As mentioned above, PMs have become the most dangerous pollutants in some countries. Chemical species of PMs, derived from the available data over China included SO42-, NO3-, NH4+, organic carbon, and elemental carbon, which were in a range of 2.2–60.9, 0.1–35.6, 0.1–29.8, 1.5–102.3, 0.2–37.0 μg cm⁻³ in PM_2.5, and 1.6–104.6, 0.5–46.6, 0.2–31.0, 1.7–98.7, and 0.3–26.8 μg cm⁻³ in PM₁₀, respectively (Zhou et al., 2016). PM_2.5 is the major component of PM₁₀, accounting for 65%. PMs are also composed of microorganisms. In a study of PMs in Jeddah, Saudi Arabia (Alghamdi et al., 2014), the average concentrations of PM₁₀ and PM_2.5 were 159.9 and 60 μg cm⁻³, respectively and the concentrations of O₃, SO₂, and NO₂ averaged 35.73, 38.1, and 52.5 μg cm⁻³, respectively. Microbial loads were higher in PM₁₀ than PM_2.5. Aspergillus fumigatus and Aspergillus niger were the common fungal species associated with PMs. Microbes were also found in PMs in Austria (Haas et al., 2013), including fungi from genera Aspergillus, Cladosporium, and Penicillium and aerobic mesophilic bacteria. Using metagenomic methods, Cao et al. (2014) identified 1,315 distinct bacterial and archaeal species from 14 PM samples collected from Beijing, China. The most abundant phyla were Actinobacteria, Proteobacteria, Chloroflexi, Firmicutes, Bacteroidetes, and Euryarchaeota. Among them, an unclassified bacterium in the nitrogen fixing, filamentous bacteria genus Frankia was the most abundant, and the most abundant classified bacterial species appeared to be Geodermatophilus obscures. The abundance of airborne bacteria was reported to be in a range from 10⁴ to 10⁶ cells m⁻³ depending on environmental conditions (Bowers et al., 2011), and materials of biological origin might account for up to 25% of the atmospheric aerosol (Jaenicke, 2005). Ammonia oxidizing archaea (AOA), ammonia oxidizing bacteria (AOB), and complete ammonia oxidizers (Comammox) were identified in PM_2.5 collected from the Beijing-Tianjin-Heibei megalopolis, China (Gao et al., 2016). Of which Nitrosopumilus subcluster 5.2 was the most dominant AOA, Nitrosospira multiformis and Nitrosomonas aestuarii were the most dominant AOB, and the presence of Comammox was revealed by the occurrence of Candidatus Nitrospira inopinata. The mean cell numbers of AOA, AOB, and Ca. N. inopinata were 2.82 × 10⁴, 4.65 × 10³, and 1.15 × 10³ cell m⁻³, respectively. The average maximum nitrification rate of PM_2.5 was 0.14 μg (NH4⁺-N) [m³ air h]⁻¹ (Gao et al., 2016). AOA might account for most of the ammonia oxidation, followed by Comammox, while AOB were responsible for a small part of ammonia oxidation. The assay of nitrification activity was performed in laboratory conditions (Gao et al., 2016). However, the nitrification potential of such bacteria in PMs after being deposited on leaf surfaces is unknown. We hypothesize that the nitrification process could be more active once such PM-containing bacteria settled on leaves. Further investigation on nitrification of PM-associated bacteria in the phyllosphere could provide insight into how the phyllosphere could potentially act as manufactories in the nitrification of ammonia.

The current literature regarding phylloremediation of PMs has been primarily focused on plant leaves. Plant canopy is a sink for PMs. This is due to the fact that leaves are in the air and they span more than 4 × 10⁸ km² on a global scale, which is about 78.4% of the total surface area of the earth; leaves thus physically act as a natural carrier for PMs. Leaves differ greatly in surface structure and metabolic secreted substances as well as microbial composition. The amount of surface waxes and compositions show different capacity to retain and embrace PMs. Sbø et al. (2012) studied leaves of 22 trees and 25 shrubs in accumulation of PMs in Norway and Poland and found that PM accumulation differed by 10 and 15 folds depending on plant species in the two locations and also positive correlations occurred among PM accumulation, leaf wax contents, and leaf hair density. Thirteen woody species were examined by Popek et al. (2013) during a 3-year period, and total amount of PMs captured by leaves ranged from 7.5 mg cm⁻² by Catalpa bignonioides to 32 mg cm⁻² by Syringa meyeri. Leaf wax contents were significantly correlated with the amount of PMs on leaves. Among the PMs captured, 60% was washable by water, and 40% could be washed by chloroform only, suggesting that the PMs were embraced in waxes. Using two photon excitation microscopy (TPEM), Terzaghi et al. (2013) investigated leaves of stone pine (Pinus pinea), cornel (Cornus mas), and maple (Acer pseudoplatanus) in capture and encapsulation of PMs. The authors found that particles ranging from 0.2 to 70.4 μm were visualized on leaves, of which PM_2.6 was the dominant size across plant species. Particle less than 10.6 μm were encapsulated in the cuticle. Plant species differed in particle retention and encapsulation, which were attributed to leaf characteristics, cuticle chemical composition and structure.

Leaf physical characteristics such as leaf shape, hairs or trichomes, and stomata significantly affect PM accumulation. Needle leaves were reported to accumulate more PM_2.5 than broad leaves (Terzaghi et al., 2013; Chen et al., 2017). The effectiveness was attributed to the higher capture efficiency and higher Stoke's numbers of needles compared to those of broad leaves (Beckett et al., 2000). Additionally, small individual leaf area and abundant wax layer also contribute to the effectiveness (Chen et al., 2017). Leaf trichomes have been shown to increase PM_2.5 accumulation. The trichome density was positively correlated with amount of PM_2.5 accumulated on leaves, and plant species with abundant hairs, such as Catalpa speciosa, Broussonetia papyrifera, and Ulmus pumila were able to retain more PM_2.5 than those with fewer hairs (Chen et al., 2017). The adaxial surface of leaves accumulated more PMs than the abaxial leaf surface (Baldacchini et al., 2017), which is probably due to the fact that the abaxial surface in general has few trichomes and less rough surface. Stomata may play some roles in accumulation of PMs. The length of stomata ranges from 10 to 80 μm and densities varies from 5 to 1,000 mm⁻² depending on plant species and environmental conditions (Hetherington and Woodward, 2003). Stomatal pore areas range from 46 to 125 μm² (Peschel et al., 2003; Dow et al., 2014), thus stomata could retain or adsorb either PM_2.5 or PM₁₀. A study of PM deposition on leaves of five evergreen species in Beijing, China showed that PM diameter up to 2 μm was in the stomatal cavity (Song et al., 2015). Rai (2016) studied the effects of PMs on 12 common roadside plant species and found that stomatal sizes were reduced due to air dust deposition, but plant growth was not affected, suggesting the potential of plants in adsorbing air pollutants.

Growing evidence has suggested that plant leaves are able to capture PMs and act as biofilters. On average, the upper leaf surface of 11 plant species intercepted 1,531 particles per mm⁻² (Wang et al., 2006). Needles of Pinus sylvestris accumulated 18,000 mineral particles per mm² (Teper, 2009). Upper leaves of Hedera helix captured about 17,000 particles per mm² (Ottele et al., 2010). Trees removed 1,261 tons of air pollutants in Beijing, of which 772 tons were PM₁₀(Yang et al., 2005). In New Zealand, urban trees removed 1,320 tons of particular matter annually due to the existence of woodlands in Auckland (Cavanagh and Clemons, 2006). Nowak et al. (2014) showed trees within cities removed fine particles from the atmosphere and consequently improved air quality and human health. Tree effects on PM_2.5 concentrations and human health are modeled for 10 U.S. cities. The total amount of PM_2.5 removed by trees varied from 4.7 tons in Syracuse to 64.5 tons in Atlanta in the U.S annually. All the reported removal of PMs is attributed to plant leaves. It is unknown at this time if phyllosphere microbes could break down the PMs on leaves and if mineral elements released from the broken PMs could become plant nutrients. Considering the fact that the microbes can biodegrade a wide range of substances including petroleum, we hypothesize that some microbes should be able to break down PM. Future research in this regard will be conducted, and identified microbes could be used for PM reduction.

Sulfur dioxide (SO₂) was among the first air pollutants identified to harm human health and ecosystems. The combustion of fossil fuels has substantially increased SO₂ in the air. China has contributed to about one-fourth of global SO₂ emission since 1990 (Zhang et al., 2013). The emission of SO₂ from Guangdong province totaled 1,177 Gg in 2007, of which 97% was emitted by power plants and industries (Lu et al., 2010). SO₂ can be oxidized photochemically or catalytically to sulfur trioxide (SO₃) and sulfate (SO42-) in the air (Bufalini, 1971). With the presence of water, SO₃ is converted rapidly to sulfuric acid (H₂SO₄), which is commonly known as acid rain. While in sulfur assimilation, SO42- is reduced to organic sulfhydryl groups (R-SH) by sulfate-reducing bacteria, fungi, and plants. Sulfur oxidizing bacteria such as Beggiatoa and Paracoccus are able to oxidize reduced sulfur compounds like H₂S to inorganic sulfur, and thiosulfate to form sulfuric acid (Pokoma and Zabranska, 2015). Sulfate reducing bacteria like Archaeoglobus and Desulfotomaculum can convert sulfur compounds to hydrogen sulfide (H₂S). Oxidation of H₂S produces elemental sulfur (S°), which is completed by the photosynthetic green and purple sulfur bacteria and some chemolithothrophs. Further oxidation of elemental sulfur produces sulfate. Sulfate is assimilated through the sulfate activation pathway, which is consisted of three reactions: the synthesis of adenosine 5′-phosphorylation of (APS), the hydrolysis of GTp, and the 3′-phosphorylation of APS to produce 3′-phosphoadenosine 5′-phosphosulfate (PAPS) (Sun et al., 2005). In Mycobacterium tuberculosis, the entire sulfate activation pathway is organized into a single complex (Sun et al., 2005). Additionally, sulfate reducing bacteria have been shown to use hydrocarbons in pure cultures, which can be used for bioremediation of benzene, toluene, ethylbenzene, and xylene in contaminated soils (Muyzer and Stams, 2008). Such bacteria may also colonize leaf surfaces and could be used for remediation of air pollutants.

Plant leaves absorb SO₂ via stomata. At apoplastic pH, it is hydrated and oxidized successively to sulfite and sulfate, both of which can inhibit photosynthesis and energy metabolism if they accumulate to a high concentration. Such inhibition can cause SO₂ toxicity. Symptoms include interveinal chlorosis and necrosis in broad-leaved species, and chlorotic spots and brown tips in pine conifers (Rennenberg, 1984). Until the 1970s, SO₂ was considered to be a key contributor of acid rain causing forest dieback (Bloem et al., 2015). Interestingly, when the Clean Air Acts came into action in the 1980s, the reduction in atmosphere SO₂ resulted in sulfur (S) deficiency in crops, particularly Brassica species. The S deficiency was responsible for the increased incidence of disease caused by Pyrenopeziza brassicae (Bloem et al., 2015). The explanation is that plants could become injured in a SO₂ concentration range from 131 to 1,310 μg m⁻³; plants, however, can rapidly assimilate SO₂ and H₂S into reduced sulfur pools such as cysteine and sulfates as illustrated in Figure 1. A recent transcriptome analysis of Arabidopsis responses to SO₂ showed that plant adaptation to SO₂ evokes a comprehensive reprogramming of metabolic pathways including NO and reactive oxygen species (ROS) signaling molecules, and also plant defense response pathways (Zhao and Yi, 2014). The importance of this study revealed that plant responses to SO₂ stress is at the transcription level with initial activation of cross tolerance and followed by sulfur assimilation pathways. Cysteine metabolism in particular is associated with the network of plant stress responses, thus improving plant growth in soils where sulfur supply is limited (Bloem et al., 2015). It has been shown that an atmospheric level of 79 ng m⁻³ SO₂ could contribute to 10–40% of leaf sulfur assimilation (De Kok et al., 2007; Zhao et al., 2008). Elevated SO₂ concentrations around natural CO₂ springs have been documented to enhance accumulation of sulfur metabolites and proteins in surrounding vegetation (Rennenberg, 1984). Therefore, plants can be selected for growing in SO₂ polluted environments (Chung et al., 2010). In 2000, about 42.62 Mg of SO₂ was removed from the atmosphere by urban trees in Guangzhou, China (Zhang et al., 2013). Additionally, S metabolism can be genetically engineered for improving plant resistance to SO₂. Transgenic tobacco plants overexpressing cysteine synthase or serine acetyltransferase gene were highly tolerant to SO₂ and sulfite (Noji et al., 2001).

There are several oxides of nitrogen (N) in the atmosphere: nitrogen dioxide (NO₂), nitric oxide (NO), nitrous oxide (N₂O), nitrogen trioxide (N₂O₃), and nitrogen trioxide (N₂O₅). Among them, the USEPA regulates NO₂ only because it is the most prevalent form of NO_x generated anthropogenically (USEPA, 1999). NO₂ also participates in the formation of ozone (O₃) and NO. NO_x emissions in China increased rapidly from 11.0 Mt in 1995 to 26.1 Mt in 2010. Power plants, industry, and transportation were major sources of NO_x emissions, accounting for 28.4, 34.0, and 25.4% of the total NO_x emissions in 2010, respectively (Zhou et al., 2013). The total NO_x emissions in China are projected to increase 36% based on the 2010 value by 2030.

A group of bacteria like Azotobacter and Rhizobium and fungi such as mycorrhizas are capable of fixing atmospheric N. Cyanobacteria are able of using a variety of inorganic and organic sources of combined N, like nitrate, nitrite, ammonium, urea or some amino acids. These microbes are often associated with plant roots. Nitrifying bacteria including species from the genera Nitrosomonas, Nitrosococcus, Nitrobacter, and Nitrococcus oxidize ammonia to hydroxylamine, and nitrite oxidoreductase oxidizes nitrite to nitrate. Nitrifying bacteria thrive in soils, lakes, rivers, and streams with high inputs and outputs of sewage, wastewater and freshwater because of high ammonia content. Phyllosphere diazotrophic bacteria, like Beijerinckia, Azotobacter, and Klebsiella and also Cyanobacteria, such as Nostoc, Scytonema, and Stigonema can use atmospheric dinitrogen (N₂) as a source of nitrogen (Whipps et al., 2008). N₂ is fixed by the nitrogenase enzyme encoded by nif genes, and the gene nifH has been widely used for analysis of their community structure (Fürnkranz et al., 2008; Rico et al., 2014). The abundance of N₂-fixing bacteria was also reported to improve drought tolerance, suggesting their adaptability to plants grown in different environmental conditions (Rico et al., 2014).

Plants absorb gaseous NO₂ more rapidly than NO because NO₂ reacts rapidly with water while NO is almost insoluble. The uptake of NO₂ per unit leaf area was reported to be nearly three times that of NO when the two gases occurred in the same concentration (Law and Mansfield, 1982). As a result, NO₂ has been considered to be more toxic than NO. Visible symptoms resulting from NO₂ exposure are relatively large, irregular brown or black spots. However, phytotoxicity of NO₂ is rare and much less than SO₂ and O₃. This is due to the fact that NO_x are plant nutrients. When NO and NO₂ are absorbed and dissolved in the extracellular solution of leaves, they form nitrate (NO₃) and NO₂ in equal amounts and proton (H⁺). NO₃ is then utilized by plants in the same way as it is absorbed from roots and used as a nitrogen source for synthesizing amino acids and proteins (Figure 1). Foliar absorption of NO₂ varies widely depending on plant species. Morikawa et al. (1998) studied 217 herbaceous and woody species in uptake of NO₂ and found that plant species differed by 657 folds in NO₂ uptake and assimilation. The most efficient woody plants included Eucalyptus viminalis, Populus nigra, Magnolia kobu, and Robinia pseudoacacia, and the most herbaceous plants include Erechtites hieracifolia, Crassocephalum crepidioides, and Nicotiana tabacum (Morikawa et al., 1998).

Nitrogen dioxide could be a plant signal molecule that improves plant growth. Morikawa et al. (2004) reported that about one-third of NO₂-derived N absorbed by leaves was converted into a previously unknown Kjeldahl-unrecoverable organic nitrogen, which comprise a novel heterocyclic Δ2 1,2,3 thiadiazoline derivative and nitroso- and nitro-organic compounds (Miyawaki et al., 2004; Morikawa et al., 2005). These results indicate that NO₂ is not only known as a pollutant or a supplemental source of N, but also acts as an airborne reactive nitrogen species signal (Morikawa et al., 2004, 2005). This is in agreement with the reports that endogenously produced NO_x such as NO act as a vital plant signal (Wendehenne et al., 2001; Neill et al., 2003). To further analyze atmospheric NOx effects on plants, Morikawa et al. (2003) determined if plants could use NO₂ as a fertilizer and concomitantly reduce NO₂ concentrations. The authors found that application of 282 μg m⁻³ NO₂, equivalent to the heavily polluted urban air, to plants for 10 weeks almost doubled the biomass, total leaf area, the contents of carbon (C), N, S, phosphorus (P), potassium (K), calcium (Ca), and magnesium (Mg) as well as free amino acid contents and crude proteins (Morikawa et al., 2003). The mass spectrometric analysis of the ¹⁵N/¹⁴N ratio showed that N derived from NO₂ comprised less than 3% of total plant N, meaning that the contribution of NO₂-N to total N was relatively low. These results imply that NO₂ could be a multifunctional signal to stimulate plant growth, nutrient uptake, and metabolism (Takahashi et al., 2005).

Anthropogenic O₃ is primarily generated from the reaction of atmospheric O₂ with ground-state O (3P) radicals that result from the photolytic dissociation of ambient NO₂. Thus, the presence of NO and NO₂ in the lower atmosphere is closely linked with ground-level of O₃. In China, O₃ levels increased at a rate of 2.2 μg m⁻³ per year from 2001 to 2006. Average O₃ concentrations in Beijing varied from 45 to 96.2 μg m⁻³ depending on locations (Wan et al., 2014). In Shanghai, 1-h average concentration of O₃ was 54.2 μg m⁻³. O₃ level increased during spring, reached the peak in late spring and early summer, and then decreased in autumn and finally dropped in winter. The highest monthly average O₃ concentration (82.2 μg m⁻³) in June was 2.7 times greater than the lowest level (30.4 μg m⁻³) recorded in December (Zhao et al., 2015).

Ozone is considered an effective antimicrobial agent against some bacteria and fungi (Sharma and Hudson, 2008). There have been no reports on microbial-mediated O₃ reduction. However, in a study of O₃ effects on phyllosphere fungal populations, Fenn et al. (1989) found that a chronic exposure of mature Valencia orange trees (Citrus sinensis) to O₃ or SO₂ for 4 years decreased populations of phyllosphere fungi. In a same experiment conducted by the authors, a short-term fumigation of O₃ to giant sequoia (Sequoiadendron giganteum) and California black oak (Quercus kelloggii) did not significantly affect the numbers of phyllospere fungi. Plant absorption of O₃ is mainly through stomata, O₃ is easily dissolved in water and reacts with apoplastic structures and plasma membranes to form reactive oxygen species (ROS), such as O2-, H₂O₂, and OH radical. The O₃ or ROS can disturb cell membrane integrity and attack sulfhydryl (SH) groups or ring amino acids of protein, thus causing phytotoxicity. Injury symptoms include white, yellow or brown flecks on the upper surface of leaves. The threshold concentrations that cause a 10% reduction in yield are 80μg m⁻³ for sensitive crops and 150 μg m⁻³ for the most resistant crops. Adaptation of plants to O₃ stress has resulted in plants developing mechanisms against O₃ toxicity. First, O₃ can be removed from the air by chemical reactions with reactive compounds emitted by vegetation, particularly monoterpenes (Di Carlo et al., 2004). Second, semi-volatile organic compounds, such as different diterpenoids exuded by trichomes on leaves are an efficient O₃ sink (Jud et al., 2016). Tobacco leaves can secret diterpenoid cis-abienol, which acts as a powerful chemical protection shield against stomatal O₃ uptake by depleting O₃ at the leaf surface. As a result, O₃ flux through the open stomata is strongly reduced (Jud et al., 2016). As to O₃ absorbed by leaves, an oxidative burst occurs as the initial reaction to O₃, followed by activation of several signaling cascade and plant antioxidant systems including ascorbate-glutathione cycle and antioxidant enzymes to alleviate the oxidative burden resulting from O₃ exposure (Vainonen and Kangasjarvi, 2015).

VOCs are organic chemicals that have a low boiling point and a high vapor pressure at room temperature causing large numbers of molecules to evaporate into the surrounding air. VOCs are numerous and ubiquitous including naturally occurring and anthropogenic chemical compounds. VOCs participate in atmospheric photochemical reactions contributing to O₃ formation and also play a role in formation of secondary organic aerosols, which are found in PMs. The strong odor emitted by many plants consists of green leaf volatiles, a subset of VOCs called biogenic VOCs, which emit exclusively from plant leaves, the stomata in particular. Major species of biogenic VOCs include isoprene, terpenes, and alkanes.

Anthropogenic VOCs include large groups of organic chemicals, such as formaldehyde, polycyclic aromatic hydrocarbons (PAHs), and BTX (benzenes, toluene, and xylenes). The most significant sources of formaldehyde are engineered wood products made of adhesives that contain urea-formaldehyde (UF) resins. BTX come from painting and coating materials used for interior decoration and refurbishment. Motor-vehicle exhausts, tobacco smoke, and heating also contribute to the presence of VOCs. A great concern over VOCs has been indoor air quality. Indoor formaldehyde in recently renovated homes ranged from 0.14 to 0.61 mg m⁻³, and benzene, toluene, and xylenes were 124.0, 258.9, and 189.7 μg m⁻³, respectively (Hao et al., 2014). The formaldehyde concentration is 65–100% higher than indoor air quality standards of China. Formaldehyde and BTX as main indoor VOCs contribute to the so-called “sick building syndrome” (Brown et al., 1994; Wieslander et al., 1996; Wargocki et al., 2000; Berg et al., 2014). This review regarding VOCs is thus emphasized on indoor air quality.

As early as in the 1970s, NASA (U.S. National Aeronautics and Space Administration) conducted research on the use of foliage plants for remediation of air quality in space shuttles. Foliage plants are those with attractive foliage and/or flowers that are able to survive and grow indoors (Chen et al., 2005). Results showed that foliage plants removed nearly 87% of air pollutants from sealed chambers within 24 h (Wolverton et al., 1984, 1989; Cruz et al., 2014a). For example, each plant of peace lily (Spathiphyllum spp. ‘Mauna Loa’) removed 16 mg of formaldehyde, 27 mg of trichloroethylen, and 41 mg of benzene from sealed chambers after a 24-h exposure to the respective chemical. Generally, plants absorb gaseous pollutants via leaf stomata. Some of the VOCs are recognized as xenobiotics by plants, and they are detoxified through xenobiotic metabolism, involving oxidoreductase or hydrolases, bioconjugation with sugars, amino acids, organic acids, or peptides, and then removed from the cytoplasm for deposition in vacuoles (Edwards et al., 2011). In addition to plant leaves, rhizosphere microbes also contribute to reduction of VOCs under interior environments (Llewellyn and Dixon, 2011). Using a dynamic chamber technique, Xu et al. (2011) investigated formaldehyde removal by potted foliage plants and found that formaldehyde removal was attributed not only to the formaldehyde dehydrogenase activities of plant leaves but also to the absorption and metabolism by microorganisms in the rhizosphere. Such bacteria have been isolated from soils, water, and different tissues of plants in polluted environments. Many pure cultures of bacteria, including various strains of P. putida, have been evaluated for biodegradation of air pollutants. Some fungi strains are also able to use volatile aromatic hydrocarbons as sole source of carbon and catalyze degradation reactions (Prenafeta-Boldú et al., 2001; Kennes and Veiga, 2004; Jin et al., 2006). Here we mainly discuss phylloremediation of formaldehyde, benzene, toluene, and xylene as well as phenols and PAHS.

Plants should be selected from four categories: (1) trees, (2) shrubs or small tress, and (3) ground cover plants for use in outdoor environments as well as (4) foliage plants for indoor environments. Trees are referred to as perennial plants with elongated stems or trunks, supporting branches and leaves. Shrubs (or small trees) are those small to medium-sized woody plants that grow under some degree of shaded conditions. Ground covers are any plants that can grow over an area of ground and they can grow below the shrub layer including turfgrass and other woody and herbaceous selections. Foliage plants are those which can grow and survive indoors for interior decoration.

Plant species not only differ greatly in adsorption, absorption, and assimilation of air pollutants but also vary significantly in pollution tolerance. Air pollution tolerance index has been used for evaluation of plants specie in response of pollutants (Singh et al., 1991). Information generated by the index is useful, but the index may require revision for better reflecting the ability of plants in tolerance of air pollutants. An initial large-scale evaluation of plants from the four categories should be conducted for identifying candidate species that are able to tolerate PMs, O₂, SO₂, NO_x, and VOCs individually or collectively and can also substantially retain or assimilate these pollutants. Plants should also tolerate abiotic stresses, such as drought, heat, and cold, and biotic stresses like plant pathogens. Leaves of plants should be able to support one or more selected microbes. Trees should have a relatively fast growth rate. Needle-leaved plants should be particularly considered. As mentioned before, needles are rich in waxes for capturing PMs, and they are also used as as passive bio-samplers to determine polybrominated diphenyl ethers (Ratola et al., 2011). Broad-leaved plants should have more hairs or trichomes and more stomata with a large canopy. Leaf water and nutritional contents, leaf cuticular wax composition, hairs or trichomes, and surface physical characteristics should be suitable for microbial colonization. Shrubs and ground cover plants should have similar leaf physical and chemical properties but be able to tolerate slight shade. For foliage plants, they should substantially tolerate shade and can survive and grow under indoor low-light conditions.

Plant species possessing the aforementioned traits should be selected from particular regions where plants survive and thrive under heavily polluted environments. The rationale is that plants that are able to grow in the polluted environments may develop mechanisms for adaptation to the stressful conditions. Thus, some regions of China and India could be ideal locations for initial selection of plant species. Plants have been documented to tolerate multiple stresses, which include induced cross tolerances and the ability of particular variants to resist multiple distinct stresses. Reactive oxygen species are key molecular signals produced in response to multiple stresses, which are aimed at the maintenance of cellular equilibrium (Perez and Brown, 2014). Glutathione-S-transferase (GST) genes play an important role in the maintenance of ROS equilibrium. Salicylic acid, jasmonic acid, and ROS interplay in the transcriptional control of multiple stresses. Additionally, omics technologies should be used for identifying molecular mechanisms in regulation of plant responses to multiple stresses. Such information, particularly transcriptional factors, key regulatory genes or enzymes should be incorporated into the plant selection processes.

Genetic engineering is an option for improving plants to remediate air pollutants (Abhilash et al., 2009). Genes listed in Table 4 can be used for generating transgenic plants. Cysteine synthase is a key enzyme to utilize H₂S and SO₂ as a sulfur source to synthesize cysteine. Overexpression of cysteine synthase in rice was shown to enhance sulfur assimilation upon exposure to a high level of H₂S (Yamaguchi et al., 2006). Nitrite reductase catalyzes the six-electron reduction of nitrite to ammonium. Transgenic Arabidopsis plants bearing chimeric spinach NiR gene enhanced nitrite reductase activity and NO₂ assimilation (Takahashi and Morikawa, 2001). Cytochrome P450 2E1 has strong and specific capacity of decomposing organic pollutants in animal bodies. Transgenic tobacco plants overexpressing CYP2E1 gene showed increased ability to detoxify broad classes of pollutants such as chlorinated solvents and aromatic hydrocarbons (James et al., 2008). Unlike tobacco, poplar (Populus tremula × Populus alba) plants are a fast-growing tree species with large canopies. Poplar plants overexpressing a mammal CYP2E1 exhibited increased metabolism and enhanced removal of organic pollutants from hydroponic solution and the air (Doty et al., 2007). Some genes from microbes can also be used for engineering transgenic plants for phylloremediation. The ribulose monophosphate (RuMP) pathway is one of the formaldehyde-fixation pathways found in microorganisms (Orita et al., 2006). The key enzymes of this pathway are 3-hexulose-6-phosphate synthase (HPS), which fixes formaldehyde to D-ribulose 5-phosphate (Ru5P) to produce D-arabino-3-hexulose 6-phosphate (Hu6P) and 6-phospho-3-hexuloisomerase (PHI), and then converts Hu6P to fructose 6-phosphate (F6P) (Orita et al., 2006; Chen et al., 2010). Co-expression of HPS and PHI in tobacco plants resulted in 20% reduction of formaldehyde compared to the control plants (Chen et al., 2010). In another study, a chlorocatechol 1,2-dioxygenase gene (tfdC) derived from the bacteria Plesiomonas was introduced into Arabidopsis thaliana (Liao et al., 2006). Transgenic plants showed enhanced tolerances to catechol, an aromatic ring. Transgenic plants were also able to remove a large amount of catechol from their media and highly efficient in convertion of catechol to cis, cis-muconic acid, suggesting that degradative genes derived from microbes can be used to produce transgenic plants for bioremediation of aromatic pollutants in the environment (Liao et al., 2006).

Selected plants should be evaluated in controlled environmental chambers to measure their capacity for tolerance and also assimilation of air pollutants. Seedlings could be exposed to particular pollutants or a mixture of pollutants in different concentrations and durations. Plant responses to the exposures could quickly evaluated based on stomatal conductance, net photosynthetic rate, the maximum quantum efficiency of photosystem II using the new LI-COR6800. Their morphological appearance, i.e., leaf greenness, leaf size, and plant height and canopy dimension compared to control treatments should be evaluated. The ability of plants to remove pollutants should be tested using GC-MS. For evaluation of plant responses to PM, in addition to the mentioned plant characteristics, leaf morphology, particularly leaf surface characters should be examined under microscopes and stomatal size and density recorded. If needed, isotopic labeling techniques could be used to track the fate of particular compounds. The evaluation results once analyzed and compared, plants that tolerate stresses and are able to adsorb or absorb or assimilate pollutants could be identified from each type of plants for subsequent compatiablity tests with selected microbes.

Cultivable bacteria only account for a small fraction of the total diversity in the phyllosphere, which has greatly hampered the use of some valuable microbes. New approaches, such as the use of improved culture and advanced devices (i-Chip), co-culture with other bacteria, recreating the environment in the laboratory, and combining these approaches with microcultivation should be employed to convert more uncultivable bacteria into cultured isolates in the laboratory (Nichols et al., 2010; Stewart, 2012; Müller and Ruppel, 2014). Similar to plant selection, initial microbial selection could be carried out in areas where plants have been contaminated by air pollutants. In coordination with plant selection, microbes could be isolated from leaves of plants identified in plant selection. This is because the pollutants may exert selective pressures to phyllosphere microbial diversity. For example, bacterial communities hosted by Platanus × acerifolia leaves from different locations of Milan (Italy) were analyzed by high throughput sequencing. The results showed that biodiversity of bacterial communities decreased but hydrocarbon-degrading populations increased along the growing season, which suggest that air contaminants might play an important role in the selection of phyllospheric populations in urban areas (Gandolfi et al., 2017).

A particular attention should be given to endophytic microbes. There are about 300,000 plant species on the earth; each plant could host one or more endophytes (Petrini, 1991; Strobel and Daisy, 2003). Endophytes are resided inside plant tissues and generally have no harmful effects on plants. Endophytic bacteria that colonize leaves could be particularly desirable as they could not be washed away by precipitation. Recent advances in endophyte-assisted remediation have been reviewed (Khan and Doty, 2011; Stepniewska and Kuzniar, 2013; Ijaz et al., 2016; Syranidou et al., 2016). Endophytic B. cereus ZQN5 isolated from natural Zamioculcas zamiifolia leaves enhanced ethylbenzene removal rate on sterile Z. zamiifolia (Toabaita et al., 2016). Microbes could also be isolated from the rhizosphere of plants contaminated by air pollutants as more endophytism occurs in roots (Ijaz et al., 2016). Some of leaf endophytes could be initially established in roots and subsequently transported to shoots. Khaksar et al. (2016a) reported that some microbes isolated from roots can also colonize leaf surfaces. An endophytic strain of B. cereus ERBP from roots of Clitoria ternatea was able to colonize the leaf surface of Z. zamifolia. During a 20-d fumigation with formaldehyde, the inoculation of ERBP did not interfere with the natural shoot endophytic community of Z. zamiifolia. ERBP inoculated Z. zamiifolia exhibited a significantly higher formaldehyde removal efficiency when compared to the non-inoculated plants.

Microbes, once identified and cultured, could be engineered to improve phylloremediation capacity (Table 5). A pTOM toluene-degradation plasmid from B. cepacia G4 was introduced into Bacillus cepacia L.S.2.4, a natural endophyte from yellow lupine (Lupinus arboreus; Barac et al., 2004). After the engineered bacteria were inoculated into aseptic lupine seedlings, the recombinant endophytics degraded 50–70% more toluene and provided much more protection against the phytotoxic effects of toluene than that obtained from soil bacteria (Barac et al., 2004). Horizontal genes can transfer among plant-associated endophytic bacteria in plants. Poplar was inoculated with the yellow lupine endophyte B. cepacia VM1468, which contains the pTOM-Bu61 plasmid coding for constitutively expressed toluene degradation (Taghavi et al., 2005). Inoculated plant growth was enhanced in the presence of toluene, and the amount of toluene release via evapotranspiration was also reduced. Although no inoculated strains were detected in the endophytic community, there was horizontal gene transfer of pTOM-Bu61 to different members of the endogenous endophytic community (Taghavi et al., 2005). The TCE-degrading strain P. putida W619-TCE also can be engineered via horizontal gene transfer in poplar plants (Weyens et al., 2009b).

Efforts on microbe selection should also be placed on the identification of microbes that could remediate PM, SO₂, NO₂, and O₃. As mentioned above, a group of microbes can assimilate SO₂ and NO₂, further research should explore those microbes for effective assimilation of the two pollutants. Thus far, it appears that no information is available regarding microbial remediation of PM and O₃, which may not be the case in the nature. Extensive research should be conducted to determine if nature has offered microbes that can break down PMs and can also biodegrade or biotransform O₃.

Selected microbes could be domesticated by growing them in different cultures varying in pH, carbon source, temperature, and O₂ to identify appropriate culture media and conditions for maximizing their growth. Morphological characterization and internal transcribed spacer rDNA analysis should be conducted to determine their phylogenetic relationships with other microbes. Their ability to biodegrade particular or a group of air pollutants should be evaluated in the laboratory. Microbial characteristics including their utilization of organic compounds, decomposition rate of pollutants, adaptability, competition, and growth rate should be recorded and analyzed. Competitive strains that show promise in bioremediation should be identified. A series of bacterial and filamentous fungal genomes have been sequenced recently. More than hundreds of bacterial and fungal transcriptomic and proteomic datasets are available. With the advent of increasingly sophisticated bioinformatics and genetic manipulation tools, mechanisms underlying the biodegradation or transformation of pollutants by the isolated microbes could be elucidated. This information, in turn, will significantly improve our understanding of the microbes and provide us with molecular bases for manipulation of the microbes for enhancing phylloremediation.

Plants selected from the four categories should be inoculated with selected microbes to determine the compatibility of each selected microbe with each selected plant species. The test could begin first in laboratory settings using entire leaves in designated chambers or utilizing young seedlings in relative large growth chambers to evaluate if inoculated microbes could grow on leaf surfaces and if the specific inoculation affects plant growth. Compatible combinations would be exposed to pollutants at different concentrations and durations to determine the potential for pollutant reduction. A microbe that is compatible with one plant species may not be compatible with another. For example, B. cereus ERBP isolated from roots of C. ternatea was compatible with the leaf surface of Z. zamifolia but not with the leaf surface of Euphorbia milii. ERBP-colonized Z. zamifolia grew well and showed high efficiency in removal of formaldehyde, but ERBP-colonized E. milii were less effective in removal formaldehyde and the plants exhibited stress symptom (Khaksar et al., 2016a). Laboratory evaluation will generate a large number of plant-microbe combinations that are specifically effective in removal of a particular pollutant or a particular group of pollutants. Bacteria would be propagated using bioreactors and corresponding plants would be propagated through either cuttings or tissue culture. The plants would be transplanted into greenhouses or specific regions with air pollution for testing the effectiveness of the combinations in real-world situations.

Plants and microbe combinations that pass the real-world test will be investigated using the next-generation sequencing (NGS) technologies (metagenomics, metatranscriptomics, metaproteomics, and metabolomics) and the rapid evolution of SIP (Stable isotope probing) for identifying molecular mechanisms underlying microbial and plant interactions in facilitation of phylloremediation. The compatibility evaluation and molecular analysis would ultimately result in the development of protocols for culturing microbes and producing corresponding plants. Some protocols will be catered to trees, others used for shrubs or small trees. Some would be effective for improving groundcover plants, and some will be used for indoor foliage plants. Effectiveness of each protocol in remediation of particular or general pollutants would be determined using the model described by Nowak et al. (2006). If the test is to be conducted in a large scale, satellite image acquisition and analysis should be used. The analysis of the data will finally validate the protocols, i.e., particular plants can be inoculated with a specific group of microbes for use in remediation of a particular pollutant or a mixture of pollutants.

The protocols will be implemented for phylloremediation. We propose three types of plantscape: (1) manufactory plantscape, (2) urban plantscape, and (3) interior plantscape. The plantscape for manufactories and cities should have three levels of greening: the sky with trees, the ground with groundcover plants, and shrubs in between. Additionally, climber plants can be used to build green walls and small trees and shrubs as well as groundcovers can be used to build green roofs. For interior plantscape, each room should have a minimum of one potted foliage plant. Foliage plants can also be used to install green walls in interior environments for enhance remediation of indoor air pollutants.

The implementation of phylloremediation technologies should also take landscape design concepts into consideration, resulting greenbelts, green parks, green walls that fulfill roles not only for air remediation but also for recreation. Depending on the occurrence of pollutants and the scale and degree of the overall pollution, relevant protocols to the particular situations would be implemented. The remediation efficiency could be monitored over time using specific models in connection with satellite imagine data to determine how much of individual pollutants have been removed.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.