PCR-fingerprinting methods have previously been used to follow the population dynamics in sugar cane-based distilleries over a fermentation season (de Souza et al. 2005; Silva-Filho et al. 2005a
b; Basílio et al. 2008) and during wine fermentation (Xufre et al. 2011). In the present study, a similar method was applied to demonstrate that a resident S. cerevisiae strain, different from the inoculum strain, can dominate and repeatedly take over the fermentation process in a multistage continuous SSL fermentation plant. Whereas sugar cane juice and molasses, obtained from the processing of sugar cane, contain already considerable amounts of readily fermentable sugars (Basso et al. 2011), lignocellulosic-based feedstock requires the hydrolysis of polysaccharides to obtain fermentable sugars. As a consequence, inhibitory compounds are also released during the pretreatment step (Almeida et al. 2007), which implies that the obtained strain may differ considerably from previously isolated strains from sugar cane distilleries (Silva-Filho et al. 2005a; Basso et al. 2008). In the two isolation processes that were carried out before and after a regular plant re-inoculation with commercial baker’s yeast (BY), BY was never identified. Instead, all isolates displayed very similar molecular and physiological profiles that were distinct from BY. The identified contaminant yeast strain, named TMB3720, clearly fermented better undiluted softwood SSL than BY, with 4-fold higher maximum specific ethanol productivity and 1.8-fold higher ethanol yield. It also displayed similar ethanol yield and 1.6-fold higher maximum specific ethanol productivity than the previously reported tolerant industrial strain TMB3500 (Almeida et al. 2009).
TMB3720 clustered in a different clade that the strains isolated from the same SSL ethanol plant in the early 90’s (Lindén et al. 1992). The two isolation processes are separated by more than 15 years and by the implementation of four stainless steel fermentation tanks in the SSL plant, which indicates that the modifications in the yeast population result from differences in environmental conditions as well as the possible input of novel wild yeasts species from the raw material (Silva-Filho et al. 2005b). Chronical contamination episodes are actually expected in large scale facilities as bioethanol fermentations are not designed to be carried out under sterile conditions (Skinner and Leathers 2004). However such episodes are often associated with negative fermentation performances. Notably, contamination by lactic acid bacteria is regarded as one of the major problems in ethanol production facilities (Hynes et al. 1997; Narendranath et al. 1997). A reduction up to 7.6% of ethanol concentration was, for example, demonstrated when 109 CFU of lactic acid bacteria/ml were added to wheat mash, as a result of lactic acid production and possible competition with yeast for essential growth factors (Narendranath et al. x1997). Also, an economically relevant decrease of ethanol yield was observed when the level of the yeast Dekkera bruxellensis, which was consistently identified as the main contaminant in bioethanol distilleries in Brazil, increased up to almost 50% of the yeast population (de Souza Liberal et al. 2007). In some cases however, microbial contamination did not give detrimental effects. For example, a stable ethanol-producing consortium consisting of D. bruxellensis and Lactobacillus vini showed the same efficiency and productivity as the inoculated commercial baker’s yeast in a wheat starch-based alcohol production process (Passoth et al. 2007).
In the present study, the contaminating strain, TMB3720, showed even clearly superior traits than the inoculated commercial BY strain in fermenting undiluted softwood SSL, which indicates that the lack of sterility can also be an advantage for the selection of better performing wild yeast strains for SSL fermentation processes.
Acetic acid tolerance and furaldehyde reduction capacity may explain why TMB3720 strain became the resident yeast in the SSL-based fermentation plant. At the plant operation pH, pH 5, the maximum specific growth rate of TMB3720 was not affected by the presence of acetic acid whereas a reduction of 6% was observed for BY. At this pH, 42% of the total acetic acid (6 g l-1) remains in undissociated form which diffuses through the cell membrane. Once inside the cell, acetic acid dissociates, thereby releasing protons and acidifying the cytosol. To be able to maintain the intracellular pH level, the excess of protons have to be transported out of the cell using ATP, therefore additional ATP needs to be synthesized. If the ATP production rate becomes limiting, less ATP is available for biomass formation (Pampulha and Loureiro-Dias 2000). Therefore, in the presence of acetic acid, TMB3720 might be able to cope with the extra ATP demand without reducing the biosynthetic capability to produce biomass. Also, TMB3720 showed higher in vitro reductase activity for furfural and HMF as compared to BY. The lower reductase capacity of BY together with the fact that it could not be found during the second isolation process, indicates that this strain was not able to cope with the inhibitory conditions of the fermentation plant. Moreover, since yeast is being recirculated, the cell lysis of BY could serve as a source of fresh nutrients for TMB3720.
During the cell propagation step for SSL fermentation and also in the evaluation of the acetic acid tolerance, TMB3720 heavy flocculated when grown in liquid media. A formation of flocs with the subsequent rapid sedimentation from the medium was observed regardless of the type of medium used (defined or rich). The same process of aggregation took place when the cells were grown in the presence of softwood SSL. In parallel, TMB3720 had a significantly lower specific growth rate (0.26 ± 0.01 h-1) compared to two other tested industrial strains (0.44 ± 0.00 h-1 and 0.45 ± 0.00 h-1 for BY and TMB3500 respectively). This fact could be linked to mass transfer limitations inside the floc due its high flocculation capability. However, the variability in the size flocs, together with the fact that they could not be dispersed completely by the use of EDTA, might also lead to an underestimation of the maximum specific growth rate. Flocculation has been suggested to be a social behaviour and to act as a mechanism of survival under nutrient limitation or high ethanol concentrations (Soares 2010). Yeast flocculation depends on the expression of specific flocculation genes such as FLO1, FLO5, FLO8 and FLO11 (Verstrepen et al. 2003). When comparing the non-flocculent S288C S. cerevisiae strain and its FLO1-overexpressing variant under ethanol stress, the number of the cell survival of the flocculent strain was two-fold greater compared to the wild type, and it was suggested that cells enclosed within the floc would induce changes at physiological level that would promote stress resistance (Smukalla et al. 2008). Therefore, flocculation may also partly explain why TMB3720 predominates under the harsh conditions of the SSL plant. On the one hand, the external part of the floc might act as a protectant that also detoxify compounds for the inside cells. On the other hand, the possible cell lysis could be a provision of nutrients for the metabolic active cells present inside the floc.
In the course of the study, Pichia
galeiformis yeast contaminant was also isolated in the last fermentation tanks. Xylose-utilising Pichia species have previously been identified in the same SSL plant (Lindén et al. 1992). Their presence is most probably connected to the fact that S. cerevisiae cannot naturally consume xylose, which would leave this sugar fraction available for xylose-utilising contaminant species at any point of the process line. Also, it is known that another Pichia species, Pichia
(Scheffersomyces) stipitis, is inhibited by compounds present in lignocellulosic biomass (Lohmeier-Vogel et al. 1998). Therefore, we can speculate that the detoxification process performed by S. cerevisiae in the fermentation line would reduce the toxicity of the SSL and allow less tolerant contaminant species to be present only at the end of the process line. Pichia isolates found at the end of the fermentation line would, however, not be tolerant enough to be a contaminant able to take over the entire fermentation line. The presence of inhibitory compounds, and notably sulfur-derivate compounds (du Toit et al. 2005) may also explain why Dekkera species were not found in any of the isolation processes, although it has been identified as a major contaminant species in sugar cane juice distilleries (de Souza Liberal et al. 2007; Basílio et al. 2008; Basso et al. 2008) and is also believed to be the cause of reported stuck fermentations in corn mash-based fermentation processes (Abbott et al. 2005).
In conclusion, TMB3720 is an adapted high-productive and robust S. cerevisiae strain that can be employed as regular inoculum in SSL plants. More generally, the isolation and characterization of tolerant high ethanol producing resident yeasts from lignocellulosic biomass plants will increase the knowledge on tolerance to highly inhibitory compounds as well as generate background strains for further improvement by targeted and/or evolutionary engineering.