Effect of the inoculation strategy of non-Saccharomyces yeasts on wine malolactic fermentation

Interest in some non-Saccharomyces yeasts has increased recently, because they have been associated with an improvement in wine quality. Nevertheless, little attention has been paid to the effect that the use of these yeasts may have on malolactic fermentation (MLF). In this study, the strains Torulaspora delbrueckii Biodiva and Metschnikowia pulcherrima Flavia were evaluated by co-inoculation and sequential fermentation with S. cerevisiae QA23. A fermentation with S. cerevisiae as a single starter was also performed as a control, then MLF was performed inoculating Oenococcus oeni PSU-1 in all wines. Finally, the wines obtained after alcoholic fermentation and MLF were characterised. The results of the coinoculated fermentations were similar to those of the S. cerevisiae control fermentations. Nevertheless, significant differences were observed in sequential fermentations in terms of lower content of acetic, L-malic and succinic acids. These differences were particularly noticeable in fermentations carried out with T. delbrueckii.


INTRODUCTION
There has been an increasing interest in inoculating grape musts with non-Saccharomyces yeasts to complement the traditional usage of Saccharomyces cerevisiae as a sole starter, as they improve product quality and complexity (Ciani et al., 2010;Comitini et al., 2011;Contreras et al., 2014;Jolly et al., 2003Jolly et al., , 2014Whitener et al., 2015;Zott et al., 2011). These other yeast species have little or moderate fermentation power and S. cerevisiae must be inoculated to finish the alcoholic fermentations (AF) . Thus, a new trend in winemaking uses mixed starter cultures of non-Saccharomyces and S. cerevisiae (Belda et al., 2015;Ciani et al., 2010) or a sequential inoculation of S. cerevisiae after non-Saccharomyces (Contreras et al., 2014;González-Royo et al., 2015). Among the different species of non-Saccharomyces, Torulaspora delbrueckii and Metschnikowia pulcherrima show the most promising results for global wine quality, such as low production of volatile acidity (Renault et al., 2009) or a notable mannoprotein release ability, which increases the mouthfeel properties of wine (Belda et al., 2016).
Malolactic fermentation (MLF) consists of the decarboxylation of L-malic acid to L-lactic acid by the lactic acid bacteria (LAB), mainly Oenococcus oeni. In addition to decreasing wine acidity, MLF induces other changes such as microbiological stability or organoleptic improvement (Bartowsky, 2005).
The performance of MLF by LAB is affected by the intrinsic properties of wine, which are mostly determined by yeasts (Balmaseda et al., 2018). The effects of yeasts on MLF can be either inhibitory, for example the production of ethanol or the nutrient exhaustion (Arnink and Henick-Kling, 2005), or stimulating, such as the production of citric and pyruvic acids (Liu et al., 2016). These effects depend on the concentration of the compounds in wine, which, in turn, depends on species and strains (Balmaseda et al., 2018).
The aim of this study was to determine the effect of the species of non-Saccharomyces with interesting oenological traits (T. delbrueckii and M. pulcherrima) on the MLF, by evaluating and comparing the inoculation strategies of co-inoculation (non-Saccharomyces and S. cerevisiae), or their sequential inoculation at different times.

Experimental fermentations
Fermentations were performed in 500 mL flasks containing 400 mL of sterile must, prepared using white grape concentrated must (65.4 ° Brix; Mostos Españoles S.A., Tomelloso, Spain) and sterile MilliQ purified water to obtain a sugar concentration of 200 ± 10 g/L.  Each yeast was inoculated for a population of 10 6 cells/mL. All fermentations were carried out in triplicate. Samples were taken every 24 h to monitor the evolution of sugar consumption and yeast population. YPD agar was used to calculate the total number of yeast cells, and lysine agar medium (Oxoid LTD, England) was used to quantify the non-Saccharomyces (Wang et al., 2016) after incubation at 28 ºC for 48 h. AF was considered to have finished when the sugar concentration was below 1 g/L. To eliminate all yeasts, the resulting wines were filtered (MF-Millipore TM 0.45 μm, Merck Millipore, Madrid, Spain).
Next, each wine (100 mL) was inoculated with O. oeni for a population of 2 × 10 7 cells/mL. These fermentations were also carried out in triplicate. Samples were taken every 24 h to monitor the evolution of L-malic acid consumption and the bacterial population. Samples were plated on MRSmf and incubated at 27 °C in a 10 % CO 2 atmosphere for 7 days. MLF was considered to have finished when L-malic acid was below 0.05 g/L

Statistical analysis
Statistical software XLSTAT version 2018.4.51298 (Addinsoft, Paris, France) was used. The data obtained was submitted to one-way ANOVA with subsequent analysis using the Tukey test, with a confidence interval of 95 % and significant results with a p-value of ≤ 0.05. Principal component analysis (PCA) was also performed to determine differences between the wines.

Alcoholic fermentation
Control fermentation with only S. cerevisiae was completed in 10 days, with a sugar consumption rate of 31.25 g/L·day (Table 1). This was the fastest fermentation performed because, unlike coinoculated and sequential fermentations, there was neither synergy nor competition for the substrate between yeasts. Coinoculated Td-Sc and Mp-Sc fermentations was completed in 11 days with a lower consumption rate (Table 1). Table 1 shows that sequential fermentations of T. delbrueckii (Td.24 h, 48 h, 72 h) took longer than the control fermentations, largely because the final stages were slower (Figure 2 left), while fermentations of M. pulcherrima had slow early stages but finished at the same time as the control. In fact, sugar consumption was not significant until S. cerevisiae was inoculated (Figure 2, right). Nevertheless, during the initial days the non-Saccharomyces populations were stable and did not decrease until the S. cerevisiae inoculation (data not shown).  Values shown are the means of triplicates ± SD. *Calculation based on consumption speed of sugar (AF) and L-malic acid (MLF) considering the period of exponential decrease of these compounds. a-d , vales are significantly different at p ≤ 0.05, according to a Tukey post-hoc comparison test.

Malolactic fermentation
No significant differences were observed in MLF between Sc and non-Saccharomyces wines (considering the exponential decrease in L-malic acid), except for Mp 72 h, which was slower (Table 1 and Figure 3). Nonetheless, there were significant differences in terms of MLF duration between non-Saccharomyces species. MLF was slower in wines produced from Mp sequential fermentations and Td-Sc.

Changes in wine composition
As expected, the inoculation strategy had an impact on wine composition after AF ( Table 2). The most relevant changes were observed in sequential inoculation, which were related to the longer persistence of non-Saccharomyces populations, as introduced above. Td fermentations showed no significant differences in ethanol content with respect to the control. Although T. delbrueckii has been reported to produce less ethanol than S. cerevisiae (Contreras et al., 2014), several other authors have found almost no difference (Belda et al., 2015) and thus it may depend on the strain and the conditions (Benito, 2018).
On the contrary, Mp sequential fermentations were found to have significantly lower ethanol content, especially in Mp.24 h. The production of glycerol was observed to be similar in Sc and Td wines. However, Mp fermentations presented a significantly higher content at the end of AF, as has been observed elsewhere (Contreras et al., 2014). This may be due to the higher capacity of non-Saccharomyces yeasts (Mp) to use the glycopyruvic pathway instead of the usual pyruvate-to-ethanol pathway (Jolly et al., 2014).   28.32 ± 5.33ᵇ 19.97 ± 3.87ᵇ 7.00 ± 0.00 3.33 ± 1.53 29.67 ± 2.89ᵇᶜᵈ 13.33 ± 1.53ᵃᵇᶜ 2.33 ± 0.58ᵃᵇ pH Citric acid (g/L) Acetic acid (g/L) Glycerol ( The pH was significantly higher in Td wines with sequential inoculation, and was closer to initial must pH. In addition, Mp.48 h and Mp.72 h showed significantly higher pH than control Sc. A higher pH can be an attenuating factor for the inhibitory effect of ethanol on O. oeni, but MLF was not shorter in Td sequential wines. Despite this, under non-sterile cellar conditions, a pH close to 4 or higher may promote the development of other LAB, such as Pediococcus spp. and threaten wine quality (Wade et al., 2019).
Both S. cerevisiae and non-Saccharomyces consumed some L-malic acid during AF (Table 2). In Mp sequential fermentations, consumption was higher when it took longer to inoculate Sc, although all non-Saccharomyces consumed less L-malic acid than Sc. Nevertheless, high values of L-malic acid tend to ensure a good MLF. L-lactic acid production depended on L-malic acid consumption (data not shown).
Citric acid content did not vary during AF. No differences were found in its metabolisation by O. oeni, except for Td-Sc and Td.24 h, in which it was not completely consumed, unlike other fermentations. In Td fermentations, MLF was slower when O. oeni did not totally consume the citric acid.
Acetic acid concentration after AF was up to 60 % lower in Td sequential AF wines. It has been observed elsewhere that some non-Saccharomyces can decrease acetic acid concentration (Chen et al., 2018). Data obtained from Mp wines appeared to be similar to the control data, although Mp-Sc was higher, probably due to the early imposition of Sc. After MLF, as expected, the acetic acid concentration was higher due to citrate consumption.
In agreement with other studies, succinic acid production decreased in non-Saccharomyces AF (Contreras et al., 2014). These differences were most noticeable in Mp fermentations, in which succinic acid decreased by up to 10 % more than Sc fermentation. Succinic acid can act as a competitive inhibitor of the malolactic enzyme (Lonvaud-Funel and Strasser de Saad, 1982), which has a negative effect on MLF, although this inhibition has not been observed in the present study.
Td sequential fermentations consumed the least α-NH 2 -the free alpha-amino nitrogen that is equivalent to available amino acids ( Table 2). The coexistence of the two yeast populations may have resulted in higher nitrogen consumption. This data is in agreement with other studies reporting competition for nitrogen sources between yeasts (Gobert et al., 2017). No significant differences in ammonium consumption by yeasts were observed here.
One of the main products of the antagonistic interactions between yeasts and O. oeni is SO 2 (Nehme et al., 2008). Some non-Saccharomyces strains can produce significant amounts of SO 2 (Wells and Osborne, 2011). In this study, differences were found between Td wines and the others. Sequential Td wines showed that total SO 2 production was lower and the content of free SO 2 was similar. The lack of any difference between Mp and Sc wines may be the consequence of an The values shown are the mean of triplicates early imposition of the latter. The values of total SO 2 were clearly lower after MLF than before it (Table 2). This could be explained by the known reduction of bound SO 2 levels due to degradation of acetaldehyde and other binding compounds by O. oeni during MLF (Davis et al., 1985;Jackowetz and Mira de Orduña, 2012).
Considering the variables studied, the PCA (Figure 4) confirmed the differences between yeast species. It can be observed that in wines after AF (Figure 4 A), Td sequential fermentations are clustered in one group and Mp sequential fermentations are grouped in another.
The first group consumes less α-NH 2 and has a higher pH, while the second has a lower ethanol but higher glycerol content. This shows that there are similarities between wines fermented with the same non-Saccharomyces species, regardless of the time of inoculation with S. cerevisiae.
After MLF (Figure 4 B), the clusters were maintained with slight differences. Mp-Sc wine is clustered with the other sequential fermentations of Mp. In addition, all wines are closer in the PCA, indicating a homogenisation of wines after MLF due to the metabolism of the O. oeni strain used.

CONCLUSIONS
This study shows that the impact of non-Saccharomyces was greater on sequential AF than on coinoculated AF. Differences were observed between T. delbrueckii and M. pulcherrima. When T. delbrueckii was used, it had a positive effect on O. oeni and MLF due to lower acidity, succinic acid and SO 2 , even though MLF was slightly slower than in S. cerevisiae wines. M. pulcherrima decreased ethanol content during AF, which minimised its negative effect on O. oeni, yet MLF was slower than in control wines. Thus, other compounds must have a negative effect on O. oeni.
Further research is required for a better understanding of the impact of non-Saccharomyces on MLF