The molecular biology of fruity and floral aromas in beer and other alcoholic beverages

Higher alcohols

Higher alcohol production by yeast occurs through the Ehrlich pathway either from amino acids transported over the cell membrane or through de novo biosynthesis of amino acids and their α-ketoacid intermediates. The Ehrlich pathway includes three steps: (1) deamination of the amino acid to an α-ketoacid, (2) decarboxylation and (3) reduction of aldehyde to alcohol by aldehyde reductase activity (Hazelwood et al.2008). The most important substrates for beer flavor are the branched-chain amino acids leucine (yielding isoamyl alcohol, 3-methylbutanol), isoleucine (yielding active amyl alcohol, 2-methylbutanol) and valine (yielding isobutanol, 2-methylpropanol). An overview of the metabolic pathways involved in higher alcohol and ester production is shown in Fig. 5.

Open multimedia modal

De novo synthesis of higher alcohols starts with the common metabolic intermediate pyruvate originating from glycolysis. During the formation of higher alcohols, the mitochondrial branched-chain isoleucine-leucine-valine (ILV) pathway metabolizes pyruvate into α-ketoisovalerate through acetolactate synthase (Ilv2), acetohydroxyacid reductoisomerase (Ilv5) and dihydroxyacid dehydratase (Ilv3) (Kohlhaw 2003). Next, α-isopropylmalate synthase (Leu4 and Leu9) condenses the α-ketoisovalerate produced in the mitochondria with acetyl-coA and H₂O to yield α-isopropylmalate. Leu4 exist in two isoforms, the full length isoform of Leu4 is targeted to the mitochondria, while the short cytosolic isoform of Leu4 is transcribed from a downstream start codon lacking the N-terminal mitochondrial target signal (Kohlhaw 2003). Biosynthesis of α-isopropylmalate is feedback inhibited through the C-terminal regulatory domain of Leu4 by leucine and free coenzyme A (mediated by Zn²⁺) (Satyanarayana, Umbarger and Lindegren 1968; Kohlhaw 2003; Koon, Squire and Baker 2004; Oba et al.2014). This inhibitory effect can be abolished by selecting for resistance to the toxic leucine analog 5,5,5-trifluoro-DL-leucine, which forces a higher flux through de novo synthesis (Casalone et al.1997; Cavalieri et al.1999). These resistant strains produced 3- to 4-fold more isoamyl alcohol in a saké fermentation (Ashida et al.1987), a 2-fold increase in wine fermentations (Casalone et al.1997), and a 2- to 3-fold increase in low alcohol beer produced with resistant S. pastorianus strains leading to a distinct banana flavor (Strejc et al.2013). In a similar fashion, strains that acquire resistance to DL-thiaisoleucine also produce higher levels of both active amyl alcohol and propanol from α-ketoglutarate (Fukuda et al.1993).

Pyruvate is also a precursor for acetaldehyde and ethanol production and for acetyl-coA, which is used as a substrate by Leu4, via the pyruvate dehydrogenase (PDH) complex and the PDH bypass pathway (Pronk, Steensma and Van Dijken 1996; Boubekeur et al.1999^, 2001; Krivoruchko et al.2015). Under anaerobic conditions, the E3 subunit of the mitochondrial PDH complex (encoded by the PDA1 gene) is inhibited by the reductive (high) NADH/NAD⁺ ratios, and the cells depend on the cytosolic PDH bypass pathways to provide the acetyl-coA (van Rossum et al.2016). The PDH bypass consists of the pyruvate decarboxylase (PDC), acetaldehyde dehydrogenase (ALD) and the acetyl-coA synthase genes ACS1 and ACS2 (Pronk, Steensma and Van Dijken 1996). The ACS1 gene is highly induced during growth with non-fermentable carbon sources and under aerobic conditions, and repressed by glucose, leaving ACS2 to provide the cytosolic acetyl-coA under anaerobic and high sugar conditions present in alcoholic fermentations (Krivoruchko et al.2015). Frick and Wittmann (2005) observed that cytosolic flux into acetyl-coA was even higher during respirative growth (increased biomass) than during fermentation. They hypothesized that the PDH bypass therefore not only occurs as an anabolic reaction to provide cytosolic acetyl-coA, but also to provide acetyl-coA for the mitochondria. The formation of acetyl-coA from pyruvate is compartmentalized, as the molecule is impermeable to the inner membrane of the mitochondria, and transport of acetyl-coA therefore occurs via the carnitine shuttle (Strijbis and Distel 2010). Jouhten et al. (2008) used a controlled supply of oxygen to investigate the formation of acetyl-CoA under fermentation conditions, and similarly found that the cytosolic flux of acetyl-coA was increased in more oxidative conditions. Admittedly, they did not supplement the media with L-carnitine (which cannot be synthesized by S. cerevisiae), and did therefore not include the inward flux into the mitochondria (Jouhten et al.2008). There appear to be very low levels of L-carnitine in barley seeds (Panter and Mudd 1969), which suggests that intramitochondrial production is sufficient for production of ‘normal’ levels of higher alcohols. Fluctuations in acetyl-coA levels will be discussed later, as this is also a substrate for acetate ester biosynthesis.

The second and third steps in isoamyl alcohol biosynthesis are carried out in the cytosol by isopropylmalate isomerase and β-isopropylmalate dehydrogenase encoded by LEU1 and LEU2, respectively, yielding the direct precursor of leucine, α-ketoisocaproate.

In the Ehrlich pathway, α-ketoisocaproate is produced directly by deamination of leucine by branched-chain amino acid transferase (BCAATase). There are two BCAATases involved, Bat1, located in the mitochondria, and Bat2, located in the cytosol (Eden, Simchen and Benvenisty 1996). They are encoded by paralogous genes that are expressed in an opposite manner during growth. BAT1 is highly expressed during logarithmic phase and repressed during stationary phase, while BAT2 is expressed during stationary phase and repressed during logarithmic phase (Eden, Simchen and Benvenisty 1996). Both gene products are involved in the Ehrlich pathway, and deletion or overexpression of BAT1/2 has a pronounced impact on the concentrations of higher alcohols and the overall aroma profile produced by S. cerevisiae (Eden et al.2001; Styger, Jacobson and Bauer 2011; Styger et al.2013). However, their role seems to be highly dependent on medium composition as Bat2 acted as the major BCAAT in laboratory yeast fermenting in synthetic medium (Eden et al.2001; Styger, Jacobson and Bauer 2011; Styger et al.2013), while Bat1 also affected production of higher alcohols in model wine fermentations (Lilly et al. 2006b; Rossouw, Naes and Bauer 2008). Overexpression of either resulted in yeast with a fruitier fermentation profile (Lilly et al. 2006b; Rossouw, Naes and Bauer 2008).

After deamination of the amino acid or de novo synthesis of the equivalent α-keto acid, they are decarboxylated and reduced to the higher alcohols. For isoamyl alcohol, α-ketoisocaproate is catabolized by the Thi3/Kid1 decarboxylase and to a lesser extent by the Aro10 aromatic decarboxylase into isoamyl aldehyde, which is subsequently reduced to isoamyl alcohol with aldehyde reductase activity of several different dehydrogenase enzymes (Adh1, 2, 3, 4 and 5, Sfa1) (Dickinson et al.1997; Dickinson, Salgado and Hewlins 2003). A similar pathway has been shown for isobutanol and active amyl alcohol with decarboxylation also carried out by the pyruvate decarboxylases Pdc1, Pdc5 and Pdc6 (Dickinson, Harrison and Hewlins 1998; Dickinson et al.2000). Saccharomyces cerevisiae strains engineered with co-compartmentalized branched chain amino acid and Ehrlich pathways in the mitochondria showed a significant increase in isobutanol production (Avalos, Fink and Stephanopoulos 2013). This clearly illustrates the crucial role of the mitochondria in the production of important beer aroma compounds.

Using radiolabeled [C13] valine, Oshita et al. (1995) found that when valine was abundant, it was catabolized to isobutanol, while when its levels were scarce, the majority of isobutanol was biosynthesized de novo as by-product of the valine anabolism pathway. In both cases, the level of isobutanol at the end of the fermentation was similar. Thus, the lack of amino acids in the medium can be compensated for by biosynthesis of the substrates feeding into the Ehrlich pathway. For isoamyl alcohol, the source from which it is made depends both on the level of free leucine in the wort and the glycolytic state of the yeast. The latter was proposed to be determined by the deamination step carried out by BCAATases that require the amino-acceptor α-ketoglutarate and NADH co-factor, which are produced in high levels during glycolysis (Espinosa Vidal et al.2015). While the production rate of isobutanol was increased upon oxygenation, isoamyl alcohol production does not appear to be affected significantly by oxygen availability (Espinosa Vidal et al.2015).

2-Phenylethanol. Production of 2-phenylethanol (‘rose’ and ‘honey’ aromas) occurs via catabolism of L-phenylalanine transported from the medium into the cell and via de novo production of L-phenylalanine and intermediates in the Shikimate pathway (Fig. 5). The Shikimate pathway commences with erythrose-4-phosphate derived from the pentose phosphate pathway and phosphoenolpyruvate derived from glycolysis. They are converted into 3-deoxy-D-arabino-heptulosonate-7-phosphate (DAHP) via DAHP synthase. There are two isoforms of DAHP synthase, encoded by the ARO3 and ARO4 genes. The shikimate pathway further consists of the multifunctional enzyme Aro1 that carries out the five subsequent reactions leading to 3-enolpyruvyl-shikimate-5-phosphate, the phosphate group is removed by chorismate synthase (carbon-oxygen lyase) Aro2 to produce the final metabolite in the pathway, chorismate (Braus 1991). Chorismate is further processed to anthranilate for tryptophan biosynthesis or to prephenate by chorismate mutase, Aro7, for production of phenylalanine and tyrosine. Prephenate is converted into phenylpyruvate by the prephenate dehydratase Pha2 or into 4-hydroxyphenylpyruvate by the prephenate dehydrogenase Tyr1. The aromatic amino acid transferases, Aro8 and Aro9, carry out the subsequent amination that forms the end products, phenylalanine and tyrosine. The transamination reaction carried out by Aro8 and Aro9 is reversible and can therefore also mediate breakdown of phenylalanine to phenylpyruvate. The subsequent steps in the Ehrlich pathway towards 2-phenylethanol production include decarboxylation of phenylpyruvate into phenylacetylaldehyde, mainly carried out by aromatic decarboxylase Aro10 and dehydrogenation by alcohol dehydrogenase (Fig. 5). Overexpression of the branched chain amino acid transferase genes BAT1 and BAT2 could only weakly influence 2-phenylethanol levels (Lilly et al. 2006b). This indicates that ARO8 and ARO9 are the major genes involved in deamination in the Ehrlich pathway. ARO9 and ARO10 are controlled at the transcriptional level by the aromatic amino acid biosynthesis regulator Aro80 and the amino acid starvation sensitive kinase Gcn2 (Staschke et al.2010; Lee and Hahn 2013), through mechanisms that involve both TORC1 signaling and binding of uncharged tRNAs to Gcn2 (Conrad et al.2014). Polygenic mapping of genes conferring high 2-phenylethyl acetate (and 2-phenylethanol) formation in beer production has been used to identify superior mutations in TOR1 and FAS2 (Trindade de Carvalho et al.2017). The superior TOR1 allele for 2-phenylethyl acetate production contained a nonsense mutation (E216*) that results in formation of a truncated and supposedly inactive Tor1 kinase, which is consistent with the Aro80-dependent induction of ARO9 and ARO10 upon exposure to rapamycin (Staschke et al.2010; Lee and Hahn 2013). The yields obtained are therefore dependent on the type of nitrogen source and the concentration of aromatic amino acids in the fermentation medium (Etschmann et al.2002; Vuralhan et al.2005). Surprisingly, however, ARO8 deletion enhances 2-phenylethanol production under ammonium-repressed conditions by increasing expression of ARO10 (Romagnoli et al.2015), encoding the major phenylpyruvate decarboxylase (Vuralhan et al.2003^, 2005). This underscores the high importance of ARO10 for 2-phenylethanol production.

The three aromatic amino acids, phenylalanine, tyrosine and tryptophan, exert feedback inhibition on multiple enzymes of the pathway. The phenylalanine-dependent DAHP synthase, Aro3, is sensitive to feedback inhibition by phenylalanine, while the tyrosine-dependent DAHP synthase, Aro4, is feedback inhibited by tyrosine (Fukuda, Watanabe and Asano 1990; Etschmann et al.2002; Tzin et al.2012; Zhang et al.2014). In a similar fashion to the LEU4 mutants in the leucine biosynthesis pathway, the inhibitory effects exerted on Aro3 and Aro4 by the end products, phenylalanine and tyrosine, can be released by selecting for resistance to fluorinated phenylalanine analogs (Fukuda, Watanabe and Asano 1990; Fukuda et al.1990). These mutants produced three to six times elevated levels of 2-phenylethanol in Sake fermentation trials. The fluorinated phenylalanine analogs p-fluoro-DL-phenylalanine and o-fluoro-DL-phenylalanine have also been used for selection of ARO4 and TYR1 mutants that produced up to 20-fold more 2-phenylethanol and 2-phenylethyl acetate than the wild-type wine yeast (Cordente et al. 2018). The isolated ARO4 mutants contained mutations in specific amino acid residues that release Aro4 from suppression by tyrosine (Hartmann et al.2003) leading to higher levels of the three aromatic amino acids intracellularly, whereas mutants in TYR1 accumulate phenylalanine extracellularly and show reduced intracellular tyrosine formation, indicating a lower activity of the prephenate dehydrogenase (Cordente et al.2018). The TYR1 mutants showed the highest increase in 2-phenylethanol levels, suggesting that the flux from prephenate to phenylpyruvate and the release of feedback inhibition on Aro4 by lower intracellular tyrosine levels are determining factors for the production of 2-phenylethanol.

Ester biosynthesis in yeast

Acetate ester biosynthesis in yeast. The word ‘ester’ was introduced probably as a reference to ethyl acetate, acetic ether, or in German: essig-ether (Gmelin 1848). In its most basic form, ethyl acetate is formed by a dehydration reaction between acetic acid and ethanol with the loss of a water molecule, or degraded into acetic acid and ethanol in hydrated acidic or alkaline conditions (Riemenschneider and Bolt 2000). Nevertheless, early studies of ester formation in brewing conditions concluded that acetate ester formation during brewing was not simply due to spontaneous esterification of acetate and alcohols, but instead involved enzymatic activity of S. cerevisiae through the energy rich thioester acetyl-coA (Nordström 1961^, 1962; Howard and Anderson 1976). This led to believe that acetyl-coA, an important metabolite in many pathways in living cells, was the determining factor for their production. However, it was observed that factors such as addition of unsaturated fatty acids and provision of oxygen, which inhibit acetate ester production, do not lower the levels of acetyl-coA or change the affinity constants for acetate synthase enzyme(s), highlighting possible involvement of other factors, such as the availability or expression level of the AATase as a major factor for their production (Malcorps et al.1991).

Surprisingly, most acetate esters are formed by the activity of a single alcohol acetyl-coA transferase (AATase) enzyme, encoded by the ATF1 gene (Fujii et al.1994; Fujii, Yoshimoto and Tamai 1996; Verstrepen et al. 2003d). Because of the importance of the ATF1 gene for the aroma profile of beer, saké and wine, its regulation during fermentation has been investigated in detail. The gene is expressed at low levels in brewing yeast with a peak at 12–36 h into the beer fermentation, when the yeast is still actively growing and fermenting under anaerobic conditions (Verstrepen et al.2003c). Beer made with lager yeast engineered with a constitutively overexpressed ATF1 gene contained four times more isoamyl acetate and six times more ethyl acetate compared to a wild-type strain and showed a very distinct fruity (‘banana’ and ‘pineapple’) and solvent-like character (Verstrepen et al. 2003). Thus, the limiting factor for ester production is not only the level of the alcohols, but to an even greater extent the expression level of the AATase enzyme. Interestingly, all the amyl esters in banana are formed during the ripening phase after ethylene treatment (Marriott 1980; Engel, Heidlas and Tressl 1990), which is accompanied by increased flux through glycolysis, increased levels of leucine and isoleucine (together with the higher alcohols), and induced expression of the banana AATase BanAAT gene (Seymour 1993; Jayanty et al.2002), thus indicating that fruit ripening potentially has some resemblance with the rapid formation of isoamyl acetate by yeast during fermentation. Also, in addition to the major AATase gene, ATF1, S. cerevisiae and S. pastorianus contain a paralogous gene, ATF2, which plays a minor role in the production of acetate esters during fermentation. Overexpression of ATF2 in a laboratory yeast with a low basal level of AATase activity causes a significant increase in the production of isoamyl acetate and ethyl acetate (Verstrepen et al. 2003d), while its overexpression in wild-type beer, saké and wine yeasts has more subtle effects (Verstrepen et al. 2003c; Lilly et al. 2006a; Sahara et al.2009). This led to the proposal of using an ATF2 overexpression strain for saké production in order to avoid production of overly solvent-like and banana-like flavors (Sahara et al.2009). Recently, an ethanol acetyl-coA transferase, Eat1, has been isolated from Wickerhamomyces anomalus (Kruis et al.2017). Saccharomyces cerevisiae contains a mitochondrial EAT1 ortholog (YGR015C), which has been shown to enhance ethyl acetate production when overexpressed in aerobic conditions. Polygenic analysis of ethyl acetate production in the absence of the ATF1 gene led to discovery of natural variants of EAT1 and SNF8 having a mutation that causes an early termination of translation (eat1^K179fs and snf8^E148*), and which lower ethyl acetate production (Holt et al. 2018a). However, modification of three unrelated yeast strains (wine, sake and ale yeast) showed highly strain-dependent effects of the two mutations, which underscores the importance of evaluating such mutations in the proper strain background, both for breeding and genetic engineering.

Substrate availability. The affinity constant (Km) of the major AATase (Atf1) for isoamyl alcohol (measured in partially purified cell extracts) was 25 mM (∼2204 mg/L) (Malcorps and Dufour 1992), which is roughly 20-fold higher than the average substrate level. This substrate limitation is consistent with the fact that there is generally a good correlation between isoamyl alcohol and isoamyl acetate levels during the course of a yeast fermentation (Calderbank and Hammond 1994; Quilter et al.2003). However, the variability in acetate ester production cannot only be explained by the availability of the higher alcohols (or ethanol), and the influence of acetyl-coA has been poorly explored in S. cerevisiae. The Atf1 enzyme contains transmembrane domains which fix the protein in the membrane of lipid droplets that bud out from the endoplasmic reticulum (ER), whereas Atf2 is localized in the membranes of the ER (Verstrepen et al.2004; Tiwari, Koffel and Schneiter 2007). Therefore, both enzymes depend on the cytosolic pool of acetyl-coA, which is generated by the PDH bypass (discussed previously for higher alcohols). Acetyl-coA metabolism has been engineered in Escherichia coli with significantly improved isoamyl acetate production (Vadali et al.2004; Vadali, Bennett and San 2004).

The cytosolic level of acetyl-coA is regulated by phosphorylation of acetyl-CoA carboxylase Acc1 by Snf1/AMPK kinase, which inactivates the enzyme and limits the production of malonyl-coA from acetyl-coA and CO₂, resulting in higher levels of acetyl-coA (Zhang, Galdieri and Vancura 2013). The enhanced acetyl-coA pool goes hand in hand with increased acetylation of histones and other acetylated proteins. During initiation of growth, the acetyl-coA levels increase from ∼3 to ∼30 μM in exponentially growing cell cultures, which might represent the metabolic state of the cell in acetylation/deacetylation reactions (Cai and Tu 2011; Cai et al.2011; Weinert et al.2014). The Atf1 enzyme has an affinity (Km) for acetyl-coA of 25 μM (Malcorps and Dufour 1992), but since the cytosolic pool of acetyl-coA has not been measured in the course of alcoholic fermentation, it remains unclear whether fluctuations in its concentration significantly affect AATase productivity.

ATF1 orthologs in Saccharomyces sensu stricto species and S. pastorianus. An S. eubayanus ortholog of the ATF1 gene (designated Lg-ATF1) is also present in the S. cerevisae/eubayanus hybrid lager yeast S. pastorianus. The activity of the Lg-ATF1 encoded enzyme towards production of isoamyl acetate is lower than that of the S. cerevisiae ATF1 encoded enzyme. Its overexpression in commercial lager yeast leads to a 2-fold increase of isoamyl acetate in synthetic YP medium (Verstrepen et al. 2003c). The kinetic properties of Atf1 from S. uvarum and S. kudriavzevii have been determined after overexpression in an S. cerevisiae laboratory strain in absence of the known AATase genes, ATF1 and ATF2, and the isoamyl hydrolase gene, IAH1 (Stribny, Querol and Perez-Torrado 2016). The affinity constants (Km) of isoamyl alcohol for the Atf1 orthologs in S. uvarum and S. kudriavzevii were 92.9 and 57.4 mM vs. 32.3 mM for S. cerevisiae with the same experimental setup. With the close homology of 99% similar amino acids between the S. uvarum and S. eubayanus ATF1 gene products, we can speculate that the Lg-ATF1 gene product in lager yeast has similar properties. Interestingly, 2-phenylethyl acetate production by S. cerevisiae in wine fermentations was 2-fold higher when the S. cerevisiae ATF1 ORF was replaced with ORFs from either S. uvarum or S. kudriavzevii, which illustrates some diversity in the enzyme specificities, even between Saccharomyces sensu stricto species.

The complex regulation of Atf1. The yeast AATase activity is affected by a range of different fermentation conditions, including carbon to nitrogen (C/N) ratio, sugar content of the wort, CO₂/hydrostatic pressure, fermentation temperature, levels of unsaturated fatty acids and oxygen, and pitching rate (Engan 1974; Calderbank and Hammond 1994; Verstrepen et al. 2003b; Dekoninck et al.2012). The molecular basis for the regulation of ATF1 transcript levels by unsaturated fatty acids, oxygen and nutrients is based on both direct and indirect evidence discussed below, while the effect of CO₂/hydrostatic pressure is discussed separately. A simplified model of the complex regulation of ATF1 is shown in Fig. 6.

Open multimedia modal

The multifaced activator/repressor Rap1. Some of the multiple effects can be explained by the protein kinase A (PKA) and Sch9-mediated fermentable growth medium (FGM) pathways, which have been shown to affect the expression of ATF1 (Fujiwara et al.1999; Verstrepen et al. 2003a). A truncated TOR1 allele was shown to enhance the levels of acetate esters, and in particular the ‘rose’ and ‘honey’-like ester phenylethyl acetate (Trindade de Carvalho et al.2017). Tor1 is a well-known regulator of nitrogen catabolite repression genes (Conrad et al.2014) and is therefore likely to affect the expression of the ATF1 gene. However, it is difficult to interpret these results in direct relation to fermentation conditions, as strains attenuated or disrupted in these pathways are known to be affected in broad stress-related responses. A more targeted approach was used by Fujiwara et al. (1999), which showed that the repressor-activator protein Rap1 is responsible for both transcriptional activation in anaerobic conditions and repression by unsaturated fatty acids. Rap1 is an essential transcription factor that regulates silencing and chromosome end maintenance by recruiting Sir and Rif proteins to the telomeres through its C-terminal RCT protein interaction domain (Chen et al.2011b). Interestingly, Rap1 and the nitrogen catabolite repression transcription factor Gcn4 interact directly and repress ribosomal protein expression under nitrogen starvation (Joo et al.2011), which could also explain why low nitrogen content hampers ATF1 expression. The transcription factor Gcr1 also interacts directly with Rap1, and is able to induce genes of the glycolytic pathway (Tornow et al.1993), which explains why ATF1 is upregulated in high gravity fermentations although it does not contain a Gcr1 binding site. Based on the observation that Rap1 can bind directly to zinc finger transcription factors and repress or induce expression of target genes, it is possible that many of the effects can be explained through a coupled interaction. This raises the possibility that other stress response zinc-finger transcription factors, such as Msn2/Msn4, could alter Rap1 transcription efficiency, explaining the complex regulation of the ATF1 gene.

In addition to the possible coupled interactions, Rap1 availability is itself regulated by external factors such as oxygen (Dastidar et al.2012). Interestingly, transcriptional changes occur in cells that stop growing, which disperses Rap1 from telomeric loci to bind to other parts of the genome (Platt et al.2013). As Rap1 is a major regulator of ribosomal protein expression, which consumes a large fraction of the cellular energy production (Lieb et al.2001), it can be regarded as one of the general activators of growth. A rule of thumb among brewers says that vigorous production of acetate esters is a good indicator of the happiness of the yeast. This conclusion fits with the role of Rap1 in induction/repression of ATF1 transcription and with the effect of stress responses that negatively affect ATF1 transcription.

Activation by Mga2. Early research revealed that the melting point of the unsaturated fatty acids is correlated with the repression of ATF1 transcription in a similar way as OLE1 fatty acid desaturase expression (Fujiwara et al.1998). Ole1 transforms the saturated fatty acids into Δ9-unsaturated acids (such as oleic acid) that are required for growth. The paralogous transcription factors Mga2 and Spt23 are necessary for induction of OLE1 transcription. They have complementing functions as an mga2Δ spt23Δ double deletion strain is auxotrophic for unsaturated fatty acids as opposed to the single deletion strains (Zhang, Skalsky and Garfinkel 1999; Chellappa et al.2001; Kandasamy et al.2004). Apart from being tightly regulated by unsaturated fatty acids, Mga2 activation is repressed by oxygen and induced in response to low temperatures (Nakagawa et al.2002). Yeast responds to cold by increasing membrane fluidity as an adaptation to control membrane rigidness, whereas oxygen limitation requires upregulation of OLE1 for more efficient desaturase activity (Aguilar and de Mendoza 2006). Cold induction of Mga2 fits with increased fruitiness of beers produced at low temperature. However, the temperature effect on acetate ester production is strain dependent (Verstrepen et al. 2003b), and fits with counteraction by decreased growth rate via regulation by Rap1.

Mga2 and Spt23 are tethered in their inactive forms to the membrane of the ER via a C-terminal hydrophobic domain. Activation occurs by proteolytic cleavage of the C-terminal domain after which they enter the nucleus and activate their transcription targets (Hoppe et al.2000; Rape et al.2001). The processing and release of Mga2 and Spt23 from the ER into the nucleus is strongly inhibited by unsaturated fatty acids (Hoppe et al.2000). Transcriptional activation by the soluble Mga2 factor, but not Spt23, retains repression by unsaturated fatty acids even without the C-terminal transmembrane domain (Chellappa et al.2001). Thus, different roles have been assigned to Spt23 and Mga2, in which Mga2 regulates both expression and degradation of the OLE1 transcript (Kandasamy et al.2004). Only re-addition of soluble Mga2 could restore the transcript to wild-type levels (Chellappa et al.2001), indicating that Mga2 plays a major role.

A mutant Mga2 allele with a premature stop codon (Ser706*) was shown to release ATF1 mRNA transcription from repression by unsaturated fatty acids (Takahashi et al.2017). The Mga2 mutation was identified in Saké yeasts that produce 2.6-fold more isoamyl acetate after acquiring resistance to the antibiotic aureobasidin A (Takahashi et al.2017). The mechanism of action of this truncated form of Mga2 could occur in multiple ways. For example, it could cause a permanent nuclear localization through which it could activate its targets constitutively.

Repression by Rox1. The strong repression of the ATF1 gene by oxygen occurs mainly through the Rox1 transcription factor (Fujii et al.1997; Fujiwara et al.1999). Rox1 is itself regulated by another transcription factor Hap1, which responds to oxygen through sensing of the heme level in the cells (Kastaniotis and Zitomer 2000). When oxygen is abundant, Hap1 upregulates Rox1 and represses targets that are not required in aerobic conditions. For efficient repression by Rox1, the Ssn6-Tup1 repressor complex binds to Rox1 (and other DNA binding proteins) and facilitates gene repression (Smith and Johnson 2000). YAP1 mutants with a truncated Yap1 C-terminus have been isolated, which showed up to 3-fold elevated production of isoamyl acetate in wine fermentations (Cordente et al.2013). These mutants were isolated based on cerulenin resistance. Yap1 is a zinc finger transcription factor that is induced under oxidative stress (Gulshan, Thommandru and Moye-Rowley 2012). Acetic acid induces an oxidative stress response in S. cerevisiae by several different mechanisms, which is necessary for cell survival, and in which Yap1 is a central regulator (Semchyshyn et al.2011; Morano, Grant and Moye-Rowley 2012). It has been shown that Rox1 represses gene expression of the FET4 high-affinity iron transporter gene through interaction with Yap1 (Caetano et al.2015). A yap1Δ strain also produced higher transcript levels of Rox1, indicating several modes of inhibition.

The inhibition of Atf1 by CO₂ and hydrostatic pressure. One of the most pressing issues for brewers is the lack of isoamyl acetate production in the very tall cylindroconical fermentors used in commercial breweries. The amount of isoamyl acetate produced by yeast AATase activity is negatively correlated with the depth of the fermentor. This is due to the high levels of dissolved CO₂ present under high hydrostatic pressure and to the physical agitation effect by the CO₂ bubbles accumulating at high depths (Vrieling 1978; Meilgaard 2001). CO₂ is essential as substrate in low levels during rapid fermentation as it feeds into the TCA cycle through the anaplerotic reactions and for the biosynthesis of fatty acids, purines and the charged amino acids aspartate and glutamate (Aguilera et al. 2005b). However, at higher levels, it inhibits growth, fusel alcohol production and, to an even larger extend, the formation of acetate esters (Renger, Hateren and Luyben 1992; Shen et al.2004). When the yeast is subjected to high CO₂ pressure for about 30 h, the uptake of any remaining branched chain amino acids through the cell membrane is reduced (Knatchbull and Slaughter 1987). This is correlated with lower production of fusel alcohols and esters as well as higher diacetyl production, increased cell size and lower viability, but it is not directly correlated with the ethanol production rate, at least not when the CO₂ overpressure remains limited (e.g. 1 bar CO₂ pressure) (Knatchbull and Slaughter 1987; Slaughter, Flint and Kular 1987). The growth inhibitory effects of CO₂ are to some extend caused by direct inhibition of metabolic enzymes, lowering of the intracellular pH and impairing mitochondrial function once the CO₂ (and bicarbonate) is inside the cells (Jones and Greenfield 1982; McIntyre and McNeil 1998; Aguilera et al. 2005a; Garcia-Gonzalez et al.2007). As a consequence of the combined effects caused by high CO₂ pressure, the cells have lower ATP levels upon long-term exposure (Richard, Guillouet and Uribelarrea 2014), but the major molecular biological factors involved in growth inhibition and reduction of the AATase activity have not yet been clarified.

A lead into a better understanding of the molecular-genetic mechanisms involved in suppression of ester production may come from the Sch9 and cAMP/PKA pathways that have been implicated in ATF1 expression regulation (Fujiwara et al.1999; Verstrepen et al. 2003a) and CO₂ sensing. Saccharomyces cerevisiae can sense CO₂ and downregulate the enzyme carbonic anhydrase to limit bicarbonate levels under CO₂ exposure. This occurs through sphingolipid Pkh1/2-mediated activation (phosphorylation) of the Sch9 protein kinase, with subsequent activation (phosphorylation) of the Cst6 transcription factor, responsible for expression of carbonic anhydrase (Pohlers et al.2017). On the other hand, high CO₂ levels activate adenylate cyclase (Cyr1) in Candida albicans, where it acts as a chemosensor for virulence in the human body (Klengel et al.2005). An activated Cyr1 enzyme will produce a surge in cyclic AMP (cAMP) which can thus explain CO₂-mediated activation of PKA (Vandamme, Castermans and Thevelein 2012). However, Jungbluth, Mosch and Taxis (2012) could not find direct evidence for Cyr1 being regulated by CO₂/bicarbonate in S. cerevisiae, and while CO₂ exposure causes filamentous growth (hypha formation) in some Candida species, it does not have this effect in S. cerevisiae (no pseudohyphae formation), suggesting that different pathways are involved in the two species (Gancedo 2001; Vandamme, Castermans and Thevelein 2012) The mitochondrial CO₂/cAMP/PKA pathway is a well-characterized regulatory pathway in mammalian cells (Valsecchi et al.2013), and has been investigated in S. cerevisiae where a mitochondrial spike in cAMP levels upon bicarbonate exposure led to the proposition that mitochondrially localized Cyr1/PKA is responsive to CO₂ during non-fermentative growth (Hess et al.2014). The involvement of the cAMP/PKA pathway in sensing CO₂ pressure in industrial anaerobic fermentation conditions is therefore not yet clear.

The biological function of Atf1. Atf2 has been shown to function primarily in the sterol detoxification cycle together with the deacetylase Say1 (Tiwari, Koffel and Schneiter 2007). On the other hand, a phenotype has not been detected for the ATF1 null mutant except for strongly reduced acetate ester production. Acetate ester production can have an impact on the dispersal of yeasts in nature. This was shown by purging the headspace from semi-anaerobic fermenting yeast cultures into a closed gas box containing fruit flies. The fruit flies showed a higher frequency of gathering towards the headspace gas originating from a yeast with artificially engineered overexpression of the ATF1 gene (Christiaens et al.2014), and were especially attracted to ethyl acetate. It has been known for a long time that there are many non-Saccharomyces yeasts typically isolated in much higher titers from fruits and other sugar-rich natural habitats that produce much higher quantities of ethyl acetate (Tabachnick and Joslyn 1953; Rojas et al.2001; Sabate et al.2002; Fleet 2003; Raspor et al.2006; Li et al.2010). For example, the corn sap beetle was highly attracted by W. anomalus (Pichia anomala) that produces abnormally high levels of ethyl acetate, even exceeding that of an ATF1 overexpression strain in S. cerevisiae (Nout and Bartelt 1998). The production of esters by the Pichia and Saccharomyces genera must be affected differently in nature, as the Saccharomyces genes are repressed by oxygen while Pichia is an aerobic yeast that produces esters at high levels in the presence of oxygen, a condition typically encountered in most natural niches. In addition, a range of volatile aroma-active metabolites, including acetic acid and beer phenolic off-flavors, such as 4-vinylguaiacol, also function as attractants for fruit flies (Drosophila melanogaster) (Dzialo et al.2017). It remains elusive whether the Atf1 enzyme performs other biological functions besides the production of acetate esters. Interestingly, the S. cerevisiae Atf1 AATase has very broad substrate specificity and can acetylate long-chain fatty alcohols (Ding et al.2016). In mammalian cells, the non-volatile acetate ester 12-O-Tetradecanoylphorbol-13-acetate disrupts actin filaments and increases mucin secretion, which increases the extracellular viscosity and adhesiveness, through regulation by the protein kinase C (PKC)-MAPK kinase pathway (Shiba, Sasaki and Kanno 1988; Hong, Forstner and Forstner 1997; Lee et al.2002; Bansil and Turner 2006). Yeast mutants resistant to aureobasidin A and producing increased isoamyl acetate have been isolated (Takahashi et al.2017). Aureobasidin A disrupts cortical actin patches and inhibits inositol phosphorylceramide synthase AUR1, which triggers signaling through the cell-wall-integrity PKC pathway (Endo et al.1997; Zhong, Murphy and Georgopapadakou 1999; Jesch et al.2010). In S. cerevisiae, the integral membrane mucin Msb2 serves to stimulate pseudohyphal growth via MAPK activation (Cullen et al.2004) and inhibition of sphingolipid synthesis via PKC-MAPK induces telomeric silencing (Lee et al.2013). Both pathways are poorly characterized under anaerobic fermentation conditions. A yet unexplored possibility is that Atf1 acetylates a yet unidentified lipid messenger, which regulates a subtle phenotype or a phenotype that is only prominent under specific natural conditions. More studies should be carried out especially under anaerobic conditions to investigate the unresolved issues concerning Atf1 functionality in S. cerevisiae.

Selection for improved acetate ester production and esterase activity. Isolation of strains resistant to various drugs has been employed to improve the production of acetate esters and increase the fruity aroma of saké. As previously mentioned, the toxic leucine analog 5,5,5-trifluoroleucine has been used to increase the isoamyl alcohol level, which also leads to a higher isoamyl acetate level. It is, however, not desirable to have high levels of isoamyl alcohol as it also gives a heavy (‘alcoholic’, ‘vinous’, ‘sweet’) aroma. Isoprenoid anticancer drug 1-farnesylpyridinium (Hirooka et al.2005), copper (Hirooka et al.2010), ergosterol inhibitor econazole (Asano et al.1999), monofluoroacetate (Watanabe et al.1993) and isoamyl monochloroacetate (Watanabe, Nagai and Kondo 1995) have all been used specifically for improvement of the isoamyl alcohol to acetate conversion ratio. These compounds work either by increasing the AATase activity or by reducing esterase catalyzed degradation. The major esterase in S. cerevisiae, isoamyl acetate hydrolase encoded by the IAH1 gene, has been cloned and characterized (Fukuda et al.1996^, 2000). Loss of function of this gene significantly increases the yield of isoamyl acetate and to a lesser extent that of the other acetate esters. It is therefore the balance between AATase and esterase activities that determines the final level of acetate esters produced by brewing yeast (Fukuda et al.1998).

Ester biosynthesis and role of the fatty acid synthase complex. The apple esters ethyl hexanoate and ethyl octanoate are present above their threshold in certain beers and thus contribute to the fruity aroma. Ethyl hexanoate, the most prominent apple ester in beer, is produced by the ethanol acyl-coA transferases Eht1 and Eeb1 that condense ethanol with a medium-chain fatty acid charged with coenzyme A (acyl-coA), with Eeb1 being particularly active (Saerens et al.2006). Moreover, an allele of the phospholipase B gene, PLB2, with in vitro acyltransferase activity (Merkel et al.1999), was isolated by polygenic analysis of ethyl ester production, and confirmed to be a strong effector of ethyl octanoate production by gene disruption (Steyer et al.2012). Reverse esterase activity has also been observed with the purified Iah1 enzyme, but only for ethyl hexanoate formation (Kuriyama et al.1986).

The expression levels of Eht1 and Eeb1 do not show a strong correlation with the final levels of ethyl hexanoate (Saerens et al.2008), which is consistent with the finding that the bottleneck of their formation is the level of the acyl-coA substrate (Saerens et al.2006). Ethyl hexanoate has received considerable interest in saké brewing because it is a key aroma compound in high-quality saké. Yeast selected for cerulenin resistance showed 3.7-fold increased production of ethyl hexanoate in saké fermentations (Akada et al.1999). The mutation responsible for cerulenin resistance (Gly1250Ser) was found to be in the FAS2 encoded subunit of the fatty acid synthase (FAS) complex (Inokoshi et al.1994). The cerulenin resistance conferring mutation, which is located in the middle of the 3-ketosynthase domain, may compromise cycling of the fatty acids and confer leakage of medium-chain fatty acid substrates, such as hexanoic acid, from the FAS complex. This specific mutation, causing enhanced production of ethyl hexanoate, has been engineered into saké yeasts currently in commercial use (Aritomi et al.2004). Furthermore, a FAS2 allele isolated from an Ale yeast has been shown to improve acetate ester production in a way that appears to be unrelated to cerulenin resistance (Trindade de Carvalho et al.2017). The molecular mechanisms of mutations leading to the altered acetate ester production have not been elucidated.

Non-Saccharomyces yeasts with superior ester profiles. Non-Saccharomyces yeast (often termed non-conventional yeast) contains an untapped reservoir of flavor production that can enhance or complement the flavor profiles of Saccharomyces yeasts. In particular, co- or sequential fermentations with P. kluyverii or Cyberlindnera fabianii and brewing yeast shows potential for enhancing the ester profile (van Rijswijck et al.2017; Holt et al. 2018b). While P. kluyverii produces a very high level of acetate esters (resembling a Saccharomyces ATF1 overexpression strain), C. fabianii forms a high level of ethyl esters (‘apple’ like). However, the high levels of acetate esters were repressed in consecutive fermentations with brewing yeast, leading to the hypothesis that esterases are activated upon fermentation with S. cerevisiae (Holt et al. 2018b). A better understanding of how they interact in co- and sequentially inoculated cultures will therefore not only be interesting in an ecological context, but might also be beneficial for development of compatible strains.

Biosynthesis of S-conjugate precursors in hops

The biosynthesis of glutathionylated precursors in grapevine is believed to occur through a glutathione-S-transferase, which condenses reactive and stress-induced aldehydes (such as trans-2-hexenal) with glutathione (to from glutathione-S-3-hexan-1-ol) (Thibon et al.2011). Formation of the glutathione-S-conjugated precursor in grapevine leaves was induced by UV radiation, drought stress and botrytis infections, which correlates with the expression of the two glutathione-S-transferase genes, VvGST3 and VvGST4, that were shown to form glutathione-S-3-hexan-1-ol in vitro with the substrates trans-2-hexenal and glutathione (Kobayashi et al.2011). The high UV index in the southern hemisphere is believed to induce high levels of polyfunctional thiols and their precursors in Sauvignon Blanc grapes from New Zealand, although post-harvest UV treatments of defrosted grapes do not appear to affect the levels (Parish-Virtue et al.2019). Clearly, more research is necessary to conclusively determine the impact of UV radiation. The formation of glutathione-S-3-hexan-1-ol is induced in grapevine cultivated in high nitrogen conditions. However, the transcript levels of the major grapevine glutathione-S-transferase enzymes were not induced under this condition, which suggests the presence of other yet unidentified enzymes responsible for their formation under high nitrogen conditions (Helwi et al.2016).

Lactones in beer

The formation of γ-lactones is dependent on the yeast strain (Loscos et al.2007) and the composition of the wort, particularly the fatty acid levels originating from both malt and hops. Biological formation of lactones typically starts with oxidation of a fatty acid substrate by lipoxygenases (LOX), which introduces a hydroxy group from hydrogen peroxide in a specific position, and subsequent breakage of the hydroxy acid in the β-oxidation pathway (Romero-Guido et al.2011; Joo and Oh 2012; El Hadi et al.2013). Next, the shortened fatty acid is released from coenzyme-A and forms a ring structure by esterification of the hydroxy group with the carboxylic acid (called lactonization). Lactonization is partly dependent on the yeast strain used during the fermentation and is favored at low pH in the fermentation media by spontaneous lactonization of fatty acids excreted from the cell (occurs naturally around pH 3.5) (Muller, Kepner and Webb 1973; Endrizzi et al.1996). Lactonization in beer is therefore likely to occur either by heating to form Maillard products during mashing and kilning of the malt or by enzymatic catalysis in the wort. During Whiskey production, high concentrations of lactic acid bacteria can occur spontaneously. These bacteria can cross-feed hydroxylated oleic and palmitoleic acid to distillers and brewer's yeast, which transforms these compounds into γ-decalactone and γ-dodecalactone (‘peach’, ‘fatty’, ‘butter’) (Wanikawa, Hosoi and Kato 2000).

Barley contains two LOX enzymes (LOX-1 and LOX-2), of which only LOX-2 is active during germination in the malting process (Yu et al.2014). The activity of this LOX enzyme is not only responsible for the formation of γ-nonalactone that contributes with ‘sweet’ and ‘coconut’ aroma, but also of trans-2-nonenal, a ‘cardboard’ off-flavor, that is formed mainly by hydroxylated fatty acid degradation during aging in the bottle and negatively affects flavor stability (Baert et al.2012). Therefore, LOX-negative barley strains have been constructed by the Tsingtao brewery (work commenced by Heineken and Carlsberg); they lack the LOX genes, do not form significant amounts of trans-2-nonenal, and result in beer with an increased shelf-live (Yu et al.2014). γ-Nonalactone also originates from hop precursors (Sakuma et al.1996), but with no active barley LOX enzyme, the conversion is dependent on autooxidation (Baert et al.2012). The fatty acids linoleic and linolenic acid are released during mashing and subsequently hydroxylated, which indicates both lipase and LOX activities (Kobayashi et al.1993). The LOX enzymes are heat-labile and are partly inactivated during the mashing (Kobayashi et al.1993; Baert et al.2012).

The formation of γ-butyrolactone has been studied in film-forming yeasts (S. fermentati), and derives from the metabolism of the amino acid glutamate, resulting in the formation of 4-oxobutyric acid (Muller, Kepner and Webb 1973; Wurz, Kepner and Webb 1988). The formation of lactones has been particularly well studied in the lipophilic yeast Yarrowia lipolytica that produces much higher levels than S. cerevisiae (Wache et al.2003). Also, it was shown with radiolabeling experiments that γ-decalactone and γ-dodecalactone could be produced from oleic acid (C18) in Sporobolomyces odorus, which is consistent with both oleic and linolenic as the natural precursors (Haffner and Tressl 1996).

Furanones formed during malting and mashing

Strawberry furanone is formed as a Maillard product during kilning of malt and mash boiling and is found in higher concentrations in roasted malt varieties (Schieberle 1991; Slaughter 1999). Its level also increases during fermentation and depends on the yeast strain used (Sakuma et al.1996). It is derived partly from α-ketoglutarate catabolism and de novo synthesis of the amino acid threonine during the fermentation (Tressl et al. 1978a).

Terpenoid synthesis in hops

Terpenoids are the largest and most diverse group of flavor compounds. In particular, two of the plant species that produce terpenoids are Vitis vinifera (grapes) and Humulus lupulus (hops), important raw ingredients of wine and beer industries, respectively. They are synthesized de novo in the plastids from pyruvate and glyceraldehyde-3-phosphate in the methylerythritol 4-phosphate (MEP) pathway, or from acetyl coenzyme A (acetyl-coA) in the cytosol, the ER and the peroxisome through the mevalonate (MVA) pathway (Singh and Sharma 2014). The MEP pathway genes are highly expressed and have for each enzyme several paralogs in hop, compared to the MVA pathway genes (Nagel et al.2008; Wang et al.2008). It is therefore thought to be the major pathway responsible for provision of isoprenoid precursors for terpenoid synthesis. This pathway has seven sequential enzymatic steps, requiring carbon influx and energy charged molecules (ATP, CTP) (Fig. 7). Hydroxymethylbutenyl 4-diphosphate, the end product of the MEP pathway, is transformed into the isoprenoid precursors isopentenyl diphosphate (IPP, C5) and dimethylallyl diphosphate (DMAPP, C₅) by IPP/DMAPP synthase and can be interconverted by IPP/DMAPP isomerase. Next, IPP and DMAPP are condensed by prenyltransferases to geranyl diphosphate (GPP, C₁₀) and neryl diphosphate (C₁₀), farnesyl diphosphate (FPP, C₁₅) and geranylgeranyl diphosphate (GGPP, C₂₀). These are the basic units for further processing by terpene synthase enzymes into monoterpenes (C₁₀), sesquiterpenes (C₁₅) and diterpenes (C₂₀) using mono-, sesqui- and diterpene synthases (MTS, STS and DTS), respectively (Chen et al. 2011a). The wealth of different terpenoids in plant products is due to terpene synthases in different plant species yielding multiple different products from the same substrate and the subsequent addition of functional groups (Tholl 2006; Nagegowda 2010; Chen et al. 2011a). Notably, the functional groups can also come directly from the terpene synthase activity. For example, linalool synthase cleaves the oxygen from the phosphate of GPP, and finally yields β-linalool, which contains an alcohol group (Cseke, Dudareva and Pichersky 1998). Addition of functional groups (alcohols, ketones, aldehydes, etc.) can occur either by enzymatic reactions in the hop plant or by yeast in the brewing process. The latter is discussed in detail later.

Open multimedia modal

Hop terpenoid synthesis has been investigated by Wang and colleagues (Wang et al.2008; Wang and Dixon 2009). Using a cDNA library made from RNA of the glandular trichomes (hairs) and flowers of the hop plant, GPP synthase, two monoterpene synthase and two sesquiterpene synthase homologs were identified, expressed in Escherichia coli and enzymatically characterized. The GPP synthase consists of a small and a large subunit that form a heterodimer producing GPP (59.5%) and GGPP (40.5%) when incubated with the substrates IPP and DMAPP, and uniquely GGPP when IPP and GPP or GFP is supplied in vitro (Wang and Dixon 2009). The large subunit is the catalytic entity that on its own has a higher tendency to form GGPP (68.2%) than GPP (26.9%). Expression levels of the small subunit of GPP synthase are highest in glandular trichomes, and are directly correlated with the myrcene concentration, whereas the large subunit is also expressed in stems, leafs and flowers (Wang and Dixon 2009). The MTS enzyme, Mts2, catalyzes the formation of myrcene from GPP, while the STS enzyme, Sts1, synthesizes β-caryophyllene and α-humulene from FPP (Wang et al.2008). Among these, only myrcene is a powerful aroma-active terpenoid. The terpene synthases responsible for formation of some of the most aroma-active terpenoids have not been characterized yet. Interestingly, Wang and colleagues identified a putative β-linalool synthase predominantly expressed in flowers, but did not investigate it further. Comparative RNA-Seq analysis showed a higher expression of the MTS1 gene in the hop cone of the cultivated hop variety cv. Shinshu Wase compared to a wild hop variety (Natsume et al.2015). This indicates that a selection occurred in the cultivar for upregulated specialized flavor metabolism, leading to increased aroma production. Recently, nine terpenoid syntheases from glandular trichomes of Cannabis sativa, belonging to the same Cannabaceae family as hops, were functionally characterized and linked to the production of β-myrcene, (E)-β-ocimene, (–)-limonene, (+)-α-pinene, β-caryophyllene and α-humulene (Booth, Page and Bohlmann 2017). These terpenoid synthases, as well as the already characterized hop terpene synthases Mts1, Mts2, Sts1 and Sts2, belonged to the TPS-b and TPS-a subfamilies, which typically contain mono or sesquiterpenoid syntheases, respectively.

Some aroma-active terpenoids, termed norisoprenoids (e.g. β-ionone, ‘artificial raspberry’, ‘cedarwood’, ‘violet’), are formed from oxidative cleavage of carotenoids (C40, tetraterpenoids) by carotenoid cleavage dioxygenases (CCDs) (El Hadi et al.2013). Interestingly, some yeasts and fungi can also produce this compound from oxidation and cleavage of β-carotene (a major carotenoid compound) (Zorn et al.2003a^, b; Romero-Guido et al.2011). There is also a high interest in norisoprenoids from the flavor and fragrance industry due to their pleasant aroma and low sensory thresholds. This has led to the construction of a transgenic S. cerevisiae strain that produced considerable amounts of β-ionone by introduction of genes for both the formation and oxidative degradation of β-carotene (López et al.2015). In wine production, carotenoids are typically absent in musts. The presence of β-ionone is likely due to the activity in the grapevine of the CCD enzyme VvCCD4b that is induced during ripening of grapes, when the level of carotenoids decreases steadily (Mendes-Pinto 2009; Lashbrooke et al.2013). A study on Pinot grapevine showed that the degradation of carotenoids during ripening is well correlated with the formation of some norisoprenoids, such as β-damascone (‘blackcurrant’, ‘raspberry’, ‘menthol’), but not linked to others such as β-ionone (Yuan and Qian 2016), suggesting that some of these compounds form during fermentation or during maturation.

Genomic regions (quantitative trait loci, QTLs) linked to the production of 17 mono- and sesquiterpenoids in hops showed mostly complex pleiotropic interactions between QTLs with linkage for multiple compounds and also some unique QTLs for specific compounds (humulene, δ/γ-cadinene and ρ-cymene) (McAdam et al.2013). This overlap between genomic regions for several terpenoids may be due to selection for regulatory elements responsible for high expression of terpenoid synthases or common genes in the biosynthesis pathway as described above. However, the large QTLs, typically identified in plant studies, do not allow to pinpoint the specific genes responsible without further targeted molecular analysis, such as genome editing facilitated by CRISPR/Cas (Ricroch, Clairand and Harwood 2017). This technique is currently being explored in the genetically related cannabis plant (Maxmen 2018), but still needs to be implemented for hops. Important work therefore remains to be performed regarding functional and genetic characterization of the remaining terpenoid synthases, their regulation and transformation by brewing yeast.

Another important aspect of terpenoid biosynthesis is the formation of flavorless glucosides of terpenoids that are being produced by glucosyltransferase activity in hops, and later being hydrolyzed during the brewing process into flavor-active terpenoids. The glycosylation of terpenoids occurs through transfer of the sugar molecule from the nucleoside diphosphate-activated group [e.g. uridine diphosphate (UDP)-glucose] to the functional group of the terpenoid (e.g. an alcohol or acid) (Tiwari, Sangwan and Sangwan 2016). Glycosylation is known to aid detoxification by solubilizing these hydrophobic compounds, which allows transport over the cell membrane (Zhao et al.2011). Screening for glycosylation activity by in vitro assays with glycosyltransferases from the plant model organism Arabidopsis thaliana identified 27 glycosyltransferases with diverse activities towards the aroma-active monoterpenoid alcohols geraniol, linalool, menthol, α-terpineol, β-citronellol and the sesquiterpenoid farnesol (Caputi, Lim and Bowles 2008). Recently, four hop terpenoid glycosyltransferases were shown to have high activity towards UPD-glucose together with linalool and geraniol, which was patented for purpose of reducing their activity, potentially leading to a higher level of free flavor-active terpenoids (Ono and Nobuo 2013). In grapevine, the two UDP-glucose monoterpenol glycosyltransferases, VvGT7 and VvGT14, were shown to be induced by ripening, which correlated with glycoside formation of a range of monoterpenoids, including geraniol and linalool (Li et al.2017). UDP-glucose:cinnamate glycosyltransferase expression in strawberry was also shown to be correlated with fruit ripening, specific for fruit-forming tissue, downregulated by auxin and induced by oxidative stress (Lunkenbein et al.2006). The molecular mechanisms of regulation of glycosyltransferases related to flavor-active terpenoids have not been investigated to the same level as conjugates with flavonoids and hormones (Gachon, Langlois-Meurinne and Saindrenan 2005; Tiwari, Sangwan and Sangwan 2016), and it is therefore not clear how the environmental signals are transmitted.

Hop terpenoids in the brewing process

The most abundant terpenoids found in the essential oil of hops, namely the sesquiterpenes: α-humulene, β-caryophyllene and β-farnesene and the monoterpenoid myrcene, do not directly shape the flavor profile of most beers. Instead, they undergo significant chemical and functional modifications during the brewing process (thermal reactions, oxidation, hydrolysis or isomerization) (Fig. 8) as well as biochemical transformations by yeast during fermentation (esterification, ester exchange and enzymatic cleavage). That said, these effects on the hop oil components have not yet been fully investigated. Generally, it is known that the extraction of terpenoids from hops depends on the time that the hop was boiled. The amount decreases with increasing boiling time due to evaporation (Kishimoto et al.2005). Therefore, late, whirlpool or dry hopping is advised to achieve richer flavor and aroma profile in final beers (Sharp, Steensels and Shellhammer 2017). The impact of boiling parameters (heat, light or oxygen) on the components of essential oils, in terms of their stability and possible reactions, has been reviewed elsewhere (Turek and Stintzing 2013). In general, it is known that oxygenated derivatives of the terpenoids (e.g. in the form of alcohols) have a much higher probability to remain in the final beer due to their higher relative solubility.

Open multimedia modal

The behavior and fate of each aroma-active terpenoid is unique and compound-specific. The loss of the above-mentioned most abundant hop-derived terpene hydrocarbons during fermentation of hopped wort was explained by their adsorption to the hydrophobic yeast cells, by binding with yeast cell wall components and by migration to the foam layer (Lam, Foster and Deinzer 1986; King and Dickinson 2003; Praet et al.2012). However, it was not originally attributed to bioconversion by yeast, because no yeast transformation products of myrcene, α-humulene and β-caryophyllene could be detected (King and Dickinson 2003). An early study reported that the hop-derived monoterpenoids linalool and geraniol survived the entire brewing process and that β-citronellol was formed during fermentation from a hop precursor (most likely geraniol). This suggested that yeast has an ability to transform certain terpenoids into other terpenoids enzymatically (Lam, Foster and Deinzer 1986). Many subsequent reports focused on the use of mono- and sesquiterpene alcohols as potential precursors for bioconversion reactions performed by yeast (King and Dickinson 2003; Takoi et al.2010a^,b; Praet et al.2012). Regardless of the hop cultivar used, the same tendencies were observed during fermentation: slight decrease in linalool and α-terpinol, followed by an increase in β-citronellol and nerol and finally an increase in geraniol (Takoi et al. 2010a). The proposed metabolic reactions for transformation of monoterpene alcohols by Torulaspora delbrueckii and Kluyveromyces lactis, as well as S. pastorianus lager and S. cerevisiae ale yeast, are outlined in Fig. 9. However, the enzymes responsible for these transformations have not been identified yet, also very little is known about the genes involved. In general, it was observed that yeasts were able to convert hop-derived monoterpene alcohols, specifically geraniol into linalool or β-citronellol, nerol into geraniol, α-terpinol or linalool and the latter into α-terpinol (King and Dickinson 2000). Monoterpene alcohols can be further metabolized by yeasts, through esterification or hydrolysis, particularly acetylation of monoterpenes by AATase enzymes (King and Dickinson 2000) or by cleavage of monoterpene esters by esterases or lipases (Lam, Foster and Deinzer 1986; Chatterjee, Chatterjee and Bhattacharyya 2001). Using gene overexpression and gene deletion analysis, it was demonstrated that the AATase enzyme, Atf1, is responsible for acetylation of the major monoterpenoids (geraniol, citronellol and nerol) in S. cerevisiae S288c (Steyer et al. 2013). The same effect was not observed for Atf2. Notably, the non-conventional yeast Williopsis saturnus was able to produce higher concentrations of acetate ester derivatives of monoterpenoids (geranyl- citronellyl-) and retain better the original terpenoid content compared to brewer's yeast (Liu and Quek 2016).

Open multimedia modal

Geraniol is a major precursor for yeast transformation, and its initial level correlates with the final content of the other monoterpene alcohols in beer (Takoi et al. 2010a). During brewing, the rates of bioconversion reactions were dependent on yeasts’ growth phase and the timings of hop addition. Several researchers have reported that geraniol is mainly converted to β-citronellol during the first 2–4 days in model fermentations producing late-hopped beers (Takoi et al. 2010a). This phenomenon was explained by Yuan et al. (2011), who revealed the involvement of NADPH dehydrogenase 2, encoded by OYE2 (old yellow enzyme 2), in geraniol to citronellol reduction (Fig. 9). Overexpression of S. cerevisiae OYE2 increased the reduction of geraniol to β-citronellol to 87% in comparison to the 50% obtained with the control S. cerevisiae strain (Steyer et al. 2013). Deletion of the OYE2 or ATF1 genes, both involved in endogenous conversion of geraniol to other terpenoids, improved geraniol production by 1.7-fold or 1.6-fold in batch fermentation, respectively (Zhao et al.2017). Apart from Oye2, Atf1 and Atf2, the remaining enzymes involved in the biotransformation reactions of monoterpene alcohols in S. cerevisiae have not yet been identified. However, monoterpene degradation pathways, including the transformation between linalool and geraniol, have been more intensively studied in bacteria (Fig. 9). The bifunctional linalool dehydratase-isomerase (encoded by the LDI gene) was isolated from the bacterium Castellaniella defragrans, which catalyzes the hydration of β-myrcene to linalool in the thermodynamically unfavorable direction (dehydratase reaction). It further catalyzes the isomerization to geraniol (geraniol isomerase). These constitute the initial steps in anaerobic β-myrcene biodegradation (Brodkorb et al.2010). Geraniol can be further converted to geranial and geranic acid by geraniol- and geranial dehydrogenases, respectively (Luddeke et al.2012). In C. defragrans, geraniol dehydrogenase (NAD⁺-dependent) is encoded by the GEOA gene, while geranial dehydrogenase (NAD⁺-dependent) by the GEOB gene. Myrcene was also transformed by Pseudomonas aeruginosa to α-terpineol with limonene as intermediate (Esmaeili and Hashemi 2011). It has been suggested that limonene hydroxylase enzyme is possibly involved in conversion of limonene into linalool, perillyl alcohol and α-terpineol (Yang, Park and Chang2007). Besides, other enzymes involved in monoterpene transformations of microbial origin have been reviewed elsewhere (Marmulla and Harder 2014).

Among microorganisms, yeasts usually employ the MVA pathway for terpene synthesis, whereas bacteria mainly use the MEP route, except for some species, which use either the MVA route or both. In S. cerevisiae, firstly IPP (C₅) is synthetized from mevalonate by the enzymes mevalonate kinase (Erg12/Rar1), phosphomevalonate kinase (Erg8) and diphospho-mevalonate decarboxylase (Mvd1/Erg19). Then, IPP (C₅) can be rearranged to DMAPP (C₅) catalyzed by the IDI1-encoded enzyme IPP isomerase. Next, two reactions leading to the formation of GPP (C₁₀) and then FPP (C₁₅) are conducted by the bifunctional Erg20 enzyme, which has both GPP synthase and FPP synthase activities. Firstly, GPP synthase catalyzes condensation of IPP (C₅) with DMAPP (C₅) into GPP, followed by a second condensation of GPP with another IPP that is performed with FPP synthase to form the FPP molecule. GPP and FPP are key precursors for further mono- and sesquiterpenoid synthesis (Cordente et al.2012).

Normally, yeasts do not synthesize significant quantities of terpenoids de novo, due to lack of the MTS in the monoterpene synthesis pathway. In addition, they also lack a specific GPP synthase, and this metabolite only occurs as an intermediate in the synthesis of FPP, the precursor of several essential metabolites (Grabinska and Palamarczyk 2002). A relatively low level of the monoterpenoids, geraniol and linalool, could be synthesized de novo by Hanseniaspora uvarum and S. cerevisiae (1.2 and 4 μg/L, respectively) (Carrau et al.2005). However, new strategies in genetic modification helped to overcome these constraints. This included enginieering and optimization of the upstream isoprenoid pathway flux (prior to mevalonate synthesis) together with introduction of an appropriate plant-originated terpenoid synthase. For example, dynamic control of ERG20 expression combined with minimized endogenous downstream metabolism (redistributed precursor GPP flux) contributed to the 3.4-fold improvement in geraniol production in S. cerevisiae with introduced geraniol synthase from valerian flowers (Valeriana officinalis) (Zhao et al.2017). Moreover, the combination of dynamic control of ERG20 expression with OYE2 deletion increased geraniol production up to 1.69 g/L (Zhao et al.2017). The enzyme hexaprenyl pyrophosphate synthetase, encoded by the COQ1 gene, which is known to catalyze the first step in ubiquinone (coenzyme Q) biosynthesis, is involved in monoterpene synthesis in S. cerevisiae. Overexpression of COQ1 together with a mutated FPP synthetase, Erg20^K197E, resulted in significantly higher levels of linalool (above 750 μg/L), geraniol, α-terpineol and the sesquiterpenes, farnesol and nerolidol (Camesasca et al.2018). Recently, a methodology was developed to create brewer's yeasts capable of biosynthesizing monoterpenes to similar levels as those typically found in finished beer, and importantly, beer brewed with this yeast was characterized by a hoppy aroma in sensory analysis, without the addition of any external hops (Denby et al.2018). The successful strain contained a unique combination of strong promoters driving expression of the four modulated genes, which ensured variation in the ratio of linalool to geraniol: (1) a truncated form of yeast HMG acetyl-coA reductase lacking the regulatory domain, Hmg1^M1_S511del; (2) the mutated FPP synthase with enhanded activity, Erg20^K197E; (3) a truncated linalool synthase from mint (Mentha citrata), McLIS^M1_E67del without the plastidic targeting peptide; and (4) a geraniol synthase from basil (Ocimum basilicum) ObGES. In fermentation experiments, it was verified that the selected genes indeed control monoterpene production since the total production of monoterpenes was correlated with Hmg1^M1_S511del and Erg20^K197E abundance, linalool correlated with McLIS^M1_E67del abundance, but ObGES abundance was not promotional with geraniol synthesis. Although these engineered strains can improve fruity and floral notes in beer, genetically modified brewing yeasts often still lack acceptance by the consumer.

Release of terpenoids from glycosides

Due to the described transformations, terpenoids present in wort and beer undergo dynamic changes and interactions, usually leading to their decrease under traditional brewing conditions (Takoi et al. 2010a; Liu and Quek 2016). However, in beer brewed with hop pellets accumulation of linalool and β- damascone throughout the fermentation of wort has been reported (Kishimoto et al.2005; Kollmannsberger, Biendl and Nitz 2006). The observed increase in monoterpenoid levels is due to release from glycosidically conjugated flavor-inactive precursors (glycosides) of terpenoids and norisoprenoids found in hopped wort, which undergo hydrolysis during fermentation with release of their aromatic counterparts (aglycones) (Fig. 10). Plant materials used in the food industry contain up to five times more flavor compounds bound to glucose than free, aroma-active flavor compounds (Vervoort et al.2016). The cleavage of glycosides originating from hops or fruits is most commonly observed in spontaneously fermented beers with fruit maceration, such as Lambic and Gueuze base beers with cherries (Kriek), or in sour beer varieties (Daenen and De Clerck 2012). Principally, two mechanisms of hydrolysis are responsible for their release from the flavor-inactive glucosides. In sour beers, at low pH (3.0–3.5), acidic hydrolysis by terpene glycosidases can induce rearrangement of monoterpene alcohols into other compounds (Maicas and Mateo 2005). First, acidic hydrolysis (pH 3.0) is mainly responsible for release of linalool and α-terpineol, α-ionol and α-damascone in sour cherry beers (Daenen et al.2008). The second mechanism, enzymatic hydrolysis, can be initiated by addition of exogenous enzymes or by enzymatic activity of yeast, although the enzymatic activity in conventional brewing yeast is generally very low. Commercially available enzymes used in winemaking for aroma enrichment but also in fruit and vegetable processing for hydrolysis of glycosides act sequentially in two steps, as shown in Fig. 10 (Daenen and De Clerck 2012; Wen et al.2014). First, disaccharide glycosides are cleaved by the action of α-L-rhamnosidase, α-L-arabinosidase and β-D-apiosidase liberating the corresponding terpene monoglucosides by cleavage at the (1→6) glycoside bond. In the second step, the aglycone-carbohydrate bond is cleaved by β-D-glucosidase releasing monoterpene alcohols such as linelool. For example, the enzyme preparation Rapidase AR2000, containing pectinase and glycosidases derived from Aspergillus niger, has been successfully applied to release aglycones from glycosides in hops, hop products and beer (Kollmannsberger, Biendl and Nitz 2006; Daenen and De Clerck 2012), in cherry juice (Wen et al.2014) and sour cherry Kriek beer (Daenen et al.2008; Daenen et al.2008; Daenen and De Clerck 2012; Vervoort et al.2016).

Open multimedia modal

Yeasts with β-glucoside hydrolase activity can also release flavor-active compounds from hop-derived glucosides (Daenen et al.2008; Vervoort et al.2016; Sharp, Steensels and Shellhammer 2017). Daenen et al. (2008) first reported that the Saccharomyces brewing yeasts that they tested did not show 1,4-β-D-glucosidase (BGL) activity, only a strain-dependent low-to-moderate activity of exo-1,3-β-glucanase (EXG). The major exoglucanase enzyme in S. cerevisiae, encoded by the EXG1 gene, has also been shown to have non-specific activity, which can act on glucose polymers and smaller glycosidically conjugated substrates (Larriba et al.1993; Suzuki et al.2001; Schmidt et al.2011). Thus, Saccharomyces strains characterized by high EXG activity demonstrated BGL-like activity against hop glycosides during fermentation (Daenen et al.2008), and their BGL-like activity varied widely among different strains (Sharp, Steensels and Shellhammer 2017). The Exg1 protein is proteolytically hydrolyzed by the Kex2 protease and subsequently released into the extracellular space via the secretion (Sec) pathway, through the endoplamatic reticulum, the golgi apparatus and the periplasmic space from where it diffuses into the extracellular medium (Cenamor et al.1987; Larriba et al.1993; Cappellaro, Mrsa and Tanner 1998). Release of active Exg1 into the extracellular medium is enhanced in a strain lacking the ß-glucan synthase Kre6, which has compromised cell integrity (Wang et al.2015). The EXG1 gene shows high expression under its native promoter during exponential growth, but is repressed in low hypoxic conditions that occur during fermentation (Lu-Chau et al.2004). Overexpression of the EXG1 gene in a Saccharomyces wine yeast strain under the constitutive ACT1 promoter showed 20- to 40-fold higher EXG activity and enabled release of significantly higher levels of terpenoids than with a control strain from glycoside-rich grapevine extract in Muscat wine fermentations (Gil et al.2005). Interestingly, the levels of yeast-derived esters and higher alcohols were also enhanced significantly in the wines, similarly to what was observed upon overexpression of the ß-glucosidases from A. nidulans and Saccharomycopsis fibuligera (Ganga et al.1999; Van Rensburg et al.2005). This is consistent with differences observed in the yeast-derived ester profile in beers hopped with four different hop varieties (Steyer et al.2017), indicating an uncharacterized interaction between hop compounds and regulation of yeast aroma pathways.

Some of the non-conventional Brettanomyces yeasts possessed active BGL with the highest activity found in Brettanomyces custersii LD72, isolated from Lambic beer (Daenen et al.2008). In refermentation experiments with addition of amygdalin (cyanogenic glycoside present in the seeds of the sour cherry), B. custersii LD72 demonstrated higher release of sour cherry glycosides compared with Saccharomyces ale strains with low and high BGL-like activity (Daenen et al.2008). The strain B. custersii LD72 was further tested in different refermentation conditions with whole cherries, cherry pulp, cherry juice and cherry stones. Similar patterns of released linalool, α-terpineol, geraniol and α-ionol were observed in all conditions. Out of 428 different yeasts that were screened for BGL activity, B. anomalus YV396 and B. bruxellensis YV397 showed exceptionally high activities (Vervoort et al.2016). Their genomes were sequenced, and the β-glucosidase-encoding genes were codon-optimized and expressed in E. coli. The β-D-glucosidase from B. anomalus was thoroughly characterized for its technological and industrial application for instance fruity cherry beer production. Isolating Brettanomyces β-D-glucosidase enzymes and adding them during the brewing process without the Brettanomyces yeast can enhance the fruitiness of beer without production of the characteristic off-flavors that typically develop with Brettanomyces yeasts during fermentation (Vervoort et al.2016).