Versatility of microbial consortia and sensory properties induced by the composition of different milk and pea protein-based gels

Journal Pre-proof Versatility of microbial consortia and sensory properties induced by the composition of different milk and pea protein-based gels Salma Ben-Harb, Françoise Irlinger, Anne Saint-Eve, Maud Panouillé, Isabelle Souchon, Pascal Bonnarme PII:

S0023-6438(19)31062-X

DOI:

https://doi.org/10.1016/j.lwt.2019.108720

Reference:

YFSTL 108720

To appear in:

LWT - Food Science and Technology

Received Date: 14 June 2019 Revised Date:

27 September 2019

Accepted Date: 7 October 2019

Please cite this article as: Ben-Harb, S., Irlinger, Franç., Saint-Eve, A., Panouillé, M., Souchon, I., Bonnarme, P., Versatility of microbial consortia and sensory properties induced by the composition of different milk and pea protein-based gels, LWT - Food Science and Technology (2019), doi: https:// doi.org/10.1016/j.lwt.2019.108720. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier Ltd.

1 2

Versatility of microbial consortia and sensory properties induced by the composition of different milk and pea protein-based gels

3 4 5

Salma Ben-Harb, Françoise Irlinger, Anne Saint-Eve, Maud Panouillé, Isabelle Souchon, Pascal Bonnarme* UMR GMPA, AgroParisTech, INRA, Université Paris-Saclay, 78850 Thiverval-Grignon, France

6 7

*Corresponding author: [email protected]

8

Abstract

9

The use of pea protein is limited in food by the persistence of off-flavours. Fermentation by

10

microbial consortia could be a way to circumvent this problem. The aim of our study was to

11

investigate the adaptation potential and metabolic features of consortia as a function of the matrix

12

composition (pea and/or milk proteins). Three designed consortia were investigated for their ability

13

(i) to grow in gels containing 100% pea proteins, 50% pea-50% milk proteins and 100% milk

14

proteins; (ii) to reduce off-flavours and increase aroma perception. VEGAN is suited for the pea gel

15

(PG): it better colonises than in the mixed (PMG) and milk gels (MG), except for lactic acid bacteria

16

that grow better with lactose. In PG, aromatic notes are mainly smoked/onion/garlic, while those in

17

MG and PMG are perceived as dairy/cheese. MEGAN is adapted to MG and PG: it did not mask

18

“green notes” in PMG and generates cheesy and fruity notes in MG. The ExEco consortium is not

19

suited in PG: it does not mask “green notes”. Cheese, yogurt and fruits (apple, apricot) notes were

20

perceived in PMG and MG. Our study shows that fermentation could be applied to develop pea-

21

based products with diversified sensory characteristics.

22

Keywords: pea proteins; fermentation; microbial communities; olfactometry; volatilome.

23

1

24

1. Introduction

25

Foodborne microorganisms play a major role in fermented foods with respect to

26

physicochemical properties, e.g., aroma, acidification, texture, sensory properties (Ravyts, De Vuyst,

27

& Leroy, 2012). It is generally estimated that about one-third of our diet is composed of fermented

28

foods (Campbell-Platt, 1994). Because of their considerable importance and the development of

29

molecular techniques, they have been the subject of extensive investigation to identify microbial

30

communities and to understand the mechanisms of adaptation, interaction and assemblage within

31

various matrices, revealing new facets and potential uses of microbial consortia (Monnet, Landaud,

32

Bonnarme, & Swennen, 2015; Irlinger, Layec, Hélinck, & Dugat-Bony, 2015; El Sheikha & Hu,

33

2019; Ben Harb et al. 2019). Fermented foods are influenced by a range of biotic and abiotic factors.

34

Thus, milk used for fermented dairy products (e.g., yogurt, kefir, cheese) has a heterogeneous

35

physicochemical composition (lactose, casein, fat) since the matrix changes its characteristics during

36

the fermentation process. Such changes result from the growth of multiple microbial populations,

37

including lactic acid bacteria (LAB), yeasts, filamentous fungi and cheese-ripening acid-sensitive

38

bacteria. The microorganisms have to adapt themselves to the presence of other microorganisms with

39

which they may have positive or negative interactions (Mounier, Monnet, Jacques, Antoinette, &

40

Irlinger, 2009; Monnet, Landaud, Bonnarme, & Swennen, 2015).

41

Whereas our knowledge of species involved in the key functions of dairy fermentation

42

processes is extensive, we know little about the key functions, steps, and microbial players involved

43

in the fermentation processes of new legume based-foods. Pea protein isolate, composed of major

44

pea storage proteins including legumin, vicilin and convicilin, is considered to be a good alternative

45

for the formulation of emerging healthy and sustainable food products (Sirtori et al., 2012). In a

46

similar context, a recent study has highlighted ways in which pea proteins could be used in new food

47

products by proposing different gelation techniques to create gels made from pea and milk proteins

48

(Ben-Harb et al., 2017). Lactic fermentation by Lactobacillus plantarum and Pediococcus 2

49

pentosaceus strains improved pea protein extract aroma by decreasing the n-hexanal content and

50

reducing or masking "green-note" off-flavours (Schindler et al., 2012). Since pea proteins may

51

provide a wide range of nutrient sources for a variety of proteolytic microorganisms, pea protein

52

suspensions were or were not mixed with milk and subjected to fermentation with mixed

53

allochthonous microbial consortia consisting of LAB, Proteobacteria, Actinobacteria, yeasts and

54

moulds. Stable and reproducible fermentation induced by these consortia significantly affected the

55

content of aroma compounds, including those responsible for green off-flavours, with consequences

56

on consumer perception (Ben-Harb et al., 2019).

57

In this work, we chose to study three previously and rationally designed microbial consortia

58

(Dugat-Bony et al., 2015; Ben-Harb et al., 2019) containing five to nine microbial dairy strains in

59

three well-characterized and controlled environments: (i) milk (ExEco consortium); (ii) pea protein

60

suspension (VEGAN consortium); and (iii) a mix of pea proteins and milk suspension (MEGAN

61

consortium) for their impact on the aromatic properties of fermented products. In previous

62

publications, the microbial ExEco consortium was shown to be able to reproduce the complex

63

metabolic pattern of cheese maturation (Bonaïti, Irlinger, Spinnler, & Engel, 2005; Mounier et al.,

64

2008; Dugat-Bony et al., 2015). The main objective of our study was to investigate the adaptation

65

potential and metabolic features of the microbial consortia as a function of the matrix composition

66

(pea and/or milk proteins). The growth capacities of each strain and the impact of fermentation on

67

the sensory quality of the fermented matrices (in particular, off-flavour perception and production of

68

aroma compounds) were evaluated.

69

2. Materials and methods

70

2.1. Ingredients and raw materials

71

Pea protein isolate (PPI) (NUTRALYS® S85F) was obtained from Roquette Frères (Lestrem,

72

France). Skim milk powder (SMP) was purchased from Lactalis (Bourgbarré, France) and rapeseed 3

73

oil (Fleur de Colza, Lesieur, France) was purchased from a local supermarket. Glucono delta-lactone

74

(GDL) was used for coagulation (Sigma Aldrich, Steinheim, Germany).

75

2.2. Microbial strains and growth conditions

76

The 15 strains composing the three model microbial communities (MEGAN, VEGAN and

77

ExEco) are listed in Table S1. All these strains were originally isolated from cheeses. Lactic acid

78

bacterium S3 was grown for 24 h at 30°C under static conditions in M17 lactose (0.5%) broth

79

(Biokar Diagnostics, Beauvais, France). All other bacteria were grown under aerobic conditions:

80

rotary shaker at 150 rpm at 25°C for 48 h in 50-mL conical flasks containing 10 mL of brain heart

81

infusion broth (Biokar Diagnostics). Yeasts were grown under the same conditions, except that

82

potato dextrose broth (Biokar Diagnostics) was used as the growth medium.

83

2.3. Preparation of gels

84

Three types of emulsions were prepared with 10% proteins, containing (i) 100% pea proteins

85

(pea emulsion); (ii) a mixture of pea proteins and milk powder (50/50) (mixed emulsion); and (iii)

86

100% milk powder (milk emulsion). Pea and mixed emulsions were prepared using a previously

87

described method (Ben-Harb et al., 2019). Milk emulsions were prepared like other types of

88

emulsions with minor modifications: a heating step (95°C/30 min) was used for these types of

89

emulsions. Three types of gels were formed by chemical coagulation using GDL (glucono delta-

90

lactone). To obtain gels, suspensions were mixed with GDL (0.1 , 0.5 and 1.5% w/v for pea, mixed

91

and milk emulsions, respectively) and with each defined community (MEGAN, VEGAN and

92

ExEco), with a cell density equivalent to 6.0 log CFU/g for each bacterial strain and 4.0 log CFU/g

93

for each fungus. The mixtures obtained were then incubated at 25°C for 24 h for coagulation and at

94

16°C for 2 and 6 days for fermentation (Fig. 1). Gels were prepared in triplicate.

4

95

2.4. Microbial analyses, cell enumeration and pH measurements

96

Samples were taken on days 0, 3 and 7. Day 0 corresponds to the non-fermented gel control

97

before inoculation. Three independent gels, i.e., the three biological replicates, were analysed at each

98

time point. The gels were crushed and homogenised with sterile knives and forks. Serial dilutions

99

were prepared in 0.9% NaCl from one gram of gel and plated in triplicate on agar, as previously

100

described (Ben-Harb et al., 2019).

101

Gel fermentations were carried out in triplicate from independent batches, and data for pH

102

and growth values were averaged. Distribution and growth were expressed as log N/N0 (N: growth

103

of strains at three (D3) and seven (D7) days of fermentation; N0: initial cell density). The pH values

104

were the arithmetic means of three measurements made with a Blue Line 27 surface electrode

105

(Schott).

106

2.5. Sensory evaluation using the Check-All-That-Apply (CATA) method

107

The odour of the samples was evaluated using the CATA (Check-All-That-Apply) method.

108

This methodology consists in selecting all the attributes that we consider appropriate to characterise

109

the perception of each sample from a list of sensory attributes (Valentin, Chollet, Lelièvre, & Abdi,

110

2012). This method was performed as previously described (Ben Harb et al., 2019). Sensory analysis

111

was performed to characterise the sensory properties of fermented gels (from control and inoculated

112

fermentations) after seven days of fermentation. All sensory analyses were carried out in an air-

113

conditioned room (20°C) in individual booths under white light. The odour of samples was evaluated

114

using the method of Ben-Harb et al. (2019) with minor modifications. Twenty-five panellists were

115

asked to select the attributes that best described each sample from a list of 40 attributes (Table S2).

116

The products were presented with a Latin square order. Thirteen products were evaluated per session.

117

All gels were diluted (1/10) with Milli-Q water (Merck Millipore, Merck KGaA, Germany) at 4°C

118

and then stirred for 30 s with a Stomacher apparatus. They were presented in glass bottles labelled

5

119

with randomly selected three-digit numbers. All samples were analysed in triplicate (three

120

independent batches).

121

2.6. Extraction and identification of volatile flavour components from fermented gels

122

Volatile compounds of the gels were extracted and identified using the GC-MS method, as

123

previously described by Ben-Harb et al. (2019). They were extracted from the samples using a water-

124

jacketed purge and trap concentrator (Tekmar-Dohrman 3100, Tekmar, USA) at 40°C (purge:

125

40°C/15 min; desorb: 225°C/2 min) coupled to a gas chromatograph (Agilent Technologies 3800,

126

USA) and a mass spectrometer detector (MSD 5975C, Agilent Technologies, USA). The apparatus

127

was equipped with a DB-5 polar capillary column (30 m × 0.25 mm; film thickness: 0.25 µm;

128

Agilent 122-5532, USA). Compounds were identified based on their linear retention indexes and

129

mass spectra using a standard MS database (Wiley Registry of Mass Spectral Data). Volatile

130

compounds were quantified based on total peak area (TPA) of each volatile compound. All samples

131

were analysed in biological duplicates. The results are average values.

132

2.7. Statistical analysis

133

Data analysis was performed using the statistical software programmes XLSTAT (Addinsoft,

134

Paris, France, 2009) and Fizz Data Treatment (Biosystemes® 1999). Data sets were analysed with

135

analysis of variance (Anova) or Cochran tests to compare the growth of strains in each microbial

136

community in the different gels (pea, mixed and milk). Cochran's Q test (Cochran, 1950) was

137

performed to determine whether the proportions of selection by panellists for individual terms of the

138

CATA question differed as a function of the products. If there was a significant difference among the

139

variables, post hoc multiple pairwise comparisons were performed (Newman-Keuls test). In addition,

140

correspondence analysis (CA) based on chi-square distance was used to visualise associations

141

between the CATA terms and the products. Significant terms determined by Cochran's Q test were

142

used for correspondence analysis. A statistically significant difference was said to exist when P <

6

143

0.05.

144 145

A multiple factor analysis (MFA) was also used to study the relationships between sensory

146

features, growth of the microbial starters and composition of volatile compounds in the samples.

147

Each strain was associated with a microbial starter (for example, Lactobacillus rhamnosus is present

148

in VEGAN and MEGAN, and Lactobacillus rhamnosus-MEG and Lactobacillus rhamnosus-VEG

149

can be found in the map presentation). The RV coefficient (multivariate generalisation of the squared

150

Pearson correlation coefficient) was calculated between the first two axes of the partial configuration

151

of samples in order to analyse similarity between sample configurations). A hierarchical ascendant

152

cluster analysis (HCA) was carried out to group trials according to their similarity, measured by the

153

Pearson correlation, whereas cluster aggregation was based on the average linkage method.

154

3. Results

155

In the present study, three protein-based gels (pea gels (PG), pea and milk gels (PMG) and

156

milk gels (MG)) were fermented with three microbial communities (ExEco, VEGAN and MEGAN)

157

(Fig. S1). We first investigated the effect that fermentation - and the choice of the microbial

158

community - could have on sensory properties, our target being the degradation and/or masking of

159

green notes by fermentation. The volatile compounds produced in the fermented products were then

160

investigated, more specifically those possibly impacting off-flavour perception and/or flavouring

161

properties. The adaptation of each microbial community was then investigated through the

162

measurement of microbial growth in the gels.

163

3.1. Sensory properties after fermentation

164

The sensory profiles of the gels fermented by the different microbial communities MEGAN,

165

VEGAN and ExEco after seven days and those of the non-inoculated control gels are shown in Fig.

166

2. Correlation maps of Correspondence Factorial Analysis (CFA) with the significant sensory 7

167

descriptors were used and identified by the Cochran test (P < 0.05). A hierarchical ascendant cluster

168

analysis (HCA) was carried out to group the gels according to their similarity.

169

3.1.1. Sensory properties of gels after fermentation by the VEGAN community

170

Sensory properties of the different gels fermented by the VEGAN community and control

171

gels (non-inoculated) are shown in a sensory map in Fig. 2A. The first two principal axes (F1 and

172

F2) explained 77.9% of the variance. Gels were well discriminated by the panel according to their

173

olfactory perception. The F1 axis explains differences between gels according to the presence or

174

absence of pea proteins in the gels. It separated fermented pea, mixed control and pea control from

175

milk control and fermented milk fairly well, with the exception of mixed gel fermented by VEGAN,

176

perceived to be closer to milk gel than to pea gel. The odour notes such as cut grass and pea were

177

loaded on the negative side of axis 1 and described pea-based controls and pea fermented by

178

VEGAN. On the positive side of axis 1, milk-based gels fermented by VEGAN were characterised

179

by yogurt, fresh milk, cream and honey odours. MG, MPG and PG were fermented by the same

180

community (VEGAN), so that the proximity of MPG and MG indicates close similarity in terms of

181

sensory properties. Conversely, the PG is clearly separated from them as well as from the control

182

gels. This delineation is along axis 2 on which garlic/onion, pea and coffee are loaded on the positive

183

dimension, while dried fruit and cut grass odours (pea control), are loaded on the negative

184

dimension.

185

3.1.2. Sensory properties of gels after fermentation by the MEGAN community

186

The sensory map of MEGAN gels (Fig. 2B) explained 82.2% of the information in axes 1

187

and 2. Closely related to each other along axis F1, the fermented pea gel was characterised by odour

188

attributes similar to the non-inoculated control pea and mixed gels, confirming the results obtained

189

by HCA and showing the highest scores for pea, woody, dried fruit and cut grass notes. Conversely,

190

the other gels fermented by MEGAN (milk and mixed gels) were both significantly clustered with

8

191

distinctive aroma profiles. The fermented milk gel, loaded positively on the F1 and F2 axes, was

192

characterised by sensory attributes close to fermented fruit and cheese notes after seven days of

193

fermentation. In addition, the mixed fermented gel, loaded negatively on the F2 axis, was

194

characterised by notes of sour, melted butter and yogurt odour perception.

195

3.1.3. Sensory properties of gels after fermentation by the ExEco community

196

The sensory map of ExEco gels (Fig. 2C) explained 76.2% of the information on the first two

197

dimensions. The gels fermented by the ExEco community were clustered into three groups

198

characterised by different aroma profiles. The first was composed of non-inoculated pea-based

199

controls and fermented pea gel, characterised by pea, cut grass and roasted/grilled notes. The second

200

group included both fermented mixed and milk gels, characterised by fruity and cheesy notes. The

201

last group in the non-inoculated milk gel control was characterised by fresh milk notes.

202

3.2. Production of volatile compounds after fermentation

203

Identification, quantification and comparison of volatile compounds produced before and

204

after fermentation (for seven days) in PG, PMG and MG, fermented by the three microbial

205

communities, were carried out. These compounds were subsequently classified into the families of

206

alcohols, aldehydes, aromatic hydrocarbons, alkanes, alkenes, ketones, sulphur compounds and

207

esters. Aldehydes were classified in two groups to differentiate aldehydes potentially responsible for

208

green notes ("green aldehydes") and other aldehydes ("other aldehydes") responsible for malted,

209

grilled and roasted notes. Figure 3 lists the overall distribution of these compounds, their area values

210

depending on the type of gel and the microbial community used, showing significant differences. As

211

shown in Fig. 3A, the total peak areas (TPA) of volatile compounds detected in control pea, mixed

212

and milk gels were 1.36 x 108, 1.67 x 108 and 8.27 x 108, respectively. All non-fermented gel

213

samples were mainly associated with aliphatic green aldehyde compounds (hexanal, nonanal and

214

octanal), arising from lipid oxidation (Table S3).

9

215

3.2.1. Production of volatile compounds after fermentation with the VEGAN community

216

Regarding the main volatile compounds produced in the three different gels by the VEGAN

217

community, the highest abundance (given as TPA) of volatile compounds was found in MG and

218

PMG (Fig. 3B), both reaching a value of 2.5 x 108, while the fermented PG remained at a level

219

comparable to the control PG. Nevertheless, some changes in the distribution of volatile compounds

220

detected were observed for the fermented pea gel (Table S3). After seven days of fermentation, a

221

decrease in the TPA of green aldehydes was observed, whereas the quantities of alcohols (2-methyl-

222

1-butanol) and other aldehydes (3-methyl butanal, 2-methyl butanal and 2-methylpropanal)

223

increased. In the gels containing milk (PMG and MG), 27 and 38 volatile compounds were detected,

224

respectively, both characterised after fermentation by a slight decrease of green aldehyde, an increase

225

of other aldehydes and considerable production of ketones, including butane-2,3-dione and 6-

226

methylhept-5-en-2-one.

227

3.2.2. Production of volatile compounds after fermentation with the MEGAN community

228

When using the MEGAN community for the fermentation of the three gels, the highest

229

abundance of volatile compounds was found in MG (Fig. 3C), reaching a value of 8.0 x 108, while

230

the PG and PMG reached 1.8 x 108 and 2.4 x 108, respectively. Significant changes in the

231

composition of volatile compounds were observed in PG, PMG and MG when fermented by the

232

MEGAN community, with esters being abundantly produced in MG (Table S3). In PG, a significant

233

decrease in the TPA of green aldehydes was found, while an increase of other aldehydes such as

234

(2E,4E)-hepta-2,4-dienal, 3-methybutanal and 2-methylpropanal was observed compared to the non-

235

fermented control, as well as for aromatic hydrocarbons (2-pentylfuran and styrene), alkanes

236

(undecane) and ketones (3-methylbutan-2-one). In the mixed gel, fermentation led to a significant

237

production of esters (ethyl acetate), ketones (butane-2,3-dione) and alcohols (2-methylpropan-1-ol,

238

and 3-methylbutan-1-ol). In fermented milk gel, substantial amounts of esters (mainly ethyl acetate),

10

239

acetaldehyde and alcohols (ethanol, 2-methylpropan-1-ol, 3-methylbutan-1-ol), together with ketones

240

(heptan-2-one, nonan-3-one, and butane-2,3-dione) and aromatic hydrocarbons (benzene, toluene,

241

benzaldehyde, styrene and 2-ethylfuran) were detected.

242

3.2.3. Production of volatile compounds after fermentation with the ExEco community

243

Concerning the main volatile compounds produced by the ExEco community in three

244

different gels, their abundance remained relatively stable in the set of gels after 7 days of

245

fermentation (Fig. 3D). Unlike the VEGAN and MEGAN communities, which were able to produce

246

specific ketones depending on the type of fermented gel, the ExEco community did not produce this

247

family of volatile compounds, favouring the production of alcohols in the pea gel, aldehydes such as

248

acetaldehyde in the mixed gel, alkanes such as pentane in the milk gel, and esters in the gels with

249

milk.

250

3.3. Microbial growth after fermentation

251

The populations reached in the different gels after three and seven days of fermentation are

252

shown in Table 1. The initial pH values of PG, PMG and MG were 6.71, 6.29 and 5.68, respectively.

253

After fermentation, the pH of PG gels remained fairly stable, whereas PMG and MG gels were

254

characterised by acidic pH values. The initial cell density in the inoculated gels was 6.0 log CFU/g

255

for each bacterium and 4.0 log CFU/g for each fungus.

256

3.3.1. Microbial growth of the VEGAN community

257

Fermentation by the VEGAN microbial community showed that the pH decreased or

258

remained stable, depending on the gels tested (with or without milk) (Table 1A). In general, the

259

acidic gels containing milk (MG and PMG) reached the highest microbial numbers, except for

260

Geotrichum candidum and Hafnia alvei, which grew quite efficiently on PG compared to MG and

261

PMG, and for Yarrowia lipolytica that essentially grew on PG, since cell counts in this condition

262

were 2.7 log higher than in both gels containing milk. Brevibacterium casei and Brevibacterium 11

263

antiquum, initially inoculated in all of the gels, were not detected or present to only a slight degree

264

after three and seven days of fermentation.

265

3.3.2. Microbial growth of the MEGAN community

266

Fermentation by the MEGAN microbial community essentially exhibited the same pH

267

profiles as for the VEGAN community (Table 1B). Regardless of the type of substrate used, growth

268

of the MEGAN community was strong and stable in all gels, reaching cell counts of about 109 CFU/g

269

for LAB and 107-108 CFU/g for yeasts after seven days of fermentation. It is noteworthy that, in

270

milk, Kluyveromyces lactis showed the highest growth in MG (4.1 to 4.4 log CFU/g) compared to

271

PG (1.5 to 2.8 log CFU/g) and PMG (2.2 to 2.8 log CFU/g). For bacteria such as LAB that are likely

272

to grow rapidly on both media containing lactose (MG and PMG), it is likely that the maximum

273

population will be reached before 3 days and that differences between the values observed for two

274

strains on the same medium can result either from an actual higher growth or from a decrease in the

275

cultivable population during incubation. The results also suggest the presence of interactions. Thus,

276

the behaviour of the Lactobacillus rhamnosus strain common to the VEGAN and MEGAN consortia

277

significantly differs depending on the consortium in which it is found.

278

3.3.3. Microbial growth of the ExEco community

279

Fermentation by the ExEco microbial community, initially designed for the manufacture of

280

Livarot-type cheese (Bonaïti et al., 2005; Dugat-Bony et al., 2015), followed the same pH profiles as

281

for VEGAN or MEGAN consortia, where a strong decrease in pH was observed in MG and PMG

282

(Table 1C). In the most acidic gels, namely MG and PMG, the highest microbial cell counts were

283

reached, except for Lactococcus lactis, and Hafnia alvei, which developed maximally in PG.

284

Glutamicibacter arilaitensis and Staphylococcus equorum, known to be acid-sensitive bacteria,

285

developed in PG (pH 6.5-7), with cell counts 3.2 and 1.3 log higher than in PMG and MG (pH 4.5-

286

4.7). Corynebacterium casei and Brevibacterium aurantiacum were not detected or were present

12

287

only to a slight extent in every type of gel.

288

4. Discussion

289

This work is a comparative study of different gels (PG, MPG and MG) fermented by different

290

microbial starters (VEGAN, MEGAN and ExEco) involving sensory evaluation together with

291

identification of aromatic profile and microbial growth. A correlation map from Multiple Factorial

292

Analysis (MFA) with the sensory attributes, microbial communities and aroma compounds produced

293

after seven days of fermentation is shown in Fig. 4.

294

PG fermented by the VEGAN consortium was characterised by notes like "dried fruits",

295

"roasted/grilled", "coffee", and “smoked". Geotrichum candidum, Hafnia alvei and Yarrowia

296

lipolytica were the dominant species after seven days of fermentation. These strongly proteolytic and

297

lipolytic microorganisms may play an important role in the production of aroma compounds (Suzzi

298

et al., 2001; Boutrou & Guéguen, 2005; Dugat-Bony et al., 2015). Among the aroma compounds

299

produced by the above three species, volatile sulphur compounds have been reported (Arfi, Spinnler,

300

Tâche, & Bonnarme, 2002) and were detected - namely dimethyl disulphide (DMDS) and dimethyl

301

sulphide (DMS). Fermented MG and PMG were characterised by different aromatic profile such as

302

“fresh milk”, “cooked milk”, “honey”, “fresh cream” and “yogurt”. Contrary to PG, milk-based gels

303

were characterised by the considerable growth of LAB (Lactobacillus rhamnosus and Leuconostoc

304

lactis). On PG, fermentation was without sugars, which promoted alkaline fermentation. In contrast,

305

lactose fermentation by LAB allowed acid fermentation in PMG and MG. LAB may also contribute

306

to the aroma and flavour of fermented products. This is in agreement with the study of Yousseef,

307

Lafarge, Valentin, Lubbers, & Husson (2016) in which fermentation by seven species of LAB of

308

mixes with different milk-to-pea proteins ratios were compared. They showed that fermented

309

products tended to have higher intensities for positive descriptors such as creamy, dairy and sweet,

310

and lower intensities for negative descriptors such as plant, soil and vinegar.

13

311

The correlation map with the MEGAN consortium shows that it behaves similarly in all types

312

of gels, with a preference in the presence of milk (PMG and MG). MEG-7-MILK and MEG-7-

313

MIXED samples were characterised by fermented and fruity notes like “fermented fruit”, “solvent”,

314

“ethanol” and “pungent” notes. This is to be correlated with the production of esters (mainly ethyl

315

acetate) - most probably by K. lactis - which is well known to produce such aroma compounds (Arfi

316

et al., 2002; Liu, Holland, & Crow, 2004). Pea-MEG gels, loaded positively on the F1 and F2 axes,

317

were characterised by green notes like “woody” and “cut grass”.

318

Considering the ExEco consortium, mixed and milk gels exhibited fermented fruit and

319

pineapple notes that are typical of esters, e.g., ethyl acetate, corresponding to the presence of the

320

yeast K. lactis, which did not grow in pea gel. The occurrence of cheese, cheese/garlic and rind notes

321

is a signature of volatile sulphur compounds (e.g., DMS, DMDS), the latter being produced by

322

Hafnia alvei and/or Geotrichum candidum.

323

Our data clearly show an increase in TPA of the aromatic compounds in the fermented gels,

324

compared to the non-fermented control gels (Fig. 3). Interestingly, the total amounts of several

325

aldehydes (e.g., hexanal, heptanal, nonanal, octanal, (E)-2-ethylbut-2-enal) potentially responsible

326

for off-flavours could be significantly reduced by fermentation, especially using VEGAN or

327

MEGAN consortia (Table S4). This degradation of such aldehydes by an as yet unknown

328

mechanism would require further investigation. Hexanal, heptanal, nonanal and octanal are important

329

aroma compounds that contribute the undesirable “green notes” of vegetables and that belong to the

330

family of aliphatic aldehydes. They could be derived from either enzymatic or auto-oxidation of fatty

331

acids, mainly linoleic and linolenic acids (Murat, Bard, Dhalleine, & Cayot, 2013). Hexanal and

332

heptanal are also important aroma compounds that contribute to good flavour in fermented milk

333

products (Dan et al., 2017). (2E,4E)-hepta-2,4-dienal was detected in pea MEGAN gels. It was also

334

detected in pea protein fermented by LAB (Lactobacillus plantarum L1047 and Pediococcus

335

pentosaceus P113) (Schindler et al., 2012) and in fermented milk by Lactobacillus delbrueckii subsp.

14

336

bulgaricus and Streptococcus thermophilus (Dan et al., 2017).

337

Amino acids released through proteolysis serve as a major source of energy (and alternatively

338

as precursors of flavour compounds) for the growth of bacteria and fungi via oxidative deamination

339

and/or transamination reactions (Molimard & Spinnler, 1996). Such enzymatic reactions

340

(aminotransferase and/or carboxylase) are found notably in G. candidum. Consequently, a number of

341

the volatile compounds identified in fermented gels, including 3-methylbutanal and benzaldehyde,

342

can be recognised as products of amino acid catabolism. The conversion of leucine, isoleucine and

343

valine takes place via transamination of the amino acids to the corresponding α-keto acids and,

344

subsequently, via a chemical or enzymatic decarboxylation step, to 3-methylbutanal, 2-

345

methylbutanal and 2-methylpropanal, respectively (Yvon, Berthelot, & Gripon, 1998; Ayad,

346

Verheul, Engels, Wouters, & Smit, 2001). These volatile molecules are in the other aldehyde class.

347

They increased after fermentation in all gels, with malt, chocolate and roasted coffee notes (Smit,

348

Smit, & Engels, 2005). Those aroma compounds are produced mainly by LAB and yeasts (e.g.,

349

Geotricum candidum, Yarrowia lipolytica), through the Ehrlich degradation pathway (Mollimard &

350

Spinnler, 1996). In our study, a high concentration of 3-methylbutanal was present in the volatile

351

fraction of all gels fermented by all microbial communities.

352

MG fermented by the ExEco consortium did not show the production of volatile compounds

353

identified as ketones, unlike MG fermented by MEGAN and VEGAN, which were characterised by

354

the production of butane-2,3-dione (= diacetyl) (for mixed and milk gels), pentan-2-one, 2-butan-2-

355

one and 3-methylbutan-2-one (for pea MEGAN gels), and nonan-3-one (for milk MEGAN gel).

356

Diacetyl is a product of citrate metabolism that positively contributes to the perception of buttery and

357

creamy flavours in butter and some fermented milk products (Cogan & Hill, 1993). It is an important

358

metabolite for LAB used in the dairy industry and results from pyruvate excess in the cell where

359

pyruvate is converted to diacetyl via α-acetolactate. Other molecules can be derived from lipolytic

360

activity, and compounds such as pentan-2-one and butan-2-one have been described as having a wine

15

361

or acetone-like odour, and a sweet apricot-like odour, respectively (Burdock, 2009).

362

Styrene was detected primarily in MEGAN and ExEco gels. It may have originated from

363

packaging, normal metabolism of the raw product or microbial metabolism. It can be formed, for

364

example, during the fermentation of grapes (Steele et al., 1994), by microorganisms during the

365

ripening or storage of mould-ripened cheeses (Spinnler, Grosjean, & Bouvier, 1992), or by the

366

decomposition of chemically-related food additives by microorganisms. It has been proven to be the

367

result of starvation of certain fungi (Adda, Dekimpe, Vassal, & Spinnler, 1989). Styrene appears to

368

be synthesized from phenylalanine by phenylalanine ammonia lyase activity, followed by a

369

decarboxylation catalysed by a cinnamic acid decarboxylase (Pagot, Belin, Husson & Spinnler,

370

2007). A number of aromatic hydrocarbons were found, including toluene, benzene and 2-

371

pentylfuran. They were probably the result of the oxidation of unsaturated fatty acids and may have

372

affected the characteristic aroma and taste of the products. These molecules were detected in some

373

pea cultivars (Azarnia et al., 2011). The possible influence of such compounds on the flavour of

374

fermented gels remains unclear. Future research will target these aromatic hydrocarbons by

375

examining samples to obtain a better understanding of their possible roles in fermented pea gels.

376

5. Conclusion

377

Our results suggest that metabolic activity and growth of the microbial consortia are

378

essentially driven by the composition of the gel. Hence, a mosaic of aromatic notes was perceived,

379

depending on the type of gel and the microbial consortium used. The microbial composition of each

380

consortium varies during the fermentation process, which may lead to potential interactions of

381

various natures (e.g., biotic, metabolic) within the food matrix. The nature of such interactions

382

together with adaptive microbial metabolism are currently under investigation by combining sensory,

383

biochemical, physiological and meta-omic analyses, to give a full view of the functioning of the

384

microbial consortia in a pea gel. One striking feature of our work is that microbial consortia can

16

385

efficiently reduce off-notes, which we found to be in good agreement with a decrease in the

386

concentrations of "green aldehydes" and an increase in volatile flavouring compounds in gels

387

enriched with pea proteins. This work shows that fermentation could be successfully applied to

388

develop plant-based protein food products with diversified sensory characteristics. Our approach,

389

including matrix and microbial consortia design, could be applied to other sustainable sources of

390

food protein to generate a new generation and more diversified fermented products with target

391

functionalities (e.g., improved taste and digestibility).

392 393

Conflict of interest

394

The authors declare no conflict of interest.

395

Acknowledgments

396

The authors would like to acknowledge the Carnot Qualiment Institute for its financial

397

support. The authors also thank Anne Sophie Sarthou, Malou le Corronc-Nuzum (Université de

398

Technologie de Compiègne, Compiègne, France), Manon Surin and David Forest for their technical

399

support. S. Ben-Harb gratefully acknowledges the Tunisian Ministry of Higher Education for her

400

PhD fellowship.

401

References

402

Adda, J., Dekimpe, J., Vassal, L., & Spinnler, H.E. (1989). Production de styrène par Penicillium camemberti

403

Thom. Le Lait, 69, 115-120. https://doi.org/10.1051/lait:198928

404

Arfi, K., Spinnler, H.E., Tâche, R., & Bonnarme, P. (2002). Production of volatile compounds by cheese-

405

ripening yeasts: requirement for a methanethiol donor for S -methyl thioacetate synthesis by

406

Kluyveromyces

407

https://doi.org/10.1007/s00253-001-0925-0

lactis.

Applied

Microbiology

and

Biotechnology,

58,

503-510.

408

Ayad, E.H., Verheul, A., Engels, W.J., Wouters, J.T.M., & Smit, G. (2001). Enhanced flavour formation by

409

combination of selected lactococci from industrial and artisanal origin with focus on completion of a

17

410

metabolic pathway. Journal of Applied Microbiology, 90, 59-67. https://doi.org/10.1046/j.1365-

411

2672.2001.01219.x

412

Azarnia, S., Boye, J.I., Warkentin, T., Malcolmson, L., Sabik, H., & Bellido, A.S. (2011). Volatile flavour

413

profile changes in selected field pea cultivars as affected by crop year and processing. Food Chemistry,

414

124, 326-335. https://doi.org/10.1016/j.foodchem.2010.06.041

415

Ben-Harb, S., Panouillé, M., Huc-Mathis, D., Moulin, G., Saint-Eve, A., Irlinger, F., et al. (2017). The

416

rheological and microstructural properties of pea, milk, mixed pea/milk gels and gelled emulsions designed

417

by

418

https://doi.org/10.1016/j.foodhyd.2017.09.022

thermal,

acid,

and

enzyme

treatments.

Food

Hydrocolloids,

77,

75-84.

419

Ben-Harb, S., Saint-Eve, A., Panouillé, M., Souchon, I., Bonnarme, P., Dugat-Bony, E., et al. (2019). Design

420

of microbial consortia for the fermentation of pea-protein-enriched emulsions. International Journal of

421

Food Microbiology, 293, 124-136. https://doi.org/10.1016/j.ijfoodmicro.2019.01.012

422

Berthier, F., Beuvier, E., Dasen, A., & Grappin, R. (2001). Origin and diversity of mesophilic lactobacilli in

423

Comté cheese, as revealed by PCR with repetitive and species-specific primers. International Dairy

424

Journal, 11, 293-305. https://doi.org/10.1016/S0958-6946(01)00059-0

425

Bonaïti, C., Irlinger, F., Spinnler, H.E., & Engel, E. (2005). An iterative sensory procedure to select odour-

426

active associations in complex consortia of microorganisms: application to the construction of a cheese

427

model. Journal of Dairy Science, 88, 1671-1684. https://doi.org/10.3168/jds.S0022-0302(05)72839-3

428 429 430 431 432 433 434

Boutrou, R., & Guéguen, M. (2005). Interests in Geotrichum candidum for cheese technology. International Journal of Food Microbiology, 102, 1-20. https://doi.org/10.1016/j.ijfoodmicro.2004.12.028 Burdock, G.A. (2009). Fenaroli’s handbook of flavor ingredients (6th ed.). In CRC Press, Boca Raton, Fla. https://doi.org/10.1201/9781439847503 Campbell-Platt, G. (1994). Fermented foods – A world perspective. Food Research International, 27, 253257. Cochran, W.G. (1950), The comparison of percentages in matched samples, Biometrika, 37, 256-266.

18

435

Cogan, T.M., & Hill, C. (1993). Cheese Starter Cultures. In Fox, P.F. (Ed.), Cheese: Chemistry, Physics and

436

Microbiology.

Vol.1.

General

Aspects

437

https://doi.org/10.1007/978-1-4615-2650-6_6

(pp.

193-255).

Springer

US,

Boston,

MA,.

438

Dan, T., Wang, D., Wu, S., Jin, R., Ren, W., & Sun, T. (2017). Profiles of Volatile Flavor Compounds in

439

Milk Fermented with Different Proportional Combinations of Lactobacillus delbrueckii subsp. bulgaricus

440

and Streptococcus thermophilus. Molecules, 22(10), 1633. https://doi.org/10.3390/molecules22101633

441

Dugat-Bony, E., Straub, C., Teissandier, A., Onésime, D., Loux, V., Monnet, C., et al. (2015). Overview of a

442

Surface-Ripened Cheese Community Functioning by Meta-Omics Analyses. PLOS ONE, 10, e0124360.

443

https://doi.org/10.1371/journal.pone.0124360

444 445 446

El Sheikha, A.F., & Hu, D.M. (2019). Molecular Techniques Reveal More Secrets of Fermented Foods. Critical Reviews in Food Science and Nutrition (CRFSN). DOI: 10.1080/10408398.2018.1506906 Forquin-Gomez, M.P., Weimer, B.C., Sorieul, L., Kalinowski, J., &

Vallaeys, T. (2014). The Family

447

Brevibacteriaceae. In Rosenberg, E., DeLong, E.F., Lory, S., Stackebrandt, E., Thompson, F. (Eds.), The

448

Prokaryotes:

449

https://doi.org/10.1007/978-3-642-30138-4_169

Actinobacteria

(pp.

141-153).

Springer

Berlin

Heidelberg,

Berlin,

Heidelberg.

450

Gavrish, E.Y., Krauzova, V.I., Potekhina, N.V., Karasev, S.G., Plotnikova, E.G., Altyntseva, O.V., et al.

451

(2004). Three New Species of Brevibacteria, Brevibacterium antiquum sp. nov., Brevibacterium

452

aurantiacum sp. nov., and Brevibacterium permense sp. nov. Microbiology, 73, 176-183.

453

https://doi.org/10.1023/B:MICI.0000023986.52066.1e

454

Irlinger, F., Layec, S., Hélinck, S., & Dugat-Bony, E. (2015). Minireview: Cheese rind microbial

455

communities:

456

https://doi.org/10.1093/femsle/fnu015

457 458

diversity,

composition

and

origin.

FEMS

Microbiology

Letters,

362,

1-11.

Liu, S.Q., Holland, R., & Crow, V.L. (2004). Esters and their biosynthesis in fermented dairy products: a review. International Dairy Journal, 14, 923-945. https://doi.org/10.1016/j.idairyj.2004.02.010

459

Molimard, P., & Spinnler, H.E. (1996). Compounds involved in the flavor of surface mold-ripened cheeses:

460

origins and properties. Journal of Dairy Science, 79, 169-184. https://doi.org/10.3168/jds.S0022-

461

0302(96)76348-8

19

462

Monnet, C., Landaud, S., Bonnarme, P., & Swennen, D. (2015). Minireview: Growth and adaptation of

463

microorganisms

on

the

cheese

464

https://doi.org/10.1093/femsle/fnu025

surface.

FEMS

Microbiology

Letters,

362,

1-9.

465

Mounier, J., Monnet, C., Vallaeys, T., Arditi, R., Sarthou, A.-S., Helias, A., et al. (2008). Microbial

466

Interactions within a Cheese Microbial Community. Applied Environmental Microbiology, 74, 172-181.

467

https://doi.org/10.1128/AEM.01338-07

468

Mounier, J., Monnet, C., Jacques, N., Antoinette, A., & Irlinger, F. (2009). Assessment of the microbial

469

diversity at the surface of Livarot cheese using culture-dependent and independent approaches.

470

International Journal of Food Microbiology, 133, 31-37. https://doi.org/10.1016/j.ijfoodmicro.2009.04.020

471

Murat, C., Bard, M.H., Dhalleine, C., & Cayot, N. (2013). Characterisation of odour active compounds along

472

extraction process from pea flour to pea protein extract. Food Research International, 53, 31-41.

473

https://doi.org/10.1016/j.foodres.2013.03.049

474

Pagot, Y., Belin, J.M., Husson, F., & Spinnler, H.E. (2007). Metabolism of phenylalanine and biosynthesis of

475

styrene

476

https://doi.org/10.1017/S0022029906002251

477

in

Penicillium

camemberti.

Journal

of

Dairy

Research,

74,

180-185.

Prillinger, H., Molnár, O., Eliskases-Lechner, F., & Lopandic, K. (1999). Phenotypic and genotypic

478

identification

of

yeasts

from

cheese.

479

https://doi.org/10.1023/A:1001889917533

Antonie

Van

Leeuwenhoek,

75,

267-283.

480

Ravyts, F., De Vuyst, L.D., & Leroy, F. (2012). Bacterial diversity and functionalities in food fermentations:

481

Bacterial diversity and functionalities in food fermentations. Engineering in Life Sciences, 12, 356-367.

482

https://doi.org/10.1002/elsc.201100119

483

Schindler, S., Zelena, K., Krings, U., Bez, J., Eisner, P., & Berger, R.G. (2012). Improvement of the aroma of

484

pea (Pisum sativum) protein extracts by lactic acid fermentation. Food Biotechnology, 26, 58-74.

485

https://doi.org/10.1080/08905436.2011.645939

486

Sirtori, C.R., Triolo, M., Bosisio, R., Bondioli, A., Calabresi, L., De Vergori, V., et al. (2012).

487

Hypocholesterolaemic effects of lupin protein and pea protein/fibre combinations in moderately

20

488

hypercholesterolaemic

489

https://doi.org/10.1017/S0007114511004120

490

individuals.

British

Journal

of

Nutrition,

107,

1176-1183.

Smit, G., Smit, B., & Engels, W. (2005). Flavour formation by lactic acid bacteria and biochemical flavour

491

profiling

of

cheese

products.

492

https://doi.org/10.1016/j.femsre.2005.04.002

FEMS

Microbiology

Reviews,

29,

591-610.

493

Spinnler, H.E., Grosjean, O., & Bouvier, I. (1992). Effect of culture parameters on the production of styrene

494

(vinyl benzene) and 1-octene-3-ol by Penicillium caseicolum. Journal of Dairy Research, 59, 533-541.

495

https://doi.org/10.1017/S0022029900027205

496

Steele, D.H., Thornburg, M.J., Stanley, J.S., Miller, R.R., Brooke, R., Cushman, J.R., et al. (1994).

497

Determination of styrene in selected foods. Journal of Agricultural and Food Chemistry, 42, 1661-1665.

498

https://doi.org/10.1021/jf00044a015

499

Suzzi, G., Lanorte, M.T., Galgano, F., Andrighetto, C., Lombardi, A., Lanciotti, R., et al. (2001). Proteolytic,

500

lipolytic and molecular characterisation of Yarrowia lipolytica isolated from cheese. International Journal

501

of Food Microbiology, 69, 69-77. https://doi.org/10.1016/S0168-1605(01)00574-8

502

Valentin, D., Chollet, S., Lelièvre, M., & Abdi, H. (2012). Quick and dirty but still pretty good: A review of

503

new descriptive methods in food science. International Journal of Food Science and Technology, 47,

504

1563-1578. https://doi.org/10.1111/j.1365-2621.2012.03022.x

505

Yousseef, M., Lafarge, C., Valentin, D., Lubbers, S., & Husson, F. (2016). Fermentation of cow milk and/or

506

pea milk mixtures by different starter cultures: Physico-chemical and sensorial properties. LWT - Food

507

Science and Technology, 69, 430-437. https://doi.org/10.1016/j.lwt.2016.01.060

508

Yvon, M., Berthelot, S., & Gripon, J.C. (1998). Adding α-ketoglutarate to semi-hard cheese curd highly

509

enhances the conversion of amino acids to aroma compounds. International Dairy Journal, 8, 889-898.

510

https://doi.org/10.1016/S0958-6946(99)00011-4

21

Table legend Table 1 Distribution and growth (expressed as log N/N0, N: growth of strains at three (D3) and seven (D7) of fermentation; N0: initial cell density) of microbial species and pH values during the fermentation of pea gel (PG), pea and milk proteins (PMG) and milk gel (MG) by three microbial communities (VEGAN, MEGAN and ExEco) for three and seven days of fermentation. VEGAN consortium; MEGAN consortium; ExEco consortium. Values presented are the means of three replicate trials. a-c, mean values, referring to growth, within a row with different superscript letters are significantly different (p ≤ 0.05). Green: (N/N0>1); Orange (0.5
VEGAN consortium pH Time of fermentation (days)

Lactobacillus rhamnosus Leuconostoc lactis Lactococcus lactis Geotrichum candidum Candida catenulata Kluyveromyces lactis Debaryomyces hansenii Yarrowia lipolytica Brevibacterium antiquum Brevibacterium casei Brevibacterium aurantiacum Hafnia alvei Staphylococcus equorum Corynebacterium casei Glutamicibacter arilaitensis

D0 D3 D7

Pea 6.71±0.10 6.39±0.04 6.80±0.06

D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7 D3 D7

0.73 c 0.22 a 0.73 c 0.22 a 3.03 a 4.07 b 0.71 a 0.06 a 0.95 a 2.75 b 0.15 a 0.38 a 0.00 a 0.07 a 2.99 b 3.34 -

a

Mixed 6.29±0.14 5.19±0.17 4.79±0.01 a

1.60 b 2.13 a 1.60 b 2.13 a 3.08 a 3.40 b 0.68 a 0.21 a 0.57 b 0.00 b 0.14 b 0.00 a 0.00 a 0.07 b 2.61 c 2.90 -

MEGAN consortium Milk Pea Mixed 5.68±0.53 6.71±0.10 6.29±0.14 5.00±0.02 6.46±0.06 4.43±0.04 5.46±0.04 6.36±0.01 4.30±0.01 Species log CFUs/g* a c a 1.03 2.04 2.99 a c a 2.49 2.08 3.58 a 1.03 a 2.49 b a 1.68 3.53 b a 1.93 3.40 b ab a 2.08 2.45 2.65 b b a 3.58 2.96 3.34 a b a 1.14 0.61 1.90 a b a 0.22 2.28 2.95 c b 1.48 2.16 b b 2.82 2.80 a 0.35 b 0.00 a 0.58 b 0.00 a 0.00 a 0.31 c 2.21 a 3.46 -

a-c, mean values, referring to growth, within a row with different superscript letters are significantly different (p ≤ 0.05). Green: (N/N0>1); Orange (0.5
ExEco consortium

Milk 5.68±0.53 3.99±0.01 3.90±0.03 b

2.29 b 2.70 b 1.78 c 1.56 b 2.19 b 2.99 ab 1.17 c 1.30 a 4.12 a 4.39 -

Pea 6.71±0.10 6.52±0.09 6.98±0.21

Mixed 6.29±0.14 4.55±0.01 4.65±0.03

Milk 5.68±0.53 4.52±0.01 4.70±0.01

b 1.83 a 3.50 a 1.89 b 2.74

ab 2.29 c 2.65 a 2.07 a 3.19

a 2.52 b 2.92 b 1.32 b 2.93

c

1.77 a 2.50 a 0.48 b 0.00 a 0.00 a 0.00 c 1.12 c 1.39 a 0.00 b 0.00 a 0.00 a 0.00 a 0.00 b 0.00

a

1.08 ab 1.54 a 0.99 a 2.24 a 0.00 a 0.00 b 2.01 b 2.06 a 0.00 b 0.00 a 0.03 a 0.00 a 0.00 b 0.04

0.00 b 0.00 a 0.20 b 0.00 a 0.04 a 0.00 a 2.89 a 3.44 a 0.90 a 1.25 a 0.00 a 0.03 a 0.68 a 3.2

b

Figure legends Figure 1 Protocol for the preparation of fermented gels. Figure 2 Representation of the different fermented (after seven days of fermentation) and non-fermented gels, significant descriptors (p ≤ 0.05) in the first two dimensions of CFA (Correspondence Factorial Analysis) of data from Call All That Apply (CATA). A hierarchical ascendant cluster analysis (HCA) was carried out to group the tests according to their similarity, measured by the Pearson correlation, whereas cluster aggregation was based on the average linkage method (black circles). Data presented are the means of three replicate trials. A: Gel fermented with the VEGAN consortium; B: Gels fermented with the MEGAN consortium; C: Gels fermented with the ExEco consortium. Red: significant descriptors cited by the panel; blue: samples. Figure 3 Total Peak Area (TPA) of the different aromatic compounds of non-fermented (control samples) or fermented gels of pea, mixed, and milk proteins. Gels were fermented with three microbial consortia. A: control samples; B: VEGAN consortium; C: MEGAN consortium; D: ExEco consortium. Data (expressed as Total Peak Area) are the means of two replicate trials. Figure 4 Multiple Factorial Analysis (MFA) of fermented gels, sensory attributes, microbial communities and volatile compounds. Red: microbial communities; green: sensory attributes; purple: volatile compounds; blue: fermented samples. Data from the sensory and microbial analysis presented are the means of three replicate trials. Data of volatile compounds (expressed as Total Peak Area) are the means of two replicate trials.

Pea emulsion

Mixed emulsion

Milk emulsion

Coagulation with GDL (24h/25°C)

Fermentation (16°C/2 and 6 days)

Pea gel (PG)

Pea and milk gel (PMG)

Figure 1. Protocol for the preparation of fermented gels.

Milk gel (MG)

..

(axes F1 and F2: 82.17%)

(axes F1 and F2: 77.88%) 2

1.5

Neutral

A

B

VEGAN

MEGAN

1.5 1

1

Milk control

Ethanol Pea control Dried fruit Milk Fermented fruit Pea Cut herb Spicy Woody Pea Cheese Fresh curd Mixed control Solvent potato Mixed Sour

0.5

F2 (14.15%)

Pea

Coffee Pea

0.5

Smoked

Fresh cream

Onion/Garlic

0

Cut herb Woody Mixed control Dried fuit Pea control

Cooked milk

Yogurt Mixed Fresh curd Cheese Milk

F2 (20.77%)

Potato

0

-0.5

Honey -0.5

Rancid

Rind

Fresh milk Milk control

-1

Yogurt Melted butter

Neutral -1 -1.5

F1 (61.40%)

F1 (63.72%)

-2

-1.5 -1.5

-1

-0.5

0

0.5

1

1.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

(axes F1 and F2: 76.17%) 2.5

ExEco

C 1.5

Fermented fruit Honey Rind Dried fruit Mixed Cheese Pea control Fresh curd Woody Pea Roasted/Grilled Solvent Pea Milk Garlic/onion Cut herb Mixed control Apple Apricot Potato Yogurt

F2 (20.27%)

0.5

-0.5

Fresh milk Milk control -1.5

Melted butter F1 (55.91%) -2.5 -2.5

-1.5

-0.5

0.5

1.5

2.5

Figure 2. Representation of the different fermented (after seven days of fermentation) and non-fermented gels, significant descriptors (p ≤ 0.05) of the samples in the first two dimensions of CFA (Correspondence Factorial Analysis) of data from Call All That Apply (CATA). A hierarchical ascendant cluster analysis (HCA) was carried out to group the tests according to their similarity, measured by the Pearson correlation, whereas cluster aggregation was based on the average linkage method (black circles). Data presented are the means of three replicate trials. A: Gel fermented with the VEGAN consortium; B: Gels fermented with the MEGAN consortium; C: Gels fermented with the ExEco consortium. Red: significant descriptors cited by the panel; blue: samples.

A-Control

8.E+08

Ester Sulphur compounds

7.E+08 6.E+08

Ketones

5.E+08

Alkenes

4.E+08

Alkanes

3.E+08

Aromatic hydrocarbons

2.E+08

Aldehydes _Other

1.E+08

Aldehydes _green

0.E+00 Pea control

Mixed control

Milk control

Alcohols

C-MEGAN

B-VEGAN

D-ExEco

8.E+08

8.E+08

8.E+08

7.E+08

7.E+08

7.E+08

6.E+08

6.E+08

6.E+08

5.E+08

5.E+08

5.E+08

4.E+08

4.E+08

3.E+08

3.E+08

2.E+08

2.E+08

1.E+08

1.E+08

4.E+08 3.E+08 2.E+08 1.E+08 0.E+00 Pea-VEG-D7

Mixed-VEG-D7

Milk-VEG-D7

0.E+00

0.E+00 Pea-MEG-D7

Mixed-MEG-D7

Milk-MEG-D7

Pea-Ex-D7

Mixed-Ex-D7

Milk-Ex-D7

Figure 3. Total Peak Area (TPA) of the different aromatic compounds of non-fermented (control samples) or fermented gels of pea, mixed, and milk proteins. Gels were fermented with three microbial consortia. A: control samples; B: VEGAN consortium; C: MEGAN consortium; D: ExEco consortium. Data (expressed as Total Peak Area) are the mean values of two replicate trials.

strains

Variables (axes F1 and F2: 48,26%)

Observations (axes F1 and F2: 48,26%) 4

pH 1

sensory descriptors

Mixed gel

Lactobacillus rhamnosusVEG Geotrichum candidum VEG Yogurt Fresh cream

Molecules 0.75

Leuconostoc lactis-VEG

Milk gel Brevibacterium caseiVEG

Cooked milk 0.5

VEG7Milk 3

VEG7Mixed1

Honey Fresh milk

Aldehydes _Other

Pea gel

3

Hafnia alvei-VEG

VEG7Mixed2 2

Candida catenulata-VEG

VEG7Milk 1

VEG7Mixed3

Sulphur compounds

VEG7Milk 2 Ketones

0.25

Rancid Melted butter

F2 (17,11%)

Apricot Apple 0

Yarrowia lipolytica-VEG Brevibacterium antiquum -VEG Garlic/Onion Smoked Coffee Dried fruit

Brevibacterium aurantiacum-EX

VEG7Pea1

pH

Debaromyces hanseniiEX cut-herb Corynebacterium caseiSolvent Fresh curd -0.25 EX Ethanol Alkenes Alkanes Wooded Sour Aldehydes _green Esters Lactococcus lactis-EX Aromatic hydrocarbons Potato Pea Lactobacillus rhamnosusStaphylococcus Neutral MEG equorum-EX Candida catenulata -0.5 Arthobacter arilaitensisKluyveromyces lactisMEG EX Hafnia alvei-EX Lactococcus lactis-MEG MEG Geotrichum candidum MEG

VEG7Pea3

Ex7Mixed2 MEG-A7Mixed1

MEG-A7Mixed2 MEG-A7Mixed3

-1 MEG-A7Milk1

Ex7Milk 2 Ex7Milk 3 Ex7Milk 1

Ex7Mixed3

MEG-A7Milk2

Kluyveromyces lactis -EX

Spicy

-0.75

-1 -1.00

Ex7Mixed1

0

Griiled/Roasted

Alcohols

Fermented fruit

VEG7Pea2

1

F2 (17,11%)

Rind Cheese

MEG-A7Pea 3 MEG-A7Pea 2

MEG-A7Milk3

Ex7Pea1 Ex7Pea2 Ex7Pea3

MEG-A7Pea 1

-2

Geotrichum candidum EX

-3

-4 -0.75

-0.50

-0.25

0.00

F1 (31,16%)

0.25

0.50

0.75

1.00

-4

-3

-2

-1

0

1

2

3

4

F1 (31,16%)

Figure 4. Multiple Factorial Analysis (MFA) of fermented gels, sensory attributes, microbial communities and volatile compounds. Red: microbial communities; green: sensory attributes; purple: volatile compounds; blue: fermented samples. Data of sensory and microbial analysis presented are the means of three replicate trials. Data of volatile compounds (expressed as Total Peak Area) are the means of two replicate trials.

Highlights -

Microbial consortia were tested for the fermentation of pea protein-based matrices The design of microbial consortia adapted to the matrix could reduce off-flavors Metabolic activity and growth of the microbial consortia is driven by the matrix composition Our procedure can be used to produce diversified aroma and sensory profiles in peabased products

Conflict of interest The authors declare no conflict of interest.