Visualized analysis of cellular fatty acid profiles of Vibrio parahaemolyticus strains under cold stress

Authors


Abstract

Vibrio parahaemolyticus is a common foodborne bacterial pathogen, which survives in cold environments and is sometimes difficult to culture. Fatty acid analysis under cold stress was conducted for several V. parahaemolyticus strains using gas chromatography/mass spectrometry, and the results were compared with those of the controls. All the fatty acid profiles obtained were visualized by multidimensional scaling (MDS) and self-organized map (SOM). It was observed that the fatty acid profiles of V. parahaemolyticus substantially changed under cold stress. The percentage of methyl palmitate remarkably decreased and that of methyl palmitoleate (except for two strains) and methyl oleate increased. These findings demonstrate the role of fatty acids in cold stress. The changes in the fatty acid profiles illustrated by MDS and SOM could differentiate strains under cold stress from the controls and can potentially lead to a method of detecting injured cold-stressed V. parahaemolyticus.

Introduction

Vibrio parahaemolyticus lives in brackish water and can contaminate seafood, causing gastrointestinal illness in humans. In minimally processed foods, bacteria may encounter stress conditions (Abee & Wouters, 1999), such as cold shock, heat shock, weak acids, starvation, high osmolarity, or high hydrostatic pressure. Although these stress conditions are considered to hinder bacterial proliferation (Chiang et al., 2008), they can also result in injured stressed bacteria that are difficult to detect by classical culture method or can enhance the virulence and resistance of the bacteria to subsequent processing conditions (Weichart & Kjelleberg, 1996; Bang & Drake, 2002; Lin et al., 2004), known as stress hardening. For example, cold shock has been reported to reduce the susceptibility of V. parahaemolyticus to disinfectants (Lin et al., 2013) and enable the bacterial cells to survive better in the subsequent exposure to low temperatures and crystal violet applications (Lin et al., 2004). In addition, V. parahaemolyticus can be induced into a viable but nonculturable (VBNC) state by starvation at low temperatures (Jiang & Chai, 1996; Mizunoe et al., 2000), and the cold-induced VBNC cells have been observed to be highly resistant to thermal, low salinity, or acid inactivation (Wong & Wang, 2004). As a result of their high survival ability along with high virulence, the VBNC cells are a threat to public health (Wong et al., 2004); thus, stressed V. parahaemolyticus poses a potentially higher risk to food safety.

Cold stress is perhaps the most common environmental stress condition during food processing and storage. Fatty acid composition plays an important role in the process of resistance to stress conditions (Chiang et al., 2005) and is a characteristic for the identification of bacteria, particularly bacterial strains (Tighe et al., 2000); however, there is relatively limited literature available on the fatty acid composition of V. parahaemolyticus under cold stress. In addition, some researchers have studied the influence of environmental factors on the changes in the bacterial fatty acid profiles (Hamamoto et al., 1994; Fouchard et al., 2005) and have proposed that the fatty acid profiles can be used to monitor the impact of growth conditions on bacteria, because the composition of fatty acids is closely linked to the growth environment (Fouchard et al., 2005). In the present study, we analyzed the fatty acid compositions of several strains of V. parahaemolyticus under cold stress in comparison with those of their counterparts as controls and then visualized these samples by classical multidimensional scaling (MDS) and accomplished unsupervised classification of these samples by self-organized map (SOM). Thus, this study on the characteristics of the fatty acid profiles of V. parahaemolyticus under cold stress could help to explain the mechanism of bacterial resistance to cold stress and contributes to the identification of cold-stressed injured V. parahaemolyticus cells which may be difficult to culture.

Materials and methods

Cultures

Four V. parahaemolyticus strains were isolated from seafood during our routine work. All the V. parahaemolyticus strains were identified by Vitek 2 Compact GN systems (bioMérieux, France), and the presence of DNA gyrase subunit B gene (gyrB) (Cai et al., 2006), transmembrane regulatory protein gene (toxR) (Takahashi et al., 2005), thermostable direct hemolysin gene (tdh) (Blackstone et al., 2003), and thermolabile hemolysin gene (tlh) (Ward & Bej, 2006) had been previously detected by real-time PCR. Some information and characteristics of the V. parahaemolyticus strains used in this study are presented in Table 1.

Table 1. Strains of Vibrio parahaemolyticus used in this study
StrainsSourcesProbabilties of results from Vitek 2 compact (%)Genes
gyrB toxR tlh tdh
Vp6Frozen cuttlefish96+++
Vp15Live oyster94++++
Vp20Frozen scallop99+++
Vp26Frozen fish99+++

Preparation of cold-stressed and control cells

The stock cultures of all the bacteria were maintained at −70 °C in physiological saline solution with 15% glycerol. A total of 10 μL of the stock culture of each strain were streaked onto the trypticase soy-yeast extract agar (TSA-YE; Difco, Becton Dickinson) plate and incubated overnight at 37 °C. For each strain, a single colony was picked from the TSA-YE plate and then streaked on a TSA-YE slant. These slants were incubated at 37 °C for 24 h, and c. 0.04 g of pure culture on each slant was collected and subjected to fatty acid analysis. Furthermore, the remaining cultures on the slants were placed in a refrigerator at 4 °C for cold treatment. After 1 month, these slants were taken out, and the cells were subjected to fatty acid analysis. All the fatty acid analyses were performed in triplicate strictly according to the following procedure.

Analysis of cellular fatty acids

The cellular fatty acids were extracted, and fatty acid methyl esters (FAMEs) were derived according to the procedure described previously (Sasser et al., 2005). In the present study, FAMEs were analyzed by GC/mass spectrometer (MS) using Agilent 5975C Series Gas Chromatograph/Mass Selective Detector Systems complemented by 7890A GC. For all the cultures, the procedure for the analysis of FAMEs was implemented as uniformly as possible. The FAME mixture standards were Nestle 37 (Nu-Chek Prep, INC.) containing 37 FAMEs (C4-C22). The FAMEs were separated in an Agilent CP-Sil 88 capillary column (100 m × 0.25 mm × 0.2 μm, part number: CP 7489) using high-purity nitrogen as the carrier gas.

The settings of GC analysis were as follows: two stages of increase in the column temperature (the first was an increase from 170 to 260 °C at a rate of 5 °C min−1, and the second was an increase to 310 °C at a rate of 40 °C min−1, holding for 1.5 min), injector temperature of 250 °C, detector temperature of 300 °C, carrier gas flow rate of 2 mL min−1, high-purity nitrogen as the make-up gas at a flow rate of 30 mL min−1, pressure of 68.95 kPa, injection volume of 1 μL, and split ratio of 100 : 1.

The settings of MS were as follows: ion source using electron impact (EI), ion source temperature of 230 °C, quadrupole temperature of 150 °C, EI energy of 70 eV, and transfer line temperature of 280 °C. The MS was recorded in full scan mode with a mass-to-charge ratio ranging from 50 to 600 units. The results of the fatty acid analysis were expressed as FAME percentages of total FAMEs. The significances of the changes in the major fatty acids were investigated by t-test with < 0.05.

MDS plot

MDS is a data reduction algorithm for expressing the original data as points in low-dimensional graph, while the interpoint relationship is maintained to the greatest extent possible. It has been widely used for data visualization in various research areas such as geography, ecology, and linguistics. All the procedures of MDS were implemented in matlab R2013a (The Mathworks Inc., Natwick, MA). First, for all the cultures, the results of the fatty acid analyses were organized into a matrix with rows representing fatty acids and columns indicating cultures. Second, the pairwise distances between every two cultures were calculated according to the following Euclidean distance equation:

display math(1)

where xi and xj are each pair of the columns in the matrix, R is the number of all the detected fatty acids, and dij is the distance between xi and xj. The distance matrix was then submitted to MDS function in matlab, and a corresponding configuration matrix was produced. The rows of the configuration matrix were the coordinates of the cultures in high-dimensional space. Next, the most important dimensions were selected based on the product of the configuration matrix and its transpose, which is known as eigenvalue. Normally, the MDS process is accompanied by information loss because of the decrease in dimension. Hence, the prevalence of the first two eigenvalues needs to be calculated to determine the proportion of variance of the scaled data. If the prevalence of the first two eigenvalues is above 80%, the first two columns of the configuration matrix could be used as the alternative of the original matrix. Lastly, the scatter plot of the first two dimensions of the cultures was graphed based on the first two columns of the configuration matrix.

SOM classification

SOM network is a type of artificial neural network that is trained using unsupervised learning to produce a lower dimensional, discretized representation of the input space of the training samples. Such network can learn to detect regularities and correlations in their input by competitive network and adapt their future responses to that input accordingly in such a way that the neurons physically near each other in the neuron layer respond to similar input vectors.

In the present study, the SOM classification of the strains based on the fatty acid profiles was implemented in matlab. The 2 × 2 and 5 × 5 neurons of the network were, respectively, developed, because the 2 × 2 SOM is a relatively smaller SOM, which can divide objects into a low number of categories, whereas the 5 × 5 SOM can contain higher number of categories, which were a maximum of 24 for the fatty acid profiles in this study. The training epoch was 200.

Results and discussion

Cellular fatty acid profiles

The results showed that the major FAMEs of V. parahaemolyticus under cold stress were most often methyl palmitate (16:0), methyl palmitoleate (16:1), and methyl oleate (18:1), whereas those of the controls were most often methyl palmitate and methyl palmitoleate (Table 2). Furthermore, under cold stress, the percentage of methyl palmitate decreased, while that of methyl palmitoleate except for strains Vp6 and Vp26 and methyl oleate remarkably increased (< 0.05). Although some relative standard deviations appeared very high, up to hundreds of percent, they were reasonable and acceptable because these fatty acid percentages were very low, reaching < 1%.

Table 2. Comparison of fatty acid percentages of Vibrio parahaemolyticus strains under cold stress and their counterparts as controls (repeats = 3)
Fatty acidV. parahaemolyticus as controlsV. parahaemolyticus under cold stress
Vp6Vp15Vp20Vp26Vp6Vp15Vp20Vp26
m (%)RSD%m (%)RSD%m (%)RSD%m (%)RSD%m (%)RSD%m (%)RSD%m (%)RSD%m (%)RSD%
  1. m, mean of three repeats of analysis; RSD, relative standard deviation; –: not available because the denominator is 0.

C10:00.3151.46.132.400.7105.20.830.802.18.71.79.9
C12:03.7299.08.44.224.62.011.43.119.53.013.56.32.63.412.4
C13:000000.790.51.310.83.92.90.98.1
C14:04.9214.08.84.94.33.63.15.41.51.810.58.00.95.53.9
C15:06.5170.1227.14.922.81.940.34.43.40.1110.21.016.64.46.3
C16:047.6151.13.446.4346.64.536.40.69.90.410.21.335.73.2
C16:1T1.113201.315.22.321.55.12.60.5371.07.82.399.9
C16:132.4429.21.334.62.239.83.131.81.554.31.641.93.336.81.5
C17:03.52303.7261.9124.53.10.59.71.126.23.55.2
C17:10000000.644.90
C18:000.3129.200.8112.60.911.70.1112.80.9140.374.9
C18:10000.581.56.93.228.54.422.91.85.59.9

For decades, the fatty acid profiles have been used for microbial identification. It has been proved that the fatty acid profiles are unique from one species to another (Tighe et al., 2000; Ivanova et al., 2001; Slabbinck et al., 2010; Li et al., 2011) and vary under different growth conditions (Hamamoto et al., 1994; Day & Oliver, 2004; Wong et al., 2004; Chiang et al., 2005; Fouchard et al., 2005; Ibragimova et al., 2012). The major fatty acids of the Vibrio strains, are observed most often, are hexadecenoic (16:1), hexadecanoic (16:0), and octadecenoic (18:1) acids (Urdaci et al., 1990; Carballeira et al., 1995). Similar results were noted in the present study. In addition, we observed that the percentage of unsaturated fatty acids (16:1 and 18:1) increased under cold stress. It has been proposed that at low temperatures, the bacterial cells increase the percentage of unsaturated fatty acids to maintain fluidity (Allen et al., 1999). In a previous study, when a psychrophilic Vibrio sp. strain was grown at low temperature, the relative amount of unsaturated docosahexaenoic acid (22:6) was observed to be increased in the cells, to maintain membrane fluidity at that temperature (Hamamoto et al., 1994). Similarly, in another study, after ethanol shock, the percentage of 18:1 was increased in V. parahaemolyticus, and the proportion of 16:0 and the ratio of saturated fatty acids to unsaturated fatty acids decreased (Chiang et al., 2008). Thus, it can be concluded that not only low temperatures but also other stress conditions could induce an increase in the percentage of unsaturated fatty acids.

However, there are minor discrepancies between the findings of the present study and those reported previously. The fatty acid profiles of V. parahaemolyticus controls obtained in the present study were noted to be different to some extent, mainly in the percentages of 18:1, from those of the strains in Culture Collection in University of Göteborg (CCUG, http://www.ccug.se/). The 18:1 percentages of our control cells were significantly lower than those of the CCUG strains. The reasons for this difference could perhaps be the variations in the strains, culture media, incubation temperature, and extraction and analysis methods. The CCUG strains were isolated from river water (CCUG 34903) or seafood (CCUG 34775 and CCUG 14474 T) and incubated on marine agar or blood agar at 30 °C, indicating that these strains that occur in cold marine environment are different from those used in the present study. Interestingly, the 18:1 percentages of our cold-stressed cells were similar to those of the CCUG strains. Hence, it can be presumed that the CCUG strains might have been stressed by cold environment, which could have resulted in the high 18:1 percentages. In addition, the results obtained in the present study for cold-stressed strains were quite different from those reported by Wong et al. (2004) for cells in VBNC state. The reason for this difference might be the modified Morita mineral salt-0.5% NaCl medium and induced VBNC state in the study by Wong et al. It must be noted that the VBNC cells are usually distinctly different from the culturable ones.

MDS plot

MDS was performed on this Euclidean distance matrix to achieve a spatial representation of the interrelatedness of the bacterial cultures in low-dimensional space (Fig. 1). The sum of the first two eigenvalues accounted for 95.55% (> 80%) of the total, then the cultures could be plotted onto a two-dimensional graph with little information loss, and the graph was an excellent representation of the original data matrix. The actual distance between every two data points represented their relationship in fatty acid profiles.

Figure 1.

Bidimensional representation of the fatty acid profiles of Vibrio parahaemolyticus strains under cold stress and their controls obtained by MDS (Filled markers: strains under cold stress; Unfilled markers: strains as controls).

From the graph, it was obvious that there was a spatial division between the cold-stressed bacterial cultures and the controls. After cold stress, the strains, especially strains Vp15 and Vp20, seemed to move toward the left of the graph. The repeats of the same strain lying close to one another indicated that the variations in the same strain were much lesser than those between the strains, although strains Vp6 and Vp20 as controls overlapped to some extent. Furthermore, the four strains under cold stress were separated more far away than their controls, which suggested that the variation in the former was much greater than that in the latter. Thus, it can be concluded that the fatty acid profiles of the cells were significantly changed after cold stress.

The data of the fatty acid profiles are multivariate with a dozen of dimensions (fatty acids). Much of the impetus for the analysis of multivariate data came from the problem of handling the data matrix. Such matrices cannot be readily analyzed visually. For the fatty acid analysis, similarity index is commonly applied to express the classification results (Sasser et al., 2005), followed by cluster analysis (Ivanova et al., 2001; Fouchard et al., 2005; Slabbinck et al., 2010; Li et al., 2011). In the present study, we proposed a MDS to detect the cold-stressed cells based on the fatty acid profiles, which provided a unique visual representation of their relationships, similar to the principal component analysis (PCA; Whittaker et al., 2003; Xu et al., 2003); however, unlike PCA, our proposed MDS did not provide a linear dimensionality reduction method (Raje et al., 2010). As a result, the 12-dimensional fatty acid data were successfully reduced to a two-dimensional configuration representation with little information loss.

SOM classification

After the SOMs were trained, the 24 fatty acid profiles were classified into categories (Fig. 2). Each hexagon in the graph of the SOM sample hits represented a neuron in the SOM network and was called a category. The number in a neuron denoted the quantity of the samples classified into this category, and the neuron containing a number was a winning neuron. As the SOMs are trained with input vectors in a random order, starting with the same initial vectors does not guarantee identical training results (winning neurons), but renders identical classification results. Hence, different neurons won when a new training finished. For the 2 × 2 SOM, there were three categories containing 12, 6, and 6 bacterial culture samples, respectively, which were easily identified in matlab (Table 3). The numbers in Table 3 denote the winning neurons and indicate that their corresponding samples belong to the same categories. All the controls were obviously classified into the same category, and cold-stressed strains were classified into the other two categories, which were the category containing cold-stressed Vp6 and Vp26 and that containing cold-stressed Vp15 and Vp20. This 2 × 2 SOM supported two conclusions. First, the repeatability of the fatty acid analysis was perfectly accomplished, because all the triplicates shared the same categories. Second, as noted previously, the variation in the cold-stressed V. parahaemolyticus strains was greater than that in their corresponding controls. For the 5 × 5 SOM, there were 11 categories containing samples as shown in Table 3. Obviously, the larger (5 × 5) SOM provided more intensive classification than the smaller (2 × 2) one. The controls shared closely related categories because the category numbers were close (17–25), whereas the cold-stressed strains belonged to different categories (even a triplicate shared different categories; category number 1–10). Thus, it can be concluded that the cold-stressed V. parahaemolyticus strains were different from the controls, although there were obvious variations in both the cold-stressed samples and the controls.

Table 3. Classification results of Vibrio parahaemolyticus strains under cold stress and their counterparts as controls
SOM classification numbersV. parahaemolyticus strains as controlsV. parahaemolyticus strains under cold stress
Vp6Vp15Vp20Vp26Vp6Vp15Vp20Vp26
  1. SOM, self-organized map.

  2. All strains were in triplicates. The number denotes the winning neuron (classified category). Close numbers mean close distances between the two neurons.

2 × 2 SOM444444444444222111111222
5 × 5 SOM1717172121212318182425251114441010101477
Figure 2.

SOM sample hits of the fatty acid profiles of Vibrio parahaemolyticus strains under cold stress and their controls. (a) 2 × 2 SOM sample hits; (b) 5 × 5 SOM sample hits.

Detection of stressed bacteria

As mentioned earlier, stressed bacteria are injured and sometimes difficult to be detected by classical culture method. In the present study, from the MDS or SOM graph, the bacterial samples were distinctly shown to have characteristic fatty acid profiles owing to their separation from one another. After a classification model built on a analysis techniques, if an unknown bacterial sample, which is difficult to be detected by classical culture method owing to stress damage, is analyzed using the same techniques, it could be placed within this model to determine whether it was akin to those species/strains already analyzed (Thorn et al., 2011), and thus can be identified. However, in our opinion, the reliability of this technique for detecting bacteria in unusual stress states is directly linked to two aspects: (1) the discrete fatty acids are characteristic to stress states and chosen, and (2) the methods of purification and enrichment of stressed target bacteria in food are established. The former aspect is supported by the present study as well as some previous ones. For example, it has been reported that the fatty acid profiles were closely linked to the growth phase and growth temperature (Fouchard et al., 2005). However, considering the latter aspect, it is currently quite difficult for this technique to be used in detecting the bacterial state in routine work of food testing. Under stress, the target bacteria can survive simultaneously in various states or die, and the present approaches for sample preparation such as immune magnetic separation cannot differentiate these stressed cells from the others, so in fact, the collection of pure stressed bacteria in food is difficult and even impossible. Nevertheless, although there are some difficulties, we consider that this technique could bring forward a promising thought for detecting bacteria in unusual state.

In conclusion, the technique described in the present study, which is based on fatty acid profiles using MDS or SOM could have potential use in the detection of stressed bacteria, including those in an unusual stress state like VBNC state; however, its reliability should be established based on the two above-mentioned aspects.

Acknowledgements

This work was supported by the Chinese State High-Tech Development Plan (863 program) (2012AA101605) and the Science Foundation of General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China (2009IK176 and 2012IK305).

Ancillary