Most vascular plants constitutively emit volatile organic compounds (VOCs), but emissions may substantially increase and diversify under conditions of abiotic and biotic stress (Holopainen and Gershenzon, 2010). Feeding by herbivores was found to induce the emission of novel volatile compounds often referred to as herbivore-induced plant volatiles (HIPVs) (Hare, 2011) that attract natural enemies of the herbivores. This was shown for the first time in seminal studies conducted with spider mites and predatory mites by Dicke and Sabelis (1988) and with moth larvae and parasitic wasps by Turlings et al. (1990). Since the first studies of plant volatiles, interest in the synthesis and control of volatiles by plants and their ecological and atmospheric functions have increased substantially (Dicke and Loreto, 2010) and several other ecophysiological and ecological effects of constitutive and inducible plant volatiles have been described (Laothawornkitkul et al., 2009; Holopainen and Gershenzon, 2010; Loreto and Schnitzler, 2010; Peñuelas and Staudt, 2010).

HIPVs are often expected to increase the fitness of the emitting plant either directly or indirectly (Dicke, 2009). Direct defense reduces herbivore approach and attack or decreases the herbivore's consumption rate, but the roles of HIPVs could be complicated. For instance, the ratios of compounds in the typical HIPV profiles of Quercus robur trees correlate with the tree's susceptibility to herbivore damage (Ghirardo et al., 2012). Trees representive of an herbivore-resistant phenotype emitted HIPVs that included the sesquiterpenes α-farnesene and germacrene D and were avoided by females of the defoliating moth Tortrix viridian (Ghirardo et al., 2012). However, in the same outbreak area trees emitting other typical HIPVs including the monoterpene β-ocimene and homoterpene (E)-4,8-dimethyl-1,3,7-nonatriene (DMNT) were susceptible and were largely defoliated (Ghirardo et al., 2012).

Indirect defense involves the recruitment of natural enemies of herbivores to increase predation or parasitism rates and eventually reduce damage. Many laboratory reports have given support to this hypothesis (Hare, 2011). However, there is scarce field-based -evidence that attraction of natural enemies by HIPVs actually reduces herbivore populations (Kessler and Baldwin, 2001). Furthermore, studies to show improved Darwinian fitness in plants emitting HIPVs, i.e., more offspring in the next generation, are lacking (Hare, 2011). A current ecological view of the role of HIPVs is as components of a wider infochemical web, including e.g., pollinators and root synergists, that overlay the food webs of a community rather than simply defending against attackers or attracting carnivores (Dicke and Baldwin, 2010). Ghirardo et al. (2012) concluded that for Q. robur, the strategy of emitting herbivore-repellent rather than natural enemy attracting HIPVs, appears to be the better mechanism for avoiding defoliation. However, when plants are influenced by a diverse community of chewing and sucking herbivores, a single HIPV compound could be an efficient repellent against one herbivore, but act as an attractant of another herbivore and many of the community's predators and parasitoids (Xiao et al., 2012).

HIPVs are not only emitted by aboveground parts of plants. Many plant species have an extensive root system where HIPV-releasing resins are stored (Kivimaenpää et al., 2012) or HIPVs synthesized (Degenhardt et al., 2009) and released in the soil air space and eventually to the atmosphere. Root feeding by herbivores induces the emission of HIPVs by the root system, which act as belowground attractants for parasitic nematodes (van Tol et al., 2001; Rasmann et al., 2005). Aboveground herbivory also has a systemic impact on belowground HIPV production and vice versa (Erb et al., 2009). Defoliation of Pinus syvestris by diprionid sawflies induced substantial HIPV emissions from the shoots, but resulted in significant reduction in monoterpene and sesquiterpene emissions from the root system (Ghimire et al., 2013). This was expected to be related to reduced carbon allocation to below-ground parts after defoliation. Arbuscular mycorrhizal (AM) infection of bean plant roots affected the HIPV composition emitted by foliage by making it less attractive to predators (Schausberger et al., 2012), whereas an ectomycorrhizal (EM) root symbiont did not affect the terpene pool of pine needles (Manninen et al., 1998). These studies highlight the complex and systemic nature of HIPVs and the need for a holistic view of a plant's volatile emissions and their various related roles.

HIPV compounds typically have relatively short atmospheric lifetimes after release from plants, which may limit the efficiency with which they attract natural enemies of herbivores and mediate other ecological interactions (Yuan et al., 2009). However, reactive VOCs have various functions in the atmospheric processes, such as formation of ozone in NO_x polluted atmospheres (Atkinson and Arey, 2003), formation of OH-radicals (Mentel et al., 2009), formation of organic nitrates (Pratt et al., 2012) and formation of secondary aerosols (SOA) (Joutsensaari et al., 2005; Kiendler-Scharr et al., 2009; Mentel et al., 2009; Virtanen et al., 2010). Laothawornkitkul et al. (2009) divided the various functions of plant VOCs into three broad categories; biological, chemical and physical.

In this review we focus on the different roles and fate of inducible VOC molecules after release from the VOC synthesizing plant including biological, chemical and physical aspects. Furthermore, we discuss why biosynthesis and emission of short-lived volatile compounds has evolved as a general response to external biotic damage to plants. We will pay attention to the biological and ecological role of VOCs post-emission and their atmospheric reaction products. These effects may take place in their gaseous, liquid and solid organic particulate forms in the atmosphere, but probably also after deposition on plant surfaces. We also discuss how the post-emission reaction products of VOCs may improve plant fitness. Finally we try to present the potential routes that the carbon fixed by a plant and bound in VOC molecules will ultimately take.

The majority of the typically documented herbivore-induced plant volatiles can be classified in three major chemical groups based on their biosynthesis pathways or their known within-plant functions (Holopainen and Gershenzon, 2010). First, and the dominating group of constitutively emitted VOCs and HIPVs in many plant species, are the terpenoids, which are produced by two separate pathways, one active in plastids (MEP) and one (MVA) in the cytosol (Loreto and Schnitzler, 2010; Maffei, 2010). The volatile terpenoids (Figure 1) include the five carbon (C5) isoprene molecule and a range of molecules comprising various multiples of this basic C5 unit, including monoterpenes (C10), homoterpenes (C11 or C16) and sesquiterpenes (C15). The second group is the C6 lipoxygenase (LOX) products better known as Green Leaf Volatiles (GLVs). GLVs, such as (Z)-3-hexenal and (Z)-3-hexenyl acetate are compounds released after mechanical or other destructive damage to cell membranes (Maffei, 2010; Holopainen, 2011). The third group is the volatile aromatic compounds such as methyl salicylate and indole produced by the shikimate pathway and containing an aromatic ring (Maffei, 2010). In addition to these three main groups there are a multitude of other volatile compounds that are specific to varying degrees such as to an order, genus or species. The volatile plant hormone ethylene has often been considered an inducible volatile and has several functions in plant physiological processes and growth, while membrane-bound ethylene receptors are well known (Holopainen and Gershenzon, 2010).

HIPVs can also be classified based on their volatility or the atmospheric life times of the compounds in atmospheres with standardized levels of reactive scavengers such as ozone (O₃), nitrate (NO₃) or hydroxyl (OH⁻) radicals (see Holopainen, 2011; Holopainen et al., 2013 and Table 1). These scavengers appear in higher concentrations in polluted air, but they are common in ambient air and are involved in important chemical and physical processes in the atmosphere. Significantly, plant VOCs also participate in their formation (Hallquist et al., 2009). A greater number of C-C double bonds in the VOC molecules (Figure 1) will make them more prone to reactions with atmospheric radicals, degrade faster and form more particles (Hoffmann et al., 1997; Atkinson and Arey, 2003; Pinto et al., 2010).

When a plant is attacked by a herbivore, there is a succession of different inducible volatile compounds appearing in the emission bouquet. Inducible defenses of herbivore-attacked plants involve herbivore perception, transcriptional responses, protein formation and biosynthetic responses (Dicke, 2009). In plant species storing a constitutively synthesized volatile mixture, these compounds will volatilize very rapidly upon rupture of storage structures such as glandular trichomes and constitute the first response to external damage (Jansen et al., 2011). In this case the emission is the result of mechanical injury, which will occur before the attacker induces the biosynthesis of volatiles in growing plant tissues. The GLV emissions have a time lag between herbivore-feeding and compound emission of from just a few seconds to several minutes and these compounds show very rapid response to mechanical or biological damage to cell membranes. The C18 fatty acids of membranes are cleaved to C12 and C6 compounds by hydroperoxide lyases with 3-Z-hexenal (aldehyde) being the first C6 GLV compound synthesized by the lipoxygenase (LOX)/lyase pathway (Maffei, 2010). This compound is then converted to other common C6 GLVs such as (E)-2-hexenal (aldehyde), 3-hexenol (alcohol) and (Z)-3-hexenyl acetate (ester) (Shiojiri et al., 2006). Geometrid moth feeding induces emissions of GLVs, which peak soon after larval feeding starts on deciduous tree foliage (Blande et al., 2007), while on-line monitoring of HIPVs shows that GLV peaks can even reveal the timing of larval feeding periods (Schaub et al., 2010).

Continuous mechanical injury (Mithofer et al., 2005), oral secretions of herbivores (Turlings et al., 1990), plant cell membrane damage by biotrophic fungal leaf pathogens (Toome et al., 2010; Jansen et al., 2011) or bacterial pathogens (Yi et al., 2009) elicit signal transduction pathways that are mediated by phytohormones such as jasmonic acid (JA) and ethylene in the case of chewing herbivores, salicylic acid (SA) and ethylene in the case of fungal pathogens and result in the synthesis of typical HIPV terpenoids and aromatic compounds (Jansen et al., 2011). There is variability in the succession of different herbivore-induced terpenoid emissions, which could be partly due to variable allocation of precursors into the different biosynthesis pathways. Biosynthesis of the homoterpene (E)-DMNT, for example, could originate predominantly from the MVA-pathway in herbivore-stressed plants, while the fungal elicitor alamethicin stimulates the biosynthesis of (E)-DMNT via the MEP-pathway (Bartram et al., 2006). In Alnus foliage damaged by geometrid moths, emissions of (E)-β-ocimene and (E)-DMNT peaked on day 3 (Copolovici et al., 2011). In the same study the emission kinetics of the sesquiterpene (E,E)-α-farnesene tended to be biphasic with peaks on days 2 and 4 after the start of larval feeding. Emission rates of the induced LOX products, (E)-β-ocimene and (E,E)-α-farnesene were positively correlated with the number of larvae feeding (Copolovici et al., 2011).

Variation in the feeding strategies of herbivores can result in profound variation in the volatiles emitted by damaged plants. Herbivores that feed via stylets, such as aphids, inflict apparently minor mechanical damage, but still induce the emission of a rich blend of volatiles including both GLVs and terpenoids (Gosset et al., 2009). Sustained feeding by aphids and colony growth can also result in large increases in emission of methyl salicylate (Blande et al., 2010), which can take several days to start appearing in emission bouquets. The volatiles induced by chewing herbivores can vary with the life stage of the herbivore, with early instars (first to fourth) of Pseudaletia separata larvae inducing different volatile bouquets to larger more advanced larvae (fifth to sixth instars) (Takabayashi et al., 1995). In this case, foraging parasitoids are able to distinguish between the volatile blends induced by potential host larvae (the younger instars) and larvae that are too old to be used as hosts and may actually constitute a threat through their aggressive defensive behaviors (Takabayashi et al., 1995). In an alternative system, with Pieris brassicae feeding on Brussels sprouts, Cotesia glomerata parasitoids do not appear to determine larval instar through volatile emissions, but can determine presence of suitable hosts through other cues on infested leaves, without necessarily contacting the host itself (Mattiacci and Dicke, 1995). Elicitors in the saliva of the herbivores are responsible for alterations in the herbivore-induced blend, the specificity and range of which can vary (Mattiacci and Dicke, 1995; Takabayashi et al., 1995; Roda et al., 2004). It has also been shown that deposition of eggs by the Brassica specialising Lepidopteran Pieris brassicae can induce changes in the expression of hundreds of genes (Little et al., 2007; Fatouros et al., 2008) and emission of volatiles that are attractive to its parasitic wasps (Fatouros et al., 2012).

In Salix hybrid plantlets infected with Melampsora epitea leaf rust fungi, the total monoterpene emissions did not change although a stress-signaling compound (Z)-β-ocimene showed an increase in infected plants on several days. The infection also increased the emission of sesquiterpenes and LOX products by factors of 175-fold and 10-fold, respectively (Toome et al., 2010). The induced VOCs showed two clear peaks during the experiment; at 6–7 and 12 days post-infection, whereby the relative volatile emission signal increased to about 6-fold that of uninfected plants. Peak emission periods were directly connected to rust infection with day 6 corresponding with the appearance of the first rust pustules on the leaves and day 12 corresponding with necrosis developing around several pustules (Toome et al., 2010).

Isoprene -a major biogenic VOC released from vegetation—and some monoterpenes are constitutively emitted, but are induced by elevated temperatures, which can greatly enhance the overall emission of these compounds (Loreto and Schnitzler, 2010). Of these compounds, isoprene in particular is not induced by fungal pathogens or insect feeding. Toome et al. (2010) reported that isoprene emissions from Salix hybrids with rust-infected leaves decreased 3-fold compared to controls, Ghirardo et al. (2012) found that Tortrix viridiana larval feeding did not affect isoprene emissions from Quercus robur and Blande et al. (2007) found that emission of isoprene in Populus hybrids did not respond significantly to geometrid moth or leaf weevil feeding.

Several studies have implicated the blend of volatiles emitted by plants either constitutively, or after herbivore-damage, to play an important role in the behavior of foraging insects. Aphids in particular have been shown to utilize blends of volatiles emitted by undamaged plants as host location cues (Bruce and Pickett, 2011; Webster, 2012), while parasitoids have also been shown to utilize chemical blends to locate their hosts (Pareja et al., 2009; Clavijo McCormick et al., 2012). It is clear that the blend emitted by plants can evolve as the degree of herbivore-induced stress changes. Differences in degradation rates of certain chemicals could result in rapid changes to the blend, by reducing the proportions of compounds relative to each other (Pinto et al., 2007a), which could render the blend less effective as a cue for foraging insects (Pinto et al., 2007b).

The ecological functions of HIPVs described above have all been established experimentally. All rely mechanistically on the responses of a receiver organism to volatile emissions from a plant that has been subjected to a degree of stress or stimulation. While the volatile compounds remain intact they theoretically constitute a signal that can be detected and potentially perceived by organisms of the surrounding community. However, as soon as the chemicals leave the plant there are a range of potential fates or roles that could be played out. In the following section those potential adaptive roles have been classified as roles for intact HIPVs and for reaction products of HIPVs (Figure 2). In practice, in the atmosphere HIPVs and constitutively emitted VOCs cannot in most cases be separated from each other and the following description concerns the mixture of both. The following fates of VOCs and the corresponding functions are suggested:

Intact volatiles travel in air currents and facilitate the interactions detailed above

Production of HIPVs, many of which are highly reactive in the atmosphere, is an evolved response in many plant species under biotic stress. When considering the chemical and physical properties of the HIPV compounds, a longer atmospheric life time of the compounds could possibly attract natural enemies over longer distances. However, the molecular concentration in the atmosphere will in any case be rapidly reduced due to increased distance from the point source of a damaged plant. The selection of highly reactive compounds as attractive signals in indirect plant defense does not necessarily maximize the attraction capacity of HIPVs (McFrederick et al., 2009) or improve plant fitness (Hare, 2011). Therefore, additional potential traits which can be linked to production of reactive HIPVs may give an alternative explanation for the type of HIPV compounds that evolved. Conversely, Peñuelas and Llusia (2004) proposed that an increase in biological complexity of plant physiological processes during evolution is one of the causes of the diversity of VOC emissions and their emission is just an unavoidable trait as a result of their volatility. Volk (2003) also noted that feedback loops in the biosphere contain segments based solely upon by-products of organisms' metabolisms. According to Volk (2003) these were not metabolically evolved by organisms to be sent out into the environment for altering its chemistry and do not represent a trait that was selected during evolution by natural selection.

Increased emission of reactive HIPV compounds from plants under biotic stress will inevitably lead to SOA formation as shown experimentally (Joutsensaari et al., 2005) and in a modeling study in insect outbreak areas (Berg et al., 2012). SOA particles can grow by aggregation and by VOC vapor uptake to form cloud condensation nuclei (CCN), which will have an impact on light quality reaching vegetation, influence precipitation and affect the amount of sunlight reflected to space (Riipinen et al., 2012). The importance of SOA and other aerosols in controlling global temperature and radiation balance has been convincingly shown (Arneth et al., 2009). Recently, the results from large scale measurement campaigns (Paasonen et al., 2013) suggested that VOCs emitted from vegetation substantially increase sun-screening SOA formation under climate warming. This biogenic VOC based growth mechanism produces roughly 50% of particles at the size of cloud condensation nuclei across Europe (Paasonen et al., 2013). An important question that remains to be answered, is whether HIPV have evolved to act as a part of a biosphere-atmosphere feedback system that improves the abiotic growth conditions of the plants under attack? This question prompts a further question of whether a biogenic SOA related change in environmental conditions to indirectly better defend against biotic attackers comes anywhere close to the expenditure in terms of VOCs and carbon used to create the change? If the costs of a HIPV based biosphere-atmosphere feedback system are higher than the improvement in fitness of the HIPV emitting individuals, selection of the trait does not fall in line with natural selection in a strict Darwinian sense (Moody, 2012), except perhaps in clonal plants with a population distributed across large areas (De Woody et al., 2009).

Recent evolutionary models incorporate phenotypes expressed in the external environment; however, there is still debate whether such traits generate dynamics that alter evolution (Bailey, 2012). Such models of extended phenotypes predict that the individual carrying genes for the trait, which has an impact on the physical environment, may have effects on conspecifics including an emitting individual's own offspring or siblings, but also other species (Bailey, 2012). These kinds of extended effects are particularly created by species known as ecosystem engineers (Jones et al., 1994, 2010; Hastings et al., 2007) which by niche construction modify their own niche and those of other organisms e.g., through behavior or metabolic processes.

We suggest that high HIPV emission rate capacity of Boreal conifer forest trees could be such a trait and that it may have an impact on SOA formation in insect outbreak areas (Berg et al., 2012). Higher atmospheric SOA density and associated cooling effect could provide a stabilizing feedback (Lenton, 1998; Paasonen et al., 2013) to protect conifer ecosystems against factors such as global warming, insect outbreaks and spread of invasive deciduous tree species related to warming (Kellomaki et al., 2001). It is expected that current global warming will increase the frequency of forest pest outbreaks at higher latitudes (Niemelä et al., 2001), and warmer temperatures could substantially increase the HIPV emission rates of affected conifer trees (Heijari et al., 2011; Amin et al., 2012). However, it has not yet been shown if there really is a negative feedback loop (Lenton, 1998; Lovelock, 2003) related to insect outbreak areas, i.e., the cooling effect of HIPVs through enhanced albedo by HIPV-induced SOA formation (Berg et al., 2012), CCN formation (Paasonen et al., 2013) and finally by improved cloud albedo. This cooling feedback loop would reduce the frequency of forest pest outbreaks and relieve heat stress of vegetation. More efficiently dispersed light may improve photosynthesis, but cooling would counteract the light effect, thus reducing photosynthesis rate and additionally reducing VOC production and the protective role of HIPVs against pests and pathogens.

To return to our original question; in this review we have tried to demonstrate that the carbon fixed by a plant and then, particularly under biotic stress, released back to the atmosphere as volatile organic compounds, will have important roles in chemical, physical and biological processes during their life time. These may be facilitated by the HIPV compounds originally synthesized and emitted by a plant, or in the form of other chemical compounds after atmospheric reactions of the HIPVs. So far, we know only a fraction of the highly diverse potential routes and functions that the VOCs emitted by plants may have. For example, the carbon of highly volatile C10 monoterpenes and semivolatile C15 sesquiterpenes in the atmosphere could be bound to freshly nucleated SOA particles as a result of ozonolysis during the day time or oxidized by reactions with NO₃ at night. The secondary organic aerosol particles formed from the same HIPV may have different chemical composition and different biological functions depending on the time of day during SOA formation. The size of SOA particles in the atmosphere may also grow by adsorption of other organic vapors to their surface (Kroll and Seinfeld, 2008) or the particles could be influenced by solar UV irradiation leading to the photolysis and formation of several oxygenated C1–C3 compounds (Pan et al., 2009). The reactions could also be reversible and low molecular mass carbon compounds may react again to form larger carbon-based molecules. Finally the carbon bound to the HIPV compounds will be oxidized to CO or CO₂ in the atmosphere (Kroll and Seinfeld, 2008) and may again be utilized by plants in the process of photosynthesis, while it could alternatively end up as organic polymers in the sediments of terrestrial or aquatic ecosystems.