Adaptive laboratory evolution of Rhodosporidium toruloides to inhibitors derived from lignocellulosic biomass and genetic variations behind evolution

Using lignocellulosic biomass hydrolysate for the production of microbial lipids and carotenoids is still a challenge due to the poor tolerance of oleaginous yeasts to the inhibitors generated during biomass pretreatment. In this study, a strategy of adaptive laboratory evolution in hydrolysate-based medium was developed to improve the tolerance of Rhodosporidium toruloides to inhibitors present in biomass hydrolysate. The evolved strains presented better performance to grow in hydrolysate medium, with a significant reduction in their lag phases, and improved ability to accumulate lipids and produce carotenoids when compared to the wild-type starting strain. In the best cases, the lag phase was reduced by 72 h and resulted in lipid accumulation of 27.89 ± 0.80% (dry cell weight) and carotenoid production of 14.09 ± 0.12 mg/g (dry cell weight). Whole genome sequencing analysis indicated that the wild-type strain naturally contained tolerance-related genes, which provided a background that allowed the strain to evolve in biomass-derived inhibitors.


Introduction
In the last years, oleaginous yeasts have attracted increased attention due to their ability to produce lipids and carotenoids from different carbon sources. In addition, the accumulation of lipids in oleaginous yeasts can reach 70% of their dry cell weight under nutrient starvation conditions (Probst et al., 2016). Lipids produced by oleaginous yeasts mainly include palmitic acid (C16:0), oleic acid (C18:1) and linoleic acid (C18:2), which have a composition similar to cocoa butter and palm oil, being good alternatives to replace plant oils to produce fatty acid esters based chemicals such as soaps, cleansers and personal care products (Vasconcelos et al., 2019;Liu et al., 2020). In addition, fatty acids produced by oleaginous yeasts can be used for biodiesel production, since the cetane number and clod filter plugging points of methyl esters of those fatty acids, which are important parameters to evaluate the quality of biodiesel, are in the range of EU standards (Bonturi et al., 2015;Ramos et al., 2009).
Besides lipids, some oleaginous yeasts can also produce carotenoids, among of which, β-carotene, γ-carotene, torulene and torularhodin are the most common (Chandi and Gill, 2011). Carotenoids present antioxidant and anti-tumor activities, and help reduce the risk of many diseases (Novoveská et al., 2019). Therefore, these compounds have been widely utilized in food, feed, pharmaceutical and cosmetic industries, and their market is estimated to reach $2.00 billion by 2026 (Markets and Markets, 2020). At present, edible plants are the main sources of carotenoids. Using oleaginous yeasts to replace edible plants to produce lipids and carotenoids can help reduce the conflict with food production. Then, seeking for cheap and renewable carbon sources for cultivation of oleaginous yeast is a key strategy to reach this sustainable goal.
Previous studies have demonstrated that agricultural and agroindustrial residues, wastewaters, and crude glycerol can be used as carbon sources for oleaginous yeast cultivation (Arous et al., 2016;Kitcha and Cheirsilp, 2013;Liu et al., 2020). Among them, lignocellulosic biomass is highly attractive because it is non-edible, has a low cost and is abundant around the world. To be used in fermentation processes, a pretreatment step is required to recover sugars from biomass structure. However, besides sugars, other compounds, which are toxic to microorganisms, are also released from the material's structure or formed during their partial degradation. Such toxic/inhibitory compounds are divided into 3 main groups: furan aldehydes, aliphatic acids, and phenolic compounds. Furan aldehydes, such as furfural and hydroxymethylfurfural (HMF), result from the degradation of pentose and hexose sugars, respectively. The further degradation of HMF lead to the formation of formic acid and levulinic acids, which are aliphatic acids. Phenolic compounds including vanillic acid, syringic acid, 4-hydroxybenzaldehyde, p-coumaric acid, vanillin and syringaldehyde, among others, are obtained from lignin degradation (Mussatto and Roberto, 2004). The presence of these inhibitory compounds in biomass hydrolysate can significantly influence the growth and product accumulation in oleaginous yeast (Liu et al., 2020).
Although different methods have been proposed to reduce the concentration of inhibitors in biomass hydrolysate in order to improve the microbial performance during fermentation, detoxification methods result in additional costs to the overall process (Mussatto and Dragone, 2016). Other strategies that can be used to improve the tolerance of oleaginous yeasts to inhibitors derived from lignocellulosic biomass include adaptive laboratory evolution (ALE) and genetic engineering. When compared with random engineering, ALE is easier to perform and mutant strains with desired characters can be obtained by cultivating generations of strains under selective pressure conditions. In the sequence, whole genome sequencing analysis can help to unravel the mutation mechanism and provide knowledge for genetic engineering.
So far, the use of the ALE technique in hydrolysate-based medium has been little explored to improve the performance of oil producing microorganisms. In the present study, an ALE strategy was designed to improve the tolerance of the oleaginous yeast Rhodosporidium toruloides NRRL Y-1091 toward inhibitors derived from biomass pretreatment. The growth profile, lipid and carotenoid production of wild-type and evolved strains were then compared. Finally, a whole genome sequencing analysis of wild-type and evolved strains was performed to explain the mutation mechanism and genetic variations behind evolution.

Microorganism and cultivation conditions
The wild-type oleaginous yeast Rhodosporidium toruloides NRRL Y-1091 was used in this study. The strain was obtained from the Agricultural Research Culture Collection, USA and stored at − 80 • C in 20% glycerol. In the pre-cultivation step, one loop of yeast cells was transferred to 100 mL YPD medium in 250 mL flasks and cultivated at 30 • C, 250 rpm for 72 h. To obtain single colonies, the YPD broth was diluted 20.000-folds with autoclaved distilled water. After that, 100 µL of diluted YPD broth was spread on YPD agar plates and incubated at 30 • C for 5 days.

Preparation of wheat straw hydrolysate
Wheat straw was provided by the Danish Technological Institute (Denmark). The material was milled to a size of about 180-1800 mm and then submitted to hydrothermal pretreatment and enzymatic hydrolysis in order to produce a sugar-rich hydrolysate, which was also decolorized to be used in the ALE experiments. The conditions used to produce the hydrolysate were previously described (Liu et al., 2020). Table 1 summarizes the composition of the hydrolysates (decolorized and non-decolorized) used for the ALE experiments.
The concentrations of glucose, xylose and acetic acid in the hydrolysates were quantified using a HPLC system (Dionex Ultimate 3000, Germany) equipped with an Aminex HPX-87H column (300 × 7.8 mm, Bio-Rad, USA) eluted with 0.005 M H 2 SO 4 at 50 • C for 50 min with a flow rate of 0.6 mL/min. The concentrations of furfural, HMF, levulinic acid, vanillic acid, syringic acid, 4-hydroxybenzaldehyde, p-coumaric acid, vanillin and syringaldehyde were measured by using the same HPLC system, but equipped with a UV detector at 280 nm and a Zorbax eclipse plus C18 column (Agilent, USA) eluted with mobile phases composed by A: 0.05% (v/v) acetic acid in water and B: acetonitrile at a flow rate of 1 mL/min. The gradient elution started with 5% of B and linearly increased to 60% over 5 min and further increased to 90% of B over 0.5 min. This gradient was kept at this ratio for 2 min and returned to 5% of B until 10 min. Total phenolic compounds were determined by colorimetric method (Ballesteros et al., 2014). All the standards used for analysis were purchased from Sigma-Aldrich.

Adaptive laboratory evolution (ALE)
The ALE experiment was performed in an automated liquid handling robot platform in which the strains were incubated in 15-mL plastic tubes with caps in a heat block kept at 30 • C and aerated by magnetic stirring at 1200 rpm. The growth of the strains was monitored by absorbance reading at 600 nm using a Teach Sunrise microplate reader (Tecan, Männedorf, Switzerland) connected to the automated platform. Continuous cultivations of 10 replicates of the wild-type strain were performed in medium containing increasing concentration of wheat straw hydrolysate. To obtain different concentration of inhibitors for the experiments, the hydrolysate was diluted with synthetic medium composed by the same amount of sugars, supplemented with 1.79 g/L yeast nitrogen base without amino acid and ammonium sulfate (Sigma-Aldrich) and 1 g/L NH 4 Cl, pH 4.9.
The driving force for the ALE experiments was based on increasing the concentration of inhibitor gradually and making the mutant strains enrich in population. The whole evolution process was divided in two rounds. In the first round, the decolorized hydrolysate was used and a screening of the wild-type strain in 25%, 50%, 75% and 100% hydrolysate was performed to decide the starting point of the evolution. When the growth rate of the strains in certain level of the decolorized hydrolysate increased compared to the growth rate observed in the screening experiment, then 10% of the culture was transferred to a higher concentration of decolorized hydrolysate. This process was repeated until all replicates of the strains were able to grow well in 100% decolorized hydrolysate. Then, the ALE strategy was repeated using non-decolorized hydrolysate until the strains could grow well in 100% hydrolysate.

Comparison of wild-type and evolved strains
The evolved and wild-type strains were cultivated in the synthetic medium described above in shaking flasks at 30 • C, 250 rpm for 72 h. Afterward, the cells were collected by centrifugation (5000 rpm, 5 min), washed with sterilized distilled water and suspended in non-decolorized wheat straw hydrolysate supplemented with 1 g/L NH 4 Cl to obtain a cell concentration of 1 g/L. The cell mass was estimated using a calibration curve between absorbance of cells in distilled water at 600 nm and dry weight of cells.
The assays for comparison of the wild-type and evolved strains were then performed in 24 deep well plates with 3 mL working volume in each well. Every 24 h, 2 well samples per strain were taken to measure the cell mass and • Brix (total soluble solids in aqueous solution). Lipid and carotenoid contents were determined at the end of the cultivation (120/144 h).

Analysis of lipids and carotenoids produced by wild-type and evolved strains
For the cell disruption step, cells from 2 mL fermentation broth were collected by centrifugation (5000 rpm, 5 min), washed twice with distilled water, then suspended in 1 mL distilled water and transferred to a lysing matrix E tube (6914-500, MP Biomedicals, France). The lysing tubes were beaten on a bead beater (Precellys 24, Bertin Technologies, France) for 3 cycles of 60 s at 5000 rpm, with 60 s break after each cycle. After that, all cell debris, beads and liquid in the beating tubes were transferred into new glass tubes with screw caps.
For total lipid quantification, the disrupted cells were extracted 3 times with a solution of chloroform:methanol (1:1.25). After centrifugation (3500 rpm, 5 min), the bottom phase was withdrawn and dried under N 2 atmosphere for total lipid analysis. Before gas chromatograph analysis, the fatty acids were methylated. The methylation and analysis were conducted according to the method described by Breuer et al. (2013) using a gas chromatograph system (Agilent 7890, USA) equipped with a Supelco Nucol TM 25357 (30 m × 530 μm × 1.0 μm) column.
For carotenoid quantification, 2 mL of chloroform were added to each tube containing the disrupted cells and the mixture was vortexed for 10 min. Then the tubes were centrifuged at 3500 rpm for 5 min and the bottom phase was taken into another clean glass tube. The extraction was repeated 3 times and the combined chloroform fractions were dried under N 2 flow. The extracted carotenoids were redissolved in 2 mL acetone containing 0.2% (w/v) BHT (butylated hydroxytoluene) and inserted in a brown HPLC glass vial for further analysis. A Shimadzu Nexera UHPLC system equipped with a D2&W 190-800 nm PDA detector and a C30 reversed-phase carotenoid column (3 µm, 250 × 4.6 mm -YMC, Japan) was used for carotenoid quantification. The HPLC conditions for carotenoid detection were adapted from the method described by Søltoft et al. (2011). The carotenoid standards β-carotene and γ-carotene (Sigma-Aldrich) were detected at their maximum absorbance wavelength (450 and 461 nm, respectively). Torulene was quantified using γ-carotene as standard (Liu et al., 2021).

Whole genome sequencing
To identify the common mutations and important genes responsible for the mutations, the whole genome of the 7 best evolved strains (identified as CH1, CH3, CH4, CH5, CH6, CH8, and CH10) were sequenced, and then compared with the genome sequence of the wildtype starting stain. Single colonies of all these 8 strains were picked up from YPD agar plates and cultivated in synthetic medium at 30 • C, 250 rpm for 72 h. The DNA of the strains was extracted using a Quick-DNA™ fungal/bacterial miniprep kit (Zymo Research, USA). To prepare the reference genome, the DNA of the wild-type strain was sequenced by the PacBio RS platform (10-kb SMRT library) and Illumina HiSeq 2000 platform at 2 × 150 paired-end reads. The genome of the evolved strains was sequenced by the same Illumina platform at 2 × 150 paired-end reads.

De novo assembly and function annotation
After removing low quality reads, the clean data was corrected using the softwares Pbdagcon and Falcon Consensus. The corrected reads were assembled using softwares such as Celera Assembler, Falcon and SMRT Analysis. Then, the best assembly result was chosen and the single reads of it was corrected with Illumina Hiseq data using GATK software. The gap between contig and scaffold was filled with long inset size pairedend reads using SSPACE Basic. To determine the location of the genes, the homology prediction was conducted with Gene Wise. Then, de novo prediction tools including SNAP, Augustus, Gene Markers were used to predict the gene models. Non-coding RNA was identified by RNAmmer and tRNAscan software. The transposon sequence was found by aligning the assembly with the known transposon sequence database (Repeat-Masker, RepeatProteinMasker, buildXDFDatabase). Gene function annotation was completed by blasting genes with different databases including P450, VFDB, ARDB, TF, CAZY, KINASE, SWISSPROT, NOG, COG, CARD, CWDE, NR, KEYWORD, DBCAN, TCDB, T3SS, TREMBL, IPR, PHI, KEGG, GO, PHOSPHATASE, KOG databases.

Mutation detection
After removing low quality reads, the clean data of the evolved strains was mapped to the reference using the BWA software. The Picard tool was used to filter repeat reads. The single nucleotide polymorphisms (SNPs) and insertion and deletion of bases into genome (InDels) of evolved strains were detected by the GATK software and compared with the reference strain.

Statistical analysis
Significant differences among samples were verified by analysis of variance at a significance level of 0.05, using the software OriginPro 2018b (OriginLab, USA).

ALE strategy and fitness changes of strains during evolution
Biomass hydrolysates usually contain a mixture of different inhibitory compounds and their synergistic effect may strongly affect the microbial performance during fermentation (Mussatto and Roberto, 2004). Therefore, the ALE strategy used in the present study consisted in using biomass hydrolysate as culture medium for evolution in order to improve the ability of the strain to grow in the presence of different compounds present simultaneously in the medium. To improve the tolerance of R. toruloides to the inhibitory compounds, two rounds of ALE were performed: a first-round, where the yeast was evolved in decolorized hydrolysate, and a second-round, where the previously evolved yeast was now evolved in non-decolorized (raw) hydrolysate. It is worth highlighting that, although the decolorization step partially removed some inhibitors from the hydrolysate, the decolorization removed mainly compounds contributing to the color of the medium, and a significant amount of inhibitors was still present in the hydrolysate after this step (Table 1).
To boost the tolerance of the strain to the toxic compounds, the level of toxicity to be applied must be in an appropriate range since a very low concentration of inhibitors may not be enough to promote selective pressure for the strain to evolve, while a very high concentration of inhibitors may cease the growth of the strain. Therefore, before the ALE experiments, a screening step was performed by cultivating the yeast in different concentrations of decolorized hydrolysate. It was observed in these experiments that the growth rate of the yeast cultivated in 25% decolorized hydrolysate was comparable to that observed in synthetic medium (0.27 h − 1 ). So, 25% of decolorized hydrolysate was used as starting concentration for ALE. The concentration of the hydrolysate was increased when the growth rate of the strains reached approx. 0.15 h − 1 . After 21 days of ALE, all replicates were able to grow in 100% decolorized hydrolysate (Table 2) and their growth rates were similar to that observed in synthetic medium.
Before starting the second-round of ALE in non-detoxified hydrolysate, another growth screening was conducted, and it was observed that, except CH1, CH2, CH3 and CH4, the other replicates were already able to grow in 80% non-decolorized hydrolysate. Therefore, the starting concentration of hydrolysate for the second-round ALE for these 6 populations was 80%. For the other 4 populations, the starting concentration of hydrolysate was adjusted to 30%, 70%, 10% and 70% for CH1, CH2, CH3 and CH4, respectively. After 41 days of ALE, most of the strains were able to grow in 100% non-decolorized hydrolysate ( Table 2). Confirmation tests revealed that, except the strains CH1 and CH3, all the other evolved strains were able to grow in 100% hydrolysate with fitness similar to that observed in synthetic medium.
These results allow concluding that the ALE strategy was successfully applied to improve the tolerance of the oleaginous yeast to biomass inhibitors. In future studies, analysis of the cell viability measurement by colony forming units (CFU) could also be incorporated to further refine the ALE monitoring, giving indications on the number of living cells during the evolution process. This can be especially relevant in media where color can affect the OD measurement, which was not the case of the present study. Fig. 1 shows the growth profile of the wild-type starting strain and all the 10 evolved strains cultivated in non-decolorized hydrolysate. As can be seen, all the evolved strains (except CH3) presented shorter lag phase when compared to the wild-type strain (Fig. 1A). Among them, CH6 and CH10 started to grow after 24 h only, while CH2, CH4, CH5, CH7 and CH8 started to grow after 48 h. The lag phase for CH9 lasted 72 h, while the CH3 and the wild-type strains started to grow after 96 h. However, it was found that sugars in the hydrolysate were not totally consumed for any of the strains during the 120 h of cultivation (data not shown), and the cell mass accumulated was also relatively low for all the strains. The low sugars consumption and cell mass accumulation could be an indication that the nitrogen source in wheat straw hydrolysate was not high enough to favor the yeast propagation. Therefore, in a second step, the strains were cultivated in hydrolysate supplemented with 1 g/L NH 4 Cl. After nitrogen supplementation (Fig. 1B), sugars consumption and cell mass accumulation of wild-type and evolved strains were improved. However, similar as before, the wild-type strain entered in the exponential phase after 96 h, while the evolved strains had a shorter lag phase. When observing the strains, the performance of CH6 was relatively consistent in both media (supplemented or not with nitrogen) with a lag phase of only 24 h and more cell mass accumulated than the other strains in both cases. Such result indicates that the hydrolysate supplementation with nitrogen had a more pronounced effect on the cell mass accumulation than in the lag phase.

Comparison of lipid and carotenoid accumulation for the wild-type and evolved strains
After the evolution, the ability of the wild-type and evolved strains to accumulate lipids and carotenoids was also evaluated and compared. Quantifications were performed after 144 h of fermentation using nondecolorized hydrolysate supplemented with nitrogen (1 g/L NH 4 Cl). As can be seen in Fig. 2, significant differences (p < 0.05) in terms of product accumulation were observed for the wild-type and evolved strains. Except CH2, all the evolved strains were able to accumulate more lipids than the parental strain. CH4 was the best lipid producer, accumulating 27.89 ± 0.80% of the cell weight in lipids. To better understand the effect of the evolution on lipid production by the yeast, the composition of fatty acids produced by the wild-type and evolved strains was also determined and compared. As shown in Table 3, palmitic acid (C16:0) and oleic acid (C18:1) were the main fatty acids produced by all the strains. This result reveals that although the ability of the strain to accumulate lipids was improved after ALE, the evolution did not change the lipid synthesis pathway of the yeast.
Carotenoids were also accumulated in significantly higher amount (p < 0.05) in the evolved strains than in the wild-type ( Fig. 2A). The best carotenoid producers were CH2 and CH7, which accumulated 13.89 ± 0.09 mg/g and 14.09 ± 0.12 mg/g of cell weight in carotenoids, respectively (the carotenoid accumulation by these two strains was Z. Liu et al. similar at p < 0.05). All the strains produced β-carotene, γ-carotene and torulene; with γ-carotene and torulene being the carotenoids produced in higher amount (Table 4). So, similarly to what was observed for lipids, the ability of the strain to accumulate carotenoids was improved after ALE, but the evolution did not change the profile of the carotenoids produced by the yeast. The difference observed on the production of carotenoids by the evolved strains was also reflected on the color of the cells at the end of the cultivation, which was darker for the evolved strains.
In oleaginous yeasts, carotenoids are located in the fat rich region in yeast cells, and play a protective role of resistance to light and oxidation damages from the cultivation environment (Avalos and Carmen Limón, 2015). To improve the production of carotenoids, radiation or oxidation   . 2. Lipid and carotenoid accumulation by the wild-type and evolved strains of R. toruloides. Carotenoid accumulation in mg/g cell and lipid accumulation in % (w/w). Bars marked with two stars have a significant difference (p < 0.05) compared with the value of wild-type strain in each plot.

Table 3
Fatty acids produced by the wild-type (WT) and evolved (CH) strains of R. toruloides cultivated in non-decolorized hydrolysate supplemented with nitrogen source.

Strains
Fatty acids (mg/g of cells) C14:0 C16: conditions can be used during the ALE process, for example (Reyes et al., 2014). In the present study, ALE of R. toruloides was performed in wheat straw hydrolysate (which contains a mixture of inhibitory compounds) and also resulted in an improved production of carotenoids. Such improvement could be the response of the yeast to oxidation stress. Phenolic compounds present in the hydrolysates can enhance the generation of reactive oxygen species including H 2 O 2 , O 2 − and OH − , which cause enzyme denaturation, cytoskeleton damages and cell death Ibraheem and Ndimba, 2013). Furfural, which is another inhibitor usually found in hydrolysates, may also cause oxidation damages to yeast (Allen et al., 2010).

Genome sequencing and annotation for the wild-type strain
The combination of Illumina HiSeq and PacBio RS sequencing and de novo assembly of the wild-type R. toruloides yielded a 21.5 Mb genome with an average 101 × coverage. This genome was composed of 263 contigs and 68 scaffolds, in which 8387 coding genes and 236 ncRNA were distributed. Regarding to the annotation results, 98.85% of genes could be mapped to definitive functions via blasting against CAZy, Kinase, Swissprot, NOG, KEGG, KOG, Tremble, GO, T3SS and others total 23 databases.
The sequencing analysis revealed that the wild-type strain naturally contains tolerance-related genes, which provided a background that allowed the strain to evolve in biomass-derived inhibitors. From the annotation result, 25 alcohol dehydrogenases encoding genes, 18 aldehyde reductases encoding genes, and 4 alcohol oxidases were found in this yeast, which is in agreement with the study of Wang et al. (2016). Other functional genes, such as stress response genes and other tolerance related genes were also identified.

Mutations detected in evolved strains
In this step of the study, the genome of the 7 best performing evolved strains were aligned to the reference genome of the wild-type starting strain with the aim of identifying the mutations resulting from the ALE process. Mutation types and numbers in each evolved strain were summarized in Fig. 3. Missense nonsynonymous SNPs was the mutation that most often occurred, being observed around 2500 times in each evolved strain. This mutation can change the amino acid sequence of protein. Similarly, insertions and deletions occurred in coding regions, resulting in frame shifted, have significant effects on the function of genes.
To identify the most important mutations related to inhibitor tolerance, common SNPs and InDels and their functions were displayed in Fig. 4. It was then found that the common mutations occurred in the 7 evolved strains included 48 missense SNPs and 15 InDels resulting in frame shifted. It is interesting to note that among those 48 missense SNPs, 25 occurred at different positions of a gene, which is aligned to the one encoding reverse transcriptase (RHTO_02155) in R. toruloides NP11 (Zhu et al., 2012). However, there are no previous studies explaining the role of this gene in tolerance; so, the function of this mutant gene should be further identified. Another common missense mutation happened in a gene predicted as RHTO_01747 (coding for MFS transporter, quinate permease). Major facilitator superfamily (MFS) transporters are responsible for the bonding and transport of essential nutrients, ions and other compounds in a wide range of microorganisms (Rédei, 2008). A plasma membrane transporter belonging to the MFS family in S. cerevisiae was reported to show resistance to short-chain monocarboxylic acid and quinidine (Tenreiro et al., 2002). Deletion of genes encoding for MFS proteins was reported to promote a 3-fold reduction of coniferyl aldehyde in yeast (Sundström et al., 2010). The mutation of this gene may change the structure of membrane proteins, resulting in an improved tolerance of yeast to the inhibitors.
Among the 15 common InDels causing frame shifted, Δ1bp deletion was observed in RHTO_05744, a RIM9 protein encoding gene. RIM9 protein has been reported as relevant to a variety of responses to stress and to the yeast-hyphal transition (Yan et al., 2012). Another common Δ1bp deletion was found in RHTO_04597 (coding for ATP-binding cassette transporter, ABC), which contains a large family of proteins with a diversity of physiological functions (Bouige et al., 2005). Fungal ABC transporters have been reported to confer multiple drug resistance, stress response, cellular detoxification and resistance to weak organic acid (Wolfger et al., 2001;Prasad and Goffeau, 2012;Piper et al., 1998). In addition, ABC transporters play an important role to export toxic Table 4 Carotenoids produced by the wild-type (WT) and evolved (CH) strains of R. toruloides cultivated in non-decolorized hydrolysate supplemented with nitrogen source.

Strains
Carotenoids (mg/g of cells)  Z. Liu et al. compounds and metabolites from intracellular environment, which is powered by TCA processing (Liu, 2011). This could explain why the wild-type strain of R. toruloides consumed all the carbon source during the cultivation but produced less lipids and carotenoids than the evolved strains, as it probably used more energy to export inhibitors from the intracellular environment. On the other hand, due to the improved resistance to the inhibitors, evolved strains had more energy available for lipid and carotenoid production. Remarkably, two Δ1bp insertions were also found in different positions of an oxidoreductase encoding gene domain (IPR000572) belonging to GO:0016491 in these 7 mutant strains. In this domain, xanthine dehydrogenase, aldehyde oxidase, nitrate reductase and sulphite oxidase can be coded (Wootton et al., 1991). An evolved strain of S. cerevisiae was found to have 1.92-fold higher vanillin reduction ability than the wild-type strain due to up-regulated genes enriched in oxidoreductase activity (GO:0016491) (Shen et al., 2014). Those frame shifted InDels may cause changes/loss of gene functions; however, it was reported that disrupted multiple genes in S. cerevisiae could result in higher tolerance and shorter lag phase in media containing lactic acid (Suzuki et al., 2013). Although the mutations found in intergenic regions do not change the amino acid sequence, it may still affect the gene splicing, transcription factor binding, mRNA degradation and gene expression. In the present study, 149 common intergenic SNPs and 44 common intergenic InDel were found in the 7 evolved strains of R. toruloides, which largely contributed to the genome variations of the evolved strains. An investigation on stress-tolerance of S. cerevisiae Fig. 4. Common mutations detected in the 7 evolved strains of R. toruloides. Z. Liu et al. demonstrated that intergenic variants have significant impact on expression levels of stress tolerant genes, resulting in stress resistant phenotypes of strains (Gan et al., 2018). To better understand the impacts of intergenic mutations in R. toruloides, further studies should be done to link the function of intergenic regions, relevant gene expression, and phenotype of strains. Finally, to obtain strains with better potential for industrial application, concentrated hydrolysates (containing higher concentration of sugars) should also be considered in subsequent ALE studies.

Conclusion
Adaptive laboratory evolution in hydrolysate-based medium was an effective strategy to obtain strains with improved ability to grow in medium containing biomass-derived inhibitory compounds. In fact, the adaptation strategy resulted in evolved strains of R. toruloides able to grow faster in hydrolysate containing inhibitory compounds, and also with better performance to accumulate lipids and carotenoids. Whole genome sequencing analysis allowed identifying common mutations in the evolved strains, giving useful indications on the mechanisms used by the strain to evolve. These findings are also relevant for the development of new strains with improved ability to convert biomass hydrolysates into valuable compounds.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.