Introduction

During the last few decades, a wide variety of novel products containing probiotics have been developed and marketed in the European countries. They can be grouped into three main categories: (1) conventional foods such as fermented products containing probiotic bacteria, consumed primarily for nutritional purpose; (2) food supplements or food formulations mostly used as a vehicle for probiotic bacteria; and (3) dietary supplements in the form of capsules (Fasoli et al. 2003; Simmering and Blaut 2001).

A number of definitions of the term ‘probiotic’ have been used over the years, but the one derived by the Food and Agriculture Organization of the United Nations/World Health Organization (FAO/WHO 2002) and endorsed by the International Scientific Association for Probiotics and Prebiotics (Reid et al. 2003) best exemplifies the breadth and scope of probiotics as they are known today: “Live microorganisms, which when administered in adequate amounts, confer a health benefit on the host” (Reid 2006). The most commonly used organisms in probiotic preparations are lactic acid bacteria (LAB) and Bifidobacterium, because LAB are presumed to impart beneficial effects on the host such as improving intestinal tract health, enhancing the host’s immune system, reducing the symptoms of lactose intolerance, and reducing the risk of certain cancers (Parvez et al. 2006).

Viability is generally considered a prerequisite for optimal probiotic functionality (Maukonen et al. 2006). Many studies have shown that the viability of bacteria is not a simple question of cells being dead or alive (Dodd et al. 1997; Bloomfield et al. 1998; Kell et al. 1998; Barer and Harwood 1999; Colwell 2000; Yamamoto 2000; Nystrom 2001; Bogovič Matijašić and Rogelj 2006).

Traditionally, plate counting has been the method of choice for viability determination, but there are obvious disadvantages, including the relatively long times needed for the growth of colonies. The viable plate count method can be frustrated by clumping, inhibition by neighbouring cells, and composition of the growth media used (Breeuwer and Abee 2000). In addition, many possible probiotic effects of bacteria depend on activity rather than culturability, and even dead cells can have some probiotic effect, such as immunomodulation (Ouwehand et al. 2000). Therefore, obtaining information about all individual bacteria and their physiological status is relevant (Bunthof and Abee 2002). Kell et al. (1998) have suggested four terms to describe the different stages of microorganisms: viable (active and readily culturable), dormant (inactive but ultimately culturable), active but nonculturable, and dead (inactive and nonculturable; Lahtinen et al. 2005). Bunthof and Abee (2002) reported that such dormant population might exist in probiotic products and dairy starters. A similar stage of bifidobacteria occurring after exposure to stress conditions was later demonstrated by Ben Amor et al. (2002). Further studies showed that probiotic bacteria may become dormant in fermented products during prolonged storage (Lahtinen et al. 2005, 2006). Therefore, there is an increasing interest in the development of rapid methods for cell viability determination (Breeuwer and Abee 2000; Grattepanche et al. 2005). Although culture-independent molecular methods for identification and quantification of probiotic bacteria have been extensively developed during recent years, they are still not widely used in routine laboratories. Among possible alternatives, quantitative real-time polymerase chain reaction (PCR) as well as fluorescent techniques have a potential to replace conventional enumeration of probiotic bacteria (Furet et al. 2004).

However, at the moment, there are still some limitations to overcome before the introduction of such methods in routine analysis. Real-time PCR does not enable the distinction between DNA arising from dead or live cells; therefore, the DNA from non-viable cells contributes to the result. In contrast, fluorescent techniques such as fluorescent in situ hybridisation, combined with flow cytometry (FCM) using commercial bacterial viability kits, enable the separate counting of live and dead cells, but do not enable species- or strain-specific quantification in mixed preparations.

The most important criterion for distinguishing between viable and irreversibly damaged cells is membrane integrity. Live cells with intact membranes are able to exclude DNA-binding dyes that easily penetrate dead or membrane-compromised cells. This principle is increasingly applied in FCM (Nebe-von-Caron et al. 2000). With fluorescent stains, a dual approach involving the staining of intact cells with one stain followed by the counterstaining of membrane-compromised or all cells with another stain is commonly used (Breeuwer and Abee 2000; Alakomi et al. 2005). In this work, the use of LIVE/DEAD® BacLight™ bacterial viability kit containing SYTO9 and propidium iodide (PI) in combination with FCM was evaluated. SYTO9 stains all cells fluorescent green, while PI penetrates only the cells whose cell membrane has been damaged, staining them red (Boulos et al. 1999).

Recently, treatment with ethidium monoazide (EMA) and subsequent PCR analysis was reported to be an easy-to-use alternative to microscopic or flow cytometric determination of bacterial viability (Rudi et al. 2005; Wang and Levin 2006; Hein et al. 2007; Lee and Levin 2007). Because of the lack of selectivity and overall applicability of EMA, a propidium monoazide (PMA) was introduced, which is more selective in penetrating dead bacterial cells with compromised membrane integrity. PMA is a DNA-intercalating dye with the azide group, which enables covalent binding to DNA upon exposure to bright visible light and, consequently, strongly inhibits PCR amplification. Subjecting a bacterial population to PMA treatment before PCR therefore results in selective amplification of DNA from cells with intact membranes only (Nocker et al. 2007a, b, 2009). The species already tested with PMA-PCR were Escherichia coli 0157:H7, Listeria monocytogenes, Micrococcus luteus, Mycobacterium avium, Pseudomonas syringae, Salmonella typhimurium, Serratia marcescens, Staphylococcus aureus (Nocker et al. 2006), and Clostridium perfringens (Wagner et al. 2008).

In the present study, we assessed the viability of two probiotic strains in a lyophilised form by plate counting, real-time PCR of PMA-treated samples, and differential fluorescent staining in combination with FCM. We examined the suitability of PMA to selectively bind to genomic DNA of membrane-compromised cells from pure bacterial cultures, as well as from lyophilised product, and evaluated the possibility to use PMA in combination with real-time PCR with SYBR Green I chemistry for selective quantification of live Lactobacillus acidophilus and Bifidobacterium animalis ssp. lactis in the lyophilised product. Furthermore, the viability of probiotic bacteria in lyophilised product was followed by all three methods during 3 months storage aiming to evaluate the capability of these methods in supporting stability studies of probiotic products.

Materials and methods

Bacterial strains and growth conditions

Pure strains of L. acidophilus LA-5 and B. animalis ssp. lactis BB-12 were obtained from Chr. Hansen’s Laboratories A/S (Horsholm, Denmark). The strains were cultured overnight at 37°C in de Man, Rogosa, and Sharpe (MRS) broth (Merck, Darmstadt, Germany) with cysteine (0.05% w/v, Merck, Darmstadt, Germany). The viable bacterial counts of individual strains were determined in duplicate by standard plate counting on MRS agar with cysteine after 3 days of anaerobic incubation at 37°C.

Lyophilised probiotic product

Two batches of the lyophilised probiotic product produced at different times were analysed. Microorganisms declared on the label were L. acidophilus LA-5 and B. animalis ssp. lactis BB-12 at a concentration of 109 colony-forming units (CFUs) of each strain per capsule. Apart from bacteria, the probiotic products also contained the following excipients: Beneo synergy (73%), saccharose (11%), dextrose anhydrous (10%), and excipients in traces (referred to in the following as filler).

Determination of bacteria concentration by plate count method

One gram of lyophilised probiotic product was homogenised thoroughly in 99 g of sterile Ringer solution (Merck 1.15525), and further tenfold serially diluted for determination of viable bacteria by standard pour plate procedures. MRS agar with added cysteine hydrochloride (0.05%, Merck), dichloxallin (0.5 mg/l, Merck), and LiCl (0.1%, Merck) was used for bifidobacteria, and MRS agar with clindamycin (0.1 mg/l, Merck) and ciprofloxacin (10 mg/l, Merck) for lactobacilli. The plates were incubated at 37°C for 3 days under anaerobic conditions. The total viable count of LAB was performed using MRS agar with cysteine hydrochloride (0.05%, Merck) under anaerobic incubation for 3 days.

Optimisation of real-time PCR for selective quantification of L. acidophilus LA-5 and B. animalis ssp. lactis BB-12

Sample preparation

The content of three capsules of each product was well resuspended in Ringer solution to prepare 1% (w/w) solution. One millilitre of 1% lyophilised product suspension was centrifuged (3,500 g, 10 min), and DNA was extracted from pellet by two different methods as described below. Samples were three-, five-, and tenfold serially diluted prior to PCR analysis.

DNA extraction method

Two methods of DNA extraction were used: Wizard genomic DNA purification kit (Promega) and MaxwellTM 16 cell tissue DNA purification kit applied with MaxwellTM 16 instrument.

Before application of Wizard kit, 600 µl of ethylenediaminetetraacetic acid (50 mM) with lysozyme (1 mg/100 µl) was added to the cell pellet and incubated for 45 min at 37°C for bacterial cell lysis. Further procedure was conducted according to the manufacturer’s protocol. DNA was finally resuspended in 100 µl of water.

Before isolation by Maxwell system, 400 µl of TE buffer and 100 µl of lysozyme (25 mg/ml) with mutanolysine (10 U/ml) was added to the pellet, which was further resuspended and incubated for 2 h at 37°C for bacterial cell lysis. The whole sample (500 µl) was transferred into the first well of the cartridge and further treated according to manufacturer’s instructions. DNA was finally resuspended in 300 µl of elution buffer with added 1.5 µl RNAse (4 mg/ml).

Quantitative real-time PCR

PCR amplifications were performed in a 25-µl reaction volume, containing Platinum SYBR Green qPCR Super Mix UDG (11733; Invitrogene, Carlsbad, CA, USA), 0.2 µM of each primer, and 5 µl of genomic DNA extract. Primers used are listed in Table 1. The PCR amplification was performed with an MX3000P (Stratagene, La Jolla, CA, USA) instrument. The amplification programme for lactobacilli was 50°C for 2 min and 95°C for 2 min, 35 cycles of 95°C for 30 s, 60°C for 15 s, 72°C for 20 s, and then 95°C for 1 min and 55°C for 30 s. The amplification programme for bifidobacteria was 50°C for 2 min and 95°C for 2 min, 35 cycles of 95°C for 20 s, 56°C for 20 s, 72°C for 30 s, and then 95°C for 1 min and 55°C for 30 s. All samples were subjected also to melting-curve analysis in order to establish the specificity of the amplification.

Table 1 Oligonucleotide primers used in polymerase chain reaction amplification of fragments of 16S rDNA of Lactobacillus acidophilus and Bifidobacterium animalis ssp. lactis
Full size table

Standard curves relating cell numbers to real-time PCR Ct values for B. animalis ssp. lactis BB-12 and L. acidophilus LA-5

For standard curves, bacterial cells of L. acidophilus or B. animalis ssp. lactis were mixed with 1 ml of 1% (w/v) suspension of filler ingredients in order to imitate the original lyophilised product. Eighteen-hour cultures were centrifuged (3,500 g, 5 min), and approximately 1.5 g of pellet was resuspended in 15 ml of Ringer solution. This initial suspension was further diluted in duplicate tenfold serial dilutions for plate counting. In three parallel series of 1 ml, 1:10 dilution of initial suspension was prepared and centrifuged (3,500 g, 10 min). Filler suspension (1% w/v) was added (1 ml) to the cell pellet. Samples were vortexed, centrifuged (3,500 g, 10 min), and pellet further processed by two different DNA extraction protocols. Ten-, five-, or threefold dilution series from target species genomic DNA preparations were amplified by real-time PCR. The correlation between Ct values and CFU/ml was determined by Stratagene MX3000P system’s programme.

PMA treatment and real-time PCR

PMA treatment of product, standard samples, and heat-treated samples

Twenty millimolar stock solution of PMA (phenanthridium, 3-amino-8-azido-5-[3-(diethylmethylammonio)propyl]-6-phenyl dichloride; Biotium, Inc., Hayward, CA, USA) was prepared in 20% dimethyl sulfoxide (Fisher Chemical D128-500) and stored at −20°C in the dark. To 1 ml of product (1% w/w in Ringer solution) or standard samples (in 1% w/v filler suspension) aliquots (final concentration of 50 µM), 2.5 µl of PMA was added in light-transparent 1.5 ml microcentrifuge tubes. Following an incubation period of 5 min in the dark with occasional mixing, samples were light-exposed for 2 min using a 650-W halogen light source (sealed beam lamp, FCW 120 V, 3,200 K). The sample tubes were laid horizontally on ice to avoid excessive heating and placed about 20 cm from the light source. After photoinduced cross-linking, cells were pelleted at 3,500 g for 10 min prior to DNA isolation by the above-described protocols.

In order to test the efficiency of PMA treatment of membrane-compromised bacterial cells, duplicate tubes (9 ml) containing bacteria from pure cultures of L. acidophilus or B. animalis ssp. lactis in 1% (w/v) filler suspension (approximately 2 × 108 CFU/ml) or lyophilised product in Ringer solution (1% w/w) were heated at 95°C in a water bath for 5 min. The heat-treated samples were then cooled to room temperature and subjected to DNA isolation. In addition, the loss of viability was examined by streaking 100 µl of heat-treated samples on the corresponding agar plates followed by incubation at the optimal growth temperature.

LIVE/DEAD BacLight™ bacterial viability assay in combination with FCM

Preparation of samples

Bacterial cultures were grown in MRS broth for 20–24 h (stationary phase), and cells were harvested by centrifugation at 10,000 g for 15 min. The buffer that was used throughout the experiment was Ringer solution. Cells were washed four times and resuspended in Ringer solution. One of two aliquotes of bacterial suspension was exposed to 70% isopropyl alcohol for 1 h in order to permeabilize cell membranes to cause cell death. Different proportions of live and dead cells were mixed to obtain cell suspension containing five different ratios, i.e., 100:0, 75:25, 50:50, 25:75, and 0:100 to create a standard curve.

Fluorescent staining method

The fluorescent stains SYTO9 and PI of LIVE/DEAD BacLight bacterial viability kit (Molecular Probes, Inc., Eugene, OR, USA) were prepared as described by the manufacturer. Double staining with PI and SYTO9 was performed by incubating the samples with 1.50 µM PI and 250.5 nM SYTO9 at room temperature for 15 min. From SYTO9 versus PI plot, populations of live and dead cells were gated for analysis. For instrument adjustment, single-colour controls were used. Non-stained cells were used as a background control. FCM analysis was performed with a FACSCalibur flow cytometer (Becton Dickinson, Inc., Franklin Lakes, NJ, USA) and data analysis software. Microsphere standards with a diameter of 6.0 µm were included in the FCM samples for enumeration of cells (1.0 × 106 beads/sample).

Statistical analysis

The Ct values were automatically generated by the Invitrogene software. One-way analysis of variance (ANOVA) with post-hoc Tukey test was used to compare results obtained by real-time PCR with and without PMA treatment and plate counts at each storage time. A linear regression analysis was performed to evaluate the stability of probiotic product during storage. The SAS programme for windows 6.12 software (SAS Institute, Inc., Cary, NC, USA) was used to determine the statistical significance of results obtained.

Results

PCR reaction optimisation

Considering extraction procedure, two DNA extraction methods were evaluated, namely, Wizard Genomic and Maxwell DNA extraction system. They were compared for extraction efficiency in real-time PCR reaction of PMA-treated and non-treated samples, using Bif-F/Bif-R and LactoR′F/LBFR primer pairs. As can be seen from Table 2, a higher average Ct value corresponding to lower DNA concentration in the sample was detected in the extracts obtained by Wizard protocol. Furthermore, higher standard deviation among simultaneously extracted samples was observed. The DNA extraction procedure with Maxwell RM 16 instrument (Promega) based on MagneSil paramagnetic particles, proved to be more reliable regarding extraction efficiency and precision. It was consequently used in further analyses.

Table 2 Comparison of Wizard Genomic and Maxwell DNA extraction efficiencies, obtained from parallel DNA extracts (A and B) for two batches of probiotic product (I, II), using Bif-F/Bif-R primers
Full size table

The ability of PMA to inhibit amplification of DNA derived from damaged bacterial cells was investigated on bacteria from pure cultures of L. acidophilus or B. animalis ssp. lactis in 1% (w/v) filler suspension (approximately 2 × 108 CFU/ml) and on probiotic product (1% w/w). For information of filler composition, see M and M. Cells were killed by heat treatment of the samples at 95°C for 5 min. The samples were then treated with PMA (50 µM) and subjected to DNA isolation. Results in Fig. 1 show that PMA was effective in preventing real-time PCR amplification of the target sequences of DNA released from heat-killed bacteria in the suspension of filler, since the differences of the Ct values between the samples of heat-treated and non-treated cells were 16.9 and 15.8 for lactobacilli and bifidobacteria, respectively. Similar results were obtained also in the case of probiotic product treated in the same way, where the differences obtained were 17.3 and 15.6 for lactobacilli and bifidobacteria, respectively. It was concluded that filler ingredients did not have a negative impact on the efficiency of the PMA treatment of bacterial cells.

Fig. 1
figure 1

Results of amplification by real-time polymerase chain reaction of DNA from heat-treated (95°C/5 min) or non-treated bacterial cells from pure cultures of Lactobacillus acidophilus LA-5 and Bifidobacterium animalis ssp. lactis BB-12 in 1% (w/v) filler suspension and from probiotic product. Lactobacilli were amplified with LactoR′F/LBFR primers and bifidobacteria with Bif-F/Bif-R primers

Full size image

The two primer pairs used in this study were tested for possible cross reactivity. No amplification of DNA of bifidobacteria was observed in PCR reactions with LactoR′F/LBFR using a previously published amplification programme (Songjinda et al. 2007). The unspecific amplification of lactobacilli by Bif-F/Bif-R primers, however, was observed. Therefore, we had to increase the original annealing temperature from 56°C (Rinttilä et al. 2004) to 60°C.

Following the PCR reaction optimisation, standard curves were constructed from known concentrations (CFU/ml) of viable bacterial cells of L. acidophilus and B. animalis ssp. lactis mixed with the suspension (1%) of capsule’s filler ingredients in order to simulate the product. Aliquots of the standard samples prepared for DNA extractions were plate counted. On the basis of standard curves, reaction efficiencies were determined, and the possible inhibition of PCR reactions was examined. The first two dilutions in each dilution series exhibited PCR inhibitory effect possibly because of too high DNA concentration or the presence of some other interfering substances, and were therefore, excluded from standard curves. The best standard curves’ parameters were obtained by threefold dilution series: R 2 = 0.9940, amplification efficiency 98.5%, and slope −3.359 for bifidobacteria, and R 2 = 0.9790, amplification efficiency 96.9%, and slope −3.399 for lactobacilli (Fig. 2). The investigated samples were treated the same way, giving the best results in fivefold dilution series, while a similar PCR inhibitory effect was observed with the first two dilutions.

Fig. 2
figure 2

Standard curves between CFU/ml of standard samples or diluted product samples and the Ct values detected with the real-time polymerase chain reaction method for (a) Lactobacillus acidophilus LA-5 and (b) Bifidobacterium animalis ssp. lactis BB-12. The concentrations of the tested samples derived from the standard curves refer to 500-fold diluted probiotic product; triangle unknown samples, square standard samples

Full size image

The specificity of PCR reaction was evaluated by the melting-curve analysis carried out in conjunction with each real-time PCR assay showing eventual amplification of unspecific fragments or primer-dimers. For L. acidophilus LA-5 and B. animalis ssp. lactis BB-12, a single, sharp peak was observed, confirming that only a specific PCR product was generated with each pair of primers.

Analysis of probiotic product by real-time PCR with and without PMA treatment

Bacterial counts derived from real-time PCR determination of PMA-treated and non-treated samples from two batches of examined probiotic product were compared to the plate counts and statistically evaluated at each time point in the course of a stability test using one-way ANOVA (Table 3). Bacterial counts (log CFU/g) decreased for less than 0.5 log in 90 days, suggesting that bacteria preserved very well the culturability, bacterial DNA remained intact, and that no cell lysis occurred. The statistical treatment revealed no significant differences (α = 0.05) in all but one situation between the methods, indicating that the methods are relatively comparable for quantification of examined LAB strains under the conditions of this study.

Table 3 The number of microorganisms in two batches of the tested product (A and B) performed by the plate count method, real-time polymerase chain reaction (PCR), and propidium monoazide real-time PCR
Full size table

Analysis of probiotic product by the FCM method

In order to introduce the FCM method, all traces of growth medium before staining were removed by washing the bacterial pellet, staining was optimised, and background was evaluated. Also, the two dye components provided with LIVE/DEAD BacLight viability kit had to be balanced first. The background was defined as a labelled imitation of filler ingredients without bacteria and beads. A threshold for the FCM signal was determined in a way that all particles that were similar in size to or larger than bacterial cells were included, while smaller particles were excluded. The buffer and filler themselves did not produce any background.

The FCM was first tested on the pure cultures of probiotic strains which were alcohol-treated or live (untreated). The dead and live bacterial cells were mixed in different ratios. The discrimination between live and dead probiotic bacteria was satisfactory (Fig. 3). The relationships between added and detected live cells are presented in Fig. 4. The results show that the detected ratio of cells correlated well (R 2 = 0.9946 and 0.9989) with the actual ratio in the suspensions. A nucleic acid double-staining assay thus allowed us to distinguish intact bacteria from membrane-compromised ones. As expected, alcohol-treated bacteria with a compromised membrane were stained red; whereas, cells with an intact membrane stained green. The percentage of permeabilized cells was further determined in the product and evaluated in the course of a 90-day stability study.

Fig. 3
figure 3

The flow cytometry analysis of mixed live and alcohol-killed Bifidobacterium lactis cells showing PI binding versus SYTO9 staining. Different proportions of live and dead cells were mixed to obtain cell suspension containing different ratios, i.e., 0/100, 25/75, 50/50, 75/25, and 100/0 (in percent) to create a standard curve. The percentages inside the quadrants show the measured percentage of injured cells (upper right quadrant) and live cells (lower right quadrant) for each mixture of live and alcohol-killed cells

Full size image
Fig. 4
figure 4

Relationship between known proportions of live bacteria (prepared by mixing live and heat-killed bacterial suspensions) and measured viability. Standard calibration curves for (a) Lactobacillus acidophilus and (b) Bifidobacterium animalis ssp. lactis

Full size image

The results of FCM presented in Table 4 show that approximately 10–15% of the cells in probiotic product were membrane-compromised at the time of first sampling. Only a slight increase of dead cells was found after 30 days. However, after 90 days, more cells were permeabilized (approximately 20–26%), indicating a substantial increase in the ratio of dead cells. Although the results of FCM cannot be directly compared with those of PMA real-time PCR, they have in common that after 90 days, a decrease in intact cell ratio was observed. This could be assigned to the decrease of L. acidophilus as detected by PMA real-time (ca. 0.5 log from the time 0), and not to bifidobacteria.

Table 4 Comparison of live and dead percentage for two batches of probiotic product determined by flow cytometry method
Full size table

Discussion

The main objective of this study was to evaluate the possibility to use PMA in combination with real-time PCR using SYBR Green I chemistry for selective enumeration of viable LAB in lyophilised probiotic product. By use of the PMA live-dead distinction method, we can get some information about cell membrane integrity, which is one of the important criteria for bacterial viability. In addition, the ratio of total live vs dead bacterial cells was assessed by differential staining and fluorescence detection by FCM.

The results of real-time quantification of PMA-treated samples were compared with those obtained by two other quantification methods, namely, FCM and pour plate method. The lyophilised probiotic product was also subjected to an accelerated stability test in order to determine the stability properties of probiotics. The first step towards this goal was a proper design and optimisation of the PMA real-time PCR.

When developing the molecular-based quantification method such as real-time PCR of PMA-treated probiotic bacteria, a number of considerations must be taken into account. In this respect, it is important to evaluate the DNA extraction method, possible interference of formulation constituents, and reagents with fluorescence emissions, as well as PCR reagent suppression and inhibition of enzymatic reaction, effectiveness of PMA reagent to inhibit amplification of DNA from dead cells, and a proper primer selection. The influence of these parameters, including standard preparation and choice of a PCR protocol, on the specificity and efficiency of the PCR reaction was evaluated.

The results obtained during optimisation of the real-time PCR method using pure cultures confirmed the method feasibility for quantitative evaluation of both tested strains in pure cultures. Furthermore, the experiments showed good potentials for PMA to selectively inhibit amplification of DNA derived from heat-killed L. acidophilus LA-5 and B. animalis ssp. lactis BB-12 cells included in the matrix simulating the filler of tested probiotic product, as well as in the product. The optimised method, with and without PMA treatment, was then applied to two batches of the examined probiotic product in order to determine the bacterial content and to evaluate the quality of the product.

To our knowledge, the PMA real-time has not yet been tested for the analysis of lyophilised probiotic products. The matrix of complex samples such as environmental samples or food can negatively influence the efficiency of PMA treatment and DNA isolation; therefore, the evaluation of a particular matrix in this regard is needed. Wagner et al. (2008) for instance applied PMA-based real-time PCR for quantification of bacteria in anaerobic fermentor sludge. They found the treatment of PMA not to be effective enough and speculated that the matrix might inhibit the cross-linking step since the radiation could not penetrate the liquid.

The efficient removal of genomic DNA from dead cells by PMA, however, was described for E. coli and for the representatives of different genera, G+ and G− in pure cultures (Nocker et al. 2006), and recently also for LAB in fermented milk products (García-Cayuela et al. 2009).

In this study, a greater loss of overall membrane integrity of the cells in the probiotic product or reduction in culturability was not observed during the storage by either of the methods applied. The observation that plate counts in general did not differ significantly from the real-time-based results was quite surprising, since we expected at least some viable but not culturable (VBNC) or dormant bacteria in such a product, if not bacteria with compromised membrane or DNA released from lysed cells. Particularly, the manufacturing process usually results in the damage to or death of part of the population. VBNC or dormant bacteria, however, still have intact membrane but are not able to grow on the plates (Lahtinen et al. 2006; Bloomfield et al. 1998). Regarding the effects of PMA treatment, the presence of the membrane-compromised bacteria in the product would have resulted in the higher values of bacterial quantification for the samples which were not treated with PMA than for those treated with PMA. As already mentioned, this was, however, not observed, and the absence of significant differences could not be due to the inefficiency of PMA treatment of the bacteria in the product since the efficiency was confirmed by heat treatment of the product, which resulted in reduced amplification of the DNA isolated from such a sample. We may conclude that the manufacturing process was optimal so that only negligible death or damages of the cells occurred or that the DNA released from dead or membrane-compromised cells was quickly degraded and could not influence the result of quantification by real-time PCR. Masco et al. (2007) reported total bifidobacteria quantification in several probiotic products in a lyophilised form by real-time PCR, using genus-specific primers and the SYBR Green I chemistry. Using LIVE/DEAD flow cytometric assay, they also found a very low ratio of dead cells in the products (1.07 ± 0.25% of dead cells; 95.18 ± 0.96% intact), which is comparable to our results (on average 12% dead and 87% intact).

There may be several other reasons for the above-mentioned observations. To our knowledge, there are no reports addressing this problem, but it is possible that the number of bacteria determined by real-time PCR was underestimated due to different extraction efficiency from the cells of fresh cultures, used for standard samples, and from those lyophilised, which are present in the product. In order to exclude this possibility, the standard curves could be generated from the bacterial cells, lyophilised before DNA extraction. Furthermore, insufficient selectivity of selective agar media and, consequently, overestimated plate counts are also possible. This problem has already been discussed by several researchers (Masco et al. 2005; Van de Casteele et al. 2006). The comparison of plate count and real-time PCR-based enumeration of LAB in six commercial fermented milk products showed that the loss of cultivability during storage was strain dependent (Furet et al. 2004). For representatives of Lactobacillus casei group for instance, the results of real-time PCR quantification did not differ from the results of plate counting, while for most of the strains of Lactobacillus delbrueckii, ten to 100 times lower values were detected by plate counting, and two- to fivefold differences for L. acidophilus in favour of real-time PCR determined concentration levels.

In a recently published study, García-Cayuela et al. (2009) also enumerated L. acidophilus LA-5 and B. animalis ssp. lactis BB-12 by using real-time PCR combined with PMA treatment and plate count method, but in a commercial fermented milk product, during shelf life (28 days) and another 60 days after the expiration. The high bacterial counts of B. lactis observed by PMA real-time PCR as well as by plate count, which remained practically unchanged all 90 days, are in accordance with the observations of our study. The stability of L. acidophilus, however, was much better in our lyophilised product than in the fermented milk, which may be a result of different matrix, conditions, or physiological state of the microbial cells.

The results of our study indicate that PMA real-time PCR and FCM are capable of distinguishing between populations of cells with compromised membranes and live cells present in the matrix of lyophilised probiotic product. In the present study, both methods showed a low ratio of membrane-compromised bacteria in the product at the beginning of storage and only a little reduction of bacterial number during the storage. The main advantage of PMA real-time PCR over the FCM is that the first one enables absolute quantification; whereas, FCM detects a ratio of intact vs total bacteria. Moreover, the PMA real-time PCR is species-specific, while FCM using commercial DEAD/LIVE viability assay enables determination on the level of total bacteria only. Additionally, FCM also has bias due to cell clumping, and the optimisation of the instrument for mixed populations is difficult and tedious. According to the literature, it is not yet possible to differentiate more than four populations of LAB according to fluorescence characteristics. This is usually not sufficient since probiotic products often contain more than four strains. Furthermore, the method may overestimate the number of viable cells compared to traditional CFU counts in the case of stressed cells (Pedersen 2008).

Viability is not easily defined in terms of a single physiological or morphological parameter (Barer and Harwood 1999; Bloomfield et al. 1998). Even viable cells may be active or inactive with respect to metabolic activities such as substrate uptake, conversion, and respiration, and still, there is no alternative to determine the active fraction of cells only. The functionality of dormant probiotic cells in probiotic products for example has not been studied yet, and it is not known whether they are able to contribute to the proposed health effects. Also, membrane integrity is no proof of activity. Under certain conditions, bacteria having compromised membranes may recover and reproduce, even though such bacteria may be scored as dead by FCM assay and/or by PMA real-time PCR.

In conclusion, because of the complexity of the bacterial population in probiotic products, including different physiological states of bacteria, the combination of different methods would represent a significant improvement of the analysis of probiotics content. The treatment of samples with PMA, followed with real-time PCR analysis, as presented in this study, appears a promising approach for routine monitoring of quality of a probiotic product examined, as well as for stability studies.