Microbial lipid production from crude glycerol and hemicellulosic hydrolysate with oleaginous yeasts

Background Crude glycerol (CG) and hemicellulose hydrolysate (HH) are low—value side-products of biodiesel transesterification and pulp—and paper industry or lignocellulosic ethanol production, respectively, which can be converted to microbial lipids by oleaginous yeasts. This study aimed to test the ability of oleaginous yeasts to utilise CG and HH and mixtures of them as carbon source. Results Eleven out of 27 tested strains of oleaginous yeast species were able to grow in plate tests on CG as sole carbon source. Among them, only one ascomycetous strain, belonging to Lipomyces starkeyi, was identified, the other 10 strains were Rhodotorula spec. When yeasts were cultivated in mixed CG/ HH medium, we observed an activation of glycerol conversion in the Rhodotorula strains, but not in L. starkeyi. Two strains—Rhodotorula toruloides CBS 14 and Rhodotorula glutinis CBS 3044 were further tested in controlled fermentations in bioreactors in different mixtures of CG and HH. The highest measured average biomass and lipid concentration were achieved with R. toruloides in 10% HH medium mixed with 55 g/L CG—19.4 g/L and 10.6 g/L, respectively, with a lipid yield of 0.25 g lipids per consumed g of carbon source. Fatty acid composition was similar to other R. toruloides strains and comparable to that of vegetable oils. Conclusions There were big strain differences in the ability to convert CG to lipids, as only few of the tested strains were able to grow. Lipid production rates and yields showed that mixing GC and HH have a stimulating effect on lipid accumulation in R. toruloides and R. glutinis resulting in shortened fermentation time to reach maximum lipid concentration, which provides a new perspective on converting these low-value compounds to microbial lipids. Supplementary Information The online version contains supplementary material available at 10.1186/s13068-021-01916-y.


Background
There is increased need for recycling of waste products from food, wood industry and agriculture in recent years. Vast quantities of organic residues are a challenge that is being addressed by converting waste and low purity sugars into higher quality products [1].
Crude glycerol (CG) is a side product of biodiesel production, which is released during transesterification of vegetable oils. CG is a highly problematic side product; it contains methanol, soap and ash. Industrial application of CG requires extensive purification, making its application quite expensive [2][3][4][5]. In recent years there have been developments to utilse glycerol and convert it to useable biomass and for example turn it into microbial fuel cells [6], or to convert glycerol into valuable fatty acids using Yarrowia lipolytica [7]. It was also tested as an additive with other waste substrates to make microbial oil and other valuable products [8][9][10][11][12][13].
Hemicellulose is, besides of cellulose and lignin, one of the major polymers of the plant cell wall. It is a heteropolysaccharide, which composition differs between different groups of plants. Hardwood hemicellulose consists mainly of xylans, while softwood is mainly built of glucomannans. Hemicellulose of lignocellulosic materials of monocoyledons like wheat straw consist of a xylan backbone with side chains of arabinose and glucuronic acid [14,15]. In pulp and paper industry, hemicellulose is a side product and used for steam generation. However, it is rather inefficient for this application, due to its low heating value (13.5 MJ/ kg) [5]. In lignocellulose-based bioethanol production, separating hemicellulose from the cellulose fraction can be advantageous, since conversion of the pentoses, which are the major sugar monomers of hemicellulose to ethanol is still a challenge [7,8]. When lignocellulose is thermochemically pretreated by acid hydrolysis, hemicellulose is solubilised and hydrolysed, resulting in the release of monosaccharides [16]. Since the major parts of cellulose and lignin are still in the solid phase, it is easy to separate the hemicellulose hydrolysate from the other components of lignocellulose hydrolysate. Hemicellulose hydrolysates contain, in addition to released sugars, a variety of inhibitors such as furaldehydes, weak organic acids and phenolic compounds [17], and frequently it is necessary to remove those inhibitors or to dilute hemicellulose hydrolysate to enable bioconversion of hemicellulose hydrolysates [13,18,19].
Many oleaginous yeasts, i.e., yeasts that can accumulate more than 20% of their biomass as lipids, can convert both glycerol as well as sugars and organic acids derived from hemicellulose to oil. This oil has a similar composition as some vegetable oils [20]. Production of vegetable oil can have considerable greenhouse gas potential [21] and thus, replacement of vegetable oil by yeast oil may result in more sustainable biodiesel [22,23] or animal feed production [24]. However, production costs of microbial oil are still too high to be competitive with oil from fossil resources or vegetable oils [25,26]. Therefore, it is important to explore the conversion of low-value side products such as lignocellulose hydrolysate, hemicellulose hydrolysate, or crude glycerol to microbial oils. Another approach to improve cost efficiency of microbial lipid production is to identify strains with high productivity when converting the mentioned residual products to lipids and to understand their physiology, to identify targets for strain optimisation [27,28].
Converting a mixture of CG and HH to yeast oil may provide a possibility of simultaneously converting these side streams to a high-value products. At the same time, inhibitors present in only one of these side streams might be diluted and thus less inhibitory. It was our aim to get a survey about the diversity of strains when converting CG and HH to lipids. We also intended to investigate effects of mixing these two problematic substrates. We tested several of strains belonging to the genera Lipomyces (ascomyceteous oleaginous yeasts) and Rhodotorula (basidiomycetous oleaginous yeasts) for their ability to grow and synthetise lipids on a mixture of CG and HH.

Crude glycerol and hemicellulose growth tests
The ability of 27 yeast strains to utilise CG as sole carbon source and their resistance against inhibitors in CG was tested on solid media. A droplet of undiluted crude glycerol or a filter soaked with CG was placed at the centre of YNB plates inoculated with the tested strains. Out of the 27 tested strains, 11 strains showed visible growth close to the droplet/filter and thus were able to tolerate CG, and to utilise it as carbon source (experiments were performed in duplicates). Interestingly, only one ascomycetous strain (L. starkeyi 7786) showed substantial growth on the plates with crude glycerol. These strains were also tested in liquid shake flask cultures with 40% hemicellulose hydrolysate as sole carbon source, and in cultures with different concentrations of crude glycerol (up to 120 g/l) as sole carbon source. After 96 h of cultivation almost all red yeast strains grew well in hemicellulose hydrolysate (OD 20-40) and in 120 g/L CG (OD > 40). R. minuta CBS 8013 grew slower than other Rhodotorula strains in CG concentrations higher than 60 g/L. L. starkeyi CBS 7786 grew comparatively well in hemicellulose hydrolysate and 30 g/L CG medium, but had lower growth compared to the tested red yeast strains in higher concentrations of CG.

Testing of growth in a mixture of crude glycerol and hemicellulose
To evaluate the effect of mixing CG and HH in the cultivation media we tested five strains, L. starkeyi CBS 7786, R. toruloides CBS 14, R. glutinis CBS 3044, R. glutinis CBS 2889 and R. mucolaginosa CFSQE 63, for their growth on these mixed substrates and compared it to cultivations with either glucose (positive control) or crude glycerol as sole carbon source (Figs. 1-4). The strains were chosen among those that showed good growth on crude glycerol and hemicellulose hydrolysate media. There were varying responses of the strains to the different carbon sources. In the glucose control medium (Control media), R. toruloides CBS 14 had the lowest OD 600 after 96 h of cultivation (21), while R. mucolaginosa CFSQE 63 reached an OD 600 more than twice as high as R. toruloides after 96 h. In mixed media (sHH40CG60 media), results were almost the opposite. R. toruloides had the second highest OD 600 after R. glutinis CBS 3044. Out of the red yeasts, R. mucolaginosa CFSQE 63 had the lowest OD 600 . In media with crude glycerol (sCG60 media) only, R. glutinis CBS 2889 had the highest OD 600 and R. glutinis CBS 3044 reached the second highest OD 600 . Interestingly, faster growth was observed for the red yeasts when cultivated using a mixed medium, compared to cultivations using pure CG.
L. starkeyi CBS 7786 showed slow growth on both CG alone and the mixture of CG and HH (Fig. 1). In contrast to the red yeasts, mixing CG with HH had no stimulating effect but was rather inhibiting the growth. In sHH40CG60 media, from starting glycerol concentration of approximately 50 g/L, Lipomyces starkeyi CBS 7786 consumed only around 25% of this, with 38.2 g/L glycerol still remaining after 72 h cultivation (Fig. 2). It was also only able to assimilate a small proportion of xylose, which was consumed by the other tested strains within 30 h.
In R. toruloides CBS 14 cultivations, the remaining glycerol concentration was 26 g/L after 70 h of cultivation. Xylose consumption started after 30 h (Fig. 3). R. For R. mucolaginosa CFSQE 63 and R. glutinis CBS 2889 it turned out to be difficult to measure the concentration of carbon sources in the medium, because during filtration of all samples taken after inoculation, filter membranes rapidly got clogged. Although the supernatants were centrifuged before filtration as described before [19], it was in principle not possible to filter the supernatant and for the little amount of liquid that passed through the filter, results were not reproducible (results not shown). Production of exopolysaccharides has been shown in several strains of Rhodotorula spec. [29][30][31], and the clogging of the filters may be due to production of those polysaccharides in these two strains. Due to the difficulties in analysing the metabolites in the fermentations, we did not further investigate these strains in this study.

Fermentation in bioreactors
Two of the investigated strains, R. toruloides CBS 14 and R. glutinis CBS 3044, were tested under controlled conditions in 500 ml fermenters for their abilities to utilise crude glycerol and hemicellulose hydrolysate and mixtures of them.
Cultivations were performed in 55 g/L of CG (CG55) as sole carbon source for the control and in a mixture of 55 g/l glycerol with 10% HH (HH10CG55). The C/N ratio of HH10CG55 mixture was 53.6, and CG50 C/N ratio was 48 which are slightly below the reported optimum for lipid production of 60-100 [32]. To obtain an impression on the lipid accumulation kinetics, the intracellular lipid content during fermentation was checked over the whole fermentation time using our recently established FT-NIR-based method for intracellular lipid determination [33].
In the cultivation of R. toruloides CBS 14, glycerol was faster consumed in the HHCG mixture, compared to the CG control. After 72 h, more than 75% of all glycerol was consumed in the mixed medium, and after 96 h no glycerol was detectable any more. In the control with only CG as carbon source, even after 120 h of cultivation 4.7 g/L glycerol was left in the culture and complete consumption was only found after 144 h. Moreover, the highest cell dry weight (21.2 g/L) and lipid concentration (12.5 g/L) were reached in the HHGC mixture after 96 h cultivation (Fig. 5b). In the culture with CG as sole carbon source, the maximum dry weight (19.1 g/L) and lipid concentration (8.81 g/L) were detected after 140 h cultivation (Fig. 5a).
Similarly, an activating effect was observed for R. glutinis CBS 3044 when cultivated in HHGC compared to CG as sole carbon source (Fig. 5c, d). The glycerol was almost finished after 72 h, while in the control with only CG, more than 40% of the initial carbon source was present at this time. R. glutinis CBS 3044 in CG consumed glycerol slightly faster than R toruloides CBS 14 -all of it was consumed by 120 h. In the culture with CG as sole carbon source, the maximum dry weight (20.3 g/L) and lipid concentration (7.6 g/L) were detected after 120 h cultivation (Fig. 5c). Highest lipid concentration (7.6 g/L) was reached in the HHGC mixture after 96 h cultivation (Fig. 5d). The highest cell dry In addition, we analysed the specific lipid production rates in 55 g/L CG media (Additional file 1: Fig. S1). The lipid accumulation was highest in both red yeast strains in media containing HH. R. toruloides had similar specific lipid production rates compared to R. glutinis (maximum specific rates of 0.014 vs 0.013 g/g/h), but growth was delayed by 24 h. According to a fixed effect linear model the hemicellulose hydrolysate additive had a significant impact on R. toruloides lipid accumulation (p < 0.001) but not on R. glutinis (p = 0.0625). The volumetric production rates followed the same trend as the specific rates. R. toruloides CBS14 in mixed medium had its maximum production rate (0.23 g/L lipid per hour) between 48 and 72 h of cultivation, in CG medium R. toruloides CBS14 had the highest rate (0.16 g/L lipid per hour) between 96 and 120 h. R. glutinis CBS 3044 in mixed medium had its maximum production rate (0.15 g/L lipid per hour) between 24 and 48 h of cultivation, in CG medium R. glutinis CBS 3044 had the highest rate (0.12 g/L lipid per hour) between 72 and 96 h.
These results confirmed under controlled conditions the activating effect of mixing HH and CG, which we already observed for red yeasts in shake flasks. To further verify this effect, we performed a variety of further controlled fermentations with varying concentrations of CG and HH (Additional file 2: Fig. 2 and Additional file 3:   3). For all tested concentrations (with a C/N ratio from 38.3 to 72) we found a more rapid glycerol consumption and lipid formation in cultures with a mixed carbon source. Lipid content, concentration, yield and biomass at the end of fermentation times are presented in Table 1. In most cases, the final lipid yields per g consumed carbon source were not significantly different between the cultivations with mixed carbon source and CG alone, the major effect was due to more rapid initial glycerol consumption. The highest lipid content was measured in R. toruloides CBS 14 grown in HH40CG60 medium (C/N ratio 72.2), which was probably due to the higher concentration of carbon source and higher C/N ratio. In addition, a comparatively high lipid yield (0.23 g lipids per g consumed carbon source) was observed. However, in this cultivation, not all glycerol was consumed and it turned out that the glycerol consumption rate decreased towards the end of the culture (Additional file 3: Fig. 3). In the cultivation with HH10CG55, all carbon source was consumed and the yield was even higher (0.25). The lipid yields of R. glutinis CBS 3044 were slightly lower compared to R. toruloides CBS 14.

Fatty acid analysis
The average fatty acid composition for R. toruloides CBS 14 is presented in Table 2. In all cases, there were four dominant fatty acids; oleic acid, palmitic acid, linoleic acid and stearic acid. Oleic acid (C18:1) was the most common fatty acid with the highest proportion of 46.0% of total identified fatty acids. The second major compound was palmitic acid (C16:0) with its highest proportion of 28.0%. The third major fatty acid was linoleic acid (C18:2), the highest proportion of which was 16.3% of total identified fatty acids. The last major fatty acid was stearic acid (C18:0). Its highest proportion was 12.6% of total identified fatty acids. Proportions of the remaining fatty acids can be seen in Table 2. The difference in proportion of the fatty acid concentration in samples that were taken from same cultures at 96 h and 144 h was not larger than four percent. The biggest changes were oleic acid for palmitic acid in HH10CG50 media and the opposite for CG50 after 48 h. It should also be noted that behenic acid (C22:0) was above detectable levels only

Discussion
Crude glycerol proved to be a problematic substrate for oleaginous yeasts.
Only one third of all tested strains showed substantial growth in plate-cultivation tests with CG as sole carbon source, all except one being basidiomycetes. All positively tested yeast strains were able to grow in medium containing up to 120 g/L CG, although there were differences in growth rates between the strains. All strains showed comparatively good growth up to CG concentrations of 60 g/L, which is in accordance with results presented by Chatzifragou et al. [34].
Only one of the tested Lipomyces strains showed substantial growth on crude glycerol, but considerably slower than the tested red yeasts, in fact, its growth was already retarded at CG concentrations of 60 g/L. The reason for low tolerance of Lipomyces strains to crude glycerol in all cases is unknown as the literature clearly shows that they can utilise glycerol effectively [9]. Most likely some inhibitors from ash were the cause of low tolerance. Other inhibitors in trace amount like soaps, free fatty acids and residual catalysts might also be present in our samples due to differences in purification processes of each biodiesel production plant. Another observation was that mixing both CG and HH had a negative effect on both growth and assimilation of glycerol and xylose for L. starkeyi. The presence of xylose had a slightly negative effect on lipid accumulation in other oleaginous yeasts [35]. However, from the current data we cannot draw any valid conclusion about the physiological background of the different behaviour of the red yeasts and L. starkeyi in mixtures of CG and HH.
In contrast, for the two in detail tested red yeast strains, we found a considerable activating effect of mixing HH with CG. Cells started consuming glycerol and producing lipids much earlier than in culture without HH. At least for R. toruloides CBS 14 there was a clear effect on specific lipid production after the additional carbon sources were consumed. Thus, the higher glycerol conversion was not only due to an increased biomass because of the additional carbon sources but to an increased metabolic activity. The impact on specific lipid production was less pronounced in R. glutinis CBS 3044, still, the activating effect of adding HH was clearly visible for this strain as well.
From the results it can be concluded, that an addition of HH to the culture media has a positive effect on lipid production rate for Rhodotorula species. The addition of HH also had an improving effect on biomass production. An addition of just 10% of HH (approximately 2.6 g/L xylose, 0.6 g/L glucose, 0.8 g/L acetic acid) resulted in consumption of all available carbon sources within 96 h of cultivation. Cultures without HH needed at least 24 h more to consume all glycerol in the media. The addition of HH did not impact specific glycerol uptake significantly but it accelerated reaching maximum lipid production rate in R. toruloides. This increased rate of lipid production by supplementing crude glycerol media with hemicellulose hydrolysate has not been reported before for yeast, and this is worth further investigations. In previous studies mixed substrates of lignocellulosic hydrolysates supplemented with glycerol have been performed. Shen et al. found a similar positive effect in Trichosporon fermentans of faster glycerol metabolism when adding sweet potato waste, which contained 30 g/L of reducing sugars. Gong et al. described increased lipid yield and productivity in Cutaneotrichosporon curvatum (synonym Cryptococcus curvatus) when cultivating it in a mixture of corn stover hydrolysate and glycerol [10,36]. However, in another study with the oleaginous zygomycete Cunninghamella echinulata, where tomato hydrolysate was used with glycerol no such effect has been observed [37]. Our results showed higher total lipid concentrations when compared to recently published studies, which investigated mixtures with xylose and crude glycerol of higher quality (90-95% vs our 80%) as growth substrates [35]. This may be due to better controlled conditions in bioreactors compared to shake flask cultures.
The 40% HH supplement to the culture media used to grow R. toruloides also resulted in the highest average lipid content. Even the 10% HH supplement apparently decreased the time in which lipid contents of > 40% were achieved. Without HH addition, lipid content in CG media (CG50 and CG60) was similar to that in experiments with other R. toruloides strains grown using glycerol as carbon source [38]. It is possible that with proper adjustment of the ratio between HH and CG an even higher lipid concentrations could be achieved in less than 96 h. Shortening the fermentation time has been described as a crucial aspect of an efficient process, as aerobic fermentations, in contrast to anaerobic processes, require substantial energy input for aeration [22]. The calculated yields of lipids per consumed substrate in R. toruloides in mixed media (0.18-0.25) were higher or comparable to previously reported ones using glucose, which were 0.21 g/g [39] or pure glycerol with R. toruloides 0.16 g/g [35]. In C. curvatum in cornstover hydrolysate, lipid yield was 0.18 g/g [36].
In our hands, it was necessary to add additional nitrogen to the hydrolysates when cultivating red yeasts, as the cells stopped growth and lipid accumulation during cultivation, probably due to too low nitrogen contents. For industrial processes, cheap and abundant nitrogen sources have to be identified. Food wastes are examples of such nitrogen sources, and recent results demonstrated that for instance waste milk or protein hydrolysates from chicken byproducts can be used for the production of yeast biomass [40,41].
The activating effect by the addition of HH seems to be specific for red yeasts, as in L. starkeyi CBS 7786 the addition of HH had a rather inhibiting effect. The metabolism of xylose and glycerol are both generating glyceraldehyde-3-phosphate (G-3-P). In red yeasts, glycerol assimilation is possibly activated due to an increased amount of G-3-P. Glycerol is also required to form triacylglycerol (TAG), the major storage lipid in oleaginous yeasts. It is possible that the cell activates glycerol uptake to produce TAG from the additional carbon sources. Acetic acid could also play an important role in this process, as it is converted to acetyl-CoA, the precursor of fatty acid production [42]. Nevertheless, for the moment we can only speculate on the mechanism of the activation of the lipid metabolism by HH in Rhodotorula species. Further investigations are required to understand this phenomenon, for instance studies of the transcriptome of the yeasts.
The fatty acid profiles achieved for R. toruloides in this study are similar to those of vegetable oils like palm oil, and match biodiesel profile requirements [20,43,44]. Moreover, the yeast oil can also be used in other applications, such as for the replacement of vegetable oil in fish feed [24]. Interestingly, the fatty acid profile did not vary much despite that different cultivation media were used and the time of cultivations had varied.

Conclusions
This study proved that the low-value side products CG and hemicellulose can be used for production of lipids. Addition of HH to CG considerably increased lipid production rate in R. toruloides and to some extent in R. glutinis. Based on the results of our study it seems to be possible to add value to side streams of biodiesel production and pulp and paper industry. Glycerol utilisation and shorter cultivation times are crucial factors to improve the efficiency of biodiesel production based on oleaginous yeasts [22]. In addition, this study has proven that R. toruloides CBS 14 has a good potential for microbial lipid production utilising CG and HH as carbon sources during cultivation.

Yeast strains
Strains used are presented in Table 3.
All cultivation media were sterilised by autoclaving for 20 min at 121 °C and 4 bars unless stated otherwise. Fragile media components were sterile filtered using 0.2 µm syringe filters (Sarstedt, Germany). Hemicellulose hydrolysate was filtered using 0.45 µm bottle top unit filters followed by 0.2 µm bottle top unit filters (VWR, Belgium).
For preparation of inoculum, cultures were collected into sterile 50 mL falcon tubes and centrifuged at 3000× g for 5 min. The supernatant was discarded and the pellet washed with a solution of sterile filtered NaCl (9 g/L). This process was repeated once. After refilling with the NaCl solution, OD 600 was measured to calculate inoculation volume to reach a starting OD 600 of ~ 1.
Media used in growth tests contained-1.7 g/L Yeast Nitrogen Base (YNB) w/o amino acids and 2 g/L ammonium sulphate (DifcoTM, Becton-Dickinson and Company, USA), with different concentrations of crude glycerol, hemicellulose hydrolysate or combinations of both, together with 0.1 M Potassium phosphate buffer pH 6. Tests were performed in 100 mL Erlenmeyer flasks, incubated at 25 °C on a rotary shaker at 120 rpm. In all shake flask tests, OD 600 measurement was used to quantify cell growth. All media used are presented in Table 4.
C/N ratios were determined as described in [36]. For yeast extract a C/N ratio of 3.6:1 was assumed as described in [45], which also was confirmed by own analyses (unpublished results).
Hemicellulose hydrolysate was based on wheat-straw subjected to acid-based steam explosion which was processed by the Department of Biochemical Process Engineering Luleå University of Technology, Sweden, as described previously [46]. According to HPLC-measurements, HH contained 26.2 g/L xylose, 7 g/L glucose, 6.6 g/L acetic acid and trace amounts of arabinose (< 0.5 g/L). The pH of the HH was set to 6 by addition of 5 M NaOH. In the experiments, the HH was diluted with water.

Crude glycerol and hemicellulose growth test
To test strains for the ability to grow on CG, plates with YNB and agar (16 g/L) were used without carbon source. The inoculation cultures were diluted to reach a density of ~ 0.5 McFarland and plated on the testing plates. Testing plates were (I) a CG drop was placed in the centre of the plate, (II) a sterile disk soaked in 50% CG was placed in the centre of the plate and (III) negative control without any carbon source. All the plates were incubated for 72 h in 25 °C.
For confirming the growth on mixed CG and HH, six liquid media were used: Control, sCG3, sCG6, sCG9, sCG12 and sHemi (Table 4). 20 ml of media was used in 100 ml Erlenmeyer flasks. Cultures were grown for 96 h at 25 °C on a rotary shaker at 120 rpm, with OD 600 measurements after 72 h and 96 h.
A mixture of CG and HH (sHH40CG60, C/N ratio ~ 78) was compared to CG alone (sCG60, C/N ratio ~ 58) or a control (Table 4). All samples were set up in duplicates. Cultures were grown for 72 h with OD 600 measurements each 24 h and subsequently every 6 h.