Supplementary material

Medium optimization for starting-up stage

The amount of the main medium component (i.e. glucose, yeast extract and alcohol) was optimized based on the growth curve with dynamic modelsmethodology. The growth curves were obtained at various concentration of the optimized objectexperiment with experiment. The growth curve was fitted with Logistic model (Seq. 1) by least square method to find the maximal biomass in theory (Xm) to the given concentration ofthe corresponding medium composition.

(Seq. 1)

Where X0: the biomass at the start in theory, Xm: themaximal biomass in theory, w: half-peakwidth, μm: thespecificgrowthrate in theory.

The maximal biomass in theory value (Xm) to the concentrationof the corresponding medium component (S) was further fitted with peak function such as Gaussianfunction (Seq. 2), Logistpk function (Seq. 3)and Lorentzfunction (Seq. 4) to obtain the optimal value of the medium component. (Yu L, Fei X, Lei T. Dynamic models for L-lacticacid batch fermentation by Lactobacillus rhamnosus.Science and Technology of Food Industry, 2008, 29: 100-102). To the main medium component (ie glucose, yeast extract and alcohol), the corresponding experiment data and the data analysis were given in the following.

(Seq. 2)

(Seq. 3)

(Seq. 4)

Glucose optimization

Glucose concentrations were2.0,3.0, 4.0, 5.0 and 6.0 g/L.

S.Fig. 1 Growth curves and fitting of biomass with Logistic equation with different concentration of glucose

Glucose concentration (S): 2.0,3.0, 4.0, 5.0, and 6.0 g/L.

S.Table 1 Parameter values of the Logistic modelwith Seq. 1 (glucose)

Glucose Concentrationg/L / X0 g/L / Xmg/L / µm /h / σ / R2
2.0 / 0.001 / 0.765 / 0.106 / 0.002 / 0.985
3.0 / 0.001 / 0.759 / 0.145 / 0.001 / 0.989
4.0 / 0.001 / 0.770 / 0.150 / 0.002 / 0.987
5.0 / 0.001 / 0.850 / 0.129 / 0.002 / 0.988
6.0 / 0.001 / 0.776 / 0.128 / 0.002 / 0.983

S.Fig. 2 Fitting curves of Xmto various glucose concentrations (S)with Logistpk function, Gaussian function and Lorentz function

To Logistpk function, Gaussian function and Lorentz function,the fittingdegree(R2) are respectively 0.986, 0.987 and 0.980. Based on the R2, Gaussian equation is the most suitable. When Xmwas thehighest, the corresponding parameter values were X0=0.762，Sm=5.066，A=0.108，w=1.144. According it, the optimum value for glucose concentration in the medium is about 5.066 g/L, inpractice it was 5.1 g/L.

Nitrogen Sourceoptimization

Yeast extract was applied as the nitrogen source. 23.0, 24.0, 25.0, 26.0 and 27.0g/L of yeast extract were applied.

S.Fig. 3 Growth curves and fitting of biomass with Logistic equation withvarious concentration of yeast extract

Yeast extractconcentration (S):23.0, 24.0, 25.0, 26.0, and 27.0g/L

S.Table2 Parameter values of the Logistic mode with Seq. 1 (yeast extract)

yeast extract g/L / X0 g/L / Xmg/L / µm /h / σ / R2
23.0 / 0.002 / 0.896 / 0.154 / 0.003 / 0.984
24.0 / 0.001 / 0.900 / 0.158 / 0.002 / 0.992
25.0 / 0.001 / 0.910 / 0.164 / 0.002 / 0.988
26.0 / 0.003 / 0.969 / 0.136 / 0.008 / 0.963
37.0 / 0.002 / 0.924 / 0.147 / 0.002 / 0.992

S.Fig. 4 Fitting curves of Xmto various yeast extract concentrations (S) with Logistpk function, Gaussian function and Lorentz function

The fitting degree (R2) is 0.999, 0.991 and 0.995 for Lorentz function, Gaussian function and Logistpkfunction respectively. Accord to the fittingdegree (R2), the Lorentz function is the best. The parameter values were X0=0.894, Sm=26.207, w=1.188, A=0.158. whileX=Xm, X have thehighest. So, the optimum value for yeast extract amount is about 26.207 g/L, 26.2 g/Lpractice.

Ethanol concentrationsoptimization

Ethanol concentrations were 10.0, 20.0, 30.0, 40.0, 50.0 g/L.

S.Fig. 5 Growth curves and fitting of biomass with Logistic equation withvariousconcentration of ethanol

Ethanol concentrations (S):0, 10.0, 20.0, 30.0, and 40.0g/L.

S.Table3 Parameter values of the Logistic mode withSeq. 1 (ethanol)

ethanol g/L / X0 g/L / Xmg/L / µm /h / σ / R2
0.0 / 0.001 / 2.307 / 0.134 / 0.001 / 0.993
10.0 / 0.002 / 4.162 / 0.189 / 0.005 / 0.986
20.0 / 0.001 / 3.331 / 0.068 / 0.010 / 0.879
30.0 / 0.001 / 0.841 / 0.031 / 0.001 / 0.967
40.0 / 0.002 / 0.855 / 0.028 / 0.001 / 0.993

S.Fig. 6 Fitting curves of Xmto various ethanol concentrations (S) with Logistpk function, Gaussian function and Lorentz function

The fitting degree (R2) isrespectively 0.914, 0.823 and 0.872 forGaussian function, Logistpkfunction andLorentz function.Accord to the fittingdegree (R2), the Gaussian equation is the best. The parameter values were X0=0.658, Sm=11.927, w=1.809, A=4.018. So, the optimum value for ethanol amount is about 11.927 g/L, in practiceit was 11.9 mL/L.

Confirm the optimized conditions for the starting-up stage

To confirm the optimized medium for the starting-up stage, the AAB was culture in seeds medium YG1and inoculated into the optimized YG2 medium (5.1 g glucose, 26.2 g yeast extract, 0.5 g KH2PO4, 0.5 g MgSO4 and 0.5 g NaH2PO4, 11.9gedible alcohol in 1L of distilled water). The corresponding growth curve was measured. The results were given in S.Fig. 7.

S.Fig. 7AAB growth curves with the optimized YG2 medium