De Man, Rogosa and Sharpe (MRS) broth and agar and peptone water were from Scharlau Microbiology (Scharlab, S.L., Barcelona, Spain). The yeast extract and phosphate buffer were from Oxoid Ltd. (Basingstoke, UK). Sweet whey was a product of a local company (Sopó, Colombia). The composition of the sweet whey (in % by mass) was as follows: protein 11.67, lipids 2.0, carbohydrates 51.64, and ashes 10.9. The maltodextrin and gum arabic were from Shandong WNN Industrial Co., Ltd. (Weifang, PR China). Bile salt was from Sigma-Aldrich, Merck (St. Louis, MO, USA). Phosphorus pentoxide was from Merck (Darmstadt, Germany).

Lactobacillus fermentum K73 (GenBank KP784433.1) was stored at –80 °C with 20% sterilized glycerol as the cryoprotectant in MRS broth (26). The microorganism was grown in MRS broth for 24 h at 37 °C under aerobic conditions before the experiment.

Biomass production was performed in a 1-litre bioreactor (BioFlo 110; New Brunswick Scientific Co., Inc., Edison, NJ, USA) with a workload of 800 mL at 37 °C and agitation at 3354×g for 10 h. The culture medium contained 8% sweet whey and 0.22% yeast extract, and was adjusted to pH=5.5 and sterilized at 121 °C for 15 min. Lactobacillus fermentum K73 was inoculated at 10% (by volume). Cells were counted following fermentation, as described below.

All suspensions used in the model (Table 1) were prepared at a final mass fraction of 40% soluble solids (27). The culture medium containing growing microorganisms (10 h of incubation) was used to hydrate each mixture used in the experimental design (Table 1). The mixtures were homogenized using a magnetic stirrer (1.8K; BT Lab Systems, Geno Technology, Inc, St. Louis, MO, USA) for 30 min at 130 rpm, and the initial cell counts were determinated. Next, the mixtures were incubated for 24 h at (37±2) °C. After incubation, cells were counted, and the bacterial change cycles under simulated gastrointestinal conditions (acid pH and bile salts) were evaluated. All experiments were performed in triplicate. The final cell count after homogenization was (10.29±0.18) log CFU/mL.

The optimal mixture experimental design for obtaining the proportion of carrier material that improves bacterial survival under simulated gastrointestinal conditions was determined using Design Expert v. 8.0.7.1 (28). The design examined 12 mixtures (runs), with mixtures 2, 3, 9, 10 and 15 using one repetition (runs 16, 8, 14, 13 and 17, respectively) (Table 1), thus evaluating total of 17 runs. Maltodextrin, gum arabic and sweet whey were selected as carrier materials (4, 22, 29). The bacterial change cycles after incubation (log CFU/mL) with the mixture proposed by the design and after simulated gastrointestinal conditions (log CFU/mL) were the response variables.

The selected optimal mixture enabled the microorganisms to tolerate the simulated gastrointestinal conditions and showed a desirability equal or close to 1 (30). The criterion of desirability is a general approach in which the value of each response variable is transformed to a measurement ranging from 0–1; values close to 1 represent maximization processes, while values close to 0 represent minimization processes (30, 31).

For the experimental design of the drying, two factors were considered: (i) temperature of the outlet air, (90±3)–(110±3) °C, which was controlled by regulating the inlet flow, and (ii) atomizing air pressure, 0.08–0.15 MPa. The inlet air temperature for each run varied between 190 and 200 °C. The optimal mixture was a feed containing 40% total solids in the pilot-scale spray-dryer (GEA Process Engineering, Mobile Minor^TM, GEA Niro, Düsseldorf, Germany) for drying. The equipment was operated with a pneumatic co-current two-fluid nozzle as atomizer system (diameter of 1 mm) with an evaporation capacity of 5 L/h. The response variables to be evaluated were: bacterial change cycles after drying (log CFU/g), bacterial change cycles (log CFU/g) after exposure of the capsules to simulated gastrointestinal conditions (acid pH and bile salts), physicochemical properties, and some morphometric features. The design included 12 runs with five replicates (run 1=16, run 4=9, run 7=13, and run 8=10=14) (Table 2). Analysis of variance (ANOVA), the adjusted determination coefficients, and contour plot generation were performed using Design Expert v. 8.0.7.1 (28). The conditions that maximized the survival of L. fermentum K73 after drying and minimized the effects on cell viability after exposure to simulated gastrointestinal tract conditions, with a desirability equal or close to 1, were selected. The culture medium with growing microorganisms was dried under the optimized spray drying conditions to evaluate its function as carrier during spray drying. Cells were counted before and after drying and the results are in Table 2.

The cell count of L. fermentum K73 in all proposed experiments was determined by plate counting in MRS agar after cultivation at (37±2) °C for 24 h under aerobic conditions (32). Serial 1:9 dilutions in peptone water (0.1%) were prepared. To count encapsulated cells, the capsule was dissolved (29). First, 0.1 g of powder was added to 9.9 mL sterile phosphate buffer (0.1 M, pH=7.0) and then the capsule was hydrated for 15 min followed by vortexing for 10 min. The bacterial change cycles were determined as:

Thus, the bacterial change cycles could show positive or negative values.

The simulated gastrointestinal conditions were prepared using MRS broth with some modifications. The MRS broth was adjusted to pH=2.0 with 6 M HCl (33) to evaluate the tolerance to gastric pH. The resistance to bile salt was evaluated by supplementing the media with 0.3% bile salt (34, 35). Both experiments were conducted by inoculating 100 µL of the culture medium or mixture or 0.1 g of powder in modified MRS broth for 2 h at 37 °C. The initial and final cell counts were considered to determine the cell viability and bacterial change cycles.

The morphology of the culture medium powder and the powder obtained under optimal conditions were observable by scanning electron microscopy (SEM, Phenom G2 Pro; Eindhoven, the Netherlands). Powders were dried in phosphorus pentoxide for 2 days and fixed on carbon tape (Ted 57; Ted Pella, Inc., Redding, CA, USA). SEM was conducted at an acce-lerating voltage of 10 kV at 500× and 1500× magnification.

The morphometric features were determined using the SEM images at a 500× and 1500× magnification (40). The 500× images were converted to grey scale with brightness values between 0 and 106 and saved as bitmap files (41). These files were then converted to binary images (black and white) using ImageJ v. 1.34 (42). The tool “Analyze particles” from ImageJ was used to calculate the particle area, this value served to estimate the mean diameter (d/μm) of the particle.

The shifting differential box-counting method (SDBC plugin for the ImageJ software) was used to determine the fractal dimension texture (FDt). The 1500× magnification images were converted to bitmap files and cropped to obtain a square area of 22.86 μm×18.29 μm. The FDt was calculated using this file (40).

Culture medium and powder obtained under optimal conditions were packed into Flex Up metallic bags (140 µm; Alico SA, Bucaramanga, Colombia). The bags were vacuum-closed (0.0009 MPa, C200; Multivac, Barcelona, Spain) and stored at 4, 25 and 37 °C for 36 days (43). Cell count and water activity in the bag content were determined every 5 days.

The plot of relative viability (log N_t/N₀) versus storage time was fitted to a first order equation (44):log(N_t/N₀)=k_m · t /2/where N_t is the total viability of bacteria at time t (day) during storage, N₀ is the viability of bacteria at time zero, and k_m is the cellular viability loss specific rate (day^–1) at the three evaluated temperatures.

The Napierian logarithm of the moisture content and water activity (a_w) values were plotted versus storage time. The values were fitted to a linear model:ln(w or a_w)=A+k_{w or aw} · t /3/where ln(w) is the Napierian logarithm of moisture content, ln(a_w) is the Napierian logarithm of water activity, k_x is the constant of moisture content change rate, k_w is the constant of water activity change rate, A is the intercept of the curve and t is time of storage. The effect of storage temperature on k_m was fitted to the Arrhenius equation (45):

where T_ref is the storage temperature, E_a is the activation energy (J/mol), R is the universal gas constant (J/(mol·K)) and k_m is the cellular viability loss specific rate (day^–1).

The values from Eq. 3 were used in Eq. 4 to predict the storage duration under the three study temperatures (45):t(storage)=[(ln(E_a)–ln(N(bacteria))]/[exp(A_a+k_y/T_ref)] /5/where N(bacteria)=10⁶ CFU/g is the amount of probiotic in a functional food before ingestion by the final consumer (46), and A_a is the intercept of the curve ln(k_m) versus (1/T_ref) (R²= 0.965 for culture medium powder curve and R²=0.981 for optimal condition powder curve).

One stage of probiotic microencapsulation is mixing the microorganism with the carrier material (47). In this study, the microorganism grew on an average (1.25±0.03) log CFU/mL in all suspensions, and thus, the model was not significant (p-value=0.1772, Table 1) for the response variable bacterial change cycles after the mixting. This may be because lactose in the whey from the carrier material was used as a carbon source (6) and small peptides or amino acids in the medium are delivered to the culture medium as metabolizable substrates (51). Liong and Shah (52) used maltodextrin in culture medium. They found that Lactobacillus casei ASCC 292 metabolized maltodextrin by fermentation, which enhanced the production of propionic acid. Thus, maltodextrin can be used as a carbon source for L. fermentum K73 in the carrier material. Using maltodextrin in the powder may have advantageous effects on taste according to in vivo assays. Olano-Martin et al. (53) used a three-stage continuous culture cascade system that simulates different regions of the large intestine containing human gut microorganisms. They reported that after incubation with maltodextrin as a carbohydrate source, the population of bifidobacteria and lactobacilli increased. Therefore, maltodextrin may have probiotic effects. Gum arabic is a dietary fibre that improves the growth of bifidobacteria (20, 54) and Enterococcus faecium NCIMB 30183 (20), but does not have growth-stimulating effects on L. paracasei NFBC338 (19). In contrast to the results of Desmond et al. (19), L. fermentum K73 grew for 2.28 and 2.29 cycles (runs 15 and 17 respectively, Table 1). This suggests that the microorganism uses gum arabic and should be further examined.

Additionally, the mass fraction of maltodextrin, gum arabic, and sweet whey influenced the survival of L. fermentum K73 under simulated gastrointestinal conditions. The response of bacterial change cycles to gastric pH conditions was fitted to a cubic mixture model, and the response of bacterial change cycles to the presence of bile salt was fitted to a quadratic model (Table 3).

The models were highly significant (p<0.0001) with a non-significant lack of fit (p=0.2729 and 0.1283, respectively). The regression coefficients and significance test results are shown in Table 3. The combination of components maltodextrin+gum arabic, maltodextrin+sweet whey, and the third-order interaction between the compounds had a significant effect (p<0.0001) on the tolerance of the microorganism to an in vitro simulated gastric pH environment. Additionally, the interaction between maltodextrin+whey (p=0.0001) and gum arabic+whey (p=0.0003) influenced the survival of L. fermentum K73 in the medium simulating the concentration of bile salt in the small intestine. The regression equations of the model with the interactions between maltodextrin, gum arabic and whey (C) are shown in Table 3.

Fig. 1 shows the simplex plots for the response variables evaluated under in vitro simulation of gastrointestinal conditions. Fig. 1a shows that the lowest difference between the final and initial cell count under gastric pH conditions was observable in the binary mixtures of maltodextrin+whey and maltodextrin+gum arabic. However, when analyzing the estimated coefficients of the three factors (Table 3), the results showed that the maltodextrin and sweet whey helped the survival of the microorganism under gastric pH conditions. This effect was largely observed when estimating second-order effects found on the side of the simplex between vertices A and C. Fig. 1b shows, in apex A of the simplex, the strong and positive effect of maltodextrin and whey on the microorganism under bile salt conditions, which agrees with the coefficient analysis and regression coefficient shown in Table 3.

The model showed that the interactions between maltodextrin and sweet whey had a protective effect on the microorganism under gastrointestinal conditions (after incubation in carrier material and during spray drying). This can be attributed to the abilities of maltodextrin and whey lactose to bind to polar residues of proteins and maintain the integrity of cell membranes (55). However, the mixture reconstituted with culture medium, which contains denatured whey following sterilization, exposes a higher number of functional groups that interact with the non-denatured whey and maltodextrin, forming a strong crosslink between the compounds. Its materials also interact with cell membrane components, generating a viscous layer on the microorganism and reducing the mobility of water across the membrane. This mixture protects the phospholipid of the cell membrane during spray drying (9) and prevents intracellular acidification by HCl and colic acid, which are a product of the hydrolysis of sodium taurocholate and sodium glycolate, and thus maintains the viability of microorganisms in the carrier material (43, 56).

Desirability analysis determined that –0.19 log CFU/mL indicated the decrease of the bacterial change cycles at gastric pH; in the presence of bile salts, the increase in the cell count was 0.02 log CFU/mL, with corresponding mixture ratio of maltodextrin/whey 0.61:0.39 for a desirability of 0.97. The results were corroborated experimentally using the optimal mixture. The decrease in population at gastric pH was by (–0.12±0.04) log CFU/mL, and the increase in population in the presence of bile salt was by (0.15±0.05) log CFU/mL. These results largely agree with those of the model predictions. Thus, this mixture was used in the spray drying experiments.

After selecting the carrier material, the spray drying using the optimal mixture was conducted. The RSM design was performed to determine the drying conditions (outlet temperature and atomizing pressure) that improved the protective effect of the selected mixture.

ANOVA of the experimental design for the bacterial change cycles after drying showed significant results with p<0.0001 and a determination coefficient R²=0.95. The linear effect of outlet temperature, quadratic effect of the outlet temperature and quadratic effect of atomizing pressure were significant within the model (p≤0.05).

Fig. 2 shows the effects of outlet temperature and atomizing pressure on the capacity of the microorganism to tolerate the drying process and their influence on probiotic properties. Fig. 2a shows that viability decreased as the outlet temperature increased and atomizing pressure decreased. As outlet temperature increased, the microorganism viability decreased, which was related to cell damage caused by DNA and RNA denaturation, ribosome damage, cytoplasm dehydration and cell membrane damage. This depends on the speed at which heat and mass transfer occur from the inside to the outside of the cell (36, 47, 57). This may explain the highest linear effect (F-value=149.04) of the outlet temperature on the viability of L. fermentum K73 after drying. Furthermore, the low pressure decreased cell viability, possibly because the atomizing pressure directly affected capsule flexibility, causing more resistance of the carrier material mixture to the migration of water, both in the cell and the capsule itself (58).

The results showed that under all drying conditions, the cell count decreased when the powder was exposed to simulated gastrointestinal conditions. The bacterial change cycles under gastric pH and bile salt conditions were fitted to cubic models with p-values of 0.013 and 0.0008 and R²=0.92 and 0.97, respectively. In both models, the quadratic effect of atomizing pressure (F-value_{gastric pH conditions}=26.4, F-value_{bile salt conditions}=37.6) was the most influential value on the capacity to tolerate gastrointestinal conditions, followed by the quadratic effect of outlet temperature (F-value_{gastric pH conditions}=20.66, F-value_{bile salt conditions}=29.52). The three types of interaction between the outlet temperature and atomizing pressure influenced the tolerance to gastric pH. In contrast, the outlet temperature and its interaction with the quadratic effect of atomizing pressure influenced the capacity of L. fermentum K73 to tolerate bile salt.

The contour diagram in Fig. 2b shows that at an atomizing pressure between 0.13 and 0.15 MPa and outlet temperature close to 90 °C, the microorganism was able to tolerate gastric pH. In contrast, the contour diagram in Fig. 2c shows that at a pressure between 0.08 and 0.12 MPa and outlet temperature of 110 °C the probiotic was able to tolerate bile salt. This agrees with the coefficients and magnitudes expressed in the regression equations in Table 4. Therefore, after drying, the capsule protects the microorganism from the gastrointestinal conditions. The protection may be attributed to physicochemical properties, such as crosslinking between whey proteins and maltodextrin. The increased outlet temperature and atomizing pressure result in the formation of a compact powder, and thus, the microcapsule membrane does not facilitate diffusion of the substances (14) such as HCl or colic acid into the microcapsule, with the microcapsule acting as a physical barrier. Therefore, these conditions enhanced the survival of the microorganism under simulated gastrointestinal conditions, although the outlet temperature of 110 °C injured the cell membrane, as described above.

The optimal conditions for spray drying were: atomizing pressure of 0.117 MPa, outlet temperature of 90.8 °C, and inlet temperature of 175 °C (Table 2), with a desirability of 0.999. The experimental data validated the model and showed a decrease of (–2±0.2) log CFU/g after drying or (–0.61±0.08) log CFU/g after exposure to in vitro conditions of gastric pH, and decrease of (–0.2±0.0) log CFU/g after the addition of bile salt.

The culture medium was dried under optimal drying conditions and examined if it could act as a carrier material. Table 2 shows the bacterial changes after drying and after the exposure of the microcapsules to simulated gastrointestinal conditions in vitro when the culture medium was used as the carrier material. Compared with the powder obtained under optimal conditions, the culture medium powder showed the highest viability after drying (–0.96±0.04 log CFU/g) and under simulated gastrointestinal conditions (gastric pH: (–0.58±0.08) log CFU/g, bile salt: (–0.67±0.04) log CFU/g). The low content of solids (8%) in the culture medium form a feed liquid with low viscosity, which at the beginning of spraying in the drier may increase the retention time and expose the microorganism to a higher temperature in the drying chamber (59). However, this common interpretation contrasts the results of our study. The most likely explanation of the successful use of the culture medium as a carrier is as follows: the hydrophobic and sulfhydryl groups are exposed because whey proteins from the culture medium are denatured by sterilization, and these groups begin irreversible aggregation (60), and thus the culture medium is a heterogeneous feed liquid because agglomerates have dense and irregular structures (Fig. 3) (61). When the culture medium containing microorganisms was sprayed in the drier, the inlet and outlet temperature and atomizing pressure removed the water from the feed liquid (culture medium) and compacted the denatured whey aggregates, giving the microcapsules a rigid surface (62); however, the temperature in the drying chamber was not sufficient to evaporate the water (92% wet basis) and the core of the microcapsules had a high moisture content ((12.7±0.2) % wet basis, Table 2). Çabuk and Harsa (63) also reported this high moisture content of >10% (wet basis), with high survival rates of L. acidophilus NRRL B-4495, and they concluded that it is not possible to establish a correlation only based on the moisture content. However, the capsules with high moisture content may create a microenvironment in which the microorganism can repair the damage to the cell membrane after spray drying. This hypothesis requires further analysis. Additionally, the whey proteins have a buffering effect at gastric pH to improve microorganism survival (64). This may cause low bacterial change cycles at acidic pH.

The powder obtained under optimal drying conditions and culture medium powder (Table 2) were characterized based on the water activity (a_w), humidity, solubility, hygroscopicity and bulk density. The a_w and moisture content were 0.09–0.19 and 0.66–1.0%, respectively (Table 2). The values of a_w were adjusted to a linear model, with p-value<0.0001 and R²=0.81 (Table 5). The outlet temperature showed a statistically significant effect (p<0.0001) on a_w. The moisture content was adjusted to a cubic model with p=0.0043 and R²=0.94 (Table 5). The outlet temperature and cubic effect of the outlet temperature had the greatest influence on the moisture content in the microcapsules. Therefore, higher outlet (110 °C) and inlet temperature (200 °C) produced a powder with lower moisture content than under the other evaluated temperatures, thereby suggesting temperature to be the most important factor (14), which agrees with the results of the contour plot in Fig. 4. Figs. 4a and 4b show that the lowest values of a_w and moisture content were observed at 110 °C, which is consistent with the coefficients shown in the linear equation and regression equation (Table 5).

The solubility measurement of the powders was adjusted to a quadratic model (Table 5), as shown by ANOVA, with p=0.003 and R²=0.80. The outlet temperature and its quadratic effect influenced the ability to reconstitute the powder. In accordance with the regression coefficients, the longest time to solubilize the sample occurred at temperatures close to 90 °C (Fig. 4c).

The hygroscopicity values were adjusted to a cubic model (p=0.0008) with an R²=0.97 (Table 5). The first-order interaction between the outlet temperature and atomizing pressure, quadratic effect of the outlet temperature and atomizing pressure, second-order interaction between the outlet temperature and atomizing pressure squared, and cubic effect of both factors were significant for the model. Fig. 4d shows that at temperatures of 105–110 °C, the hygroscopicity of the capsules increased. The optimal hygroscopicity for spray drying was (15.6±0.5) g/100 g.

The data obtained for bulk density were adjusted to a cubic model with p=0.0050 and p<0.0001 and R²=0.94 and 0.99, respectively. For both response variables, the quadratic and cubic effects of the outlet temperature showed significant effects (p≤0.05) (Table 5). The bulk density results showed that higher outlet and inlet temperatures decreased the density of powders (Fig. 4e) because of more rapid drying of the particle. This fixed the particle dimensions before most of the water content had evaporated, which agrees with the results of a previous study (65).

The final product of the spray drying under optimal conditions was characterized according to the physicochemical properties. In this study, we obtained an optimal product for the food industry. The moisture content and a_w were <4% and <0.2, respectively. The isosmotic conditions generated between the wall and cell membrane and the water monolayer that maintains the shape and structure of enzymes and proteins are maintained (66). In contrast, a low moisture content and a_w prevent the growth of the accompanying microorganisms in the powder, thereby extending its lifetime (67).

The time to complete the dissolution in water, including the time of (126.99±5.92) s obtained from the dried powder under optimal conditions, was comparable to those reported by Fritzen-Freire et al. (37) and Pinto et al. (38), who suggested that during spray drying, the hydrophilic groups in lactose and whey proteins are more exposed. Moreover, the absence of fat on the surfaces of the particles makes the powder more hydrophilic, causing its water solubility to increase.

The hygroscopicity of the samples, including the microcapsules obtained under the optimal spray drying conditions, (15.6±0.5) g/100 g, was comparable to those of microcapsules containing 20–30% maltodextrin, as reported by Tonon et al. (68), who attributed this property to the high concentration of maltodextrin, which has low hygroscopicity, confirming its effectiveness for use as a carrier material.

The physicochemical properties of the culture medium powder were compared with those of the powder obtained under optimal conditions. The time to complete dissolution of the culture medium powder ((166.66±29.2) s) was higher than that of the optimized powder ((127.0±5.9) s). This can be explained by the hydrophobic properties of the denatured whey proteins, which decrease microparticle solubility. The difference between the hygroscopicity of culture medium powder and that obtained under optimal conditions ((36.57±0.05) and (37.65±0.13) g/100 g, respectively) was low. The hygroscopicity value of the culture medium powder is attributed to the greater number of hydrophobic amino acid residues from the denatured whey proteins, which affect water adsorption (37). The bulk density of the powder obtained under optimal conditions was (0.55±0.00) g/mL, whereas that of the culture medium powder was (0.29±0.02) g/mL. Felix et al. (14) reported that adding maltodextrin to the carrier material increased the final bulk density.

Finally, the microencapsulation process proposed in this study was performed using pilot-scale equipment. Performing this type of study using robust equipment would enable the development of microencapsulation processes easily adaptable to the food industry, without the need for large changes in the drying technology, preparation of the carrier material and preparation of the biomass acquisition process.

ANOVA revealed no significant differences between the diameter and FDt values (p=0.761 and 0.399, respectively) (Table 2). The powder obtained under optimal drying conditions showed a spherical shape with few fractures and concavities on the surface (Fig. 3a), with a smoother surface than the culture medium powder. The culture medium powder had rougher particles (Fig. 3b). Adding maltodextrin decreases the roughness because the maltodextrin molecules join with denatured whey proteins, creating stronger steric hindrance against protein aggregation (69) and producing particles with a smoother surface, as observable in this study.

The diameter values of the optimized powder and culture medium powder were (<13.7±0.3) μm (Table 2). In industrial production, capsule sizes below 100 μm prevent gritty or sandy undesirable textural properties in the food product (63). Therefore, the powder product obtained in this study had the correct size for use in functional foods.

Both powder obtained under optimal conditions and culture medium powder were evaluated at 4, 25 and 37 °C for 36 days to determine the cell count, moisture content and water activity. Fig. 5a shows L. fermentum K73 behaviour at 4, 25 and 37 °C spray-dried with and without carrier material. The results showed that k_m, k_x and k_w increased with increasing storage temperature and with the use of maltodextrin/sweet whey (0.61:0.39) (Table 6). The storage of powders showed the highest viability losses at 37 °C (Fig. 5a), while lower k_m occurred at 4 °C (Table 6).

The viability of the microorganisms dried using culture medium as the carrier material was similar at 4 and 25 °C (Fig. 5a) and showed similar k_m values (Table 6), although the k_{m25 ºC} =–0.04 days^–1 was lower than k_{m4 ºC}=–0.04 days^–1. Additionally, the predicted shelf life (Table 6) showed that the viability of microorganisms was better at 25 °C (57.51 days) than at 4 °C (54.89 days). The observed increase in viability at 25 °C may have been because the moisture content was high ((12.7±0.2) %) in the culture medium capsules but stable during storage (Fig. 5b). This stability may help maintain L. fermentum K73 viability at 25 °C, as described above. The high moisture content may help microorganisms repair the injured areas on the cell membrane.

In contrast, the k_m values of the viability loss kinetics (Fig. 5a) of L. fermentum K73 with carrier material increased with storage temperature (Table 6). The moisture content of powder obtained under optimal conditions was low (<8.2±0.5, wet basis, Fig. 5b) and constant at all evaluated temperatures, while the a_w values were high (<0.370±0.002), Fig. 5c. Therefore, the viability loss and short shelf life (23 days at 4 °C, Table 6) may be attributed to the a_w values. It has been suggested that a_w values greater than 0.25 considerably increase the k_m of lactic acid bacteria, such as L. fermentum K73, likely by stimulating its metabolism that is affected by high molecular mobility in the carrier material (4).

Shelf life results showed that using culture medium as a carrier material protected L. fermentum K73 not only from the spray drying conditions, but also from the storage temperatures. The best conditions were at 25 °C, followed by those at 4 °C. The moisture content may determine the viability of L. fermentum K73, but these data must be interpreted with caution. This may occur because of the repair of the damage to the cell membrane at high moisture content after drying, but further studies are needed to confirm this.