A whiff of the future: functions of phenylalanine-derived aroma compounds and advances in their industrial production

Aroma and humans

Aroma compounds are involved in our everyday lives. Smelling allows us to evaluate our surroundings and is used for hazard avoidance, reproduction, and social interactions. Food is a major aspect of the human–volatile relationship as aroma compounds convey the most accurate information on a food's availability and quality (Fig. 1a). Perhaps, the most important aspect is the contribution of aroma compounds to foods' flavour, and it is no wonder humans have been using them for this purpose since ancient times (Schwab et al., 2008). For example, residues of vanillin were identified in ceramic juglets in Tel Megiddo, Israel, dating back ca. 3500 years (Linares et al., 2019). The connection between food and aroma goes even deeper as aroma compounds can shape humans' food preference and even influence their eating behaviour (Rolls et al., 1996; Small and Green, 2011). For example, aroma intensity of vanilla in a custard dessert affected food intake, with higher intensity resulting in smaller bite size (de Wijk et al., 2012). In addition, aroma compounds are suggested to possess health benefits; Phe-derived compounds like vanillin and cinnamaldehyde have demonstrated anticarcinogenic properties, and RK aids in weight loss (Ahn et al., 2015; Ayseli and Ipek Ayseli, 2016; King et al., 2007; Morimoto et al., 2005). Due to their immense contribution, Phe-derived compounds, for example, eugenol (clove), cinnamaldehyde (cinnamon), phenylacetaldehyde (rose), benzaldehyde (almond), RK (raspberry) and of course vanillin (vanilla), are highly desired in the food and beverage and pharma industries. The specific rise in demand for ‘natural’ aroma compounds of biological origin has led several large food companies to pledge complete removal of artificial flavours from their products (Bomgardner, 2016). This has ignited extensive research into finding platforms for the production of these compounds (see Microbes and plants as platforms for aroma compound production). In addition, in crops that are consumed fresh, such as tomato, basil, and strawberry (Fan et al., 2021; Tieman et al., 2017; Walters et al., 2021), much effort is being made to identify flavour-contributing Phe-derived aroma compounds and the levels that appeal to consumers. Integrating knowledge on the interplay between humans and aroma molecules may enable generating healthier and tastier foods and crops to improve humans' well-being.

Aroma and pollinators

As sessile organisms, plants are dependent on external factors for sexual reproduction; ca. 87.5% of angiosperms are pollinated by animals, including birds, mammals and reptiles, with insects being the dominant taxa (Ollerton, 2017). Flowers advertise themselves to pollinators by various cues, olfactory ones playing a major role (Fig. 1b; Joffard et al., 2020). Phe-derived volatile compounds are widespread among floral blends, and benzaldehyde, methyl salicylate (MeSa), benzyl alcohol, and 2-phenylethanol are the most common (Knudsen et al., 2006). These aroma compounds strongly contribute to specificity in plant–pollinator interactions. For example, loss of benzaldehyde emission in Capsella underlies the transition from animal-mediated pollination to selfing (Jantzen et al., 2019; Sas et al., 2016). In moth-pollinated petunia, loss of benzaldehyde production led to a shift in pollinator to hummingbirds, which depend largely on their eyesight (Amrad et al., 2016). Directed evolution by bumble bee pollination of Brassica rapa led to increased floral emission of benzyl nitrile (Ramos and Schiestl, 2019), another Phe-derived volatile (Liao et al., 2020b). From a human perspective, animal pollination is especially important because it is responsible for 35% (by volume) of crop production (Klein et al., 2007). Perils such as global environmental changes present a tangible threat to volatile-mediated plant–pollinator mutualism (Byers and Chang, 2017; Glenny et al., 2018). A standing question is whether advanced breeding techniques can be harnessed to improve (or at least maintain) pollination efficiencies to overcome future problems (Dudareva et al., 2013). Success has been extremely limited so far, as numerous complexities still need to be addressed. For example, floral bouquets are typically comprised of complex blends of numerous volatiles originating from various biochemical pathways. Increasing the metabolic flux into multiple distinct branches belonging to often competing pathways is extremely complicated, especially when they are negatively correlated, compete for the same substrates, or affect other agriculturally important traits (Iijima et al., 2004). Furthermore, pollinators often perceive blends as separate from the components that make up the blend (Wycke et al., 2020). Understanding pollinator-specific perceptual characteristics of components and blends is especially important when planning which compound/s to target. Solving these pending problems is a prerequisite for success in this endeavour.

Aroma as defence signals within and between plants

In addition to pollination, plants also make use of volatile compounds as part of their defence mechanism, and they are used in both intra- and inter-plant communication (Fig. 1c,d) (Ninkovic et al., 2020; Vlot et al., 2020). Phe-derived aroma compounds, such as eugenol, vanillin and guaiacol, have been shown to possess antimicrobial properties (Fitzgerald et al., 2004; Friedman et al., 2002; Walsh et al., 2019; Wang and Fan, 2014). The volatile methyl ester of salicylic acid (SA), MeSA, is a signalling phytohormone that upon conversion to bioactive SA, regulates or activates resistance mechanisms (Park et al., 2007; Shulaev et al., 1997; Vlot et al., 2009). MeSA is a mobile signal that has been shown to travel through the phloem from infected to uninfected tissues, enabling activation of systemic acquired resistance (Park et al., 2007). In addition, MeSA is suggested to act as an airborne signal between neighbouring plants; MeSa produced in response to viral infection of tobacco plants activated defence mechanisms in its healthy, uninfected neighbours (Shulaev et al., 1997). Although some possible pathways have been suggested, the exact mechanisms of volatile signal perception in plants remain to be investigated (Ninkovic et al., 2020). Furthermore, MeSA has been shown to serve indirectly against herbivores by recruiting their natural predators (Ament et al., 2010; Rowen et al., 2017).

Generating plants with modified emission levels of MeSA and other defence volatiles to improve the defence efficiency and priming of plant populations against pathogens and herbivores seems like a logical goal. Nevertheless, tampering with the defence mechanism might be a double-edged sword because predicting all layers of volatile-mediated plant–plant and plant–pathogen interactions is extremely complex. For instance, synthetic lures containing MeSA placed in tomato fields improved resistance to both pathogens and herbivores, but increased their susceptibility to colonization by thrips (Rowen et al., 2017). Taken together, although the ability to generate tailor-made metabolically engineered plants with superior volatile-based defence mechanisms or alert systems is an attainable goal, a better understanding of the many aspects of plant volatiles' interactions with the environment is required.

Thus, future goals of harnessing Phe-derived aroma compounds for human needs are not an easy task, because these volatiles are entangled in many of the plant's interactions with the environment that are vital to humans, pollinators, and plants. Hence, the engineering approach to reach these goals needs to be more ‘holistic’ and less ‘atomistic’, due the many layers of aroma function. For instance, MeSA and guaiacol are disliked by consumers and in tomatoes, reducing their levels has been suggested for flavour improvement (Tieman et al., 2017). Yet, lowering their levels might compromise the plants and their population defence and priming mechanisms, rendering the plants more susceptible to pathogens and herbivores. MeSA is also a prevalent compound in floral bouquets with a proven role in pollinator attraction (Hoballah et al., 2005; Knudsen et al., 2006), hence perturbing its synthesis might also have adverse effects on pollination efficiency. Taken together, it is clear that future strategies for metabolic engineering of aroma compounds in plants should ideally consider integration of flavour, pathogen and herbivore resistance, pollination efficiency and more.

Phenylpropanoid pathway

The phenylpropanoid pathway produces myriad metabolites involved in numerous processes of utmost importance in the plant's life cycle, such as lignin, flavonoids, and aroma compounds (Table 1; Vogt, 2010). These metabolites are produced via an intricate and strongly interlinked network of enzymes and regulators. The precursors of the pathway are the aromatic amino acids (AAAs). Biosynthesis of AAAs begins in the plastid-localized shikimate pathway. In the first step, the enzyme 3-deoxy-D-arabino-heptulosonate-7-phosphate (DAHPS) condenses the pentose-phosphate-pathway-derived erythrose-4-phosphate (E4P) and glycolysis-derived phosphoenolpyruvate (PEP) (Fig. 2). Further enzymatic steps lead to the formation of the pathway's final product—chorismate. Chorismate mutase 1 (CM1) subsequently converts chorismate to prephenate, from which Phe may be biosynthesized in plants mainly via the arogenate pathway (Lynch and Dudareva, 2020). However, Phe biosynthesis is not limited to the plastids; as in microbes, it may also occur in the cytosol via the phenylpyruvate pathway with the aid of chorismate mutase 2 (CM2) (Qian et al., 2019). Phenylpropanoid production begins with Phe being directed to two branches mediated by competing enzymes, yielding four aroma compound classes based on their carbon backbone: benzenoids (C6–C1), phenylpropanoid-related compounds (C6–C2), phenylpropenes (C6–C3), and phenylbutanones (C6–C4) (Fig. 2). Metabolic flow towards benzenoids, phenylpropenes, and phenylbutanones is mediated by the action of L-phenylalanine ammonia-lyase (PAL), which in the committed step deaminates Phe to trans-cinnamic acid (CA). This constitutes a main transition point of carbon from primary (amino acid) metabolism to phenylpropanoid-specialized metabolism. Benzenoids production occurs via the peroxisome-localized β-oxidative and/or cytosolic non-β-oxidative pathways (Widhalm and Dudareva, 2015). CA conversion to p-coumaric acid by cinnamate 4-hydroxylase (C4H) followed by multiple enzymes, including CoA ligases like 4-coumarate-CoA ligase (4CL), directs metabolism toward phenylpropenes, phenylbutanones, lignins, and flavonoids. Recently, a bifunctional Phe/L-tyrosine ammonia-lyase (PTAL), which also converts tyrosine to p-coumaric acid, was identified in Brachypodium distachyon (Barros et al., 2016; Barros and Dixon, 2020). In addition to route through PAL, Phe may be channelled towards the production of C6–C2 phenylpropanoid-related aroma compounds via different pathways (Fig. 2), as revealed in tomato, petunia, rose and poplar (Farhi et al., 2010; Günther et al., 2019; Hirata et al., 2016; Kaminaga et al., 2006; Tieman et al., 2006).

Vanillin and raspberry ketone biosynthesis

Biosynthesis of the phenylbutanone RK [4-(4-hydroxyphenyl)butan-2-one], which defines the aroma of ripe raspberry fruit, has been fully elucidated (Fig. 2) (Abe et al., 2001; Koeduka et al., 2011). CA is converted to p-coumaric acid by C4H, a step also shared with vanillin production, followed by the action of 4CL to yield p-coumaroyl-CoA. In the penultimate step, benzalacetone synthase (BAS) condenses p-coumaroyl-CoA and malonyl-CoA, generating 4-hydroxybenzalacetone. Reduction of 4-hydroxybenzalacetone to RK is catalysed by raspberry ketone/zingerone synthase 1 (RZS1).

Formation of vanillin (4-hydroxy-3-methoxybenzaldehyde), the main aroma compound in cured extract of Vanilla planifolia pods (Gallage and Møller, 2018), has been suggested to occur via several possible routes. Similar to C6–C3 compounds, p-coumaric acid is the precursor for vanillin (Fig. 2) (Kundu, 2017). Chemically, vanillin is a phenolic aldehyde belonging to the C6–C1 group because it undergoes C2 side-chain shortening (Kundu, 2017). This latter reaction is suggested to be catalysed by the vanillin synthase recently isolated from V. planifolia (VpVAN), directly from ferulic acid (Gallage et al., 2014, 2018; Kundu, 2017). However, VpVAN's ability to directly generate vanillin was questioned by Yang et al. (2017), who reported its involvement in vanillin biosynthesis through conversion of p-coumaric acid to p-hydroxybenzaldehyde (Kundu, 2017). The debate continues, with studies by Chee et al. (2017) and Gallage et al. (2018) further suggesting VpVAN's activity via ferulic acid, although detailed enzymological data, such as kinetics, are lacking. For comprehensive reviews of Phe-derived volatile biosynthesis, we suggest Vogt et al. (2010), Muhlemann et al. (2014) and Widhalm et al. (2015).

Regulation of Phe-derived aroma compound biosynthesis and emission

Microbes

Microorganisms, most commonly the bacterium E. coli and the yeast Saccharomyces cerevisiae, have been widely used as heterologous systems for the production of numerous specialized metabolites. Microbes often lack the complete biochemical pathways for biosynthesis of plant aroma compounds. Therefore, the main approach for metabolic engineering relies on heterologous expression of enzymes to complete the necessary chemical steps, complemented with mutation or deletion of specific genes to direct metabolite flow towards the target while suppressing competing metabolic shunts and preventing unwanted modifications to the product. Microbes have many advantages for metabolic engineering, such as rapid growth, secretion of metabolites to the media for easy harvest, simple transformation procedures and inexpensive carbon and nitrogen sources (Chemler and Koffas, 2008; Miralpeix et al., 2013; Yang et al., 2020)). Another major advantage of the production of plant metabolites in microbes is that competing branches are often non-existent, and there is less chance of feedback inhibition by non-target downstream products, for example, flavonoids in the case of PAL (see earlier). Moreover, both European and US food legislation classifies microbial-produced aroma compounds as GRAS (generally recognized as safe) (Food and Drug Administration, 2018; Kallscheuer, 2018).

Industrially applicable fungi and bacteria have been successfully metabolically engineered to generate vanillin from different substrates, such as ferulic acid, eugenol and isoeugenol, reaching yields of >10 g/L (Gallage and Møller, 2015). Microalgae can also produce vanillin, albeit at low levels, via biotransformation with ferulic acid (Tripathi et al., 2002). Nonetheless, glucose and other carbon sources are considered more ideal substrates as they are readily utilized, non-toxic, and inexpensive (Gallage and Møller, 2015). Vanillin production from glucose was achieved in E. coli harbouring a mutated shikimate dehydrogenase (aroE) combined with modifications that allowed bypassing the phenylpropanoid pathway entry points that exist in plants. However, the modified E. coli was only able to generate vanillic acid and the reduction to vanillin was performed in vitro (Fig. 3a) (Li and Frost, 1998). To eliminate the costly and complicated in vitro reaction, yeast with aromatic carboxylic acid reductase (ACAR) and phosphopantetheinyl transferase (PPTase), necessary for ACAR activation, were generated. This created the complete route for vanillin production from glucose, yielding 65 and 45 mg/L vanillin in Schizosaccharomyces pombe and S. cerevisiae (Fig. 3b), respectively (Hansen et al., 2009). The ACAR-based strategy also proved effective for heterologous production from glucose of the aroma compounds cinnamaldehyde and cinnamyl alcohol (Gottardi et al., 2017). However, in this case, the authors ectopically expressed PAL from Arabidopsis to generate CA from Phe (instead of dehydroshikimic acid path), enabling de novo biosynthesis of the target molecules. Similarly, Ni et al. (2015) introduced a synthetic vanillin pathway into E. coli aimed at mimicking the one in plants, to avoid diversion of substrates from the shikimate pathway. They used tyrosine ammonia-lyase, which converts tyrosine directly to p-coumaric acid and not PAL, to overcome the low activity of C4H in E. coli due to the absence of a suitable NADPH reductase (Fig. 3a) (Watts et al., 2004). The constructed pathway allowed the biogenesis of vanillin with media containing tyrosine, reaching a titre of 97.2 mg/L. Furthermore, the modified bacterium could also utilize glucose, xylose, and glycerol (Fig. 3a) (Ni et al., 2015).

To construct a pathway for de novo biosynthesis of RK in yeast, PTAL, with the ability to use both Phe and tyrosine as substrates, was expressed together with plant C4H, 4CL, and BAS (Fig. 3b) (Lee et al., 2016). There was no need to add the enzyme responsible for the final step, benzalacetone reductase (BAR), as the yeast demonstrated endogenous activity (Beekwilder et al., 2007). Maximum RK yield was 2.81 mg/L when 4CL and BAS were physically fused, an approach suggested to allow substrate funnelling (Aalbers and Fraaije, 2019; Albertsen et al., 2011; Lee et al., 2016). Indeed, multiprotein complexes known as ‘metabolons’ have been identified in plants, including PAL and C4H (Achnine et al., 2004; Zhang and Fernie, 2021). However, fusion of PTAL and C4H in yeast led to an inverse effect, greatly reducing the final yield of RK (Aalbers and Fraaije, 2019; Lee et al., 2016). To our knowledge, de novo RK biosynthesis has not been achieved in E. coli; however, substantial amounts have been obtained by growing E. coli engineered with 4CL, BAS and raspberry ketone/zingerone synthase 1 (RZS1) on fatty acids, reaching a titre of 180.94 mg/L (Fig. 3a) (Chang et al., 2021).

The importance of the shikimate and AAA pathways as the primary source for downstream metabolites has established them as a key target for metabolic engineering (Averesch and Krömer, 2018; Peled-Zehavi et al., 2015). Both pathways are tightly regulated, with significant bottlenecks due to feedback inhibition of two key enzymes—DAHPS and CM—by their products, the AAAs (see Regulation of Phe-derived aroma compound biosynthesis and emission). Use of feedback-insensitive forms of these enzymes has proven successful in alleviating these bottlenecks in microbes and plants (Graf and Altenbuchner, 2014; Liu et al., 2019; Luttik et al., 2008; Oliva et al., 2017; Tzin et al., 2012, 2013; Xie et al., 2016). This approach was also taken for the de novo biosynthesis of vanillin in microbes. Li and Frost (1998), Kunjapur et al. (2014) and Ni et al. (2015) all introduced a feedback-insensitive DAHPS into their modified E. coli for improved carbon flux. In yeast, strains overexpressing an insensitive DAHPS (Aro4) exhibited a three-fold increase in AAAs, together with an increase in the Phe-derived aroma compounds 1-phenylethanol and phenylacetate (Liu et al., 2019; Luttik et al., 2008). Combining Aro4 with an insensitive CM (Aro7) further increased flux towards the AAA, manifested by enhanced production of p-coumaric acid, a precursor of both vanillin and RK (Liu et al., 2019). E4P of the pentose phosphate pathway is another limiting factor in microbes, because DAHPS has low affinity to it and its endogenous pool is low. To overcome this, Liu et al. (2019) expressed two phosphoketolases to divert the carbon flux from glycolysis towards E4P. Indeed, combining phosphoketolases together with Aro4 and 7 further increased the levels of p-coumaric acid (Liu et al., 2019).

One of the problems in attempting to enhance metabolite production in microbes is conversion of the target molecule to unwanted side products. More specifically, as microbes rapidly convert aldehydes to alcohols, vanillin and its precursor protocatechualdehyde may be converted to the unwanted by-products, such as vanillyl alcohol and protocatechuic alcohol respectively. A rational approach to overcoming this problem is to delete or knock out the genes encoding the enzymes with this activity. In this way, Kunjapur et al. (2014) generated E. coli with 55 times higher titres of vanillin (119 mg/L) compared with E. coli without deletion of the respective reductases (Fig. 3a). These strains are also capable of accumulating the benzenoid benzaldehyde, as its reduction to benzyl alcohol was also decreased (Kunjapur et al., 2014). Although successful in this case, deletion or silencing of ‘unwanted’ genes is not always an applicable strategy as these may be involved in other pathways that are crucial for the organism's survival. For example, in S. cerevisiae, an enoyl reductase that is involved in the biosynthesis of fatty acids also converts p-coumaroyl-CoA (pCA) to phloretic acid, resulting in carbon loss when attempting to enhance phenylpropanoid (i.e., flavonoid) production (Lehka et al., 2017). To avoid pCA conversion while preserving fatty acid production, two alternative approaches were explored: a mutation in the active site of enoyl reductase responsible for phloretic acid production, and incorporation of a plant gene that lacks this activity. The latter approach was more effective in preventing carbon loss (Lehka et al., 2017).

One may think that when it comes to enhancement of aroma compounds, the goal is always ‘the more the merrier’. Nevertheless, aroma compounds in general, and vanillin and RK in particular, become toxic to cells at certain threshold concentrations (Beekwilder et al., 2007; Fitzgerald et al., 2004; Friedman et al., 2002; Hansen et al., 2009). Vanillin was shown to inhibit growth of S. cerevisiae cells at concentrations above 500 mg/L (Hansen et al., 2009). Similarly, vanillin demonstrated bacteriostatic properties in E. coli via disruption of the membrane, ultimately impairing respiration (Fitzgerald et al., 2004); thus, the rapid reduction of vanillin to the by-product vanillyl alcohol in that organism is probably a detoxification step (Graf and Altenbuchner, 2014). Interestingly, membrane disruption was also reported in petunia cells in which the Phe-derived volatile release mechanisms were impaired (Adebesin et al., 2017; Liao et al., 2020a). Another approach to lowering a compounds' toxicity is by modification, such as methylation or glycosylation. Indeed, in Vanilla pods, vanillin is stored as vanillin β-D-glucoside, and S. cerevisiae cells grown on media supplemented with vanillin β-D-glucoside at 25 g/L (50-fold more than vanillin) do not exhibit impaired growth (Hansen et al., 2009). To obtain glycosylation of vanillin in yeast, a glycosyltransferase from Arabidopsis was incorporated. Indeed, it successfully converted almost all of the vanillin to this less toxic water-soluble form, yielding 500 mg/L of vanillin β-D-glucoside (Fig. 3b) (Brochado et al., 2010; Hansen et al., 2009). Nonetheless, when the desired product is the aglycone, and not the glycoside, another step is required to break the glycoside residue.

The studies described in this section demonstrate the progress achieved in metabolic engineering of bacteria and yeast for the production of the commercially important aroma compounds vanillin and RK. These advances are due to the improved understanding of metabolic fluxes in host microbes, together with identification of specialized genes from different organisms. Perhaps, the best proof of this progress is the fact that biologically derived vanillin generated from glucose by microbial fermentation is already being sold (Waltz, 2020).

Plants

With our growing understanding of the metabolic fluxes within and between target pathways, their feedback loops, wide-scale functional gene analyses, big data analyses, tools for directed evolution of target enzymes, etc., plants are an attractive system for the production of target metabolites. This is further strengthened by the progress that has been made in the application of precise genome-editing tools (Bortesi and Fischer, 2015; Pott et al., 2019; Yang et al., 2014). Moreover, plants have some inherent advantages over microorganisms as platforms for the enhancement of endogenous pathways or as hosts for heterologous ones (Owen et al., 2017). From an economic standpoint, the production process in plants is autotrophic and more cost-effective than in microbes, scale up is relatively easy, and there is no need for culture media or special growth facilities (Fu et al., 2018). Moreover, as flowers and fruit are, in many cases, the end product, they present an interesting target for metabolic engineering to improve food flavour and nutritional qualities (Ashokkumar et al., 2020; Hoffmann et al., 2011; Lobato-Gómez et al., 2021; Yu et al., 2021).

When dealing with Phe-derived pathways, plants present the advantage of possessing all of the basic components of the target pathway, as opposed to microbes. Thus, expression of dedicated transcription factors for the activation of multiple genes in a pathway(s) is a routine approach in plant metabolic engineering (Koeduka et al., 2021; Skaliter et al., 2019; Zhang et al., 2015; Zvi et al., 2008, 2012). In some cases, the genetic components are not actively producing target products because the substrate is missing, a phenomenon known as ‘silent metabolism’ (Fang et al., 2021; Hoffmann et al., 2011; Lewinsohn and Gijzen, 2009). For example, lisianthus flowers that are normally unscented produced the Phe-derived methyl benzoate and 2-phenylethanol when precursor availability was increased (Fang et al., 2021). Another advantage is that the final biosynthetic steps of many specialized metabolites are not fully detailed, and plants offer the possibility to complete these uncharacterized reactions (Dong et al., 2012; Jacobowitz and Weng, 2020; Zhou et al., 2018). Another main advantage of plants, as eukaryotic organisms, over microbes, is the presence of organelles for example, improved sequestration or accumulation of target molecules, or storage of toxic compounds. On the other hand, until recently, there was no efficient way to achieve targeted plant genome modifications, while growth of transgenic plants and the use of products extracted from them are restricted by complex regulations.

As the pathway/enzymes leading to vanillin production have yet to be fully characterized, hot chili pepper (Capsicum frutescens) is an example of a potential target for vanillin production, since vanillin is a precursor of capsaicins (pungency factor in hot pepper) (Chee et al., 2017; Gururaj et al., 2012). Pepper is highly recalcitrant to transformation, and attempts to produce vanillin in pepper have focused on work with calli. Silencing of the aminotransferase responsible for transamination of vanillin to vanillyl amine (the direct precursor of capsaicin) in pepper calli led to increased levels of the former (Fig. 4a) (Gururaj et al., 2012; Weber et al., 2014). Ectopic expression of VpVAN following particle bombardment of chili pepper callus led to substantially enhanced vanillin production (Fig. 4a); transformed calli generated ca. 573.39 μg/g fresh weight (FW) of vanillin, which translates to a 190-fold increase compared with non-trasnformed calli (Fig. 4a). Levels of vanillin β-D-glucoside also increased ca.16-fold (110.32 μg/g tissue), revealing an additional pool that might be tapped in the future. It would be interesting to combine both approaches, that is, aminotransferase silencing and VpVAN expression. VpVAN was also tested in Nicotiana benthamiana and barley, leading to the accumulation of vanillyl alcohol glucoside (Fig. 4b; Gallage et al., 2014).

An alternative approach to vanillin production based on Pseudomonas fluorescens 4-hydroxycinnamoyl-CoA hydratase/lyase (HCHL), which converts the lignin precursor feruloyl-CoA to vanillin in two steps, was employed in beet (Beta vulgaris) hairy root cultures (Fig. 4c; Gasson et al., 1998; Singh et al., 2015). Vanillin yield was similar to that in pepper callus, reaching 520 μg/g FW. Feruloyl-CoA is formed from ferulic acid by 4CL (Barros et al., 2019). Supplementation of ferulic acid to hairy root lines enhanced vanillin production to 2090 μg/g FW, reaching the yield of the vanilla orchid and revealing ferulic acid as a major production bottleneck. However, ferulic acid is an expensive substrate, rendering this approach cost-ineffective (Gallage and Møller, 2015). Ectopic expression of HCHL was also attempted in tobacco, which synthesizes vanillin naturally, albeit in trace amounts (Kundu, 2017; Mayer et al., 2001). Transgenic HCHL-expressing plants accumulated increased amounts of the vanillin derivatives, that is, vanillic acid and vanillyl alcohol (Fig. 4d). However, the alterations in metabolic flux resulted in multiple phenotypic abnormalities, such as phloem fiber deformation. These results demonstrate the negative ramifications of perturbing crucial metabolic pathways (Mayer et al., 2001). Vanillin is also a component of scent bouquets of flowers (Knudsen et al., 2006; Oliva et al., 2015, 2017; Skaliter et al., 2021). In petunia flowers, production of vanillin and its derivatives was enhanced by introducing an activator of the phenylpropanoid pathway—the Arabidopsis transcription factor PRODUCTION OF ANTHOCYANIN PIGMENT1 (PAP1) (Fig. 4e) (Skaliter et al., 2019; Zvi et al., 2008). Similarly, the strategy taken in microbes of alleviating metabolic bottlenecks by harnessing the feedback-insensitive enzymes AroG and PheA, bacterial DAHPS and CM/prephenate dehydratase, respectively, also yielded enhanced levels of vanillin in petunia (Fig. 4e; Oliva et al., 2015, 2017).

In all of the above studies in plants, a single genetic element was used. However, as shown in microbes, due to the complex nature of biochemical pathways leading to aroma compound production, to optimize carbon channelling, multiple genes need to be targeted. Thanks to advances in synthetic biology, multigene expression is now a routine practice in plants—it is no longer restricted to microbes. This strategy of using gene combinations proved successful in strawberry. Expression of petunia isoeugenol synthase (IGS) in strawberry fruit did not affect the concentration of either t-anol or isoeugenol, which was nearly zero (Hoffmann et al., 2011). However, when in addition to IGS expression, flavonoid biosynthesis was blocked via suppression of the branchpoint enzyme CHS, a substantial increase in both compounds (ca. 400 ng/g FW) was achieved (Cna'ani et al., 2014; Hoffmann et al., 2011; Oliva et al., 2017; Zvi et al., 2008). Recently, Koeduka et al. (2021) also successfully implemented multigene expression for RK production in plants (Fig. 4f). Ectopic expression of BAS together with RZS1 in tobacco led to the accumulation of ca. 2 μg/g FW of RK and its glycoside. However, when metabolic flow in these BAS/RZS1-expressing plants was enhanced via suppression of chalcone synthase (CHS), which competes with BAS for the substrate pCA, or by expression PAP1, a further three-fold increase in RK production was obtained in the tobacco flowers, similar to levels in raspberry fruit (Koeduka et al., 2021). Coexpression of PAP1 with BAS/RZS1 enabled RK biosynthesis in the leaves as well, representing the plant's main biomass, probably through activation of phenylpropanoid pathway genes by PAP1 (Koeduka et al., 2021; Skaliter et al., 2019; Zvi et al., 2008).

The strategy of integrating numerous genetic elements, opening new routes and blocking unwanted fluxes (Fu et al., 2018) has yet to be used for vanillin engineering in plants. This strategy can be applied to suitable plant hosts, such as tobacco and pepper, and logically, to the vanilla orchid, once transformation and regeneration of this latter plant become available. The vanilla orchid is the most efficient natural producer of vanillin, despite the long pollination, growth and curing processes; and the recent publication of its genome (Hu et al., 2019) makes its use especially promising. Once it is possible to apply gene-editing techniques, such as CRISPR to Vanilla, it will be possible to direct the flux towards phenylpropenes by, for example, knocking out negative regulators, such as the vanilla ortholog of PhMYB4 and CHS, and enhancing total metabolic flux in the phenylpropanoid pathway by introducing feedback-insensitive AroG and PheA with transcription factors, such as EOBI/II, ODO1, PAP1, and AtMYB12 (Fig. 5; Boersma et al., 2022; Fu et al., 2018; Oliva et al., 2015, 2017; Peled-Zehavi et al., 2015; Spitzer-Rimon et al., 2010, 2012; Zhang et al., 2015). However, overall, despite the potential of multigene combinations, they are not always straightforward (Tzin et al., 2015; Xie et al., 2016), and the crosstalk between different transgenes and its effect on metabolic flow should be carefully monitored and assessed.