Genetic engineering of brewing yeast to reduce the content of ethanol in beer


  • Elke Nevoigt,

    Corresponding author
    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
      *Corresponding author. Tel.: +49 (30) 31 42 75 97, or +49 (30) 45 08 02 78; Fax: +49 (30) 31 42 75 92, E-mail address:
    Search for more papers by this author
  • Rita Pilger,

    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author
  • Edeltraud Mast-Gerlach,

    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author
  • Ulrike Schmidt,

    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author
  • Silke Freihammer,

    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author
  • Martin Eschenbrenner,

    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author
  • Leif Garbe,

    1. Institut für Biotechnologie, Fachgebiet Chemisch-technische Analyse, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author
  • Ulf Stahl

    1. Institut für Biotechnologie, Fachgebiet Mikrobiologie und Genetik, Technische Universität Berlin, Seestrasse 13, D-13353 Berlin, Germany
    Search for more papers by this author

*Corresponding author. Tel.: +49 (30) 31 42 75 97, or +49 (30) 45 08 02 78; Fax: +49 (30) 31 42 75 92, E-mail address:


The GPD1 gene encoding the glycerol-3-phosphate dehydrogenase was overexpressed in an industrial lager brewing yeast (Saccharomyces cerevisiae ssp. carlsbergensis) to reduce the content of ethanol in beer. The amount of glycerol produced by the GPD1-overexpressing yeast in fermentation experiments simulating brewing conditions was increased 5.6 times and ethanol was decreased by 18% when compared to the wild-type. Overexpression of GPD1 does not affect the consumption of wort sugars. Only minor changes in the concentration of higher alcohols, esters and fatty acids could be observed in beer produced by the GPD1-overexpressing brewing yeast. However, the concentrations of several other by-products, particularly acetoin, diacetyl and acetaldehyde, were considerably increased.


The reduction of ethanol in alcoholic beverages, especially beer, is of great commercial interest. Consumer demand for these beverages is continuously increasing due to both increased awareness for health and stricter laws regarding drinking and driving.

Current methods producing low-alcohol beer, i.e. manipulated fermentation or post-fermentation removal of ethanol, result either in a worty taste or a loss of aroma components [1]. In addition, the elimination of ethanol based on distillation or dialysis is expensive and labour-intensive.

An alternative technique to produce beer with a reduced ethanol content could be established by providing the breweries with a genetically modified yeast strain that forms less ethanol during complete fermentation of wort sugars. The reduction of ethanol production could be achieved by metabolic engineering of the carbon flux in yeast resulting in an increased formation of other fermentation products such as glycerol. However, only by-products that do not disturb the taste of beer are acceptable. It is worth mentioning in this context that the full-bodied character of low-alcohol beer was improved when it was supplemented with glycerol [2].

Glycerol is the most important by-product when yeast ferment sugar to ethanol. It is synthesized by reducing the glycolytic intermediate dihydroxyacetone phosphate (DHAP) to glycerol-3-phosphate (G3P) that is catalysed by the glycerol-3-phosphate dehydrogenase (GPD), followed by dephosphorylation of G3P by glycerol-3-phosphatase (GPP). Two isoenzymes of GPD encoded by the genes GPD1 and GPD2 have been characterized in Saccharomyces cerevisiae[3–6]. The isoenzymes of GPP are encoded by GPP1 and GPP2[7,8].

Glycerol production during sugar catabolism plays an important biological role in maintaining cytosolic redox balance, especially during anaerobic growth [6,9–11]. As ethanol formation from sugar is redox-neutral, the cytosolic NADH produced in surplus during other cellular processes such as biomass production and excretion of oxidized metabolites [9,11–13] is reoxidized by GPD during glycerol formation.

Glycerol is also involved in the osmoregulation of S. cerevisiae and acts as the main compatible solute in this organism [14]. Osmotic stress results in glycerol overproduction mainly controlled by induction of GPD1 (for reviews refer to [15–17]).

A shift of carbon flux towards glycerol at the expense of ethanol formation in yeast was achieved by simply increasing the level of GPD [18–21]. For example, the amount of glycerol produced by a laboratory yeast strain overexpressing the GPD1 gene was increased 8.1 times while only 63% of ethanol was formed compared to the reference strain. Hence, overproduction of GPD seemed to be a suitable approach to create yeast strains for producing low-alcohol beer. Here, we describe the overexpression of GPD1 in lager brewing yeast and the characterization of this strain in fermentation experiments simulating brewing conditions.

2Materials and methods

2.1Strains and culture conditions

Industrial lager brewing yeast strain Sa-07256 (S. cerevisiae ssp. carlsbergensis) was provided by the strain collection of the ‘Institut für Gärungsgewerbe Berlin’. Laboratory S. cerevisiae strain YSH 1.1.-6B (MATα leu2-3 leu2-112 ura3-52 trp1-92) [18] was employed to compare plasmids YEpKmGPD1 and YEpTKmGPD1 with regard to overproduction of GPD. Yeast strains were maintained and grown in YPD medium (2% peptone, 1% yeast extract and 2% glucose).

Escherichia coli NM522 was used for cloning experiments. Cultivation of E. coli and preparation of media were performed as previously described [22].

2.2Construction of plasmids

The plasmids YEpTKm and YEpTKmGPD1 were obtained from YEpKm and YEpKmGPD1[18], respectively. The kanamycin resistance gene (Km) controlled by the original bacterial promoter was replaced by the Km linked to a strong yeast promoter (TEF1) using the BamHI sites. The first fragment was obtained from the plasmid pUCTK23 [23].

The plasmid YEpTKmA-GPD1 used to overexpress the GPD1 gene controlled by the ADH1 promoter is a derivative of the plasmid YEpTKm. An EcoRV fragment of the plasmid pGEM-Ex-A-GPD1 (Mast-Gerlach, unpublished data), containing the promoter of the S. cerevisiae ADH1 gene, the coding sequence of GPD1 and the terminator of the S. cerevisiae TRP1 gene, was inserted into the EcoRV site of YEpTKm.

2.3Transformation of yeast

Yeast was transformed by the electroporation method [24] which was slightly modified. The yeast culture was diluted to 2×106 cells ml−1 and further grown in a culture volume of 400 ml until a cell concentration of approximately 6×106 cells ml−1 was obtained. The amount of plasmid DNA used for one transformation was about 500 ng. The electroporated cells of the brewing yeast strain Sa-07256 were incubated in 5 ml YPD overnight, washed with 0.85% NaCl and spread on YD agar medium (1% yeast extract, 2% glucose, pH 6.3) containing 15 μg ml−1 G418. Transformants of the strain YSH 1.1.-6B were selected using a G418 concentration of 100 μg ml−1.

2.4Determination of specific enzyme activities

The specific activity of GPD was measured as previously described [18]. The specific activity of BDH was determined in crude yeast extracts following a modified procedure according to Heidlas et al. [25]. The assay was performed in a total volume of 1 ml containing 100 mM phosphate buffer (pH 6.7), 0.4 μmol NADH and 100 μl of the crude cell extract or an appropriate dilution. The reaction was started by adding 10 μmol acetoin. To ascertain the oxidation of NADH, the decrease of absorbance at 340 nm was measured at 25°C.

2.5Primary and secondary fermentation under brewing conditions

Brewers’ wort with an original gravity of 11.38°P [26] was kindly provided by a German brewery. The sugar composition of the wort was as follows: maltose (52.8 g l−1), maltotriose (16.3 g l−1), glucose (9.7 g l−1), sucrose (3.0 g l−1), fructose (2 g l−1), non-fermentable sugars which are mainly dextrins (28.8 g l−1).

Brewing yeast transformants were precultured in 100 ml wort containing 100 μg ml−1 G418 by shaking at room temperature. After 24 and 48 h, 700 ml of fresh wort supplemented with G418 (100 μg ml−1) were added, respectively, and the precultures were cultivated under the same conditions for additional 48 h. Then, the cultures were incubated without shaking at 12°C for 24 h. The cells were harvested and washed once in cold wort to remove G418.

Wort was inoculated at a density of approximately 1×107 cells ml−1 for primary fermentation. Fermentation experiments were carried out to a scale of 3 l in a glass fermenter (Jenaer Glas AG, Jena, Germany) with stirring (50 rpm) at 12°C. The fermenter was closed with air locks, which ensured the exclusion of oxygen but allowed the release of gases. The primary fermentation was completed when wort sugars were nearly consumed, i.e. 7 days after inoculating the fermenters. Refer to Fig. 1C, D for the actual concentration of maltose and maltotriose.

Figure 1.

Fermentation performance of a brewing yeast strain (S. cerevisiae ssp. carlsbergensis Sa-07256) overexpressing GPD1 (●) in comparison to the corresponding reference strain (◯). Experiments were carried out in wort simulating brewing conditions at a 3-l scale. The results of one of three independent representative experiments are shown. A: Concentration of non-sedimented cells in fermentation broth. B: Course of pH values. Consumption of maltose (C) and maltotriose (D), the main fermentable sugars in wort.

The broth obtained from the primary fermentation, i.e. green beer, underwent a maturation process (secondary fermentation). The fermenters containing the green beer were kept without stirring at 10°C for 24 h. Aliquots of 1.2 l were taken from the fermenters and incubated in Schott flasks equipped with pressure-controlled valves at 1.0 bar and 10°C for 10 days, followed by a further 10 days at 0°C.

2.6Fermentation under laboratory conditions

Transformants of YSH 1.1.-6B were precultured in YPD medium containing 100 μg ml−1 G418 at 30°C. Cells were washed once with 0.85% NaCl before being inoculated into the fermentation medium at a density of 4×107 cells ml−1. The batch fermentation was performed in 250 ml Erlenmeyer flasks containing 150 ml of YPD medium, supplemented with 100 μg ml−1 G418. The cultures were stirred continuously at 300 rpm after closing the vessels with air locks.

The glucose in the YPD medium was replaced by 19 g l−1 maltose for fermentation experiments in YPM medium.

2.7Analytical methods

Ethanol and glycerol concentrations in beer were determined by HPLC on a 300×6.5 mm Polymeric Column (ERC, Interaction 28042). The compounds were eluted at 40°C with 0.25 mM H2SO4 at a flow rate of 0.6 ml min−1. Detection was performed by means of a refraction index detector (ERMA, ERC-7511). In order to quantify the signals, external standards prepared from pure compounds (Merck) were applied. Acetate, succinate and pyruvate were enzymatically assayed using test kits or single reagents and the corresponding protocols (Enzymatic BioAnalysis/Food Analysis, UV method) supplied by Boehringer Mannheim (Mannheim, Germany).

To measure the concentration of 2,3-butanediol, 12 g K2CO3 were dissolved in 10 ml of the beer sample, 3 ml CHCl3 and 10 mg 1,4-butanediol in 0.2 ml CHCl3 (internal standard) were added and stirred overnight at room temperature. The resulting emulsion was subjected to centrifugation (5000 rpm) and the organic layer was dried using Na2SO4. Pyridine (0.1 ml) and acetic acid anhydride (0.1 ml) were added to 0.2 ml of the dried sample. After incubation in a sealed vessel at 80°C for 4 h, the samples were directly analysed by gas chromatography on a DBWAX fused silica capillary column (polyethylene glycol) of 60 m length installed in a Fisons instrument, Model GC 8000 (8165). The column temperature was programmed from 70°C (5 min) to 230°C (4°C/min); the carrier gas was nitrogen.

Determinations of wort sugars and all ingredients of beer listed in Table 1, except for those already mentioned, were performed according to internationally standardized methods edited by the European Brewery Convention [26].

Table 1.  Analysis of green beer, i.e. beer after the primary fermentationa, produced with an industrial lager brewing yeast strain (Sa-07256) overexpressing GPD1 and the corresponding reference strain. Besides ethanol and glycerol, flavour-relevant by-products have been analysed.
  1. aFermentation experiments were carried out in original brewers’ wort simulating brewing conditions at a 3-l scale (see Section 2). Concentrations of beer components were determined after primary fermentation, i.e. when sugars were nearly depleted. Except as noted, all results shown are mean values of two or three experiments including standard deviations.

  2. bSingle experiment.

  3. cData for diacetyl relate to the sum of free diacetyl and its precursor (α-acetolactate).

 Reference plasmidGPD1 overexpression
Ethanol (g l−1)37.0±1.330.4±1.5
Glycerol (g l−1)1.57±0.178.73±0.65
Acetaldehyde (mg l−1)5.1±1.9226±50
Acetoin (mg l−1)2.93±1.521329±301
2,3-Butanediol (mg l−1)b160211
Organic acids
Pyruvate (g l−1)b0.050.05
Acetate (g l−1)b0.030.45
Succinate (g l−1)b0.060.27
Vicinal diketones
2,3-Pentanedione (mg l−1)0.07±0.040.17±0.09
Diacetylc (mg l−1)0.11±0.042.70±0.36
Fusel alcohols
Propyl alcohol (mg l−1)14.2±4.114.7±2.9
Isobutyl alcohol (mg l−1)6.8±0.38.4±1.5
Isoamyl alcohol (mg l−1)40.3±2.734.8±5.2
2-Phenylethyl alcohol (mg l−1)10.4±2.39.0±2.5
Ethyl acetate (mg l−1)10.6±3.36.4±1.4
Isobutyl acetate (mg l−1)0.02±0.010.02±0.01
Butyl acetate (mg l−1)0.00±0.000.00±0.00
Isopentyl acetate (mg l−1)0.50±0.010.32±0.15
2-Phenylethyl acetate (mg l−1)0.14±0.000.09±0.04
Ethyl formiate (mg l−1)0.43±0.000.46±0.15
Ethyl butyrate (mg l−1)0.04±0.000.02±0.01
Ethyl caproate (mg l−1)0.10±0.010.08±0.02
Ethyl caprate (mg l−1)0.18±0.020.12±0.03
Ethyl caprylate (mg l−1)0.04±0.000.03±0.02
Fatty acids
Isovaleriate (mg l−1)1.92±0.341.27±0.09
Caproate (mg l−1)1.39±0.161.24±0.10
2-Ethyl capronate (mg l−1)0.07±0.000.08±0.00
Caprate (mg l−1)3.03±0.402.60±0.13
Caprylate (mg l−1)0.41±0.120.31±0.04

3Results and discussion

3.1Overexpression of GPD1 in brewing yeast

In an initial experiment, the industrial brewing yeast strain Sa-07256 was transformed with plasmid YEpKmGPD1 previously used to overproduce GPD in a laboratory strain [18]. However, no transformants could be obtained. In accordance to previous studies [23,27], the ability for plasmid YEpKmGPD1 to be transformed into yeast was improved by replacing the bacterial promoter of the Km gene by the TEF1 promoter of S. cerevisiae (refer to Section 2.2). Sa-07256 transformants carrying YEpTKmGPD1 and YEpTKm (reference plasmid) were selected, respectively. The specific activity of GPD in these transformants, harvested 72 h after starting the primary fermentation under brewing conditions, was increased from 0.16±0.06 U mg−1 protein in the reference strain to 0.68±0.04 U mg−1 protein in the GPD overproducer.

3.2Fermentation performance of a brewing yeast overexpressing GPD1

The performance of the GPD1-overexpressing brewing strain in fermentation experiments simulating brewing conditions was characterized by the following parameters: (i) the time course of numbers of non-sedimented cells (Fig. 1A), (ii) the time courses of the pH values (Fig. 1B), and (iii) consumption of maltose and maltotriose (Fig. 1C, D), the most important fermentable sugars in brewers’ wort (refer to Section 2.5). It should be noted here that sugar consumption under standard brewing conditions is recorded by determining the specific gravity of fermentation broth [26]. However, this method is not suitable to study sugar consumption in strains overproducing glycerol, as the high gravity of this metabolite would strongly falsify the data. Therefore, only the most important wort sugars maltose and maltotriose were detected by HPLC in this study.

The growth of the cells was negatively affected by the overproduction of GPD under simulated brewing conditions (Fig. 1A). A similar result was previously described for the strain YSH 1.1.-6B under laboratory conditions [18].

At the end of fermentation, the GPD-overproducing brewing yeast showed a faster sedimentation rate than the reference strain (Fig. 1A). This characteristic seems to be strain-dependent since overexpression of GPD1 in other brewing yeast strains did not have this effect (data not shown).

The time courses of the pH values as well as the consumption of maltose and maltotriose were not influenced by GPD1 overexpression (Fig. 1B–D).

There seems to be a discrepancy as to the unaffected time courses of sugar consumption of the GPD overproducer in spite of the slower growth observed. This could be explained by a higher specific rate of sugar metabolization caused by GPD1-overexpression. This assumption is supported by the fact that the laboratory strain YSH 1.1.-6B overexpressing GPD1 showed a higher specific rate of glucose consumption [18]. In addition, engineered wine yeast strains exhibited a significant increase in the CO2 production rate during the stationary phase [20].

3.3Reduction of ethanol formation by overproduction of glycerol

The aim of this study was to reduce the content of ethanol in beer by redirecting the carbon metabolism towards glycerol. The amount of glycerol in beer produced by the GPD1-overexpressing yeast was increased 5.6 times in comparison to the reference strain while ethanol was decreased by 18% (Table 1). The extent of redirecting carbon flux towards glycerol in the brewing yeast overexpressing GPD1 was, hence, less effective than previously shown for the laboratory yeast where the amount of ethanol was reduced by 37% and glycerol was 8.1 times increased [18]. There are several reasons which could account for this difference.

(1) The plasmids used for overexpression were slightly different in both studies and in fact, laboratory yeast YSH 1.1.-6B transformed with the plasmid YEpTKmGPD1 showed significant lower specific activity of GPD than the YEpKmGPD1 transformant (Fig. 2). The use of the yeast-based TEF1 promoter which controls the selective marker gene (Km) (refer to Section 3.1) might have reduced the number of plasmid copies per cell in YEpTKmGPD1 transformants during growth under selective pressure (G418). This could have led to less effective overproduction of GPD.

Figure 2.

Specific activities of GPD of GPD1 multicopy transformants of the laboratory S. cerevisiae strain YSH 1.1.-6B by employing either the plasmid YEpKmGPD1 or plasmid YEpTKmGPD1 for overexpression of GPD1. YEpTKmGPD1 contains a S. cerevisiae promoter (TEF1) instead of the original bacterial promoter upstream from the Km gene. The specific GPD activity was measured during exponential growth in a fermentation under laboratory conditions (YPD medium). Mean values and standard deviations of three independent experiments are shown.

(2) Cells were exposed to G418 only during the preculturing phase in the current study which was in contrast to the fermentation experiments under laboratory conditions [18]. This could also have contributed to a reduced number of plasmids per cell and, therefore, to a lower amount of glycerol formed by the brewing yeast transformant.

(3) Lager brewing yeast strains differ genetically from S. cerevisiae laboratory strains. It is largely assumed that the former are genetic hybrids of S. cerevisiae and at least one other Saccharomyces species [28].

(4) Maltose and maltotriose are the main carbon sources in brewers’ wort. Since strain YSH 1.1-6B was not able to metabolize the sugars in wort, glucose (in YPD medium) was used as the sole carbon source in previous studies analysing the laboratory strain overexpressing GPD1. As shown in Fig. 3, based on equimolar amounts of glucose and maltose, the GPD1 overexpressing brewing yeast produced about twice the amount of glycerol when glucose was used (in YPD medium) instead of maltose (in YPM medium). Hence, these results indicate that the carbon source used has a strong impact on the redirection of carbon flux towards glycerol in a GPD overproducer.

Figure 3.

The effect of maltose instead of glucose on the metabolic shift from ethanol towards glycerol production in a brewing yeast strain (Sa-07256) overproducing GPD1 (YEpTKmGPD1). Fermentation under laboratory conditions was performed as described in Section 2.

All in all, our data present another example that results from laboratory yeasts cannot be directly transferred to practical conditions, a fact well known from other approaches to improve industrial yeast strains [29,30].

In order to optimize the reduction of ethanol in the GPD1-expressing brewing yeast, the stronger ADH1 promoter was substituted for the original (GPD1) promoter to control GPD1 expression (YEpTKmA-GPD1). In this case, ethanol formation under brewing conditions was indeed slightly reduced compared to the results obtained with plasmid YEpTKmGPD1, but still did not by any means reach the level of reduction known from the laboratory yeast (data not shown).

3.4Determination of flavour-relevant by-products in green and mature beer

Depending on their specific flavour thresholds, many by-products of the yeast metabolism are known to decisively contribute to the flavour and the body of beer [31,32]. Therefore, other components in addition to ethanol and glycerol were analysed in the beer produced by the GPD1-overexpressing yeast. The analysis of the concentration of fusel alcohols, esters and fatty acids after primary fermentation showed only minor differences in comparison to the reference beer (Table 1). However, there was a moderate increase in the concentration of acetate, succinate and 2,3-butanediol, but a dramatic increase in the levels of acetaldehyde, acetoin and diacetyl (Table 1).

A very similar pattern of by-products was obtained when GPD1 was overexpressed in wine yeast strains for the purpose of increasing the level of glycerol in wine to improve body and sweetness [19,20]. The substantial changes in the by-product pattern of the GPD overproducers were attributed to the need of balancing NADH metabolism on the one hand and to acetaldehyde detoxification on the other hand [19].

Interestingly, the level of acetoin formed by the GPD-overproducing brewing yeast was considerably higher than previously observed in engineered wine yeast strains [20]. A possible explanation might be that acetoin in wine yeast is reduced to 2,3-butanediol to a higher extent than in brewing yeast. In accordance with this hypothesis, we found that 2,3-butanediol dehydrogenase (BDH) [25,33] activity was considerably lower in brewing yeast than in wine yeast (Pilger et al., in preparation).

The primary fermentation process in brewing practice is usually followed by secondary fermentation of the so-called green beer. During this maturation process at a low temperature the concentration of many components normally found in green beer (e.g. diacetyl, acetoin, acetaldehyde, organic acids) are known to be reduced to a moderate extent. Therefore, we studied the influence of a secondary fermentation on the concentration of the most undesired by-products, i.e. acetaldehyde, diacetyl and acetoin, present in beer obtained by employing the GPD overproducer. As expected, the concentration of acetaldehyde and diacetyl decreased during secondary fermentation (Fig. 4), but could not be reduced below the flavour thresholds of these substances in beer [34]. The concentration of acetoin formed by the GPD overproducer was even increased after the maturation process (Fig. 4). Possibly, the elimination of acetaldehyde in the GPD-overproducing yeast [19] continued during beer maturation, resulting in a decrease in acetaldehyde accompanied by an increase of acetoin.

Figure 4.

Effect of the secondary fermentation on the concentration of acetaldehyde, acetoin and diacetyl in beer, produced with a brewing yeast strain overexpressing GPD1 and the corresponding reference strain, respectively. The secondary fermentation simulating brewing conditions was performed at 10°C for 10 days, followed by a further 10 days at 0°C. Mean values and standard deviations of two independent experiments are shown.

Several persons tasted the mature beer produced by the GPD overproducer and described its flavour as ‘sherry-like’. A comparison of the by-product pattern of our beer and sherry revealed similarities regarding the high concentrations of acetoin, diacetyl and acetaldehyde that are responsible for the typical taste.


Using a brewing yeast overexpressing GPD1 we were able to reduce the content of ethanol in beer by 18%. Removing alcohol from beer produced by the GPD1-overexpressing brewing yeast could be less expensive when commercial techniques are employed. Less ethanol has to be removed allowing the plant to achieve higher throughput. Possibly, the GPD-overproducing brewing yeast could be of interest for some breweries for creating a beverage with a special type of taste due to the ‘sherry-like’ flavour of the beer. The future goal of metabolic engineering will be to combine the optimization of ethanol reduction and by-product formation, thus facilitating the direct use of modified brewing yeast to produce low-alcohol beer.


We are grateful to D. Müller and T. Hirschberger for technical assistance; to M. Treuner, U. Schmidt, U. Donalies, and R. Bensmann for proofreading the manuscript and critical discussion. This work was supported by grants from the ‘Bundesministerium für Wirtschaft, Arbeitsgemeinschaft für industrielle Forschung (AiF) Otto V. Güricke’, Cologne, Germany, and by the ‘Stiftung Industrieforschung’, Bonn, Germany.