Due to the vast surplus of crude glycerol accompanying the thriving biodiesel industry, glycerol has become an attractive substrate for biotechnological processes [1–3]. There are several reasons why glycerol can be an interesting alternative to sugars as a source of carbon and energy. One reason is that glycerol is more reduced than sugars, providing more reducing power for the production of small molecules with a high degree of reduction [4]. Another reason is that glycerol is metabolized in a fully respiratory manner in most microorganisms including the popular yeast Saccharomyces cerevisiae. This implies that the use of glycerol does not exert a Crabtree effect in S. cerevisiae, thereby circumventing the need to precisely control the sugar concentration in processes aiming at the production of biomass and growth-related molecules [5].

Saccharomyces cerevisiae is an attractive production organism in industrial biotechnology, and the palette of compounds currently produced by this organism by far exceeds the traditionally known products bioethanol and carbon dioxide. The preferred status of S. cerevisiae in industrial biotechnology is the result of the enormous body of knowledge available on its physiology and genetics, the intense experience with this organism in industrial bioprocesses, its robustness under process conditions, and the ease to genetically engineer the organism [6]. In particular, strains of the CEN.PK family have become a popular platform in both fundamental research and industrial applications as they offer a good compromise between physiological properties (e.g., the growth characteristics in shake-flask cultures) and genetic properties (e.g., transformation efficiency) [7]. CEN.PK strains have been applied in numerous metabolic and evolutionary engineering studies, such as for the production of lactate and pyruvate [8, 9], isoprenoids [10, 11], C₄-dicarboxylic acids [12], ornithine [13], n-butanol [14], and 3-hydroxypropionic acid [15], and for the fermentation of pentose sugars [16–18]. Although some of these processes could profit from utilizing glycerol as a carbon source due to its high reducing power, wild-type strains of the CEN.PK family cannot utilize glycerol at all in synthetic medium, particularly when complex medium supplements such as yeast extract or peptone are omitted [19, 20].

A previous study conducted in our laboratory showed that certain S. cerevisiae strains are able to grow in synthetic glycerol medium with maximum specific growth rates (µ_max) up to 0.15 h⁻¹ [20], confirming the previously reported high intraspecies diversity with regard to this phenotype [21]. The polygenetic basis of the glycerol growth phenotype in a haploid meiotic segregant (CBS 6412-13A) from the glycerol-utilizing isolate CBS 6412 in comparison to the non-growing, well-established laboratory strain CEN.PK113-1A was recently unraveled, and the knowledge was used for reverse engineering of the latter strain [22]. The identified CBS allele UBR2 _CBS showed the highest impact on glycerol growth performance in strain CEN.PK113-1A after allele replacement, resulting in a µ_max on glycerol of 0.04 h⁻¹. An additional positive effect was obtained by also replacing the GUT1 _CBS allele (µ_max of 0.08 h⁻¹) and the SSK1 _CBS allele (µ_max of 0.09 h⁻¹) in the strain CEN.PK113-1A UBR2 _CBS. The growth rate of the final reverse-engineered strain was 68% of the rate obtained for the wild-type CBS 6412-13A strain (µ_max of 0.13 h⁻¹) [22].

The results obtained by the above-described reverse engineering approach have been very promising and raised the question whether and how even higher maximum specific growth rates could be achieved in order to obtain strains attractive for real industrial applications. It is noteworthy in this context that several previous studies focusing on adaptive laboratory evolution (ALE) for improving growth on glycerol of S. cerevisiae strains have reported growth rates of evolved CEN.PK strains up to 0.22 h⁻¹. ALE employs evolutionary mechanisms for improving a specific industrially relevant phenotype of a chosen microbial strain. Cells are subjected to precisely defined conditions for prolonged periods of time, allowing selection for improved phenotypes that have arisen by natural mutations occurring throughout the course of cultivation [23]. Using repeated batch cultivation on glycerol, Merico et al. [19] achieved a growth rate of 0.17 h⁻¹ for the strain CEN.PK113-7D, while Ochoa-Estopier et al. [5] reported a growth rate of 0.20 h⁻¹ for the same original strain and a growth rate of 0.22 h⁻¹ for the industrial strain Ethanol Red. Although the evolved S. cerevisiae strains show even higher growth rates on glycerol than the above-mentioned reverse-engineered derivatives using the superior alleles of the CBS 6412-13A wild-type isolate [22], the mutations underlying the glycerol growth phenotype of the evolved strains have not been reported so far.

The goal of the current study was to find additional or superior genetic targets for reverse engineering of CEN.PK strains by scrutinizing the underlying genetic basis of the improved glycerol growth phenotype in evolved CEN.PK derivatives. The first part of the work focuses on generating CEN.PK113-1A derivatives with improved growth in synthetic glycerol medium via ALE and comparing their growth performance with that of the strain JL1, a derivative of the strain CEN.PK113-7D that was previously evolved for glycerol utilization by Ochoa-Estopier et al. [5]. The second part concerns the identification of the crucial mutations in the evolved strains and the reverse engineering of a CEN.PK wild-type strain.

The goal of this study was to identify the crucial mutations enabling CEN.PK derivatives to utilize glycerol after adaptive laboratory evolution. Ochoa-Estopier et al. [5] generated a respective strain (JL1) that showed a µ_max on glycerol of 0.20 h⁻¹ when analyzed directly after the ALE experiment, i.e., without any intermediate subcultivation in glucose-containing medium. The same strain only exerted a µ_max on glycerol of 0.16 h⁻¹ in the current study after it was subjected to five subcultivations in glucose-containing medium. This discrepancy could have resulted from the use of slightly different growth conditions and media and/or the possibility that non-genetic adaptations might have appeared during the evolution, which contributed to the higher growth rate directly measured after the ALE experiment. One example of a non-genetic adaptation can be due to an increased mitochondrial activity resulting from a fully respiratory mode of carbon catabolism during growth on glycerol [41]. Another example has been suggested by Brickner [42] that yeast cells have cellular and molecular mechanisms to mark and promote previously expressed genes, a phenomenon the author coined “adaptive transcriptional memory” [42]. Certain carbon utilization-related genes are found relocalized to the nuclear periphery upon the shift of carbon source and retained stably for several generations. Such peripheral localization allows for a more robust reactivation of the genes and confers epigenetic memory of the previous transcriptional state [42, 43]. In the current study, the fact that among 28 segregants tested after backcrossing of strain JL1, no single segregant could be identified exerting the same µ_max as the parental strain JL1 supports the assumption of non-genetic adaptations occurred in the evolved strain JL1. Moreover, the growth rate of the best performing segregant obtained from each backcross gradually decreased from generation to generation (Fig. 2), eventually leading to a µ_max of 0.128 h⁻¹ in the strain F3. This growth rate is comparable to the one of the reverse-engineered strain CEN.PK113-7D GUT1 _JL1 UBR2 _JL1 (0.130 h⁻¹). In spite of these indications, we cannot completely exclude that this result is due to several minor advantageous mutations concertedly contributing to the higher growth rate of strain JL1 that were gradually lost during the backcrossing.

In the course of this work, the two crucial mutations in strain JL1 could be allocated to the genes UBR2 and GUT1. It should be noted that one of the two crucial mutations could only be identified when analyzing the raw sequencing data against the sequence of the well-known S. cerevisiae strain S288C as a reference genome. In fact, only the point mutation within the UBR2 gene was identified when performing the complete analysis of the raw sequencing data using the published assembly of the parental CEN.PK113-7D strain [32] as a reference. As the presence of a second crucial mutation was predicted with high confidence by the backcrossing results, we searched for this mutation in a new data analysis using the S288C reference sequence with a higher sequencing depth than the one from CEN.PK113-7D. It turned out that the region where the GUT1 SNP is located was not sequenced in the CEN.PK113-7D assembly. It can be concluded that the quality of the reference sequence used was eventually more important for the quality of the analysis than the level of identity between the reference genome and the genome of the analyzed strain.

It was surprising to find that the same genes (UBR2 and GUT1) were mutated in the three strains obtained in the independent ALE experiments for growth on glycerol (i.e., in the evolved strains JL1, PW-1, and PW-2). Several previously published adaptive evolution approaches of S. cerevisiae (such as for acetic acid and n-butanol tolerance) have shown that strains evolved in parallel experiments can have mutations affecting different sets of genes [44, 45]. Therefore, the results of the current study strongly suggest that UBR2 and GUT1 are absolutely crucial in the establishment of glycerol utilization capability in CEN.PK strains. This is also supported by the above-mentioned fact that the genes UBR2 and GUT1 have also been identified as targets to establish glycerol utilization in the strain CEN.PK113-1A when using a completely different approach [20, 22]. In the latter studies, we sought to identify the genetic basis of the difference in glycerol utilization capability between the S. cerevisiae wild-type isolate CBS 6412-13A (µ_max of 0.13 h⁻¹) and the strain CEN.PK113-1A (no growth at all) by applying the method of pooled-segregant whole-genome resequencing [46]. The alleles of UBR2, GUT1, and SSK1 were identified to be the genetic determinants of the glycerol growth phenotype of strain CBS 6412-13A. Interestingly, no mutation in SSK1 was found in the genomes of the evolved strains (neither in the coding sequence nor in the promoter), and thus cannot be responsible for the above-mentioned difference in glycerol growth performance between strain JL1 and the reverse-engineered strain CEN.PK113-7D GUT1 _JL1 UBR2 _JL1.

Diversity, either natural or artificial, can both guide strain improvement via reverse engineering [47]. Both ALE experiments and the analysis of evolutionary distant strains by the pooled-segregant whole-genome resequencing approach have their pros and cons. Although ALE is only applicable when a selectable phenotype is in demand, two advantages of this method become obvious when comparing the current study with our previous genetic mapping approach [22]. Apart from being much less time-consuming, ALE has a higher potential to directly yield the causative mutations within crucial genes. To illustrate this, one has to note that 18 SNPs in the coding sequence and 10 SNPs within the respective promoter and terminator regions of UBR2 were identified when comparing the strains CBS 6412-13A and CEN.PK113-1A [22]. In contrast, only a single-point mutation per allele (including the coding sequence as well as the promoter and terminator regions) was found in all evolved strains. In general, we only expected about 1 to 10 genome-wide mutations in the strain JL1 compared to strain CEN.PK113-7D. The reason is that Ochoa-Estopier et al. [5] applied only about 50 generations in their ALE experiment and the natural mutation rate in S. cerevisiae strains ranges from 10⁻⁹ to 10⁻¹⁰ mutations per base pair per generation [48, 49].

In the study of Swinnen et al. [22], the respective UBR2 _CBS allele had the strongest impact on glycerol growth performance in strain CEN.PK113-1A, and the crucial SNP was a change of a premature stop codon into an amino acid encoding codon resulting in a longer Ubr2 protein. The same modification was found in the current study in all three independently evolved strains. It was therefore intriguing to check whether the sole presence of a full-length UBR2 gene sequence correlates with the ability of S. cerevisiae strains to grow on glycerol. However, alignment of the UBR2 _CEN allele against the corresponding alleles from other S. cerevisiae strains available on the SGD website (www.yeastgenome.org/) unveiled that the truncated UBR2 allele is a specific characteristic of the CEN.PK family, and a full-length allele is not the sole criterion for glycerol utilization ability. In fact, the genomes of the well-known strains S288C, W303, and Ethanol Red encode for a full-length Ubr2 protein, but these strains still cannot grow at all in synthetic glycerol medium [20]. Moreover, the replacement of the UBR2 allele in strains W303 and S288 by the UBR2 _CBS allele did not establish growth on glycerol (unpublished data).

As GUT1 encodes for glycerol kinase (GK), the first enzyme in the glycerol catabolic pathway of S. cerevisiae [50], it seems obvious that a higher in vivo activity of this enzyme could affect the efficiency of glycerol utilization. The respective mutation in the GUT1 coding sequence might have resulted in modified regulatory properties of the enzyme or a different V_max, e.g., by changes in posttranslational modifications or in a higher protein abundance caused by increased mRNA stability, translational efficiency, or protein stability. To test if a simple increase in Gut1 abundance is able to establish growth in synthetic glycerol medium, an additional copy of GUT1 (driven by the strong TEF1 promoter) was integrated in the genome of the wild-type strain CEN.PK113-7D. This, however, did not result in a detectable improvement of the growth on glycerol (data not shown).

Protein abundance can also be checked in vitro by measuring specific GK activities in crude cell extracts. As the strain CEN.PK113-7D with the wild-type GUT1 allele cannot grow at all in synthetic glycerol medium, we used two compromises in order to compare its specific GK activity with that of the evolved strain JL1, and the reverse-engineered strain CEN.PK113-1A GUT1 _JL1 UBR2 _JL1. In the first approach, the specific GK activities of glucose-pre-grown cells were recorded at 1 and 4 h after shifting them from synthetic glucose to glycerol medium. In the second approach, CSM was added to the synthetic glycerol medium allowing the wild-type CEN.PK113-7D strain to grow [20] and to analyze the specific GK activities of all three strains during exponential growth on glycerol. As shown in the Additional file 1: Figure S1, we could not detect any significant differences regarding the respective specific GK activities between the wild-type strain and the reverse-engineered strain regardless the experimental setup used. Hence, it can be concluded that the identified mutation in GUT1 does not seem to affect the abundance of the gene product in the cell, but rather might influence its regulatory properties causing an increased in vivo flux from glycerol to glycerol 3-phosphate. To experimentally verify this hypothesis, studies regarding enzyme kinetics have to be performed with purified wild-type and mutated Gut1 protein.

In contrast to the Gut1 protein, no direct involvement of Ubr2 in glycerol transport or catabolism is known so far. In general, Ubr2 is an E3 ubiquitin ligase, which usually acts together with a ubiquitin activating enzyme (E1) and a ubiquitin conjugation enzyme (E2) in the so-called ubiquitin proteasome system (UPS) which is a key mechanism for the spatiotemporal control of metabolic enzymes or disposal of misfolded and damaged proteins [51, 52]. In the UPS, the 76-amino acid ubiquitin is conjugated to lysine residues in target proteins through a series of reactions catalyzed by the above-mentioned E1–E2–E3 cascade of enzymes. The ubiquitin acts as a signal for the proteasome-mediated degradation of the respective target enzyme and this mechanism is particularly important for the acclimation of cells to changes in their environment. For a detailed review regarding the UPS system in yeast, the reader is referred to [51, 53]. It has been shown that Ubr2 has an overlapping function with its paralog, Ubr1, in the clearance of misfolded cytosolic proteins [54]. By far, the most prominent substrate identified for Ubr2 ligase is Rpn4, a major stress-induced transcription factor responsible for induction of proteasome subunits [55, 56]. The deletion of UBR2 leads to Rpn4 stabilization, resulting in an elevated abundance of the proteasome that is suggested to positively correlate with longevity of the cells [57, 58].

It might be tempting to conjure that the truncated Ubr2 protein found in CEN.PK strains is a loss-of-function allele based on the result obtained from this study. Nonetheless, the result of Swinnen et al. [22] showed that the deletion of Ubr2 from the strain CEN.PK113-1A (i.e., the allele encoding the truncated Ubr2) in a CEN.PK113-1A × CBS 6412-13A hybrid background still resulted in an elongated lag phase and a reduced µ_max as compared to the hybrid strain CEN.PK113-1A × CBS 6412-13A [22]. Therefore, the exact mechanism by which the truncation of this allele affects the glycerol utilization in CEN.PK strains remains to be elucidated. Another obvious question is whether Ubr2 could be directly involved in the protein turnover of the three proteins predominantly involved in glycerol catabolism, that is Stl1, Gut1, and Gut2 [50]. However, these proteins do not belong to the so far identified substrates for ubiquitination according to the S. cerevisiae Ubiquitination Database (http://scud.kaist.ac.kr/). Moreover, there are experimental findings which might indicate that Ubr2 rather affects the glycerol utilization pathway downstream of dihydroxyacetone phosphate (DHAP). In fact, the replacement of UBR2 also positively affected glycerol utilization in the CEN.PK 113-1A genetic background when the native glycerol catabolic pathway from glycerol to DHAP (‘glycerol 3-phosphate pathway’) was replaced by the ‘dihydroxyacetone phosphate’ pathway. Notably, the latter pathway is independent of Gut1 and Gut2 enzyme activities and glycerol is assumed to be predominantly taken up via a heterologous glycerol facilitator [59].

The current work confirmed the importance of UBR2 and GUT1 as targets for establishing glycerol utilization in strains of the CEN.PK family. In addition, it shows that a growth rate on glycerol of 0.130 h⁻¹ can be established in reverse-engineered CEN.PK strains by solely replacing two single amino acids within the coding sequence of Ubr2 and Gut1. Future work will focus on studying the molecular basis for the positive impact of the single-point mutations on the glycerol growth phenotype.