To investigate the effects of fermentation parameters on the cell growth and on the resistance to each step of the freeze-drying process of Lactococcus lactis subsp. lactis TOMSC161, a natural cheese isolate, using a response surface methodology.
Methods and Results
Cells were cultivated at different temperatures (22, 30 and 38°C) and pH (5·6, 6·2 and 6·8) and were harvested at different growth phases (0, 3 and 6 h of stationary phase). Cultivability and acidification activity losses of Lc. lactis were quantified after freezing, drying, 1 and 3 months of storage at 4 and 25°C. Lactococcus lactis was not damaged by freezing but was sensitive to drying and to ambient temperature storage. Moreover, the fermentation temperature and the harvesting time influenced the drying resistance of Lc. lactis.
Lactococcus lactis cells grown in a whey-based medium at 32°C, pH 6·2 and harvested at late stationary phase exhibited both an optimal growth and the highest resistance to freeze-drying and storage.
Significance and Impact of the Study
A better insight on the individual and interaction effects of fermentation parameters made it possible the freeze-drying and storage preservation of a sensitive strain of technological interest. Evidence on the particularly damaging effect of the drying step and the high-temperature storage is presented.
Lactic acid bacteria (LAB) are widely used as starters to manufacture fermented food (dairy, meat and vegetable products), probiotic products and in green chemistry applications. Among LAB, Lactococcus lactis is extensively used as an industrial starter to manufacture a wide range of fermented dairy products (Champagne et al. 2009; Turchi et al. 2013). This micro-organism has many interesting technological properties such as lactic acid and flavour production and texturing (e.g. exopolysaccharide production) (Ozkalp et al. 2007; Gutierrez-Mendez et al. 2008; Kristo et al. 2011). The strain TOMSC161 is particularly interesting for its cheesemaking properties and is used for producing nonscaled curd pressed cheeses (Tomme de Savoie).
Before being used in industrial applications, the preparation of starter requires production (fermentation, concentration) and preservation (freezing or freeze-drying) techniques that maximize the viability and the activity of bacterial cells during long-term storage.
Freeze-drying is a commonly used technique for the long-term preservation of LAB due to its low storage and transportation costs compared with freezing and its easy handling in countries where the management of low temperatures is not well established. It involves freezing the aqueous solution containing bacteria, followed by primary drying to remove ice by sublimation and, finally, secondary drying to remove unfrozen water by desorption. Even if low temperatures are applied, not all strains survive the process and among those that survive, very poor viability recovery are recurrently reported (Carvalho et al. 2003b; Savini et al. 2010). It is well known that resistance to freeze-drying and storage varies among strains and Lc. lactis TOMSC161 is a particularly sensitive one, thus exhibiting significant acidification activity loss following stabilization processes. The freeze-drying process involves two main events that damage bacteria: freezing and drying. Freezing is at the origin of the cryoconcentration of solutes and the formation of ice crystals, which can lead to osmotic stress and cell dehydration. Drying, especially the removal of unfrozen water associated with cell, can lead to changes in the physical state of membrane lipids and in the structure of sensitive proteins due to breakage of hydrogen bonds (Leslie et al. 1995; Castro et al. 1997).
Lyoprotectants such as polymers, sugars, amino acids and antioxidant molecules are usually added to limit stress and to enhance LAB resistance to freeze-drying (Carcoba and Rodriguez 2000; Carvalho et al. 2002). Moreover, operating conditions such as the cooling rate, the thermal history of the product during drying, the residual moisture content and the subsequent storage conditions, including temperature, water activity and atmosphere, have an impact on bacterial survival (Font de Valdez et al. 1985; Schoug et al. 2006; Higl et al. 2007; Trelea et al. 2009; Zhao and Zhang 2009; Passot et al. 2012).
Bacterial cells can also be prepared to cope with the different stresses of freeze-drying by varying the growth conditions before freeze-drying. Fermentation conditions, including medium composition, temperature, pH, harvesting time, gas atmosphere and other pre-adaptation treatments, reported for LAB are detailed in Table 1. Most of these studies used a ‘one-factor-at-a-time’ testing approach, which requires a large number of experiments and may still fail to induce the optimal responses, especially when the factors jointly, not just individually, influence the cell responses. Furthermore, the freeze-drying process is often considered as a ‘black box’, and the storage stability of freeze-dried LAB is hardly ever studied. Finally, the reported experimental procedures were applied in conditions very different from those found in industry using rich growth media and unadapted protective molecules and analysing the freeze-drying resistance just in terms of survival but not in terms of activity preservation, such as acidification activity, which is an important parameter for industrial applications.
Table 1. Overview of studies on the influence of fermentation conditions on the freeze-drying survival of LAB
The aim of this work was therefore to investigate the combined effects of three important fermentation parameters, temperature, pH and harvesting time, on the resistance of Lc. lactis TOMSC161 at each step of the production process of freeze-dried cells. A response surface methodology (RSM) was chosen to study relationships between fermentation parameters and bacterial cultivability and acidification activity losses (Box and Behnken 1960). The experimental design used attempted to reach a better understanding of the physiological impact of fermentation conditions on freeze-drying and storage resistance and to identify the key steps of the bacterial preservation process. This study was conducted in conditions close to those of an industrial process (simple medium), thus making it possible to recommend fermentation conditions to maximize cell survival and acidification activity after freeze-drying and storage of a strain of technological interest but sensitive to freeze-drying.
Materials and methods
The experimental approach used during this study and the main parameters investigated are summarized in Fig. 1.
Lyophilized starter production process
Strain and inoculum preparation
Lactococcus lactis subsp lactis TOMSC161 (Savoîcime, France), a natural cheese isolate, was studied because of its sensitivity to freeze-drying. Frozen cells were stored at −80°C in M17 medium (Biokar Diagnostics; Beauvais, France) supplemented with 15% (v/v) glycerol (VWR; Leuven, Belgium). One millilitre of frozen aliquot was precultured in 100 ml of M17 medium at 30°C for 4·5 h before inoculation in the bioreactor.
The culture medium was composed of 60 g l−1 lactose (VWR; Leuven, Belgium) and 15 g l−1 yeast extract (BD; Sparks, MD). After adjusting the pH at 6·8 with a 30% KOH solution, the medium was introduced into a 5-l working volume bioreactor (Biostat® A plus, Sartorius-Stedim; Aubagne, France) and sterilized at 120°C for 20 min. The inoculation was performed at a low initial concentration of approx. 100 CFU ml−1. An agitation speed of 200 rev min−1 was applied to permit homogenization. Temperature and pH were set at different values that depended on the experimental design (Fig. 1). The pH was controlled by the automatic addition of 15 mol l−1 KOH solution (Merck; Darmstadt, Germany) in the bioreactor. The weight of the KOH solution was monitored throughout fermentation, and the consumption rate (dm/dt, in g min−1) was calculated with WCidus software (INRA, Thiverval-Grignon, France). The culture was stopped at three different times of the stationary phase (Fig. 1). The base consumption rate was used to determine the beginning of the stationary phase fixed to 1·5 h after the maximum acidification rate.
Concentration and lyoprotection
Cells were harvested at three different times depending on the experimental design and concentrated 10 times by centrifugation (Avanti® J-E centrifuge; Beckman Coulter; Fullerton, CA) at 17 000 g for 20 min at 4°C. Concentrated cells were resuspended in protective medium at a ratio of 1 : 2 (1 g of concentrated cells for 2 g of protective medium) prior to freeze-drying. The protective solution was composed of 60 g l−1 sucrose (VWR; Leuven, Belgium), 140 g l−1 maltodextrin (dextrose equivalent = 6; M = 3000 g mol−1) (Glucidex 6D, Roquette; Lestrem, France) and 9 g l−1 NaCl and was sterilized at 110°C for 20 min. The lyoprotected cell suspensions were distributed in 5-ml vials (Verretubex; Nogent-Le-Roi, France) with a volume of 1 ml prior to freezing at −80°C (freezing rate of 3°C min−1).
Freeze-drying and storage
The samples were frozen at −80°C in a cold air chamber and then transferred to a precooled shelf at −50°C in a prototype pilot freeze-dryer built by Telstar (Terrassa, Spain). After a holding step of 2 h at −50°C, the chamber pressure was decreased to 20 Pa and the shelf temperature was increased from −50 to −20°C at a heating rate of 0·25°C min−1 to initiate the sublimation step. After 40 h of sublimation, the shelf temperature was increased to 25°C at a heating rate of 0·25°C min−1 to initiate the secondary drying step. After 8 h of desorption, the vacuum was broken by injecting air and the vials were capped. Finally, freeze-dried samples were packed under vacuum in aluminium bags and stored at 4 and 25°C for 3 months to evaluate the stability of starters.
A standard response surface methodology known as the Box-Behnken design (Box and Behnken 1960) was performed to investigate the linear, interaction and quadratic effects of multiple fermentation parameters (temperature, pH and harvesting time) on cell growth and on freeze-drying and storage resistance. Each independent variable was coded at three levels between −1, 0 and +1, corresponding to the low, middle and high level, respectively. The experimental ranges of the three independent variables were set to 22, 30 and 38°C for the temperature, 5·6, 6·2 and 6·8 for the pH, and early stationary phase, 3 and 6 h of stationary phase for the harvesting time. In the case of temperature and pH, the middle level was determined according to the literature on optimal temperature and pH conditions for Lc. lactis growth (Hofvendahl et al. 1999; Adamberg et al. 2003). Low and high levels were chosen to create stressful conditions for Lc. lactis cells, but were still compatible with acceptable growth. Finally, the harvesting time had to be performed in the stationary phase to have maximum biomass. The low level corresponded to the early stationary phase and the high level was chosen to avoid carbon (lactose) starvation in the fermentation medium. This experimental design thus involved a three-factor-three-level pattern with 15 design points (13 combinations with the three replications of the centre point), and the entire design was carried out in random order.
Cell growth properties
Growth kinetics measurement
Cell growth was monitored by measuring the optical density every hour at 650 nm (OD650) in a DU® 640B spectrophotometer (Beckman Coulter). The specific growth rate (μmax in h−1) was calculated according to the following equation (Eqn (1)):
where N is the cell density (OD650) and t is time in hour.
Cell dry weight (Xmax) was determined by filtering 10 ml of culture sample through nitrocellulose filters (pore size: 0·22 μm) and drying the filters at 80°C for 24 h. The results are given as dry weight per volume of culture (g l−1). The cell dry weight was measured at the early stationary phase in triplicate.
Substrate and metabolite analysis
Lactose, glucose, galactose, lactic acid, succinic acid, acetic acid and ethanol were quantified using high-performance liquid chromatography (HPLC, Waters Associates, Millipore; Molshein, France) following the experimental procedure previously described by Rault et al. (2008). All quantifications were performed at least in triplicate. The substrate and metabolite concentrations were measured at the beginning of the fermentation, at the early stationary phase and at the harvesting time as well.
Biological and physical properties of starters
The bacterial cell concentration was determined using the agar plate count method. The frozen cells were thawed for 15 min at room temperature. The freeze-dried cells were rehydrated in 1 ml of saline water (NaCl, 9 g l−1) and stirred for 5 min at room temperature. Cell suspensions were serially diluted in saline water, plated into M17 agar (Biokar Diagnostics) and incubated at 30°C for 48 h in anaerobiosis. The cell count was expressed in CFU ml−1, and the measurements were taken in triplicate. The cultivability was measured after concentration and lyoprotection, freezing, drying and after 1 and 3 months of storage at 4 and 25°C.
The determination of the cultivability loss after each step of the process was calculated using the following equation (Eqn (2)):
where i and i + 1 are two successive steps of the process.
Acidification activity measurement
The Cinac system (AMS; Frépillon, France) was used to measure acidification activity of bacterial suspensions (Spinnler and Corrieu 1989). Acidification was measured in triplicate at 30°C using reconstituted skim milk with 100 g l−1 skim milk powder (EPI Ingredients, Ancenis, France), heat treated at 110°C for 15 min in 120-ml flasks and stored at 4°C before use. The flasks were inoculated with 500 μl of bacterial suspensions. The inoculated milk pH was continuously measured by the Cinac system and led to the determination of the time necessary to obtain a decrease of 0·7 pH units (tpH0·7, in min). The descriptor tpH0·7 was used to characterize the acidification activity of bacterial suspensions. The higher the tpH0·7 value was, the longer the latency phase was and the lower the acidification activity was. The acidification activity was measured after concentration and lyoprotection, freezing, drying and after 1 and 3 months of storage at 4 and 25°C.
As a result, the acidification activity loss was calculated as follows (Eqn (3)):
where i and i + 1 are two successive steps of the process.
An increase in the dtpH0·7 value corresponds to an increased loss of acidification activity during the process step considered.
Water activity and water content measurement
The moisture content of the samples was measured after freeze-drying and after 1 and 3 months of storage by the Karl Fisher titration method using a Metrohom KF 756 apparatus (Herisau, Switzerland). At least 20 mg of powder was mixed with 2 ml of dry methanol and titrated with Riedel-deHaen reagent (Seelze, Germany) until the end point was reached.
The water activity of the samples was measured after freeze-drying using a LabMasteraw aw-meter (Novasina, Precisa, Poissy, France). Three freeze-dried samples were pooled and water activity was measured at 25°C. The instrument was calibrated with saturated pure salt solutions (aw: 0·04, 0·06 and 0·11) (SAL-T, Novasina).
Multiple regression analyses were performed to examine the experimental data obtained by the Box-Behnken design and to quantify the effect of the independent variables (fermentation conditions) on the responses (cell growth and freeze-drying and storage resistances) using Statgraphics®Plus 3.0 (Statistical Graphics Corp., Rockville, MA). A second-order polynomial model was generated using least square method to represent each response variable (μmax, Xmax, dlog(CFU)i+1, dtpH0·7i+1).
The 15 runs of the complete Box-Behnken design were used to represent the cultivability loss and the acidification activity loss during each step (dlog(CFU)i+1, dtpH0·7i+1) as a function of fermentation temperature, fermentation pH and harvesting time using the following equation (Eqn (4)):
where Y is the response (cultivability loss or acidification activity loss during one process step), X1,X2 and X3 are uncoded independent variables (fermentation temperature and pH and harvesting time), and β0, β1, β2,β3, β11, β22, β33, β12, β13 and β23 are intercept, linear, quadratic and interaction constant coefficients, respectively.
Concerning the cell growth properties (specific growth rate and final biomass concentration), only the fermentation temperature and the fermentation pH were included in the model with the following equation (Eqn (5)):
where Y is the response (specific growth rate and final biomass concentration), X1 and X2 are uncoded independent variables (fermentation temperature and pH) and β0, β1, β2,β11, β22 and β12 are intercept, linear, quadratic and interaction constant coefficients, respectively.
anova tables were also generated to evaluate the significance of each independent variable. A significance level of 95% (P-value lower than 0·05) was considered.
The adequacy of the models developed was tested by calculating the coefficient of determination which is a measure of the percentage of the total variation explained by the model.
To visualize the relationships between the responses and variables, response surface plots were generated from the fitted quadratic polynomial equations (Eqns (4) and (5)) obtained.
Finally, a numerical optimization technique was applied to determine the optimal fermentation conditions for the various responses involved in the freeze-dried starter production using Matlab® R2010b software (MathWorks; Natick, MA).
Effect of fermentation conditions (temperature and pH) on bacterial growth
Firstly, the substrate and metabolite concentrations were analysed at the early stationary phase and at the harvesting time for all the studied fermentation conditions. Growth of Lc. lactis TOMSC161 at different temperatures and pH resulted in the same end product formation, namely lactate. This indicates that this strain used only the homolactic pathway within the studied conditions. Moreover, regardless of the fermentation conditions applied, the media always contained residual lactose (>1 g l−1), thus confirming that the beginning of stationary phase was not due to carbon starvation.
Secondly, the bacterial growth was examined for all the studied fermentation conditions. The results of the multiple regression analyses are summarized in Table 2. They enable to quantify the linear, interactive and quadratic effects of the independent variables (fermentation temperature and pH) on the cell growth responses: specific growth rate (μmax) and final biomass concentration (Xmax). The coefficients (β) of Eqn (5) as well as the P-value of each variable of the models for both responses are presented. It should be noted that the models explain more than 92% of the variability in the response variables according to the coefficients of determination (R2) of the multiple regressions, thus indicating that the models accurately represent the specific growth rate and the final biomass concentration. Figure 2 shows the three-dimensional representation of the quadratic models within the experimental domain and enables the visualization of both the response surface of μmax (Fig. 2a) and Xmax (Fig. 2b) and the conjugated effect of fermentation temperature and pH.
Table 2. Regression analysis for the specific growth rate and the final biomass concentration (X1: fermentation temperature; X2: fermentation pH) (Eqn (5))
Estimated coefficient (β)
Specific growth rate (Y)
R2 = 95·42%
Standard error = 0·019 h−1
Final biomass concentration (Y)
R2 = 92·62%
Standard error = 0·68 g l−1
The specific growth rate (μmax) was influenced by the linear effect of temperature, the quadratic effect of pH and the interactive effect of temperature and pH (P-value < 0·05) (Table 2). The surface response of the specific growth rate (Fig. 2a) displayed a bell-shaped form with a maximal value of 0·49 h−1 for temperature and pH values of 32°C and 6·1, respectively. Minimal values of 0·15 and 0·24 h−1 were obtained for low values of temperature (22°C) and pH (5·6) and high values of temperature (38°C) and pH (6·8), respectively. The interaction between temperature and pH was clearly visible in Fig. 2a when considering extreme fermentation conditions. At low temperature (22°C) and low pH (pH < 6·2), the specific growth rate sharply increased with pH (from 0·15 to 0·30 h−1, from pH 5·6 to pH 6·2). Conversely, the specific growth rate sharply decreased with pH (from 0·42 to 0·24 h−1) at high temperature (38°C) and high pH (pH varying from 6·2 to 6·8).
The final biomass concentration (Xmax) was influenced by linear effect of pH and quadratic effect of temperature (P-value < 0·05). Increasing pH also increased the final biomass concentration (Fig. 2b). For example, at 30°C, increasing the pH from 5·6 to 6·8 increased the final biomass by 44% (from 6·8 to 9·8 g l−1). The model predicted an optimal value of final biomass concentration of 9·8 g l−1 at 29°C and a pH value of 6·8.
Quantification of the cultivability and acidification activity losses after each step of the freeze-drying process and storage
The cultivability and the acidification activity of the bacteria were determined after each step of the preservation process: (i) after concentration and lyoprotection (initial values), (ii) after freezing, (iii) after drying (sublimation and desorption) and (iv) after 1 and 3 months of storage at 4 and 25°C. Regardless of the fermentation conditions applied (pH, temperature, harvesting time), the concentrated and lyoprotected samples presented the same cultivability (1·5 1011 ± 0·3 1011 CFU ml−1) and acidification activity (400 ± 24 min) before freeze-drying. In Fig. 3, the losses of cultivability (Fig. 3a) and acidification activity (Fig. 3b) are displayed after each step of the process. For the sake of clarity, only the data concerning the centre point of the Box-Behnken design (runs 13, 14, 15; Fig. 1) were plotted. The same tendencies were observed, regardless of the fermentation conditions applied.
No significant losses of cultivability were observed after freezing, drying and storage at 4°C. In contrast, high losses of cultivability were measured after 1 and 3 months of storage at 25°C with loss values of 1·21 ± 0·3 log and 2·43 ± 0·31 log, respectively. A total cultivability loss of 2·75 ± 0·25 log was observed after freeze-drying and 3 months of storage at 25°C.
Acidification activity is an important technological property for industrial application. Low losses of acidification activity were observed after freezing and storage at 4°C. However, significant loss was measured after the drying step with an increase of 112 ± 5 min of the tpH0·7 (i.e. a 30% increase in the tpH0·7 value obtained before freeze-drying). As expected, storage at 25°C resulted in high losses of acidification activity with an increase of 208 ± 17 min and 379 ± 17 min of the tpH0·7 after 1 and 3 months of storage, respectively. The freeze-drying process followed by 3 months of storage at 25°C induced a total loss of acidification activity of 499 ± 6 min (i.e. a 125% increase in the tpH0·7 value obtained before freeze-drying).
Water content and water activity of the freeze-dried starters were identical regardless of the fermentation conditions applied: 0·43 ± 0·05% and 0·024 ± 0·003, respectively. No modification of the water content was observed during storage at 4 and 25°C.
Effect of fermentation conditions (temperature, pH and harvesting time) on cultivability and acidification activity losses throughout the starter freeze-drying process
Multiple regression analyses were applied to the 15 runs of the Box-Behnken design to evaluate the linear, quadratic and interactive effects of the fermentation conditions (temperature, pH and harvesting time) on the losses of cultivability and acidification activity (dlog(CFU)i+1 and dtpH0·7i+1) after each step of the preservation process. Only the acidification activity loss after the drying step was significantly influenced by the culture conditions.
The results of the regression analysis concerning the acidification activity loss during the drying step are summarized in Table 3. The coefficients of Eqn (4) are shown in Table 3 and used to plot the corresponding three-dimensional surface responses for the three harvesting times studied (early stationary phase in Fig. 4a; 3 h of stationary phase in Fig. 4b; 6 h of stationary phase in Fig. 4c). The high value of R2 (0·85) indicated that the second-order polynomial equation is capable of representing quite well the loss of acidification activity during drying under the given experimental domain. The loss of acidification activity during drying was significantly influenced by temperature and harvesting time (linear effect) (P-value < 0·05).
Table 3. Regression analysis for the acidification activity loss during drying (X1: fermentation temperature; X2: fermentation pH; X3: harvesting time) (Eqn (4))
Estimated coefficient (β)
R2 = 84·87%
Standard error = 25 min
The response surfaces are different according to the harvesting time and present an inverted bell-shaped surface only for a harvest at 6 h of stationary phase. Increasing the harvesting time resulted in decreasing the loss of acidification activity during drying. For example, increasing the harvesting time from the early stationary phase to 6 h of stationary phase resulted in a decrease in acidification activity loss during drying of 50% (from 185 to 92 min) for a fermentation at 22°C and pH 6·2. Linear effect of fermentation temperature indicates that increasing temperature values were related to reduce losses of acidification activity. However, even if no significant interaction was observed between factors, temperature effect appeared dependent on harvesting time (Fig. 4). For a harvest at early stationary phase and at 3 h of stationary phase, increasing the temperature also resulted in a decrease in the loss of acidification activity loss during drying. For example, increasing the temperature from 22 to 38°C resulted in a decrease in acidification activity loss during drying of 52% (from 185 to 88 min) for a fermentation at pH 6·2 and a harvest at the early stationary phase (Fig. 4a). Consequently, high fermentation temperatures (38°C) were the optimal conditions found for a harvest at early stationary phase and at 3 h of stationary phase. However, this linear effect of temperature decreased with increasing harvesting time (Fig. 4b) and changed the optimal fermentation temperature after late stationary phase (6 h) thus identified by numerical optimization at 32°C (Fig. 4c). As the fermentation pH did not influence freeze-drying resistance, pH has to be chosen to optimize growth parameters. As low pH (5·6) induced low final biomass concentration and high pH (6·8) led to low specific growth rate, fermentation optimal pH was fixed at 6·2. Fermentation at 32°C, pH 6·2 and a harvest at 6 h of stationary phase therefore made it possible to optimize both the growth parameter and the freeze-drying resistance of Lc. lactis TOMSC161.
To confirm the relevance of the predicted optimized conditions, three experiments were carried out close to predicted optimized conditions and results were compared with predicted values of the responses using the model Eqns (4) and (5). The complementary experiments were therefore conducted at 30°C, pH 6·2 and harvested after 6 h of stationary phase. When the bacterial cells of Lc. lactis TOMSC161 were produced and lyophilized in these conditions, the specific growth rate, the biomass concentration and the acidification activity loss during dehydration were 0·47 ± 0·01 h−1, 9·6 ± 0·1 g l−1 and 71 ± 15 min, respectively. The corresponding predicted values were 0·49 h−1, 9·8 g l−1 and 60 min, respectively. Predicted and experimental values were very close, thus confirming the optimized conditions and the relevance of the models proposed.
Industrial starter production practices currently focus on the optimization of biomass and activity at the end of fermentation and on the addition of protective molecules to limit degradation during the freeze-drying process and storage. This approach remains empirical and the fermentation parameters are rarely established to maximize the bacterial activity recovery after freeze-drying. However, optimizing the biomass concentration after the fermentation step often resulted in a decrease in the bacterial activity after the freeze-drying step. The objective of this work was thus to better investigate the relationships between the fermentation conditions (temperature, pH and harvesting time), the biomass concentration and the bacterial resistance to the freeze-drying process. It should be emphasized that conditions close to those found in industry were chosen in terms of composition of the fermentation medium and of the nature of the protective molecules added before freeze-drying. Moreover, to the best of our knowledge, this is the first study to investigate the influence of three fermentation conditions on cell growth and on LAB resistance to each step of the production of freeze-dried cells.
Optimal fermentation conditions led to growth kinetics close to those previously reported for other strains of Lc. lactis (Hofvendahl et al. 1999). Specific growth rates were similar for Lc. lactis ssp. lactis ATCC 19435 cultivated in rich medium (0·62 h−1) and for Lc. lactis ssp. lactis TOMSC161 in simple medium close to industrial conditions (0·49 h−1). Fermentation pH strongly influenced the final biomass concentration, whereas the loss of acidification activity during freeze-drying was influenced by the fermentation temperature and the harvesting time. Increasing the harvesting time from the early to the late stationary phase did not damage the initial quality of the starter (i.e. final biomass concentration after fermentation, cultivability and acidification activity before freeze-drying) but had a strong impact on reducing the loss of acidification activity during freeze-drying. After 6 h of stationary phase, the lower acidification activity loss was obtained at 32°C and corresponded to the optimal temperature for bacterial growth within the experimental domain. Similarly, increasing the fermentation temperature decreased the loss of acidification activity during drying but only for a harvest at the early stationary phase and at 3 h of stationary phase. The positive impact of increasing harvesting time, increasing fermentation temperature or applying a heat shock after fermentation on the freeze-drying resistance of lactic acid bacteria was reported by several authors (Broadbent and Lin 1999; Schwab et al. 2007; Schoug et al. 2008; Hua et al. 2009). These effects may result from physiological changes induced in the bacteria. It was largely reported that modifying fermentation conditions changes physiological responses at the membrane and cytosolic levels and demonstrated that there were correlations between membrane lipid composition of bacteria and their freeze-drying tolerance (Schoug et al. 2008; Hua et al. 2009; Li et al. 2009). In our study, no effect of pH was observed on the resistance of Lc. lactis to freeze-drying and storage. This result differs from some literature reporting that cells grown under acidic conditions improved the freeze-drying survival (Palmfeldt and Hahn-Hagerdal 2000; Hua et al. 2009). For example, Palmfeldt and Hahn-Hagerdal (2000) observed that Lactobacillus reuteri cells grown at pH 5 yielded a higher freeze-drying survival rate than at pH 6. The different genus studied and protective solutions applied can explain these differences. The efficiency of the protective solution used in our study (mixture of sucrose and maltodextrin) could have indeed lowered or hidden the pH effect. However, in accordance with Palmfeldt work, low pH decreased the final biomass concentration and the specific growth rate of Lc. lactis TOMSC161. Consequently, optimum pH must come out from a compromise between biomass production and survival rate after freeze-drying and storage.
The strain Lc. lactis TOMSC161 did not appear to be sensitive to the freezing step but sensitive to the drying step of the freeze-drying process. Low losses of cultivability (<0·6 log CFU ml−1) were also previously reported for Lc. lactis (Broadbent and Lin 1999). The low sensitivity of cocci has been related to their small size and spherical shape resulting in high cell surface area to volume ratio, thus facilitating water efflux from cell during freezing (Mazur 1977; Fonseca et al. 2000). Besides, a cold adaptation during the concentration step (centrifugation at 4°C) could also explain the great survival of Lc. lactis TOMSC161 to the freeze-drying, in accordance with previous work on Lc. lactis (Broadbent and Lin 1999).
Despite the addition of sucrose, an efficient protective molecule, significant loss of acidification activity was observed after rehydration. The removal of the unfrozen water during desorption with the breakage of the hydrogen bonds thus destabilized the biological membrane leading to cell damages. However, better resistance to the freeze-drying process was observed compared with the results obtained by Broadbent and Lin (1999). The higher loss of cultivability could be ascribed to the fact that no protective molecule was added before freeze-drying.
This study highlighted drying and storage at 25°C as the two critical steps during preservation. Storage at 25°C appeared as the most important step inducing considerable cultivability and acidification activity losses. Many degradation reactions may occur during storage at high temperatures, including the Maillard reaction and oxidation. Furthermore, sugar crystallization could occur, thus leading to water release, increasing water activity (Jouppila et al. 1997) and cell destabilization (Passot et al. 2012). However, the high maltodextrin content of the protective solution and the low water content (0·43%) and water activity (0·024) values observed for all freeze-dried starters led to high glass transition temperature (>100°C). Consequently, the molecular mobility and the diffusion controlled degradation reactions were limited by embedding bacteria in a solid glassy matrix (Santivarangkna et al. 2011; Foerst et al. 2012; Tymczyszyn et al. 2012). In this case, the high loss of bacterial cultivability and activity during storage at 25°C could not be explained by Maillard reaction or sugar crystallization but could be ascribed to oxidation reactions. The very low value of water activity of the freeze-dried matrix would enhance the oxidation reaction, especially in the absence of an antioxidant in the protective medium (Ruckold et al. 2001).
This work proved that the Box-Behnken design was useful for optimizing industrial freeze-dried starter production and for improving our knowledge about the influence of fermentation conditions on cell resistance throughout the process with a reduced number of experimental runs. The Box-Behnken design is slightly more efficient for the quadratic model compared with the central composite design that has only 15 runs compared with the 27 runs required for the central composite design. Validation experiments indeed confirmed the success of the process optimization. The optimized fermentation conditions (32°C, pH 6·2 and harvest at 6 h of stationary phase) made it possible to highly reduce acidification activity losses during freeze-drying and to eliminate cellular and acidification activity loss during storage at refrigerated temperature. Lactococcus lactis TOMSC161 is therefore robust enough for industrial processing. In a future study, an integrative approach combining membrane fluidity and fatty acid composition measurements, proteomic and transcriptomic methods in extreme conditions in terms of freeze-drying resistance will lead to a better understanding of the cell mechanisms responsible for bacterial degradation during the critical stages of the process: drying and ambient temperature storage.
This work was supported by the Centre National Interprofessionnel de l'Economie Laitière (Paris, France). The authors thank Savoîcime for providing the reference strain. They authors also thank Professor Marc Danzart (AgroParisTech, Paris, France) for his support and expertise in statistics.