the best technique for detection of disease plants but not for identification of volatiles. Overall, the E-nose is a simple, portable, easy-to-use device that can be used for early detection of stressed plants (Fig. 1).

Overall, researchers can choose appropriate methods in plant disease detection based on their criteria such as sensitivity, specificity, speed, in situ application, non-destructive and real-time application. Vast array of volatile analysis methods can provide high speed, non-invasive and in field detection of plant disease in early stage of disease development. But, molecular methods such as real-time PCR are quantitative and highly specific in detection of pathogen species. It is also noteworthy that volatiles emission is a dynamic phenomenon, under control of ecological stimulus and physiological conditions. Volatiles emission could be affected spatially and temporally by plant phenological stage, damage severity, circadian clock, type of plant-pathogen interaction and saprophytic colonization of necrotized tissues (Attaran et al., 2009; Giacomuzzi et al., 2017; Sharifi et al., 2018). So, standardization of sampling conditions has important role as same as choosing proper instrumental tools. It is just possible to compare results from same plant phonological stage in same diurnal time and same stage in disease development. This seem to be important issue in case of hemibiotrophic pathogens which have both biotrophic and necrotrophic interaction with plant.

Most research into the activities of plant-growth-promoting BVCs utilize a bipartite Petri dish (referred to as I-plate) design (Bailly and Weisskopf, 2012), with plants on one side of the plate and the emitter bacteria on the other side, both growing on MS media (Ryu et al., 2003). Several media have been used for growth of volatile-producing bacteria. In general, volatiles from bacteria grown on poor media enhance plant growth, whereas volatiles from bacteria grown on rich media inhibit plant growth (Asari et al., 2016; Blom et al., 2011). Bacteria produce HCN, H₂S, and NH₃, all volatile nonorganic compounds, in glycine and protein-rich media, and these have inhibitory effects on plants (Piechulla and Schnitzler, 2016). However, these studies were all obtained using Petri dish tests, and production of BVCs in soil is affected by several factors such as nutrient and oxygen availability, pH and temperature of soil, soil texture (Effmert et al., 2012; Wheatley, 2002), and bacterial GacS/GacA regulatory system (Cheng et al., 2016; Ossowicki et al., 2017). Further studies are needed to evaluate the types and concentrations of BVCs or individual volatiles that improve plant growth under natural conditions. Park et al. (2015) designed an experiment to investigate the effects of volatiles under soil conditions. A single colony of Pseudomonas fluorescens SS101 was placed in a chamber at the bottom of a pot and the chamber was covered with a filter. The pot was filled with soil and tobacco seeds were sown. Volatiles produced by P. fluorescens approximately doubled the plant fresh weight in that experiment. The same experiment was performed on tomato plants (Tahir et al., 2017), and B. subtilis SYST2 BVCs significantly increased tomato growth and photosynthesis. Together, these results indicate that bacterial volatiles could be effective even under natural conditions.

Production of BVCs is affected by several environmental conditions. As a consequence, results from application of emitter bacteria in natural conditions are not always reproducible. Experimental application of pure volatiles can yield more reproducible results, though it should be noted that the plant growth activity of BVCs is usually concentration dependent. VOC concentrations can vary in the 30-200 ng/g range on different soil types (Minnich, 1993) and plants can sense and respond to trace amounts of volatiles at levels below 120 parts per trillion (Shiojiri et al., 2012). Some pure volatiles are active in this range (Piechulla et al., 2017). Ryu et al. (2003) showed that 100 ng of 2,3-butanediol was the optimum amount for increasing Arabidopsis leaf surface area. The effects of higher concentrations of 2,3-butanediol were not significantly different from those of water. A similar result was obtained for 0.63 ng of indole (Yu and Lee, 2013), 50 ng of 13-tetradecadien-1-ol (Park et al., 2015), and 0.75 μM dimethylhexadecylamine (Velázquez-Becerra et al., 2011). Optimum concentrations are specific to each BVC/plant combination. For example, optimal concentrations of dimethyl hexadecylamine were not the same in sorghum and alfalfa (Castulo-Rubio et al., 2015; Velázquez-Becerra et al., 2011). Application of pure BVCs has been tested in the field. The pure volatiles 3-pentanol and 2-butanone increased fruit production in cucumber by 6- and 4-fold, respectively, under field conditions (Song and Ryu, 2013), and soaking seedlings in 1 mM 3-pentanol enhanced pepper resistance to two pathogens under field conditions (Choi et al., 2014). These results indicate that plant growth-promoting rhizobacteria (PGPR) volatiles have great potential to enhance plant growth and health under field conditions. BVCs might thus be considered as emerging gaseous fertilizers in modern agriculture.

Volatiles generally have high vapor pressure and are highly diffusible in air and water (Chung et al., 2016), characteristics that are advantageous for intra- and inter-specific communications. However, these properties are the main challenges for creating formulations of BVCs that can be used in agricultural applications. Volatiles are unstable and prone to react with highly reactive chemicals, and also have a short half-life and evaporate rapidly under natural conditions (Bakry et al., 2016). However, long-term exposure to BVCs seems to be necessary for plants to induce beneficial effects (Xie et al., 2009).

Blends of volatile compounds can be biological indicators of the taxonomic position of an organism. However, the volatile blends produced by organisms also change in response to biotic and abiotic stresses. Conspecific neighbor and ecological counterpart species have learned to monitor the blends of volatile compounds and interpret the information therein during evolutionary history. If the language of volatile signals can be understood, these compounds can be exploited to improve agricultural productivity. Careful exploration of this chemical language is necessary to develop blends that are beneficial and non-harmful to the environment or human health. Safe applications include the non-invasive detection of plant diseases, but field applications of bacterial volatiles to control pests and diseases will require a more cautious approach. In conclusion, the chemical language of VOCs has great potential for exploitation in modern agriculture.

This work was supported by the Strategic Initiative for Microbiomes in Agriculture and Food, Ministry of Agriculture, Food and Rural Affairs, Republic of Korea (as part of the (multi-ministerial) Genome Technology to Business Translation Program) (918017-4) and the Advanced Biomass R&D Center (ABC) of the Global Frontier Project funded by the Ministry of Science and ICT (ABC-2015M3A6A2065697), and the KRIBB Initiative Program, South Korea.