An approach to screen diverse cheeserelated bacteria for their ability to produce aroma compounds T. Pogačić 1,2,3, A. Yee 1,2, MB. Maillard 1,2, A. Leclerc 4, C. Hervé 4, V. Chuat 1,2, F. Valence 1,2, A. Thierry 1,2 Rennes 1 INRA, UMR1253 Science and Technology of Milk and Eggs, Rennes, France 2 Agrocampus Ouest, UMR1253, Rennes, France 3 Laboratoires Standa, F-14000 Caen, France
Why? Flavour an important property for consumers choosing food products is the result of the correct balance and concentration of a wide variety of volatile flavour compounds Mulder (1952) Kosikowski and Mocquot (1958) Flavour of fermented food products, including cheese, largely results from the production of aroma compounds by microorganisms The production of flavour compounds by micro-organisms is highly strain-dependent, thus requiring screening studies
How? A metabolomics-based approach VARIABLES to[1] Metabolomics: a comprehensive approach aims at identifying and (semi-)quantifying all the metabolites present and accessible to the analysis (Fiehn 2002) follows a specific workflow divided into several steps SELECTION OF SIGNALS OF INTEREST INTERPRETATION SAMPLE PREPARATION EXTRACTION AND ANALYSIS MULTIVARIATE STATISTICAL ANALYSES (PCA) EXPERIMENTAL DESIGN DATA PROCESSING SAMPLES C vs T1 vs T2 vs T4 ESI+.M3 (OPLS/O2PLS-DA), C vs T1 vs T2 vs T4 esi + t[comp. 1]/to[XSide Comp. 1] Colored according to classes in M3 1 2 3 4 35 30 25 MICROBIAL CULTURES 20 15 10 C C T1 T4 5 0 C T1 T1 T4 T2 T4 T4 T2 T2 T2-5 -10-15 -20-25 C C T1 T2-30
A metabolomics-based approach: first steps EXPERIMENTAL DESIGN MICROBIAL CULTURES
.05
Experimental design ~100 strains tested: 10 genera, 30 species Lb. fermentum Lb. rhamnosus Lb. parabuchneri Lb. paracasei Lb. plantarum Lb. crustorum Lb. sakei Lb. gallinarum Lb. sanfranciscensis Lb. helveticus Lc. lactis Leuc. lactis Leuc. mesenteroides Leuc. fallax Leuc. citreum Leuc. carnosum P. freudenreichii P. jensenii P. thoenii P. acidipropionici P. cyclohexanicum Lut. japonicus Micr. gubbeenense Cor. casei Brev. aurantiacum Brev. linens Brac. alimentarium Brac. tyrofermentans Brac. articum Hafnia alvei Lactic acid bacteria Propionibacteria Others (coryneforms and G - bacteria)
Growth of bacterial strains L. sakei L. sakei L. paracasei L. paracasei L. paracasei L. rhamnosus L. rhamnosus L. rhamnosus L. fermentum L. fermentum L. helveticus L. helveticus L. helveticus L. citreum L. citreum L. Leuc_lactis L. mesenteroides L. mes crem P. freudenreichii P. freudenreichii P. freudenreichii P. freudenreichii P. freudenreichii B. articum B. tyrofermentans B. aurantiacum B. aurantiacum M. gubbeenense H. alvei H. alvei Culturable population, enumerated at 24 h and after 3 weeks incubation at 15 C CFU/g 1.0E+10 1.0E+08 1.0E+06 1.0E+04 1.0E+02 1.0E+00 maximal populations > 4.10 7 CFU/g for most strains
SAMPLE EXTRACTION AND ANALYSIS Volatiles extracted and analyzed by gas chromatography-mass spectrometry (GC-MS) Headspace Trap GC-MS EXPERIMENTAL DESIGN MICROBIAL CULTURES
SAMPLE EXTRACTION: Head space methods Extraction is a key step! Head space-related methods widely used for study of food flavor static headspace Head space sample reliable poor sensitivity Other head space-related methods - Solid Phase Micro-Extraction (SPME): - the most widely used for food flavour Small trap capacity risk of competition between the volatiles - Head Space Trap (Perkin Elmer) trap dynamic headspace sensitive automation
SAMPLE PREPARATION EXTRACTION AND ANALYSIS: Head Space Trap-GC-MS Head space Trap Gas chromatograph Mass spectrometer 10
SAMPLE PREPARATION EXTRACTION AND ANALYSIS: Head Space Trap Headspace Trap (Perkin Elmer patent ) Optimization: four parameters tested - T : 50-60 C - Time: -P 15-30 min trap - Nr of repeated -T extractions: 1-2 - Sample -Time size: 2.5-5 g/vial -Nr of repeated for two types extractions of samples: cheese and a mixture of standard compounds example for the extraction of diacetyl from cheese
Volatile extraction: optimisation of extraction 65 C 2.5 g 2 cycles 65 C 2.5 g 2 cycles 65 C 5 g 2 cycles 65 C 5 g 2 cycles Abundance of diacetyl as function of the extraction parameters 65 C 2 cycles of extraction 65 C 2.5 g 1 cycle 65 C 2.5 g 1 cycle 65 C 5 g 1 cycle 65 C 5 g 1 cycle 50 C 2.5 g 2 cycles 50 C 2.5 g 2 cycles 50 C 5 g 2 cycles 50 C 5 g 2 cycles 50 C 2.5 g 1 cycle 50 C 2.5 g 1 cycle 50 C 5 g 1 cycle 50 C 5 g 1 cycle 50 C 1 cycle of extraction X 3.8 T and Nr of extraction cycles significantly recovery rates Similar results for the other volatiles and for both types of samples 0.0E+00 5.0E+07 1.0E+08 1.5E+08 2.0E+08 2.5E+08 Limit of quantification : 15 ng/g neutral volatiles, 200 ng/g for acids Linearity range: 0-1000 ng/g Pogacic et al, Food Microbiology, 2015
A metabolomics-based approach: data processing VARIABLES SAMPLE PREPARATION EXTRACTION AND ANALYSIS EXPERIMENTAL DESIGN DATA PROCESSING SAMPLES MICROBIAL CULTURES using the XCMS package of R software
100 GC-MS DATA PROCESSING 6.34e9 14.29 % 11.96 time m/z 3.97 0 intensity tem_6h 100 9.70 14.38 15.65 18.74 19.02 1: Scan EI+ TIC 6.34e9 intensity 14.35 % m/z 11.96 19.19 m/z 9.68 19.39 3.97 6.52 10.78 16.20 0 Time 0.39 5.39 10.39 15.39 20.39 25.39 30.39 16.78 GC-MS chromatogram time peak: one ion (m/z) at one time GC-MS data are 3-D files containing the abundance of all the ions detected at each retention time The manual extraction of the information (integration of each peak) from GC-MS data is very time-consuming
GC-MS DATA PROCESSING Input: Series of GC-MS raw files (format.cdf) sample 1 sample 2 sample 3 sample 4 sample 5 Several softwares developed to facilitate data analysis convert the raw data to time- and massaligned chromatographic peaks Peak picking Retention time alignment across samples Use of the open source package XCMS of R software (Smith et al., 2006) initially developed for LC-MS metabolomics name Fill in missing peak data Ion 1 time 1 ion 2 time 1 ion 3 time 1 ion 1 time 2 ion 2 time 2 ion 3 time 2 sample 1 sample 2 Output: A unique table for all samples with the abundance of each ion detected at each time sample 3 sample 4 ~2500 lines ~ 50-80 volatile compounds
A metabolomics-based approach: examples of results VARIABLES to[1] INTERPRETATION SAMPLE PREPARATION EXTRACTION AND ANALYSIS SELECTION OF SIGNALS OF INTEREST EXPERIMENTAL DESIGN DATA PROCESSING SAMPLES MULTIVARIATE STATISTICAL ANALYSES (PCA) C vs T1 vs T2 vs T4 ESI+.M3 (OPLS/O2PLS-DA), C vs T1 vs T2 vs T4 esi + t[comp. 1]/to[XSide Comp. 1] Colored according to classes in M3 1 2 3 4 35 30 MICROBIAL CULTURES 25 20 15 10 5 0 C C C T1 T1 T1 T4 T4 T4 T2 T4 T2 T2 T2-5 -10-15 C T1 C -20-25 -30-35 -34-32 -30-28 -26-24 -22-20 -18-16 -14-12 -10-8 -6-4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 t[1] T2
First example of results: 11 cheese-related bacterial species 53 volatile compounds identified, of which 50 varied significantly in abundance between bacteria Sulfur compounds Hafnia alvei Biological replicates closely situated Different volatile profiles according to the species Free fatty acids and esters Control medium Brachybacterium Pogacic et al, Food Microbiology, 2015
Dim 2 (14.99%) Second example of results : 78 propionibacteria strains branched-chain aldehydes (malty) Individuals factor map (PCA) -4-2 0 2 4 Ctrl Pj_S40B Pt_C1429 Pj_C1657 Pj_C1658 Pt_C1508 Pj_C1661 Pt_C1510 Pt_C1662 Pt_C1434 Pt_C1329 Pf_S13 Pt_C1663 Pf_S14 Pj_C1432 Pf_S209 Pj_C133 Pj_S36B Pj_C1659 Pj_C1506 Pj_C1430 Pf_S23 Pf_C1406 Pf_S101 Pf_S108 Pf_S24 Pj_C1655 Pj_C1505 Pj_C1660 Pj_S39 Pj_S37 Pj_S38 Pt_C1509 Pf_S31 Pf_S210 Pj_S61 Pj_S58 Pf_S100 Pf_S56 Pa_C1652 Pa_C1656 Pm_C677 Pf_S18A Pc_C698 Pf_S10 Pf_C123 Pf_C1412 Pf_C141 Pf_S110 Pa_C1425 Pa_S96 Pf_C6 Pf_S219 Pf_S26 Pf_S9 Pf_C1397 Pf_S214 Pf_S11S Pf_S3F Pf_S22 Pa_C1653 Pa_C1664 Pf_C1499 Pf_S102 Pf_C1413 Pf_S28 Pf_S3S Pf_S51S Pj_C1507 Pf_C1497 Pf_S19 Pa_C1427 Pf_C1426 Pf_S223 Pa_C1428 Pf_S51F Pf_C1414 Pf_S1F P. acidipropionici P. thoenii free fatty acids (rancid) branched-chain acids (socks) esters (fruity) -6-4 -2 0 2 Yee et al, 4 Int J. Food Microbiol, 2014 Dim 1 (45.02%) P. jensenii P. freudenreichii Production of compounds with a common origin generally correlated Large inter- and intraspecies diversity of propionibacteria
Take-home messages The metabolomics-based workflow (data processing using XCMS followed by multivariate analyses) facilitates data analysis The HS-trap method used is an efficient tool for volatile extraction and an alternative to SPME This approach can be applied with different objectives to evaluate very different genera of cheese-associated bacteria for their ability to produce aroma volatile compounds in pure and mixed cultures to determine the volatile profile ( volatilome ) of spoilage micro-organisms to analyze cheeses manufactured under different conditions or with different starters/adjuncts
Thanks to Tomislav POGAČIĆ Alyson YEE Marie-Bernadette MAILLARD Victoria CHUAT Florence VALENCE Centre of Biological Ressources THANK YOU FOR YOUR ATTENTION Please visit http://www6.rennes.inra.fr/stlo_eng.020