Improving truffle mycelium flavour through strain selection targeting volatiles of the Ehrlich pathway

Truffles (Tuber spp.) are the fruiting bodies of symbiotic fungi, which are prized food delicacies. The marked aroma variability observed among truffles of the same species has been attributed to a series of factors that are still debated. This is because factors (i.e. genetics, maturation, geographical location and the microbial community colonizing truffles) often co-vary in truffle orchards. Here, we removed the co-variance effect by investigating truffle flavour in axenic cultures of nine strains of the white truffle Tuber borchii. This allowed us to investigate the influence of genetics on truffle aroma. Specifically, we quantified aroma variability and explored whether strain selection could be used to improve human-sensed truffle flavour. Our results illustrate that aroma variability among strains is predominantly linked to amino acid catabolism through the Ehrlich pathway, as confirmed by 13C labelling experiments. We furthermore exemplified through sensory analysis that the human nose is able to distinguish among strains and that sulfur volatiles derived from the catabolism of methionine have the strongest influence on aroma characteristics. Overall, our results demonstrate that genetics influences truffle aroma much more deeply than previously thought and illustrate the usefulness of strain selection for improving truffle flavour.

Truffle fruiting bodies are heavily colonized by bacteria and to a lesser extent by yeasts, filamentous fungi and viruses 15 . The origin of many truffle odorants is hence unclear, as they might be synthesized by the truffle itself or by its microbiome. The contribution of microbes to truffle aroma has indeed only been demonstrated in a single case, which links bacteria inhabiting the fruiting body of T. borchii to the production of thiophene derivatives (i.e. 3-methyl-4,5-(2 H)thiophene) 7,12 . Additionally, indirect evidence suggests that bacteria might be exclusively responsible for the emission of 2,4-dithiapentane in T. magnatum as well 15 . By contrast, the origin of odorants that are more common among truffles (i.e. dimethyl sulfide, 2-methylbutan-1-ol and 3-methylbutan-1-ol and their aldehydes derivatives, just to cite a few examples) remains elusive since, theoretically, they may be synthesized by both truffles and their microbiomes 15 .
Uncertainties currently exist not only about the origins of truffle odorants, but also about their precursors. Indeed, the identity of aroma precursors in truffles is to a large extent based on speculation and indirect evidence. Following the sequencing of the genome of the black truffle T. melanosporum 16 , it has been suggested that numerous truffle odorants were produced from amino acid catabolism through the Ehrlich pathway. In this pathway, an amino acid is first deaminated into an α-keto acid, followed by a decarboxylation into an aldehyde and either a reduction or oxidation into an alcohol or acid, respectively 10,16,17 . In this way, leucine, isoleucine, phenylalanine and methionine are respectively transformed into 3-methylbutanal, 2-methylbutanal, 2-phenylacetaldehyde and 3-methylsulfanylpropanal and their corresponding alcohols and acids 17 . Supplying axenic cultures of truffle mycelium with leucine, isoleucine, phenylalanine and methionine was shown to induce numerous volatiles of the Ehrlich pathway [18][19][20] , suggesting that these amino acids were either the direct precursors of those volatiles or indirectly induced them. Nevertheless, demonstrating the existence of the Ehrlich pathway beyond a reasonable doubt shall ultimately require feeding experiments with isotopically-labelled amino acids, which has not been performed to date.
Differences in the aroma profiles of truffles do not only exist among species and/or developmental stages as highlighted earlier, but also within truffles of the same species. Indeed, a major variability in the concentration of four and eight carbon-containing volatiles (i.e. oct-1-en-3-ol and 2-butanone) has been documented for Tuber aestivum fruiting bodies collected a few centimeters apart in the same truffle orchards 21,22 . This aroma variability has been linked to genetic differences 21,22 . Considering that strain selection has been successfully performed in microbes to improve the final flavour of fermented food products (i.e. cheese and wine) or to eliminate off-flavour compounds [23][24][25][26] , a similar approach might improve the characteristics of truffle flavour produced through axenic truffle mycelium cultures.
The main aim of this study was to assess whether strain selection could be used for improving human-sensed truffle flavour produced through mycelial fermentation. To answer this, nine strains of the white truffle T. borchii were tested in various feeding experiments and sensory tests (see Fig. 1 for the experimental design). This species, endemic to Europe and introduced in New Zealand 5 was chosen because of the good growth of its mycelium compared to other truffle species 27 . Specifically, in the first set of experiments, the extent of aroma variability among strains was assessed and volatiles responsible for this variability were identified. Subsequently, the existence of the Ehrlich pathway was tested with isotopically labelled amino acids and finally, sensory tests were performed to assess whether the human nose was capable of differentiating among strains.

Results
T. borchii mycelia vary in their volatile profiles. The aim of this first experiment was to assess the variability in aroma profiles of mycelial cultures of T. borchii. For this purpose, the aroma profiles of nine strains (axenic cultures grown in malt extract) and one fruiting body (included here for completeness) of T. borchii (Table 1) were analysed by solid-phase microextraction gas chromatography-mass spectroscopy (SPME-GC/MS). Volatile profiles, generated for three independent replicates of all strains/fruiting body, were processed for peak realignment with the Tagfinder software 28 . This resulted in a data matrix of specific TAGs (volatiles) in each sample (Supplementary Table S1 illustrates the raw data of TAGs normalized to the total ion current (TIC)). Principle Component Analysis (PCA) applied to the volatile profile in Supplementary Table S1 could explain 59% of the data variability as seen in Fig. 2A, highlighting differences among strains.
TAGs which significantly differed in concentrations among strains were identified using the non-parametric Kruskal-Wallis statistical test performed in R 29 . Those TAGs could be assigned to 29 compounds, which are represented in a heat map in Fig. 2B. These include one sulfur-containing volatile, alkenes, alcohols, aldehydes, ketones, aromatic compounds and eight unidentified volatiles. Both qualitative and quantitative differences were detected among strains. Interestingly, one-third of the volatiles that varied in concentrations among strains might be products of the Ehrlich pathway. Common volatiles produced by most T. borchii strains included 2-and 3-methylbutanal, benzaldehyde, 2-phenylethan-1-ol whereas other volatiles such as 2-phenylacetaldehyde, 2-and 3-methylbutan-1-ol and 2-methylpropan-1-ol were specific to two or three strains and 3-methylsulfanylpropanal was detected from a single strain (strain 2) only (Fig. 2B).
Our data demonstrated that the largest part of aroma variability among strains was due to quantitative differences in volatiles possibly derived from amino acids catabolism (Ehrlich pathway).
Confirming the existence of the Ehrlich Pathway in T. borchii through feeding mycelia with 13 C labelled amino acids. Considering the important variability in volatiles possibly derived from amino acid catabolism, we tested the existence of the Ehrlich pathway in truffles. With this aim in mind, three mycelial strains (strains 2, 3 and 5) were supplied with four amino acids separately, namely leucine, isoleucine, phenylalanine, and methionine (unlabelled and 13 C labelled) and their aromas were profiled by pressure-balanced headspace-GC/MS. Pressure-balanced headspace extraction and subsequent trapping on a charcoal cartridge was chosen in favour of SPME for its improved reproducibility. In all strains, leucine induced 3-methylbutanal and 3-methylbutan-1-ol; isoleucine: 2-methylbutanal and 2-methylbutan-1-ol; phenylalanine, 2-phenylacetaldehyde, ScIentIFIc RepoRts | (2018) 8:9304 | DOI:10.1038/s41598-018-27620-w 2-phenylethan-1-ol, and benzaldehyde; methionine, 3-methylsulfanylpropanal, dimethyl disulfide (DMDS), dimethyl trisulfide (DMTS) (Fig. 3). Comparison of the mass spectra of all three mycelial strains supplemented with 13 C labelled and unlabelled amino acids gave comparable results for all strains and showed that all the labelled carbon atoms were fully incorporated in target volatiles ( Fig. 3 and Supplementary Fig. S1), hence confirming the existence of the Ehrlich pathway in T. borchii.

Testing the induction of Ehrlich volatiles in T. borchii.
Having confirmed the existence of the Ehrlich pathway in truffles, and documented an important concentration variability in resulting volatiles, we next questioned whether strains differed in their ability to produce those volatiles in the presence of amino acids. With this aim, nine mycelial strains were supplemented with non-labelled amino acids (leucine, isoleucine,   Sensory tests with T. borchii mycelial strains. Considering the differences in aroma profile exemplified among strains of T. borchii in our earlier experiments, we questioned if these differences would impact human-sensed aroma perception. Indeed, to have a possible impact on the overall aroma, a volatile compound needs to be detectible by the human nose (some volatile compounds are odourless) and be present above its detection threshold. A total of three sensory tests were performed to address various questions. The first test was carried out to investigate the ability of the human nose to perceive induced volatiles in T. borchii strains upon amino acid supplementation. A second test aimed at exploring the capability of the human nose to differentiate between aromas induced by a single amino acid or a mixture using a single strain. In the third test, the ability of the human nose to distinguish among strains supplemented with the same amino acid was examined. All tests were performed by 14 panellists and consisted of a triangle test (discrimination test) and an assessment of aroma attributes as described in the material and methods section.
Induced sulfur containing volatiles are perceived by the human nose. In the first test, mycelial strain 2 was supplemented with methionine or water (control). The result of the triangle test illustrated that the panellists were able to distinguish between unsupplemented and methionine supplemented samples (p = 0.02 < α, one-sided binomial proportions test with α = 0.05), hence confirming that sulfur containing volatiles (hereafter "sulfur volatiles") of the Ehrlich pathway were perceived by the human nose.
In terms of aroma attributes, differences between samples could be ascribed to sulfur and fermented/roasted notes that were highest in methionine supplemented samples and floral notes that were the highest in unsupplemented samples (Fig. 6A). Samples also differed in terms of intensity with methionine supplemented samples being the most intense (Fig. 6A). The samples were further analysed by pressure-balanced headspace-GC/ MS to relate volatile profiles to the aroma impressions of the panellists (Fig. 6A). The marked sulfurous/garlicky notes reported by the panellists in the methionine supplemented sample, corresponded to the induction of sulfur volatiles, whereas the flowery notes of the unsupplemented sample corresponded to the induction of 2-phenylacetaldehyde and 2-phenylethan-1-ol. A second triangle test was performed to estimate the overall importance of sulfur volatiles compared to other (non-sulfur) aroma compounds. Essentially, the aromas induced in a single strain (strain 2 as used in the first trial) by methionine were compared to those induced by a mixture of amino acids that included methionine and also leucine, isoleucine and phenylalanine. The triangle test of the second trial illustrated that the human nose was unable to distinguish among samples supplemented with methionine only compared to those supplemented with a mixture of amino acids (that included methionine as well) (p = 0.86 > α, one-sided binomial proportions test with α = 0.05). Similarly, no significant difference in overall aroma intensity among the samples was detected (Fig. 6B). However, quantification of volatiles by pressure-balanced headspace-GC/MS highlighted the marked differences in DMDS and benzaldehyde, that were significantly higher in the sample supplemented with the amino acid mixture (Fig. 6B).
The human nose can differentiate among strains. A third triangle test was performed to examine the ability of the human nose to distinguish between strains 3 and 9 supplemented here with phenylalanine only.
The triangle test results demonstrated that the human nose could differentiate between strains supplemented with phenylalanine (p = 0.004, one-sided binomial proportions test, α = 0.05). These findings were corroborated by the sensory test since higher floral notes were attributed to strain 3 (Fig. 6C). Profiling of volatiles revealed that the concentrations of six volatiles significantly differed among strains 3 and 9, with 2-phenylacetaldehyde, 2-phenylethan-1-ol and benzaldehyde being the highest in strain 3 (p < 0.05, Kruskal-Wallis test with α = 0.05) (Fig. 6C).

Discussion
Genetics is a powerful tool to improve flavour and aroma of food products and/or eliminate off-flavours [23][24][25][26] . The present study demonstrated how strain selection could be used to improve truffle mycelium flavour. Type of strain however is not the only means of flavour improvement in truffles since Xiao et al. 20 illustrated that desirable sulfurous, mushroom and earthy aromas could also be induced in the black truffle T. melanosproum by repeated freeze-thaw cycles.
The aroma emitted by axenic cultures of three truffle species (T. borchii, T. melanosporum, T. formosanum) have been described earlier by various groups 13,14,19,20,30,31 , yet none of those studies investigated aroma variability within the same species. Here, we used nine strains of T. borchii for the latter purpose. Of the 21 volatiles identified from T. borchii mycelia grown on malt extract (Fig. 2B), only five have been previously reported from two T. borchii strains including 3-methylbutan-1-ol, oct-1-en-3-ol, 2-phenylethan-1-ol, phenylmethanol and 2,4-bis(1,1-dimethylethyl)-phenol 13,14 . Tirillini et al. 14 could nevertheless detect 28 additional volatiles including dimethyltrisulfide through dynamic headspace sampling. Similarly, ten volatiles reported in Fig. 2B have been described from axenic cultures of T. melanosporum or T. formosanum, which also produced numerous volatiles not reported here 30,31 . These similarities and discrepancies in volatile emission among our results and literature might be attributed to various factors. For example, cultural conditions might have a major impact on fungal volatiles 32,33 . In the case of axenic cultures of different truffle species (T. melanosporum, T. sinense, T. indicum, and T. aestivum), Tang et al. 34 demonstrated that cultural conditions indeed had a stronger influence on the composition of emitted volatiles than species itself. Besides media composition, volatile sampling techniques might also influence volatile profiles 35 , and so does sample processing (i.e. freezing) 1,20,36 . Nevertheless, considering that in our study, all strains of T. borchii were handled in the same way, the aroma variability observed among strains ( Fig. 2A,B) was without a doubt attributed to genetic differences among strains.
The influence of genetics on truffle aroma had been reported earlier in T. aestivum 21,22 fruiting bodies for specific groups of volatiles, namely four (C 4 ) and eight (C 8 ) carbon-containing compounds (i.e. C 8 : oct-1-en-3-one, oct-1-en-3-ol and t-2-octenal; C 4 : butan-2-one and butan-2-ol). Our data support the influence of genetics on the 29 volatiles reported in Fig. 2B, and particularly on volatiles of the Ehrlich pathway, which could not be shown in earlier work by our group 21,22 . This apparent contradiction between our earlier results and those presented here can be explained by differences in study systems (axenic mycelial cultures versus fruiting bodies). Indeed, fruiting bodies are heavily colonized by microbes (i.e. bacteria and yeasts), which, in addition to truffles, are able to produce volatiles from the Ehrlich pathway 15 . The production of the Ehrlich-derived volatiles 2-methylbutan-1-ol, 3-methylbutan-1-ol along with some sulfur volatiles (3-methylsulfanylpropanal, dimethyl sulfide (DMS), DMDS and DMTS) have been reported from yeasts directly isolated from truffle fruiting body 37 . Besides microbes, other factors might influence fruiting body aroma (i.e. maturation 38 , geographical location 39 ). Many of those factors often co-vary (i.e. genetics and geography), thus using fruiting bodies for investigating the factors behind aroma variability in truffles can only lead to limited conclusions.
The Ehrlich pathway, first described in the budding yeast Saccharomyces cerevisiae, consists of a three-step process, in which an amino acid is first deaminated into a α-keto acid, followed by a decarboxylation into an aldehyde and a reduction or oxidation into an alcohol or acid 17 . We demonstrated here the existence of this pathway in T. borchii using amino acids with 13 C labelled carbons, yet, proving the origin of non-carbon atoms (i.e. sulfur, oxygen) was beyond the scope of our study. According to the "conventional Ehrlich pathway", phenylalanine leads to two volatiles, 2-phenylacetaldehyde and 2-phenylethan-1-ol 17 . In our data (Fig. 3), the incorporation of 13 C labels into a third volatile (benzaldehyde) hints at possible pathway differences between Saccharomyces yeasts and truffles. Considering that benzaldehyde might be produced from phenylalanine through both enzymatic and chemical reactions in some bacteria 40 , suggests that this might similarly occur in truffles. Identifying the exact pathway in truffles shall be the focus of further studies.
Methionine catabolism leads to the formation of sulfur volatiles through two hypothetical pathways for which candidate enzymes have been identified in the genome of T. melanosporum 10,16 . Indeed, in the "standard" Ehrlich pathway, methionine might be converted into 4-methylthio-2-oxobutyric acid (KMBA) via transamination. Subsequently, KMBA decarboxylation might lead to 3-methylsulfanylpropanal which might further non-enzymatically decompose to methanethiol (MTL). The latter might also be produced from KMBA through both chemical and enzymatic reactions by C-S lyase enzymes (i.e. cystathionine βand γ-lyases). In the alternative "non-standard" Ehrlich pathway, methionine might directly be converted into MTL by enzymatic reactions (i.e. methionine γ-lyase). In both standard and non-standard pathways, MTL can spontaneously oxidize to DMS, DMDS and DMTS 10 . The induction of 3-methylsulfanylpropanal observed in our data (Fig. 3) supports the existence of at least the standard Ehrlich pathway for methionine catabolism in T. borchii. By contrast, the absence of DMS, a typical sulfur volatile of many truffle species 10 from our mycelial cultures might be interpreted in two ways. On one side, DMS might not be induced by the cultural conditions used here, and different conditions might have favoured it as reported for T. melanosporum and T. formosanum 30,31 . On the other side, DMS in truffle fruiting bodies might actually be exclusively produced by bacteria (and not by the truffle mycelium), as demonstrated for some cyclical sulfur volatiles in T. borchii 12 . Either hypothesis will require a more detailed characterization of methionine catabolism pathway to be tested.
Volatiles produced through the Ehrlich pathway are well-known food odorants 6 . For example, taken as pure compounds, 2-phenylethan-1-ol has a rose-like odour, 3-methylbutan-1-ol is reminiscent of cheese and whiskey and sulfur volatiles (i.e. DMS, DMDS, DMTS and 3-methylsulfanylpropanal) smell like garlic, rotten food, and cooked potatoes 8,15 . Aroma perception of specific food depends on the interplay among odorants and on their respective concentrations 6 , yet, typically, only a small fraction of all volatiles contribute to human-perceived aroma. Our sensory tests illustrated the importance of methionine in the formation of sulfur containing odorants in axenic truffle cultures. Indeed, sulfur volatiles are also major odorants in truffle fruiting bodies 7,8,15 . In our sensory tests, panellists predominantly used sulfur/garlicky and fermented/roasted notes to describe the aroma of mycelium supplemented with methionine. Even if not determined here, this strongly suggests that the concentration of sulfur volatiles were higher than their detection threshold (odour activity value (OAV) > 1) in methionine supplemented samples. This is further supported by the fact that sulfur volatiles tend to have much lower odour thresholds compared to other volatiles of the Ehrlich pathway (i.e. odour threshold in water (part per billion) of sulfur volatiles DMDS (0.    Panellists could further differentiate among mycelial strains supplemented with the same amino acids. This illustrates that genetics does not only influence volatile production in truffles as highlighted earlier, but that those differences have an impact on the overall aroma characteristics. Similar findings of strains influence on aroma characteristics have been also demonstrated for strawberries and baker's yeast 46,47 , among numerous other examples. Specifically, in domesticated strawberries, an insertional mutation was shown to result in flavour loss compared to wild strawberries 46 . Similarly, some non-conventional yeast strains markedly vary in aroma profile (i.e. concentrations of 2-phenylacetaldehyde and 2-phenylethan-1-ol) compared to the S. cerevisiae strain traditionally used in bakery 47 .
Our findings illustrate how strain selection might be used to identify truffle strains with more desirable aroma and flavour. This is specially promising since T. borchii is the first species that has been successfully used to mycorrhize seedlings with axenic mycelial cultures instead of ascospores 48 . Overall, our results might also lead to the production of better quality truffle flavour and a higher consumer acceptance of truffle-flavoured food products (i.e. truffle-flavoured olive oil) that currently predominantly contain synthetic flavours.

Materials and Methods
Biological material. A total of 9 strains (axenic cultures) of T. borchii and one fruiting body (FB) were used in this study as described in Table 1.
For preparation of liquid cultures, T. borchii mycelia were pre-grown on malt extract (ME) agar (pH 7.0) [formula used per liter: malt extract 10 g (Difco & Becton Dickinson, Heidelber, Germany) and agar 15 g (Carl Roth, Karlsruhe, Germany)] for two months as described by Splivallo et al. 21 , and half a colony transferred to 100 ml Erlenmeyer flasks containing 30 ml of ME broth (1%, pH 7.0). Liquid cultures were homogenized with a sterile grinder at 17,000 rpm (T 18 digital ULTRA-TURRAX, IKA, Staufen, Germany) for 10 s, and incubated at 23 °C for two months. Sample handling for subsequent experiments is illustrated in Fig. 1. along with the number of replicates used for each experiment.

Comparing the volatile profiles of different T. borchii mycelial cultures.
Two-month-old mycelial cultures and malt extract broth without mycelium (negative controls) were homogenized at 17,000 rpm as described above, an aliquot of 5 ml in a 50 ml tube was subsequently pelleted by centrifugation (12,000 g for 10 min at 4 °C). The supernatant was discarded and a biomass of 200 ± 0.002 mg (wet weight) was transferred to a 20 ml SPME vial sealed with a screw cap and a silicon/polytetrafluoroethylene (PTFE) septum (VWR, Germany). Medium control samples were prepared by transferring 100 µl of malt extract to SPME vials. One fruiting body sample (originally frozen, but allowed to equilibrate to room temperature for 2 h before analysis) was also included for volatile analysis (300 ± 0.002 mg per SPME vial).
Volatiles extraction was performed by SPME in an autosampler (PAL RSI 85, CTC Analytics AG, Switzerland). Samples were first preheated to 60 °C for 20 min prior to extraction. Extraction was then done with an SPME fibre (PDMS/DVB/CARB Agilent Technologies, Waldbronn, Germany) exposed for 15 min in the headspace of the sample. Empty SPME vials were regularly inserted between samples to make sure no carry-over occurred. Volatiles were profiled by GC/MS (Agilent 7890B GC system equipped with a 5977B quadrupole MS detector (Agilent Technologies, Waldbronn, Germany). The SPME fibre was desorbed in the GC inlet (250 °C) in splitless mode and volatiles were separated on a capillary column (HP-5MS Agilent 19091S-433UI 0.25 mm × 30 m × 0.25 μm) by using the following program: start at 40 °C and hold for 5 min, ramp at a rate of 3 °C min −1 to 160 °C, ramp at a rate of 50 °C min −1 to 260 °C, hold for 1 min (total run time: 53 min). The carrier gas was helium with a flow rate of 1.2 ml min −1 . MS parameters were adjusted to a mass-to-charge ratio (m/z) scan range from 50 to 350, in an etune mode, ion M + , with electron energy of 70 eV (MS source 230 °C, MS quad 150 °C).
Chromatograms were visualized with the Agilent Mass Hunter Qualitative Analysis software (version: B.07.00). To eliminate the shift in retention time due to the machine drifts, peaks were aligned using Tagfinder software version 4.1 28 . The intensity threshold was set to 3,000 but other parameters were the same as previously described in Sherif et al. 49 . As an output, Tagfinder created a matrix with sample names in columns, and TAGs (specific m/z values within a specific retention time window) in rows. Background noise (three times the values of the medium control samples (ME without mycelium)), was subtracted from all samples and TAGs for each sample were normalized by dividing their intensity by the total ion current (TIC) (Supplementary Table S1). Based on Supplementary Table S1, principal component analysis (PCA) was produced using the Past software version 3.04 50 . To create the heat map in Fig. 2B from Supplementary Table S1, a non-parametric test in R (version 3.2.3) 29 was used to identify TAGs that significantly differed among samples (p < 0.05, Kruskal-Wallis with α = 0.05). Moreover, to avoid a biased representation in the heatmap towards tags with the highest intensities, TAG in single rows were divided by their maximum TAG intensity in that row. This resulted in relative concentrations between 0 and 1.

Confirming the Existence of the Ehrlich Pathway in T. borchii with 13 C labelled amino acids.
Supplementation experiments with 13 C isotope labelled amino acids were performed with three mycelial strains (2, 3 and 5 in Table 1 After washing, mycelial pellets were dissolved in 5 ml minimal medium and 1.5 ml of this culture was transferred to an SPME vial. Vials were supplemented with 0.5 ml of an aqueous solution (sterile filtered) containing 13 C labelled L-methionine, L-leucine, L-isoleucine and L-phenylalanine (labels are shown in Fig. 3, final concentration of 5 mM in SPME vials) (Cambridge isotope laboratories, Andover, USA). Samples were incubated at room temperature on a 100 rpm orbital shaker (Neolab, Heidelberg, Germany) for 24 h in the dark. After incubation, mycelial volatiles were profiled by pressure-balanced headspace-GC/MS as described hereafter.
Volatiles were profiled using a Clarus 680 GC coupled to quadrupole Clarus SQ 8 C MS detector (PerkinElmer, MA, USA) equipped with a pressure-balanced headspace-autosampler (Headspace Sampler TurboMatrix 40, Perkin Elmer, MA, USA). Volatiles were extracted by pressure-balanced headspace in trap mode using the following program: 20 ml SPME vials sealed with a septum were pre-heated for 20 min at 70 °C (to increase the volatility of the analytes), then the vials were pressurized with helium (to 105 kPa for 2 min). 4-Bromofluorobenzene (100 µl of a 27.2 ppm IS in N2 (mol/mol)) was used as an internal standard (IS). The IS gas was injected into the vials prior to volatile adsorption by the trap (TurboMatrix air monitoring trap, Perkin Elmer) (loop load: 0.5 min, loop equilibration: 0.4 min, inject time 0.5 min). Volatiles were thermally desorbed from the trap and injected in the GC by increasing the temperature from 30 °C to 280 °C. Separation was performed on a capillary column Elite-5MS (30 m × 0.25 mm i.d., 1.00 μm film thickness, Perkin Elmer). The GC oven temperature program was: start at 50 °C and hold for 1 min, 15 °C min −1 to 180 °C and hold for 1 min, 100 °C min −1 to 300 °C and hold for 5 min (total run time: 16.87 min). The carrier gas was helium with constant pressure at 75 kPa. MS operated electron impact ionisation (EI) conditions (70 eV) with scan range from m/z 50-300 (MS source 200 °C). Turbo Mass software (version: 6.1.0, PerkinElmer) was used to visualize the chromatograms and the mass spectrums. Mass spectra of mycelial strains supplemented with 13 C labelled amino acids was compared to the mass spectra of non-labelled amino acids supplemented mycelia to inspect whether the labels were integrated into the targeted volatiles (see Supplementary Fig. S1).

Testing the induction of Ehrlich volatiles in T. borchii. The supplementation experiment with
non-labelled amino acids was carried out using all the nine mycelial strains of Table 1. Mycelial cultures and controls with no mycelium in ME broth were homogenized with a sterile grinder and were divided into aliquots of 2 ml in SPME vials. Vials were supplemented with 0.4 ml of an aqueous solution containing non-labelled methionine, leucine, isoleucine and phenylalanine (sterile filtered, with a final concentration of 20 mM in SPME vials) (Carl Roth, Karlsruhe, Germany) or water (control). Samples were kept at room temperature on a 100 rpm orbital shaker for 24 h in the dark. Mycelial volatiles were profiled as for the samples containing labelled amino acids.
To determine the mycelial biomass, 3 ml of each homogenized culture was transferred to 50 ml Falcon tube, the mycelium was separated by centrifugation from its supernatant and wet weight was recorded. Peak areas (PA) of target volatiles were subsequently normalized to the peak area of the internal standard and to the biomass of each sample (Fig. 5).
Sensory tests with T. borchii mycelial strains. Three sensory tests were performed in total, aimed at various objectives. A first test was done to evaluate if the human nose perceived induced volatiles in T. borchii strains supplemented with methionine (comparison: strain 2 either unsupplemented (water control)) or supplemented with 5 mM methionine. The second test was performed to investigate whether the human nose could differentiate between the aromas induced by methionine (strain 2, 5 mM) or a mixture of amino acids (strain 2 and leucine, isoleucine, phenylalanine, and methionine, each 5 mM) in the same strain. A third test assessed the ability of the human nose to distinguish between strains (strains 3 and 9, both supplemented with 5 mM phenylalanine).
For all the sensory tests, panellist (14 individuals, 6 women and 8 men) were presented with a questionnaire comprising two parts: a discriminatory test ("triangle test" as described in O'Mahony et al. 51 ) and assessment of aroma attributes. For the triangle test, each panellist was presented with three samples (two identical and one "different") and asked to identify the two samples judged the most similar by their smells. Results of the triangle test were analysed using a one-sided binomial proportions test 52 . For the assessment of aroma attributes, each panellist was presented with four samples (two replicates of each sample). They were asked to indicate the presence of the following aroma attributes; sulfur/garlic, mushroom, floral, straw/malt, fermented/roasted, followed by rating the "overall aroma intensity" using a four-point category scale (0 indicating 'no smell' and 3 'very intense smell'). These aroma attributes were provided to the assessors as representative aroma attributes typically used to describe the volatiles of Fig. 5. For each aroma descriptor, frequencies were expressed as a percentage of total possible counts (14 panellist × 2 replicates = 28 counts), yielding values between zero and one. Volatile profiles from the samples used for sensory analysis were generated by pressure-balanced headspace -GC/MS as previously described.
Ethics statement. Use of human subjects for this study was reviewed by the Ethics Committee of the medical department of the University of Frankfurt and was granted exempt status. Informed consent was obtained from all participants. Data Availability. The data generated or analysed during this study are included in this published article (and its Supplementary Information files).