Metabolic Profiling of Lactococcus lactis Under Different Culture Conditions

Gas chromatography mass spectrometry (GC-MS) and headspace gas chromatography mass spectrometry (HS/GC-MS) were used to study metabolites produced by Lactococcus lactis subsp. cremoris MG1363 grown at a temperature of 30 °C with and without agitation at 150 rpm, and at 37 °C without agitation. It was observed that L. lactis produced more organic acids under agitation. Primary alcohols, aldehydes, ketones and polyols were identified as the corresponding trimethylsilyl (TMS) derivatives, whereas amino acids and organic acids, including fatty acids, were detected through methyl chloroformate derivatization. HS analysis indicated that branched-chain methyl aldehydes, including 2-methylbutanal, 3-methylbutanal, and 2-methylpropanal are degdradation products of isoleucine, leucine or valine. Multivariate analysis (MVA) using partial least squares discriminant analysis (PLS-DA) revealed the major differences between treatments were due to changes of amino acids and fermentation products.


Changes of Amino Acids and Fermentation End Products
Multivariate analysis was performed by hierarchical clustering analysis (HCA) to examine the variations of the amino acids and fermentation end products under stresses of temperature and agitation. As shown in Figure 3, some of the aspartate family (aspartate, glutamine, glycine, serine and threonine) and shikimate-derived amino acids (phenylalanine and tyrosine) are more abundant (red) under the 30 °C agitated condition. In general, aspartate is the precursor to several amino acids including threonine, and isoleucine, while the shikimate pathway via chorismate is essential for the aromatic compounds biosynthesis. Furthermore, greater abundance of phenylalanine and tyrosine particularly under agitated condition suggested a role of phosphoenolpyruvate and erythrose-4phosphate in the activation of shikimate pathway for the production of these aromatic amino acids. Under non-agitated conditions, ornithine was more abundant at 30 °C, compared with cysteine at 37 °C. The relationship between larger amounts of ornithine and the fermentation end products ethanol and acetate are likely to be associated with carbon limitation. During exponential phase that lasted 5 to 6 h (data not shown), when the carbon source started to exhaust, there might be a shift toward mixed-acid fermentation. The modification of pyruvate metabolism via pyruvate dehydrogenase activity could explain the production of ethanol and acetate. Furthermore, it was suggested that ornithine is a result from arginine catabolism via the arginine deiminase (ADI) pathway [14]. ADI is responsible for the conversion of arginine into ornithine via citrulline. The activation of ADI was influenced by the carbon starvation and changes of pH. In general, ornithine is not a constituent of casein, thus the presence of ornithine indicates the activation of specific enzyme or pathway to generate the amino acid.
A greater amount of cysteine at 37 °C is consistent to that reported effects at elevated temperature, which affect the incubation time, viable bacterial counts and pH changes [15]. Cysteine plays important roles in protein folding, assembly and stability via the formation of disulfide bonds [16]. The greater amount of lactate at 37 °C suggested a metabolic shift to lactate by pyruvate during carbon limitation. This was indicated by little changes in alanine, which is produced from pyruvate that also influences the production of glycine and serine.

Changes of Amino Acids in Response to Temperature and Agitation
The HCA in Figure 3 indicated that there are several clusters (horizontal dendrogram) of amino acids changes with two big clusters, larger one (upper cluster) divided into three groups. The lower clusters that grouped serine, glycine, aspartate, tyrosine, phenylalanine and isoleucine and threonine and glutamine indicated that specific pathways were more activated. As shown in Figure 4, the changes of amino acids that were observed at 30 °C with agitation indicated role of 3-phosphoglycerate (3PG) and phosphoenol pyruvate (PEP). The larger amounts of valine and leucine suggest the presence of more pyruvate during agitation stresses while the larger amount of isoleucine would require more threonine via 2-oxobutanoate which is derived from oxaloacetate (OAA). Observation of amino acids changes at 30 °C and 37 °C suggested role of ribose-5-phosphate (R5P) and oxaloacetate (OAA) due to changes of histidine and asparagine, methionine, and threonine. . Schematic representation of precursor relationship between amino acids and the central carbon metabolism. R5P represents ribose -5-phosphate, 3PG represents 3-phosphoglycerate and PEP represents phosphoenol pyruvate. OAA represents oxaloacetate and AKG represents 2-oxoglutarate while ADI represents the arginine deiminase.

Comparison of Metabolites Detected Using TMS, MCF and HS Analysis
TMS is routinely employed in gas chromatography (GC) to increase the chemical volatility and stability of organic metabolites containing active hydrogen [17]. The derivatization is based on the methoximation and silylation enabling the detection of a wide range of metabolite groups, including sugar derivatives, organic acids, fatty acids and amino acids (Table 1). However, the procedure usually requires a long heating treatment and must be carried out under anhydrous conditions [12,13]. The use of microwave-assistance (MA) in TMS derivatization procedure of MSTFA and BSTFA has significantly increased the detection of metabolites while shortening the heating period of methoximation and silylation steps [13,18,19]. The use of microwave irradiation for methoximation and silylation prior to GC-MS analysis has been widely used in biological samples including environmental analysis, herbicides and industrial related processes [19][20][21][22][23][24][25]. In this study, MSTFA which works well with microwave application was used instead of other TMS reagent.
The use of MCF derivatization in GC-MS was first introduced by Husek et al. [26]. The derivatization does not require special sample preparation or multiple reaction steps or heating treatment [27]. The alkyl chloroformates based derivatization favors detection of carboxyl group (-COOH) containing metabolites. As shown in Table 2, detected metabolites are mainly amino acids and organic acids, including metabolites from the citrate cycle and fatty acids.

Headspace (HS) Analysis
Direct analysis using dynamic headspace (HS) coupled to GC-MS was carried out to detect any volatile metabolites that react less with the derivatization reagents. Branched chain methyl aldehydes (2-methylbutanal, 3-methylbutanal and 2-methylpropanal) that are commonly produced by L. lactis during cheese manufacturing [9,28,29] were successfully detected, including pentanal and sulfur-based compounds ( Figure 5). In brief, aldehydes are the most abundantly produced metabolites by L. lactis, especially in the cheese manufacturing process [28,29]. The production of the branched-chain methyl aldehydes was associated with the lower amounts of branched-chain amino acids of leucine, isoleucine and valine, while pentanal was likely from degradation of unsaturated fatty acids [28]. No production of 2-methylbutanal is observed under the agitation condition, although a high isoleucine response was detected under those conditions.

Figure 5.
Bar chart representing the relative abundances of detected branched-chain aldehydes using dynamic headspace (HS) coupled to GC-MS according to the three conditions. Condition with agitation showed less production of branched chain aldehydes. HS_30 represents condition of 30 °C, HS_37 represents condition of 37 °C and HS_30150 represents condition with agitation (150 rpm).

Partial Least Square Discriminant Analysis
Discriminant analysis using supervised PLS-DA for samples derivatised using TMS ( Figure 6A) and MCF ( Figure 6B) indicated clear separation between the 30 °C, 30 °C with agitation and 37 °C.
As MCF derivatization favours the detection of amino acids, the PLS-DA score plot ( Figure 6B) is likely to be influenced by the amino acids response. This is supported by the hierarchical clustering analysis (HCA) on the particular metabolites that revealed amino acids to influence the discrimination ( Figure 6C).

Growth Estimation
Optical density (OD) at 600 nm was used to provide a measure biomass and constructing growth curve for each of cultivation under the specific conditions.

Extraction of Extracellular Samples
Approximately 15 mL of fermented culture medium was taken and filtered using a cellulose acetate membrane filter (0.2 µm pore size) to remove the microbial cells. Filtered culture medium was then separated into 1 mL aliquots (n = 5) followed by the addition of dH 2 O (10 mL) and internal standard (0.2 µmol of 10 mmol solution of 2,3,3,3-d4 D,L-alanine) to each of the samples. Samples were then freeze dried under low temperature (−56 °C) and stored at −20 °C. Un-inoculated M17 broth (WT) was also prepared as control.

Sample Derivatization Using TMS
The TMS derivatization method was based on the optimized protocol described by Villas-Boas et al. [25] and Rossner et al. [31] Briefly, freeze dried samples were resuspended in methoxyamine hydrochloride solution in pyridine (80 µL, 2 g/100 mL), followed by incubation in a domestic microwave (Panasonic NN-K544WF) with multimode irradiation set to 500 W and 50% of exit power for 2.48 min. MSTFA was then added (approximately 80 µL), followed by incubation in the domestic microwave for 3 min, under same conditions as previously mentioned. The final mixed incubation sample was then transferred to a GC-MS vial and analyzed by GC-MS.

Sample Derivatization Using MCF
The MCF derivatization method was based on protocol described by Smart et al. [27] and Villas-Boas [25]. Briefly, freeze dried samples were resuspended in NaOH (1M), followed by addition of methanol, pyridine, MCF and chloroform and sodium bicarbonate. The upper aqueous layer in the sample was discarded, and a small portion of anhydrous sodium sulphate was added to dry the remaining reagents. Finally, the dried solution was transferred into a GC-MS vial and analyzed by GC-MS.

Sample Preparation for Headspace Analysis (HS)
Briefly, freeze dried samples were resuspended in 200 µL of dH 2 O and homogenized under temperature of 40 °C before being loaded in headspace apparatus and analyzed using GC-MS.

GC-MS Parameter for Samples Prepared by TMS
The GC-MS parameter used was optimized based on Villas-Boas et al. [25] and Rossner et al. [31]. GC-MS analysis was performed using the GC-MS Perkin Elmer Turbo Mass Clarus 600 coupled to a quadruple mass selective detector on electron ionization (EI) operated at 70 eV. An aliquot of approximately 1-μL was injected into an Elite-5MS capillary column coated with 5% diphenyl crosslinked and 95% dimethylpolysiloxane (30 m × 0.25 mm i.d. × 0.25 μm thickness) in split mode (50:1). The injection temperature was set to 250 °C, and the ion source temperature was adjusted to 200 °C. The GC method was set from 70 °C to 300 °C with helium gas flow constantly at 1.1 min −1 . The measurements were made in the full scan mode (m/z 45-600).

GC-MS Parameters for Samples Prepared by MCF
The GC-MS parameter used was described by Smart et al. [27] and Villas-Boas et al. [25]. GC-MS analysis was performed using Agilent GC-MS coupled to a quadruple mass selective detector on electron ionization (EI) operated at 70 eV. An aliquot of approximately 1-μL was injected into J&W 1701 column (30 m × 250 mm i.d. × 0.15 mm) (Folsom, CA). The injection temperature was set to 250 °C, and the ion source temperature was adjusted to 200 °C. The GC method was set from 45 °C to 280 °C with helium gas flow constantly at 1.0 min −1 . The measurements were made in the scan mode of 38-650 m/z at 1.47 scan per sec.

GC-MS Parameters for Headspace Analysis (HS)
A Perkin Elmer TurboMatrix Headspace Sampler 40XL connected to a GC-MS Perkin Elmer Turbo Mass Clarus 600 was used for volatile compounds analysis. A minimal of three duplicates were subjected to helium purge and concentrated in a Tenax trap, kept at 40 °C. Line temperature was adjusted to 180 °C, while helium flow was set at 40 mL/min. Sample temperature was 80 °C, with dry purge time was 1min and desorbed temperature was 200 °C. Desorbed time was 1min, and injection port temperature was set to 200 °C. GC-MS analysis was performed using electron ionization (EI) operated at 70 eV. An aliquot of approximately 1 μL was injected into an Elite-5MS capillary column coated with 5% diphenyl crosslinked and 95% dimethylpolysiloxane (30 m × 0.25 mm i.d. × 0.25 μm thickness) in split mode (50:1). The injection temperature was set to 250 °C, and the ion source temperature was adjusted to 200 °C. The GC method was set from 45 °C to 220 °C with helium gas flow constantly at 1.0 mL min −1 . The measurements were made in the scan mode of m/z 33-220.

Data Analysis and Validation
The general approach used for data analysis and validation was performed according to Smart et al. [27] and Villas-Boas et al. [25]. In summary, detected metabolites were identified using in-house TMS and MCF MS library of derivatised pure standard developed by Villas-Boas. For peaks that have not been identified was identified using NIST mass spectral database library (NIST 2008) with cut-off similarity of 90%. The value of height of the peak was used to represent the detected metabolites. The values were firstly normalized by total sum of GC height and internal standard followed by log transformed. One-way Analysis of Variance (ANOVA) was used to statistically validate the values followed by comparison using Fisher's least significant difference (LSD) method with significance levels of P < 0.05, P < 0.01 and P < 0.001 [32]. Visualization of the clean, validated data was then carried out using Principal Component Analysis (PCA) and PLS-DA of Simca-P+ version 12.0 (Umetrics AB, Ume, Sweden) for group classification and discrimination analysis with Q 2 value > 50%. The heatmap with hierarchical clustering analysis was performed using R script (http://www.r-project.org) with the ward method.

Conclusions
GC-MS and HS analysis on the metabolites produced by L. lactis in response to temperature and agitation contribute to the understanding of metabolic changes during environmental stresses.
Meanwhile, the use of TMS derivatization provides a wider range of metabolites detection compared to MCF derivatization which specifically targets amino acids. The PLS-DA derived analyses indicate a strong relationship between fermentation end products of lactate, ethanol, acetate and amino acid changes according to temperature. Finally, these specific responses can serve as optimization factors useful for dairy food production which uses L. lactis as starter culture.