Yeast strains used in this study are listed in Table 1. Stock cultures were grown in YPD (10 g.L⁻¹ Bacto yeast extract, 20 g.L⁻¹ Bacto peptone and 20 g.L⁻¹ glucose) until stationary phase, supplemented with sterile glycerol [final concentration 30% (v/v)] and stored at −80°C as 1 ml aliquots until further use.

Chemostat and bioreactor batch cultures were grown on synthetic medium (SM) containing 3.0 g.L⁻¹ KH₂PO₄, 5.0 g.L⁻¹ (NH₄)₂SO₄, 0.5 g.L⁻¹ MgSO₄, 7 H₂O, 1 mL.L⁻¹ trace element solution, and 1 mL.L⁻¹ vitamin solution (Verduyn et al., 1990). A sugar mixture (Dried Glucose syrup C plus 01987, Cargill, Haubourdin, France; sugar content (w/w): 2.5% glucose, 28.0% maltose, 42.0% maltotriose, 26.4% higher saccharides) was added to SM to a final concentration of fermentable sugars of 20 g.L⁻¹ (SM-Mix). For cultivation on solid medium, SM or SM-Mix was supplemented with 2% (w/v) agar. SM used for anaerobic bioreactor cultivation was supplemented with ergosterol and Tween-80 (0.01 g.L⁻¹ and 0.42 g.L⁻¹, respectively; Verduyn et al., 1990) and with 0.15 g.L⁻¹ of antifoam C (Sigma-Aldrich, Zwijndrecht, The Netherlands). In all cases, the pH was adjusted to 5.0 with 1 M HCl. For assays of maltose and maltotriose transport, yeast cells were pregrown in YPM (10 g.L⁻¹ Bacto yeast extract, 20 g.L⁻¹ Bacto peptone and 20 g.L⁻¹ maltose).

For propagation and cultivation in E.B.C. (European Brewery Convention) tall-tubes (1977), strains were grown in aerated (18 ppm DO) 15° Plato industrial wort. Wort was produced from barley and wheat malt at the VTT Pilot Brewery (VTT Technical Research Centre of Finland Ltd, Espoo), and contained an extract of 15.0° Plato (62 g.L⁻¹ maltose, 22 g.L⁻¹ maltotriose, 16 g.L⁻¹ glucose, and 4.6 g.L⁻¹ fructose) and a free amino nitrogen (FAN) content of 269 mg.L⁻¹ (Gibson et al., 2013). Pilot fermentation experiments were carried out in 1,300 L stainless steel tanks filled with 1,000 L of 15° Plato industrial wort. Prior to inoculation, the wort was aerated with 25 ppm of O₂ and supplemented with zinc to a final concentration of 1.2 mg.L⁻¹. The culture was pitched with 1.0 × 10⁷ cells.mL⁻¹. Precultures were prepared by sequentially propagating cells from frozen stocks, into 50, 500 mL, 12, and 1,000 L pre-cultures on wort. All inoculum preparation stages were carried out at 20°C, except the last stage performed at 15°C.

Single cell lines were isolated from prolonged chemostat experiments by plating on solid SM-Mix and incubating plates anaerobically at 20°C. After three consecutive restreaks, one single-colony isolate from each evolution experiment was selected and characterized (Gonzalez-Ramos et al., 2016).

Optical density of cultures (OD₆₆₀) was determined with a Libra S11 spectrophotometer (Biochrom, Cambridge, UK) at 660 nm. Biomass dry weight was determined by filtering duplicate 10 mL culture samples over preweighed nitrocellulose filters with a pore size of 0.45 μm. Filters were washed with 10 mL demineralized water, dried in a microwave oven (20 min at 350 W) and reweighed. Off gas of bioreactor fermentations was cooled to 4°C in a condenser and CO₂ concentration was continuously monitored with an NGA 2000 analyser (Rosemount Analytical, Orrville, OH). For metabolite analysis, bioreactor culture samples were centrifuged and supernatants were analyzed with a Waters Alliance 2690 HPLC (Waters Co., Milford, MA) containing a Waters 2410 refractive-index detector, a Waters 2487 UV detector, and a Bio-Rad HPX-87H column (Bio-Rad, Hercules, CA) which was equilibrated and eluted with 0.5 g.L⁻¹ H₂SO₄ at 60°C at a 0.6 mL.min⁻¹ flow rate. Sugar concentrations in samples from tall tube fermentations were analyzed with a Waters HPLC with a Waters 2695 separation module and a Waters 2414 differential refractometer (Waters Co., Milford, MA) and an Aminex HPX-87H organic acid analysis column (Bio-Rad, Hercules, CA) equilibrated at 55°C and eluted with 5 mM H₂SO₄ in water at a 0.3 ml.min⁻¹ flow rate.

Cell count and viability were determined using a NucleoCounter® YC-100™ (Chemometec A/S, Allerod, Denmark) according to the manufacturer's recommendations. The specific gravity, alcohol content and pH in pilot-scale experiments and bottled beer were analyzed after filtration using an Anton Paar density meter (Anton Paar GmbH, Graz, Austria). Concentration of fermentable sugars and volatile compounds (acetaldehyde, acetone, ethylformate, ethylacetate, methanol, ethylpropanoate, propanol, isobutanol, isoamylacetate, ethylcaproate, dimethyl sulfide, diacetyl, and 2,3-pentanedione) in these samples were determined by ultra-performance liquid chromatography (UPLC) (Waters Co) and gas chromatography, respectively.

To determine if two sets of data were significantly different from each other unpaired two-sample Student's t-test was used in Prism 4 (version 4.03) (Graphpad software Inc., La Jolla, CA).

Sugar-uptake kinetics by yeast cell suspensions were determined with [¹⁴C] labeled maltose and maltotriose [0.1 mCi.mL⁻¹] (American Radiolabelled Chemicals, St Louis, MO). Prior to use ¹⁴C-labeled maltotriose was treated to remove impurities as described previously in Dietvorst et al. (2005). Transport rates were calculated from the radioactivity remaining inside the yeast cells after washing (Lucero et al., 1997). Yeast strains were pregrown in 100 mL YPM at room temperature and harvested at an OD₆₀₀ between 3 and 7. After centrifugation, yeast pellets were washed with ice-cold water, followed by washing with ice-cold 0.1 M tartrate-Tris buffer (pH 4.2) and resuspended to a cell concentration of 200 mg yeast wet weight.mL⁻¹ in the same buffer. Yeast suspensions were equilibrated to the 20°C assay temperature in a water bath. Experiments for maltose and maltotriose uptake rates were performed at different sugar concentration ranged from 1 to 25 mM. Sugar uptake was stopped after 60 s by addition of 5 mL ice-cold water and immediate filtration through a HVLP membrane (0.2 μm) (Millipore, Merck Life Science, Espoo, Finland). After rinsing the membrane with another 5 mL of ice-cold water, it was transferred to a scintillation cocktail (Optiphase Hisafe 3, PerkinElmer, Waltham, MA) and the radioactivity was counted in a scintillation counter (Tri-Carb 2810TR Low Activity Liquid Scintillation Analyzer, PerkinElmer). K_m and V_max were derived from Eadie-Hofstee plots fitting of the measured rates (Eisenthal and Cornish-Bowden, 1974). The statistical significance of observed differences between CBS1483 and IMS0493 strains was assessed by unpaired two-sample Student's t-test in Prism 4 (version 4.03) (Graphpad software Inc., La Jolla, CA).

DNA was isolated from strains S. pastorianus IMS0493 and CBS1483 as previously described in van den Broek et al. (2015). Whole-genome sequencing was performed by Novogene (HK) Company Limited (Hong Kong, China). A DNA library was produced with the TruSeq DNA PCR-Free Library Preparation kit (Illumina, San Diego, CA). Paired-end libraries with 354 and 370-bp inserts were prepared for strains CBS1483 and IMS0493, respectively, and sequenced with an Illumina HiSeq2500 sequencer (Illumina). A total of 4.3 Gb and 4.6 Gb of 150-bp paired-end fragments were generated for IMS0493 and CBS1483 respectively. Chromosome copy numbers were estimated using the Magnolya software tool (Nijkamp et al., 2012a; Oud et al., 2013). For visualization purposes, assembled contigs and their copy number were mapped on an illumina reads based-assembly of S. pastorianus CBS1483 (ASM80546v1) (van den Broek et al., 2015). The copy number estimation of individual maltose transporter genes was performed by concatenating sequences of SeMALT1 (GenBank Accession number: XM_018363333.1), SeMALT2 (XM_018364778.1), SeMALT3 (XM_018367005), SeMALT4 (XM_018368187.1), ScMAL11 (AJ012752.1), ScMAL31 (NM_001178646.1), ScMAL32 (NM_001178647.3), SNF5/YBR289W (Z36158.1), MSH3/YCR092C (M96250.1), MES1/YGR264C (Z73049.1), SeBRN1/SeYBL097W (XM_018363344.1) SePDA1/SeYER178W (XM_018364769.1) SeATF1/SeYOR377W (XM_), SeGTR1/SeYML121W (XM_018367018.1), SePLC1/SeYPL268W (XM_018368196.1) into an artificial contig onto 30.10⁺⁶ sequencing reads from IMS0493 or CBS1483 were mapped using Burrows Wheeler Aligner BWA (Li and Durbin, 2010) with default parameters. For each alignment 100 bp-average windows were calculated. For each gene 100-bp average windows data were collected and CBS1483 and IMS0493 data were statistically assessed for significant difference using unpaired two-sample Student's t-test in Prism 4 (version 4.03) (Graphpad software Inc., La Jolla, CA). Per gene average and standard deviation were calculated and plotted.

The raw sequencing data of strains IMSS0493, IMS0495 and IMS0508 are searchable at NCBI Entrez (http://www.ncbi.nlm.nih.gov/) under BioProject number PRJNA393253. The data of the parental strain CBS1483 can be found in BioProject PRJNA266750 (accession number SRP049726) (van den Broek et al., 2015).

In brewing, the term “attenuation” describes the extent to which brewing yeast completely converts wort extract into ethanol, CO₂, yeast biomass and flavor compounds. Optimal attenuation requires efficient consumption of glucose, fructose, maltose as well as maltotriose and is a highly sought-after phenotypic trait in industrial S. pastorianus strains. Despite their history of domestication in wort, many lager yeasts exhibit incomplete or slow fermentation of the two major wort α-oligosaccharides, maltose and maltotriose.

Since preliminary results indicated that the lager-brewing strain S. pastorianus CBS1483 could be an interesting model to study low attenuation due to suboptimal maltotriose fermentation kinetics in lager brewing strains, its growth and fermentation performance in 15° Plato wort was quantitatively analyzed in cylindrical 2-L stainless steel vessels at 15°C. Indeed, after 15 days of fermentation, residual concentrations of maltose and maltotriose remained at 3.2 ± 0.1 g.L⁻¹ and 14.2 ± 0.1 g.L⁻¹, respectively (Figure 1A). In these experiments, the fermentable sugars in wort were sequentially consumed. Glucose was fermented first, followed by maltose, while maltotriose consumption only began after 30 h. To investigate whether the incomplete utilization of maltose and maltotriose was caused by a limiting amount of available nitrogen in the wort, additional experiments with S. pastorianus CBS1483 were performed in stirred bioreactors on synthetic medium with excess nitrogen and lower overall sugar concentration (20 g.L⁻¹ of a sugar mixture containing 2.5% glucose, 28.0% maltose, 42.0% maltotriose and 26.4% higher dextrins). As observed in the wort fermentations (Figure 1A), a sequential utilization of the fermentable sugars was observed (Figure 1B and Figure S1). Maltotriose consumption slowed down severely when its concentration reached ca. 1 g.L⁻¹, and 0.25 g.L⁻¹ maltotriose was left at the end of fermentation.

These data indicated that, irrespective of the initial sugar concentration and nitrogen source availability, S. pastorianus CBS1483 could not completely consume maltotriose. This phenotype was stronger under the higher gravity fermentation conditions representative of industrial brewing. It was therefore decided to use strain CBS1483 as a model to explore an evolutionary engineering strategy for improving the kinetics of maltotriose fermentation.

The batch cultivation experiments on wort and on maltotriose-enriched sugar mixtures (Figure 1) clearly indicated suboptimal kinetics of maltotriose fermentation in S. pastorianus CBS1483. In particular, a deceleration of fermentation at low maltotriose concentrations contributed to a poor attenuation in these mixed-substrate cultures. The ability of microorganisms to maintain high biomass-specific substrate conversion rates (q_s) at growth-limiting concentrations of a substrate (C_s), is referred to as the affinity for that substrate. When growth kinetics obey the Monod equation (q_s = q_{s, max} (C_s/(C_s + K_s))), affinity can be defined as q_{s, max}/K_S (Button, 1985), in which q_{s, max} is the maximum biomass-substrate uptake rate and K_s is the substrate concentration at which q_s equals 50% of q_{s, max}. Nutrient-limited chemostat cultivation confers a strong selective advantage to spontaneous mutants with an improved affinity for the growth limiting nutrient. During prolonged chemostat cultivation, a gradual “take over” of cultures by mutants with an improved affinity for the growth-limiting nutrient is often reflected by a progressive decrease of its residual concentration in the cultures (for examples see Jansen et al., 2004, 2005). To explore whether this concept can be applied to improve the affinity and fermentation kinetics of S. pastorianus CBS1483 for maltotriose, four independent chemostat experiments were performed. In these carbon-limited chemostat cultures, which were grown at a dilution rate of 0.05 h⁻¹ and at 16°C, the carbon source consisted of a sugar mixture comprising 2.5% glucose, 28.0% maltose, 42.0% maltotriose, and 26.4% higher saccharides (w/w). Continuous cultivation was started when, in an initial batch cultivation stage on the same sugar mixture, maltotriose was the sole remaining carbon source in the bioreactor.

After 5 volume changes, concentrations of glucose and maltose in the four chemostat cultures were below detection level (<0.2 g.L⁻¹ and <0.1 g.L⁻¹, respectively), while maltotriose concentrations had not decreased below 3.0 g.L⁻¹ (Figure 2A and Table S1). After approximately 30 generations, the residual maltotriose concentration in all four chemostat experiments gradually decreased until, after 80 generations, a reduction of about 70% was reached. In chemostat experiment 7R (Figure 2, green symbols) the maltotriose concentration decreased from 3.0 g.L⁻¹ at the beginning of the culture to 0.56 g.L⁻¹ after 132 generations (Figure 2A). The higher utilization of maltotriose at the end of this fermentation experiment was accompanied by increases of the biomass and the ethanol concentrations by 17% (Figure 2B) and 13% (Figure 2C) respectively.

After 130 generations of selective chemostat cultivation on a sugar mixture, the fermentation performance of the evolved cultures could either reflect the characteristics of a mixed population or of a single, dominant enriched clonal population. To investigate whether pure cultures isolated from the evolution experiments exhibited an improved fermentation performance, single colonies were isolated from each evolution experiment. Four of these strains were grown on 15° Plato wort in cylindrical 2 L stainless-steel tall tubes, operated at 15°C. Single-colony isolates IMS0493, IMS0495, IMS0407, and IMS0508 clearly improved fermentation kinetics relative to their common parental strain CBS1483. In particular, residual maltotriose concentrations were significantly 2.8- to 3.7-fold lower (Student's t-test p-value = 0.0011) than observed for their common parental strain CBS1483 (Figures 1A, 3A, Figure S2). Although differences in residual maltose concentration were less pronounced and not statically significant (Student's t-test p-value = 0.5844), the evolved strains consistently showed a higher residual maltose concentration (18–66% increase relative to the parental strain). Of the four single-colony isolates, strain IMS0493 showed the best performance, with a 1.18-fold increase and a 3.7-fold decrease of the residual maltose and maltotriose concentrations, respectively, relative to S. pastorianus CBS1483 (Figures 1A, 3).

To test whether strain IMS0493 phenotypically resembled the evolving population from which it had been isolated, its growth and metabolite profile were compared with those of its parental strain CBS1483 in carbon-limited mixed-sugar (SM-Mix) chemostat cultures grown at a dilution rate of 0.05 h⁻¹ (Figure 4, Table 2). Chemostat cultures of strain IMS0493 rapidly reached a steady state in which extracellular metabolite profiles resembled those observed at the end of the evolution experiment, whilst strain CBS1483 showed the same high residual maltotriose concentrations that were observed at the beginning of the evolution experiments. These results confirmed that, during chemostat-based evolution, strain IMS0493 had acquired an improved affinity for maltotriose which contributed to lower residual sugar concentrations and higher biomass concentrations in steady-state cultures (Table 2) and better attenuation in batch cultures on wort (Figure 3). A faster and more complete conversion of maltotriose by strain IMS0493, relative to it parental strain CSB1483, was also observed in controlled bioreactor batch cultures, grown at 16°C on SM-Mix (Figure 3). In these cultures, CO₂ production profiles of the non-evolved strain CBS1483 exhibited two distinct peaks, reflecting diauxic utilization of first maltose and then maltotriose (Figure S1). In contrast, the CO₂ profile of the evolved isolate IMS0493 showed a single peak (Figure S1) which, as confirmed by analysis of sugar concentrations, reflected co-utilization of glucose, maltose and maltotriose (Figure 3 and Figure S1).

To investigate the role of maltotriose transport kinetics in the improved affinity of the evolved strain IMS0493 for this trisaccharide, sugar-uptake assays were performed at several [¹⁴C-] labeled maltose or maltotriose concentrations. An Eadie-Hofstee fit of the sugar rates revealed that the maximum maltose-uptake rates of the strain IMS0493 and CBS1483 were similar (Figure 5). Substrate-saturation constants (K_m) of strains IMS0493 and CBS1483 were also very similar for maltose (3.9 and 3.6 mM, respectively) but also for maltotriose (6.5 and 7.2 mM, respectively). In contrast, the V_max for maltotriose uptake of the evolved strain IMS0493 was over 4-fold higher (Student's t-test p-value = 0.03697) than that of the parental strain CBS1483 [23.5 and 4.9 μmol.min⁻¹.(g_{dry biomass})⁻¹ respectively; Figure 5]. The transport capacity for maltose of both strains was approximately 23 μmol.min⁻¹.(g_{dry biomass})⁻¹ (Figure 5). These results are consistent with an evolutionary adaptation to sugar-limited chemostat cultivation in which a higher affinity (q_{s, max}/K_m) for maltotriose primarily resulted from increased expression and/or an improved transport capacity (mol_maltotriose.(mol _transporter)⁻¹.s⁻¹) of one or more maltotriose transporters in the yeast plasma membrane.

The alloploid genomes of S. pastorianus strains, including the reference strain CBS1483 used in the present study (van den Broek et al., 2015), complicate identification of point mutations by short-read resequencing technologies. However, chromosome copy number analysis (Nijkamp et al., 2012a) indicated that, in the evolved strain IMS0493, a 2-fold amplification had occurred of a large part of CHRIII, while copy numbers of chromosomes ScI, ScVIII, SeI, SeIX and a segment of the right arm of ScXIV had decreased. The copy number of the affected part of CHRIII had increased from four copies in strain CBS1483 to eight copies in strain IMS0493 (Figure 6A). S. pastorianus CBS1483 carries only one version of this chromosome, which is composed of the left arm, centromere and a large part of the right arm of S. eubayanus CHRIII and the end of the right arm of S. cerevisiae CHRIII (van den Broek et al., 2015). The MAL2 locus, which is located on the subtelomeric region of the S. cerevisiae right arm of CHRIII, seemed not to be affected by the amplification. However, the copy number estimation of individual maltose transporter encoding gene was disrupted by the inability to accurately assemble paralogous genes (e.g., SeMALT2 and SeMALT4 or MAL31, MAL21, MAL41 and MAL61) with short sequencing reads. Thus, an attempt to get quantitative estimation of maltose transporter genes was performed by mapping sequencing reads from CBS1483 and IMS0493 on an artificial contig composed of the sequences of all four S. eubayanus maltose transporter genes (SeMALT1, SeMALT2, SeMALT3 and the truncated SeMALT4, Baker et al., 2015), two S. cerevisiae MALx1 genes (MAL11, MAL31,). To standardize the mapping results, control genes (SNF5/YBR289W, MSH3/YCR092C, MES1/YGR264C, SeBRN1/SeYBL097W, SePDA1/SeYER178W, SeATF1/SeYOR377W, SeGTR1/SeYML121W, SePLC1/SeYPL268W) located on chromosomes harboring a MAL gene and having no paralogs were selected and together with the highly conserved maltase gene MAL32 present in all S. cerevisiae MAL loci added to the artificial contig. The sequence reads mapping from IMS0493 and its ancestor CBS1483 on these concatenated sequences corroborated the amplification of a large fraction of CHRIII as coverage of MSH3 was deemed statistically different and the average 1.8-fold higher supporting a doubling of ScCHRIII. All other tested markers did not reveal any significant difference, suggesting that maltose transporter encoding genes were neither gained nor lost in the evolved strain IMS0493 (Figure 6B and Figure S3).

Laboratory evolution is a powerful tool to develop strains with improved, innovative traits. Although often performed in academic studies (Novick and Szilard, 1950; Ferea et al., 1999; Wisselink et al., 2009; Gonzalez-Ramos et al., 2013; Oud et al., 2013), these are rarely accompanied by assessment at pilot or full industrial scale. To investigate the industrial relevance of the evolutionary engineering strategy described above, the evolved strain IMS0493 and its parental strain CBS1483 were each tested in duplicate industrial pilot-scale (1,000 L) beer-fermentation experiments on a high-gravity 15° Plato wort (Figure 7). These experiments showed that the acquired phenotypes of the evolved strain were also expressed under industrial conditions. In particular, IMS0493 fermentation showed a significantly lower residual concentration of total fermentable sugars at the end of fermentation (Student's t-test p-value = 0.021944) (Table 3). Total fermentable sugars left in the green beer fermented by IMS0493 were 53% lower than in reference fermentations with its non-evolved parent strain (Table 3). This improvement was predominantly caused by a more complete conversion of maltotriose, whose residual concentration was reduced by 72% in comparison to residual concentration measured in beer fermented by CBS1483 (Table 3). Although residual maltose concentrations remained below 2 g.L⁻¹, they were 83% higher in experiments with the evolved strain.

To explore whether the improved maltotriose fermentation kinetics of evolved strain IMS0493 affected other brewing-related strain characteristics, beers from the pilot fermentations were analyzed. Consistent with the analysis on the 1,000-L fermentation experiments, the ethanol concentration in beer brewed with strain IMS0493 was 6% higher than in beer made with the reference strain, while the concentration of residual fermentable sugar was 48% lower (Student's t-test p-value = 0.043339 and 0.021944 respectively) (Table 3). These results confirmed that the increased fermentation efficiency seen during laboratory experiments could be transferred to an industrial setup (Huuskonen et al., 2010). This increase in ethanol concentration would account for a gain of 7.4% in volumetric productivity if the green beer were standardized to 5% (v/v) ethanol (Table 3). To investigate whether such standardized beers would comply with lager beer quality standards, aroma compounds were measured (Table 3). Total higher alcohols concentration in standardized bottled beer made with the evolved strain were 9% higher than in similar beer brewed with S. pastorianus CBS1483 but remained within specifications. Additionally, a 40% lower diacetyl concentration was observed in bottled beer produced with the evolved strain IMS0493 (Table 3). The level measured in IMS0493 (14 μg.L⁻¹) was far below the sensory threshold of 20 μg.L⁻¹. In comparison the level measured in the non-evolved CBS1483 were just above the threshold (23 μg.L⁻¹). Altogether, the aroma profile of beers produced with the two strains met common quality standards for lager beers.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. It worth mentioning that JG and NK are employed by HEINEKEN Supply Chain, Global Innovation & Research (Zoeterwoude, NL).