Microbial Ecology and Food Safety of Fermented Carrot Juice

Home , Carrot juice, Cucumber juice, Vegetable juice

MICROBIAL ECOLOGY AND FOOD SAFETY OF FERMENTED CARROT JUICE

Cédric Verschueren Student number: 01611321

Promotor(s): Prof. dr. ir. Mieke Uyttendaele (Ghent University), Prof. dr. ir. Sarah Lebeer (University of Antwerp)

Tutor: MSc. Wannes Van Beeck (University of Antwerp)

Master’s Dissertation submitted to Ghent University in partial fulfilment of the requirements for the degree of Master of Science in Bioscience Engineering: Food Science and Nutrition

Academic year: 2018 - 2019

MICROBIAL ECOLOGY AND FOOD SAFETY OF FERMENTED CARROT JUICE

Cédric Verschueren Student number: 01611321

Promotor(s): Prof. dr. ir. Mieke Uyttendaele (Ghent University), Prof. dr. ir. Sarah Lebeer (University of Antwerp)

Tutor: MSc. Wannes Van Beeck (University of Antwerp)

Master’s Dissertation submitted to Ghent University in partial fulfilment of the requirements for the degree of Master of Science in Bioscience Engineering: Food Science and Nutrition

Academic year: 2018 - 2019

Permission of use De auteur en de promotor geven de toelating deze masterproef voor consultatie beschikbaar te stellen en delen van de masterproef te kopiëren voor persoonlijk gebruik. Elk ander gebruik valt onder de beperkingen van het auteursrecht, in het bijzonder met betrekking tot de verplichting de bron uitdrukkelijk te vermelden bij het aanhalen van resultaten uit de masterproef. The author and the promotor give permission to use this thesis for consultation and to copy parts of it for personal use. Every other use is subject to the copyright laws, more specifically the source must be extensively specified when using results from this thesis.

Ghent, September 2019

Promotor(s): Tutor: Author:

Prof. dr. ir. Mieke Uyttendaele, MSc. Wannes Van Beeck Cédric Verschueren Prof. dr. ir. Sarah Lebeer

Preface “non scholae sed vitae discimus” (Inversion of Annaeus Seneca, Lucius. Epistulae morales ad Lucilium, CVI) With the submission of this Master’s Dissertation my academic journey comes closely to an end. All the hard work is ultimately rewarded. Despite the hard work, this Master programme was very interesting and instructive in all its aspects. I am convinced that the acquired knowledge and experience only can bring added value to the next phases of my life. This is the perfect moment to thank the people who have supported me in recent years and also the people who have helped me bringing this Master’s Dissertation to a good end. First of all I would like to thank my promotors and tutor, Prof. dr. ir. Mieke Uyttendaele, Prof. dr. ir. Sarah Lebeer and MSc. Wannes Van Beeck. They were always there providing useful feedback and answers to my questions and managed this dissertation in the right directions. I would also like to thank the lab of environmental ecology and applied microbiology at the University of Antwerp and the lab of food microbiology and food preservation at Ghent University for the daily pleasant atmosphere. In particular dr. Inge Van der Linden and Ann Dirckx for their help in the lab and provision of necessary material for my research. Finally, I would also like to thank my parents, family and friends for their support during the past years, they made it possible to bring this chapter to a successful end. Ghent, September 2019

Cédric Verschueren

List of contents

Permission of use ______ii

Preface ______iv

List of Abbreviations ______I

Abstract (EN) ______1

Abstract (NL) ______2

Introduction ______3

Literature______5

The art of fermentation ______5 1. Fermentation ______7 Lactobacillus genus complex______7 Fermentation in general ______8 Fermentation types ______10 Lactic acid fermentation ______10 Vegetable fermentation ______11 1.5.1. Sauerkraut ______13 1.5.1.1. Production process ______13 1.5.1.2. Microbiology ______13 1.5.2. Kimchi ______13 1.5.2.1. Production process ______13 1.5.2.2. Whole-cabbage kimchi ______14 1.5.2.3. Microbiology ______14 Fermented vegetable juices ______14 Spontaneous vs. Starter cultures ______15 2. Food safety ______16 Food safety in general ______16 2.1.1. Global ______16 2.1.2. Europe ______17 2.1.3. Belgium ______17 Food safety of vegetables ______18 Food safety of fermented foods ______19 Pathogens ______20 2.4.1. Listeria monocytogenes ______21 2.4.2. Salmonella spp. ______22 2.4.3. Pathogenic STEC (Shiga-toxin E. coli): O157 and non-O157 ______22 2.4.4. Yersinia enterocolitica and pseudotuberculosis ______22 2.4.5. Opportunistic bacteria ______23 3. Why fermented carrot juice & what about the food safety? ______24

Material & methods ______25

1. Bacterial strains & culture conditions ______25 Strain selection ______25 Stock culture ______25 Working culture ______26 Selective media for pathogens ______26 1.4.1. Agar Listeria according to Ottaviani and Agosti (ALOA) ______26 1.4.2. Xylose-Lysine-Desoxycholate (XLD) Agar ______26 1.4.3. Cefixime Tellurite - MacConkey Sorbitol agar (CT-SMAC) ______27 Media for microbial community ______27 1.5.1. MRS (Man, Rogosa & Sharpe) ______27 1.5.2. VRBG (Violet Red Bile Glucose) ______28 1.5.3. YPD (Yeast Extract-Peptone-Dextrose) ______28 Good Laboratory Practices (GLP) ______29 Plating out ______29 2. Protocol optimisation ______29 Pathogen recognition & acid effect on culture media ______29 Pathogen survival in fermented carrot juice ______30 Cucumber juice mix ______30 Whole carrot community ______30 3. Fermentation and cucumber/cooling ______32 Fermentation set-up ______33

aw ______34 Pathogen inoculation ______34 Cucumber juice and cooling ______34 Fermentation progress analysis ______34 3.5.1. Plating ______34 3.5.2. Freeze-stocks ______35 3.5.3. pH ______35 RNA-based 16S amplicon (V4) sequencing ______35 3.6.1. RNA extraction and analysis______35 3.6.2. Routine DNase treatment ______35 3.6.3. First-strand cDNA Synthesis ______36 3.6.4. Barcoded PCR ______36 3.6.5. PCR clean-up ______37 3.6.6. Qubit ______37 3.6.7. Size selection using Gel-extraction ______37 3.6.8. Illumina ______38 3.6.9. Sequence analysis______38 4. Pathogen detection by enrichment procedures ______39 Detection of Listeria monocytogenes ______40 Detection of Salmonella spp. and E. coli O157 ______40 4.2.1. E. coli O157 ______41 4.2.2. Salmonella spp. ______41 5. Robustness of the spontaneous carrot juice fermentation ______41

Results ______43

1. Protocol optimisation ______43 Effect of acid matrix on the plates ______43 Effect of acid matrix on the counts ______44 Cucumber juice ratios ______45 Original carrot community ______45 2. Fermentation (20°C) ______46 pH evolution ______46

aw ______46 Pathogen development ______47 RNA analysis ______47 2.4.1. Community bar plots ______48 2.4.2. Alpha diversity ______48 2.4.3. Beta diversity ______50 3. Refrigeration (7.5°C) and possible addition of cucumber juice ______50 pH evolution ______51 Pathogen development ______51 RNA analysis and statistics ______52 3.3.1. Community bar plots ______53 3.3.2. Alpha diversity ______54 3.3.3. Beta diversity ______56 Community ______56 4. Enrichment procedures ______57 5. Robustness test ______58 pH evolution ______58 Pathogen development ______59

Discussion ______61

1. Strain selection ______61 2. Protocol optimisation ______61 3. Fermentation (20°C) ______62

4. Refrigeration (7.5°C) and addition of cucumber juice ______67

Conclusion ______71

Further research ______73

References ______75

Addendum ______A

List of Abbreviations ADP Adenosine Di-Phosphate ALOA Agar Listeria Ottaviani & Agosti ANOVA Analysis of Variance ASV Amplicon Sequence Variant ATP Adenosine Tri-Phosphate aw Water activity BC Before Christ BDA British Dietetic Association BHI Brain Heart Infusion BOGK Bundesverband der Obst-, Gemüse- und Kartoffelverarbeitenden BPW Buffered Peptone Water CAJ Carrot Juice CDC Centre for Disease Control cDNA complementary DNA CFU Colony Forming Unit CIN Cefsulodin, Irgasan, Novobiocin CMET Center for Microbial Ecology and Technology CO2 Carbon dioxide CT-SMAC MacConkey Agar with Sorbitol, Cefixime, and Tellurite CUJ Cucumber Juice DADA2 Divisive amplicon denoising algorithm 2 DALY Disability adjusted life years div_inv_simpson Inverse Simpson Diversity DMSO Dimethyl sulfoxide DNA Deoxyribonucleic acid dNTP nucleoside triphosphate DTT dithiothreitol EC European Commission EDTA Ethylenediaminetetraacetic acid EFSA European Food Safety Authority EFSA BIOHAZ EFSA panel on Biological Hazards EHEC Enterohemorrhagic E. coli EPEC Enteropathogenic E. coli ETEC Enterotoxigenic E. coli EURL European Union Reference Laboratory FASFC Federal Agency for the Safety of the Food Chain FCJ Fermented Carrot Juice FCJL2 Fermented Carrot Juice containing L2 pathogens FDA Food and Drug Administration g Gravitational force GAP Good Agricultural Practices GHP Good Hygiene Practices GLP Good Laboratory Practices GMP Good Manufacturing Practices GRAS Generally recognized as safe H2 Hydrogen gas H2O Water HUS Hemolytic-Uremic syndrome ILVO Instituut voor Landbouw-, Visserij-, en VoedingsOnderzoek

L2 Pathogen of L2 safety level LAB Lactic Acid Bacteria LFMFP Lab of Food Microbiology and Food Preservation LOD Limit Of Detection LOQ Limit Of Quantification MM Phusion HF Buffer Master Mix Phusion High-Fidelity Buffer MRS De Man, Rogosa and Sharpe agar NaCl Sodium chloride (Salt) NAD+ Nicotinamide adenine dinucleotide NADH Nicotinamide adenine dinucleotide H NB Nutrient Broth OTU Operational Taxonomic Units PBS Phosphate-buffered saline PCoA Principal Coordinates Analysis PLA Poly Lactic Acid Psi pound-force per square inch RIVM Rijksinstituut voor Volksgezondheid en Milieu RNA Ribonucleic acid RPM Revolutions per minute rRNA Ribosomal ribonucleic acid RT-PCR Reverse transcription polymerase chain reaction RVS Rappaport-Vassiliadis Soya Peptone Broth SciCom FASFC Scientific Committee FASFC STEC Shiga-Toxin producing E. coli TAE Tris-Acetate-EDTA TPP Thyamine Pyrophosphate TSA tryptone soya agar Tukey’s HSD Tukey’s Honestly Significant Difference UA University of Antwerp UK United Kingdom USA United States of America VRBG Violet Red Bile Glucose WHO World Health Organisation WIV-ISP Wetenschappelijk Instituut Volksgezondheid - Institut Scientifique de Santé Publique XLD Xylose Lysine Deoxycholate YOPI Young, Old, Pregnant, Immune deficient YPD Yeast extract Peptone Dextrose

Abstract (EN) Introduction: Fermenting is an age-old process. In the past, it was especially used as a preservation method. Today, many fermentations are also carried out to obtain specific tastes, flavours and possible probiotic features. Not only at household level but also in Michelin-star restaurants, food fermentations are regaining popularity. Therefore, it is of great importance to provide evidence on the safety and robustness of the fermentation process, especially for vegetable fermentations, which seem to be understudied. Purpose: This study evaluated the suppressing qualities of a spontaneous carrot juice fermentation (2.5% NaCl at 20°C) on three food-borne pathogens, and the impact of the presence of these pathogens on the microbial community. Also, the refrigerating effect and addition of cucumber juice were investigated.

Methods: Contaminations were simulated at the start of the fermentation by adding Listeria monocytogenes, Salmonella Typhimurium and Escherichia coli O157 (10³ CFU/ml each). The pathogen development was monitored through time, as well as the pH evolution and the fermenting microbial community (by 16S V4 amplicon sequencing). After 30 days, the fermented carrot juice was mixed with fresh cucumber juice for organoleptic reasons and stored under refrigeration (7.5°C). During this mixing step, also a post-contamination was simulated by adding a ‘fresh’ pathogen cocktail. In a subsequent experiment, contamination at the start of the fermentation was repeated by adding 105 CFU/ml of each pathogen, this to investigate the robustness of the fermentation at higher contamination levels. Results: During the main experiment the fermentation process had a suppressive effect on all pathogens, especially on Listeria monocytogenes, which was non detected (< limit of detection or LOD 10 CFU/ml) after less than 3 days of fermentation (pH 4.6). The other two fell under the detection limit in less than 15 days (pH 4.0). The addition of pathogens did not affect the pH evolution nor the fermenting community (Lactobacillus plantarum dominated among the microbial community after 15 days of fermentation.). The mix of cucumber and fermented carrot juice (pH 4.4) suppressed the presence of L. monocytogenes but the numbers of S. Typhimurium remained constant during refrigeration, also E. coli O157 was still detected after 8 days of refrigeration. Repeating the initial contamination step with higher pathogen levels during a subsequent experiment, showed that the fermentation process still had a suppressive effect on all pathogens. L. monocytogenes did not show growth above 107cfu/ml, while the other two pathogens reached higher amounts, the first 24 hours of fermentation. After 6 days (pH 4.6) L. monocytogenes and S. Typhimurium numbers were below the LOD (10 CFU/ml), E. coli O157 fell under the LOD after 9 days (pH 4.0). Considering the growth (> 7 log) firstly and absence after 9 days, a 6-log reduction could be confirmed for S. Typhimurium and E. coli O157 due to the fermenting conditions. The conditions were probably even more arduous for L. monocytogenes, that is why no growth occurred. Significance: Spontaneous fermentation of carrot juice, for at least 9 days, enables a reduction of L. monocytogenes at least with 5 log. For S. Typhimurium and E. coli O157, a 6-log reduction can be achieved. Fermented carrot juice (as such or mixed with fresh cucumber juice + refrigeration) was shown to be a robust and stable microbial environment not supporting the growth of pathogens.

Abstract (NL) Introductie: Fermentatie is een eeuwenoud proces. In het verleden werd dit vooral gebruikt als conserveermethode. Vandaag de dag, wordt deze ook toegepast om specifieke smaken te bekomen en mogelijkse probiotische effecten. Niet enkel thuis maar ook in Michelin-ster restaurants worden deze fermentaties steeds frequenter toegepast. Door stijgende populariteit is het uitermate belangrijk aan te tonen dat het fermentatieproces veilig en robuust is, zeker op vlak van groentefermentaties, die tot op heden nog ondermaats onderzocht zijn. Doelstelling: De studie evalueert of de spontane wortelsapfermentatie (bij 20°C) een voldoende onderdrukkend karakter heeft op drie verschillende voedselinfectanten. Tevens wordt de impact van de infectanten op de microbiële gemeenschap onderzocht. Methoden: Een contaminatie werd gesimuleerd aan het begin van de fermentatie door het toevoegen van Listeria monocytogenes, Salmonella Typhimurium en Escherichia coli O157 (10³ CFU/ml elks). De ontwikkeling van de pathogenen werd opgevolgd doorheen de tijd, net als het pH verloop. De microbiële gemeenschap werd geanalyseerd via 16S V4 rRNA amplicon sequenering. Na 30 dagen werd het gefermenteerde wortelsap gemengd met vers komkommersap, dit voor organoleptische redenen, en bewaard onder koeling (7.5°C). Tijdens het mengen werd een mogelijkse nabesmetting gesimuleerd door toediening van ‘verse’ pathogenen (10³ CFU/ml). In een volgend experiment werd de contaminatie aan het begin van de fermentatie herhaald, maar ditmaal in hogere concentraties, i.e. 105 CFU/ml elks. Hierdoor werd de robuustheid van de fermentatie nog meer op de proef gesteld, en kon een mogelijkse 6 log reductie aangetoond worden. Resultaten: Tijdens het eerste experiment had de fermentatie een onderdrukkend effect op alle drie de pathogenen, in het bijzonder op L. monocytogenes, deze was afwezig (< detectielimiet 10 CFU/ml) na minder dan 3 dagen (pH 4.6). S. Typhimurium en E. coli O157 vielen onder de detectielimiet na minder dan 15 dagen (pH 4.0). Het toevoegen van de verschillende pathogenen had geen effect op het pH verloop van de fermentatie, noch op de microbiële gemeenschap (Lactobacillus plantarum dominant na 15 dagen). De mix van gefermenteerd wortelsap met komkommersap (pH 4.4) bewerkstelligde afsterving van L. monocytogenes, in tegenstelling tot S. Typhimurium die nog steeds in dezelfde aantallen aanwezig was na 8 dagen koeling. Ook E. coli O157 bleek aanwezig na 8 dagen koeling maar in lage aantallen (bevestigd met aanrijkingsmethoden). Bij het herhalen van het experiment met hogere aantallen aan pathogenen (105 CFU/ml) had de fermentatie nog steeds een onderdrukkend effect op alle drie de pathogenen. S. Typhimurium en L. monocytogenes waren niet meer detecteerbaar met de klassieke telplaten na 6 dagen (pH 4.6), E. coli O157 viel onder de detectielimiet van 10 CFU/ml na 9 dagen fermenteren (pH 4.0). Door groei tot aantallen boven 107 CFU/ml kon een 6 log reductie bevestigd worden voor S. Typhimurium en E. coli O157. Tijdens dit experiment was L. monocytogenes opnieuw het gevoeligst aan het fermentatieproces, en vertoonde geen groei, hierdoor kon enkel een 5 log reductie vastgesteld worden. Impact: Een spontane wortelsapfermentatie, indien het fermentatieproces langer dan 9 dagen verloopt bij kamertemperatuur, zal minstens een 5 log reductie veroorzaken voor L. monocytogenes. Voor S. Typhimurium en E. coli O157 werd in die tijdsperiode een 6 log reductie aangetoond. Gefermenteerd wortelsap op zich, of gemengd met komkommersap bewijst hierbij een stabiele microbiële gemeenschap te bevatten en robuust te zijn met betrekking tot borging van voedselveiligheid waarbij de fermentatiecondities nefast zijn voor de groei/overleving van de drie onderzochte voedselinfectanten. 2

Introduction This research finds its roots, in a Citizen Science project called ‘ Pekes!’ and the doctoral dissertations of Sander Wuyts and Wannes Van Beeck. The project originated at the University of Antwerp by a collaboration between Prof. Lebeer and the Michelin-star chef, Kobe Desramaults, who shared a joint interest in fermented foods as a way to increase people’s contact with relative large doses of beneficial microbes such as lactic acid bacteria (LAB) (Lebeer, 2018). At the start of this collaboration, and also during Ferme Pekes (Wuyts et al., 2018), the main interest was the identification of the different bacteria present in spontaneous fermented carrot juice and the dynamics of the microbial community during the fermentation, with focus on the lactic acid bacteria. Because various questions related to food safety can be raised for spontaneous food fermentations, this Master Thesis aimed to combine the knowledge of the carrot juice fermentation at the University of Antwerp, with the food safety expertise from the Ghent University. This resulted in the current Master Thesis topic which dealt with the study of the growth and survival of pathogens during the carrot juice fermentation and their potential impact on the microbial dynamics during the fermentation. Supplementary, the effect of the addition of non-pasteurised cucumber juice to the fermented carrot juice and further storage under refrigeration was investigated in relation to its effect on pathogen’s presence, survival and composition of microbial ecology. For the main experiment, a carrot juice fermentation was inoculated with a cocktail of three food- borne pathogens (Listeria monocytogenes LFMFP 394, Escherichia coli O157 (-stx gene negative) LFMFP 884 and Salmonella enterica subsp. enterica Typhimurium LFMFP 689). The set-up simulates a possible initial contamination of the fermentation. The development of the pathogens was monitored during the fermentation process, as well as their effect on the microbial community and the pH evolution. Classical plating techniques were used in combination with a 16S (V4) rRNA amplicon sequencing approach. At the end of the carrot juice fermentation (which occurred at 20°C), cucumber juice was added for organoleptic reasons as typically used in Michelin-star restaurants. The cucumber juice was also artificially inoculated with pathogens to simulate a possible post-contamination route. Afterwards, the obtained juice mixtures were transferred from room temperature (ca. 20°C) to the fridge (ca. 7.5°C). The addition of this cucumber juice will increase the pH of the acid (fermented) carrot juice and the cooling will probably inhibit the growth of the present bacteria (e.g. LAB). The latter conditions may enable the pathogens to survive or might facilitate pathogen growth. The monitoring of the pathogens and the microbial ecology of the mixed juice (fresh cucumber juice and fermented carrot juice) will allow us to assess whether the pathogens are still being suppressed under these conditions.

The contamination of three food-borne pathogens at the start of the fermentation was repeated in a subsequent experiment with higher inoculum concentrations to investigate the robustness of the spontaneous carrot juice fermentation process in ensuring food safety.

On the basis of this research, more insights will be gained on:

 The suppressing qualities of a spontaneous carrot juice fermentation on three major food pathogens.  Robustness of the fermented carrot juice towards contamination with food-borne pathogens as a final ready-to-use beverage when challenged with a pH increase (addition of fresh cucumber juice) and temperature decrease (7.5°C in refrigerator).  Effect of the addition of fresh cucumber juice (pH increase) and/or storage at lower temperatures on the microbial community in the final ready-to-use fermented carrot juice or juices’ mix.

The results could lead to a formulation of guidelines or warnings to those who are fermenting vegetables (in specific carrots) without using starter cultures (artisans, restaurants, home produced) to inform them on possible microbial safety issues. Furthermore, from the experiment’s information could be provided on the stability of the fermented juice as such or as the basis in a mix with other fresh vegetables juices.

Literature

The art of fermentation Fermentation of food is one of the oldest food processing techniques (used as preservation method) known to man. All kind of methods for the fermentation of milk, vegetables and meats have been reported, with earliest records dating back to 6000 BC. Food fermentation originated in the so called, 'Fertile Crescent', between the Tigris and Euphrates, present known as Iraq (Fox, 1993). There is a myth about the origin of cheese production for example. The legend says that an Arabian trader travelled through the desert with his milk stored inside a bag made of a sheep stomach. After a while he discovered that his milk had changed into a curd (Choi & Han, 2015). He thus discovered cheese by accident, but in that way, he could store his milk for a longer time, and increased its digestibility. In the past, all processes were artisanal in nature and there was no knowledge about the role of microorganisms. Nevertheless, several traditional methods were developed by which the handling and storage of raw materials were adjusted. The final fermented food also maintained a better quality for a longer time, far superior than the original substrate. There was also an effect on the taste, and other organoleptic characteristics (e.g. smell and texture).

‘The art of fermentation’, was handed down from generation to generation within local communities (Caplice & Fitzgerald, 1999). Traditional fermentation processes remained similar for a long time, until the beginning of the 19th century. At this time, two main events occurred that changed everything. Firstly, the industrial revolution, during which the focus shifted to mass productions and global trade. This led to a decrease of locally produced fermentations. The local trader could not compete against these large-scale industrial economies (Clark, 2010). Secondly, during the same time period the field of microbiology was flourishing, especially in the area of bacteriology, with great names like Pasteur and Koch (Wainwright, 2003). When The global market asked for more consistent quality and uniform products, the microbiology provided the solution in the form of starter cultures for food fermentations (Kotzé, 2003). Nowadays, people are living in a paradox of internationalization, where on the one hand, there is globalisation and international market, but on the other hand there is a mind-set to strive for authenticity (van Ittersum, 2002). Also the awareness and debate about health and sustainability is omni-present (Eurostat, 2017). By fermenting food shelf-life of a substrate will be prolonged and, in that way, reduce food waste. Food fermentations can also be used as a delivery method for probiotics. Probiotics are defined by the World Health Organization as “live microorganisms that, when administered in adequate amounts confer a health benefit on the host” (Hill et al., 2014; Kok & Hutkins, 2018; Şanlier, Gökcen, & Sezgin, 2017). Spontaneous fermentations are easily applicable at a household level and strive in that way to the sense of authenticity. Fermented food fits for those reasons perfectly in the modern-day mind-set of our society. Today the two main types of food fermentations are the ethanol fermentation and the lactic acid fermentation. The production of wine, beer, and other kinds of alcoholic beverages are based on the ethanol fermentation, also called the alcohol fermentation. The lactic acid fermentation is omnipresent within all kind of food substrates common in the western world e.g. cheese, yoghurt,… The list beneath (Table 1) gives a small overview of well-known and less-common fermented food products fermented by lactic acid bacteria (LAB) utilizing the lactic acid pathway. The country of origin and the microorganisms who dominate the fermentation are also indicated (Caplice & Fitzgerald, 1999; Tamang, Thapa, Tamang, Rai, & Chettri, 2015).

Table 1: Overview of different fermented products by lactic acid bacteria (country of origin, dominating LAB & substrate used)

Product Country Microorganisms Substrate Source Cheese International LAB (e.g. Lactococcus lactis) Milk (Caplice & Fitzgerald, 1999) Fermented Europe, USA Lactobacilli, pediococci Meat (Caplice & sausage Fitzgerald, 1999) Idli India Ln. mesenteroides Rice (Caplice & Fitzgerald, 1999) Kimchi Korea Lactic acid bacteria (LAB) Cabbage, (Caplice & vegetables, Fitzgerald, 1999) seafood, nuts Koumiss Russia Lb. bulgaricus, Lb. bucherni mare's milk (Tamang et al., 2015) Mahewu South-Africa LAB Maize (Caplice & Fitzgerald, 1999) Ogiri Nigeria LAB Melon seeds (Tamang et al., 2015) Olives Mediterranean Ln. mesenteroides, Lb. Green olives (Caplice & plantarum Fitzgerald, 1999) Pickles International Lb. plantarum Cucumber (Caplice & Fitzgerald, 1999) Sauerkraut International Ln. mesenteroides, Lb. brevis, Cabbage (Caplice & Lb. plantarum, Lb. curvatus, Fitzgerald, 1999) Lb. sake Shiokara Japan LAB Squid (Tamang et al., 2015) Sourdough Europe, USA, Lb. sanfranciscensis, Lb. casei Rye, wheat (Tamang et al., bread Australia 2015) Yoghurt International Streptococcus thermophilus, Milk, milk solids (Caplice & Lb. bulgaricus Fitzgerald, 1999)

Spontaneous fermentation is art, but can it also comply with food safety (regulations)? This dissertation combines the two aspects, on the one hand there is the fermentation part, where lactic acid bacteria (LAB) will dominate the spontaneous food fermentation process and convert the substrate to a fermented product. And on the other hand, pathogens will be inoculated and simulate a possible contamination route. These two types of microorganisms will compete each other till one group will take the upper hand. A first literature segment will handle the fermentation part, a second segment the food safety. The two parts will be reconciled in a third segment of the literature study.

1. Fermentation

Lactobacillus genus complex An enormous variety of fermented products can be obtained starting from the most different types of substrates all undergoing the same process of a lactic acid fermentation. The microorganisms responsible for the lactic acid fermentation belong to the group of lactic acid bacteria (LAB). This group is very complex in terms of taxonomy and until today adaptations are needed to formulate better relatedness within this group. Lactobacillales is industrially the most relevant order of lactic acid-producing bacteria. The order includes the genus Lactobacillus, as well as other genera like Facklamia, Granulicatella, Leuconostoc, Pediococcus and Streptococcus. All members of this group use carbohydrates as a substrate during the fermentation, which results in the production of lactic acid as a major end-product. Lactobacillus spp. are facultatively anaerobic, generally catalase-negative, gram-positive, non-spore-forming rods, that often grow better under microaerophilic conditions (Goldstein, Tyrrell, & Citron, 2015). Of interest, the host lab (University of Antwerp) could recently show that also catalase-positive species occur in this genus, namely the Lb. casei species (Wuyts et al., 2017). The gram stain morphology can be very versatile (because of the various cell wall thickness), and the colony morphology can also vary from small to medium grey colonies on blood agar. In laboratory conditions, they are generally grown on different media, including MRS (De Man, Rogosa and Sharpe) agar, where they appear as white colonies (Goldstein et al., 2015). Identification of Lactobacillus species is not straightforward by phenotypic meanings, recognition is often unreliable that way. Molecular analyses are necessary for correct identification (e.g. 16S rRNA genes) (Goldstein et al., 2015). Even by molecular meanings, the classification remains a big challenge (Ben Amor, Vaughan, & de Vos, 2007; Wuyts et al., 2017).

In the past, lactobacilli were taxonomically grouped according to their major carbohydrate metabolism. Three groups were formed: group A, containing the homofermentative bacteria, group B, who were facultatively heterofermentative and finally group C, the obligately heterofermentative ones (Claesson, Van Sinderen, & O’toole, 2007). As already mentioned, the accumulation of 16S rRNA gene sequences, and other types of genome analyses led to the realisation that the taxonomy according to phylogenetic groupings were not concordant. The genus is remarkably various, and genome-based analysis was needed (Canchaya, Claesson, Fitzgerald, van Sinderen, & O’Toole, 2006). Phylogenetic trees constructed with increasing numbers of (whole) genome sequences has shown that the Lactobacillus genus is paraphyletic, and all species descend from a common ancestor. Five other genera are also grouped within the lactobacilli as sub-clades, named Pediococcus, Weissella, Leuconostoc, Oenococcus and Fructobacillus. For example, Sun et al. have constructed such a tree by a maximum likelihood constructed from 73 core proteins, shared by 213 genomes (Figure 1). Their tree was supported by high-bootstrap values, to confirm the use of the core proteins as indicators of the evolutionary history of the lactobacilli. Pediococcus, Leuconostoc and Oenococcus have already been recognized as phylogroups within the genus of Lactobacillus for a long time, based on 16S rRNA sequencing and on phylogenomic analysis (Sun et al., 2015). In addition, Sun and co-workers could also add Fructobacillus, located between Leuconostoc and Oenococcus, and Weisella, located as a sister branch, to the Lactobacillus clade. The Lactobacillus clade includes, in other words, six different genera, and was named in an all-encompassing word, the Lactobacillus genus complex (Sun et al., 2015). 7

Classification is a dynamic process and databases should continuously be adapted to optimise their results. For example, work in the host laboratory at the university of Antwerp could recently show that the Lb. casei group consists indeed of three clades (i.e. monophyletic groups): Lb. casei, Lb. paracasei and Lb. rhamnosus. Many isolates were shown to be wrongly added to the databases as Lb. paracasei. It was suggested to reclassify the different names as a uniform group Lb. paracasei, which all contained super oxide dismutase-encoding genes. Only a minority of the genomes, as yet mentioned, the genomes with catalase–encoding genes and a proper GC content at that time should have been annotated as member of the species Lb. casei in the NCBI database (Wuyts et al., 2017).

The Lactobacillus genus is very complex which is why certain scepticism is needed, when analysing the relationships within. Analysis of origin and common ancestors can be done, using the 16S rRNA gene sequence(s). This gene is conservative, and in that way can be used as a long-term indicator. When looking into deeper and much more detailed levels, the 16S rRNA will not be distinguishable, between species. At that moment looking at proteins is a better option, those are more useful for a short-term purpose.

Figure 1: Cladogram of 452 genera from different phyla acquired by sequencing marker-genes, displaying their relation with Lactobacillus. (Sun et al., 2015) Based on the amino-acid sequences of 16 marker genes. The colours in the outer circle represent the phyla indicated in the legend. The branch-tips indicate the position of genera that are most closely related to each other.

Fermentation in general Fermentation is a general term for the anaerobic degradation of glucose or other organic nutrients to obtain energy, conserved as ATP (adenosine triphosphate) without net transfer of electrons (Nelson & Cox, 2013).

Under aerobic conditions, different groups of microorganisms catabolise complex carbohydrates to glucose. This is an important reaction and is needed as a source of energy and key substrates for biosynthesis (e.g. producing five-carbon sugars, to create nucleic acids). The reaction includes a series of phosphorylated sugars. One of those sugars is named pyruvate. There are three pathways bacteria or archaea use to breakdown glucose: first of all, the glycolysis (Embden-Meyerhof-Parnas pathway), 8

secondly the Entner-Doudoroff pathway and thirdly the pentose phosphate pathway. The exact working mechanism of these pathways will not be discussed in detail in this Master Thesis, but is summarized in Figure 2. Of greater importance is what will happen when there is a lack of free oxygen in the environment (Slonczewski & Foster, 2014).

Figure 2: Three pathways of glucose catabolism generating energy (ATP) under aerobic conditions(Slonczewski & Foster, 2014).

In the absence of oxygen or other electron accepting molecules, the pyruvate obtained from the sugar catabolism, must be transformed to an electron-acceptor. The acceptor will receive electrons from NADH (Nicotinamide adenine dinucleotide H), this to restore the electron-accepting form NAD+. Heterotrophic cells will transfer the hydrogens from NADH + H+ back onto the products of pyruvate, and form partially oxidized fermentation products, with the same redox level as the original glucose. To compensate the low yield of energy during fermentation, large quantities of substrate will be consumed, and a lot of fermentation product will be formed. A big advantage of fermentation is the rapid accumulation of acids or ethanol that can inhibit the growth of competitors (Slonczewski & Foster, 2014). As already mentioned, two major types of food fermentations are present, the alcohol (forming ethanol) and lactic acid fermentation. One could further divide the lactic acid fermentation in a homofermentative pathway were only lactic acid is formed and a heterofermentative pathway forming lactic acid and ethanol (+CO2 ). The obtained products will be excreted from the cell and can be useful to human fermentation industries, such as the production of alcoholic beverages, like wine, or lactate fermentation for the creation of cheese (Slonczewski & Foster, 2014).

A question that one could ask is, why do bacteria create waste products, which still contain a lot of energy? Under anaerobic conditions, fermentation products cannot yield energy. No ATP will be generated beyond the substrate-level phosphorylation, where a direct transfer of a phosphate group from an inorganic phosphate to ADP occurs. Much of the energy of the glucose molecule remains unspent or is lost as heat radiation. However, fermentation is more than essential in environments like for example, the animal digestive tract. Conditions in the large intestine are mostly anaerobic, here the bacteria play important roles in nutrient digestion, vitamin synthesis, energy metabolism and immune responses. In return, the bacteria are provided with steady growth conditions and a constant stream 9

of nutrients (symbiotic relationship). Even in aerated cultures, there can be a shift to fermentation, when oxygen levels are running low, often when the demand of oxygen is higher than the oxygen dissolving rate in water.

Fermentation types Microorganisms can ferment different types of substrates including polysaccharides, lipids and proteins that are affected by extracellular enzymes and broken into smaller parts, which are digested by the initial degrader or another microorganism. Fermentable monomers include sugars, on which the focus will be in this Master Thesis, but also polyols, organic acids, even less classic substrates as succinate can be used. Focusing on the fermentation of sugar, the figure below (Figure 3) gives an overview of the major pathways for all types of fermentation of sugars (not only food fermentations) including which organisms implied and end-products formed. The main focus of this Master Thesis will be on the homo- and heterolactic fermentation. Together with the ethanol fermentation it has the most outcome in food applications (Müller, 2001; Tamang et al., 2015).

Figure 3: Different fermentation pathways starting from pyruvate (Müller, 2001) Different fermenting microorganisms are given and their fermentation products

Lactic acid fermentation The lactic acid fermentation (Figure 4) is principally done by lactic acid bacteria (LAB) (Magnuson & Lasure, 2004). Depending on the products formed the LAB are allocated into two different groups, like mentioned in literature section 1.2. Lactic acid fermenters transform pyruvate to lactate, with the help of either of two enzymes ‘L -or D-lactate dehydrogenase’. Depending on the type of microorganism, the lactic acid will have this stereospecificity. Bacterial species belonging to Lactobacillus, Streptococcus, Leuconostoc, and Enterococcus are the most common producers, although fungal strains such as Mucor, Monilia, and Rhizopus also produce lactic acid (John, Nampoothiri, & Pandey, 2007).

Lactic acid shows also a great potential for non-food applications. Today a lot of commotion about sustainability and waste management arises in the media, one of the issues are the huge amounts of plastic waste. Lactic acid can be used to produce a plastic polymer, called poly lactic acid (PLA). This kind of plastic is bio-based and biodegradable, which has, after optimisation, a very high potential in future packaging applications (Peelman et al., 2013).

Figure 4: Biochemical reaction: lactic acid fermentation (Bear et al., 2016)

Some bacteria can also convert pyruvate to a blend of propionate, acetate and carbon dioxide. Also most of the propionic acid bacteria are able to transform the end product of the lactic acid fermentation, called lactate, to propionate (Müller, 2001).

Vegetable fermentation The prime focus of this thesis will be on vegetables undergoing a lactic acid fermentation (e.g. Table 1). The traditional vegetable fermentation process is based on addition of salt and inducing anaerobic conditions, to inhibit growth of detrimental microorganisms, and stimulate the LAB growth. Modifications to this process, by for example adding vinegar, have resulted in a big assortment of ready-to-eat commercial products (e.g. pickles and sauerkraut). The combination of using classical and modern preservation methods, resulted in strict definitions of the different types of fermentation. There is for example a difference between a fermented vegetable, an acidified vegetable and pickles. The definitions by Pérez-Diaz et al. are given underneath. Fermented vegetables: All vegetables that are preserved by fermentation, and is defined as follows: (a) low-acid vegetables subject to the action of acid-producing microorganisms that will naturally achieve and maintain a pH of 4.6 or lower, regardless of whether acid is added; (b) the primary acidulent(s) in the product are the acids naturally produced by the action of microorganisms. If the fermentation proceeds to completion and good manufacturing practices are applied, spoilage organisms capable of raising the pH above 4.6 are prevented from growing in the product, and pathogens of public health significance are destroyed during the process, thus making the final product safe for consumption (Pérez-Diaz et al., 2013).

Pickled and/or pickles: any fermented or acidified vegetable covered with a solution that contains vinegar (acetic acid) as the major acidifying agent (Pérez-Diaz et al., 2013).

Acidified vegetables: products in which an (organic) acid is directly added to preserve any nonfermented vegetable with an initial pH above 4.6, so that the final product pH is maintained below that initial pH, regardless of whether acetic acid is used for acidification (Pérez-Diaz et al., 2013).

As already mentioned, the spontaneous fermentation of vegetable products is liable to the activity of the natural occurring LAB. Also, yeasts and other microorganisms can have an impact on the process, depending on the salt concentration and other environmental conditions. The salt can be added in two forms, and in different concentrations, depending on the desired final product. Salt, often NaCl, can be added in dry form or as a brine. NaCl contains four major roles in the fermentation of vegetables. Firstly, it influences the characteristics of the microbial activity. Secondly it helps at the prevention of tissue softening of the vegetable. Thirdly for flavour enhancing properties. And lastly, supporting the rupturing of the vegetable membranes and in that way allowing the diffusion of all kind of compounds in the brine solution, that stimulate the growth and metabolic activities of the microorganisms (Pérez- Diaz et al., 2013).

One can conclude, traditional fermentation of food is rather straight-forward: add salt to the substrate, create advantageous growth conditions for the LAB (anaerobic conditions, temperature, moisture level) and let the microorganisms do their work.

The microbiota of vegetables is mainly populated by bacteria, more specifically aerobes, like Pseudomonas sp., Enterobacteriaceae and coryneforms. Staphylococci and other faecal bacteria can also be present, but, in normal conditions, i.e. in the absence of highly contaminated sources, those bacteria are supressed by microbial competition and are therefore not a risk for healthy human beings (Di Cagno, Coda, De Angelis, & Gobbetti, 2013). Lactic acid bacteria only make up a minority of the initial population of vegetables. However, vegetables and also their juices will undergo a spontaneous lactic acid fermentation, when applying the right conditions. For these conditions the atmosphere should be anaerobic, also the moisture levels, salt concentration and temperature have to be favourable, in that way the LAB have a competitive advantage and can start to dominate the microbial population of the vegetables (Zabat, Sano, Wurster, Cabral, & Belenky, 2018). Table 2 gives an overview of potential LAB present on raw or spontaneously fermented carrots and cucumbers, since these types of vegetables are used for this Master Thesis.

Table 2: Different lactic acid bacteria associated with carrot and cucumber (based on Di Cagno et al., 2013 & Wuyts et al., 2018)

Substrate LAB

Carrot Lactobacillus plantarum, brevis, vaccinostercus, salivarius, coryniformys, sakei, casei; Leuconostoc mesenteroides; Weissella soli Cucumber Lactobacillus plantarum, pentosus, brevis ; Leuconostoc mesenteroides ; Pediococcus pentosaceus

Globally, most vegetables are still fermented on a small-scale basis, either at home, or by entrepreneurs. The current exceptions of industrial produce are sauerkraut, cucumbers and olives in the western society. Since the breakthrough of kimchi in Asia in the twentieth century, it completes the list as a fourth member (Cheigh, Park, & Lee, 1994). Especially in Asia the fermentation of vegetables is more popular than other parts of the world and different Asian regions have their own

varieties of local produced fermented products (Karovičová & Kohajdová, 2003). These four products are today of significant commercial importance, sauerkraut and kimchi as example will be discussed in more details (Tanguler, Utus, & Erten, 2014; Terefe, 2016). Research into the fermentation of vegetables started off in the early twentieth century, but much of this (fundamental) knowledge is still valid, as few changes have occurred in the commercial preparation of those products.

1.5.1. Sauerkraut In 2017, sauerkraut worth roughly 74.9 million euros was produced in Germany (Statista, 2018). Sauerkraut fermentations are done in bulk fermentation tanks that may contain 100 tons or more of shredded or chopped cabbage, which consists mostly of large heads, typically 3.6 to 4.5 kg (Breidt, Mcfeeters, Perez-Diaz, & Lee, 2013).

1.5.1.1. Production process The substrate of this product is cabbage (Brassica oleracea), the German word literally means acid (sauer) and cabbage (kraut). The cabbage will be dry-salted. The outer green is removed and the cores are cut. The head is sliced into finely cut shreds. Afterwards, approximately 2% of salt is sprinkled on the shreds, for a uniform distribution. Next, the cabbage will be put in big vessels (up to 180 tons), and the top area will be covered. The brine forms by osmotic extraction of water out of the tissues. In a relatively short time fermentation conditions are acquired. Oxygen is depleted and carbon dioxide is formed. The fermentation time can be a few weeks to as long as a year before packing. When the levels of LAB are too high, the brine can be replaced or diluted. In the final product, unpasteurized sauerkraut is packed in glass jars or plastic bags, and stored in the refrigerator. In another approach which is commonly used, sauerkraut is first pasteurized and subsequently stored in cans (BOGK, 2012; Wood, 1997).

1.5.1.2. Microbiology The fermentation consists of two different stages. The initial stage is characterized by a heterofermentative stage, also called gaseous phase, followed by a homofermentative stage, or non- gaseous phase. The first stage is initiated by Leuconostoc mesenteroides, which is initially present in high numbers and has a short generation time, compared to other LAB. Ln. mesenteroides is a member of the heterofermentative LAB resulting in the formation of lactic acid, ethanol and carbon dioxide. The latter replaces the air within the fermentation vessel and creates the anaerobic conditions. The decreasing pH (due to the lactic acid production) and anaerobic conditions result in a transition from Ln. mesenteroides to a dominance of Lactobacillus brevis and Lb. plantarum, which initiates the non- gaseous phase. These dynamics clearly show that proceeding the fermentation under the right conditions, is important for achieving the right flavours and aromas of the end product. For example if the temperature is too high, the homofermentors will develop too fast, resulting in a shorter heterofermentative stage, the proportion of lactic acid will be bigger and the colour, flavour and texture will be poorer (Wood, 1997; Zabat et al., 2018).

1.5.2. Kimchi

1.5.2.1. Production process Cabbage and radish are the most popular substrates, for the production of kimchi, but other vegetables could also be used. The standard method for preparing kimchi is to blend minor compounds, like cereals and fruits, spices, fermented seafood, with major raw materials, mainly vegetables, like 13

cabbage and radish. The blend as a whole, is subjected to a lactic acid fermentation. There are different production processes for the preparing of kimchi, the production of whole-cabbage kimchi will be discussed below as it is the most popular and traditional kimchi in Korea (Ick, 2003; Lee, 1991).

1.5.2.2. Whole-cabbage kimchi First of all, the outer leaves of the cabbage are removed. Next the entire vegetable will be shredded. The obtained sections are soaked in a brine (3% salt), till the parts are softened. Then they are rinsed and drained off. The minor components are mixed and stuffed between the cabbage leaf layers. The stuffed cabbage is packed tightly in a crock. In a commercial plant, the mean duration of fermenting, takes roughly three days (at 20°C), depending on the temperature used. Kimchi can also be fermented at colder temperatures (4°C), this for more than a month to acquire specific taste (Hong, Lee, Kim, & Ahn, 2016). The rate of the fermentation will also be affected by some of the minor ingredients (Ick, 2003; Lee, 1991).

1.5.2.3. Microbiology Since kimchi is a composite product, the microorganisms who affect the fermentation are originated from different ingredients. LAB are mainly involved, with preponderance of anaerobic ones. Bacteria, like Lb. Plantarum, Lb. Brevis, Streptococcus faecalis, Ln. mesenteroides, Weisella and Pediococcus pentosaceus can be found as fermenting microorganisms. In the initial stage of production Streptococcus is prominent, and also the aerobic bacteria occur in high abundances. During the mid- stage, the majority are dominated by Pediococcus. In the late stage, the amount of anaerobic species increases and the lactobacilli affect the ripening of the kimchi. The type of dominant bacteria are strongly dependent of the salt concentration and the fermentation temperature (Jung, Lee, & Jeon, 2014; Lee, 1991).

Fermented vegetable juices Beside traditional fermented vegetables there are also in some extent traditional fermented non- alcoholic or low-alcoholic vegetable-based beverages consumed in European countries and Turkey. These beverages are often homemade, or local commercially available (Table 3). Kraut juice is obtained out of pressed fermented cabbage resulting in a juice where low salt content is considered to have a good taste and be healthier because of the lower sodium content (Baschali, Tsakalidou, Kyriacou, Karavasiloglou, & Matalas, 2017; Karovičová & Kohajdová, 2003). Şalgam juice, also written shalgam, is a highly popular red coloured, cloudy and sour soft beverage famous in Southern Turkey. The production methodology consists of two fermentations: dough and black carrot, where the extracts of the dough fermentation are used as an outset for the carrot fermentation (Tanguler et al., 2014).

Table 3: traditional fermented non-alcoholic vegetable-based beverages (Baschali et al., 2017)

product substrate microbiota Country of Homemade/industrial consumption

Kraut juice cabbage LAB Germany, Ukraine, Homemade + Romania, Serbia industrial

Salgam juice Black carrot, turnip LAB + yeasts Turkey Homemade + small scale Turshiena Hot peppers, LAB Bulgaria homemade chorba horseradish root

Different substrates used for the fermentation are given, also the microbiota responsible for the fermentation, the country of consumption and if the beverages are homemade or made on industrial scale.

Spontaneous vs. Starter cultures From a commercial point of view, the use of a starter culture would lead to a more uniform and valuable product over further-reaching production criteria (pH, taste, texture, …). However, the development of a good starter culture is costly and is it really necessary to replace or predominate the natural flora who is able to dominate the spontaneous fermentation? A lot of research has been done in this area, in addition to preservation and sensory improvements, the starter also can alter the chemical composition and nutritional status of a food (Bonatsou, Tassou, Panagou, & Nychas, 2017; Borresen, Henderson, Kumar, Weir, & Ryan, 2012; Illeghems, De Vuyst, & Weckx, 2015). The use of functional starter cultures is also gaining interest in the research for applications in the food industry. Functional starters are starter cultures that possess at least one inherent functional property that can contribute to food safety and offer organoleptic, nutritional, technological or other beneficial characteristics. One of the best-known usage of starters, is the manufacturing process of yogurt, where the milk is pasteurized by heat treatment, before inoculation. In that way, spoilage organisms are destroyed, the yogurt texture will be improved, due to whey protein denaturation and interaction with the caseins (Corrieu & Béal, 2016). Afterwards two types of bacteria are added, to perform the fermentation: Streptococcus thermophilus and Lactobacillus bulgaricus.

There is also an intermediate form, that could be called a ‘guided spontaneous fermentation’, where one bacterial starter culture is added to the natural substrate, this in excess. This starter culture could influence the innate microbial community. If the strain contains probiotic characteristics, an additional asset could be given to the normal fermentation process. The natural population will not be destroyed, by a pre-treatment. In that way the original texture of the substrate will be maintained (Wuyts et al., 2018). Another way to guide the fermentation is, by adding a small portion of a previous successful fermentation, for example whey or cream. This will serve as inoculum for the new fermentation. This practice is called ‘back-slopping’, and is widely used as a term in fermented sausage manufacture (Mullan, 2001). A comparable technique is used at the olive production, where part of the brine of a previous successful completed fermentation, will be added to the fresh brine of a new one (Bonatsou et al., 2017).

Pasteurising the substrate would eliminate possible unwanted bacteria (e.g. food-borne pathogens) present, but the use of a starter would be necessary since the natural presence of LAB would also be eliminated. Is it really necessary to precede the fermentation process with a pasteurisation from a food safety point of view?

2. Food safety

Food safety in general

2.1.1. Global Food-borne diseases have an enormous impact on the society, which causes a compelling bottleneck in socio-economic development worldwide (e.g. by causing high rates of morbidity and mortality). The full reach and burden of hazardous foods, especially also due to chemical and parasitic contaminants, is still not clear. Significant data and accurate information on the burden of food-borne diseases can appropriately inform policy-makers, and in that way provide the suitable food safety controls and interventions (WHO, 2015). The WHO (World Health Organisation) wrote a report based on 31 food-borne hazards, including eleven diarrhoeal disease agents, seven invasive infectious disease agents, ten helminths and three chemicals. Altogether, these hazards caused 600 million food-borne illnesses and 420,000 deaths in 2010 only (Figure 5). The effects of non-biological hazards, like chemicals, fall beyond the scope of this Master Thesis, as the focus will specifically be on the microbiological hazards, the pathogens, causing food-borne diseases and potential death. Not all individuals are at equal risk of becoming ill from pathogens in their food. There are four sensitive groups clustered together in an acronym, called YOPI. This stands for the young (under the age of five), old (65+), pregnant women and individuals with a weakened immune system or on specific medications. 15 to 20% of the overall population is estimated to belong to the YOPI-group, which is more susceptible to opportunistic food-borne illnesses than the general public (Gkogka, Reij, Gorris, & Zwietering, 2013; Stamey, 2011). The WHO report confirms that the burden of diseases is particularly present for children under five years of age. The vast majority of morbidity was provoked by infectious agents that cause diarrhoeal diseases, in particular norovirus and Campylobacter took one third of all cases. Salmonella Typhimurium accounted for 20% of the invasive infectious disease agents. Most of the mortality cases were caused by non-typhoidal Salmonella enterica, enteropathogenic E. coli (EPEC), norovirus and enterotoxigenic E. coli (ETEC) (WHO, 2015).

Figure 5: Disability adjusted life years for each pathogen acquired from contaminated food (WHO, 2015) The pathogens are ranked from lowest to highest with 95% uncertainty intervals, data from 2010

Most food-borne diseases caused by pathogens were found in African regions, followed by South-East Asian regions and Eastern Mediterranean regions. Often a lack of resources for implementation of suitable food safety controls and interventions is present here. In more developed regions, like Europe and the USA, there is a much better monitoring available and efficient food safety agencies in place.

2.1.2. Europe In 2016, most reported food-borne and waterborne outbreaks from which the cause was known, were linked to bacteria (34% of all outbreaks). The bacterial toxins (18%) were ranked second, followed by the viruses (10%), yet it is important to say that the cause was unknown for 36% of all outbreaks. Salmonella was identified as the most frequently reported agent, it accounted already for 66% and contributed with Campylobacter, for the vast majority of bacterial agents. Most of the associations were made with products of animal origin, like eggs and poultry meat. Also, fish and all kinds of crustaceans ranked high. A major drawback of the food-borne outbreaks surveillance is that for many outbreaks the causative agent stays unknown and no info is available on the suspected food vehicle (EFSA, CDC, 2017).

2.1.3. Belgium In Belgium different competent authorities are involved for the monitoring of food safety and investigation of food-borne outbreaks. The federal agency for the safety of the food chain (FASFC) deals with control on the safety of foodstuffs, monitoring of zoonotic agents (pathogens transmitted from animals via food to humans) from farm to fork and in case of a food-borne outbreak are responsible for inspection and sampling of the premises linked to the food-borne outbreak. The different communities (Flemish-, French- and German-speaking) are dealing with person related matters, e.g. monitoring of human health and epidemiological research by public health inspectors in case of food-borne outbreaks. The Scientific Institute of Public Health (WIV-ISP), nowadays called ScienSano (National Reference laboratory on Food-borne Outbreaks) analyses all suspected food 17

samples in case of a food-borne outbreak and collects all data on food-borne outbreaks and gives scientific support to the FASFC and the Public Health Inspectors. A platform on national scale brings together the different stakeholders (EFSA, 2017).

In 2016, 377 outbreaks occurred in Belgium, with 2154 illnesses and at least 78 hospitalisations, none of the human cases died. Norovirus was the most common causative agent, often detected in oysters, meat or source water. The second most recorded agent was enterotoxigenic Clostridium perfringens, and was found in a lot of stools of human cases. Other causative agents were coagulase positive Staphylococcus aureus (also types producing enterotoxins), E. coli O157:H7 and Campylobacter. Restaurants and take away or fast food outlets were the main locations of exposure (EFSA, 2017).

Food safety of vegetables The world’s largest reported vegetable borne outbreak occurred in Japan in 1996 and of the over 11,000 people affected, about 6,000 were culture confirmed. The outbreak involved the death of three school children and was caused by E. coli O157:H7 (European Commision, 2002). A lot of vegetables are minimally processed and used in ready-to-eat products. Minimal processing is accompanied with risks of possible contaminations in the context of the whole food chain. Pathogenic microorganisms can contaminate fresh produce on different stages (pre -and post-harvest) and this contamination can arise from environmental, animal or human sources (Maffei, Batalha, Landgraf, Schaffner, & Franco, 2016). Each environment from farm to fork represents a unique combination of risk factors that can influence the manifestation of possible pathogens in the chain. The implementation of food safety management systems including good agricultural practices (GAP), good hygiene practices (GHP), and good manufacturing practices (GMP), should always be the prime objective of the producer (European Commision, 2002). Multiple studies isolated pathogens including L. monocytogenes, Salmonella spp. and pathogenic E. coli from fresh and fresh cut vegetables in many countries. Also Norovirus is often associated with food-borne outbreaks of raw or minimally processed bulb and stem vegetables (Abadias, Usall, Anguera, Solsona, & Viñas, 2008; Callejón et al., 2015; Jeddi et al., 2014; Maffei et al., 2016).

The most famous example of vegetable outbreaks of the past years in Europe, was probably the outbreak of EHEC (enterohemorrhagic E. coli) infections observed in Germany, during the summer of 2011. The enormous rate of HUS (haemolytic uremic syndrome), made it the largest HUS outbreak worldwide. The roots of the outbreak strain and how the fenugreek seeds were contaminated remain unclear. The strain of EHEC O104:H4 was never isolated from the sprouts. But after the sprouts had been identified as the vehicle of the outbreak, no further outbreak clusters were found. The outbreak had huge consequences, not only for the patients but also from an economic point of view. The trade in salad and salad ingredients reduced drastically. Also Spanish cucumbers had been appointed by a local health authority as a potential source of the pathogen, with the result of a drop in the export and consumption of Spanish vegetables (Burger, 2012). For carrots specifically, 50 cases of Yersinia pseudotuberculosis occurred in Finland, 145 cases of Shigella were reported in Sweden and 2 cases of Norovirus in Belgium (EFSA BIOHAZ, 2014). In southern Finland over 400 children were affected by the infection of Y. pseudotuberculosis O:1 in 2006. The outbreak was strongly correlated with the consumption of grated carrots. Patients samples as well environmental samples from the carrot distributor showed identical serotypes and genotypes of Y. 18

pseudotuberculosis. The initial source and mechanism of contamination remained unclear, but outbreaks of Y. pseudotuberculosis linked to fresh vegetables are commonly reported in Finland. GHP and GAP instructions will be indispensable, to prevent future outbreaks (Rimhanen-finne et al., 2009).

Food safety of fermented foods Not only fresh produce or alternatively processed foods represent a certain risk for causing food-borne illness. Fermented food products, often considered as safe, are not always as impregnable as thought. There can be contaminants present in fermented foods that cause harm to human health. Multiple groups of contaminants can occur. A distinction can be made between physical-, chemical- and biological types. Different pathogens (e.g. Salmonella spp.) have already been reported in association with fermented foods (Table 4), such as cheese, fermented sausages or fish and fermented cereals. Not only pathogenic bacteria but also toxic by-products of bacterial origin, for example mycotoxins, ethyl carbamate and biogenic amines, can also have an effect on human health. But those types of hazards fall beyond the scope of this Master Thesis. Relevant pathogens associated with fermented food products are Bacillus cereus, pathogenic Escherichia coli, Salmonella spp., Staphylococcus aureus, Vibrio cholera, Listeria monocytogenes, Shigella spp., Campylobacter and for immunocompromised persons the opportunistic pathogens Aeromonas and Klebsiella (Capozzi et al., 2017).

Table 4: Fermented products reported with presence of pathogens (Capozzi et al., 2017)

Product Pathogen Country

Cheese and sausages L. monocytogenes Portugal

Doenjang B. cereus Korea Dry-cured salami E. coli O157:H7 U.S.A. Fresh pressed apple cider E. coli O157:H7 U.S.A. Sausages S. enterica Germany

Different pathogens reported in association with fermented products and the country of where the report was made. A spontaneous fermentation is a complex case, in the sense that, microbial development is desired in the product, making it thus more complicated to limit the bacterial proliferation of undesired types. Supplementary, the rising tendency of artisanal spontaneous fermentations also implies an increased risk, on the level of food safety. Even without starters, fermented foods are by far, the number one example of a multiple barrier approach (also called hurdle approach) to food preservation. The overall antimicrobial effect consists of an aggregate of different factors (so called hurdles). There are microbial effects based on the competition of the original microflora, the acidification and also a contribution of nonmicrobial barriers like water activity reduction through salting. The combination of both, microbial and nonmicrobial factors can significantly enhance inhibition (Adams & Mitchell, 2002). Processing techniques (pasteurisation, ripening, brining, …) generally used with for example fermented and ripened sausages often appear to be effective in pathogen control. Pathogen inactivation at the start of fermentation might be needed as there is evidence that incoming raw materials, can be a source of contamination of pathogens. Moreover, failures during hygiene and cleaning procedures in the production premises can still lead to pathogens entering the production line and be as such

introduced at the start of or during the on-going fermentation process. The contamination level of the pathogens is generally low, and due to the unfavourable conditions of the fermented products (presence of competing LAB or other microflora, adverse physicochemical properties) the pathogens are not able to grow. Therefore, the public health risk associated with the occurrence of the pathogen in fermented foods is rather low, but not excluded. In addition, food legislation requires often the absence of pathogens (in a 25 g multi-unit (n=5) sampling approach) in ready-to-eat products. Thus the producer has to ensure the products are not contaminated by pathogens (Barbuti & Parolari, 2002). Different bacterial challenge tests have already been done on industrial fermented vegetables like sauerkraut, kimchi and olives. Table 5 gives an overview of the different pathogens inoculated on the food products. All lactic acid fermentations had an inhibitory effect on the pathogens and resulted in an elimination of the pathogens under the limit of detection. The fermentation should always be performed properly. Acid-tolerant types of L. monocytogenes were sometimes found, investigating the presence of those strains is excessively important when using an recycling stream of for example the brine (Chang & Chang, 2011; Niksic et al., 2005; Spyropoulou, Chorianopoulos, Skandamis, & Nychas, 2001).

An important side-note to make, is the fact that many fermented products are produced and consumed in developing countries, where reporting systems for food-borne illnesses are often lacking. The available information and data, is largely related to products popular in the developed world, like fermented milks and meats (Adams & Mitchell, 2002).

Table 5: Fermented products studied by challenge tests and the pathogens inoculated

Product Pathogens

Kimchi E. coli O157:H7, S. Typhi and Staphylococcus aureus (Chang & Chang, 2011) Olives E. coli O157:H7 (Spyropoulou et al., 2001)

Sauerkraut L. monocytogenes and E. coli O157:H7 (Niksic et al., 2005)

Pathogens The most prominent food infective bacteria in Western Europe are Listeria monocytogenes, Salmonella spp., Campylobacter jejuni/coli and human pathogenic Shiga toxin-producing Escherichia coli (STEC), also referred as ‘the big four’ (Devlieghere et al., 2016). A food infection results due to intake of pathogen’s viable cells. This in contrast to a food intoxication, where the microbial toxins are the cause of the infection in the gastro-intestinal pathway. A food infection will generally be typed by a low infectious dose (10 till 106 CFU intake, depending on the pathogen) and often induce fever. An intoxication takes only place at higher dosage (at least 105 CFU/g), combined with a shorter incubation period. An example of a food-borne intoxicator is Clostridium botulinum: Although Cl. botulinum toxin production only occur in foods representing strict anaerobic conditions and if the acidification failed and thus the pH is > 4.6 enabling growth to high levels and toxin production (Devlieghere, Debevere, Jacxsens, & Uyttendaele, 2011). In particular for carrots, Yersinia species such as Y. enterocolitica or Y. pseudotuberculosis are also pathogens of concern (all pathogens are summarised in Table 6). Occasionally, some bacteria being part of the natural microbiota of vegetables may also act as opportunistic pathogens.

Table 6: Growth characteristics of different pathogens associated with fermented foods (Devlieghere et al., 2016)

Microorganism Temperature range Minimal pH Minimal aw (°C)

Listeria monocytogenes 0-45 4.4 0.92 Salmonella spp. 8-45 4.4 0.95

STEC O157 and non-O157 8-45 4.4 0.95 Yersinia enterocolitica (-2)-44 4.5 0.96 Clostridium botulinum I1 10-52 4.6 0.935

Clostridium botulinum II2 3.3-45 5.0 0.97

1 proteolytic strains, more heat resistant spores compared to type II 2 non-proteolytic strains, growth at lower temperatures possible

2.4.1. Listeria monocytogenes L. monocytogenes is a gram-positive zoonotic bacterium, which shows much more resistance to adverse conditions (minimal aw, acid pH, low temperature) compared to gram-negative pathogens. This pathogen is psychotropic and very persistent, meaning it can slowly grow in the refrigerator (4°C) and survive deep-freezing. L. monocytogenes is also resistant to brining and can survive in aerobic as well as in oxygen-poor areas. The pathogen can grow up to a lowered aw of 0.92 or up to pH 4.5 (if the other growth factors are optimal). L. monocytogenes is the most heath resistant vegetative pathogen, although a standard pasteurization treatment should be effective for a 6 log reduction (an inactivation usually accepted to render foods safe enough) (Devlieghere et al., 2016).

L. monocytogenes causes listeriosis, which causes fever and muscle aches. During outbreaks, mortality rates can reach to more than 30%. During metastasis to blood -and nervous systems, it can lead to meningitis, which can induce severe brain damage. Vulnerable population groups, the so called YOPI, are extremely sensitive to this pathogen. The pathogen is omnipresent in nature, from the soil to surface water, on plants, but also in the intestines of healthy animals and humans. The presence in food is often caused by post-contamination, so a good cleaning and disinfection procedure of the production area is necessary (Frank Devlieghere, Jacxsens, Uyttendaele, & Vermeulen, 2011; ILVO, 2005).

The microbial guide values for Listeria monocytogenes are: less than 100 CFU/g when the manufacturer can proof this threshold will not be exceeded by growth according to intrinsic and extrinsic factors of the product, i.e. the tolerance value (Commission regulation (EC) No 2073/2005). For example, also applicable in fermented vegetable products with a pH lower than 4.4. If the product supports growth of the pathogen, absence in 25 g at the end of the manufacturing process should be the target (European Commision, 2005; SciCom FASFC, 2017; Uyttendaele et al., 2018).

Carrot root tissue fluid has some antilisterial capacities. The presence of phytoalexins could work as secondary barriers to kill or prevent growth of L. monocytogenes and perhaps other food-borne pathogens and spoilers. Several natural phenolic compounds have been identified in carrots, including isocoumarin, eugenin and others (Beuchat & Brackett, 1990; Beuchat, Brackett, & Doyle, 1994).

2.4.2. Salmonella spp. Salmonella enterica subsp. Enterica includes more than 2000 serotypes which are all possible pathogens. It is a gram-negative zoonotic pathogen, which cannot grow beneath 7.5°C, destruction of the cells (at least 6 log reduction) is possible by heating (2’ 70°C equivalent). Ingestion of 10 cells of some strains, in particular in low moisture foods, can already lead to an approximately 50% chance of illness. Salmonellosis is characterised by fever, stomach-ache and diarrhoea, and can lead to hospitalization especially for YOPI. The most common serotypes are Enteritis and Typhimurium. Food- borne infections are caused by temperature abuse, or cross- or post-contamination. Respecting the cold chain is crucial to avoid Salmonella to grow (Devlieghere et al., 2011). The microbial guide values for Salmonella spp. are: absence in 25 g is always needed for fermented vegetables (SciCom FASFC, 2017 ; Uyttendaele et al., 2018). As already mentioned, Salmonella contains a lot of serotypes. That is why the guide values are applied for Salmonella spp. in general, because it is often a different serotype who causes the infection during an outbreak, also source attribution is difficult (Pires, Vieira, Hald, & Cole, 2014).

2.4.3. Pathogenic STEC (Shiga-toxin E. coli): O157 and non-O157 E. coli is a commensal and casual inhabitant of the intestines, so often it is not considered as a pathogen. E. coli in general is often used as a hygiene indicator (Devlieghere et al., 2011). For generic (non-pathogenic) E. coli, occasionally encountered in vegetables, often a target is put to < 10 CFUs/g, with a tolerance of up to 100 CFUs/g (or for leafy vegetables up to 1000 CFUs/g) (Uyttendaele et al., 2018). Some types of E. coli possess virulent genes and exhibit in that way a pathogenic character. STEC contains stx-genes and often also eae-genes, based on those specific virulence genes a distinction can be made between innocent and pathogenic E. coli. E. coli O157:H7 is the most common serotype associated with outbreaks of bloody diarrhoea and HUS (haemolytic uremic syndrome). STEC has a low infectious dose, that is why the microbial guide values strive to the absence in 25 g (SciCom FASFC, 2017). STEC cannot grow in refrigerated areas, maintaining the cold chain is primordial. To prevent STEC outbreaks, GHP and GAP in the food chain are imperative.

2.4.4. Yersinia enterocolitica and pseudotuberculosis Yersinia enterocolitica is a gram-negative rod-shaped bacterium, belonging to Enterobacteriaceae family. Up to 6 biotypes are distinguished, with over 60 serotypes, but only a few are pathogenic to humans. Pathogenicity is associated with defined types such as Y. enterocolitica serobiotype O:3/4 which carry certain virulence factors. Y. enterocolitica is ubiquitous present in the environment. The pathogen is psychrotropic, and sometimes even shows growth at -2°C, but optimal temperatures are around 30°C. Yersinia spp. are well adapted to cold conditions, but the cells are briskly killed at heating of 65°C and beyond. The symptoms of yersiniosis are fever, heavy stomach-ache and diarrhoea. Also other infections in the intestinal system have been reported, with an aftereffect causing a blood infection (Uyttendaele et al., 2010). Yersinia pseudotuberculosis infections are mainly characterised by fever and acute abdominal pain, caused by mesenteric lymphadenitis. Secondary symptoms are reactive arthritis (Kangas et al., 2008). Outbreaks associated with vegetables and Y. enterocolitica or pseudotuberculosis are sporadically reported, but when reported, often associated with carrots. This pathogen has not been considered in the control program for vegetable products, so no guide values

are present. This is mainly due to the overall low prevalence, the difficulties of differentiation between biotypes and a lack of well-performing culture media for enumeration or detection.

In meat Y. enterocolitica must be absent in 25 g (SciCom FASFC, 2017). Due to the association of Y. enterocolitica or pseudotuberculosis with carrots it could be useful to formulate some warnings or guidelines in the future.

2.4.5. Opportunistic bacteria Different Enterobacteriaceae (e.g. Pantoea agglomerans) are present on vegetables like carrots. (Wuyts et al., 2018) Nevertheless, depending on the type of bacteria, not all are considered as dangerous pathogens. They belong to the class of ‘opportunistic bacteria’ and are only of danger in unfit, vulnerable hosts. Opportunistic pathogens usually do not cause disease in a healthy, immunocompetent host. They take advantage of certain situations, for example, from compromised immune system of patients, which presents an ‘opportunity’ for the pathogen to infect. The YOPI- group should be more careful with this type of bacteria, also when using a neutropenic diet (BDA, 2016; Uyttendaele & Li, 2018). LAB are often considered as overall good bacteria often ‘General Recognized As Safe’ (GRAS). But several intrinsic properties related to their metabolism may have an effect on human safety. D-lactate and also biogenic amines are sometimes produced and can accumulate in fermented products (e.g. Lb. brevis is able to produce biogenic amines and could for that reason be a bad choice as starter culture). Despite the probiotic characteristics and wide spread presence of lactobacilli, some infections are reported (Bernardeau, Vernoux, Henri-Dubernet, & Gueguen, 2008). As a side-note it must be said, also here only adverse effects on public health are described in populations with a reduced immune function. The opportunistic infections may be complicated by antibiotic resistance of some Lb. strains. For this reason antibiotic-susceptibility assays should be performed before technological and functional use (Lebeer, Vanderleyden, & De Keersmaecker, 2008).

3. Why fermented carrot juice & what about the food safety? Spontaneous household food fermentations are regaining popularity among non-professional foodies. Advantageous reasons are post-digestive and other health benefits, but also the richness in flavour and taste (Wuyts et al., 2018).

Carrot acts as a good substrate for fermentation, due to its high amounts of sugars. 100 g of average raw carrots contains 5.6 g of carbohydrates (RIVM, 2018). After fermentation, the fermented carrot juice has a sour smell, and could be considered too acid (pH = 3.5) by consumers, for this reason cucumber juice could be added to alleviate the taste, and make the pH more neutral. Desramaults also uses a mix of different vegetable juices in his restaurant to attain specific tastes. Fermented carrot juice is characterised by a high diversity of LAB, more specific Lactobacillus and Leuconostoc species. Lactobacilli in general can accommodate probiotic characteristics. They contain properties like pathogen inhibition and restoration of microbial homeostasis, enhancement of the epithelial barrier function or the modulation of the immune responses (Lebeer et al., 2008). Nowadays, most of the probiotic foods are commercially available as dairy products (e.g. Yakult). Thus, fermented carrot juice could offer a non-dairy alternative of beneficial bacteria for people suffering from allergies to milk protein or severe lactose intolerance. Fermented carrot juice is also suitable as a model- ecosystem, as the same patterns of microbial ecology are observed, impartial to the origin of the carrots used (Wuyts et al., 2018). The consumer nowadays is more sceptical and has more attention for minimally processed foods and demands a lower amount of non-natural food additives. The food industry and scientific research investigate the application of natural compounds for the processing of food products, in order to eliminate or reduce chemical additives used as antimicrobial agents. The selection of microbial molecules, and/or bacterial strains able to produce such compounds, to be used as antimicrobials and preservatives, proved that lactic acid bacteria could be suitable candidates for such “natural purpose” (Arena et al., 2016). Fermentation has ordinarily a positive impact on the microbial food safety of a product. Due to the fermenting conditions (salt, acid pH, …). LAB develop eminently and can so inhibit the development of possible pathogens. Also, the production of certain organic acids, ethanol and other antimicrobial compounds will have an impact. From a historical point of view, the process of fermentation was used to prolong the shelf-life of a substrate (preservation method). However, a fermentation cannot exclude for sure the presence of pathogens. Certainly, at the beginning of the fermentation the preservative conditions (acid pH, lactic acid production, …) of the fermentation are not yet at their optimum, this is also the case when stored at refrigeration. Since the substrates used (carrots and cucumbers) can host pathogens, possible contamination routes are inevitable. The robustness of the fermentation, and its preserving capacity will be put up to the test. The possible development of pathogens present during and after fermentation will be investigated and also their effect on the microbial community. In that way microbial guidelines or fermentation warnings could be developed for spontaneous carrot juice fermentations as such or used as a basis of vegetable beverages.

Material & methods

1. Bacterial strains & culture conditions

Strain selection The following pathogens were chosen: Listeria monocytogenes, Salmonella enterica subsp. enterica Typhimurium and Escherichia coli O157:H7 (cured from stx-genes and thus a non-pathogenic). They are all part of the ‘big four’ food-borne infective agents, and might be found on carrots due to either faecal or environmental contamination during primary production or further handling. As mentioned already in the literature review, Yersinia enterocolitica and pseudotuberculosis, Clostridium botulinum or Norovirus were not taken up in this challenge test experiments because there are no simple standard well-performing methods for monitoring these pathogens in foods to find low numbers (amongst a lot of competing microbiota) and the time for elaborating such methods in the present Master Thesis work is limited. The strains used during the experiment were obtained from the culture collection of the LFMFP (Laboratory of Food Microbiology and Food Preservation) at Ghent University. Table 7 gives an overview of the different selected strains. Because human pathogenic STEC is a Biosafety class 3 pathogen, handling STEC in a lab or pilot scale setting is not possible. Therefore, non-pathogenic E. coli (without stx-genes) instead of pathogenic STEC may artificially be added to food to be used for process validation. Thus, the non-pathogenic E. coli acts as a so-called surrogate to study the efficiency of an inactivation treatment or preservation process. If the non-pathogenic E. coli does not survive it might indicate that also STEC will not survive.

Table 7: Pathogens used in this Master Dissertation (LFMFP number, isolation source and possible remarks)

Microorganism LFMFP number origin Remarks

Listeria 394 Cheese (Werbrouck et al., 2006) monocytogenes

Salmonella enterica 689 / Streptomycin resistance subsp. enterica (Goudeau et al., 2013) Typhimurium

Escherichia coli O157 884 Bovine Attenuated, kanamycin resistance (Dinu & Bach, 2011)

Stock culture At the LFMFP (Ghent University) cryotubes of the different pathogenic strains were taken out of the freezer (-75°C); Glass beads were transferred to non-selective broth (e.g. BHI, Brain Heart Infusion) and incubated at optimal time and temperature (pathogen dependent). A 4x4 looping out on a non- selective agar plate (e.g. TSA, Trypticase Soy Agar ) was performed, this to confirm the purity of the strain. This was also done on a selective agar plate (pathogen dependent). After incubation at the optimal time/temperature profile, and confirmation of the purity, a single colony of the selective agar plate was transferred to a non-selective plate by a 4x4 looping out and again incubated. A single colony of the non-selective agar plate was transmitted on a slant (TSA) by making a zigzag line, and once again

incubated. If necessary, an extra confirmation could be done. The slant tubes could at that moment be stored for 4 to maximum 6 weeks in the refrigerator (LFMFP, 2018).

Working culture For every pathogen separately, some bacterial colony material was transferred from the slant to a test tube filled with 10 ml nutrient broth (NB). The tubes were placed in the incubator (Memmert IN160) at 37°C for 24 h. After incubation 100 µl was transferred to a new test tube of 10 ml NB, and again incubated at the same time/temperature. After 24 h the amount of pathogens in the tubes was estimated at ± 5×108 CFU/ml (CFU, Colony Forming Unit).

Selective media for pathogens

1.4.1. Agar Listeria according to Ottaviani and Agosti (ALOA) The number of Listeria monocytogenes colonies were determined on Agar Listeria according to Ottaviani and Agosti (ALOA), completed with selective -and enrichment supplements. The selectivity of the medium depends on lithium chloride and a selective mixture of four antimicrobial components including ceftazidime, polymyxin B, nalidixic acid and cycloheximide. The differential activity is linked to the presence of a chromogenic compound X-glucoside, that works as a substrate for the detection of β-glucosidase enzyme, which is common to all Listeria species. The specific differential activity is obtained by the use of a specific substrate (L-α-phosphatidylinositol) for a phospholipase C enzyme that is present in L. monocytogenes. Due to combination of both substrates L. monocytogenes can even be distinguished from Listeria spp.. (Biolife, 2010) ALOA powder (ref 3564043, Bio-Rad Laboratories, France) was weighed and dissolved in distilled water (34.55 g/470 ml, 30 ml of supplements needed for 500 ml). The suspension was heated until complete dissolution, and autoclaved (15’ 121°C, 15psi). Afterwards the solution was brought in a warm water bath at 50°C. When cooled, 2 supplements were aseptically added. Al supplement 1 (ref 3564041, Bio- Rad Laboratories, France) was diluted with 5 ml sterile deionised water using a sterile pipet. Al supplement 2 (ref 3564042, Bio-Rad Laboratories, France) was first pre-heated for 5 minutes at 47°C. Both supplements were added to the medium and mixed. The medium was poured into petri dishes (± 20 ml per dish) and dried by the air until completely dry. Afterwards the plates were stored at 4°C until further use.

1.4.2. Xylose-Lysine-Desoxycholate (XLD) Agar Salmonella Typhimurium was detected by Xylose-Lysine-Desoxycholate (XLD) Agar. Xylose is fermented by practically all enterics except for the shigellae. This property enables the differentiation of Shigella species. Without lysine, Salmonella rapidly would ferment the xylose and be indistinguishable from non-pathogenic species. A H2S-indicator system, is included for the visualization of the hydrogen sulphide produced, resulting in the formation of colonies with black centres (Becton Dickinson, 2013a). Streptomycin was added to the XLD-medium, (1:500 from a 0.1 g/ml solution to a final concentration of 200 µg/ml) and a covering TSA-layer (TSA, Tryptone Soya Agar, CM0131, Oxoid Microbiology Products, UK), made according to the instructions of the producer, on top of the XLD-agar. The Salmonella strain is resistant to the antibiotic (streptomycin), which is added to eliminate growth of interfering bacteria. Due to the antibiotic present and additional stress due to e.g. acidified fermented 26

food matrix conditions, to prevent inhibition of growth of the Salmonella on the selective XLD-medium an extra TSA-layer will give better potential for resuscitation and enable thus to better recover the strain as was proven during this protocol development at LFMFP-UGent by postdoc Inge Van der Linden within her SPF Public Health ‘SEGERI’ project also looking into survival of Salmonella (and E. coli O157:H7) during (dry) storage of seeds for sprouting and subsequently sprouting of the seeds (and having also many lactic acid bacteria present/growing in parallel). TSA is a general, non-selective agar allowing a wide variety of microorganisms to grow. Powder (XLD-Agar, CM0469, Oxoid Microbiology Products, UK) was weighed and suspended in distilled water (53 g/L). The suspension was heated on a magnetic stirrer till boiling point, and afterwards transferred to a warm water bath at 50°C. After cooling (still fluid) the antibiotic was added, and the medium was poured into petri dishes (± 20 ml per dish) and dried by the air. When dry, a thin covering TSA layer was added and dried by the air for 10 min. Afterwards plates were stored at 4°C until further use.

1.4.3. Cefixime Tellurite - MacConkey Sorbitol agar (CT-SMAC) To detect the presence of E. coli O157 MacConkey Sorbitol agar was used. Sorbitol will be fermented by 80% of E. coli types, whereas the O157 does not ferment it. In that way distinction is possible between the types. In 1991, Chapman added cefixime to the medium to inhibit Proteus growth. Also potassium tellurite was added to increase the sensitivity of detection, by inhibiting the growth of different organisms (Biokar diagnostics, 2009). Kanamycin was added to the CT-SMAC-medium, (1:1000 from a 0.1 g/ml solution, to a final concentration of 100 µg/ml) and a covering TSA-layer (tryptone soya agar) on top. The strain used is resistant to the antibiotic (kanamycin), which is added to eliminate growth of interfering bacteria. The concentration was lower compared to the XLD plates since no interfering bacteria were spotted here during preliminary research. Powder (Sorbitol MacConckey Agar, CM0813, Oxoid Microbiology Products, UK) was weighed and suspended in distilled water (51.5 g/L). The suspension was heated on a hot plate till completely dissolved, and autoclaved (15’ 121°C, 15psi). Afterwards transferred to a warm water bath at 50°C. After cooling (down to 50°C), Cefixime Tellurite selective supplement (SR0172E, Oxoid Microbiology Products, UK) was made by resuspending it in 2ml of sterile water. The resuspended Cefixime tellurite was added to the medium alongside the antibiotic. Afterwards the medium was poured into petri dishes (± 20 ml per dish) and dried by the air. When dry, a thin covering TSA layer was added and dried by the air for 10 min. Afterwards, the plates were stored at 4°C until further use.

Media for microbial community

1.5.1. MRS (Man, Rogosa & Sharpe) MRS agar named to Man, Rogosa and Sharpe was used for the isolation of the lactobacillus genus complex (LGC) and gram-positive cocci. Essential nutrients and amino acids are present for growth together with dextrose as energy source. It contains sodium acetate, which suppresses the growth of many competing bacteria. Cycloheximide (0.1 g/L) was added to the medium (out of a 100 g/L stock). It is a highly effective antibiotic with activity against moulds, yeasts, and phytopathogenic fungi. It has been reported to inhibit the synthesis of both proteins and macromolecules, as well as affect apoptosis in eukaryotes (Merck, 2018)(Remel, 2011). 27

Powder (Difco Lactobacilli MRS Agar, ref288210, Becton Dickinson company, France) was suspended with distilled water (70 g/L), after agitation the solution was autoclaved (15’ 121°C, 15psi). When cooled down to 50°C (in a warm water bath) the cycloheximide was added, the agar solution was stirred and poured into plates. After drying the plates were stored in the refrigerator (4°C) until further use.

1.5.2. VRBG (Violet Red Bile Glucose) Violet Red Bile Glucose (VRBG) Agar was used for detecting and enumerating Enterobacteriaceae. VRBG Agar contains pancreatic digest of gelatine as a nutrient source. Glucose functions as energy supply. The bile salts and crystal violet inhibit gram-positive bacteria (Becton Dickinson, 2009). Cycloheximide (0.1 g/L) was added to the medium (out of a 100 g/L stock). VRBG powder (Violet Red Bile Agar with Glucose, X939.2, Roth, Germany) was suspended with distilled water (41.5 g/L), after agitation the solution was autoclaved (15’ 121°C, 15psi). When cooled down to 50°C (in a warm water bath) the cycloheximide was added, the agar solution was stirred and poured into plates. After drying the plates were stored in the refrigerator (4°C) until further use.

1.5.3. YPD (Yeast Extract-Peptone-Dextrose) Yeast Extract-Peptone-Dextrose (YPD) Agar was utilised for detection of possible yeasts. The medium contains only dextrose and salts, the minimal amount of nutrients creates an optimal environment for yeasts to grow, like for example Saccharomyces cerevisiae. For making faster growth possible, protein and yeast cell extract hydrolysates are present (Becton Dickinson, 2009). Chloramphenicol (0.1 g/L) was added to the medium (out of a 100 g/L stock), i.e. a bacteriostatic by inhibiting protein synthesis. YPD-medium powder (YPD broth, X970.1, Roth, Germany) was mixed with bacteriological agar powder (Bacteriological Agar, J637-500G, VWR Chemicals, Belgium) (15 g/L) and suspended in water (50 g/L), after agitation the solution was autoclaved (15’ at 121°C, 15psi). When cooled down to 50°C (warm water bath) the chloramphenicol was added, the agar solution was stirred and poured into plates. After drying the plates were stored in the refrigerator (4°C). Table 8 gives an overview of all different media used during the experiments.

Table 8: Selective media used for different microorganisms of this research and incubation characteristics

Microorganism Selective media Incubation time Incubation temperature (°C) (h)

Listeria ALOA 24 37 monocytogenes Salmonella XLD 24 37 Typhimurium Escherichia coli O157 CT-SMAC 24 37 Lactic acid bacteria MRS 48 37 Yeasts YPD 24 37

Enterobacteriaceae VRBG 24 37

Good Laboratory Practices (GLP) All lab work was done under the laminar air flow when possible, this to avoid unwanted cross- contaminations.

Plating out To neutralise the acidic sample a serial dilution was made in buffered peptone water (BPW, CM0509, Oxoid, Microbiology Products, UK) prepared according to the instructions of the producer. Diluted samples were then plated out on selected media that contained 4 to 5 sterile beads. A volume of 0.1 ml was added to the plate and afterwards shaken to spread the inoculum in a uniform way. Another approach – to obtain a 10-fold lowered detection limit of the plate count method - was to inoculate 1 ml of sample over 3 plates (0.333 ml per plate), and the amount of CFUs (colony forming units) of the 3 plates was added up, whether it was considered as one plate. All plates were put in the incubator at 37°C for 24 h except the MRS plates (48 h) and typical colonies for the pathogen or bacterial group under consideration were counted afterwards.

2. Protocol optimisation

Pathogen recognition & acid effect on culture media The aim of this first test was to recognize the different pathogens on the various media and to investigate the potential change in colour or performance of the culture media used when being inoculated with the acid (ca. pH 3.5) food matrix i.e. fermented carrot juice (FCJ). Fresh cucumber juice (CUJ) and fresh (unfermented) carrot juice (CAJ) were prepared using a commercial juicer (the ‘Juice Fountain Pro, type 843 , Solis’). It was estimated that ±2 kg of carrots and 8 cucumbers were needed for each litre of juice (the exact amount needed can be origin dependent). Also fermented carrot juice (FCJ) (30+ days old, pH = ± 3.5) was provided by the supervisor at UA (remains of prior experiments) to be used in this preparatory test. In total 6 small plastic 160 ml jars with a screw cap were filled with 160 ml of one of the three juices (FCJ, CUJ, CAJ) (2 jars for each juice). A tube of the bacterial working culture contained approximately 5x108 CFU/ml for each pathogen (but exact numbers were defined by plating). A concentration of ca. 103 CFU/ml was required for inoculation. For every type of juice 1 ml of 103 CFU/ml was prepared for the inoculation of the juices (in 160 ml jars) and 1 ml for counting the exact number of inoculum used. For this a working culture was diluted 10³ times in BPW to obtain a concentration of ca. 5x105CFU/ml. Two 160ml jars needed to be inoculated with 10³ CFU/ml, this would be 320,000 CFUs in total volume of the jars. Taking 0.64 ml (Equation 1) from the diluted working culture (1000 times) would contain theoretically this amount of CFUs.

Equation 1: Calculation of volume of working culture needed for preparatory test 2.1: 320,000 퐶퐹푈 = 0.64 푚푙 5푥10 퐶퐹푈/푚푙 A volume of 0.64ml of the diluted working culture of each of the 3 pathogens was taken and put in 1.5 ml sterile eps. The eps were centrifuged at 3,000 g for 10 minutes, supernatant was removed, washed with an identical volume of BPW and again centrifuged (3,000 g for 10’). After removal of the supernatant, the pellet of the first pathogen was resuspended in 2ml of specific juice, and transferred 29

to the following ep, resulting after transferring over all eps in a cocktail of all pathogens. A volume of 1 ml (i.e. 160,000 CFUs) of this cocktail was added to the 160 ml jar containing the corresponding juice, the other ml was used for direct plating, this to count the actual inoculum. For every juice there was 1 jar containing pathogens, and another jar that served as a control (no inoculation with pathogens). Plating of different tenfold dilutions (0 to -3) of the inoculated juice was done on day 0 and 24 h later, the 0 dilutions of non-inoculated juice were also plated. The plates were put in the incubator for 24h at 37°C.

Pathogen survival in fermented carrot juice The aim of this test was to investigate how much time the pathogens would survive, when put directly in acid FCJ (Fermented Carrot Juice), i.e. undergoing an acid shock. This time only fermented carrot juice was used (30+ days old, pH = ± 3.5). The fermented juice was present in a 250 ml glass Weck jar. 103 CFU/ml of the pathogen cocktail was prepared as described above and inoculated. Samples were taken at different time points (0, 30’ and 1 h) and plated out on selective media. The plates were put in the incubator for 24 h at 37°C and were counted afterwards.

Cucumber juice mix This experiment was set up with the aim of finding an optimal mixture of fermented carrot juice (FCJ) (30+ days of fermentation) and non-pasteurised cucumber juice (CUJ). The mixture was made after the carrot juice fermentation has finalized and can be used before serving to consumers (e.g. in restaurants). The mix could also be stored under refrigeration for later consumption. However, we needed to find the optimum mixture for organoleptic reasons (i.e. to render a less acid smell). Therefore, different mixtures of fermented carrot juice (FCJ) (30+ days) and non-pasteurised cucumber juice (CUJ) were made.

Falcons (50 ml) were filled with different ratios as mentioned in Table 9 beneath. The falcons were put in a 20°C temperature room and the pH was measured at different time points: day 0, 3, 7 and 11 (Mettler Toledo SevenCompact pH meter) and judged by smelling, by doing this at room temperature the acidification of the cucumber juice was initiated, and the limit of acid smell was tested.

Table 9: Different ratios of fermented carrot juice - cucumber juice mixes

Ratio (% FCJ/% CUJ)

Combination 1: 50/50 Combination 2: 75/25

Combination 3: 100/0 (only FCJ)

Combination 4: 0/100 (only CUJ)

Whole carrot community To have a broad idea of the natural community present on the carrots (in solid form) used for this fermentation, two methods were used for sampling and plating out to determine the groups of microbes present in the microbial community in this research (MRS, VRBG, YPD). Sampling a solid substrate is not straightforward, for this reason two different approaches were used.

First a sample was taken using a sterile filter paper dipped in sterile PBS (Phosphate buffered saline), and rubbed with a pincer over a carrot. This was repeated for three randomly picked carrots from the batch. Each filter was transferred to a 50ml falcon, containing 2 ml sterile PBS. The falcons were put for 30 minutes on a shaker (KS 260 basic, IKA) at 250 RPM. All PBS from the different falcons was brought over in one falcon and mixed. 1 ml PBS was spread over 3 plates (330 µl per plate), and this for YPD, VRBG and MRS. The plates were put in the incubator at 37°C for 24 h, with exception of MRS (48 h). During the second sampling method, pieces of different carrots were put in a sterile and waterproof bag. BPW was added till the pieces were floating (± 1.5 times the carrot mass.). The bag was fold over, but not airtight, in a plastic recipient and put in the incubator at 37°C for 18 h. Afterwards 0.1 ml was spread over MRS, VRBG and YPD (the dilution series depended on the media). The plates were put in the incubator at 37°C for 24 h and MRS for 48h (EURL, 2012).

3. Fermentation and cucumber/cooling Figure 6 provides an overview of the overall experimental set-up. For the main experiment, some of the jars with carrot juice were inoculated with a cocktail of three food-borne pathogens (room temperature (20°C) ; row A, B and C) before being subjected to fermentation. The other jars (room temperature (20°C); row S) served as a control.

Figure 6: Experimental set-up of the main experiment The figure is divided into two parts, the fermentation part at room temperature (left) & the refrigeration part (right). The different jars analysed at a certain time point are placed vertically underneath the day. Series A was used for the analysis of the fermentation part of the experiment, series B was used as starting material for the refrigeration part of the experiment. The red pathogens indicate that the carrot juice was inoculated on day 0 (before the fermentation started), in contrast to the blue pathogen dots indicating that pathogens were inoculated only at day 30 (at the end of fermentation, at the time of mixing with cucumber juice). Row S stands for spontaneous fermentation, which means the Weck jars were not inoculated with pathogens, this also applies the series B in the same (red) frame. The jars in the other (red) frame (i.e. spontaneous + pathogens) were all inoculated at day 0 with pathogens. Jars on row A, B and C were used as triplicates being sampled at each sampling day during the fermentation part of the experiment (note per day a different Weck jar was available for sampling because if samples needed to be taken the anaerobic conditions were destroyed). After 30 days CUJ was added to part of the FCJ samples, resulting in row 1, 3 and 4 of the refrigeration part. The FCJ – CUJ ratio was always 25% - 75% respectively. CUJ = cucumber juice, FCJ = fermented carrot juice

The development of the pathogens and microbial community were analysed at different time points: day 1, day 3, day 15 and day 30. Classical plating techniques were used in combination with 16S rRNA amplicon sequencing (V4 region) at every time point, and also the pH was measured.

At the end of the fermentation (day 30), fresh unpasteurised cucumber juice was added to a part of the pathogens’ inoculated fermented carrot juice samples and to some of the control (uninoculated) fermented carrot juice samples for organoleptic reasons and further stored under refrigeration (to avoid spoilage or further fermentation). The cucumber juice also contained freshly inoculated

pathogens in some samples to simulate a possible cross-contamination route occurring during manual handling of the fermented juice during mixing. These latter samples were subjected to pathogen enumeration and detection by enrichment (M&M section 4) to see whether freshly inoculated pathogens could survive or even proliferate (e.g. L. monocytogenes being a psychrotrophic organism) during storage under refrigeration (row 1 refrigerator). The other jars of fermented carrot juice were stored as such under refrigeration (without mixing with cucumber juice) (row 2, 5 Figure 6) and were used to monitor any changes in microbial community of the fermented juice due to transfer to refrigeration (row 2 Figure 6) & those that were initially inoculated with pathogens before fermentation were further subjected to pathogen detection by enrichment (M&M section 4 below) to reveal whether any low numbers of pathogens could still be present and could be resuscitated during storage under refrigeration (row 5 Figure 6). All the jars with either fermented carrot juice or as a mixture of fermented carrot juice with the fresh cucumber juice were thus kept in a temperature-controlled fridge at aiming 7.5°C (mean value = 7.478 °C ± 0.005 °C, Figure 18, addendum), and analysed on timepoints of day 31 (day 1 under refrigeration) and day 38 (day 8 under refrigeration, assumed to be the reasonable maximum shelf life), with the same plating and molecular techniques as mentioned before. Per sampling point four jars (triplicate inoculated + blank uninoculated) were taken for sampling and analysis (series A, Figure 6). Thus, in total 16 jars were filled for series A with fresh carrot juice at day 0 and subjected to fermentation (at ambient temperature, in a temperature-controlled dark room at 20 °C). Another 11 jars (5 inoculated with pathogens and 6 not) were also filled with fresh carrot juice, they were all subjected to a fermentation of 30 days and were used as a starting substrate for the refrigeration part of the experiment.

Fermentation set-up First all 27 (i.e. 16 + 11) glass Weck jars (250 ml) and the kettle that were used to collect the carrot juice from the commercial juicer were cleaned by rinsing with boiled water. Preparing the carrot juice the commercial juicer ‘Solis juice fountain Pro’ was used. In total 6.75 L fresh carrot juice was needed for this experiment. The carrots (without foliage, ± 20 cm) were bought in the supermarket. They were rinsed with tap water (no peeling occurred) and the ends were cut off. It was estimated that 2 kg was needed for 1 litre of juice. All carrot juice was assembled in a big kettle and 2.5% of salt was added. Afterwards, 250 ml was transferred to every jar (Figure 7), the lids were closed and the appropriate labelling was done.

Figure 7: Weck jar containing fermenting carrot juice

Knowing the amount of added salt, the aw could be calculated with Equation 2:

Equation 2: calculation of aw of salted carrot juice

aw = 1.00424 - βm with β = 0.702 (β is the negative of the slope of the plot aw vs. m (g NaCl/g water)) and the intercept of the regression is an approximation of 1 considered for pure water (Samapundo et al., 2010).

Pathogen inoculation In total 17 jars were inoculated with pathogens (spontaneous fermentation + pathogens, Figure 6) and an extra inoculation volume was made to enumerate the inoculum. 19 mL of pathogen mix was prepared, needed for a final concentration of 103 CFU/ml. If this would be used for 19 jars (250 ml), 4,750,000 CFU would be needed. The working culture of 5x108 CFU/ml was 100 diluted with BPW (i.e. 5x106 CFU/ml).

Equation 3: Calculation of volume of working culture needed for main experiment 4,750,000 퐶퐹푈 = 0.95 푚푙 5 푥 10 퐶퐹푈/푚푙 From the working culture (of each pathogen) 0.95 ml (Equation 3) was transferred in a sterile 1.5 ml collection tube. The eps were centrifuged at 3,000 g for 10 minutes. The supernatant was removed, washed with the same volume of BPW and again centrifuged (3,000 g for 10’). After removal of the supernatant, 2 ml of salted fresh carrot juice was added in one ep, mixed and transferred in the next ep, repeated till a cocktail of all three pathogens was acquired. The cocktail was added in a falcon (50 ml) containing 1 ml of every jar (total of 19 ml). The falcon was vortexed, and 1 ml was put over in every jar. From the remaining volume 1 ml was 200 times diluted in BPW and 0.1 ml was plated in duplicate on the different selective media for pathogens. The plates were transferred in the incubator for 24 h at 37°C, and the glass Weck jars were put in a thermostatic room at 20°C.

Cucumber juice and cooling The jars from series B (Figure 6) have been subjected to a 30-day fermentation, whether or not inoculated with pathogens at day 0. The fermented carrot juice (FCJ) from the glass Weck jars of series B was transferred to 160 ml plastic jars with a screw lid. FCJ (40 ml) was added to the plastic jars containing 120 ml cucumber juice (CUJ) (refrigeration, row3 and 4, Figure 6), and 160 ml FCJ was added in plastic jars without CUJ (refrigeration, row 2 and 5 Figure 6). The plastic jars from row 1 on Figure 6 (6 in total) containing non-pathogenic FCJ were freshly inoculated with a pathogen cocktail , using the same protocol as described in M&M section 3.3Pathogen inoculation. All 30 plastic jars (5 rows of 6 jars, Figure 6) were put in the refrigerator at 7.5°C. The temperature of the refrigerator was measured with a logging thermometer (Figure 18, addendum) this to check possible fluctuations during storage.

Fermentation progress analysis

3.5.1. Plating Different plate counts were performed as described above in M&M section Plating out1.7. The presence of pathogens was measured on XLD, CT-SMAC and ALOA plates. For timepoints day 31 and 38 plating was also done on MRS to assess the presence of the LGC (i.e. Lactobacillus genus complex).

3.5.2. Freeze-stocks Falcons (50ml) were filled with a certain amount (± 30 ml) of each sample, and kept in the freezer at -24°C. As of day 15, 25 ml sample was diluted with 25 ml BPW, this too acquire a more neutral solution when stored. The stocks were used when applying enrichment methods (M&M section 4). Cryotubes were also filled with 100 µl of sample and 100 µl glycerol. The tubes were stored at -80°C.

3.5.3. pH 30 ml sample was put in a 50ml falcon. The pH was measured with a Mettler Toledo SevenCompact pH meter. If measuring the pH directly from the sample, no further analysis could be done, due to possible cross-contamination.

RNA-based 16S amplicon (V4) sequencing

3.6.1. RNA extraction and analysis The RNA extraction was done using the QIAGEN RNeasy PowerMicrobiomeTM Kit (50), following the Quick-Start Protocol (February 2017). This kit was chosen due to its double lysis step, and in that way more chance to acquire more representative RNA. It contains a mechanical lysis, with the help of PowerBead Tubes and also a chemical lysis by adding solutions to the sample. The protocol was executed using 200 µl sample. The final elution step was performed using 100 µl RNase-free water which was added to the centre of the filter membrane (RNA yield maximalisation). The RNA samples were put on ice and labelled. The total amount was measured using the bioscreen (take3) (Synergy HTX Multi-Mode Microplate Reader, Biotek) this by adding 2 µl of each RNA-sample in the different pits. Afterwards the RNA was stored at -80°C till further analysis.

3.6.2. Routine DNase treatment For 50 µl sample RNA, 0.1 volume of 10X TURBO DNase Buffer (i.e. 5 µl) and 1 µl of TURBO DNase Enzyme was added to the 96-well plate containing the samples (TURBO DNA-free ™ Kit, AM1907, ThermoFisher scientific, Lithuania). The mix was spun down to avoid contamination (max 300 g) and incubated for 30 minutes at 37°C. After incubation, 5 µl of DNase inactivation reagent was added and the plate was centrifuged till 300 g. During 5 minutes of incubation (at room temperature) the plate was flicked every minute to disperse the DNase inactivation reagent. Afterwards the 96-well plate was centrifuged for 10 minutes at 2,000 g. The supernatant was transferred to a new 96-well plate.

A 16S rDNA PCR with DNA-dependent primers was performed to check for leftover contaminating DNA in the samples. Table 10 gives information about the content of the mastermix.

Table 10: mastermix components and volumes for 16S rDNA colony PCR samples

Component Volume (µl)

10X VWR buffer 2.5 dNTPS (10 mM) 0.5 Forward primer (10 mM) 1

Reverse primer (10 mM) 1 Taq polymerase 0.2

H2O 9.8

TOTAL 15 10 µl of template was mixed with 15 µl of mastermix. A PCR was performed comprising next program: initial denaturation at 95°C 2 min, 25 cycles of 95°C for 20 s, 55°C for 20 s, 72°C 1 min, final extension at 72°C for 10 min. Electrophoresis was performed on a 1% agar gel (50 ml TAE-buffer + 5 µl GelRed) for 30 minutes at 100 V (5 µl sample + 1 µl loading dye)(TAE = Tris base, acetic acid and EDTA).

3.6.3. First-strand cDNA Synthesis The use of the sequencing pipeline required the conversion of the resulted pure RNA to cDNA using Superscript III reverse transcriptase, in short: 1 µl primer (2 mM) was added with 1 µl dNTP and 11 µL RNA-sample (total of 13 µl). The mixture was heated to 65°C for 5 minutes and put on ice for 1 minute. After a short spin (300 g), 4 µl 5X first-strand buffer was added together with 1 µl (0.1 M) DTT, 1 µl RNase OUT Recombinant RNase inhibitor and 1 µl SuperScript III RT to the samples (SuperScript™ III First-Strand Synthesis System, catalogue number: 18080051, Invitrogen, USA). The 96-well plate was vortexed and incubated for 60 minutes at 50°C. To inactivate the reaction the plate was heated to 70°C for 15 minutes.

3.6.4. Barcoded PCR The following mastermix (Table 11) was made with Phusion High-Fidelity DNA polymerase (13 µl per sample).

Table 11: mastermix components and volumes for one barcoded PCR sample

Component For 1 reaction (µl)

5X MM Phusion HF Buffer 4

Phusion DNA pol 0.2 10 mM dNTP’s 0.4 DMSO 0.6 Molecular grade water 7.8

TOTAL 13 The mastermix was added to a 96-well plate (top and bottom row were left empty, to avoid sealing problems). 1 µl of 10 µM barcoded reverse primer (Kozich, Westcott, Baxter, Highlander, & Schloss, 2013) was added to each well. The same was done for the barcoded forward primers. cDNA (5 µl) was 36

also added, after this the 96-well plate was sealed with a PCR sealing. After a short spin (300 g) the same PCR as mentioned above was set up (M&M section 3.6.2). Electrophoresis was performed for 28 random samples on a 1% agar gel (50 ml TAE-buffer + 5 µl GelRed) for 30 minutes at 100 V (2 µl sample

+ 3 µl H2O + 1 µl loading dye). This to check if DNA of the right size was amplified.

3.6.5. PCR clean-up The products were purified by a Agencourt AMPure XP PCR Purification (Shake the Agencourt AMPure XP bottle to resuspend any magnetic particles that may have settled.). 1.8 μL AMPure XP per 1.0 μL of sample was added, by pipetting up and down 10 times. The 96-well plate was placed onto a Super Magnet Plate for 10 minutes to separate the beads (containing the DNA) from the solution. The supernatant was removed, and the beads were washed with 70% ethanol (120 µl). After 30 seconds the supernatant was removed and the washing step was repeated. The remaining ethanol was sucked away and the wells were let to evaporate the residual ethanol for 5 minutes. The 96-well plate was removed from the magnetic plate and magnetic beads were resuspended in 40 µl of dH2O. The suspension was incubated for 2 minutes and put for 1 minute back on the magnetic plate. The supernatant was transferred to a new 96-well plate.

3.6.6. Qubit N times 1 µl of Qubit reagent (Qubit dsDNA HS Assay Kit, REF Q32854, Invitrogen, U.S.A.) was mixed with N times 199 µl of Qubit buffer to create a working solution for N samples. 2 standards were made by mixing 190 µl of the working solution with respectively 10 µl of the standard. 1 µl of sample was added to 199 µl of working solution. The samples were put in the Qubit to measure the DNA concentration. According to the Qubit results, all samples were brought in one library by bringing them on equimolar concentrations. Arbitrary the lowest concentration was chosen and also the max volume to add.

3.6.7. Size selection using Gel-extraction The library was loaded in quintuplet on a 0.8% agarose gel (50 µl library + 10 µl loading dye). Electrophoresis was performed at 60 V for 50 minutes. Afterwards the bands were cut out and weighted. The DNA was extracted using NucleoSpin® Gel and PCR Clean-up from Macherey and Nagel using the standard protocol, in short: For each 100 mg agarose gel, 200 µl of Buffer NTI was added. The tubes were incubated for 10 minutes at 50°C, the samples were vortexed every 2 minutes for complete dissolution. 700 µl was loaded on a NucleoSpin® Gel and PCR Clean-up column and centrifuged at 11,000 g for 30 seconds. This step was repeated till all sample was loaded on the column. Afterwards, 700 µl of Buffer NT3 was added to the column and also centrifuged at 11,000 g for 30 seconds. The washing step was repeated a second time. next, the column was centrifuged at 11,000 g for 1 minute to remove residual buffer. At last, after transferring the column to a new tube, 20 µl of buffer NE was added. This was incubated for 1 minute at room temperature, and centrifuged for 1 minute at 11,000 g. The concentration of the acquired DNA was measured with the Qubit and after calculation diluted to 2 nM (Equation 4).

Equation 4: dilution of DNA to 2nM

. µ∗ = 24.27 nM  12 times dilution needed (in dH2O). ∗

3.6.8. Illumina A sample sheet was made for the Illumina MiSeq run and the library was brought to a sequencing facility (i.e. centre for medical genetics, University of Antwerp).

3.6.9. Sequence analysis The goal of this technique is to find which bacteria are present in the samples and in what amounts (relatively). This by amplifying and sequencing a small genomic region (i.e. amplicon). The V4 (± 250 base pairs) region of the 16S rRNA gene was used. All bacteria have this gene (16S), it has conservative regions (no horizontal transfer, good primer attachment) and variable regions (V-regions). This structure makes the gene a good target as universal primers could easily be constructed on the conservative regions with the amplicon spanning the variable regions. Over the last decade large databases were also developed making it more feasible to compare the results. After the sequencing run (MiSeq), sequencing reads were obtained (fastq files). To achieve only high-quality reads, some computational steps were done, where different quality controls were applied. Table 12 gives an overview of the different quality control steps. Using ASVs (amplicon sequence variant) has a better resolution as OUT (operational taxonomic unknit) clustering (going sub genus level is possible), but species level cannot be attained. Often an idea of possible species or related ones are given. This method has more a discriminatory purpose, if whole genome sequencing is necessary, shotgun sequencing could be applied. But for community analysis 16S amplicon sequencing is already very useful. The fastq files from the sequencing facility (MiSeq, Illumina) were processed by S. Wittouck using the DADA2 (i.e. Divisive amplicon denoising algorithm) pipeline (Table 12) (Callahan et al., 2016).

Table 12: DADA2 pipeline (Callahan et al., 2016)

0. Demultiplexing 1. Trimming 2. Quality filtering Forward and reverse reads separately

3. Dereplication 4. Denoising

5. Merging of forward and reverse

6. Chimeras removal

7. Classification

8. Non-bacterial removal Merged reads

9. Contamination removal (ASVs, amplicon sequence variant)

10. Wrong size removal

The final product of this DADA2 pipeline is list of 3 data frames, the first is a data frame in which each row corresponds to a processed sample, and each column corresponds to an non-chimeric inferred sample sequence (a more precise analogue to the common “OTU table”). The second data frame consists of taxonomical information of each denoised read also called amplicon sequence variant. The 38

final data frame consists of additional experimental data of each sample (metadata, e.g. pH , timepoint, jar). From here further analysis was done using R (version i386 3.5.1.) using the tidyamplicons package developed in house (www.github.com/swittouck/tidyamplicons). Data was used after a quality control (last three points of the pipeline) checking the taxon size, removing non-bacterial data and possible contamination (Jervis-Bardy et al., 2015). Different community compositions were visualised using bar plots containing relative abundancies of the top 11 most abundant ASVs in the different samples. Alpha diversity was analysed using inverse Simpson diversity and displayed with ggplot2 (Wilkinson, 2011). Beta diversity was analysed using the Bray-Curtis diversity index and visualised as a classical multidimensional scaling, known as Principal Coordinates Analysis (PCoA), also displayed with ggplot2. On different diversity data a Kruskal test and a possible subsequent Dunn’s test was performed to test for significant differences. Apart from this also a one-way ANOVA (Analysis of variance) was performed on the growth data of the pathogens during fermentation, with a subsequent Tukey’s HSD test (Honestly significant difference).

4. Pathogen detection by enrichment procedures The fermentation set-up was done in the lab of environmental ecology and applied microbiology at the University of Antwerp. Also, the pathogen enumeration at every time point during fermentation and RNA extraction were performed here. From different samples of the main experiment (all of the fermentation series A inoculated with pathogens and row 1 and in duplicate row 4 from the refrigeration part, Figure 6) freeze stocks were made in 50ml falcons filled with fermenting carrot juice and 50 V% of BPW to temper the acid pH. Those were kept at -24°C till the end of the main experiment. Afterwards, the samples (day 1 were frozen for almost 2 months – day 30 for almost 1 month ) were taken all by car (ca. 2h frozen storage transport) to the lab of food microbiology and food preservation at Ghent University. Here, enrichment procedures were carried out, because the majority of the selected samples contained low or absent levels of pathogens, when plated out during the main experiment. This to prove presence/absence of the pathogen in 25ml of fermented carrot juice diluted with 25ml BPW. The samples were thawed at room temperature on the counter top at Ghent University and all analysed in parallel using enrichment procedures. The carrot juice samples inoculated with pathogens (day 0) and subjected to spontaneous fermentation at ambient temperature were analysed by enrichment at all time points considered (day 1, 3, 15 and 30) .

Also, the fermented carrot juice samples, prior to fermentation not inoculated, but freshly inoculated on day 30 when mixing with the non-pasteurised cucumber juice occurred were analysed by enrichment during storage under refrigeration (day 31 and day 38, row 1, Figure 6). Furthermore, also the fermented carrot juice samples that were prior to fermentation inoculated (day 0) and mixed on day 30 with non-pasteurised cucumber juice were analysed by enrichment during storage under refrigeration (duplicate samples were available and tested on day 31 and day 38). Thus a total of 22 samples (Figure 8) were subjected to enrichment procedures and thus presence/absence testing for S. Typhimurium (M&M section 4.2.2), E. coli O157 (M&M section 4.2.1) and L. monocytogenes M&M section 4.1.

Figure 8: samples used for enrichment procedures The samples comprised in the purple ovals undergone the enrichment procedures

Enrichment was performed to assess presence/absence of S. Typhimurium, E. coli O157 and L. monocytogenes in 10 ml.

Detection of Listeria monocytogenes For the detection of L. monocytogenes 25 ml of the thawed sample (as of day 15 diluted with 50% BPW, i.e. 12.5 ml of the original carrot juice) were weighted and diluted with 225 ml demi-Fraser broth (demi-Fraser bags, REF 42 727, bioMérieux, France). If less volume was available, 1 volume of sample was diluted with 9 volumes of demi-Fraser (at least 20 ml used to confirm presence/absence in 10 ml, since 50% diluted in BPW). As recipient stomacher bags were used and heat sealed after filling. The bags were put in the incubator for 24 hours at 30°C. After incubation, 0.1 ml from the primary dilution was transferred to 10 ml Fraser broth tubes (Fraser bouillon, REF 42072, bioMérieux, France), vortexed and again incubated, for 24 h at 37°C now.

The secondary enrichment was plated out on ALOA plates (0.1 ml) and incubated for 48 hours at 37°C. This to qualitatively check the presence of the pathogen. After 24 hours the colonies can still be too tiny to see with the naked eye, so 48 h was chosen

Detection of Salmonella spp. and E. coli O157 For the detection of Salmonella or E. coli O157, 25 ml of the thawed sample (as of day 15 diluted in 50% BPW) were weighted and diluted with 225 ml BPW (Buffered Peptone Water, CM0509, Oxoid Microbiology Products, UK). If less volume was available, 1 volume of sample was diluted with 9

volumes of BPW (at least 20 ml used to confirm presence/absence in 10 ml, since diluted in BPW). As recipient stomacher bags were used and sealed after filling. The bags were put in the incubator for 18 hours at 37°C.

4.2.1. E. coli O157 After the first enrichment E. coli O157 was directly streaked on a CT-SMAC plate, and incubated for 24 hours at 37°C. To check if the presumptive E. coli O157 colonies present on the plate. Where E. coli O157 present, an agglutination test was performed using E. coli O157 LATEX Test (REF DR0620M, Oxoid Microbiology Products, UK). Some colony material of a suspect CFU was mixed with a droplet of saline, this droplet was mixed with another droplet containing a test latex. If within a minute coagulation was spotted, the CFU tested positive for O157. The same CFU would be analysed with a control latex in the same way, if this turned out negative (i.e. no coagulation), the test latex result was confirmed.

4.2.2. Salmonella spp. After the first incubation, 0.1 ml from the primary dilution was transferred to 10 ml RVS broth tubes (Rappaport Vassiliadis Soya, REF 3555773, BIO-RAD, France), vortexed and again incubated, for 24 h at 41.5°C now. The secondary enrichment was put on XLD plates applying the streak plate method for detection of Salmonella. The plates were all incubated for 24 hours at 37°C. This to qualitatively check the presence of the pathogen. The plates used for this part of the experiment did not contain antibiotics nor a TSA top layer, since selective growth methods were used. After prior enrichment in BPW overnight it is assumed that the pathogens now will be in a less stressed environment since during the enrichment the conditions are optimal for the pathogen under consideration (optimal temperature of 37°C, no selective nutrients present in BPW, …).

5. Robustness of the spontaneous carrot juice fermentation The fermentation part of the main experiment was repeated, but larger amounts of pathogens were added to the different glass Weck jars. This to simulate an excessive contamination and to investigate in that way the robustness of the fermentation in ensuring food safety also if challenged with high numbers of pathogens. Only counting of pathogens (LOD thus 10 CFU/ml) and pH monitoring was done (thus no enrichment procedures or RNA extraction to monitor microbial community). Other timepoints were chosen i.e. day 0, 1, 3, 6 and 9. The fermentation set-up and pathogen inoculation were done the same way as mentioned in M&M section 3.1 and 3.3. Only this time an inoculation level of 5 x 105 CFU/ml was attempted for each pathogen (applied as a cocktail) (instead of prior experiment where inoculation was performed with ca. 1 x 103 CFU/ml). A tube contained approximately 5 x 108 CFU/ml after one night of incubation at 37°C in nutrient broth. After washing three eps of 1 ml of every pathogen (9 eps in total) (this time a clear pellet was visible), and resuspension in 1 ml of salted carrot juice, 250 µl was added to each glass Weck jar (250 ml). Resulting in a theoretical inoculation of 5 x 105 CFU/ml in each jar.

Results Before the start of the main experiment, several preparatory tests were performed to optimize different parts of the protocol.

1. Protocol optimisation

Effect of acid matrix on the plates A first test was aimed to recognize the different pathogens (colony forming unit, CFU) on the various media and to investigate the reaction of the media with the acid substrate (i.e. FCJ, fermented carrot juice). A cocktail containing three food-borne pathogens was added to fresh carrot juice (CAJ), fresh cucumber juice (CUJ) and fermented carrot juice (FCJ) and all plated out on the selective media. The amount of CFUs were counted and summarised in Table 13.

Table 13: Pathogen counts (log CFU/ml) effect of acid matrix on the media

Day 0 Day 1 (log CFU/ml) CUJ CAJ FCJ CUJ CAJ FCJ

L. monocytogenes 2.4 2.5 < LOD 3.0 4.0 < LOD S. Typhimurium 3.0 3.3 < LOD 5.8 6.5 < LOD E. coli O157 3.0 3.2 < LOD 5.9 6.3 < LOD

Plate counts on day 0 and day 1 from different substrates inoculated with a pathogen cocktail. CUJ = cucumber juice, CAJ = carrot juice, FCJ = fermented carrot juice. LOD = 1 CFU/ml (3x333 µl) CFUs of the different pathogens were detected on the selective plates where fresh juice was cultivated (i.e. CUJ = cucumber juice, CAJ = carrot juice). No CFUs were detected on any plate containing non- diluted fermented carrot juice (FCJ, pH = ± 3.5), even using a lowered detection limit (when 1 ml FCJ was spread over 3 plates, i.e. 333 µl/plate) no colonies were found. On the XLD -and CT-SMAC plates containing non-diluted FCJ a discolouration (yellow spot) was found where the FCJ was added (Figure 9).

Figure 9: Acid trace on XLD plates (left) and interfering (yellow CFUs) CFUs on top of the plate together with Salmonella (black CFUs) on XLD plate (right).

Also, colonies of an interfering bacterium (Figure 9) were detected on several XLD plates (CUJ plate counts on day 0 and all plates containing a (non)-dilution of CUJ or CAJ on day 1). For this reason, a higher concentration of antibiotics was used when preparing the XLD plates for the fermentation set- up. Besides these interfering bacteria, also parts of edges of some XLD plates coloured blackish (Figure 9). The black edges on different XLD plates were probably caused by the wetness of the plates (i.e. condensate present). During the dispersion using sterile beads, the condensate of the plates was accumulated at the edges and created a black smear. It was important to dry the plates long enough before inoculation and incubation.

The XLD plates of the control samples (juice not inoculated with pathogens) contained also the interfering bacterium, this for CAJ and CUJ, not for FCJ, indicating the natural presence of these bacteria in the juice. To determine the number of bacteria inoculated, a subsample was taken from the pathogen cocktail and plated out. However, the plates of the inoculum could not be counted, as too much CFUs were present on the plate (> LOQ limit of quantification, > 300 CFU/plate). No typical phenotypic pathogen-looking colonies (CFUs) were present on any of the negative control plates. To counteract the effect of acidification on the plates, fermented carrot juice samples were diluted by adding 9 ml BPW to 1 ml of FCJ, resulting in a pH increase to 6.45 (Table 14). By using this dilution, the plates did not show a distinct discolouration anymore. By doing this the detection limit increased with a factor 10, to circumvent this increase 1 ml of the BPW dilution was spread out over 3 plates (LOD = 10 CFU/ml).

Table 14: Different dilutions of fermented carrot juice with BPW and corresponding pH

Dilution with BPW pH fermented carrot juice none 3.53 10 x 6.45 100 x 7.19

Effect of acid matrix on the counts As the first preparatory test did not give conclusive results for reliable recovery and count, or information on survival of pathogens upon inoculation in fermented carrot juice, a second test was performed (Table 15). The aim of this test was to investigate how much time the pathogens would survive, when put directly in acid FCJ (Fermented Carrot Juice, pH 3.53).

Table 15: pathogen counts after inoculation in FCJ on 0, 30 and 60 minutes

(log CFU/ml) 0 min1 0 min2 30 min 60 min L. monocytogenes 2.1 2.2 < LOD4 < LOD S. Typhimurium / 2.4 < LOD < LOD E. coli O157 / 1.83 < LOD < LOD 1The non-diluted FCJ was used for plating, only ALOA showed CFUs (plating 0.1 ml and 3x333 µl had the same result) 2 10x diluted FCJ was used for plating (3x333 µl); this was needed to circumvent acidification and interference thus for the selective media for Salmonella & E. coli as was noted in the preparatory test 3 Estimated number (< 10 colonies counted on one plate) 4 LOD = 10 CFU/ml

The pH of the FCJ used was ± 3.53 (measured from a duplicate jar i.e. of the same day, treated exactly the same way). When directly plating a -1 dilution (timepoint = 0), quantifiable CFUs were found. After 30 minutes the CFUs fell already beneath the LOD of the plates. After 60 minutes also no colonies were detected.

Cucumber juice ratios For organoleptic reasons cucumber juice (CUJ) was chosen to be added to the fermented carrot juice (FCJ) after 30 days. CUJ contained, like the carrot juice, also its own LAB community. Different isolates were taken from the CUJ which can be used in further research, this part was explorative and had no further application in this Master Thesis. Different plate counts were done of fermented cucumber juice (11 days old without addition of salt), 16 isolates were also picked from the MRS plates and stored at -80°C. Initiating a fermentation without salt gives also a pH drop, at day 11 only LAB were found, no Enterobacteriaceae or yeast colonies were found on respectively VRBG and YPD media (Table 16). Because no salt was added, a top layer of moulds originated at the top of the Weck jar. This proves the addition of salt is really necessary to guide the fermentation in right directions, and avoid in that way unwanted spoilage.

Table 16: community plating results of 11 days old cucumber juice (no salt added)

(log CFU/ml) Plate type Day 11 YPD < LOD VRBG < LOD MRS 7,70 ± 0.03 (n = 2) LOD = 10 CFU/ml Within the framework of the present thesis also mixes were made containing different ratios CUJ/FCJ, with the aim of finding the best organoleptic result (by odour and pH). The pH of the different mixes was measured on the different time points at room temperature (20°C). The higher the proportion of FCJ, the lower the pH. As a final ratio 25% FCJ / 75% CUJ was used for the main experiment. The results are summarised in Table 17.

Table 17: pH of different fermented carrot juice - cucumber juice mixes over time (day 0, 3, 7 and 11)

Ratio DAY pH Ratio Day pH 50% FCJ + 50% CUJ 0 3.94 50% FCJ + 50% CUJ 7 3.78 25% FCJ + 75% CUJ 0 4.21 25% FCJ + 75% CUJ 7 3.82 75% FCJ + 25% CUJ 0 3.82 75% FCJ + 25% CUJ 7 3.75 100% FCJ 0 3.73 100% FCJ 7 3.68 100% CUJ 0 5.39 100% CUJ 7 4.50

50% FCJ + 50% CUJ 3 3.87 50% FCJ + 50% CUJ 11 3.75 25% FCJ + 75% CUJ 3 3.92 25% FCJ + 75% CUJ 11 3.86 75% FCJ + 25% CUJ 3 3.74 75% FCJ + 25% CUJ 11 3.73 100% FCJ 3 3.68 100% FCJ 11 3.67 100% CUJ 3 4.51 100% CUJ 11 3.98 Addition of CUJ will elevate the pH of the FCJ. In that way - at room temperature - the pH will decrease back in time again. Considering the pH will not lower when refrigerated, the highest pH on day 0 was chosen to be the right dilution in terms of pH and acid smell.

Original carrot community On day 0, samples were plated on three selective media (YPD, VRBG, MRS) for CAJ (carrot juice containing salt) and whole carrot from the swab method (solid state, M&M section 2.4). The juice showed higher numbers of Enterobacteriaceae compared to the swab method (Table 18). Ca. 3.3 log

LAB CFUs/ml were found in the juice, while the carrots only contained few (estimated number). No yeasts were found in/on both substrates.

Table 18 community plate counts of carrot juice and whole carrot

Plate Substrate (log CFU/ml) CAJ Whole carrots YPD < LOD < 1 VRBG > 3 2.8 MRS > 31 0.62 1 CFUs > LOQ (> 300 colonies on smallest dilution (i.e. none dilution) plated (3x333 µl), thus > 900 CFUs/ml counted) 2 Mere detection of presence (less than 4 colonies counted on lowest dilution (3x333 µl non-dilution) plated, thus < 4 CFU/ml) 2. Fermentation (20°C) Different Weck jars containing fresh carrot juice and salt (2.5% W/V) were inoculated with pathogens (final concentration of ± 10³ CFU/ml each) (except for the negative control which was also prepared and sampled at each time point). The fermentation of the carrot juice was monitored over a period of 30 days at 20°C, and on each selected time point (day 1, 3, 15 and 30) the pH was measured, pathogens were counted and RNA was extracted.

pH evolution On the different time points (day 0, 1, 3, 15 and 30) the pH of the fermenting carrot juice was measured at room temperature. The results are presented in Table 19.

Table 19: pH evolution during fermentation (20°C)

Duration Spontaneous + pathogens Spontaneous (days) (pH, n=3) (pH, n = 1) 0 / 6.13 ± 0.04 (n=2) 1 5.82 ± 0.01 5.80 3 4.64 ± 0.01 4.63 15 4.06 ± 0.09 4.03 30 3.68 ± 0.04 3.67 Over 30 days of fermentation, the pH lowers from 6.13 ± 0.04 to a pH of 3.67 and 3.68 ± 0.04 respectively for the spontaneous fermentations without pathogens and the other ones containing pathogens (inoculated on day 0). Both fermentations (with or without pathogens) had a similar pH evolution (Figure 19, addendum), with the biggest pH drop between day 1 and 3.

Knowing the amount of salt added and water present in the carrot juice, the aw was calculated

(Equation 5). This to examine if the aw was a crucial factor on the pathogen development. Salt (350 g NaCl) was added to the general batch containing 14 L of carrot juice. Assuming an average carrot consists 90.1% water (RIVM, 2018), the water activity (aw) could be calculated as followed:

Equation 5: actual aw calculation of the fermenting carrot juice

() 1.00424 − 0.702 x = 0.985 = aw ()

The high aw value will be considered as non-inhibitory on the pathogen’s growth as it falls inside their normal growth range (Table 6) .

Pathogen development During the first 24 hours, all pathogens showed growth. E. coli O157 developed the best (+log 2.0), followed by respectively S. Typhimurium (+log 1.8) and L. monocytogenes (+log 1.6). At day three the number of pathogens was already declining. L. monocytogenes was even absent on the plates and fell under the limit of detection (LOD = 10 CFU /ml). The other two pathogens were still present on the plates, but already in lower numbers (Figure 20, addendum & Table 20). At day 15 and day 30 of the fermentation, no colonies were detected on any plate. At this time point also S. Typhimurium and E. coli O157 fell beneath the LOD. The die-off to actual absence per 10 ml (thus even lower than the LOD of 10 CFU/ml) of the different pathogens was confirmed later during enrichment procedures (Results section 4).

Table 20: pathogen counts during fermentation

(log CFU/ml) Day 0 Day 1 Day 3 Day 15 Day 30 L. monocytogenes 2.52 ± 0.09 3.94 ± 0.09 < LOD1 < LOD < LOD E. coli O157 2.26 ± 0.05 4.67 ± 0.04 3.77 ± 0.2 < LOD < LOD S. Typhimurium 2.79 ± 0.03 4.88 ± 0.03 3.86 ± 0.1 < LOD < LOD 1LOD is 10 CFU/ml, thus < log 1 CFU/ml (limit of detection) A one-way ANOVA was performed to check any difference between the average growth of the three food-borne pathogens (Table 21). Because the ANOVA tested significant (P < 0.05) a subsequent TukeyHSD test was done to compare the significance pairwise. The mean growth of every pathogen was significantly different compared to each other, E. coli O157 developed best, followed by S. Typhimurium and L. monocytogenes respectively.

Table 21: Statistical tests and results on pathogen growth during fermentation

Test P-value (adjusted) ANOVA 0.00019 TukeyHSD (comparison of pathogen growth) S. Typhimurium – L. monocytogenes 0.01431 E. coli O157 – L. monocytogenes 0.00016 E. coli O157 – S. Typhimurium 0.00303 P < 0.05 = significant

RNA analysis To assess the community composition by 16S (V4) rRNA sequencing, RNA was extracted and stored at -80°C. With the data gained from the RNA extraction (sequencing results), different plots (bar plot, alpha and beta diversities) were constructed in R, to obtain a better view of the fermenting community and the possible impact of different conditions on it, in this case the pathogen effect on the fermentation.

2.4.1. Community bar plots

Figure 10: Community bar plot of 11 relative most abundant species during fermentation (20°C) time points The relative abundancies of the different samples are given on the different timepoints (grey box: 1, 3, 15 and 30 days). Sample A, B and C are the samples containing pathogens (triplicate), S stands for spontaneous and has not been inoculated with pathogens. Taxon are the different genera of the 11 most abundant species.

Figure 10 displays the 11 most relatively abundant species on each time point for each sample. A number is given to the genus, if there are multiple ASVs/reads (Amplicon Sequence variation) within this genus. Sometimes different suggestions for possible species are given to the read as these map 100% to that species. Lactobacillus 1 could be determined as fabifermentans, paraplantarum, pentosus or plantarum. Lb. 2 could be brevis. Lb. 3 could be kefiri, otakiensis, parabucherni or sunkii. Leuconostoc 1 could be mesenteroides and Ln. 2 could be gelidum, inhae, miyukkimchii. The other numbers could not be specified in more detail. When looking at the bar plots, the distribution of the most abundant species on day 1 and 3 looks similar. Enterobacter 1, Leuconostoc mesenteroides and Yersinia all count for approximately 25% of the relative abundancy. Between the first two time points and the last two a big community shift is recognised where Lactobacillus 1 dominates at the end in every sample in terms of relative abundancies. The pathogens inoculated at day 0 do not belong to the 11 most abundant species. Surprisingly, not even a single read (out of 956101 reads after quality filtration) was allocated to the different pathogens inoculated at the start of the fermentation.

2.4.2. Alpha diversity To assess the overall diversity of the community per sample, the alpha diversity was calculated using the Inverse Simpson index (Figure 11). A higher div_inv_simpson means a bigger diversity within the sample. The alpha diversity has a lowering trend during the fermentation indicating a decrease in diversity on later time points. This phenomenon is clearly visible following the progress of the spontaneous fermentation (red dot) over time. A similar trend occurred analysing the samples inoculated with pathogens (blue dots), with some variation present between those samples on the same time point. The inverse Simpson diversity index

(i.e. the Simpson reciprocal index) was chosen. The diversity index takes into account the overall diversity (presence of species) as also the evenness of taxa (relative abundance within species).

Figure 11: Alpha diversities (reciprocal Simpson) during fermentation The inverse Simpson index is given for the different samples on different timepoints (day 1, 3, 15 and 30). The red dots are the spontaneous samples (S), the blue dots are the samples inoculated with pathogens at the beginning of the fermentation. A Kruskal-Wallis test was performed to check whether there was any significant difference between the alpha diversities of the different samples (over time). When significant (P < 0.05) a Dunn’s test was performed to see between which days the alpha diversity of the samples was significant. To account for multiple testing, a p-adjustment method was used (Benjamini-Hochberg correction). All results are present in Table 22. The alpha diversity differed significantly between samples of day 1 and 30 (P < 0.05). The difference was clearly not significant between day 1 and 3, and 15 and 30. This is also visible in the bar plot (Figure 10) where the relative abundancies looked almost the same. The results are now based on a small dataset, by adding more samples, more significance would probably originate from the statistical analysis. Comparing the alpha diversity of samples inoculated with or without pathogens were not significant, for this reason the statistical analyses were not incorporated in the results.

Table 22: Statistical tests and results on alpha diversities between time points of fermentation

Test P-value (adjusted) Kruskal-Wallis 0.01491 Dunn’s (comparison of days) 1 – 3 0.663 1 – 15 0.064 1 – 30 0.036 3 – 30 0.063 15 – 3 0.131 15 – 30 0.656

2.4.3. Beta diversity After calculating alpha diversity for each individual samples, beta diversity was also calculated to assess how similar samples are in community composition. To calculate this beta diversity, the Bray-Curtis distance matrix was used and visualised using a Principal Coordinates Analysis (PCoA). It takes an input matrix giving dissimilarities between pairs of items and gives Figure 12 as an output. When dots are close together it means that the community of the two samples is similar. PCoA1 and PCoA2 stands for the two axis that represents the most variation, where pcoa1 is the most important one and pcoa2 the second most.

Figure 12: beta diversities during fermentation Principal Coordinates Analysis is done for different samples during fermentation (20°C) on different time points (days 1, 3, 15 and 30). The triangles are inoculated with pathogens on day 0, while the dots are uninoculated. Samples from day 1 and 3 had a similar PCoA1 (i.e. 90%), while samples of day 15 and 30 had a similar PCoA2 (55%), with exception of some outliers. The figure (Figure 12) shows the beta diversity only applied on the samples present on the figure. When applying PCoA on all RNA samples taken for the experiment (inclusion of the refrigeration part), a cluster was formed, containing samples of day 1 and 3, the other containing samples of later days 15 until 38 were more spread (Figure 21, addendum).

3. Refrigeration (7.5°C) and possible addition of cucumber juice To investigate the robustness of the carrot juice after 30 days of fermentation towards resuscitation (of low numbers of residual pathogens present) or post-contamination and newly introduced pathogens, a follow-up challenge test was performed. During this follow-up experiment, different jars containing 30-days-old fermented carrot juice (with/without pathogen inoculated on day 0) or a mix with fresh cucumber juice were put under refrigeration conditions (i.e. 7.5°C). As mentioned above, some jars were freshly inoculated with pathogens (± 10³ CFU of each pathogen/ml) while others remained as they were (with/without pathogen inoculation on day 0)(Figure 13). The fermented carrot juice (FCJ) and mixes of fermented carrot juice with cucumber juice (FCJ + CUJ) were monitored over a period of 8 days. The pH was measured, the pathogens were counted. In addition of these FCJ or FCJ- 50

CUJ samples not inoculated with pathogens – blanc controls – also RNA was extracted to establish the effect of refrigeration and/or mixing with CUJ on the FCJ.

Figure 13: refrigeration part of the experimental set-up Blue pathogens are freshly inoculated at day 30, while samples containing the red pathogens were only inoculated at day 0. CUJ = cucumber juice, FCJ = fermented carrot juice, L2 = freshly inoculated pathogens on day 30, FCJL2 = FCJ inoculated at day 0. The different FCJ/CUJ ratios are indicated under every triplicate.

pH evolution The pH of the different samples of FCJ and mix (25% FCJ + 75% CUJ) was measured after 1 and 8 days in the refrigerator (Table 23). The pH stayed approximately constant for all jars when stored under refrigeration. The addition of fresh cucumber juice (75%) (CUJ, pH = 5.5) to the fermented carrot juice (25%) resulted in a shift from ±3.7 to ±4.4. Also, the freshly pathogen inoculated (day 30) mix (CUJ + FCJ + L2) followed the same trend. The fermented carrot juices without cucumber juice (with or without pathogen inoculation on day 0) (FCJ and FCJL2) remained throughout these further 8 days refrigerated storage at a pH below 4 (Figure 22, addendum).

Table 23: pH values during refrigeration

pH CUJ CUJ + FCJ FCJL2 FCJ CUJ + FCJ + L2 CUJ + FCJL2 Day 30 5.5 4.38 3.68 ± 0.04 3.67 - - Day 31 - 4.48 ± 0.09 3.95 ± 0.01 3.78 ± 0.09 4.44 ± 0.01 4.55 ± 0.02 Day 38 - 4.22 ± 0.04 3.68 ± 0.02 3.84 ± 0.04 4.27 ± 0.01 4.22 ± 0.05

Pathogen development Different plate counts were done on selective media for all samples (with/without pathogens, with/without cucumber juice) under refrigeration conditions on day 31 and 38 (Figure 23, addendum). This to see how the freshly inoculated pathogens would develop under those conditions as well as to see if the prior inoculated pathogens before starting the overall fermentation (thus on day 0) - if still present in very low numbers after 30 days - would resuscitate. 51

No quantification of E. coli O157 could be done in this part of the experiment, as- by accidental mistake in the lab media preparation - the CT-SMAC plates contained too high antibiotic concentrations. S. Typhimurium numbers stayed approximately constant during refrigerated storage (day 30 to day 38). L. monocytogenes numbers fell below the LOD (exception in 1 out of 3 replicates) after 8 days.

Table 24: pathogen counts during refrigeration

(Log CFU/ml)1 day 30 day 31 day 382 L. monocytogenes 2.69 ± 0.05 2.46 ± 0.02 < 13 S. Typhimurium 2.95 ± 0.01 2.67 ± 0.05 2.34 ± 0.03 1 This table contains results of the samples inoculated with pathogens at day 30. 2 All plates at this day were incubated for 48 h at 37°C. 3 Mere detection of presence (in 1 replicate, 2 CFUs counted on only ALOA for one of three plates used to lower the detection limit)

2 CFUs of L. monocytogenes were present on one of the three ALOA plates containing 330 µl of CUJ+FCJ mix, which was inoculated at day 30. The other two replicates did not contain any colony on their plates (i.e. 2x3 different plates), so 1 out of the 9 plates contained only two CFUs. All plates from day 38 were put for an extra 24 hours in the incubator, because colony-like spots were present on some plates. After 48 hours of incubation, the spots turned out to be no CFUs, but some additional L. monocytogenes colonies became visible. Table 25 gives an overview of the samples who contained L. monocytogenes after 48 h of incubation (Mere detection of presence on all plates, i.e. a maximum of 3 CFUs on one ALOA plate counted.).

Table 25: ALOA plates with presence L. monocytogenes colonies after 48h of incubation

ALOA Plate for L. monocytogenes counting Log CFU/ml CUJ + FCJ + L2 A (inoculated day 30) < 1 CUJ + FCJ (no pathogens inoculated) < 1 CUJ + FCJL2 A (inoculated day 0) < 1 CUJ + FCJL2 B (inoculated day 0) < 1 In table 24 and 25 a mere detection of CFU are found on certain ALOA plates, on a variety of plates not linked to each other, inoculated on different time points and even on a blank. The only thing they have in common is the presence of cucumber juice. For that reason, the Listeria monocytogenes retrieved here could possibly originate from the cucumbers used, which proves the usefulness of this experiment. On the other hand, an incubation time of 48h for ALOA is recommended to avoid false negatives.

RNA analysis and statistics RNA was extracted and concentration was measured as described above. With the data gained from the RNA extraction (sequencing results), different plots were constructed in R, to obtain a better visualisation of the fermenting community and the possible impact on it by altering different conditions (in this case the effect of addition of cucumber juice and refrigeration). The bacterial community was analysed on different timepoints, using relative abundancy bar plots and checking the alpha and beta diversity.

3.3.1. Community bar plots The bacterial community of FCJ as was determined on day 30 of fermentation and which served as starting point of the refrigeration part (with or without mixing with fresh CUJ) is also included here in the present figures.

Figure 14: 11 most relative abundant species in fermented carrot juice during refrigeration The relative abundancies of the different samples are given on different timepoints (grey box: day 30, 31 and 38). This relates to the samples that contain only fermented carrot juice (FCJ). Samples A, B and C are three replicates (containing no pathogens). Sample S stands as a reference, which describes the FCJ samples just before cooling. Taxon are the different genera of the 11 most abundant species. Figure 14 displays the 11 most relatively abundant species on each time point for each sample of fermented carrot juice in the cooling (no addition of cucumber juice). Sometimes different suggestions for possible species are given to the genus. For Enterobacter no suggestions were done. Lactobacillus 1 could be fabifermentans, paraplantarum, pentosus or plantarum. Lb. 2 could be brevis. Lb. 3 could be kefiri, otakiensis, parabucherni or sunkii. Leuconostoc 1 could be mesenteroides and L. 2 could be gelidum, inhae, miyukkimchii. The other numbers could not be specified in more detail.

Lactobacillus 1 loses some relative abundancy to other types of lactobacilli (Lb. 2,3 and Leuconostoc 1) and also Enterobacter 1 when stored in the refrigerator. The drop in relative abundancy for Lactobacillus 1 ranges from 80% to ±65% for sample A. The next bar plot (Figure 15) represents the relative abundancies of the (never inoculated) mix of fermented carrot juice samples mixed with cucumber juice on day 31 and 38. Also the spontaneous fermented carrot juice sample of day 30 is included as a reference of how the community was before addition of cucumber juice and refrigeration conditions.

Figure 15: 11 most relative abundant species in a mix of fermented carrot juice and cucumber juice during refrigeration The relative abundancies of the different samples are given on different time points (grey box: day 30, 31 and 38). This relates to samples (with exception of S) contain a mix of 25% FCJ and 75% cucumber juice. Samples A, B and C are three replicates (containing no pathogens). Sample S stands as a reference, which describes the fermented carrot juice just before cooling. Taxon are the different genera of the 11 most abundant species. Figure 15 displays the 11 most relatively abundant species but this time of fermented carrot juice mixed with 75% of cucumber juice in under refrigeration. Here also, for Enterobacter no suggestions were done. Lactobacillus 1 could be fabifermentans, paraplantarum, pentosus or plantarum. Lb. 2 could be kefiri, otakiensis, parabucherni or sunkii. Lb. 3 could be brevis. Leuconostoc 1 could be mesenteroides and Ln. 2 could be gelidum, inhae, miyukkimchii. The other numbers could not be specified in more detail. Lb. 1 did not lose as much relative abundancy compared to the samples of purely fermented carrot juice (Figure 14) after 8 days of refrigeration. By addition of cucumber juice, Lb. 1 seemed to keep better dominance. It can be noted that Lb. brevis (lactobacillus 2 on Figure 14), has less relative abundancy in the FCJ-CUJ mix after 8 days of cooling (max 10%) compared with the fermented carrot juice (up to 30%).

As can be noted from Figure 14 and Figure 15, in either FCJ or the FCJ-CUJ mix Yersinia already initially present in the FCJ, belongs still too the 11 most abundant species.

3.3.2. Alpha diversity The effect of cooling and addition of cucumber juice on the alpha diversity was calculated. Figure 16 shows the different alpha diversities. No clear trend is visible as was present during the fermentation (Figure 11).

Figure 16: Alpha diversities (reciprocal Simpson) during refrigeration The inverse Simpson index is given for the different samples on different timepoints (day 30, 31 and 38). The red dots are the samples containing only fermented carrot juice, the blue dots are the samples containing a mix of fresh cucumber juice and fermented carrot juice. The dots at day 30 serve as reference, and are the samples just before being put in the refrigerator. The inverse Simpson diversity does not contain a clear trend over time for both types of samples (FCJ or mix). The values fluctuate on different timepoints and are varying within the replicates. This was also confirmed with the statistical analysis where no significant results were found (Table 26: Statistical tests and results on alpha diversities between time points of refrigeration).

Table 26: Statistical tests and results on alpha diversities between time points of refrigeration

Test P-value (adjusted) Kruskal-Wallis FCJ 0.1934 Mix 0.7515 FCJ vs. mix day 31 0.5127 FCJ vs. mix day 38 0.1266

The alpha diversity of samples containing only fermented carrot juice during refrigeration was not significantly different over days (day 30, 31 and 38). The Kruskal-Wallis test resulted in a non-significant P-value. For this reason, no Dunn’s test was performed. Also, the alpha diversity of samples containing a mix of fermented carrot juice and cucumber juice during refrigeration was not significant. Comparing the difference in alpha diversity between samples with/without cucumber juice on day 31, and similar on day 38 resulted in non-significant P-values. Although the community of only fermented carrot juice versus the mix looked different on the bar plots on the two time points under refrigeration (Figure 14, Figure 15). Yet the difference was not enough to be statistically significant.

3.3.3. Beta diversity An overview of the beta diversity, i.e. the diversity between samples, under refrigeration conditions is represented in Figure 17.

Figure 17: beta diversities during refrigeration Principal Coordinates Analysis is done for different samples in cooling (7.5°C) on different timepoints (days 31 and 38). The dots are samples containing pure fermented carrot juice and the triangles are samples mixed with cucumber juice (25%). PCoA1 = 80% and PCoA2 = 12%. The beta diversity is often different even between samples which are in fact (biological) replicates, more replicates will be needed to make significant conclusions out of these results. Day 38 containing cucumber juice forms the closest cluster, this can also be confirmed when looking at the bar plots, sample 3A, 3B and 3C (Figure 15) look at that moment most similar compared to samples 2A, 2B and 2C (Figure 14) or on time point day 31.

Community During the cooling part of the experiment, plating on MRS was also done. With the aim of analysing the growth of the LAB at refrigeration. The number of LAB stayed approximately constant for the mix of CUJ and FCJ as well as for FCJ on its own (Table 27).

Table 27: community plate counts of FCJ and mix (FCJ-CUJ) during refrigeration on MRS

(log CFU/ml) day 30 day 31 day 38 CUJ + FCJ 7.33 ± 0.07 (n=6) 7.84 ± 0.06 (n=3) 8.3 ± 0.2 (n=3) FCJ 8.1 ± 1.7 (n=40)1 8.1 ± 0.2 (n=3) 8.1 ± 0.1 (n=3) 1 Wuyts et al., 2018

4. Enrichment procedures Only these FCJ samples that were initially (at day 0) challenged with pathogens and monitored during 30 days at 20°C, or even for up to 38 days (latter 8 days shifted to refrigeration) or those FCJ-CUJ mixes inoculated (challenged) with pathogens immediately after mixing prior to being put under refrigeration were subjected to pathogen testing by enrichment methods. Initial inoculation (either at day 0 for the FCJ, or at day 30 for the FCJ-CUJ mix) was done with a cocktail of 3 pathogens and an inoculum of ca. 10³ CFU/ml. To increase sensitivity in these samples where few or no colonies were found when selective plate media were used and thus pathogen numbers dropped below the LOD of the plate count method (< 10 CFU/ml) enrichment procedures were used to test presence/absence (in at least 10 ml) of any residual pathogen survivors for these samples. Table 28, Table 29 and Table 30 beneath give an overview of the results.

For the FCJ during 20°C fermentation (Day 1-30) only the samples at day 1 tested positive (as expected, also count methods showed colonies) whereas all the other days no pathogens were detected after enrichment. The fact no pathogens were found during enrichment on day 3 is peculiar, since E. coli O157 and S. Typhimurium were still found on the counting plates during the main experiment. This phenomenon will be discussed in the discussion part.

Table 28: enrichment procedures of the fermentation (pathogens inoculated at day 0)

FCJ (incubated at 20°C) Time point (days) Pathogen Presence (+)/ Absence (-) per 10ml1 L. monocytogenes +++ 1 E. coli O157 +++ S. Typhimurium +++ L. monocytogenes --- 3 E. coli O157 --- S. Typhimurium --- L. monocytogenes --- 15 E. coli O157 --- S. Typhimurium --- L. monocytogenes --- 30 E. coli O157 --- S. Typhimurium --- 1 +++ means all of the triplicate tested positive for the pathogen by enrichment procedures; --- means all three replicates tested negative.

The FCJ samples initially inoculated with pathogens on day 0, fermenting for 30 days, then mixed with CUJ and being shifted to refrigeration, also all tested negative for the enrichment method, both on day 31 and 38 (after thus 1 and 8 days refrigerated storage). Thus, this indicates that after fermentation for 30 days indeed no residual pathogens were present anymore or did not find the ability to resuscitate anymore even if shifted under refrigeration conditions. This test was only performed in duplicate since no pathogens were found already after 15 days of fermentation and were supposed to stay absent, due to even more supressing fermenting conditions (lower pH, lactic acid, …). Also, not enough BPW was available to dilute the samples in triplicate.

Table 29: enrichment procedures of refrigeration part (pathogens inoculated at day 0)

FCJ-CUJ (stored at 7.5°C) Pathogen Presence (+)/ Absence (-)per 10 ml1 Time point (days) (inoculated at Day 0) by enrichment method L. monocytogenes -- 31 E. coli O157 -- S. Typhimurium -- L. monocytogenes -- 38 E. coli O157 -- S. Typhimurium -- 1 ++ means all of the duplicates tested positive, +- means 1 out of two tested positive, … 2 Only worked in duplicate for CUJ+FCJ inoculated at day 0

The FCJ samples mixed with CUJ (i.e. FCJ-CUJ) freshly inoculated and put in the refrigerator immediately after inoculation (Day 30) tested all positive for the pathogens at day 31. But after eight days (i.e. samples day 38), no L. monocytogenes was detected anymore after enrichment whereas the other two pathogens (Salmonella and E.coli O157) tested positive for two out of three replicate samples. Thus, upon post-contamination when mixing FCJ with a non-fermented CUJ there seems to be ability for residual survival of a few cells of in particular the enteric pathogens S. Typhimurium and E.coli O157 when stored under refrigeration.

Table 30: enrichment procedures of refrigeration part (pathogens inoculated at day 30)

FCJ-CUJ (stored at 7.5°C) Pathogen Presence (+)/ Absence (-) per 10ml1 Time point (days) (inoculated at Day 30) by enrichment method L. monocytogenes +++ 31 E. coli O157 +++ S. Typhimurium +++ L. monocytogenes --- 38 E. coli O157 ++- S. Typhimurium ++- 1 +++ means all of the triplicate tested positive for the pathogen by enrichment procedures; ++- means two out of three replicates tested positive, …--- means all three replicated tested negative.

For completeness, an overview combining both the plate count results of the pathogens and the enrichment methods to detect the presence of pathogens, can be found in Table 33, addendum.

5. Robustness test The resistance of the fermenting conditions was put even more to test, by adding a higher inoculation count of pathogens. A contamination at the start of the fermentation was simulated, this time by adding a cocktail containing 105 CFU/ml of each pathogen. The pH evolution and the pathogen development were followed over time (day 0 till 9).

pH evolution Whether inoculated with pathogens or not, the pH evolution was the same. Declining from a value of 6.18 to a pH value around 4 after 9 days of fermentation. The values can be found in Table 31 and are represented in Figure 24, addendum.

Table 31: pH evolution robustness test

Duration Spontaneous + pathogens Spontaneous (days) (pH, n=3) (pH, n = 1) 0 / 6.18 1 5.73 ± 0.01 5.87 3 4.710 ± 0.006 4.55 6 4.57 ± 0.04 4.58 9 4.03 ± 0.02 4.17 Pathogens inoculated at 105 CFU/ml

Pathogen development The same pattern as during the main experiment (Table 20) was retrieved when adding higher numbers (105 CFU/ml) of pathogens to the fermentation (Table 32, Figure 25, addendum). L. monocytogenes did not show growth during fermentation and maintained numbers between 5 and 6 log CFU/ml during 3 days, as of day 6 no CFUs were found on the plates anymore (thus < 10 CFU/ml). S. Typhimurium and E. coli O157 showed growth the first day and reached numbers up to 7.8 log CFU/ml. E. coli showed again the best growth pattern. S. Typhimurium fell like L. monocytogenes under the LOD after 6 days and E. coli was not present on the plates anymore after 9 days. Even at these higher inoculation numbers the pathogens were suppressed by the fermenting conditions.

Thus, in conclusion, notwithstanding initial (2-3 log) outgrowth to high numbers (ca. 107 CFU/ml) in the first 24h, in particular after 72 h a rapid die-off was established resulting in a 6-log reduction being accomplished for S. Typhimurium and E. coli O157 after 9 Days fermentation (pH of ca. 4.0 achieved). L. monocytogenes was again most sensitive to the fermentation: only no or minor growth potential was noted in the first 24-72 h and also a die-off was established with no detectable numbers (< 10 CFU/ml) after 9 Days fermentation, but as initial growth was restricted not the often ‘targeted’ food safety objective of 6 log reduction could be established (experiments showed ca. 4.7 log reduction); however also in the latter case the fermentation process was shown to be robust and ensured a safe food.

Table 32: mean log CFU/ml of pathogen development during robustness test

n = 3 Time (days) (log CFU/ml) 0 1 3 6 9 L. monocytogenes 5.37 ± 0.02 5.68 ± 0.08 5.2 ± 0.2 < LOD < LOD S. Typhimurium 5.6 ± 0.1 7.80 ± 0.05 6.8 ± 0.2 < LOD < LOD E. coli O157 4.59 ± 0.09 7.17 ± 0.07 7.30 ± 0.05 5.0 ± 0.1 < LOD LOD = 1 log CFU/ml

Discussion

1. Strain selection In this Master Thesis we worked with model food-borne pathogens i.e. Listeria monocytogenes, Salmonella Typhimurium, Escherichia coli O157. The last member of the ‘big four’, Campylobacter jejuni/coli, was excluded from the selection, due to its more common association with meat, more specific poultry, and not with vegetables. Moreover, C. jejuni/coli is also more sensitive for growth/survival at low pH values. ( Devlieghere et al., 2016) As mentioned in literature (Wuyts et al., 2018) and confirmed now from our sequencing results (Figure 10), the presence of Yersinia spp. was detected in carrot juice. This points to a presence of high numbers of these Yersinia species as it was noted that the pathogens originally added at day 0 (10³ CFU/ml) or even when some pathogens’ growth was noted at day 1 (up to almost 105 CFU/ml) these pathogenic species were never detected during sequencing. Still, it should be noted that Yersinia sp., which was among the eleven most abundant taxon identified (Figure 10), does not necessarily include or refer to pathogenic Yersinia species. Recent reports concerning the taxonomy of the genus Yersinia show it consists of 17 distinct species, among which only three are currently recognized as human pathogens: i.e. Yersinia enterocolitica, Yersinia pseudotuberculosis, and Yersinia pestis (Biasino et al., 2018; Duan et al., 2014), In particular Y. enterocolitica and pseudotuberculosis have already been associated with food-borne outbreaks, also implying carrots (Kangas et al., 2008; Rimhanen-finne et al., 2009). Among Y. enterocolitica or pseudotuberculosis several biotypes and serotypes can be differentiated according to their biochemical characteristics and lipopolysaccharide O-antigens. Not all these subtypes are pathogenic to humans (Devlieghere et al., 2016). The pathogenicity is linked to the presence of some virulence markers (Atkinson & Williams, 2016). Y. enterocolitica bioserotype 4/O:3 and, to a lesser extent, Y. enterocolitica bioserotype 2/O:9 and Y. pseudotuberculosis O:1 are reported to be the main causative agents of food-borne human yersiniosis infections in the EU (EFSA and ECDC, 2018). Inoculating one of those pathogenic strains would have been a good idea to examine growth/survival during the carrot juice fermentation. However, due to a lack of differentiating phenotypic characteristics (typical morphology, biochemical reactions or growth characteristics) , setting up a challenge test with pathogenic Yersinia strains be difficult. There are difficulties to detect them with the present methods on the usual selective culture media (e.g. CIN agar, Cefsulodin, Irgasan, Novobiocin) as also a number of Enterobacteriaceae species, other than these Yersinia spp. may also grow on these plate media (Becton Dickinson, 2013b) and thus could generate false positives or also ‘overgrow’ the Yersinia spp. and thus generate false negative results (Van Damme et al., 2015; Van Damme, Habib, & De Zutter, 2010). For those reasons inoculation of pathogenic Yersinia sp. was not performed in the present set-up, since a lot of method optimisation for reliable monitoring and quantification of these pathogenic Yersinia strains still needs to be done before. The other pathogens, also member of Enterobacteriaceae, i.e. the Salmonella and E. coli O157 strains used during the experiment for the inoculation all carried an antibiotic resistance gene making it easier to track them throughout the fermentation.

2. Protocol optimisation Before the start of the experiment, two preparatory tests were performed to assess the effect of acidic fermented carrot juice (pH 3.5) on the pathogens and selective media.

The preparatory tests confirmed that plating non-diluted fermented carrot juice as such affected the performance of the selective plating media due to its acidity and in acidifying the medium (causing discoloration) could also generate false negatives for XLD and CT-SMAC. ALOA was the only medium that was not affected by the low pH of 3.5.

Both effects (quick die-off of the pathogens & interference on the performance of the culture media) can be attributed to the acid pH of the fermented carrot juice (FCJ). The FCJ had a pH around 3.5, this is almost 1 pH-unit below the minimal pH for growth of the pathogens used (Frank Devlieghere et al., 2016). The pH effect on XLD was already detectable with the naked eye, as an acid pH would turn the reddish plates to yellow (pH indicator methyl red)(Figure 9). To counter the pH effect on the different plates, at least a 10x dilution (pH 6.45, Table 14) was done in BPW, in that way a more neutral pH could be applied on the plates. Plating 3x333µl of the 10x dilution kept the LOD at 10 CFU/ml and was always performed at a pH < 4.0 to avoid false negatives.

An interfering bacteria causing yellow CFUs on the XLD plates (Figure 9) was found when adding nonfermented cucumber -or carrot juice with -or without pathogens. The bacteria proved to be resistant to streptomycin just like the Salmonella strain. For that reason, the streptomycin dose of the XLD plates was doubled, to suppress the bacteria of overgrowing the Salmonella strain, but not a higher dose was added, otherwise the Salmonella strain used would be under too stressed conditions. According to literature the yellow CFUs could be for example E. coli (Becton Dickinson, 2013a) for that reason, the E. coli O157 strain used for this experiment was also tested to be streptomycin resistant, the fact the CFUs were also present on non-inoculated plates could already partly exclude this hypothesis, and was confirmed with no detectable growth of the strain in nutrient broth containing the antibiotic. According to the specifications given by Oxoid UK, the yellow CFUs could be Escherichia, Enterobacter, Klebsiella, Citrobacter, Proteus or Serratia (Oxoid Microbiology Products, 2001). The plates containing fermented carrot juice did not show these interfering bacteria, and later during the main experiment, the absence of it could be confirmed as of day 3. This trend of bacteria disappearing was already confirmed in earlier research by Wuyts et al., where in that case presumptive Enterobacteriaceae were present at days 1 and 3 but showed no presence on VRBG agar medium from day 10 on (< LOD). Klebsiella and Citrobacter were both found in decent relative abundancies (16S rRNA gene sequencing) in at least 38 different fermentations, in former studies by Wuyts et al., and their relative abundancies became lower at the end of the fermentation (day 30) (Wuyts et al., 2018). This is of course only a presumption, further analysis of the CFUs needs to be done for confirmation of the strain (e.g. DNA extraction of the isolate).

3. Fermentation (20°C) The fermentation process was effective in supressing the different pathogens (inoculum of 10²-10³ CFU/ml). L. monocytogenes was the most sensitive and declined to non-detectable numbers quickly (as of day 3 of the fermentation). S. Typhimurium and E. coli O157 fell under the LOD after more than 3 days, they were not detectable anymore at day 15. Non-detectable relates to pathogens being < 10 CFU/ml. The pathogens were still not detectable after enrichment procedures (absence in 10 ml). This means, when starting from 10³ CFU/ml and the absence is confirmed in 10 ml, a 4-log reduction was obtained. It was therefore according to this experiment of importance to ferment for at least 15 days to assure that possible contamination of the different pathogens (10²-10³ CFU/ml) has been enough reduced. When comparing the results of the robustness test (inoculation 104-105 CFU/ml), L. monocytogenes still seemed to be the most sensitive out of three pathogens to the fermentation 62

conditions and did not show explicit growth during the fermentation. Since no enrichment procedures were done for the robustness test only a 5-log reduction could be confirmed for L. monocytogenes. S. Typhimurium and E. coli O157 showed growth (> 7 log CFU/ml ) at the start of the fermentation, hereby a 6-log reduction could be confirmed after growth. The 6 log reduction is commonly used in food preservation to guarantee a decent die-off/absence of vegetative pathogens (Frank Devlieghere et al., 2016), that is why striving to this reduction is of great importance to consider the fermentation as a safe process in terms of the pathogens inoculated. The guideline of 15 days could be refined to 9 days of fermentation being enough to suppress all three pathogens used.

The addition of a cocktail of pathogens (final concentration up to 105 CFU/ml of three different pathogens) did not affect the expected pH evolution (Wuyts et al., 2018) of a spontaneous CAJ fermentation. At day 3, the pH was ± 4.6, and after 9 days the pH was declining to pH-values around 4.0. At a pH-value under 4.4 no growth potential of L. monocytogenes is commonly acknowledged, without further evidence of prove, stipulated in the EU regulation (Uyttendaele, 2018)(European Commision, 2005). A rapid pH drop is recommended to avoid excessive growth of Enterobacteriaceae which are omni-present on the different vegetables used in the experiment, when looking at the VRBG- plates, even the carrot’s swab, and also confirmed in literature (Wuyts et al., 2018)(Riganakos, Karabagias, Gertzou, & Stahl, 2017). This pH effect also has an antagonistic effect on the inoculated cocktail of pathogens. Due to the pH drop, the pathogens have to invest a lot of energy in maintaining a constant internal pH. The energy normally used for growth will therefore shift towards maintenance of their own cell. Lactic acid has an inhibitory effect on the growth of the different food-borne pathogens, especially on L. monocytogenes. (Mbandi & Shelef, 2002; Ogawa et al., 2001). Additionally, undissociated lactic acid will disturb the electrochemical proton gradient, which can lead to bacteriostasis and even death (Tomé, Teixeira, & Gibbs, 2005). The fact that L. monocytogenes was more sensitive to the fermentation compared to the two other pathogens, could be attributed to the lactic acid effect. Undissociated lactic acid was evaluated as the main growth-inhibiting factor for L. monocytogenes (E. Wemmenhove, van Valenberg, van Hooijdonk, Wells-Bennik, & Zwietering, 2017; Ellen Wemmenhove, van Valenberg, Zwietering, van Hooijdonk, & Wells-Bennik, 2016). Also, in bacterial growth modelling programs like ComBase, lactic acid will be taken into account as a critical factor for growth predictions (ComBase predictor: Growth, accessed on 20 December, 2018).

The juice conditions at the start of the fermentation (pH = 6.1, aw = 0.985, T = 20°C) were within the range of growth (Table 6) of all inoculated pathogenic strains at the beginning of the fermentation. E. coli O157 showed the best growth capacity during the first 24 hours (10 times better as L. monocytogenes), followed by S. Typhimurium (5 times better as L. monocytogenes) and L. monocytogenes respectively (Table 21: Statistical tests and results on pathogen growth during fermentationTable 21). Between day 0 and 3 all pathogens reached a maximum number, which was already declining at day 3, with exception of E. coli during the robustness test. Three effects could be attributed to the declining numbers of pathogens: Firstly, the pH effect as mentioned above, but also the fact lactic acid was produced. According to literature, Escherichia coli O157 is more resistant to the effects of pH compared to Listeria monocytogenes and Salmonella spp. under typical conditions of acidified vegetables (Lu, Breidt, Perez-

Diaz, & Osborne, 2011). This finding is now also confirmed on spontaneous carrot juice fermentations according to the results of this Master Thesis. Secondly the ‘Jameson effect’: the LAB had more favourable growth conditions (anaerobic, salt) during the experiment. Due to these conditions, they were probably able to grow faster resulting in much higher amounts of LAB than the pathogenic strains. In general LAB often dominate the bacterial communities in foods, especially during fermentation, due to a growth rate advantage. The suppression observed by one species over another can be attributed to competition for nutrients, where those nutrients are needed by both organisms (i.e. Jameson effect). Certainly in the beginning of the fermentation (first 24 hours) this effect plays a big role in suppressing the growth of the pathogens (Mellefont, McMeekin, & Ross, 2008). The amount of LAB was almost twice as big as the amount of inoculated pathogens at day 0, when comparing the results of the selective media for the pathogens with the community plating on MRS (Table 18 and Table 20).

Thirdly, the possible presence of bacteriocins could explain die-off of the pathogens during fermentation. Some LAB might produce protein-like toxins to inhibit the growth of other bacteria that they are competing with for their nutrients. The most famous toxin (i.e. nisin, produced by Lactococcus lactis associated with milk) was the first commercially available bacteriocin (E234)(Delves-Broughton, 2005). Several others are on the market or still being explored (Del Rocío López-Cuellar, Rodríguez- Hernández, & Chavarría-Hernández, 2016; Soomro, Masud, & Anwaar, 2002). Bacteriocins associated with Lb. Plantarum, L. mesenteroides and other strains associated to the vegetable drink are plantaricins, leucocins and pediocins (Jimenez-Diaz, Rios-Sanchez, Desmazeaud, Ruiz-Barba, & Piard -, 1993; Settanni & Corsetti, 2007; Tenea & Barrigas, 2018). Also bacteriocins associated with carrot juice fermentation in particular were already found, showing a maximum antimicrobial activity against a broad spectrum of pathogens and especially E. coli O157 (Joshi, Sharma, & Rana, 2006). Notwithstanding the growth, the pathogens were not detected by 16S rRNA amplicon sequencing (when initial inoculated at 10³ CFU/ml), their amounts were at all times, much lower as the higher number of presumptive Enterobacteriaceae and the LAB present dominating the fermentation. The latter LAB numbers were easily up to 109 CFU/ml (Wuyts et al., 2018). So even when Salmonella reached amounts up to 105 CFU/ml at day 1 (this time when highest numbers present of the three pathogens tested), this would still account for less than 0.01% of the total community, which explains why they could never be detected by means of the current RNA sequencing method only targeting the abundant microbiota present. The faculty of bioscience engineering (Ghent University), more particular, the Lab Food Microbiology & Food preservation (LFMFP) and the Center for Microbial Ecology and Technology (CMET) did a research project in 2015 for the FASFC with similar set-up and results on cheese. As the raw milk soft cheeses also contained high numbers of LAB (ca. 108 CFU/g) and due to often low abundancies of the pathogen (i.e. L. monocytogenes < 0.01%), it could not be found by 16S rDNA amplicon (V3-V4 region) sequencing among thus the abundant LAB. And even when being in the range of 0.01-1% of the total LAB count metagenomic analysis thus did not systematically allow detecting this pathogen contamination. The study concluded, metagenomics could not offer an added value in comparison to classical plating methods to judging the public health risks by pathogen contamination of this type of fermented food. The density of the pathogen was too low, the bacterial community was inert to the presence of the pathogen, the diversity of the community was insufficient and the classical plating method performed better, with more positive samples (Boon, Uyttendaele, & Roume, 2016).

Knowing this, RNA analysis was still performed. The aim was not to find the pathogens with this technique (rather selective media and thus classical culture methods were used for this purpose), but to see if the addition of the pathogens could somehow affect the microbial community in the fermentation. Also, the effect of cucumber juice and cooling on the microbial community were investigated by RNA analysis.

To verify the residual presence of the pathogens (inoculation of 10³ CFU/ml) on time points where they fell below LOD (i.e. 10 CFU/ml), enrichment methods were used. Only the pathogens at day 1 could be confirmed. On the other days (3, 15 and 30) no CFUs were found. The negative result for Salmonella and Escherichia of day 3 are in fact false negatives, since their presence was confirmed on the counting plates. The fact pathogens were not found during enrichment could be attributed to multiple effects (synergistic effect of adverse condition executed by freezing, acid pH, bacteriocins). Firstly, the freezing effect. It should be taken into account that only as of day 15 on (and thus not yet on day 3 of the experiment) the samples taken and subjected in a later phase to enrichment were diluted (50/50) in BPW prior to freezing (-24°C). This to temper the acid pH of the fermented carrot juice or vegetable mix (FCJ + CUJ). The death of E. coli O157 during a freeze/thaw process varies depending on the strain, the method to culturing survivors and the thawing method (Sage & Ingham, 1998). Thawing was in this case done at room temperature which is supposed to be the least lethal to the pathogen. But still no pathogens were found at for example day 3, where they were supposed to. Freezing results in a significant log reduction as a result of fatal cell damage or injury. The inactivation level significantly increases when the storage time is increased (Gao, Smith, & Li, 2007). Various methods of freezing and thawing have little effect on L. monocytogenes. The freeze effect did probably not have a big influence on the survival of this type of pathogen (Kataoka, Wang, Elliott, Whiting, & Hayman, 2017). Specifically, for Salmonella, it has been reported that structurally damaged cells are unable to grow on media containing sodium deoxycholate such as XLD. The sublethal injury and thus decrease of healthy cells is significant during freezing, but in all the cases it cannot be concluded that freezing on its own is a sufficient tool for destruction of pathogens. The reductions are primarily attributed to the existence of injured cells which may not grow on selective media (Manios & Skandamis, 2015). Also, the fermentation effects as described above, had already an effect on the pathogens at day 3 and so on. The pathogens were all in a declining stage or even not present on the plates at day 3. The pH effect was slightly tempered (by addition of BPW) like mentioned above, but the pathogens were not kept under optimal storage conditions (e.g. storage in glycerol). At last the heterogeneity effect, when the pathogens are not equally divided over the product, and in that way the sample is not representative (by accident took a sample without the pathogen). Using a homogenising step (e.g. stomacher) minimizes this problem but does not eliminate it, especially with very low contamination levels (EFSA, 2013). If CFUs were found during the main experiment (counting plates), and/or the subsequent enrichment of that sample tested positive, one could conclude the pathogen was present in the sample on that specific day. However, the opposite is not true, if the pathogen was not found on the counting plates and also the enrichment tested negative, one cannot be sure about the result as it could be a false negative (presence < LOD). This chance is rather low since the methods used are well known and validated and should guaranty the absence (EFSA, 2013).

All the effects described above are disadvantageous (on the growth/presence) for the pathogens, but are not yet fully developed at the beginning of the fermentation. That is why the most critical point of supressing the pathogens was at the start of the fermentation. For this reason, the same experiment was repeated with higher inoculation levels of pathogens, to check if the fermentation could still suppress the growth at the beginning. During the fermentation all effects will be more explicit and probably form a synergistic effect, which is antagonistic to the food-borne pathogens (Allende et al., 2007; Schvartzman, Belessi, Butler, Skandamis, & Jordan, 2011). To assess the impact of the pathogen cocktail, RNA based amplicon sequencing was also performed. When analysing the bacterial community, we chose to use RNA instead of DNA, this based on previous expertise in the lab (Wuyts et al., 2019). RNA is less stable but is only present in living matter. For that reason, no bias will be present of the carrot tissues or dead bacteria, resulting in a better overview of the active population. This reasoning is only applicable on closed systems, since using an open one would drain off the dead material. Meanwhile, transforming RNA to cDNA produces an extra bias on the results. Introducing an extra processing step can always cause additional errors. For example, the tendency of various reverse transcriptases to generate spurious second-strand cDNA (Ozsolak & Milos, 2011). Looking at the bacterial community (Figure 20), the inoculation of ± 10³ CFU/ml of pathogens did not affect the l microbial community The same shift of relative abundancies was found as already demonstrated by previous research (Wuyts et al., 2018). At the start (day 1 and 3) Leuconostoc, Yersinia and Enterobacter dominate the microbial community. After 15 days a shift has occurred where Lactobacillus 1 (probably plantarum (Wuyts et al., 2018)) dominates. Lb. plantarum is known to have a broad spectrum of antimicrobial activity against a lot of spoilers and pathogens, including the pathogens tested in this study (Dinev et al., 2018). Even if this is only qualitative data (relative abundancies), the presence of Yersinia is still confirmed at day 15 and 30, so if a pathogenic strain would contaminate the fermentation, they could be present at high numbers and thus could cause possible harm to the consumer (referring as a benchmark to the inoculated pathogens that were not detected with RNA analysis, i.e. even if present up to 5 log CFU/ml). The presence of more than 25% of Yersinia (relative abundancy) in all carrot juice samples during fermentation at day 1 and also in high percentages at day 3, confirms the need of: 1) Analysing and identifying which Yersinia species is present in the carrot juice 2) Performing challenge testing with pathogenic Y. enterocolitica or pseudoturberculosis strains, to see if these pathogenic Yersinia strain could persist (and grow) in a fermenting environment. The results of the present study can only give relevant information up to genus level; one cannot conclude which species of Yersinia was present. In general Yersinia spp. causes often spoilage, depending on the presence of virulence factors pathogenicity could be attributed to certain strains (Frank Devlieghere et al., 2016; Özdemir & Arslan, 2015). But thus the 16 S rRNA analysis of this study cannot go to such deep levels (rather a broad screening for dominant microbial groups or taxa), and thus in that way the metagenetics approach used here are inadequate to confirm pathogenicity. To actually identify the Yersinia species to species level and find possible pathogens, isolates should be obtained and further characterised (by PCR targeting virulence factors, biotyping or serotyping or by whole genome sequencing). The alpha diversity (i.e. the diversity within a sample) decreases over time. At the start of the fermentation different species (belonging to the natural carrot microbial community) are present in 66

relative abundancies above 20%, shifting to a dominant ASV (Lb. 1 > 50%) as of day 15 of the fermentation. This trend occurs in every sample, and is best visible comparing the spontaneous samples (S) over time on the Figure 10, or via the red dots on Figure 11. The results show that the carrot juice fermentation is finally dominated by one ASV (i.e. Lb. 1), this trend has also been found on different other vegetable substrates, with other species dominating (Lucena-Padrós, Caballero- Guerrero, Maldonado-Barragán, & Ruiz-Barba, 2014) (Zabat et al., 2018) (Liang, Yin, Zhang, Chang, & Zhang, 2018). How much the alpha diversity will decrease during fermentation is depending on the substrate used (type of vegetable) and even the region where the substrate is derived from (Peng et al., 2018). The presence of the pathogens during this experiment did not affect the alpha diversity, where the fermentation still converged in all samples to one dominating and similar ASV belonging to the plantarum group (Wuyts et al., 2018). Statistical analysis is limited since it is not straight forward for application on this type of data. Relative abundancy results in dependent-taxa and another common problem of this type of data is the large amount of 0 (sparsity) (due to taxa being present in only 1 sample). For those reasons a visualisation approach (use of plots) was preferred over a statistical one, in order to obtain an exploratory view of the results. In this way the data could better be understood, and where possible, a statistical approach was applied to test more targeted hypotheses. The variation within samples is low, which is the result one indeed aims for when doing a spontaneous fermentation. Addition of salt, shifting to anaerobic conditions (human intervention) will favour the community to one dominating ASV.

The beta diversity showed variation between samples of the same day, but the effect was the same with all samples, inoculated with pathogens or not. The beta diversity (i.e. diversity between samples) was visualised using PCoA to reduce the high dimensionality of the data. The figures representing the PCoA variance are in fact a 2D representation of the results, where the PCoA percentages represent how well the figure has fit the multidimensionality, a value close the 100% means a good representation. Samples of day 1 and 3 clustered together according to PCoA1 and samples of day 15 and 30 showed the same pattern for PCoA2. When including the data of the samples during cooling, the figure gives a cluster containing samples of day 1 and 3 (Figure 21), at this moment the natural carrot juice community is still present, all originating from the same batch. The samples of day 15 and 30 are less clustered, because at this moment the community already shifted and more variance is present between samples. During time the fermenting conditions can vary more between samples (little temperature variations, pH fluctuations between replicates, …). The paragraph above highlights the limitations of using a PCoA method. Only a part of the variation is represented, and the representation of it is very dependent of the dataset used. The method is useful during exploratory research. Afterwards it is still possible to go further into detail with other options like clustering on basis of all taxa. However, the dataset should be a lot bigger to get significance using statistics.

4. Refrigeration (7.5°C) and addition of cucumber juice After a fermentation of 30 days, the carrot juice was transferred to the refrigerator. Some samples were mixed with fresh cucumber juice which was whether or not inoculated with pathogens to investigate a possible cross-contamination route.

Even after addition of cucumber juice (75%) the pH (maximum of 4.55) stayed beneath the food fermentation safety threshold of 4.6. The samples were stored afterwards in the refrigerator. In general LAB do not optimally grow and produce acid at these refrigeration temperatures, resulting in a relatively constant pH (Amézquita & Brashears, 2002). Plate counts on MRS during refrigeration also showed that the number of LAB (± 8.1 log CFU/ml) remained constant throughout the refrigerated storage, no outgrowth but also no die-off seemed to occur of the LAB. None of the freshly inoculated pathogens simulating post-contamination when FCJ was mixed with CUJ showed growth during refrigeration.

During the analysis of the pathogen development immediately after mixing (adding of freshly contaminated CUJ) or at intermediate time points during refrigerated storage (day 31-38) no quantification of E. coli O157 could be done as, by accident, CT-SMAC plates were prepared with too high antibiotic concentrations. Consequently, no CFUs or no representable quantities were found on the plates (this part of the experiment should be repeated in due time). At day 38 E. coli O157 was plated on CT-SMAC with the correct (non-inhibitory) kanamycin concentrations, and was not retrieved by plate counts (< LOD), but later on confirmed with enrichment procedures in two out of three samples. As considers L. monocytogenes, this pathogen died off to levels near or below the limit of the detection (< 10 CFU/ml) after 8 days. According to literature bacteriocins could still have a bactericidal effect on L. monocytogenes at refrigeration (Allende et al., 2007). Together with a low pH (and also lactic acid) a combined antimicrobial effect can be exerted (Amézquita & Brashears, 2002). Of note, Scatassa et al. (2017) performed a similar experiment where cheese was inoculated with L. monocytogenes and LAB, that showed that the pathogen fell also under the LOD after 15 days of refrigeration. Salmonella on the other hand, is more resistant to the effect of lactic acid at refrigeration compared to the other two pathogens (Anang, Rusul, Bakar, & Ling, 2007). Also, in our experiment Salmonella was the only pathogen still detectable on plate after 8 days of refrigeration (although numbers recovered were slowly decreasing throughout storage time).

Further follow-up was also done of the (initial at day 0) inoculated carrot juice, to see if any of the pathogen that died-off to non-detects (< 10 CFU/ml) during 30 days prior fermentation of CAJ could still resuscitate and maybe grow and thus be detected again when the FCJ was mixed with CUJ and kept under refrigeration. Under refrigeration, fresh pathogens were only inoculated in a mix with cucumber juice (row 1, Figure 13). L. monocytogenes was never resuscitated in refrigerated samples of pure fermented carrot juice (FCJ) (also not that FCJ that was derived from at day 0 inoculated CAJ). Only in FCJ also mixed with cucumber juice L. monocytogenes was sometimes present. In addition, these L. monocytogenes CFUs were not only detected in samples of FCJ-CUJ mix derived from at day 0 inoculated CAJ, but also on an ALOA plate from FCJ-CUJ mix derived from at day 0 uninoculated CAJ. This phenomenon most probably refers to natural occurring post-contamination originating from the cucumber juice. It is more likely that Listeria was present on the cucumbers used for juicing, compared to the resuscitation of the inoculated strain at day 0 in the carrot juice at start of fermentation (more than 30 days earlier). This also because no L. monocytogenes was found in the parallel samples of pure fermented carrot juice during refrigeration (whether inoculated or not with the pathogen at day 0). The presence of L. monocytogenes in the cucumber mix samples was only a mere detection, i.e. a low number of < 4 CFU/plate. Considering Listeria cannot grow during shelf life due to acid pH (pH < 4.4), the amounts of all samples stayed under the guideline of 100 CFU/g (Uyttendaele et al., 2018). 68

According to outbreak data available, the presence of L. monocytogenes in food represents a very low risk for the generic (non-susceptible) population group, when the concentration is below 100 CFU/g (European Commision, 1999). During the residual detection of the pathogens by enrichment of the challenge tests (using the freshly inoculated pathogens on day 30) in refrigeration, it was noted that the pathogens could be more often recovered as in the situation where the pathogens were inoculated at the start of the fermentation (inoculated day 0 in carrot juice). During the challenge test of the refrigerated storage, some were still found during enrichment procedures after 8 days of storage (while non after 3 days inoculated at day 0, i.e. undergoing the fermentation). The amount of pathogens was probably not the main reason of survival during the experiment and afterwards in the freezer (since on day 3 for example 105 CFU/ml were present during fermentation and only 103 CFU/ml on day 31 and so on during refrigerated storage, and still only those of the refrigerator part were found after enrichment.). The pathogens during refrigeration were probably in a less injured state and also stored for a shorter time at -24°C. Also, the fact that the LAB were not growing during refrigeration, and the pH was altered by addition of cucumber juice played a possible role. There was a difference in recovery depending upon the pathogen. L. monocytogenes almost disappeared completely of the plates after 8 days (day 30 till 38), and was not found after enrichment. The pathogen is very sensitive to the presence of lactic acid. It was more probable that L. monocytogenes was deteriorated by LAB during refrigeration as by the freezing step (Scatassa et al., 2017)(Kataoka et al., 2017). E. coli O157 and S. Typhimurium were both still found in two out of three samples at day 38, these pathogens are more resistant to the presence of LAB and lactic acid, this effect was also confirmed during the fermentation (Table 20). Their cells were probably less injured (shorter freezing effect) as during most of the fermentation samples and in that way still retrieved during the enrichment (Anang et al., 2007). The presence of Salmonella was already confirmed on the counting plates. None of the samples containing L. monocytogenes, which originated probably from the cucumber juice, were confirmed by enrichment. A possible post-contamination with the addition of the cucumber juice could cause a problem, only if amounts higher than the infectious dose are present. No growth was found for any of the three pathogens, and S. typhimurium seemed to be most resistant to the fermented carrot juice - cucumber juice mix (ratio 25%-75%). Even if L. monocytogenes still decreased during 8 days of refrigeration, presence in higher numbers as the infectious dose should be avoided since after mixing with cucumber juice direct consumption aimed to be possible. Besides the effect on the pathogens, the cooling seemed to have some effect on the community (Figure 14), Lactobacillus 1 lost some of its dominance after 8 days. Lb. brevis took a bigger share of the relative abundancy. And also Enterobacter and Leuconostoc gained in relative abundancy. Leuconostoc mesenteroides has already been found to grow at temperatures below 10°C and outgrowing in that case for example Lb. plantarum, this could also be here the case (Hamasaki, Ayaki, Fuchu, Sugiyama, & Morita, 2003). More species of lactobacilli were also present among the 11 most relative abundant species compared to the fermenting end conditions. Every sample originated from another Weck jar, so certain scepticism is needed analysing the results. Also, a difference was visible between the replicates. But this result gives a possible indication that Lb. plantarum, known from previous research (Wuyts et al., 2018), could lose relative abundancies due to cooling, and other species can take advantage under those conditions (7.5°C) (Hamasaki et al., 2003).

Conclusion The spontaneous fermentation of carrot juice had a suppressive effect on the different pathogens added (10³ CFU/ml each) at the start of the fermentation. L. monocytogenes was most sensitive to the fermenting conditions and fell under the detection limit (10 CFU/ml) after 3 days. E. coli O157 and S. Typhimurium showed growth after day 1 (E. coli O157 > S. Typhimurium) but these numbers declined again at day 3, and after 15 days of ambient temperature fermentation both pathogens were also non- detectable using standard enumeration methods. As of day 15, even absence in 10 ml sample was confirmed by enrichment procedures. The first trial a 4-log reduction was noted. Often in food safety management a 5-6 log reduction is required. Therefore, challenging fresh carrot juice prior to fermentation with higher numbers was established. When inoculating 105 CFU/ml the same pattern was observed where L. monocytogenes was again most sensitive (no growth) and fell together with S. Typhimurium under the LOD after 6 days of fermentation. After 9 days E. coli O157 formed also non- detects (< 10 CFU/ml). At least a 5-log reduction (6 log reduction for Salmonella & E. coli O157) was proven for all three food-borne pathogens. One can thus conclude that the spontaneous fermentation of carrot juice is robust enough to counter the survival or persistence of the three pathogens if they would be present in the juiced fresh carrots when starting the fermentation. A possible post-contamination (± 10³ CFU/ml) potentially occurring when mixing the fermented carrot juice (25%) with fresh cucumber juice (75%) was reduced to non-detectable per 10 ml in case of L. monocytogenes after 8 days of refrigeration. However, S. Typhimurium numbers remained constant during refrigeration and thus precautions should be taken to avoid post-contamination. Also, the residual presence of E. coli O157 (presence in 10 ml) was confirmed after 8 days of refrigeration. The latter two pathogens should not be tolerated in higher numbers as the infectious dose is reported to be low and thus to avoid food-borne infections, one should seek to have absence per 25 ml (the safety limit usually set for S. Typhimurium & E. coli O157). L. monocytogenes could be tolerated at somewhat higher levels if the fermented juice mix is not destined for a vulnerable group (i.e. if not served to pregnant, immunosuppressed or aging (> 65 years) persons, YOPI). If the fermented juice mix is not supporting the growth of L. monocytogenes during its shelf life and the ultimate date/time of consumption is respected (in present study after 8 days storage under refrigeration, < 7.5°C) up to 100 CFU/ml might be tolerated (Uyttendaele et al., 2018) (although by preference also this pathogen is absent per 25 ml, to ensure also safety for vulnerable persons at risk for listeriosis).

The possible contamination of these three pathogens at the beginning of the fermentation did not affect the rapid pH decline (from initial pH 6.1 to pH 3.7) and the standard shift of microbial community (when inoculating 10³ CFU/ml) during the fermentation towards a dominance of lactic acid bacteria such as Leuconostoc and Lactobacillus species, with the most abundant species noted to be as expected Lb. plantarum. The high relative abundancies of Yersinia/Enterobacter species at the beginning of the fermentation should be investigated, to see if a pathogenic strain could survive the fermentation. Refrigeration (7.5°C) and addition of cucumber juice affected the community of the spontaneous (uninoculated) fermented carrot juice. A community shift was visible but further research on this topic is needed to get more significant and representative data. Even after growth of Salmonella and E. coli O157 (up to almost 5 log CFU/ml) still not a single read could be detected using the 16S rRNA sequencing approach. Thus, to find pathogens, certainly in a dense microbial community, 16S amplicon sequencing will not be sufficient. More targeted methods are required, presently done by selective plating and enrichment procedures. 71

Relying on the results of this Master Thesis some microbiological food safety guidelines and warnings could be formulated for homemade spontaneous fermented carrot juice if used as such or as the basis in a mix for plant-based beverages:

 One should ferment the juice for at least 9 days, to ensure its suppressive (> 5 log reduction) effect on different pathogens i.e. Listeria monocytogenes, Salmonella Typhimurium and Escherichia coli O157.  When used as the basis in a mix for plant-based beverages and kept in refrigeration: verify if pH remains < 4.6 upon making mixes. Furthermore, absence of Salmonella & E. coli O157 in 25ml is required. L. monocytogenes numbers of up to max. 100 CFU/ml can be tolerated noting no growth potential during up to 1-week storage at < 7.5°C.  To prevent (post)contamination: start with good quality fresh produce (carrots, cucumbers) used for juicing, to respect GAP & take care of GHP during storage and further handling (cutting or juicing) (EFSA BIOHAZ, 2014).

If serving of fermented vegetable juice mixes to vulnerable groups, recommendations for pasteurisation of fresh juice used for mixing with the fermented carrot juice is necessary (pasteurisation by 2’ 70°C or equivalent will ensuring absence of Salmonella / E. coli but also L. monocytogenes in 25 ml as required)

Further research Amplicon sequencing results clearly show a high presence of Yersinia spp. within the carrot juice. Certainly, at the beginning of the fermentation, this bacteria is present in high relative abundancies (day 1 and 3 = ± 25%, and day 15 still around 5%). The fact that the genera of the pathogens inoculated during the main experiment (on day 0 and 30) were not detected during amplicon sequencing, indicates to a presence of Yersinia in much higher amounts. For this reason, it would be useful to optimise the detection of pathogenic Yersinia in this wide variety of natural occurring strains (e.g. by selective plating, culture independent approaches). Also, an investigation on possible pathogenic Yersinia strains (enterocolitica, pseudotuberculosis) in the natural carrot juice community could be useful (inoculation at fermentation start), this to provide some microbial safety guidelines on this pathogen too.

At the start of the fermentation the pathogens inoculated at day 0 were able to grow in the fermentation. A research on possible faster growth suppression of the fermentation by addition of a suitable starter culture (e.g. Leuconostoc spp.) should be explored. In that way the fermentation could be considered faster as safe according to this pathogens. Since it is known Leuconostoc mesenteroides plays an important role at the beginning of the carrot juice fermentation, it could be useful testing the impact of different strains on the growth inhibition of the different pathogens. Some CFUs of Listeria monocytogenes were found after addition of non-pasteurised cucumber juice. It could be useful to investigate if those CFUs originated from the cucumber juice or the original inoculated carrot juice. Addition of pasteurised cucumber juice could already provide an answer on this topic, or comparing the sequence of an isolate with the original pathogenic strain (e.g. whole genome sequencing).

If the strains of L. monocytogenes were associated to the inoculum on day 0, one could test the resistance of it by submission in a new fermentation. In that way, if a more resistant subpopulation of the original pathogen was acquired, some extra warnings could be formulated when fermentations are made using a backslopping approach.

Also, the effect of addition of cucumber juice and cooling on the fermented carrot juice microbial community should be analysed in more detail. Repeating the refrigeration part of the main experiment, could already provide additional data and in that way a better/clearer shift due to cooling and cucumber juice on the microbial community could be described.

References

Abadias, M., Usall, J., Anguera, M., Solsona, C., & Viñas, I. (2008). Microbiological quality of fresh, minimally-processed fruit and vegetables, and sprouts from retail establishments. International Journal of Food Microbiology. https://doi.org/10.1016/j.ijfoodmicro.2007.12.013

Adams, M., & Mitchell, R. (2002). Fermentation and pathogen control: A risk assessment approach. In International Journal of Food Microbiology. https://doi.org/10.1016/S0168-1605(02)00181-2

Allende, A., Martínez, B., Selma, V., Gil, M. I., Suárez, J. E., & Rodríguez, A. (2007). Growth and bacteriocin production by lactic acid bacteria in vegetable broth and their effectiveness at reducing Listeria monocytogenes in vitro and in fresh-cut lettuce. Food Microbiology. https://doi.org/10.1016/j.fm.2007.03.002

Amézquita, A., & Brashears, M. M. (2002). Competitive inhibition of Listeria monocytogenes in ready-to-eat meat products by lactic acid bacteria. Journal of Food Protection. https://doi.org/10.4315/0362-028X-65.2.316

Anang, D. M., Rusul, G., Bakar, J., & Ling, F. H. (2007). Effects of lactic acid and lauricidin on the survival of Listeria monocytogenes, Salmonella enteritidis and Escherichia coli O157:H7 in chicken breast stored at 4 °C. Food Control. https://doi.org/10.1016/j.foodcont.2006.05.015

Arena, M. P., Silvain, A., Normanno, G., Grieco, F., Drider, D., Spano, G., & Fiocco, D. (2016). Use of Lactobacillus plantarum strains as a bio-control strategy against food-borne pathogenic microorganisms. Frontiers in Microbiology. https://doi.org/10.3389/fmicb.2016.00464

Atkinson, S., & Williams, P. (2016). Yersinia virulence factors - a sophisticated arsenal for combating host defences. F1000Research. https://doi.org/10.12688/f1000research.8466.1

Barbuti, S., & Parolari, G. (2002). Validation of manufacturing process to control pathogenic bacteria in typical dry fermented products. Meat Science. https://doi.org/10.1016/S0309-1740(02)00124-9

Baschali, A., Tsakalidou, E., Kyriacou, A., Karavasiloglou, N., & Matalas, A.-L. (2017). Traditional low-alcoholic and non- alcoholic fermented beverages consumed in European countries: a neglected food group. Nutrition Research Reviews, 30(01), 1–24. https://doi.org/10.1017/S0954422416000202

BDA. (2016). Policy Statement Neutropenic Dietary Advice for Haematology Patients. Birmingham. Retrieved from https://www.bda.uk.com/improvinghealth/healthprofessionals/policy_statements/neutropenic_dietary_advice_for_ haematology_patients

Bear, R. A., Rintoul, D. A., Snyder, B. A., Smith-Caldas, M., Herren, C. D., & Horne, E. A. (2016). Principles of Biology. Kansas: New Prairie Press. Retrieved from https://books.google.be/books?id=yWVoAQAACAAJ&dq=robert+bear+principles+of+biology&hl=nl&sa=X&ved=0ahU KEwi1pJDz0P7dAhWHmbQKHVrrCEcQ6AEIKzAA

Becton Dickinson. (2009). Difco and BBL Manual: Manual of Microbiological Culture Media. (M. J. Zimbro, D. A. Power, S. M. Miller, G. E. Wilson, & J. A. Johnson, Eds.) (2nd ed.). Maryland: BD Diagnostics . Retrieved from http://www.bd.com/resource.aspx?IDX=9572

Becton Dickinson. (2013a). INSTRUCTIONS FOR USE-READY-TO-USE PLATED MEDIA: BD XLD Agar (Xylose-Lysine- Desoxycholate Agar). Heidelberg. Retrieved from http://www.bd.com/resource.aspx?IDX=8991

Becton Dickinson. (2013b). Instructions for use-ready-to-use plated media: BD Yersinia Selective Agar (CIN Agar). Heidelberg. Retrieved from http://www.bd.com/resource.aspx?IDX=8992

Ben Amor, K., Vaughan, E. E., & de Vos, W. M. (2007). Advanced molecular tools for the identification of lactic acid bacteria. The Journal of Nutrition. https://doi.org/10.1093/jn/137.3.741S

Bernardeau, M., Vernoux, J. P., Henri-Dubernet, S., & Gueguen, M. (2008). Safety assessment of dairy microorganisms: The Lactobacillus genus. International Journal of Food Microbiology. https://doi.org/10.1016/j.ijfoodmicro.2007.08.015

Beuchat, L. R., & Brackett, R. E. (1990). Inhibitory effects of raw carrots on Listeria monocytogenes. Applied and Environmental Microbiology.

Beuchat, L. R., Brackett, R. E., & Doyle, M. P. (1994). Lethality of carrot juice to Listeria monocytogenes as affected by pH, sodium chloride and temperature. J. of Food Prot. https://doi.org/10.4315/0362-028X-57.6.470

Biasino, W., De Zutter, L., Woollard, J., Mattheus, W., Bertrand, S., Uyttendaele, M., & Van Damme, I. (2018). Reduced contamination of pig carcasses using an alternative pluck set removal procedure during slaughter. Meat Science. https://doi.org/10.1016/j.meatsci.2018.05.019

Biokar diagnostics. (2009). MACCONKEY SORBITOL AGAR (CT-SMAC). Beauvais Cedex. Retrieved from https://dongnamlab.com/publics/product/854/files/MacConkey_sorbitol_agar_(CT-SMAC)_TDS_BK147_v5.pdf

Biolife. (2010). ALOA ® AGAR LISTERIA ACC. TO OTTAVIANI & AGOSTI-ALOA ® ALOA ® ENRICHMENT-SELECTIVE SUPPLEMENTS. Retrieved from www.biolifeitaliana.it

BOGK, Der Bundesverband der obst-, gemüse-und kartoffelverarbeitenden. (2012). THE 4TH INTERNATIONAL KIMCHI CONFERENCE OVERVIEW. In C. Freitag (Ed.), OVERVIEW ABOUT PRODUCTION AND MARKET OF SAUERKRAUT (p. 50). Washington D.C.

Bonatsou, S., Tassou, C., Panagou, E., & Nychas, G.-J. (2017). Table Olive Fermentation Using Starter Cultures with Multifunctional Potential. Microorganisms. https://doi.org/10.3390/microorganisms5020030

Borresen, E. C., Henderson, A. J., Kumar, A., Weir, T. L., & Ryan, E. P. (2012). Fermented Foods: Patented Approaches and Formulations for Nutritional Supplementation and Health Promotion. Nutrition & Agriculture. https://doi.org/10.2174/2212798411204020134

Breidt, F., Mcfeeters, R. F., Perez-Diaz, I., & Lee, C.-H. (2013). Fermented vegetables. Food Microbiology: Fundamentals and Frontiers (4th ed.). Washington D.C.: ASM Press. https://doi.org/10.1128/9781555818463.ch33

Burger, R. (2012). EHEC O104:H4 in Germany 2011: Large outbreak of bloody diarrhea and haemolytic uraemic syndrome by Shiga toxin-producing E. coli via contaminated food. In . Improving Food Safety Through a One Health Approach: Workshop Summary.

Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., & Holmes, S. P. (2016). DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods. https://doi.org/10.1038/nmeth.3869

Callejón, R. M., Rodríguez-Naranjo, M. I., Ubeda, C., Hornedo-Ortega, R., Garcia-Parrilla, M. C., & Troncoso, A. M. (2015). Reported Foodborne Outbreaks Due to Fresh Produce in the United States and European Union: Trends and Causes. Foodborne Pathogens and Disease. https://doi.org/10.1089/fpd.2014.1821

Canchaya, C., Claesson, M. J., Fitzgerald, G. F., van Sinderen, D., & O’Toole, P. W. (2006). Diversity of the genus Lactobacillus revealed by comparative genomics of five species. Microbiology. https://doi.org/10.1099/mic.0.29140-0

Caplice, E., & Fitzgerald, G. F. (1999). Food fermentations: Role of microorganisms in food production and preservation. International Journal of Food Microbiology, 50(1–2), 131–149. https://doi.org/10.1016/S0168-1605(99)00082-3

Capozzi, V., Fragasso, M., Romaniello, R., Berbegal, C., Russo, P., & Spano, G. (2017). Spontaneous Food Fermentations and Potential Risks for Human Health. Fermentation. https://doi.org/10.3390/fermentation3040049

Chang, J. Y., & Chang, H. C. (2011). Growth Inhibition of Foodborne Pathogens by Kimchi Prepared with Bacteriocin-Producing Starter Culture. Journal of Food Science. https://doi.org/10.1111/j.1750-3841.2010.01965.x

Cheigh, H., Park, K., & Lee, C. Y. (1994). Biochemical, microbiological, and nutritional aspects of kimchi (Korean fermented vegetable products). Critical Reviews in Food Science and Nutrition, 34(2), 175–203. https://doi.org/10.1080/10408399409527656

Choi, N.-E., & Han, J. H. (2015). How flavor works : the science of taste and aroma. Retrieved from https://books.google.be/books?id=vjitBQAAQBAJ&pg=PA103&dq=discovery+of+cheese+by+arab+trader&hl=nl&sa=X &ved=0ahUKEwi72c75x9jdAhVECewKHenkD6kQ6AEINjAC#v=onepage&q=cheese&f=false

Cifelli, C. (2017). Advice from a Nutritionist: Eat More Fermented Foods. | International Scientific Association for Probiotics and Prebiotics (ISAPP). Retrieved January 15, 2019, from https://isappscience.org/advice-nutritionist-eat-fermented- foods/

Claesson, M. J., Van Sinderen, D., & O’toole, P. W. (2007). The genus Lactobacillus - a genomic basis for understanding its diversity. Federation of European Microbiological Societies, 269, 22–28. https://doi.org/10.1111/j.1574- 76

6968.2006.00596.x

Clark, G. (2010). Industrial Revolution. In Economic Growth (pp. 148–160). London: Palgrave Macmillan UK. https://doi.org/10.1057/9780230280823_22

Corrieu, G., & Béal, C. (2016). Yogurt: The Product and its Manufacture. In Encyclopedia of Food and Health (pp. 617–624). Elsevier. https://doi.org/10.1016/B978-0-12-384947-2.00766-2

Del Rocío López-Cuellar, M., Rodríguez-Hernández, A.-I., & Chavarría-Hernández, N. (2016). LAB bacteriocin applications in the last decade. Biotechnology & Biotechnological Equipment, 30(6), 1039–1050. https://doi.org/10.1080/13102818.2016.1232605

Delves-Broughton, J. (2005). Nisin as a food preservative. Food Australia. https://doi.org/10.1080/10408399409527650

Devlieghere, Franck, Debevere, J., Jacxsens, L., & Uyttendaele, M. (2011). Levensmiddelenmicrobiologie- en conservering. Die Keure. https://doi.org/10.3795/KSME-B.2013.37.1.081

Devlieghere, Frank, Debevere, J., Jacxsens, L., Rajkovic, A., Uyttendaele, M., & Vermeulen, A. (2016). Levensmiddelenmicrobiologie en -conservering. die Keure. Retrieved from https://www.diekeure.be/nl- be/professional/7550/levensmiddelenmicrobiologie-en-conservering

Devlieghere, Frank, Jacxsens, L., Uyttendaele, M., & Vermeulen, A. (2011). Levensmiddelenmicrobiologie en -conservering. die Keure. Retrieved from https://www.diekeure.be/nl-be/professional/7550/levensmiddelenmicrobiologie-en- conservering

Di Cagno, R., Coda, R., De Angelis, M., & Gobbetti, M. (2013). Exploitation of vegetables and fruits through lactic acid fermentation. Food Microbiology. https://doi.org/10.1016/j.fm.2012.09.003

Dinev, T., Beev, G., Tzanova, M., Denev, S., Dermendzhieva, D., & Stoyanova, A. (2018). Antimicrobial activity of lactobacillus plantarum against pathogenic and food spoilage microorganisms: A review. Bulgarian Journal of Veterinary Medicine. https://doi.org/10.15547/bjvm.1084

Dinu, L. D., & Bach, S. (2011). Induction of viable but nonculturable Escherichia coli O157:H7 in the phyllosphere of lettuce: A food safety risk factors. Applied and Environmental Microbiology. https://doi.org/10.1128/AEM.05020-11

Duan, R., Liang, J., Shi, G., Cui, Z., Hai, R., Wang, P., … Wanga, X. (2014). Homology analysis of pathogenic yersinia species yersinia enterocolitica, yersinia pseudotuberculosis, and yersinia pestis based on multilocus sequence typing. Journal of Clinical Microbiology. https://doi.org/10.1128/JCM.02185-13

EFSA. (2013). Analysis of the baseline survey on the prevalence of Listeria monocytogenes in certain ready-to-eat foods in the EU, 2010-2011 Part A: Listeria monocytogenes prevalence estimates. EFSA Journal. https://doi.org/10.2903/j.efsa.2013.3241

EFSA (European Food Safety Authority). (2017a). Trends and Sources of Zoonoses and in Humans , Foodstuffs , Animals and in 2016, Belgium. EFSA Journal.

EFSA (European Food Safety Authority), E. (European C. for D. prevention and control). (2017b). The European Union summary report on trends and sources of zoonoses, zoonotic agents and food-borne outbreaks in 2016. EFSA Journal (Vol. 15). https://doi.org/10.2903/j.efsa.2016.4634

EFSA BIOHAZ (European Food Safety Panel on Biological Hazards). (2014). Scientific Opinion on the risk posed by pathogens in food of non-animal origin. Part 2 (Salmonella, Yersinia, Shigella and Norovirus in bulb and stem vegetables, and carrots). EFSA Journal, 12(12), 91. https://doi.org/10.2903/j.efsa.2014.3832

European Commision: Health & Consumer Protection Directorate-General. (1999). Opinion of the scientific committee on veterinary measures relating to public health on Listeria monocytogenes. Directorate B - Scientific Health Opinions. https://doi.org/10.1089/fpd.2011.0830

European Commision. (2002). Risk Profile on the Microbiological Contamination of Fruits and Vegetables Eaten Raw. Report of the Scientific Committee on Food. Retrieved from http://europa.eu.int/comm/food/fs/sc/scf/index_en.html%5Cnhttp://scholar.google.com/scholar?hl=en&btnG=Sear ch&q=intitle:Risk+Profile+on+the+Microbiological+Contamination+of+Fruits+and+Vegetables+Eaten+Raw#0

European Commision. (2005). Commission regulation on microbiological criteria for foodstuffs (Text with EEA relevance). 77

Official Journal of the European Union, 338/1-338/25. Retrieved from https://eur-lex.europa.eu/legal- content/EN/TXT/PDF/?uri=CELEX:32005R2073&from=EN

European Union Reference Laboratory (EURL) for Salmonella. (2012). Sampling whole melons to check for the presence of Salmonella. Retrieved from http://www.cdc.gov/salmonella/typhimurium-cantaloupe-08-12/

Eurostat. (2017). Sustainable development in the European Union MONITORING REPORT ON PROGRESS TOWARDS THE SDGS IN AN EU CONTEXT 2017 edition. https://doi.org/10.2785/842104

Fox, P. F. (1993). Cheese: An Overview. In Cheese: Chemistry, Physics and Microbiology (pp. 1–36). Boston, MA: Springer US. https://doi.org/10.1007/978-1-4615-2650-6_1

Gao, W., Smith, D. W., & Li, Y. (2007). Effects of Freezing on the Survival of Escherichia coli and Bacillus and Response to UV and Chlorine After Freezing. Water Environment Research. https://doi.org/10.2175/106143006X115426

Gkogka, E., Reij, M. W., Gorris, L. G. M., & Zwietering, M. H. (2013). The application of the Appropriate Level of Protection (ALOP) and Food Safety Objective (FSO) concepts in food safety management, using Listeria monocytogenes in deli meats as a case study. Food Control. https://doi.org/10.1016/j.foodcont.2012.04.020

Goldstein, E. J. C., Tyrrell, K. L., & Citron, D. M. (2015). Lactobacillus Species: Taxonomic Complexity and Controversial Susceptibilities. Clinical Infectious Diseases, 60(suppl_2), S98–S107. https://doi.org/10.1093/cid/civ072

Goudeau, D. M., Parker, C. T., Zhou, Y., Sela, S., Kroupitski, Y., & Brandl, M. T. (2013). The Salmonella transcriptome in lettuce and cilantro soft rot reveals a niche overlap with the animal host intestine. Applied and Environmental Microbiology. https://doi.org/10.1128/AEM.02290-12

Hamasaki, Y., Ayaki, M., Fuchu, H., Sugiyama, M., & Morita, H. (2003). Behavior of psychrotrophic lactic acid bacteria isolated from spoiling cooked meat products. Applied and Environmental Microbiology. https://doi.org/10.1128/AEM.69.6.3668

Hardy Diagnostics. (1996). XLD Agar: instructions of use. Retrieved December 18, 2018, from https://catalog.hardydiagnostics.com/cp_prod/content/hugo/xldagar.htm

Hill, C., Guarner, F., Reid, G., Gibson, G. R., Merenstein, D. J., Pot, B., … Sanders, M. E. (2014). Expert consensus document: The international scientific association for probiotics and prebiotics consensus statement on the scope and appropriate use of the term probiotic. Nature Reviews Gastroenterology and Hepatology. https://doi.org/10.1038/nrgastro.2014.66

Hong, S. P., Lee, E. J., Kim, Y. H., & Ahn, D. U. (2016). Effect of Fermentation Temperature on the Volatile Composition of Kimchi. Journal of Food Science, 81(11), C2623–C2629. https://doi.org/10.1111/1750-3841.13517

Ick, T. (2003). Factors Affecting Kimchi Fermentation. Japanese Journal of Lactic Acid Bacteria. Retrieved from https://www.jstage.jst.go.jp/article/jslab1997/14/2/14_56/_pdf

Illeghems, K., De Vuyst, L., & Weckx, S. (2015). Comparative genome analysis of the candidate functional starter culture strains Lactobacillus fermentum 222 and Lactobacillus plantarum 80 for controlled cocoa bean fermentation processes. BMC Genomics. https://doi.org/10.1186/s12864-015-1927-0

ILVO (Instituut voor Landbouw-, V. V. (2005). Mogelijke oplossingen voor het terugdringen van de ziekteverwekker Listeria monocytogenes in hoeveboter. Retrieved from https://www.ilvo.vlaanderen.be/Portals/46/Documents/listeria_artikel.pdf

Jeddi, M. Z., Yunesian, M., Gorji, M. E. haghi, Noori, N., Pourmand, M. R., & Khaniki, G. R. J. (2014). Microbial evaluation of fresh, minimally-processed vegetables and bagged sprouts from chain supermarkets. Journal of Health, Population and Nutrition. https://doi.org/10.1676/12-165.1

Jervis-Bardy, J., Leong, L. E. X., Marri, S., Smith, R. J., Choo, J. M., Smith-Vaughan, H. C., … Marsh, R. L. (2015). Deriving accurate microbiota profiles from human samples with low bacterial content through post-sequencing processing of Illumina MiSeq data. Microbiome. https://doi.org/10.1186/s40168-015-0083-8

Jimenez-Diaz, R., Rios-Sanchez, R. M., Desmazeaud, M., Ruiz-Barba, J. L., & Piard -, J. C. (1993). Plantaricins S and T, two new bacteriocins produced by Lactobacillus plantarum LPCO10 isolated from a green olive fermentation. Applied and Environmental Microbiology. https://doi.org/10.1029/2017EO071351

John, R. P., Nampoothiri, K. M., & Pandey, A. (2007). Fermentative production of lactic acid from biomass: an overview on process developments and future perspectives. Applied Microbiology and Biotechnology, 74(3), 524–534. https://doi.org/10.1007/s00253-006-0779-6

Joshi, V. K., Sharma, S., & Rana, N. S. (2006). Production, purification, stability and efficacy of bacteriocin from isolates of natural lactic acid fermentation of vegetables. Food Technology and Biotechnology. https://doi.org/10.5815/ijcnis.2014.10.05

Jung, J. Y., Lee, S. H., & Jeon, C. O. (2014). Kimchi microflora: History, current status, and perspectives for industrial kimchi production. Applied Microbiology and Biotechnology. https://doi.org/10.1007/s00253-014-5513-1

Kangas, S., Takkinen, J., Hakkinen, M., Nakari, U. M., Johansson, T., Henttonen, H., … Kuusi, M. (2008). Yersinia pseudotuberculosis O:1 Traced to raw carrots, Finland. Emerging Infectious Diseases. https://doi.org/10.3201/eid1412.080284

Karovičová, J., & Kohajdová, Z. (2003). Lactic acid fermented vegetable juices. Horticultura Science (Prague). https://doi.org/10.17221/3878-HORTSCI

Kataoka, A. I., Wang, H., Elliott, P. H., Whiting, R. C., & Hayman, M. M. (2017). Growth of Listeria monocytogenes in Thawed Frozen Foods. Journal of Food Protection, 80(3), 447–453. https://doi.org/10.4315/0362-028X.JFP-16-397R

Kok, C. R., & Hutkins, R. (2018). Yogurt and other fermented foods as sources of health-promoting bacteria. Nutrition Reviews, 76(Supplement_1), 4–15. https://doi.org/10.1093/nutrit/nuy056

Kotzé, A. D. (2003). THE SURVIVAL OF YEASTS AND PROBIOTICS AS ADJUNCT STARTERS IN CHEESE. Retrieved from http://scholar.ufs.ac.za:8080/xmlui/bitstream/handle/11660/5237/KotzeAD.pdf?sequence=1&isAllowed=y

Kozich, J. J., Westcott, S. L., Baxter, N. T., Highlander, S. K., & Schloss, P. D. (2013). Development of a Dual-Index Sequencing Strategy and Curation Pipeline for Analyzing Amplicon Sequence Data on the MiSeq Illumina Sequencing Platform. Applied and Environmental Microbiology. https://doi.org/10.1128/aem.01043-13

Lebeer, S. (2018). Role of citizen science in research on fermented foods. Retrieved January 15, 2019, from https://isappscience.org/citizen-role-fermented-foods-research/

Lebeer, S., Vanderleyden, J., & De Keersmaecker, S. C. J. (2008). Genes and molecules of lactobacilli supporting probiotic action. Microbiology and Molecular Biology Reviews : MMBR, 72(4), 728–764, Table of Contents. https://doi.org/10.1128/MMBR.00017-08

Lee, Y. C. (1991). Kimchi: The famous fermented vegetable product in Korea. Food Reviews International, 7(4), 399–415. https://doi.org/10.1080/87559129109540920

LFMFP. (2018). preparing a work stock.

Liang, H., Yin, L., Zhang, Y., Chang, C., & Zhang, W. (2018). Dynamics and diversity of a microbial community during the fermentation of industrialized Qingcai paocai, a traditional Chinese fermented vegetable food, as assessed by Illumina MiSeq sequencing, DGGE and qPCR assay. Annals of Microbiology. https://doi.org/10.1007/s13213-017-1321-z

Lu, H. J., Breidt, F., Perez-Diaz, I. M., & Osborne, J. A. (2011). Antimicrobial Effects of Weak Acids on the Survival of Escherichia coli O157:H7 under Anaerobic Conditions. Journal of Food Protection. https://doi.org/10.4315/0362-028X.JFP-10-404

Lucena-Padrós, H., Caballero-Guerrero, B., Maldonado-Barragán, A., & Ruiz-Barba, J. L. (2014). Microbial diversity and dynamics of Spanish-style green table-olive fermentations in large manufacturing companies through culture- dependent techniques. Food Microbiology. https://doi.org/10.1016/j.fm.2014.03.020

Maffei, D. F., Batalha, E. Y., Landgraf, M., Schaffner, D. W., & Franco, B. D. G. M. (2016). Microbiology of organic and conventionally grown fresh produce. Brazilian Journal of Microbiology. https://doi.org/10.1016/j.bjm.2016.10.006

Magnuson, J. K., & Lasure, L. L. (2004). Organic Acid Production by Filamentous Fungi. In Advances in Fungal Biotechnology for Industry, Agriculture, and Medicine. https://doi.org/10.1007/978-1-4419-8859-1_12

Manios, S. G., & Skandamis, P. N. (2015). Effect of frozen storage, different thawing methods and cooking processes on the survival of Salmonella spp. and Escherichia coli O157:H7 in commercially shaped beef patties. Meat Science. https://doi.org/10.1016/j.meatsci.2014.10.031

Mbandi, E., & Shelef, L. A. (2002). Enhanced antimicrobial effects of combination of lactate and diacetate on Listeria monocytogenes and Salmonella spp. in beef bologna. International Journal of Food Microbiology. https://doi.org/10.1016/S0168-1605(02)00026-0

Mellefont, L. A., McMeekin, T. A., & Ross, T. (2008). Effect of relative inoculum concentration on Listeria monocytogenes growth in co-culture. International Journal of Food Microbiology. https://doi.org/10.1016/j.ijfoodmicro.2007.10.010

Merck. (2018). Cycloheximide. Retrieved November 18, 2018, from https://www.sigmaaldrich.com/catalog/product/sigma/c7698?lang=en®ion=BE

Mullan, W. M. A. (2001). Microbiology of starter cultures. Retrieved October 22, 2018, from https://www.dairyscience.info/index.php/cheese-starters/49-cheese-starters.html

Müller, V. (2001). Bacterial Fermentation. Retrieved from www.els.net

Nelson, D. L. (David L., & Cox, M. M. (2013). Lehninger principles of biochemistry. W.H. Freeman and Company. Retrieved from https://books.google.be/books?id=n9e1NAEACAAJ&dq=lehninger+principles+of+biochemistry&hl=nl&sa=X&ved=0a hUKEwiSwOOo89rdAhUBKVAKHWiuDIgQ6AEIKzAA

Niksic, M., Niebuhr, S. E., Dickson, J. S., Mendonca, A. F., Koziczkowski, J. J., & Ellingson, J. L. (2005). Survival of Listeria monocytogenes and Escherichia coli O157:H7 during sauerkraut fermentation. J Food Prot. https://doi.org/10.4315/0362-028X-68.7.1367

Ogawa, M., Shimizu, K., Nomoto, K., Tanaka, R., Hamabata, T., Yamasaki, S., … Takeda, Y. (2001). Inhibition of in vitro growth of Shiga toxin-producing Escherichia coli O157:H7 by probiotic Lactobacillus strains due to production of lactic acid. International Journal of Food Microbiology. https://doi.org/10.1016/S0168-1605(01)00465-2

Oxoid Microbiology Products. (2001). X.L.D. Agar - Product Detail. Retrieved December 18, 2018, from http://www.oxoid.com/UK/blue/prod_detail/prod_detail.asp?pr=CM0469&c=UK&lang=EN

Özdemir, F., & Arslan, S. (2015). Genotypic and phenotypic virulence characteristics and antimicrobial resistance of Yersinia spp. isolated from meat and milk products. Journal of Food Science. https://doi.org/10.1111/1750-3841.12911

Ozsolak, F., & Milos, P. M. (2011). Single-molecule direct RNA sequencing without cDNA synthesis. Wiley Interdisciplinary Reviews: RNA. https://doi.org/10.1002/wrna.84

Peelman, N., Ragaert, P., De Meulenaer, B., Adons, D., Peeters, R., Cardon, L., … Devlieghere, F. (2013). Application of bioplastics for food packaging. Trends in Food Science and Technology. https://doi.org/10.1016/j.tifs.2013.06.003

Peng, Q., Jiang, S., Chen, J., Ma, C., Huo, D., Shao, Y., & Zhang, J. (2018). Unique microbial diversity and metabolic pathway features of fermented vegetables from Hainan, China. Frontiers in Microbiology. https://doi.org/10.3389/fmicb.2018.00399

Pérez-Diaz, I. M., Breidt, F., Buescher, R. W., Arroyo-Lopez, F. N., Jiménez-Diaz, R., Garrido Fernandez, A., … Johanningsmeier, S. D. (2013). Fermented and Acidified Vegetables. American Public Health Association, (chapter 51), 1–22.

Pires, S. M., Vieira, A. R., Hald, T., & Cole, D. (2014). Source Attribution of Human Salmonellosis: An Overview of Methods and Estimates. Foodborne Pathogens and Disease. https://doi.org/10.1089/fpd.2014.1744

Remel. (2011). MRS AGAR. Lenexa. Retrieved from https://assets.thermofisher.com/TFS-Assets/LSG/manuals/IFU1585.pdf

Riganakos, K. A., Karabagias, I. K., Gertzou, I., & Stahl, M. (2017). Comparison of UV-C and thermal treatments for the preservation of carrot juice. Innovative Food Science and Emerging Technologies. https://doi.org/10.1016/j.ifset.2017.06.015

Rimhanen-finne, R., Niskanen, T., Hallanvuo, S., Makary, P., Haukka, K., Pajunen, S., … Kuusi, M. (2009). Yersinia pseudotuberculosis causing a large outbreak associated with carrots in Finland, 2006. Epidemiology and Infection. https://doi.org/10.1017/S0950268807000155

RIVM (RijksInstituut voor Voeding en Milieu). (2018). NEVO-online. Retrieved October 30, 2018, from https://nevo- online.rivm.nl/ProductenDetailsGetabt.aspx?zoekstring=&tabid=1

Sage, J. R., & Ingham, S. C. (1998). Evaluating survival of Escherichia coli O157:H7 in frozen and thawed apple cider: potential use of a hydrophobic grid membrane filter-SD-39 agar method. J Food Prot. https://doi.org/10.4315/0362-028X- 61.4.490

Samapundo, S., Anthierens, T., Xhaferi, R., & Devlieghere, F. (2010). Development of a validated model to describe the individual and combined water activity depressing effects of water soluble salt, sugar and fat replacers. Journal of Food Engineering. https://doi.org/10.1016/j.jfoodeng.2009.08.022

Şanlier, N., Gökcen, B. B., & Sezgin, A. C. (2017). Health benefits of fermented foods. Critical Reviews in Food Science and Nutrition, 1–22. https://doi.org/10.1080/10408398.2017.1383355

Scatassa, M. L., Gaglio, R., Cardamone, C., Macaluso, G., Arcuri, L., Todaro, M., & Mancuso, I. (2017). Anti-Listeria activity of lactic acid bacteria in two traditional Sicilian cheeses. Italian Journal of Food Safety. https://doi.org/10.4081/ijfs.2017.6191

Schvartzman, M. S., Belessi, C., Butler, F., Skandamis, P. N., & Jordan, K. N. (2011). Effect of pH and Water Activity on the Growth Limits of Listeria monocytogenes in a Cheese Matrix at Two Contamination Levels. Journal of Food Protection. https://doi.org/10.4315/0362-028X.JFP-11-102

SciCom FASFC (Scientific Committee Federal Agency for the Safety of the Food Chain). (2017). Prioritering van microbiologische risico’s en richtlijnen voor het borgen van de microbiologische voedselveiligheid van rauwe en minimaal bewerkte plantaardige levensmiddelen in België. Retrieved from http://www.afsca.be/wetenschappelijkcomite/adviezen/2017/_documents/Advies11-2017_SciCom2013- 12_Groentenfruitmicrobiologie.pdf

Settanni, L., & Corsetti, A. (2007). Application of bacteriocins in vegetable food biopreservation. International Journal of Food Microbiology. https://doi.org/10.1016/j.ijfoodmicro.2007.09.001

Slonczewski, J. L., & Foster, J. W. (2014). Microbiology an evolving science. (B. Twitchell, Ed.) (third edit). W.W. Norton & company.

Soomro, A. H., Masud, T., & Anwaar, K. (2002). Role of Lactic Acid Bacteria (LAB) in Food Preservation and Human Health – A Review. Pakistan Journal of Nutrition. https://doi.org/10.3923/pjn.2002.20.24

Spyropoulou, K. E., Chorianopoulos, N. G., Skandamis, P. N., & Nychas, G. J. E. (2001). Survival of Escherichia coli O157:H7 during the fermentation of Spanish-style green table olives (conservolea variety) supplemented with different carbon sources. International Journal of Food Microbiology. https://doi.org/10.1016/S0168-1605(00)00510-9

Stamey, J. (2011). Are Your Clients YOPI’s? HealthyAndSafeFood.com. Retrieved from http://www.healthyandsafefood.com

Statista. (2018). production value of sauerkraut in Germany from 2004 to 2017. Retrieved December 21, 2018, from https://www.statista.com/statistics/521519/sauerkraut-production-value-germany/

Sun, Z., Harris, H. M. B., McCann, A., Guo, C., Argimón, S., Zhang, W., … O’Toole, P. W. (2015). Expanding the biotechnology potential of lactobacilli through comparative genomics of 213 strains and associated genera. Nature Communications, 6(1), 8322. https://doi.org/10.1038/ncomms9322

Tamang, J., Thapa, N., Tamang, B., Rai, A., & Chettri, R. (2015). Microorganisms in Fermented Foods and Beverages. In Health Benefits of Fermented Foods and Beverages. https://doi.org/10.1201/b18279-2

Tanguler, H., Utus, D., & Erten, H. (2014). Effect of black carrot size usage on the quality of shalgam (Şalgam): A traditional Turkish lactic acid fermented beverage. Indian Journal of Traditional Knowledge (Vol. 13). Retrieved from http://nopr.niscair.res.in/bitstream/123456789/29513/1/IJTK 13%284%29 647-653.pdf

Tenea, *, & Barrigas, G. N. (2018). The efficacy of bacteriocin-containing cell-free supernatant from Lactobacillus plantarum Cys5-4 to control pathogenic bacteria growth in artisanal beverages. International Food Research Journal (Vol. 25). Retrieved from http://www.ifrj.upm.edu.my/25 (05) 2018/(49).pdf

Terefe, N. S. (2016). Food Fermentation. Reference Module in Food Sciences. https://doi.org/10.1016/B978-0-08-100596- 5.03420-X

Tomé, E., Teixeira, P., & Gibbs, P. A. (2005). Anti-listerial inhibitory lactic acid bacteria isolated from commercial cold smoked salmon. Food Microbiology. https://doi.org/10.1016/j.fm.2005.05.004

Uyttendaele, mieke, Jacxsens, L., De Loy-Hendrickx, A., Devlieghere, F., & Debevere, J. (2010). MICROBIOLOGISCHE RICHTWAARDEN & wettelijke microbiologische criteria. Ghent. Retrieved from www.Food2Know.be

Uyttendaele, M., De Loy-Hendrickx, A., Vermeulen, A., Jaxsens, L., & Devlieghere, F. (2018). Microbiological guidelines support for interpretation of microbiological test results of foods. die Keure. Retrieved from https://www.diekeure.be/nl- be/professional/8831/microbiological-guidelines

Uyttendaele, M., & Li, D. (2018). HUMAN (OPPORTUNISTIC / FOOD BORNE) PATHOGENS. Ghent: Ghent University.

Van Damme, I., Berkvens, D., Vanantwerpen, G., Baré, J., Houf, K., Wauters, G., & De Zutter, L. (2015). Contamination of freshly slaughtered pig carcasses with enteropathogenic Yersinia spp.: Distribution, quantification and identification of risk factors. International Journal of Food Microbiology. https://doi.org/10.1016/j.ijfoodmicro.2015.03.016

Van Damme, Inge, Habib, I., & De Zutter, L. (2010). Yersinia enterocolitica in slaughter pig tonsils: Enumeration and detection by enrichment versus direct plating culture. Food Microbiology. https://doi.org/10.1016/j.fm.2009.09.011 van Ittersum, K. (2002). The Role of Region of Origin in Consumer Decision-Making and Choice. Retrieved from http://edepot.wur.nl/82514

Wainwright, M. (2003). An Alternative View of the Early History of Microbiology (pp. 333–355). https://doi.org/10.1016/S0065-2164(03)01013-X

Wemmenhove, E., van Valenberg, H. J. F., van Hooijdonk, A. C. M., Wells-Bennik, M. H. J., & Zwietering, M. H. (2017). Factors that inhibit growth of Listeria monocytogenes in nature-ripened Gouda cheese: A major role for undissociated lactic acid. Food Control. https://doi.org/10.1016/j.foodcont.2017.08.028

Wemmenhove, Ellen, van Valenberg, H. J. F., Zwietering, M. H., van Hooijdonk, T. C. M., & Wells-Bennik, M. H. J. (2016). Minimal inhibitory concentrations of undissociated lactic, acetic, citric and propionic acid for Listeria monocytogenes under conditions relevant to cheese. Food Microbiology. https://doi.org/10.1016/j.fm.2016.03.012

Werbrouck, H., Grijspeerdt, K., Botteldoorn, N., Van Pamel, E., Rijpens, N., Van Damme, J., … Van Coillie, E. (2006). Differential inlA and inlB expression and interaction with human intestinal and liver cells by Listeria monocytogenes strains of different origins. Applied and Environmental Microbiology. https://doi.org/10.1128/AEM.02164-05

WHO (World Health Organisation). (2015). WHO ESTIMATES OF THE GLOBAL BURDEN OF FOODBORNE DISEASES. Retrieved from http://apps.who.int/iris/bitstream/handle/10665/199350/9789241565165_eng.pdf?sequence=1

Wilkinson, L. (2011). ggplot2: Elegant Graphics for Data Analysis by WICKHAM, H. Biometrics. https://doi.org/10.1111/j.1541- 0420.2011.01616.x

Wood, B. J. B. (1997). Microbiology of Fermented Foods. Springer US.

Wuyts, S., Van Beeck, W., Oerlemans, E. F. M., Wittouck, S., Claes, I. J. J., De Boeck, I., … Lebeer, S. (2018). Carrot Juice Fermentations as Man-Made Microbial Ecosystems Dominated by Lactic Acid Bacteria. Applied and Environmental Microbiology, 84(12), e00134-18. https://doi.org/10.1128/AEM.00134-18

Wuyts, S., Wittouck, S., De Boeck, I., Allonsius, C. N., Pasolli, E., Segata, N., & Lebeer, S. (2017). Large-Scale Phylogenomics of the Lactobacillus casei Group Highlights Taxonomic Inconsistencies and Reveals Novel Clade-Associated Features. MSystems. https://doi.org/10.1128/mSystems.00061-17

Zabat, M., Sano, W., Wurster, J., Cabral, D., & Belenky, P. (2018). Microbial Community Analysis of Sauerkraut Fermentation Reveals a Stable and Rapidly Established Community. Foods. https://doi.org/10.3390/foods7050077

Addendum

Figure 18 Temperature profile during refrigeration, i.e. 7.478 °C ± 0.005 °C during the refrigeration storage

6,5

5,5 pH

4,5

3,5 day 0 day 1 day 3 day 15 day 30

spontaneous spontaneous + pathogens

Figure 19 pH evolution during 30-day fermentation (20°C)

Figure 20: Pathogen development (n=3) during fermentation (20°C) The green line indicates the limit of detection of the plate counts. Going to lower detection limits is possible using enrichment procedures. The standard error is indicated with error bars.

Figure 21 PCoA analysis containing all data of the 16S analysis (PCoA1 = 66%, PCoA2 = 25%)

Day 30 Day 31 Day 38

3 PH PH 2

0 CUJ + FCJ FCJL2 FCJ CUJ + FCJ + L2 CUJ + FCJL2

Figure 22: pH evolution during refrigeration (7.5°C), n = 3 The error bars are the standard errors of values measured in triplicate

Figure 23: Pathogen (inoculated at day 30) development during refrigeration (7.5°C) The green line indicates the limit of detection of the plate counts. Going to lower detection limits is possible using enrichment procedures. The standard error is indicated with error bars.

6,5 6 5,5 5 pH pH 4,5 4 3,5 3 0 1 3 6 9 days

spontaneous spontaneous + pathogens

Figure 24: pH evolution robustness test

Figure 25: pathogen development during the robustness test

Table 33 summary of counting and enrichment results Pathogens – Fermentation (20°C) - day 1 L. monocytogenes E. coli O157 S. Typhimurium Counting plates +++ +++ +++ Enrichment method +++ +++ +++ Samples

Pathogens – Fermentation (20°C) - day 3 L. monocytogenes E. coli O157 S. Typhimurium Counting plates --- +++ +++ Enrichment method ------samples

Pathogens – Fermentation (20°C) - day 15 L. monocytogenes E. coli O157 S. Typhimurium Counting plates ------Enrichment method ------samples

Pathogens – Fermentation (20°C) - day 30 L. monocytogenes E. coli O157 S. Typhimurium Counting plates ------Enrichment method ------samples

Pathogens – Refrigeration freshly inoculated at day 30 (7.5°C) - day 31 L. monocytogenes E. coli O157 S. Typhimurium Counting plates +++ --- +++ Enrichment method +++ +++ +++ samples

Pathogens – Refrigeration freshly inoculated at day 30 (7.5°C) - day 38 L. monocytogenes E. coli O157 S. Typhimurium Counting plates +-- --- +++ Enrichment method --- ++- ++- samples

Pathogens – Refrigeration inoculated at day 0 (8°C) - day 31 L. monocytogenes E. coli O157 S. Typhimurium Counting plates ------Enrichment method ------samples

Pathogens – Refrigeration inoculated at day 0 (8°C) - day 38 L. monocytogenes E. coli O157 S. Typhimurium Counting plates ++------Enrichment method ------samples