Advertisement
Resource
Transparent Process
Open Access

Performance and robustness analysis reveals phenotypic trade-offs in yeast

Cecilia Trivellin, Peter Rugbjerg, Lisbeth Olsson  Correspondence email
Published 30 October 2023. DOI: 10.26508/lsa.202302215

Abstract

To design strains that can function efficiently in complex industrial settings, it is crucial to consider their robustness, that is, the stability of their performance when faced with perturbations. In the present study, we cultivated 24 Saccharomyces cerevisiae strains under conditions that simulated perturbations encountered during lignocellulosic bioethanol production, and assessed the performance and robustness of multiple phenotypes simultaneously. The observed negative correlations confirmed a trade-off between performance and robustness of ethanol yield, biomass yield, and cell dry weight. Conversely, the specific growth rate performance positively correlated with the robustness, presumably because of evolutionary selection for robust, fast-growing cells. The Ethanol Red strain exhibited both high performance and robustness, making it a good candidate for bioproduction in the tested perturbation space. Our results experimentally map the robustness–performance trade-offs, previously demonstrated mainly by single-phenotype and computational studies.

Introduction

To achieve high yields while ensuring reproducibility and high-quality in bioprocesses, it is essential to address the factors contributing to process variability. The variable conditions, also referred as perturbations, can be triggered by the bioprocess environment (temperature, nutrient gradients, changes in raw materials), intracellular factors (noisy gene expression, genetic mutations) (Olsson et al, 2022), and different bioprocess steps (cell propagation, fermentation) (Tomás-Pejó & Olsson, 2015; Keil et al, 2019). The set of all perturbations present in the bioprocess is referred to as the perturbation space, and can be both predictable and stochastic. Generally, the bioprocess perturbation space has a negative impact on strain performance compared with its ideal and well-controlled lab-scale space, both in terms of specific growth rates and production of valuable metabolites. Therefore, industrial strains must exhibit consistent performance, that is, robustness, to avoid loss of product quantity and quality or larger costs (Huang et al, 2016; Mohedano et al, 2022).

Microbial robustness refers to the ability of cells to maintain a stable performance when exposed to a perturbation space (Kitano, 2004; Masel & Siegal, 2009; Levy et al, 2012; Félix & Barkoulas, 2015; Olsson et al, 2022). Microbial robustness is typically assessed for phenotypes related to industrial performance (titer, rate, and yield), but it can be extended to cell volume, cell viability, gene expression or indirect measurements of cellular parameters such as fluorescence. For simplicity, microbial robustness is referred hereafter as robustness.

Numerical assessment of robustness can be used to quantify performance stability (Steensels et al, 2014; da Conceição et al, 2015; Louis, 2016; Mestek Boukhibar & Barkoulas, 2016). We have previously proposed and validated a high-throughput methodology to quantify robustness in multiple phenotypes, resulting in a dimensionless negative number, where the theoretical zero represents a completely robust, non-changing phenotype (Trivellin et al, 2022). Our methodology is built on a subset of phenotypes (cellular functions) that can be measured experimentally. Using a series of single experimental perturbations designed to simulate bioprocess conditions, robustness measures the variation of the performance of interest with respect to its average across multiple perturbations. In addition, robustness quantification allows the exploration of a broad range of physiological phenomena (e.g., trade-offs), which would otherwise be challenging to investigate using standard methods (e.g., fermentation profiling in bioreactors).

Microorganisms appear to exhibit trade-offs between performance and robustness of one or more phenotypes (Kitano, 2007, 2010), which could be important for improving strain and process development. For example, in bacteria, antibiotic resistance and extracellular enzyme production have been shown to trade-off with the specific growth rate and a similar trade-off has been observed between melanin production and specific growth rate in fungi (Andersson, 2006; Ramin & Allison, 2019; Lovero & Treseder, 2021). To the best of our knowledge, trade-offs between performance and robustness have been investigated mostly in single-phenotype and in silico studies (Ibarra et al, 2002; Stelling et al, 2002; Fischer & Sauer, 2005). Experimental validation of trade-offs could explain why strains optimized for maximum performance are less capable of coping with environmental stresses and fluctuations (i.e., present lower robustness).

In the present study, we applied our previously developed robustness quantification method (Trivellin et al, 2022) to a large dataset of yeast responses to perturbations. The dataset contained more than 10,000 phenotypic data points obtained upon cultivation of 24 Saccharomyces cerevisiae strains under 29 different growth conditions simulating bioethanol production from second-generation biomass (perturbation space). A culture transfer step was included as additional perturbation to explore the method’s versatility in assessing robustness within different bioprocess steps.

The combination of exploratory data analysis on the large dataset with quantification of robustness allowed us to map perturbation-specific influences on performance and robustness, and identify strains with robust phenotypes. We proved that exposing a cell culture to a perturbation during pre-cultivation, significantly increased the performance associated with the specific growth rate during subsequent cultivation. Finally, correlation tests revealed trade-offs between robustness and performance (measured as biomass and ethanol yield and cell dry weight.) Although our earlier work primarily focused on the development and validation of the robustness quantification method itself, in the current study our aim was to demonstrate its practical application. Specifically, our study provides a model for integrating performance and robustness data to uncover phenotypic trade-offs in yeast, which is a critical aspect of strain engineering. Our findings demonstrate that strongly performing cells under one condition may be less robust in others, underscoring the importance of considering both factors in the design process.

Results

High-throughput characterization reveals how perturbations in medium components influence the phenotypes of S. cerevisiae strains

A total of 24 S. cerevisiae strains were examined in the present study. They included two well-characterized laboratory strains, four industrial strains employed in bioethanol production or baking, and 18 LBCM strains isolated from cachaça fermentation plants (see the Material and Methods section) (De Araújo Vicente et al, 2006). Each strain was cultivated in microtiter plates containing chemically defined Delft medium plus a single component simulating perturbations found during industrial lignocellulosic bioethanol fermentations, such as acetic acid released during the hydrolysis of hemicellulose (Jönsson & Martín, 2016). The single conditions were grouped according to similarity between the physiological responses they elicited in yeast. Specifically, the acids (lactic, levulinic, acetic, and formic acid) were grouped together, as were the pentoses (xylose and arabinose) and hexoses (galactose, glucose, and mannose). The aldehydes, including vanillin, 5-hydroxymethylfurfural, and furfural, were classified together, whereas NaCl and ethanol were considered single components. For each cultivation, five phenotypes (specific growth rate, lag phase, final cell dry weight, biomass yield, and ethanol yield) were calculated for a total of 10,295 data points.

Although it has been reported that industrial strains grow faster than laboratory strains (Kong et al, 2018; Yi & Alper, 2022) because of their ability to handle perturbations in industrial settings, no significant differences were detected between the two groups. Overall, Ethanol Red showed the highest mean performance with respect to all phenotypes except the ethanol yield, in which case PE2 attained the highest mean value (Fig 1A). The ethanol yield of the PE2 strain was more than double the average calculated across all strains. The PE2 strain has been reported to have a high CO2 production, final ethanol production (close to 95% of the theoretical yield in E. globulus wood hydrothermal hydrolysate anaerobic fermentation), fast sugar utilization and faster degradation of furfural and 5-hydroxymethyl-2-furaldehyde (Pereira et al, 2010, 2014; Soares-Costa et al, 2014). Furthermore, PE2 has also been shown to dominate and persist in Brazilian distilleries probably because of its high viability and high specific growth rate (Raghavendran et al, 2017; Araújo et al, 2018). In our study, the specific growth rate of PE2 was 20% lower than the average across all strains.

Figure 1.
Figure 1. Yeast phenotypes are impacted significantly by environmental perturbations.

(A) Phenotypic data averaged across groups of perturbations for five phenotypes (μmax, maximum specific growth rate; Yp, ethanol yield; Yx, biomass yield; CDW, cell dry weight; and lag phase) and 24 S. cerevisiae strains (X-axis). (B) Data distributions of the five measured phenotypes for each group of perturbations and 24 strains (X-axis). Data information: in panel (A), each colored dot corresponds to the mean across triplicates and groups of perturbations. In panel (B), the black line inside the colored area represents the median of the distribution. Differences between groups and hexoses were assessed with a Wilcox test (n = 10,295; ns, not significant; ****P < 1.4 × 10−5).

To assess for correlations among phenotype performances, we carried out Spearman correlation tests. The maximum specific growth rate correlated negatively with lag phase (P < 2.2 × 10−16) (Fig S1), confirming previous observations (Basan et al, 2020). Instead, positive significant correlations were observed between specific growth rate and ethanol yield, biomass yield, and cell dry weight. However, when splitting the data into groups of conditions, correlations became nonsignificant in the case of hexoses, both for ethanol and biomass yield, and in the case of pentoses and NaCl for ethanol yield. The Spearman tests revealed a positive monotonic relationship between biomass yield and end-of-cultivation cell dry weight with respect to specific growth rate. Negative correlations were observed in single strains, such as Ethanol Red and PE2; although no overall negative relationship between production (calculated as ethanol yield) and growth (maximum specific growth rate) was detected in this perturbation space.

Figure S1.
Figure S1. Scatter plots showing Spearman correlation coefficients between maximum specific growth rate (1/h) and ethanol yield (g/g), biomass yield (g/g), cell dry weight (g/liter), and lag phase (h).

Data information: Each panel corresponds to a phenotype on the Y-axis (lag phase, cell dry weight, ethanol, and biomass yield) plotted against the maximum specific growth rate (X-axis). Colored dots represent data points measured in the study and are divided according to the perturbation group, in which the yeast population was tested (acid, ethanol, aldehyde, pentose, hexose, and NaCl). Linear regression lines are included in the plots and colored based on the perturbation group mainly to investigate direction and strenght of the relationship among phenotypes. Correlation coefficients are shown above the graphs for each group, and for the whole dataset in the right upper corner of the graph. R = Spearman correlation coefficient; P = P-value (statistically significant Spearman correlations were P < 0.05); sample size (n) = 2059.

Analysis of strain performance revealed a substantial negative impact of groups of conditions on the different phenotypes, whereas no such effect was observed in growth medium containing only hexoses. The only exceptions were biomass yield in the presence of pentoses or hexoses, and maximum specific growth rate in the presence of hexoses or acids (Wilcoxon test, P = ns) (Fig 1B). The negative effect of aldehydes, acids, ethanol, and NaCl on the phenotypes confirmed earlier studies (Adeboye et al, 2014; Tekarslan-Sahin et al, 2018; Caspeta et al, 2019; Guaragnella & Bettiga, 2021). Ethanol Red, PE2, and the LBCM strains are highly tolerant towards ethanol (Demeke et al, 2013) and lignocellulosic inhibitors (Wallace-Salinas & Gorwa-Grauslund, 2013; Araújo et al, 2018; Cunha et al, 2019). Here, Ethanol Red and some LBCM strains displayed higher tolerance (higher specific growth rate) towards aldehydes (Fig 1A).

The presence of acids lowered the yields for all strains; although no significant difference in specific growth rate was observed between medium containing acid and not (Fig 1B). Considering fermentation kinetics, a decreased yield of biomass on substrate while keeping the specific growth rate constant may suggest an increase in the specific rate of substrate consumption. Weak acids are not inhibitory enough to slow down the anabolism; therefore growth can proceed at high specific rate. However, ATP is required for counteracting the effects of the acidification, which diverts the carbon source from anabolism to catabolism. This decreases the yield and increases the specific substrate consumption rate, and the specific production rate of energy-related primary metabolites. In the presence of up to 7 g/liter lactic acid, all phenotypes displayed higher or comparable performance to the control containing 20 g/liter glucose. This was likely because the pKa of lactic acid is 3.79, and at a pH of 5, most of the acid existed in its dissociated form, making it less likely to penetrate the cells. Lactic acid is encountered during bioethanol production because of contamination with Lactobacillus spp. Interestingly, Lactobacillus amylovorus has been shown to be beneficial (3% higher ethanol yields) or neutral to yeast fermentation (Senne de Oliveira Lino et al, 2021).

A higher average ethanol yield was noted across all strains when comparing pentoses and hexoses (54% higher, P < 0.001). S288C and LBCM1008 showed a significantly higher ethanol yield in medium containing 5 g/liter glucose plus various amounts of xylose or arabinose compared with medium containing only 20 g/liter glucose (Fig 1A). The other phenotypes did not exhibit an increment in the presence of pentoses (Fig 1B).

Owing to the meticulous evaluations of five different phenotypes in 24 strains, we were able to point out the best performing strains and illustrate how their performance was influenced by different inhibitors with a fast and comprehensive method in small scale.

Ethanol Red is a unique compromise for high robustness and performance

In the present study, we applied our previously developed methodology to quantify the robustness of five phenotypes in 24 strains within a set perturbation space (Equation (1), see the Materials and Methods section) (Trivellin et al, 2022).

Robustness varied according to phenotype and strain (Fig 2). LBCM1079 and LBCM1095 displayed constantly higher robustness compared with other strains; whereas LBCM1001 and LBCM1106 presented overall low robustness values. Red Star exhibited strong robustness for some phenotypes, but low robustness for others. LBCM1106 exhibited a significantly lower robustness for the cell dry weight mainly attributed to its outstanding performance in the presence of hexoses (Fig 1A). When considering all strains, robustness was generally higher for lag phase and maximum specific growth rate (−0.2 and −0.3, respectively) than for other phenotypes, such as ethanol yield (−0.9), biomass yield (−0.9), and cell dry weight (−0.9). Larger experimental variation in end-of-cultivation cell dry weight and yields contributed to the lower robustness.

Figure 2.
Figure 2. Robustness quantification reveals significant differences between strains for five phenotypes.

Data information: the five horizontal plots correspond to the five phenotypes tested in the study. Strains are shown on the x-axis. Each robustness value was calculated with Equation (1). Error bars denote the standard error of the mean (n = 3).

LBCM1095, LBCM1079, and LBCM97 showed the highest mean robustness: −0.3, −0.3, and −0.4, respectively. LBCM97 has been described as highly tolerant towards ethanol, aluminum, and a broad pH range (da Conceição et al, 2015; van Dijk et al, 2020). The mean robustness of Ethanol Red was −0.5, mostly because of low specific growth rate in the presence of ethanol. Ethanol Red is very robust when fermenting sugars from grain mash at a high temperature (Wallace-Salinas & Gorwa-Grauslund, 2013), and highly performing in fed-batch fermentation on molasses (Demeke et al, 2013). Owing to the way robustness is quantified (Félix & Barkoulas, 2015; Trivellin et al, 2022), a poorly performing phenotype would result in elevated robustness values if it behaved consistently across multiple conditions. To fully understand how a strain responds to various perturbations, a comprehensive overview of both its performance and robustness is required. When evaluating robustness and performance simultaneously, Ethanol Red stood out as the best compromise. This strain displayed an outstanding performance in four out of five phenotypes, and ranked among the highest for robustness. Therefore, even though the perturbation set chosen for the study was not specific to starch substrates, which are preferred by Ethanol Red (Cripwell et al, 2019), the genetic and physiological make-up of this strain allows it to be highly performing even in perturbation spaces other than those associated with starch fermentation.

The effect of groups of conditions on robustness helps identify key factors that significantly influence strain performance

The overall performance of some strains (e.g., LBCM37) varied greatly within the perturbation space, which led to low robustness compared to other strains (Fig 1A, spread of the data points for each strain). To better understand such variation, we tested the contribution of each group of conditions (pentoses, hexoses, NaCl, ethanol, aldehydes, and acids) to the robustness of each phenotype. The robustness calculated with Equation (1) was compared with the robustness obtained with the same equation, but excluding each group of conditions (Fig S2).

Figure S2.
Figure S2. Influence of the groups of perturbations on the robustness of S. cerevisiae strains.

Data information: the ratio between the robustness calculated by filtering out one group of perturbations and the robustness based on all the perturbations was calculated (see the Materials and Methods section). Strains are shown on the X-axis. The Y-axis is divided by group of perturbations (right) and the five tested phenotypes (left). Green dots denote ratio(R) > 1 (the group of perturbations on the right does not influence robustness) and red dots denote ratio(R) < 1 (the group of perturbations on the right has a negative impact on the robustness of the phenotype).

Overall, pentoses, acids, and NaCl had a neutral impact on the robustness, except for the robustness of the lag phase which was negatively impacted by the pentoses and the robustness of the biomass yield, negatively impacted by both NaCl and acids. Ethanol and aldehydes had a negative impact on the robustness of all phenotypes, except for some of the strains whose lag phase robustness was not affected (for example, the laboratory strains). Ethanol Red and LBCM110 were the only two strains, whose cell dry weight was not affected by aldehydes.

Calculating the influence of single groups of conditions on the robustness of a specific strain serves three purposes. First, it reveals which conditions cause the largest spread in the distribution of data (Fig 1A). Second, it suggests which conditions should be included in the tested perturbation space. Assuming all relevant stochastic and predictable perturbations (from an extracellular or intracellular environment) are included in the perturbation space, testing the influence of certain groups of conditions on robustness could reduce the number of conditions to assess. The number of tested perturbations should nevertheless be statistically significant for the robustness value to have a meaning. Third, if the robustness of a specific phenotype is not affected only in a few strains, the latter could reveal metabolic mechanisms responsible for such observation.

Negative correlations between performance and robustness confirm presumed trade-offs

Trade-offs between robustness and performance have been hypothesized previously (Kitano, 2007, 2010; Quinton-Tulloch et al, 2013). For example, in-silico studies have suggested that cells investing resources in anticipation of environmental changes display suboptimal growth (Fischer & Sauer, 2005). To determine the trade-offs between performance and robustness, a three-dimensional matrix of strains, perturbations, and phenotypes was created. Spearman’s correlation tests were carried out on robustness and performance to display monotonic relationships among the measured values (Fig 3). Negative correlations can provide evidence of phenotypic trade-offs, if the observed phenotypic response is directly caused by the environmental stimulus received (Fox & Stevens, 1991). Because complex genetic architecture and regulatory networks determine a specific phenotype, generalization is not possible. Here, we observed potential trade-offs connected to the applied perturbation space.

Figure 3.
Figure 3. Spearman correlations between the performance and robustness of five different phenotypes show potential performance–robustness trade-offs.

(A) Spearman correlation matrix between performance and robustness for five phenotypes. (B) Correlations between mean performance and mean robustness of the five measured phenotypes. Data information: in panel (B), each dot corresponds to a single strain. A linear regression line is plotted in each panel mainly to visualize the direction and strenght of the relationship. R = Spearman correlation coefficient; P = P-value; n = 24.

Negative correlations (−0.8 to −0.6) were identified between robustness and performance of cell dry weight, biomass yield, and ethanol yield (P < 0.05) (Fig 3A). This result suggested a performance–robustness trade-off, whereby low performances were associated with high robustness and vice versa. The only positive correlation between performance and robustness was observed for maximum specific growth rate (Fig 3B). Instead, no significant correlation was found for the lag phase.

The estimated standard error for the Spearman correlation between performance and robustness was around 0.1 (see the Materials and Methods section) for ethanol yield, biomass yield, and cell dry weight, and around 0.2 for specific growth rate. The error increases with small sample size and weaker (R < 0.6) correlation coefficients. Some strains displayed more evident performance-robustness trade-offs, but only for certain phenotypes, for example Ethanol Red’s lag phase had the highest performance but the lowest robustness among all the strains (Fig S3). Ethanol Red low lag phase robustness is attributed to its ability to grow, after a very long lag phase, in harsh conditions.

Figure S3.
Figure S3. Trade-offs between performance and robustness for each phenotype and strain.

Data information: strains are listed on the X-axis and phenotypes on the Y-axis. Each tile represents the mean performance (top) and robustness (bottom) across perturbations for each strain. Performance and robustness tile text was assigned based on the measured mean performance and calculated mean robusntess for each strain across conditions. The color of each tile represents normalized values between the minimum and maximum calculated separately for each phenotype across different strains.

Positive significant correlations were observed between robustness of biomass yield and maximum specific growth rate with the robustness of cell dry weight (0.8 and 0.7 respectively), and between the robustness of biomass yield with the robustness of maximum specific growth rate (0.5) (Fig 3A). Significantly negative correlations were observed instead between performance and robustness of various pairs of phenotypes, including robustness of cell dry weight and lag phase or ethanol yield, robustness of biomass yield and cell dry weight, robustness of lag phase and specific growth rate, and robustness of specific growth rate and lag phase.

Our results point to the phenotype-specific behavior of robustness (Fig 2), thereby supporting previous evidence (Barkai & Leibler, 1997; Carlson & Doyle, 2000; Félix & Barkoulas, 2015; Trivellin et al, 2022). Even though robustness cannot be considered a general property of a system, the positive correlations observed among robustness values (Fig 3A) may suggest that biomass yield, specific growth rate, and cell dry weight are regulated and stabilized by similar or shared mechanism.

In summary, the observed correlations validated the hypothesized trade-offs between robustness and performance for cell dry weight, biomass yield, and ethanol yield. Moreover, they provide a powerful tool for further investigations of possible trade-offs.

95% of the tested strains showed an increased maximum specific growth rate upon transfer to the same medium

When designing the perturbation space, it is crucial to consider all the bioprocess steps, including preculture or inoculation, as they can have a negative impact on the process outcome. In the present study, we expanded the perturbation space by including a culture transfer and investigated its impact on the specific growth rate of each strain. We compared the maximum specific growth rate in the first cultivation step, with the one in the second cultivation after transfer (see the Materials and Methods section). We calculated the improvement in maximum specific growth rate in terms of performance (Equation (2)). A strong and significant positive correlation between the first and second maximum specific growth rates was found (Fig 4A). No notable discrepancies among the groups were detected when dividing the correlations into various perturbation categories. The correlation coefficients ranged from 0.6 for ethanol to 0.8 for NaCl. The P-value was > 2.2 × 10−16 for all groups of conditions, confirming statistically strong correlations. Sugars exhibited a slightly lower correlation coefficient compared with other conditions probably because of a weaker influence of sugars on the pre-cultivation compared with harsher conditions (e.g., aldehydes). Fourteen out of 24 strains presented a minimum 20% increase in the specific growth rate of the second cultivation, with nine strains boosting performance by more than 50%. The only strain that did not show mean improvement was S288C (Fig 4B).

Figure 4.
Figure 4. Exposure to the same media improves the specific growth rate performance of tested strains during cultivation.

(A) Plot of the specific growth rate of the first cultivation and second cultivation for each strain and perturbation. (B) Percentage of improvement of the maximum specific growth rate calculated using Equation (2). Data information: in panel (A), each color represents a different group of perturbations and includes the corresponding linear regression line (mainly to check direction and strenght of the relationship). Spearman correlation coefficients are shown in the right part of the graph for each group of perturbations, together with the P-value denoting the significance of the correlation. P < 0.005 indicates strong statistical significance. In panel (B), the mean improvement in performance of the maximum specific growth rate (%P) was calculated with Equation (2) across perturbations for each strain (barplots). The SD is shown for each strain calculated from the triplicates. The colored dots correspond to the %P divided by group of perturbations for each strain.

The improved maximum specific growth rate in the second cultivation was strongly influenced by the condition and differed for each strain. LBCM1001, LBCM1017, and LBCM110 displayed outstanding improvement in specific growth in the presence of ethanol and, in the case of LBCM1017 and LBCM110, also in the presence of acids. CEN.PK113-7D was the only strain that never exhibit a decrease in its maximum specific growth rate during the second cultivation (Fig 4B). Interestingly, the higher was the concentration of certain inhibitors (acids, furfural, 5-hydroxymethylfurfural and mannose) in the medium, the larger was the percentage of improvement of the maximum specific growth rate (Fig S4). A correlation between the percentage of improvement of the maximum specific growth rate (%P) and the robustness of each of the phenotypes was not found. Overall, exposing a strain to the same stressor prior main cultivation had a positive effect on its tolerance (increased specific growth rate).

Figure S4.
Figure S4. Percentage of improvement in the maximum specific growth rate across strains.

Data information: perturbations are listed on the X-axis in g/liter unit, percentage of improvement of the maximum specific growth rate in the second cultivation on the Y-axis. Error bars were calculated with the SD among triplicates.

Discussion

In the present study, we investigated a large volume of phenotypic data from 24 S. cerevisiae strains grown in 29 different conditions, simulating the lignocellulose biomass fermentation perturbation space. Phenotypic characterization combined with robustness quantification allowed us to map the condition-specific influence on performance and robustness. Accordingly, we demonstrated the positive influence of the propagation step on performance. We also discovered and proved experimentally the trade-off between performance and robustness with respect to biomass yield, ethanol yield, and cell dry weight.

Each group of conditions affected performance in a consistent manner. Similar phenotypic responses may be linked to shared metabolic processes, such as the accumulation of reactive oxygen species under acid and aldehyde stress (Allen et al, 2010; Guaragnella & Bettiga, 2021). Using biosensors to examine the intracellular state can be a valuable approach for studying these patterns (Torello Pianale et al, 2022). By exploring the full range of phenotypes within the perturbation space, we were able to identify novel behaviors. For example, the presence of pentoses and low glucose (5 g/liter) led to an increase in ethanol yield, despite pentoses not being used as carbon and energy source by S. cerevisiae. We hypothesize that this might be attributed to an increased efficiency of some hexose transporters (specifically Hxt2p and Hxt7p [Özcan & Johnston, 1995; Reifenberger et al, 1997; Subtil & Boles, 2012]), or to the contribution of aldose reductases (Richard et al, 1999; Träff et al, 2001, 2002; Ford & Ellis, 2002).

A drop in performance is expected when cells are exposed to perturbations, as they direct nutrients and energy towards maintenance rather than growth or production (Stanley et al, 2010; Vos et al, 2016). However, our results show that poorly performing strains, such as LBCM1095 and LBCM1079, maintained a robust behavior in the tested perturbation space. We speculate that in strains with robust phenotypes, part of the energy and resources may be directed towards additional regulatory and homeostasis-maintenance pathways even when no environmental perturbations are present, at the expense of lower overall performance.

Two main mechanisms could explain the strong robustness exhibited by LCBM strains and Ethanol Red. The first is bet-hedging, whereby cell-to-cell variation within an isogenic population serves as a survival strategy under different conditions (Bagamery et al, 2020). Strains from the LBCM collection have been isolated from cachaça distilleries around Brazil and resulted from hybridization with Saccharomyces strains of varied origin (da Conceição et al, 2015; Araújo et al, 2018; Jacobus et al, 2021). Because of the introduction of new and diversified genetic material, the LBCM populations might have been able to randomly diversify their phenotypes, compared with, for example, CEN.PK113-7D, thereby responding more consistently to environmental perturbations. Bet-hedging implies sometimes a reduced mean fitness (Olofsson et al, 2009), which might explain the low performance of LBCM1079 and LBCM1095. Analysis of the Ethanol Red phylogeny has revealed its close relation to S288C, Y22-3, wine strains, and sugarcane strains, suggesting that its genetic hybridization might also be responsible for its robustness mechanisms (Gronchi et al, 2022). The second mechanism is cross-protection (Dhar et al, 2013), which arises when cells counteract new perturbations based on previous exposure to the same or different perturbations. Even though the robustness of Ethanol Red was slightly lower than that of LBCM1079 and LBCM1095, it nevertheless achieved the highest performance across strains. In Ethanol Red, robustness could also have developed because of cross-resistance mechanisms arising from exposure to different substrates, such as wine, sugarcane or starch.

Earlier investigations have shown that the inoculum might negatively influence the bioprocess outcome mainly because of variations in the physiological and metabolic states of the transferred cells (Keil et al, 2019). We observed a positive correlation between the specific growth rate in the first and second cultivations, and an overall increase of the maximum specific growth rate suggesting that exposing the cells to the same condition before fermentation had a positive effect on their performance, irrespective of perturbation type. Changes in genomic expression patterns triggered by a certain stressor are not required to survive exposure the encountered stressors, but they are fundamental to survive exposure to the same or different stress at a later time (Święciło, 2016). Furthermore, previous evidence has shown that short-term adaptation or acclimatization (van Dijk et al, 2019; Bergen et al, 2022) selects for phenotypes that are more tolerant to previously faced inhibitors (Nielsen et al, 2015). Even if an overall negative correlation was not found between the percentage of improvement of the maximum specific growth rate and the robustness, because of how robustness is defined, we speculate that an increased maximum specific growth rate in the second cultivation might be an indicator of poor robustness of the same phenotype.

The observed significant negative correlations between robustness and performance confirmed the hypothesized trade-offs between these two properties. Whereas others have suggested that robustness and performance always result in a trade-off (Whitacre, 2012), we found that trade-offs applied only to biomass yield, ethanol yield, and cell dry weight. Instead, specific growth rate exhibited a positive correlation between performance and robustness. This can be explained by specific growth rate having been selected and optimized in many environments through evolution to allow robust-faster-growing cells to dominate the population (Dragosits & Mattanovich, 2013). Natural selection could account for the robust specific growth rate phenotype in our study, and the observed positive correlation between performance and robustness. In contrast, evolutionary mechanisms do not typically favor increased yields, and this lack of optimization may explain the negative correlations we observed between performance and robustness for ethanol and biomass yields and cell dry weight. Similar trade-offs between robustness and performance have been discovered before when measuring ATP and biomass yields (Schuetz et al, 2012). In that study, researchers demonstrated that there is a greater likelihood of survival for cells that prioritize robustness (can switch fast between environments) over maximum performance (highly optimized cells).

Trade-offs appeared in two different forms in our study: first, by way of negative correlations observed between robustness and performance; second, as a reduced performance in some conditions. Yeast cells prioritize different metabolic processes depending on the environment. Because of physical and thermodynamic limitations (Niebel et al, 2019), it is impossible to maximize all metabolic processes simultaneously, resulting in trade-offs.

In conclusion, the design of microbial cell factories would benefit from studies that could reveal trade-off among cellular properties such as in silico objective optimization studies. The methodology and analysis we just presented could help in validation and integration of such simulations. Future research should evaluate more perturbation spaces and different microorganisms, including bacteria and fungi, to determine whether trade-offs follow a more general pattern or should be constrained within singular perturbation spaces and organisms. Finally, studies on evolution of performance and robustness are needed to dig into the possible molecular and genetic markers associated to the two properties, specifically highlighting differences when trade-offs are present.

Materials and Methods

Strains

A total of 24 S. cerevisiae strains were used. CEN.PK113-7D (Entian & Kötter, 2007) (Scientific Research and Development GmbH) and S288C (University of Milano Bicocca) were two representative laboratory strains. Four industrial strains included: the PE2 WT strain isolated during sugarcane-to-ethanol production (Basso et al, 2008), Ethanol Red (kindly provided by Société Industrielle Lesaffre, Division Leaf), Thermosacc (Lallemand Ethanol Technology), and Red Star (Red Star Yeast, Lesaffre). A set of strains from the LBCM collection (LBCM, Universitade Federal de Ouro Preto, laboratorio de biologia cellular e molecular) (De Araújo Vicente et al, 2006; Araújo et al, 2018) sampled from cachaça production sites in Brazil were selected: LBCM1001, LBCM1003, LBCM1008, LBCM1013, LBCM1014, LBCM1017, LBCM1030, LBCM1046, LBCM1079, LBCM1095, LBCM1099, LBCM1106, LBCM37, LBCM67, LBCM97, LBCM103, LBCM109, and LBCM110. Cachaça strains were chosen because of their ability to tolerate high concentration of ethanol and perturbations that resemble the one found in second generation biomass fermentation.

Performance measurement

Performance and robustness measurements were carried out as detailed in our previous study (Trivellin et al, 2022). In short, the yeast strains were grown in Delft chemically defined medium (Bruinenberg et al, 1983; Verduyn et al, 1992) containing 5 g/liter (NH4)2SO4, 3 g/liter KH2PO4, 1 g/liter MgSO4∙7H2O, 1 ml trace mineral solution (per L of medium), and 1 ml vitamin solution (per L of medium) (Trivellin et al, 2022). The medium was adjusted to pH 5 with KOH and buffered with 250 mM potassium hydrogen phthalate. Acids, sugars, aldehydes, NaCl, and ethanol were added to the Delft medium to simulate perturbations common during lignocellulosic bioethanol production (Table S1) (Olsson & Hahn-Hägerdal, 1996; Palmqvist et al, 1998; Hohmann, 2002; Koppram et al, 2012; Cavka & Jönsson, 2013; Favaro et al, 2013; Kim, 2018; van Dijk et al, 2019). Briefly, 10 μl glycerol stock containing the S. cerevisiae strains were inoculated in 5 ml Delft medium. Precultures were grown overnight at 30°C on a orbital shaker (orbit 1.9 cm) at 250 rpm speed. Screenings were carried out for 48 h, at 30°C, 250 rpm/50 mm orbital shaking in a growth profiler (960; Enzyscreen) using 96-well microtiter plates (CR1496dg) covered with a CO2-release cover (CR1296t), and with a starting OD600 of 0.02 in 250 μl.

The final OD600 of the cultivation was measured in a plate reader (SPECTROstar nano; BMG LABTECH). The OD values were related to the cell dry weight through a calibration curve previously built for each specific strain in Delft medium (Trivellin et al, 2022). After 48 h of cultivation, the culture broths were filtered, and ethanol and sugar concentrations were determined by enzymatic assays (K-ETOH Ethanol Assay Kit, K-GLUHK-220A D-Glucose HK Assay Kit, K-MANGL D-Mannose/D-Fructose/D-Glucose Assay kit, K-XYLOSE D-Xylose Assay Kit, and K-ARGA L-Arabinose/D-Galactose Assay Kit; Megazyme). Sugar and ethanol concentrations were used to calculate ethanol and biomass yields (Trivellin et al, 2022). Data from the growth profiler (green values) were used to calculate the maximum specific growth rate and length of the lag phase (see Data Availability).

Robustness quantification

Robustness of each strain (S) for a defined phenotype (i) across a set of perturbations (P) was calculated with the following equation:RS,i,P=−Fano factormean=−σ2x¯∙1m(1)

The Fano Factor (variance (σ2) divided by the mean (x¯) across perturbations) was normalized to the mean of the phenotype across all strains (m).

Investigation of the culture transfer on robustness and performance

To assess the impact of the culture transfer, at the end of the first screening (48 h), 5 μl of culture were reinoculated in 250 μl fresh medium as described above. The plates used in the second screening were incubated for another 48 h under the same conditions as during the first screening. At the end of the second screening, green values were used to calculate the maximum specific growth rate of the second cultivation. The percentage of improvement in maximum specific growth rate performance was measured with the following equation (μmax1: maximum specific growth rate of the first cultivation; μmax2: maximum specific growth rate of the second cultivation).%Pµmax=µmax2−µmax1µmax1*100(2)

When μmax1 was zero and μmax2 was a finite value, the ratio resulted in an infinite value, which created problems regarding the interpretation of the results. Therefore, infinite values were set to the maximum finite improvement value calculated from the dataset. Spearman correlation tests were performed between the robustness values of all the strains for each phenotype and the mean %P of each strain (scripts with in-line description are available on GitHub [See Data Availability]).

Data analysis

Dataset containing raw values was preprocessed before performance and robustness analysis. The dataset was trimmed by taking in consideration theoretical values of the yields and experimental errors. The upper boundary for the yields was set at the maximum theoretical ethanol yield on glucose (0.51 + 0.1 g/g), whereas the maximum biomass concentration (39.6 g/liter) was calculated based on the maximum biomass yield measured in chemostat cultivations (0.1 1/h dilution rate [Verduyn, 1991]) and maximum substrate used (65 g/liter glucose). A total of 145 values were excluded from the analysis (out of 10,440) (Fig S5). The Shapiro normality test revealed that the phenotypes did not follow a normal distribution (P < 1 × 10−38); instead, distribution was primarily skewed and multimodal (mostly bimodal) (Figs S6, S7, S8, S9, and S10). The quantile method (percentiles of 0.1% and 99.9%) was used to identify outliers on the trimmed dataset. Thirteen outliers were identified across all the variables corresponding to seven strains (six LBCM strains) and seven perturbations (Fig S11). The outliers were not removed from the dataset. Ethanol yields in the presence of ethanol as a perturbation were excluded from the analysis as we could not differentiate between its consumption and production during cultivation. Spearman’s rank-order correlations were performed on the data and associated with their statistical significance (P-value). The linear regression lines were included in the graphs mainly to visualize the direction and strenght of the relationships. The standard error of the correlation based on sample size was calculated for performance versus robustness correlations (Bonett & Wright, 2000). Statistical tests and plots were generated in R and scripts with line-by-line descriptions are available from GitHub (See Data Availability).

Figure S5.
Figure S5. Data distribution for all strains, perturbations, and phenotypes.

(A) Distribution of raw data for each phenotype. (B) Distribution of data trimmed to biologically reasonable intervals (see the Materials and Methods section). Data information: the horizontal line in the distributions corresponds to the median.

Figure S6.
Figure S6. Density plots of the cell dry weight (g/liter) obtained using the trimmed dataset.

Data information: The cell dry weight distribution is plotted for each strain. All perturbations tested in the study are considered. Each color corresponds to a strain group and each small vertical line on the X-axis represents a single data point.

Figure S7.
Figure S7. Density plots of the maximum specific growth rate (1/h) obtained using the trimmed dataset.

Data information: The maximum specific growth rate distribution is plotted for each strain. All perturbations tested in the study are considered. Each color corresponds to a strain group and each small vertical line on the X-axis represents a single data point.

Figure S8.
Figure S8. Density plots of the lag phase (h) obtained using the trimmed dataset.

Data information: The lag phase distribution is plotted for each strain. All perturbations tested in the study are considered. Each color corresponds to a strain group and each small vertical line on the X-axis represents a single data point.

Figure S9.
Figure S9. Density plots of the ethanol yield (g/g) obtained using the trimmed dataset.

Data information: The ethanol yield distribution is plotted for each strain. All perturbations tested in the study are considered. Each color corresponds to a strain group and each small vertical line on the X-axis represents a single data point.

Figure S10.
Figure S10. Density plots of the biomass yield (g/g) obtained using the trimmed dataset.

Data information: The biomass yield distribution is plotted for each strain. All perturbations tested in the study are considered. Each color corresponds to a strain group and each small vertical line on the X-axis represents a single data point.

Figure S11.
Figure S11. Outlier detection based on the quantile method after data trimming.

Data information: each point on the graph corresponds to one or more outliers detected and identified with the quantile method (0.001% and 0.999%). No overlap among different phenotypes was detected. Each color corresponds to a variable. The Y-axis includes seven strains (one industrial and six laboratorio de biologia cellular e molecular strains); the X-axis includes seven perturbations (two hexoses, one aldehyde, two pentoses, one acid, and one alcohol) with relative concentrations in g/liter.

Data Availability

The datasets and R code produced in this study are available on GitHub (https://github.com/cectri/Robustness_Trade-offs/tree/main). The repository contains: raw and processed data, scripts to generate phenotypic variables and preprocessing, exploratory analysis, robustness calculations, correlations, and percentage of improvement scripts. Scripts were generated using previously published R packages (Soetaert et al, 2010; Hall et al, 2014; Giner & Smyth, 2016; Wickham, 2016; McInnes et al, 2018; Bolar, 2019; Neuwirth, 2022; Schloerke et al, 2022; Wilke, 2022; Yutani, 2022; Erik Clarke, 2023; Kassambara, 2023a, 2023b; R Core Team, 2023; Tierney & Cook, 2023; Wickham et al, 2023; Xiao, 2023).

Acknowledgements

Financial support by Novo Nordisk Foundation grant DISTINGUISHED INVESTIGATOR 2019—Research within biotechnology-based synthesis and production (#0055044) is gratefully acknowledged. Société Industrielle Lesaffre, Division Leaf, is kindly acknowledged for providing the Ethanol Red strain.

Author Contributions

  • C Trivellin: conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, and writing—original draft, review, and editing.

  • P Rugbjerg: conceptualization, supervision, validation, methodology, project administration, and writing—review and editing.

  • L Olsson: conceptualization, resources, supervision, funding acquisition, methodology, project administration, and writing—review and editing.

Conflict of Interest Statement

The authors declare that they have no conflict of interest.

  • Received June 13, 2023.
  • Revision received October 20, 2023.
  • Accepted October 20, 2023.
Creative Commons logoCreative Commons logo https://creativecommons.org/licenses/by/4.0/

This article is available under a Creative Commons License (Attribution 4.0 International, as described at https://creativecommons.org/licenses/by/4.0/).

References

  1. ↵
    1. Adeboye PT,
    2. Bettiga M,
    3. Olsson L
    (2014) The chemical nature of phenolic compounds determines their toxicity and induces distinct physiological responses in Saccharomyces cerevisiae in lignocellulose hydrolysates. AMB Express 4: 46. doi:10.1186/s13568-014-0046-7
    OpenUrlCrossRef
  2. ↵
    1. Allen SA,
    2. Clark W,
    3. McCaffery JM,
    4. Cai Z,
    5. Lanctot A,
    6. Slininger PJ,
    7. Liu ZL,
    8. Gorsich SW
    (2010) Furfural induces reactive oxygen species accumulation and cellular damage in Saccharomyces cerevisiae. Biotechnol Biofuels 3: 2–10. doi:10.1186/1754-6834-3-2
    OpenUrlCrossRefPubMed
  3. ↵
    1. Andersson DI
    (2006) The biological cost of mutational antibiotic resistance: Any practical conclusions? Curr Opin Microbiol 9: 461–465. doi:10.1016/j.mib.2006.07.002
    OpenUrlCrossRefPubMed
  4. ↵
    1. Araújo TM,
    2. Souza MT,
    3. Diniz RHS,
    4. Yamakawa CK,
    5. Soares LB,
    6. Lenczak JL,
    7. de Castro Oliveira JV,
    8. Goldman GH,
    9. Barbosa EA,
    10. Campos ACS, et al.
    (2018) Cachaça yeast strains: Alternative starters to produce beer and bioethanol. Antonie van Leeuwenhoek 111: 1749–1766. doi:10.1007/s10482-018-1063-3
    OpenUrlCrossRef
  5. ↵
    1. Bagamery LE,
    2. Justman QA,
    3. Garner EC,
    4. Murray AW
    (2020) A putative bet-hedging strategy buffers budding yeast against environmental instability. Curr Biol 30: 4563–4578.e4. doi:10.1016/j.cub.2020.08.092
    OpenUrlCrossRef
  6. ↵
    1. Barkai N,
    2. Leibler S
    (1997) Robustness in simple biochemical networks. Nature 387: 913–917. doi:10.1038/43199
    OpenUrlCrossRefPubMed
  7. ↵
    1. Basan M,
    2. Honda T,
    3. Christodoulou D,
    4. Hörl M,
    5. Chang YF,
    6. Leoncini E,
    7. Mukherjee A,
    8. Okano H,
    9. Taylor BR,
    10. Silverman JM, et al.
    (2020) A universal trade-off between growth and lag in fluctuating environments. Nature 584: 470–474. doi:10.1038/s41586-020-2505-4
    OpenUrlCrossRef
  8. ↵
    1. Basso LC,
    2. De Amorim HV,
    3. De Oliveira AJ,
    4. Lopes ML
    (2008) Yeast selection for fuel ethanol production in Brazil FEMS Yeast Research. Oxford: Oxford Academic: 1155–1163.
  9. ↵
    1. Bergen AC,
    2. Kocik RA,
    3. Hose J,
    4. McClean MN,
    5. Gasch AP
    (2022) Modeling single-cell phenotypes links yeast stress acclimation to transcriptional repression and pre-stress cellular states. Elife 11: e82017. doi:10.7554/eLife.82017
    OpenUrlCrossRef
  10. ↵
    1. Bolar K
    (2019) STAT: Interactive Document for Working with Basic Statistical Analysis. https://cran.r-project.org/web/packages/STAT/index.html
  11. ↵
    1. Bonett DG,
    2. Wright TA
    (2000) Sample size requirements for estimating Pearson, Kendall and Spearman correlations. Psychometrika 65: 23–28. doi:10.1007/bf02294183
    OpenUrlCrossRef
  12. ↵
    1. Bruinenberg PM,
    2. Van Dijken JP,
    3. Scheffers WA
    (1983) An enzymic analysis of NADPH production and consumption in candida utilis. J Gen Microbiol 129: 965–971. doi:10.1099/00221287-129-4-965
    OpenUrlCrossRefPubMed
  13. ↵
    1. Carlson JM,
    2. Doyle J
    (2000) Highly optimized tolerance: Robustness and design in complex systems. Phys Rev Lett 84: 2529–2532. doi:10.1103/PhysRevLett.84.2529
    OpenUrlCrossRefPubMed
  14. ↵
    1. Caspeta L,
    2. Coronel J,
    3. Montes de Oca A,
    4. Abarca E,
    5. González L,
    6. Martínez A
    (2019) Engineering high-gravity fermentations for ethanol production at elevated temperature with Saccharomyces cerevisiae. Biotechnol Bioeng 116: 2587–2597. doi:10.1002/bit.27103
    OpenUrlCrossRef
  15. ↵
    1. Cavka A,
    2. Jönsson LJ
    (2013) Detoxification of lignocellulosic hydrolysates using sodium borohydride. Bioresour Technol 136: 368–376. doi:10.1016/j.biortech.2013.03.014
    OpenUrlCrossRef
  16. ↵
    1. Cripwell RA,
    2. Rose SH,
    3. Favaro L,
    4. Van Zyl WH
    (2019) Construction of industrial Saccharomyces cerevisiae strains for the efficient consolidated bioprocessing of raw starch. Biotechnol Biofuels 12: 201–216. doi:10.1186/s13068-019-1541-5
    OpenUrlCrossRef
  17. ↵
    1. Cunha JT,
    2. Soares PO,
    3. Romaní A,
    4. Thevelein JM,
    5. Domingues L
    (2019) Xylose fermentation efficiency of industrial Saccharomyces cerevisiae yeast with separate or combined xylose reductase/xylitol dehydrogenase and xylose isomerase pathways. Biotechnol Biofuels 12: 20. doi:10.1186/s13068-019-1360-8
    OpenUrlCrossRef
  18. ↵
    1. da Conceição LEFR,
    2. Saraiva MAF,
    3. Diniz RHS,
    4. Oliveira J,
    5. Barbosa GD,
    6. Alvarez F,
    7. Correa LfdM,
    8. Mezadri H,
    9. Coutrim MX,
    10. Afonso RJDCF, et al.
    (2015) Biotechnological potential of yeast isolates from cachaça: The Brazilian spirit. J Ind Microbiol Biotechnol 42: 237–246. doi:10.1007/s10295-014-1528-y
    OpenUrlCrossRef
  19. ↵
    1. Vicente MdA,
    2. Fietto LG,
    3. Castro IdM,
    4. dos Santos ANG,
    5. Coutrim MX,
    6. Brandão RL
    (2006) Isolation of Saccharomyces cerevisiae strains producing higher levels of flavoring compounds for production of “cachaça” the Brazilian sugarcane spirit. Int J Food Microbiol 108: 51–59. doi:10.1016/j.ijfoodmicro.2005.10.018
    OpenUrlCrossRefPubMed
  20. ↵
    1. Demeke MM,
    2. Dietz H,
    3. Li Y,
    4. Foulquié-Moreno MR,
    5. Mutturi S,
    6. Deprez S,
    7. Den Abt T,
    8. Bonini BM,
    9. Liden G,
    10. Dumortier F, et al.
    (2013) Development of a D-xylose fermenting and inhibitor tolerant industrial Saccharomyces cerevisiae strain with high performance in lignocellulose hydrolysates using metabolic and evolutionary engineering. Biotechnol Biofuels 6: 89. doi:10.1186/1754-6834-6-89
    OpenUrlCrossRefPubMed
  21. ↵
    1. Dhar R,
    2. Sägesser R,
    3. Weikert C,
    4. Wagner A
    (2013) Yeast adapts to a changing stressful environment by evolving cross-protection and anticipatory gene regulation. Mol Biol Evol 30: 573–588. doi:10.1093/molbev/mss253
    OpenUrlCrossRefPubMed
  22. ↵
    1. Dragosits M,
    2. Mattanovich D
    (2013) Adaptive laboratory evolution - principles and applications for biotechnology. Microb Cell Fact 12: 64. doi:10.1186/1475-2859-12-64
    OpenUrlCrossRefPubMed
  23. ↵
    1. Entian KD,
    2. Kötter P
    (2007) 25 yeast genetic strain and plasmid collections. Methods Microbiol 36: 629–666. doi:10.1016/S0580-9517(06)36025-4
    OpenUrlCrossRef
  24. ↵
    1. Erik Clarke SS-MD
    (2023) Categorical Scatter (Violin Point) Plots [R Package ggbeeswarm version 0.7.2]. https://CRAN.R-project.org/package=ggbeeswarm
  25. ↵
    1. Favaro L,
    2. Basaglia M,
    3. Trento A,
    4. Van Rensburg E,
    5. García-Aparicio M,
    6. Van Zyl WH,
    7. Casella S
    (2013) Exploring grape marc as trove for new thermotolerant and inhibitor-tolerant Saccharomyces cerevisiae strains for second-generation bioethanol production. Biotechnol Biofuels 6: 168. doi:10.1186/1754-6834-6-168
    OpenUrlCrossRefPubMed
  26. ↵
    1. Félix M-AA,
    2. Barkoulas M
    (2015) Pervasive robustness in biological systems. Nat Rev Genet 16: 483–496. doi:10.1038/nrg3949
    OpenUrlCrossRefPubMed
  27. ↵
    1. Fischer E,
    2. Sauer U
    (2005) Large-scale in vivo flux analysis shows rigidity and suboptimal performance of Bacillus subtilis metabolism. Nat Genet 37: 636–640. doi:10.1038/ng1555
    OpenUrlCrossRefPubMed
  28. ↵
    1. Ford G,
    2. Ellis EM
    (2002) Characterization of Ypr1p from Saccharomyces cerevisiae as a 2-methylbutyraldehyde reductase. Yeast 19: 1087–1096. doi:10.1002/yea.899
    OpenUrlCrossRefPubMed
  29. ↵
    1. Fox JF,
    2. Stevens GC
    (1991) Costs of reproduction in a willow: Experimental responses vs. Natural variation. Ecology 72: 1013–1023. doi:10.2307/1940601
    OpenUrlCrossRef
  30. ↵
    1. Giner G,
    2. Smyth GK
    (2016) Statmod: Probability calculations for the inverse Gaussian distribution. R J 8: 339–351. doi:10.32614/rj-2016-024
    OpenUrlCrossRef
  31. ↵
    1. Gronchi N,
    2. de Bernardini N,
    3. Cripwell RA,
    4. Treu L,
    5. Campanaro S,
    6. Basaglia M,
    7. Foulquié-Moreno MR,
    8. Thevelein JM,
    9. van Zyl WH,
    10. Favaro L, et al.
    (2022) Natural Saccharomyces cerevisiae strain reveals peculiar genomic traits for starch-to-bioethanol production: The design of an amylolytic consolidated bioprocessing yeast. Front Microbiol 12: 768562. doi:10.3389/fmicb.2021.768562
    OpenUrlCrossRef
  32. ↵
    1. Guaragnella N,
    2. Bettiga M
    (2021) Acetic acid stress in budding yeast: From molecular mechanisms to applications. Yeast 38: 391–400. doi:10.1002/yea.3651
    OpenUrlCrossRef
  33. ↵
    1. Hall BG,
    2. Acar H,
    3. Nandipati A,
    4. Barlow M
    (2014) Growth rates made easy. Mol Biol Evol 31: 232–238. doi:10.1093/molbev/mst187
    OpenUrlCrossRefPubMed
  34. ↵
    1. Hohmann S
    (2002) Osmotic stress signaling and osmoadaptation in yeasts. Microbiol Mol Biol Rev 66: 300–372. doi:10.1128/MMBR.66.2.300-372.2002
    OpenUrlAbstract/FREE Full Text
  35. ↵
    1. Van Dien S
    1. Huang M,
    2. Peabody G,
    3. Kao KC
    (2016) Tolerance of microbial biocatalysts to feedstocks, products, and environmental conditions. In Metabolic Engineering for Bioprocess Commercialization. Van Dien S (ed.). Cham: Springer.
  36. ↵
    1. Ibarra RU,
    2. Edwards JS,
    3. Palsson BO
    (2002) Escherichia coli K-12 undergoes adaptive evolution to achieve in silico predicted optimal growth. Nature 420: 186–189. doi:10.1038/nature01149
    OpenUrlCrossRefPubMed
  37. ↵
    1. Jacobus AP,
    2. Stephens TG,
    3. Youssef P,
    4. González-Pech R,
    5. Ciccotosto-Camp MM,
    6. Dougan KE,
    7. Chen Y,
    8. Basso LC,
    9. Frazzon J,
    10. Chan CX, et al.
    (2021) Comparative genomics supports that Brazilian bioethanol Saccharomyces cerevisiae comprise a unified group of domesticated strains related to cachaça spirit yeasts. Front Microbiol 12: 644089. doi:10.3389/fmicb.2021.644089
    OpenUrlCrossRef
  38. ↵
    1. Jönsson LJ,
    2. Martín C
    (2016) Pretreatment of lignocellulose: Formation of inhibitory by-products and strategies for minimizing their effects. Bioresour Technol 199: 103–112. doi:10.1016/j.biortech.2015.10.009
    OpenUrlCrossRefPubMed
  39. ↵
    1. Kassambara A
    (2023a) ggpubr: ‘ggplot2’ Based Publication Ready Plots. https://rpkgs.datanovia.com/ggpubr/
  40. ↵
    1. Kassambara A
    (2023b) rstatix: Pipe-Friendly Framework for Basic Statistical Tests. https://rpkgs.datanovia.com/rstatix/
  41. ↵
    1. Keil T,
    2. Landenberger M,
    3. Dittrich B,
    4. Selzer S,
    5. Büchs J
    (2019) Precultures grown under fed-batch conditions increase the reliability and reproducibility of high-throughput screening results. Biotechnol J 14: 1800727. doi:10.1002/biot.201800727
    OpenUrlCrossRef
  42. ↵
    1. Kim D
    (2018) Physico-chemical conversion of lignocellulose: Inhibitor effects and detoxification strategies: A mini review. Molecules 23: 309. doi:10.3390/molecules23020309
    OpenUrlCrossRef
  43. ↵
    1. Kitano H
    (2004) Biological robustness. Nat Rev Genet 5: 826–837. doi:10.1038/nrg1471
    OpenUrlCrossRefPubMed
  44. ↵
    1. Kitano H
    (2007) Towards a theory of biological robustness. Mol Syst Biol 3: 137. doi:10.1038/msb4100179
    OpenUrlFREE Full Text
  45. ↵
    1. Kitano H
    (2010) Violations of robustness trade-offs. Mol Syst Biol 6: 384. doi:10.1038/msb.2010.40
    OpenUrlAbstract/FREE Full Text
  46. ↵
    1. Kong II,
    2. Turner TL,
    3. Kim H,
    4. Kim SR,
    5. Jin YS
    (2018) Phenotypic evaluation and characterization of 21 industrial Saccharomyces cerevisiae yeast strains. FEMS Yeast Res 18: 1. doi:10.1093/femsyr/foy001
    OpenUrlCrossRef
  47. ↵
    1. Koppram R,
    2. Albers E,
    3. Olsson L
    (2012) Evolutionary engineering strategies to enhance tolerance of xylose utilizing recombinant yeast to inhibitors derived from spruce biomass. Biotechnol Biofuels 5: 32. doi:10.1186/1754-6834-5-32
    OpenUrlCrossRefPubMed
  48. ↵
    1. Levy SF,
    2. Siegal ML,
    3. Levy SF,
    4. Siegal ML
    (2012) The robustness continuum. Adv Exp Med Biol 751: 431–452. doi:10.1007/978-1-4614-3567-9_20
    OpenUrlCrossRefPubMed
  49. ↵
    1. Louis EJ
    (2016) Corrigendum: Historical evolution of laboratory strains of Saccharomyces cerevisiae. Cold Spring Harb Protoc 2016. doi:10.1101/pdb.corr095976
    OpenUrlCrossRef
  50. ↵
    1. Lovero KG,
    2. Treseder KK
    (2021) Trade-offs between growth rate and other fungal traits. Front Forests Glob Change 4: 197. doi:10.3389/ffgc.2021.756650
    OpenUrlCrossRef
  51. ↵
    1. Masel J,
    2. Siegal ML
    (2009) Robustness: Mechanisms and consequences. Trends Genet 25: 395–403. doi:10.1016/j.tig.2009.07.005
    OpenUrlCrossRefPubMed
  52. ↵
    1. McInnes L,
    2. Healy J,
    3. Melville J
    (2018) UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. https://cran.r-project.org/web/packages/uwot/index.html
  53. ↵
    1. Mestek Boukhibar L,
    2. Barkoulas M
    (2016) The developmental genetics of biological robustness. Ann Bot 117: 699–707. doi:10.1093/aob/mcv128
    OpenUrlCrossRefPubMed
  54. ↵
    1. Mohedano MT,
    2. Konzock O,
    3. Chen Y
    (2022) Strategies to increase tolerance and robustness of industrial microorganisms. Synth Syst Biotechnol 7: 533–540. doi:10.1016/j.synbio.2021.12.009
    OpenUrlCrossRef
  55. ↵
    1. Neuwirth E
    (2022) ColorBrewer Palettes [R Package RColorBrewer Version 1.1-3]. https://CRAN.R-project.org/package=RColorBrewer
  56. ↵
    1. Niebel B,
    2. Leupold S,
    3. Heinemann M
    (2019) An upper limit on Gibbs energy dissipation governs cellular metabolism. Nat Metab 1: 125–132. doi:10.1038/s42255-018-0006-7
    OpenUrlCrossRef
  57. ↵
    1. Nielsen F,
    2. Tomás-Pejó E,
    3. Olsson L,
    4. Wallberg O
    (2015) Short-term adaptation during propagation improves the performance of xylose-fermenting Saccharomyces cerevisiae in simultaneous saccharification and co-fermentation. Biotechnol Biofuels 8: 1–15. doi:10.1186/s13068-015-0399-4
    OpenUrlCrossRefPubMed
  58. ↵
    1. Olofsson H,
    2. Ripa J,
    3. Jonzén N
    (2009) Bet-hedging as an evolutionary game: The trade-off between egg size and number. Proc Biol Sci 276: 2963–2969. doi:10.1098/rspb.2009.0500
    OpenUrlCrossRefPubMed
  59. ↵
    1. Olsson L,
    2. Hahn-Hägerdal B
    (1996) Fermentation of lignocellulosic hydrolysates for ethanol production. Enzyme Microb Technol 18: 312–331. doi:10.1016/0141-0229(95)00157-3
    OpenUrlCrossRef
  60. ↵
    1. Olsson L,
    2. Rugbjerg P,
    3. Torello Pianale L,
    4. Trivellin C
    (2022) Robustness: Linking strain design to viable bioprocesses. Trends Biotechnol 40: 918–931. doi:10.1016/j.tibtech.2022.01.004
    OpenUrlCrossRef
  61. ↵
    1. Özcan S,
    2. Johnston M
    (1995) Three different regulatory mechanisms enable yeast hexose transporter (HXT) genes to Be induced by different levels of glucose. Mol Cell Biol 15: 1564–1572. doi:10.1128/MCB.15.3.1564
    OpenUrlAbstract/FREE Full Text
  62. ↵
    1. Palmqvist E,
    2. Galbe M,
    3. Hahn-Hägerdal B
    (1998) Evaluation of cell recycling in continuous fermentation of enzymatic hydrolysates of spruce with Saccharomyces cerevisiae and on-line monitoring of glucose and ethanol. Appl Microbiol Biotechnol 50: 545–551. doi:10.1007/s002530051332
    OpenUrlCrossRefPubMed
  63. ↵
    1. Pereira FB,
    2. Guimarães PMR,
    3. Teixeira JA,
    4. Domingues L
    (2010) Selection of Saccharomyces cerevisiae strains for efficient very high gravity bio-ethanol fermentation processes. Biotechnol Lett 32: 1655–1661. doi:10.1007/s10529-010-0330-9
    OpenUrlCrossRefPubMed
  64. ↵
    1. Pereira FB,
    2. Romaní A,
    3. Ruiz HA,
    4. Teixeira JA,
    5. Domingues L
    (2014) Industrial robust yeast isolates with great potential for fermentation of lignocellulosic biomass. Bioresour Technol 161: 192–199. doi:10.1016/j.biortech.2014.03.043
    OpenUrlCrossRefPubMed
  65. ↵
    1. Quinton-Tulloch MJ,
    2. Bruggeman FJ,
    3. Snoep JL,
    4. Westerhoff HV
    (2013) Trade-off of dynamic fragility but not of robustness in metabolic pathways in silico. FEBS J 280: 160–173. doi:10.1111/febs.12057
    OpenUrlCrossRefPubMed
  66. ↵
    1. R Core Team
    (2023) R: A Language and Environment for Statistical Computing. https://www.r-project.org/
  67. ↵
    1. Raghavendran V,
    2. Basso TP,
    3. da Silva JB,
    4. Basso LC,
    5. Gombert AK
    (2017) A simple scaled down system to mimic the industrial production of first generation fuel ethanol in Brazil. Antonie van Leeuwenhoek 110: 971–983. doi:10.1007/s10482-017-0868-9
    OpenUrlCrossRef
  68. ↵
    1. Ramin KI,
    2. Allison SD
    (2019) Bacterial tradeoffs in growth rate and extracellular enzymes. Front Microbiol 10: 2956. doi:10.3389/fmicb.2019.02956
    OpenUrlCrossRef
  69. ↵
    1. Reifenberger E,
    2. Boles E,
    3. Ciriacy M
    (1997) Kinetic characterization of individual hexose transporters of Saccharomyces cerevisiae and their relation to the triggering mechanisms of glucose repression. Eur J Biochem 245: 324–333. doi:10.1111/j.1432-1033.1997.00324.x
    OpenUrlCrossRefPubMed
  70. ↵
    1. Richard P,
    2. Toivari MH,
    3. Penttilä M
    (1999) Evidence that the gene YLR070c of Saccharomyces cerevisiae encodes a xylitol dehydrogenase. FEBS Lett 457: 135–138. doi:10.1016/s0014-5793(99)01016-9
    OpenUrlCrossRefPubMed
  71. ↵
    1. Schloerke B,
    2. Cook D,
    3. Joseph L,
    4. Briatte F,
    5. Marbach M,
    6. Edwin T,
    7. Amos E,
    8. Crowley J
    (2022) GGally: Extension to ‘ggplot2’. https://ggobi.github.io/ggally/
  72. ↵
    1. Schuetz R,
    2. Zamboni N,
    3. Zampieri M,
    4. Heinemann M,
    5. Sauer U
    (2012) Multidimensional optimality of microbial metabolism. Science 336: 601–604. doi:10.1126/science.1216882
    OpenUrlAbstract/FREE Full Text
  73. ↵
    1. Senne de Oliveira Lino F,
    2. Bajic D,
    3. Vila JCC,
    4. Sánchez A,
    5. Sommer MOA
    (2021) Complex yeast–bacteria interactions affect the yield of industrial ethanol fermentation. Nat Commun 12: 1–12. doi:10.1038/s41467-021-21844-7
    OpenUrlCrossRefPubMed
  74. ↵
    1. Soares-Costa A,
    2. Nakayama DG,
    3. Andrade LdF,
    4. Catelli LF,
    5. Bassi APG,
    6. Ceccato-Antonini SR,
    7. Henrique-Silva F
    (2014) Industrial PE-2 strain of Saccharomyces cerevisiae: From alcoholic fermentation to the production of recombinant proteins. N Biotechnol 31: 90–97. doi:10.1016/j.nbt.2013.08.005
    OpenUrlCrossRef
  75. ↵
    1. Soetaert K,
    2. Petzoldt T,
    3. Setzer RW
    (2010) Solving differential equations in R: Package deSolve. J Stat Softw 33: 1–25. doi:10.18637/jss.v033.i09
    OpenUrlCrossRefPubMed
  76. ↵
    1. Stanley D,
    2. Bandara A,
    3. Fraser S,
    4. Chambers PJ,
    5. Stanley GA
    (2010) The ethanol stress response and ethanol tolerance of Saccharomyces cerevisiae. J Appl Microbiol 109: 13–24. doi:10.1111/j.1365-2672.2009.04657.x
    OpenUrlCrossRefPubMed
  77. ↵
    1. Steensels J,
    2. Snoek T,
    3. Meersman E,
    4. Picca Nicolino M,
    5. Voordeckers K,
    6. Verstrepen KJ
    (2014) Improving industrial yeast strains: Exploiting natural and artificial diversity. FEMS Microbiol Rev 38: 947–995. doi:10.1111/1574-6976.12073
    OpenUrlCrossRefPubMed
  78. ↵
    1. Stelling J,
    2. Klamt S,
    3. Bettenbrock K,
    4. Schuster S,
    5. Gilles ED
    (2002) Metabolic network structure determines key aspects of functionality and regulation. Nature 420: 190–193. doi:10.1038/nature01166
    OpenUrlCrossRefPubMed
  79. ↵
    1. Subtil T,
    2. Boles E
    (2012) Competition between pentoses and glucose during uptake and catabolism in recombinant Saccharomyces cerevisiae. Biotechnol Biofuels 5: 14. doi:10.1186/1754-6834-5-14
    OpenUrlCrossRefPubMed
  80. ↵
    1. Święciło A
    (2016) Cross-stress resistance in Saccharomyces cerevisiae yeast—new insight into an old phenomenon. Cell Stress Chaperones 21: 187–200. doi:10.1007/s12192-016-0667-7
    OpenUrlCrossRefPubMed
  81. ↵
    1. Tekarslan-Sahin SH,
    2. Alkim C,
    3. Sezgin T
    (2018) Physiological and transcriptomic analysis of a salt-resistant Saccharomyces cerevisiae mutant obtained by evolutionary engineering. Bosn J Basic Med Sci 18: 55–65. doi:10.17305/bjbms.2017.2250
    OpenUrlCrossRef
  82. ↵
    1. Tierney N,
    2. Cook D
    (2023) Expanding tidy data principles to facilitate missing data exploration, visualization and assessment of imputations. J Stat Softw 105: 1–31. doi:10.18637/jss.v105.i07
    OpenUrlCrossRef
  83. ↵
    1. Tomás-Pejó E,
    2. Olsson L
    (2015) Influence of the propagation strategy for obtaining robust Saccharomyces cerevisiae cells that efficiently co-ferment xylose and glucose in lignocellulosic hydrolysates. Microb Biotechnol 8: 999–1005. doi:10.1111/1751-7915.12280
    OpenUrlCrossRef
  84. ↵
    1. Träff KL,
    2. Otero Cordero RR,
    3. van Zyl WH,
    4. Hahn-Hägerdal B
    (2001) Deletion of the GRE3 aldose reductase gene and its influence on xylose metabolism in recombinant strains of Saccharomyces cerevisiae expressing the xylA and XKS1 genes. Appl Environ Microbiol 67: 5668–5674. doi:10.1128/AEM.67.12.5668-5674.2001
    OpenUrlAbstract/FREE Full Text
  85. ↵
    1. Torello Pianale L,
    2. Rugbjerg P,
    3. Olsson L
    (2022) Real-time monitoring of the yeast intracellular state during bioprocesses with a toolbox of biosensors. Front Microbiol 12: 802169. doi:10.3389/fmicb.2021.802169
    OpenUrlCrossRef
  86. ↵
    1. Träff KL,
    2. Jönsson LJ,
    3. Hahn-Hägerdal B
    (2002) Putative xylose and arabinose reductases in Saccharomyces cerevisiae. Yeast 19: 1233–1241. doi:10.1002/yea.913
    OpenUrlCrossRefPubMed
  87. ↵
    1. Trivellin C,
    2. Olsson L,
    3. Rugbjerg P
    (2022) Quantification of microbial robustness in yeast. ACS Synth Biol 11: 1686–1691. doi:10.1021/acssynbio.1c00615
    OpenUrlCrossRef
  88. ↵
    1. van Dijk M,
    2. Erdei B,
    3. Galbe M,
    4. Nygård Y,
    5. Olsson L
    (2019) Strain-dependent variance in short-term adaptation effects of two xylose-fermenting strains of Saccharomyces cerevisiae. Bioresour Technol 292: 121922. doi:10.1016/j.biortech.2019.121922
    OpenUrlCrossRef
  89. ↵
    1. van Dijk M,
    2. Trollmann I,
    3. Saraiva MAF,
    4. Brandão RL,
    5. Olsson L,
    6. Nygård Y
    (2020) Small scale screening of yeast strains enables high-throughput evaluation of performance in lignocellulose hydrolysates. Bioresour Technol Rep 11: 100532. doi:10.1016/j.biteb.2020.100532
    OpenUrlCrossRef
  90. ↵
    1. Verduyn C
    (1991) Physiology of yeasts in relation to biomass yields. Antonie van Leeuwenhoek 60: 325–353. doi:10.1007/BF00430373
    OpenUrlCrossRefPubMed
  91. ↵
    1. Verduyn C,
    2. Postma E,
    3. Scheffers WA,
    4. Van Dijken JP
    (1992) Effect of benzoic acid on metabolic fluxes in yeasts: A continuous‐culture study on the regulation of respiration and alcoholic fermentation. Yeast 8: 501–517. doi:10.1002/yea.320080703
    OpenUrlCrossRefPubMed
  92. ↵
    1. Vos T,
    2. Hakkaart XDV,
    3. de Hulster EAF,
    4. van Maris AJA,
    5. Pronk JT,
    6. Daran-Lapujade P
    (2016) Maintenance-energy requirements and robustness of Saccharomyces cerevisiae at aerobic near-zero specific growth rates. Microb Cell Fact 15: 111. doi:10.1186/s12934-016-0501-z
    OpenUrlCrossRefPubMed
  93. ↵
    1. Wallace-Salinas V,
    2. Gorwa-Grauslund MF
    (2013) Adaptive evolution of an industrial strain of Saccharomyces cerevisiae for combined tolerance to inhibitors and temperature. Biotechnol Biofuels 6: 151. doi:10.1186/1754-6834-6-151
    OpenUrlCrossRefPubMed
  94. ↵
    1. Whitacre JM
    (2012) Biological robustness: Paradigms, mechanisms, and systems principles. Front Genet 3: 67. doi:10.3389/fgene.2012.00067
    OpenUrlCrossRefPubMed
  95. ↵
    1. Wickham H
    (2016) ggpolt2 Elegant Graphics for Data Analysis. Berlin: Springer: 16–260.
  96. ↵
    1. Wickham H,
    2. Hester J,
    3. Bryan J
    (2023) readr: Read Rectangular Text Data. https://readr.tidyverse.org
  97. ↵
    1. Wilke CO
    (2022) ggridges: Ridgeline Plots in ‘ggplot2’. https://wilkelab.org/ggridges/
  98. ↵
    1. Xiao Nan
    (2023) ggsci: Scientific Journal and Sci-Fi Themed Color Palettes for ‘ggplot2’. https://nanx.me/ggsci/
  99. ↵
    1. Yi X,
    2. Alper HS
    (2022) Considering strain variation and non-type strains for yeast metabolic engineering applications. Life 12: 510. doi:10.3390/life12040510
    OpenUrlCrossRef
  100. ↵
    1. Yutani H
    (2022) Highlight Lines and Points in ‘ggplot2’ [R Package gghighlight Version 0.4.0]. https://CRAN.R-project.org/package=gghighlight
Back to top
Download PDF
Email Article
Citation Tools
Performance and robustness trade-offs
Cecilia Trivellin, Peter Rugbjerg, Lisbeth Olsson
Life Science Alliance Oct 2023, 7 (1) e202302215; DOI: 10.26508/lsa.202302215
Twitter logo Facebook logo Mendeley logo
Issue Cover

In this Issue

Volume 7, No. 1
January 2024
Advertisement

Jump to section

Subjects

Related Articles

  • No related articles found.

Cited By...

  • No citing articles found.

More in this TOC Section

Similar Articles