Sonication induced particle formation in yogurt: Influence of the dry matter content on the physical properties

Accepted Manuscript Sonication induced particle formation in yogurt: Influence of the dry matter content on the physical properties Stefan Nöbel, Kristin Protte, Adrian Körzendörfer, Bernd Hitzmann, Jörg Hinrichs PII:

S0260-8774(16)30257-6

DOI:

10.1016/j.jfoodeng.2016.07.007

Reference:

JFOE 8615

To appear in:

Journal of Food Engineering

Received Date: 22 February 2016 Revised Date:

14 July 2016

Accepted Date: 18 July 2016

Please cite this article as: Nöbel, S., Protte, K., Körzendörfer, A., Hitzmann, B., Hinrichs, J., Sonication induced particle formation in yogurt: Influence of the dry matter content on the physical properties, Journal of Food Engineering (2016), doi: 10.1016/j.jfoodeng.2016.07.007. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. 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.

ACCEPTED MANUSCRIPT

Sonication induced particle formation in yogurt: Influence of the dry matter content on the physical properties Stefan Nöbel*, Kristin Protte*, Adrian Körzendörfer*, Bernd Hitzmann*, Jörg Hinrichs* Institute of Food Science and Biotechnology, University of Hohenheim, D-70593 Stuttgart, Germany *

Email address: [email protected] (Stefan Nöbel) Abstract

RI PT

Corresponding author. Tel.: +49 711 459 24208; Fax: +49 711 459 23617.

M AN U

SC

Yogurt texture can be adjusted by various compositional and processing factors. The factors protein content and type, heat treatment, and fermentation temperature are well-known and studied. In a previous study we were able to show sonication as an additional parameter affecting texture. The aim of this study was to test the hypothesis that sonication during fermentation induces large particles which, thus, have an impact on the texture perception of stirred yogurt. 26 stirred yogurt systems, differing in the fat ( 0.1  3.5% ) and protein content ( 3.8  5.2% ) were produced and a short sonication (35 kHz, 5 min) was applied within the pH range 5.1–5.2. Image analysis was performed to quantify the number and size of visually detectable particles/grains. Furthermore, the physical properties were studied by means of static light scattering and rheological measurements. Multivariate data analysis was used to discriminate between the effects of sonication and dry matter. Sonication resulted in new larger particles whereas an increase in dry matter content mainly affected the rheological properties. Above a dry matter content of 14.2%, no significant effect of the sonication was distinguishable.

Keywords: physical properties, dry matter content, acidification, graininess, sensorial evaluation, sonication

1. Introduction

EP

TE

D

Yogurt microgels are made by acidifying milk with lactic acid producing microbial starter cultures (LAB) (Courtin and Rul, 2004; Lucey, 2004; Settachaimongkon et al., 2014). In upstream processing, the milk is often concentrated prior to the fermentation or enriched with milk powder or single milk protein fractions (Sodini et al., 2004). A heat treatment improves the protein yield and results in tailored textural properties (Dannenberg and Kessler, 1988; Tamime and Robinson, 2007). During acidification a continuous gel network is formed. For stirred yoghurt, the gel network is broken up by post-fermentative mechanical processing resulting in microgel particles with diameters of about 2  100 m dispersed in milk serum (van Marle et al., 1999; Mokoonlall et al., 2016; Nöbel et al., 2014). In terms of sensorial attributes, stirred yogurt and other microgel suspensions, i.e., fresh cheese, are intended to be soft and viscous with a creamy texture, expel little whey, and a slightly acidic taste (Sonne et al., 2014).

AC C

A textural defect often occurring in fermented dairy products is graininess, which affects the visual assesment and in-mouth perception of the milk gel. Graininess arising due to sensorially detectable particles has been quantified using static light scattering (Cayot et al., 2008; Hahn et al., 2012a,b; Jørgensen et al., 2015; Krzeminski et al., 2013; Sonne et al., 2014) or image analysis (Küçükçetin et al., 2008, 2009; Remeuf et al., 2003; Sodini et al., 2005). Static light scattering analyzes the microgel particles at the micrometer-scale. In contrast, larger particles at the millimeter-scale ( 0.7  7 mm ) were identified by image analysis. Large particles have a negative impact on the visual assessment like graininess and lumpiness (Nöbel et al., 2016; Küçükçetin et al., 2009). Small microgel particle are able to aggregate beyond a certain threshold of d75,3  40 m and d90,3  100  150 m , indicating the 25 th- and 90 th-percentile of the volume weighted particle size distribution, and are perceived as grainy (Hahn et al., 2012b) and less creamy (Cayot et al., 2008) respectively. Krzeminski et al. (2013) reported that the in-mouth graininess of stirred yogurt was directly linked to the proportion of large particles. Besides the visual appearance, the large particles also affected the rheological properties since stirred yogurt is considered to be a microgel particle suspension (van Marle et al., 1999). The firmness and

ACCEPTED MANUSCRIPT flow properties are directly related to the size and distribution of the particles constituting the suspension (Hahn et al., 2015). Briefly, shear viscosity as well as moduli of soft sphere suspensions increase with an increasing effective volume fraction eff    and decreasing maximum packing fraction m    , where sticking occurs (Shewan and Stokes, 2015). An increased polydispersity of

RI PT

monomodal suspensions results in a higher maximum packing fraction and, thus, a lower viscosity of suspension. In multimodal suspensions, small particles become entrapped between large particles. Hence, the maximum packing fraction increases with the number of different particle size classes n, the size ratio of large and to small particles     and with an increasing fraction of large particles

 L    (Dörr et al., 2013). Several rheological parameters have been reported to be correlated to the

M AN U

SC

texture or microstructure of fermented dairy products (Krzeminski et al., 2013; Lucey et al., 1998; Norton et al., 2011). Non-destructive oscillatory tests, usually evaluated at a constant frequency such as f  1Hz or   10 rad s1 have been attributed to the gel strength, stiffness, or firmness (Lucey et al., 1998; Jaros et al., 2007; Norton et al., 2011). The flow properties, determined from destructive large deformation tests, resulted in a progressing structural breakdown during the measurement (AbuJdayil et al., 2013). Sonne et al. (2014) provided the most complete picture of relevant parameters by adding tribology data to conventional rheology. Further approaches endeavoured to gain insight into the microstructural levels by decomposing the mechanical spectra from rheological measurements into the contributions of the substructural elements within the microgel suspension (Pitkowski et al., 2008; Nöbel et al., 2014).

TE

D

We have shown in a previous study (Nöbel et al., 2016) that a short sonication of merely 5min during the fermentation increased the instrumental and sensorial visual graininess of stirred yogurt from skimmed milk. Different pH ranges were tested where pH 5.1–5.4 was identified as critical and causing an increasing the number and size of large microgel particles (millimeter-scale) as well as the average particle size of the small microgel particles (micrometer-scale). In Nöbel et al. (2016), we proposed two feasible mechanisms: a kinematic and a molecular approach. The first mentioned based on the intensified collision probability during ultra-sonication. Vibrations lead to additional shearing in the liquid resulting in particle motion relative to each other and particle aggregation (Johansson et al., 2016). The molecular mechanism draws on the disruption of whey proteins attached to the casein micelles. Sonication promotes the unfolding of the whey proteins and thiol groups get available (Frydenberg et al., 2016) which were reported to induce particle formation (Nguyen et al., 2015).

AC C

EP

This study is foremost intended to verify our hypothesis that oscillations during fermentation have an impact on the visual assessment of yogurt (Nöbel et al., 2016). The additional particle formation is proposed as a unique phenomenon apart from graininess occurring due to the yogurt composition under normal non-oscillating conditions. Thus, a wide range of 26 stirred yogurt systems differing in the composition were produced. Particle formation was induced by sonicating in the pH range 5.1–5.2 and studied by means of image analysis and laser diffraction spectroscopy (Nöbel et al., 2016). Various rheological parameters correlating to sensorial attributes and microstructural rearrangements were determined. A further objective of the present study is to discriminate the effects of sonication and compositional changes by means of multiple factor analysis (MFA) in order to identify the structural parameters affected the most by sonication.

2. Materials and methods 2.1. Fermentation and sonication The preparation of the milk, fermentation, sonication, and mechanical post-processing was carried out according to Nöbel et al. (2016) but altering the fat ( 0.1  3.5% ) and protein content ( 3.8  5.2% ) compared to our previous study (0.1% fat, 3.4% protein). Concisely, the fat and protein content of pasteurized milk was determined in triplicate using a mid-infrared spectrometer (LactoScope FTIR Advanced; Delta Instruments B.V., Drachten, The Netherlands). Based on the analysis, the protein content was standardized by dispersing low-heat skim milk powder Instant C (crude protein: 37% , lactose: 52% ; Schwarzwaldmilch GmbH, Offenburg, Germany). By mixing skimmed milk and high-

ACCEPTED MANUSCRIPT heated cream (90 C, 120 s) the final fat content was adjusted. After homogenizing (65 C, 150 / 30 bar ) and high heating (95 C, 256 s) the whole milk, the final composition was rechecked by FTIR.

SC

RI PT

Two liters of the milk were warmed up to 42 C, inoculated with 0.02% (w/v) Yo-Mix 215 (Danisco Deutschland GmbH, Niebull, Germany), and sealed in 100mL-glass jars. The fermentation at 42 C was carried out in (1) a standard water bath (K25-MPC, Huber Kältemaschinen GmbH, Offenburg, Germany) without sonication containing eight immersed glass jars and (2) a ultrasonic water bath (RK1028H; Bandelin electronic GmbH & Co. KG, Berlin, Germany) containing eight immersed glass jars. 300 W ultrasonic power at an excitation frequency of 35 kHz were applied to 20 L (500x300x200 mm) in the water bath. Therein the fermentation temperature was kept constant by an external recirculation (RE212, Lauda Dr. R. Wobser GmbH & Co. KG, Lauda-Königshofen, Germany). pH (BlueLine 14 pH, Schott AG, Mainz, Germany) and temperature (Pt100, Anton Paar GmbH, Graz, Austria) were recorded during the whole fermentation. The sonication was applied for about 5min after pH 5.2 was reached resulting in the pH range 5.1–5.2 as proposed by Nöbel et al. (2016). During sonication 90kJ of mechanical energy dissipated in the 20L-water bath corresponding to an average power density of 15kW / m3 in the immersed glass jars. At pH 4.6 the fermentation

M AN U

was stopped by cooling the glass jars for about 15 min on ice. After overnight storage (10 C) all samples were stirred according to a standard lab-scale procedure (Nöbel et al., 2016). No change in the fat and protein composition was expected due to the fermentation and postprocessing. Solely, the dry matter content of the final yogurt samples was determined with a gravimetric method by drying at 102 C (VDLUFA, 2010, C 35.3). The analysis was performed at least in duplicate.

2.2. Design of experiments

AC C

EP

TE

D

The influence of the preset parameters fat content, protein content, and sonication on the physical properties of stirred yogurt was analyzed using an enlarged Box-Wilson central composite design (Table 1). The dry matter content is determined by the fat and protein content as well. Owing to the large number of parameters and for the purpose of clarity abbreviations in the scheme “group.parameter” will be used in the following. Two ingredients were varied in each case at three levels (center and factorial points) and two unique levels (axial points): fat content ( ing.fat  0.1,0.6,1.8,3.0,3.5% ) and protein content ( ing.prot  3.8, 4.0, 4.5,5.0,5.2% ). Each experiment was carried out with and without sonication. The yogurt samples of each experimental point were produced and analyzed in duplicate and the center point ( ing.fat  1.8% , ing.prot  4.5% ) in triplicate in a fully randomized order. In the interest of comparability with our previous study (Nöbel et al., 2016), yogurt samples at a fat content of 0.1% and protein content of 3.4% were produced additionally in duplicate, giving a total of 26 differently composed samples (Table 1, sample code A–Z). Table 1: Sample code and composition of all yogurt samples in a fully randomized central composite design FTIRa

Sample code

Gravimetricb

Fat

Protein

Dry matterc

% (w/w)

% (w/w)

% (w/w)

A / A*

0.125±0.002 3.430±0.003 8.717±0.091

B / B*

0.108±0.002 3.540±0.002 8.792±0.057

C / C*

0.590±0.003 3.950±0.007 10.13±0.24

D / D*

0.108±0.002 4.510±0.008 10.70±0.05

0.620±0.004 4.190±0.001 10.78±0.12

F / F*

1.900±0.004 3.750±0.007 11.03±0.19

G / G*

0.620±0.014 4.450±0.088 11.43±0.05

H / H*

1.780±0.005 3.940±0.003 11.48±0.11

I / I*

0.590±0.004 4.360±0.011 11.76±0.10

J / J*

1.830±0.001 4.030±0.023 11.94±0.10

K / K*

3.030±0.015 3.800±0.183 12.57±0.09

L / L*

3.040±0.004 3.880±0.002 12.80±0.10

M / M*

d

1.810±0.003 4.520±0.005 12.84±0.13

N / N*

0.650±0.017 5.140±0.082 12.91±0.15

O / O*

0.560±0.003 5.100±0.010 12.99±0.05

P / P* d Q / Q*

1.830±0.006 4.430±0.006 13.02±0.09 d

1.780±0.005 4.470±0.005 13.24±0.05 3.570±0.008 4.260±0.011 14.03±0.06

S / S*

3.020±0.003 4.910±0.003 14.23±0.12

T / T*

3.090±0.002 4.420±0.008 14.36±0.09

U / U*

3.030±0.005 4.420±0.012 14.53±0.06

V / V*

0.140±0.004 5.640±0.001 14.57±0.06

W / W*

1.840±0.001 5.200±0.007 14.67±0.03

X / X*

1.850±0.001 5.070±0.007 14.72±0.13

Y / Y*

3.000±0.009 4.990±0.007 15.48±0.38

Z / Z*

3.320±0.003 5.520±0.002 17.35±0.09

D

TE

a

M AN U

R / R*

SC

E / E*

RI PT

ACCEPTED MANUSCRIPT

Fat and protein content were analyzed by FTIR Dry matter was analyzed by a gravimetric method (VDLUFA, 2010, C 35.3) c Samples were subsequently sorted ascending by the dry matter column d Center point of the 2-factor central composite design; repeated three times * Sonicated samples are indicated by asterisk

AC C

EP

b

2.3. Physical properties of microgel suspensions 2.3.1. Rheological characterization (rheo) Oscillatory small deformation and destructive large deformation analysis were carried out successively using stress-controlled rheometers, an AR2000 (minimum torque: 0.1μNm) or an AR2000ex (minimum torque: 0.03μNm) system (both TA Instruments Inc., New Castle, DE, USA). The chilled sample (10 C) was loaded to the cup ( D  31.1mm ) of the concentric cylinder geometry, the bob ( d  28.7 mm , l  43.0 mm ) was lowered, and the experiment was started after equilibration (10 C, 15 min). A frequency sweep was conducted from   0.05rad s 1 to 100 rad s (seven points per decade) in the predetermined linear viscoelastic region (   0.0025 ). Immediately after finishing the small deformation test, the flow curve of the same sample was recorded by linearly increasing ( 8  500s 1 ), holding ( 500s 1 ), and linearly decreasing ( 500  8s 1 ) the shear rate within 3 min per step. Shear stress values were registered every 3 s. Instrument control 1

ACCEPTED MANUSCRIPT and rheological data evaluation was performed with Rheology Advantage v5.8.0 software. All rheological measurements were performed at 10 C in triplicate.

RI PT

Two parameters from the frequency sweep and seven parameters from the flow curve were extracted and abbreviated according to the scheme rheo.parameter: (1) shear storage modulus at   10 rad s1 (rheo.modulus), (2) global minimum of the loss tangent (rheo.delta) to characterize the gel firmness and the elasticity respectively (Mokoonlall et al., 2016; Nöbel et al., 2014); shear stresses at certain characteristic points, i.e., (3) first maximum of the shear stress in the upward ramp (rheo.tauMAX), (4) corresponding shear rate at this maximum (rheo.gamMAX), (5) at   60s 1 in the upward ramp (rheo.tau60), (6) at   100s1 in the upward ramp (rheo.tau100), (7) at   500s 1 in the upward ramp (rheo.tau500), (8) apparent viscosity at   500s 1 after the holding step (rheo.eta500), and (9) apparent hysteresis loop area between upward and downward ramp (rheo.loop).

2.3.2. Image analysis (image)

M AN U

SC

The visual assessment of large particles (diameter  720 m ) in stirred yogurt was performed and analyzed according to Nöbel et al. (2016). Briefly, the yogurt samples were scratched out in a thin layer (1.2 mm) on a glass plate (120 x 90 mm) and transmission images (17.1 pixels/mm) were taken with a digital still camera (8-bit grayscale; MicroPublisher 3.3 RTV, QImaging, Surrey, Canada). Each image corresponds to approximately 13.6 g yogurt sample. Matlab 8.1 (R2013a, The MathWorks Inc., Natick, USA) was used for image processing, including contrast enhancement and binarization, and determining the number and shape of the particles. Six independent images of each sample were evaluated. Three parameters were extracted and abbreviated according to the scheme image.parameter: (1) number of particles per unit mass (image.number), (2) median diameter of the area weighted size distribution (image.d50), and (3) form factor or isoperimetric quotient (image.quot).

2.3.3. Laser diffraction spectroscopy (lds)

EP

TE

D

For the particle size distribution of the microgel particles ( d  500 m ) static light scattering (LS230, Beckman-Coulter Inc., Miami, USA) was used as described by Hahn et al. (2012a). Each sample was averaged from three consecutive runs by adjusting the obscuration to 14  16% in the instruments dispersion unit. The protein and water refractive index was 1.75 and 1.33 respectively. Instrument control and the evaluation of the volume weighted particle size distribution was performed with LS32 v3.19 software. The particle size measurement was repeated three times.

d

AC C

Four parameters describing the distribution shape and four parameters describing equivalent diameters were extracted and abbreviated according to the scheme lds.parameter: (1) second central moment as a measure of variance (lds.var), (2) third standardized moment as a measure of skewness (lds.skew), (3) fourth standardized moment as a measure of kurtosis (lds.kurt), (4) span defined as 90,3

 d10,3  / d50,3 (lds.span), (5) Sauter mean diameter or inverse specific surface (lds.d32), (6)

median diameter (lds.d50), (7) upper 25th-percentile (lds.d75), and (8) upper 10th-percentile (lds.d90) of the volume weighted distribution.

2.4. Statistical analysis All statistical calculations were performed with the R statistic language v3.1.1 (R Foundation for Statistical Computing, Vienna, Austria). Significance levels were set to   0.05 or given where appropriate. The univariate analysis of covariance (ANCOVA), Welch’s two-sample t-test, ShapiroWilk test of normality, and the correlation matrix were covered by R’s basic version. The covariance was chosen instead of a variance analysis (ANOVA) for taking the continuous nature of the parameter dry matter content (covariate) into account, covering a wide range of 8.72  17.4% (Table 1). Multivariate analysis involving all independent parameters of the experimental design and all physical properties were conducted by multiple factor analysis (MFA) with R’s package FactoMineR v1.31.3

ACCEPTED MANUSCRIPT (Lê et al., 2008). The suitable number of factors was predetermined using R’s package nFactors v2.3.3. The complete data set consisting of 52 experiments (rows), 4 independent, and 20 dependent variables (columns) can be accessed as supplementary material at [URL]. Sample codes and abbreviations in the supplemented table are identical to Table 1 and the scheme group.parameter respectively.

3. Results and discussion

RI PT

3.1. Acidification kinetics

M AN U

SC

The progress of the fermentation was compared to our previous study Nöbel et al. (2016) by means of the acidification kinetics to ensure the same time frame of the critical pH range (Figure 1). In order to induce an extensive formation of large particles during fermentation a short sonication in the pH range 5.1–5.2 was found to be sufficient. Although the same starter culture (Yo-Mix 215), milk pretreatment (95 C, 256 s), and fermentation temperature (42 C) were used, a significant shift in the characteristic sigmoid profile was observed. The critical pH range was already reached after 138 min of fermentation in this study. The final fermentation duration to pH 4.5 was within 377  384 min in both studies. Additionally, Figure 1a shows the acidification rate ( ΔpH / Δt ) from an approximated tangent in the critical pH range. The acidification rate calculated from the first order derivative of the fermentation profiles ( dpH / dt ) in this and the previous study were overlapping during the whole fermentation, indicating that mainly the lag-time was different (Figure 1b). The shifted fermentation profile might be due to the different lot of the same starter culture with an altered ratio and symbiosis of Streptococcus and Lactobacillus strains. As the acidification rate in the critical pH range was not significantly affected, the pH range 5.1–5.2 corresponding to 5 min sonication was still applied in this study.

D

Figure 1: pH value (a) and acidification rate (b) during fermentation using the starter culture YoMix 215 at   42 C; closed circle: this study, inoculation rate 0.1% (w/v), open circle: previous study (Nöbel et al., 2016), 0.02% (w/v); average points and standard errors calculated from n  13 ; gray shaded: proposed critical pH range 5.1–5.4 (Nöbel et al., 2016)

EP

TE

Slight changes in the selection and ratio of Streptococcus and Lactobacillus strains were reported to modify the shape and duration of the fermentation profile particularly with regard to higher fermentation temperatures (  40 C) (Lucey, 2004; Tamime and Robinson, 2007). A reduced initial pH and lag-time was caused by higher proteolytic activities of Lactobacillus (Courtin and Rul, 2004) whereas the overall acidification was accelerated by proteolysis due to Streptococcus strains (Settachaimongkon et al., 2014). Aghababaie et al. (2015) confirmed the cell count ratio 5:1 of Streptococcus salivarius subsp. thermophilus to Lactobacillus delbrueckii subsp. bulgaricus as most effective for the desired proto-cooperation of both species.

AC C

3.2. Formation of large particles As an example, Figure 2 shows representative transmission images with large and compact colloidal particles in the millimeter-scale, clearly observable as additional black dots in the homogeneous yogurt matrix. The formation of large particles was affected by the sonication during fermentation as well as the fat and protein content. In the reference samples without sonication few particles were observed regardless of the milk composition, e.g., 158  68 particles per 100 g yogurt (sample D; 4.5% protein, 0.1% fat). However, the number and size of the particles was markedly increased by applying sonication during fermentation in all cases, e.g., 417  76 particles per 100g yogurt (sample D*) which were significantly more particles than in the corresponding reference sample D ( P  0.01 ). Figure 2: Transmission images of scratched out yogurt samples sonicated at pH 5.1–5.2 during fermentation; (a) 3.8% and 5.2% protein at constant 1.8% fat; (b) 0.1% and 3.5% fat at constant 4.5% protein; average sample mass: 13.6g; average layer thickness: 1.2mm

ACCEPTED MANUSCRIPT

RI PT

Large amounts of particles in all sonicated samples supports our primary hypothesis that sonication during fermentation induces additional particles which, in turn, are negatively perceived in the homogeneous stirred yogurt (Nöbel et al., 2016). These particles were quantified as yet another phenomenon to the approved particle size assessment by means of laser diffraction spectroscopy (Nöbel et al., 2016). Both measurement techniques, image analysis and laser diffraction spectroscopy, covered the particle size range from 720 m to 2000 m but differ in their sensitivity towards the few large particles. Transmission images representing 13.6g of yogurt were analyzed at once (Figure 2) while the samples for the laser diffraction device had to be diluted at least by a factor of 10,000 (Hahn et al., 2012a). Graininess was strongly correlated to the upper 25th- (lds.d75) or 10th-percentile (lds.d90) of the volume weighted particle size distribution (Hahn et al., 2012a; Krzeminski et al., 2013). Hence, the effect of sonication on the visual graininess (image analysis) and on the particle size distribution (laser diffraction spectroscopy) were determined separately. Aside from the sonication induced particle formation, a detailed analysis of the phenomena in the context of the fat and protein content is required.

M AN U

SC

Further influences inducing the particle formation during the fermentation can be superimposed. An increased visual graininess and coarse texture in yogurt fermented from fortified milk was summarized by Sodini et al. (2004). Large casein particles were favored by a high dry matter content, a low casein-to-whey protein ratio, and intense heating (Küçükçetin et al., 2008, 2009). A coarse microstructure and large casein particle clusters were reported by the addition of sodium caseinate ( 5.0  5.3% protein) (Tamime et al., 1984). Yogurt samples enriched with skim milk powder were found to be softer with a smooth appearance (Tamime et al., 1984) and the lowest graininess compared to a fortification using caseinates and whey protein (Remeuf et al., 2003). Fortifying with skim milk powder in the present study provided the most smooth reference samples. Thus, sonication induced particle formation becomes as clear as possible.

3.3. Correlation of structural parameters

EP

TE

D

Multiple structural parameters were extracted from the non-destructive small amplitude oscillatory measurement (SAOS) and destructive flow curve (hysteresis loop) Figure 3 provides the entire correlation matrix of all dependent parameters that were determined for all stirred yogurt samples regardless of the ingredients and sonication. Most of the parameters in the measurement groups rheo and lds were moderate to strongly correlated within the groups indicating that these parameters were interrelated and gave a similar meaning in the context of this sample set (Figure 3). Some of them can be omitted without losing importance of the rheological or particle properties. Furthermore, the parameters were correlated frequently between the groups, except for image.d50 and lds.d32. The Sauter mean diameter (lds.d32) was exclusively moderate correlated ( r  0.297 , p  0.0327 ) to the maximum shear stress (rheo.tauMAX). Both parameters were either unique for a specific structural property which was not covered by other measures or random within all yogurt samples.

AC C

Figure 3: Correlation matrix of all dependent parameters extracted from rheology (rheo), image analysis (image), and laser diffraction spectroscopy (lds) including all yogurt samples; Pearson coefficient of correlation determined by linear regression; insignificant correlations are crossed out red ( P  0.05 ); bold face: parameters remaining for the reduced data set All dependent parameters that were already discussed in Nöbel et al. (2016) plus one selected rheological parameter from small amplitude (shear storage modulus at 10 rad s 1 , rheo.modulus) and flow curve (shear stress at 60s 1 , rheo.tau60) were included in the analysis of covariance (Table 2). The altered dry matter content (ing.dm) significantly affected the visual appearance, particle size and flow properties for all parameters ( P  0.01), except for the median diameter from image analysis (image.d50). Sonication alone or combined with the dry matter was only significant in case of the parameters image.number, image.d50, and image.quot ( P  0.05 ). In half of the test cases the linear model behind the ANCOVA exhibited an insufficient coefficient of determination (adj. r 2  0.5 ). An additional two-sample t-test was performed only on the influence of the sonication. This confirmed the highly significant effect of the sonication on the number (image.number) and median diameter

ACCEPTED MANUSCRIPT (image.d50) of the large particles ( P  0.001). The preliminary condition of normality, when comparing these parameters with regards to the sonication levels, was satisfied (Shapiro-Wilk test, p  0.05 ). Table 2: Analysis of covariance (ANCOVA) on the effect of sonication and dry matter content on selected particle and rheological parameters; alternative two-sample t-test on the effect of sonication

Image analysis image.numb er

Laser diffraction image.qu ot

lds.d5 0

lds.d7 5

lds.d9 0

rheo.modul us

rheo.tau 60

1.09

8.54 ***

11.6

56.3

104

224 ***

315 ***

**

***

***

SC

17.6 ***

Rheology

image.d 50

one-way ANCOVAb ing.dm

RI PT

Dependent parametersa

Test statistic

27.4 ***

20.9 ***

4.73 *

1.10

1.78

2.34

1.86

0.380

ing.dm

5.84 *

2.54

5.18 *

0.158

0.095 1

0.022 9

0.0282

0.270

0.494

0.296

0.238

0.159

0.514

0.664

0.815

0.857

t statistic

−4.37 ***

−4.50 ***

1.96

−0.96 1

−0.92 4

−0.87 6

−0.585

0.229

Effectiv e df

44.0

50.0

49.8

50.0

50.0

49.5

50.0



sonicati on adj. r 2

D

t-testc

M AN U

sonicati on

TE

41.8

b

EP

n  52 independent observations Linear regression model: dep. parameter  ing.dm (quantitative)  sonication (2 levels); tabulated F statistics at F 1, 48

a

Welch’s t-test of two unpaired samples with unequal variance (with vs. without sonication) adj. r 2 : Adjusted coefficient of determination df : degree of freedom *,**,*** Probability P  0.05 , P  0.01, and P  0.001 respectively

AC C

c

At a glance, the results are consistent with literature describing an overall increase of the strength characteristics of stirred yogurt by an enhanced dry matter content and constant casein-to-whey protein ratio (Cheng et al., 2002; Jaros et al., 2007; Küçükçetin et al., 2008; Lankes et al., 1998; Tamime and Robinson, 2007). However, with regards to the univariate analysis of covariance or the ttest, no clear assignment of sonication and the combined effect of fat, protein, or dry matter content can be established.

3.4. Multivariate analysis of structural parameters Factorial analysis maps the complex structure of multiple variables to a significantly lower number of dimensions. Dimensions, factors, or so called principal components in case of PCA contain the most important information but can be reliably interpreted (Pagès, 2015). Several antagonistic criteria were proposed to determine the appropriate number of factors. Figure 4 shows the scree plot from all

ACCEPTED MANUSCRIPT dependent parameters and all yogurt samples, where additionally Horn’s (parallel analysis of random data, n  3 ) and Kaiser’s criterion (eigenvalue   1 , n  4 ) were marked. Another graphical approach is to interpret the curvature of the scree plot and reject factors, where the eigenvalue λ starts to tail off. We used Horn’s criterion in this study and retained the first three dimensions ( n  3 ). A further reduction to two dimensions was probably equivalent and more comprehensible as the steep decline from   11.8 to 2.41 ceased after two factors (Figure 4).

RI PT

Figure 4: Scree plot of all factors derived from the dependent parameters; circle: observed eigenvalues, square: eigenvalues from Horn’s parallel analysis of uncorrelated random data; Horn criterion: observed eigenvalue  random eigenvalue, Kaiser criterion: observed eigenvalue  1

M AN U

SC

Multiple factor analysis (MFA) was applied to correlate rheological, image analysis and laser diffraction data with the sonication during the fermentation and the yogurt composition. MFA was used because of its mixed design including quantitative and qualitative parameters (Lê et al., 2008; Pagès, 2015). The parameters were grouped according to their source, i.e., rheology (rheo), image analysis (image), and laser diffraction spectroscopy (lds), and more in detail to their meaning, e.g., lds.shape and lds.size. An alternate linear discriminant analysis (LDA) was discarded as the basic clusters were previously unknown. For instance, sonication during the fermentation was not necessarily the main driving effect on all physical parameters of the yogurt samples. Visual assessment of the transmission images (Figure 2) and the correlation matrix (Figure 3) already indicated that parameters are interrelated and arise from multiple factors such as fat and protein content.

TE

D

Table 3 summarizes the results from the MFA including all 20 dependent parameters from rheology, image analysis, laser diffraction spectroscopy within the first three dimensions. The first constructed factor was capable to explain 43.9% of the variance, 20.5% by the second, and 9.68% by the third factor. Higher dimensions were reject by the Horn criterion (Figure 4). Focusing on the most contributing parameters, rheo.tau60, rheo.tau100, lds.d75, lds.d90, lds.var, and lds.span were positively loaded on the first factor and the parameters image.d50 and image.number from image analysis on the second factor. The first dimension was negatively correlated to the parameter rheo.delta whereas the third factor was positively correlated to the parameters lds.d32, lds.d50, and image.d50. Interestingly, the sonication was exclusively and positively correlated to the second factor that was also positively correlated to the median diameter (image.d50) of the large particles. Table 3: Correlation of the first, second and third factor from multiple factor analysis (MFA) of all dependent parameters Correlation coefficienta  ri

EP

Parameter

Factor 1

AC C

rheo.

Factor 2

Factor 3

modulusb

0.922 ***

−0.0866

−0.0722

deltab

−0.564 ***

0.417 **

0.404 **

tauMAXb

0.911 ***

−0.176

0.0788

gamMAX

−0.736 ***

0.185

0.330 *

tau60b

0.931 ***

−0.231

−0.00680

tau100

0.916 ***

−0.247

−0.0158

tau500

0.879 ***

−0.269

−0.0212

eta500

0.895 ***

−0.257

0.00249

ACCEPTED MANUSCRIPT 0.825 ***

−0.276 *

−0.110

numberb

0.589 ***

0.575 ***

0.0246

d50b

0.0227

0.736 ***

−0.266

quot

−0.520 ***

−0.357 ***

0.118

Var

0.834 ***

−0.0723

0.182

skewnb

0.683 ***

−0.0234

−0.403 **

kurt

0.446 ***

−0.0182

−0.399 **

spanb

0.931 ***

−0.0722

−0.0976

d32

0.322 *

0.123

0.827 ***

d50

0.642 ***

0.0960

0.701 ***

d75b

0.868 ***

0.0241

0.459 ***

d90b

0.929 ***

−0.0102

0.301 *

sonicationc

0.0246

0.809 ***

0.00372

20.5%

9.68%

loop image.

43.9%

a

SC

M AN U

Exp. varianced

RI PT

lds.

D

Coefficient ri and significance of correlation between factor i (column) and parameter (row) Parameters remaining for the reduced data set c Correlation of the independent parameter sonication d Proportion of the variance explained by each factor (column) bold face: 4 most contributing parameters within each factor (column) *,**,*** Probability P  0.05 , P  0.01, and P  0.001 respectively

EP

TE

b

AC C

Several parameters showed almost the same correlation with regard to each factor but with different levels of contributions (Table 3). Hence, two parameters of each measurement procedures were chosen. These parameters in the reduced data set should be either unique or highly contributing to the construction of the dimensions, namely rheo.modulus, rheo.delta, rheo.tauMAX, rheo.tau60, image.number, image.d50, lds.skew, lds.span, lds.d75, and lds.d90. As a lot of the selected parameters exhibited a asymmetric distribution with a high density on the left-hand side (data not shown), an additional logarithmic transformation was proposed: • rheo.storage.LN  log(rheo.storage) ; normality p  0.0955 • lds.span.LN  log(lds.span) ; normality p  0.413 • lds.d75.LN  log(lds.d75) ; normality p  0.0460 • lds.d90.LN  log(lds.d90) ; normality p  0.0200 The transformations, abbreviated according to the scheme group.parameter.LN, reduced the asymmetry and avoided outliers without eliminating individual observations. The remaining six parameters were not transformed. Subsequently, the assumption of normality was complied in most

ACCEPTED MANUSCRIPT cases (Shapiro-Wilk test, p  0.01). In the following, all analyses will be carried out with the reduced data set of (1) the parameters selected above, according to their contribution and uniqueness; and (2) the transformed parameters in order to ensure multivariate normality with a minimum number of outliers. Ten transformed and standardized dependent parameters remained for further analysis.

3.5. Discrimination of sonicated yogurt samples

RI PT

Results from the MFA of the reduced data set revealed that the remaining first and second factor explained 75.2% of the total variance, with the proportions 51.5% and 23.7% dedicated to first and second dimension respectively (Figure 5). The total explained inertia was enhanced compared to the analysis of the full data set in the previous section. A renewed determination of the appropriate number of factors using Horn and Kaiser criteria confirmed that two factors were adequate (data not shown). In the context of sonication and different yogurt compositions, the number of interpretable dimensions were, thus, reduced to a more comprehensible complexity.

M AN U

SC

Figure 5: Correlation of the first and second factor (a) from multiple factor analysis (MFA) of the reduced data set, dashed vector: supplementary parameters, parentheses: proportion of the variance explained by each factor; (b) Individual factor map of all trials, A–Z: capital letters referring to sample codes in Table 1, * asterisked capital letters referring to sonicated samples, supp.: supplementary data from previous study (Nöbel et al., 2016); subsequent hierarchical clustering at 3 levels (inertia criterion) All standardized parameters were well represented by the first two dimensions of the MFA (correlation coefficients r  0.6 ), except for rheo.delta ( r1  0.57 , r2  0.39 , P  0.001). Interpreting the first factor in terms of the dependent measures, mainly the laser diffraction parameters which are attributed the distribution’s width were highly correlated, e.g., lds.span.LN ( r1  0.96 ,

P  0.001) and lds.d90.LN ( r1  0.94 , P  0.001). The supplemented parameters from the yogurt composition, which were not used in the construction of the factorial analysis, were correlated to the first factor as well; the dry matter content (ing.dm) was the most representative ( r1  0.94 ,

TE

D

P  0.001) from this group. Consequently, the first factor is referred as the ingredients dimension. The fat (ing.fat) and protein content (ing.prot) were correlated to the first factor ( r1  0.62 and 0.72 respectively, P  0.001) with a similar extent and direction (Figure 5). Thus, all structural parameters

AC C

EP

related to ingredients dimension were independent whether the dry matter was altered by the fat or protein fraction. From literature a differentiated influence of the fat content was expected (Chandrapala and Leong, 2015). Owing to the acoustic cavitation fat droplets were emulsified by high-intensity sonication (Sfakianakis et al., 2015; O’Sullivan et al., 2015) as well as separated by acoustic radiation force from high-frequency sonication (Leong et al., 2014; Johansson et al., 2016). In already homogenized milk, on the other hand, the milk fat globules are covered by a protein membrane. During acidification the milk fat globules are considered to be incorprorated into the gel network (Sodini et al., 2004). In terms of the short sonication applied at pH 5.1–5.2, here, the homogenized milk fat globules contributed to the particle formation in a similar fashion as the milk proteins. The following discussion aims only on the dry matter content (ing.dm) as the representative ingredients measure. The increasing dry matter content resulted in an increased size of the microgel particles as well as an overall increase of the rheological parameters (Figure 5). The second factor was distinctly dominated by the median diameter (image.d50) calculated from the image analysis of the large particles ( r2  0.79 , P  0.001). From the independent experimental parameters, just the sonication was significantly correlated to this factor ( r22  0.80 , P  0.001). The second factor is referred as the sonication dimension below. The number of large particles (image.number) was positively correlated to both parameters, the ingredients ( r1  0.60 ) and the sonication dimension ( r2  0.57 ), indicating that image.number was enhanced by an increasing dry matter content in a similar manner as by applying sonication during the fermentation. In contrast, the size of these large particles (image.d50)

ACCEPTED MANUSCRIPT was not affected at all by an increasing dry matter content. Most of the rheological as well as the parameters from laser diffraction spectroscopy were exclusively associated to the ingredients dimensions. Despite the additional large particles, sonication had no significant effect on the flow properties of the stirred yogurt. This was maybe due to their low number compared to the microgel particles from the homogeneous gel network (Nöbel et al., 2016).

3.6. Hierarchical cluster analysis

SC

RI PT

Figure 5b maps the individual experiments (samples A – Z) within the two dimensions constructed from the reduced data set. As all of the sonicated samples (A* – Z*) were located in the upper quadrants of the plane, the partitions along the sonication dimension (factor 2) were already obvious. The observed fundamental difference within the two multifactorial dimensions was evaluated in detail by a subsequent hierarchical cluster analysis (Ward’s method; Pagès, 2015): Three clusters were well defined as a large break in the inertia from the first (5.23), to the second (2.87), and third branch (2.03) of the clustering tree was observed (data not shown). The between-clusters inertia decreased only of 0.537 by further choosing four instead of three clusters. Drawing on the classification from the cluster analysis (Figure 5b), we divided the effect of sonication into two major influences concerning the size and number of the large particles: (a) sonication during fermentation induced additional large particles and mainly increased their size (cluster 1 vs. 2), and (b) the additional number of large particles was negligible at high dry matter content (cluster 3 vs. 1 + 2).

AC C

EP

TE

D

M AN U

Clusters 1 and 2 clearly discriminated between samples with and without sonication during fermentation (Figure 5). Samples fermented without sonication were overrepresented in cluster 1 (17 of 26 experiments) whereas sonicated samples were overrepresented in cluster 2 (15 of 26 experiments). The most typical experiment (center) of cluster 1 was sample M which, at the same time, was the central point of the underlying experimental design (Table 1). Samples in cluster 1 had less particles (image.number) and lower viscosities (rheo.tauMAX, rheo.tau60), at least significantly lower than the overall average ( P  0.01). Inversely, sample F* was the most representative of cluster 2 which had significantly larger particles (image.d50) and lower viscosities (rheo.tau60) than the average ( P  0.001). The remaining samples were pooled in the third cluster, 9 reference as well as 11 sonicated samples. Cluster 3 was not particularly characterized by the sonication dimension (factor 2) or any parameter from image analysis. This cluster was composed of samples that have the highest parameters of lds.d90.LN, lds.span.LN, and rheo.tauMAX, thus, yogurt produced from the highest dry matter content ( P  0.001). A clear limit at which the dry matter content dominated on the sonication effect appeared between the samples R/S to T (cluster 1 to 3) and R* and S* (cluster 2 to 3) respectively (Figure 5b). This corresponds to a dry matter content of about 14.2% composed of approximately 3.0  3.5% fat and 4.3  4.9% protein (Table 1). Yogurt produced from a higher dry matter content was, on average, not affected by the sonication during fermentation anymore. This means, in turn, that only weak gels were sensitive to sonication in regard to forming large particles, while the gel network was formed within the pH range 5.1–5.2. Two supplementary experiments (supp.) were included in the data set presented in Figure 5. The yogurt was produced by a similar lab protocol (3.41% protein, 0.08% fat, 9.35% dry matter) in our previous work (Nöbel et al., 2016) and the parameters of the image analysis, laser diffraction spectroscopy and some from rheology were determined. Without sonication, the former reference sample (supp.) matched the interior of cluster 1 being consistent with the low dry matter samples in this work. The sonicated sample (supp.*) was located markedly increased by the second dimension and far away from the center of cluster 2, mainly due to the increased number of the large particles of 1980 per 100 g compared to 363 per 100 g of cluster 2. With increasing protein content, a general gain of negatively perceived textural attributes, such as an increased graininess, was reported by several authors (Sodini et al., 2004; Tamime and Robinson, 2007). High protein levels and high-heating (95 C, 5 min) always favored a grainy and lumpy appearance of stirred yogurts (Jørgensen et al., 2015), regardless of the casein-to-whey protein ratio. However, the authors have not investigated the effect due to an altered fat content. The disruption of weak gels by shaking during fermentation (pH 4.8–5.6) resulting in a reduced viscosity (Driessen

ACCEPTED MANUSCRIPT

RI PT

et al., 1977) was not corroborated by our results. The energy input applied in this study was low (90 kJ per 20L) compared to former studies on ultra-sonicated milk (Riener et al., 2009; Nguyen and Anema, 2010; Chandrapala et al., 2013). Thus, the number of particles and the protein content fixed within them was insufficient for affecting the rheological properties of the sonicated yogurt samples. Since our measurements showed two distinct clusters with (cluster 2) and without sonication (cluster 1), the kinematic mechanism was confirmed to be the most likely (Nöbel et al., 2016). On the one hand, the collision probability of network fragments was increased by the sonication and the low viscosity. On the other hand, the increasing dry matter content also improved the collision probability. However, above the dry matter limit of 14.2% (cluster 3) the raised bulk viscosity of the micellar suspension retards the additional shearing due to sonication. In consequence, the exact limit is valid for the specific combination of viscosity (temperature, composition, content) and energy input (excitation source, frequency range).

4. Conclusions

Acknowledgments

TE

D

M AN U

SC

The results of our study showed that sonication during fermentation reproducibly induces the formation of large particles. Furthermore, multivariate statistical analysis allows discriminating between stirred yogurts which were fermented with diffent compositions and sonication according to few rheological parameters and the visual assessment. Multivariate analysis was found to be more reliable than univariate ANCOVA, especially in case of the large number of experiments and parameters. This showed that an increasing dry matter content results in more particles as well and, above a content of about 14.2%, dominates on the sonication effect. Only weak gels were found to be sensitive to sonication in regard to forming large particles. Active mechanical treatment of yogurt can be minimized if the fermentation was carried out under idle conditions and at low dry matter content. The rheological properties and particle size distribution were just significantly affected by the dry matter content. Hence, we propose that sonication has no impact on the flow properties of the stirred yogurts. This hypothesis should be further examined, possibly by an extensive study of sensory attributes, namely graininess, creaminess, and viscosity. In addition, the fundamental threshold for the sonication induced particle formation should be determined by applying different levels of energy input and frequency ranges of excitation. Additional information about the contribution of the fat content during sonication may be obtained by substituting fat globules with milk fat fractions or by emulsifying using low molecular weight surfactants.

AC C

References

EP

This research project was supported by the German Ministry of Economics and Technology (via AiF) and the FEI (Forschungskreis der Ernährungsindustrie e.V., Bonn). Project AiF 17535 N. The authors thank Dr. Alina Sonne for advice in multivariate statistical analysis and Aryama Mokoonlall for proof reading the manuscript.

Abu-Jdayil, B., Nasser, M.S., Ghannam, M., 2013. Structure breakdown of stirred yoghurt in a circular pipe as affected by casein and fat content. Food Science and Technology Research 19, 277– 286. doi:10.3136/fstr.19.277. Aghababaie, M., Khanahmadi, M., Beheshti, M., 2015. Developing a kinetic model for co-culture of yogurt starter bacteria growth in ph controlled batch fermentation. Journal of Food Engineering 166, 72–79. doi:10.1016/j.jfoodeng.2015.05.013. Anonymous, 2010. Methodenbuch. volume VI: Milch. 7 ed., VDLUFA-Verlag, Darmstadt.

ACCEPTED MANUSCRIPT Cayot, P., Schenker, F., Houzé, G., Sulmont-Rossé, C., Colas, B., 2008. Creaminess in relation to consistency and particle size in stirred fat-free yogurt. International Dairy Journal 18, 303–311. doi:10.1016/j.idairyj.2007.06.009. Chandrapala, J., Leong, T., 2015. Ultrasonic Processing for Dairy Applications: Recent Advances. Food Engineering Reviews 7, 143–158. doi:10.1007/s12393-014-9105-8.

RI PT

Chandrapala, J., Zisu, B., Kentish, S., Ashokkumar, M., 2013. Influence of ultrasound on chemically induced gelation of micellar casein systems. Journal of Dairy Research 80, 138–143. doi:10.1017/S0022029912000696.

Cheng, L.J., Clarke, P.T., Augustin, M.A., 2002. Seasonal variation in yogurt properties. Australian

SC

Journal of Dairy Technology 57, 187–191.

Courtin, P., Rul, F., 2004. Interactions between microorganisms in a simple ecosystem: yogurt

M AN U

bacteria as a study model. Le Lait 84, 125–134. doi:10.1051/lait:2003031.

Dannenberg, F., Kessler, H.G., 1988. Effect of denaturation of β-lactoglobulin on texture properties of set-style nonfat yoghurt. 2. Firmness and flow properties. Milchwissenschaft 43, 700–704. Dörr, A., Sadiki, A., Mehdizadeh, A., 2013. A discrete model for the apparent viscosity of polydisperse suspensions including maximum packing fraction. Journal of Rheology 57, 743–765. doi:10.1122/1.4795746.

D

Driessen, F.M., Ubbels, J., Stadhouders, J., 1977. Continuous Manufacture of Yogurt – I. Optimal

TE

Conditions and Kinetics of the Prefermentation Process. Biotechnology and Bioengineering 19, 821– 839. doi:10.1002/bit.260190604.

Frydenberg, R.P., Hammershøj, M., Andersen, U., Greve, M.T., Wiking, L., 2016. Protein

EP

denaturation of whey protein isolates (wpis) induced by high intensity ultrasound during heat gelation. Food Chemistry 192, 415–423. doi:10.1016/j.foodchem.2015.07.037.

AC C

Hahn, C., Nöbel, S., Maisch, R., Rösingh, W., Weiss, J., Hinrichs, J., 2015. Adjusting rheological properties of concentrated microgel suspensions by particle size distribution. Food Hydrocolloids 49, 183–191. doi:10.1016/j.foodhyd.2015.03.020. Hahn, C., Sramek, M., Nöbel, S., Hinrichs, J., 2012a. Post-processing of concentrated fermented milk: influence of temperature and holding time on the formation of particle clusters. Dairy Science & Technology 92, 91–107. doi:10.1007/s13594-011-0046-1. Hahn, C., Wachter, T., Nöbel, S., Weiss, J., Eibel, H., Hinrichs, J., 2012b. Graininess in fresh cheese as affected by post-processing: Influence of tempering and mechanical treatment. International Dairy Journal 26, 73–77. doi:10.1016/j.idairyj.2011.12.013.

ACCEPTED MANUSCRIPT Jaros, D., Heidig, C., Rohm, H., 2007. Enzymatic Modification Through Microbial Transglutaminase Enhances the Viscosity of Stirred Yogurt. Journal of Texture Studies 38, 179–198. doi:10.1111/j.1745-4603.2007.00093.x. Johansson, L., Singh, T., Leong, T., Mawson, R., McArthur, S., Manasseh, R., Juliano, P., 2016. Cavitation and non-cavitation regime for large-scale ultrasonic standing wave particle separation

RI PT

systems – In situ gentle cavitation threshold determination and free radical related oxidation. Ultrasonics Sonochemistry 28, 346–356. doi:10.1016/j.ultsonch.2015.08.003.

Jørgensen, C.E., Abrahamsen, R.K., Rukke, E.O., Johansen, A.G., Schüller, R.B., Skeie, S.B., 2015. Improving the structure and rheology of high protein, low fat yoghurt with undenatured whey

SC

proteins. International Dairy Journal 47, 6–18. doi:10.1016/j.idairyj.2015.02.002.

Krzeminski, A., Tomaschunas, M., Köhn, E., Busch-Stockfisch, M., Weiss, J., Hinrichs, J., 2013. Relating Creamy Perception of Whey Protein Enriched Yogurt Systems to Instrumental Data by

M AN U

Means of Multivariate Data Analysis. Journal of Food Science 78, S314–S319. doi:10.1111/17503841.12013.

Küçükçetin, A., Weidendorfer, K., Hinrichs, J., 2008. Effect of heat treatment and dry matter on graininess and roughness of stirred skim milk yoghurt. Milchwissenschaft 63, 269–272. Küçükçetin, A., Weidendorfer, K., Hinrichs, J., 2009. Graininess and roughness of stirred yoghurt as

D

influenced by processing. International Dairy Journal 19, 50–55. doi:10.1016/j.idairyj.2008.07.006. Lankes, H., Ozer, H.B., Robinson, R.K., 1998. The effect of elevated milk solids and incubation

TE

temperature on the physical properties of natural yoghurt. Milchwissenschaft 53, 510–513. Lê, S., Josse, J., Husson, F., 2008. FactoMineR: An R package for multivariate analysis. Journal of

EP

Statistical Software 25, 1–18. doi:10.18637/jss.v025.i01. Leong, T., Johansson, L., Juliano, P., Mawson, R., McArthur, S., Manasseh, R., 2014. Design

AC C

parameters for the separation of fat from natural whole milk in an ultrasonic litre-scale vessel. Ultrasonics Sonochemistry 21, 1289–1298. doi:10.1016/j.ultsonch.2014.01.007. Lucey, J.A., 2004. Cultured dairy products: an overview of their gelation and texture properties. International Journal of Dairy Technology 57, 77–84. doi:10.1111/j.1471-0307.2004.00142.x. Lucey, J.A., Munro, P.A., Singh, H., 1998. Rheological Properties and Microstructure of Acid Milk Gels as Affected by Fat Content and Heat Treatment. Journal of Food Science 63, 660–664. doi:10.1111/j.1365-2621.1998.tb15807.x. van Marle, M.E., van den Ende, D., de Kruif, C.G., Mellema, J., 1999. Steady-shear viscosity of stirred yogurts with varying ropiness. Journal of Rheology 43, 1643–1662. doi:10.1122/1.551065.

ACCEPTED MANUSCRIPT Mokoonlall, A., Nöbel, S., Hinrichs, J., 2016. Post-processing of fermented milk to stirred products: Reviewing the effects on gel structure. Trends in Food Science & Technology 54, 26–36. doi:10.1016/j.tifs.2016.05.012. Nguyen, N.H.A., Anema, S.G., 2010. Effect of ultrasonication on the properties of skim milk used in the formation of acid gels. Innovative Food Science and Emerging Technologies 11, 616–622.

RI PT

doi:10.1016/j.ifset.2010.05.006. Nguyen, N.H.A., Anema, S.G., Guyomarc’h, F., Wong, M., Havea, P., 2015. The effect of the addition of thiol reagents to heated milk on protein interactions and acid gelation properties. International Dairy Journal 42, 42–50. doi:10.1016/j.idairyj.2014.10.005.

SC

Nöbel, S., Hahn, C., Hitzmann, B., Hinrichs, J., 2014. Rheological properties of microgel

suspensions: Viscoelastic modelling of microstructural elements from casein micelles to fermented

M AN U

dairy products. International Dairy Journal 39, 157–166. doi:10.1016/j.idairyj.2014.06.001. Nöbel, S., Ross, N.L., Protte, K., Körzendörfer, A., Hitzmann, B., Hinrichs, J., 2016. Microgel particle formation in yogurt as influenced by sonication during fermentation. Journal of Food Engineering 180, 29–38. doi:10.1016/j.jfoodeng.2016.01.033.

Norton, I.T., Spyropoulos, F., Cox, P. (Eds.), 2011. Practical Food Rheology – An Interpretive Approach. Wiley, Chichester.

D

O’Sullivan, J., Murray, B., Flynn, C., Norton, I., 2015. Comparison of batch and continuous ultrasonic emulsification processes. Journal of Food Engineering 167, 114–121.

TE

doi:10.1016/j.jfoodeng.2015.05.001.

Pagès, J., 2015. Multiple Factor Analysis by Example Using R. The R Series, CRC Press, Boca

EP

Raton.

Pitkowski, A., Durand, D., Nicolai, T., 2008. Structure and dynamical mechanical properties of

AC C

suspensions of sodium caseinate. Journal of Colloid and Interface Science 326, 96–102. doi:10.1016/j.jcis.2008.07.003. Remeuf, F., Mohammed, S., Sodini, I., Tissier, J., 2003. Preliminary observations on the effects of milk fortification and heating on microstructure and physical properties of stirred yogurt. International Dairy Journal 13, 773–782. doi:10.1016/S0958-6946(03)00092-X. Riener, J., Noci, F., Cronin, D.A., Morgan, D.J., Lyng, J.G., 2009. The effect of thermosonication of milk on selected physicochemical and microstructural properties of yoghurt gels during fermentation. Food Chemistry 114, 905–911. doi:10.1016/j.foodchem.2008.10.037. Settachaimongkon, S., Nout, M.J.R., Antunes Fernandes, E.C., Hettinga, K.A., Vervoort, J.M., van Hooijdonk, T.C.M., Zwietering, M.H., Smid, E.J., van Valenberg, H.J.F., 2014. Influence of different

ACCEPTED MANUSCRIPT proteolytic strains of Streptococcus thermophilus in co-culture with Lactobacillus delbrueckii subsp. bulgaricus on the metabolite profile of set-yoghurt. International Journal of Food Microbiology 177, 29–36. doi:10.1016/j.ijfoodmicro.2014.02.008. Sfakianakis, P., Topakas, E., Tzia, C., 2015. Comparative Study on High-Intensity Ultrasound and Pressure Milk Homogenization: Effect on the Kinetics of Yogurt Fermentation Process. Food and

RI PT

Bioprocess Technology 8, 548–557. doi:10.1007/s11947-014-1412-9. Shewan, H.M., Stokes, J.R., 2015. Viscosity of soft spherical micro-hydrogel suspensions. Journal of Colloid and Interface Science 442, 75–81. doi:10.1016/j.jcis.2014.11.064.

Sodini, I., Lucas, A., Tissier, J.P., Corrieu, G., 2005. Physical properties and microstructure of

SC

yoghurts supplemented with milk protein hydrolysates. International Dairy Journal 15, 29–35. doi:10.1016/j.idairyj.2004.05.006.

M AN U

Sodini, I., Remeuf, F., Haddad, C., Corrieu, G., 2004. The Relative Effect of Milk Base, Starter, and Process on Yogurt Texture: A Review. Critical Reviews in Food Science and Nutrition 44, 113–137. doi:10.1080/10408690490424793.

Sonne, A., Busch-Stockfisch, M., Weiss, J., Hinrichs, J., 2014. Improved mapping of in-mouth creaminess of semi-solid dairy products by combining rheology, particle size, and tribology data. LWT – Food Science and Technology 59, 342–347. doi:10.1016/j.lwt.2014.05.047.

D

Tamime, A.Y., Kalab, M., Davies, G., 1984. Microstructure of set-style yoghurt manufactured from cow’s milk fortified by various methods. Food Microstructure 3, 83–92.

TE

Tamime, A.Y., Robinson, R.K., 2007. Yoghurt – Science and Technology. Number 140 in

AC C

EP

Woodhead Publishing in Food Science, Technology and Nutrition. 3 ed., Woodhead, Cambridge.

ACCEPTED MANUSCRIPT

1

Table captions

2 Table 1: Sample code and composition of all yogurt samples in a fully randomized central

4

composite design

RI PT

3

5

Table 2: Analysis of covariance (ANCOVA) on the effect of sonication and dry matter

7

content on selected particle and rheological parameters; alternative two-sample t-test on the

8

effect of sonication

9

M AN U

SC

6

Table 3: Correlation of the first, second and third factor from multiple factor analysis (MFA)

11

of all dependent parameters

14 15 16 17 18 19

EP

13

Figure captions

AC C

12

TE D

10

Figure 1: pH value (a) and acidification rate (b) during fermentation using the starter culture Yo-Mix 215 at ϑ = 42 °C; closed circle: this study, inoculation rate 0.1 % (w/v), open circle: previous study (Nöbel et al., 2016), 0.02 % (w/v); average points and standard errors calculated from n ≥ 13; gray shaded: proposed critical pH range 5.1–5.4 (Nöbel et al., 2016)

1

ACCEPTED MANUSCRIPT

20

Figure 2: Transmission images of scratched out yogurt samples sonicated at pH 5.1–5.2

21

during fermentation; (a) 3.8% and 5.2% protein at constant 1.8% fat; (b) 0.1% and 3.5% fat at

22

constant 4.5% protein; average sample mass: 13.6 g; average layer thickness: 1.2 mm

RI PT

23

Figure 3: Correlation matrix of all dependent parameters extracted from rheology (rheo),

25

image analysis (image), and laser diffraction spectroscopy (lds) including all yogurt samples;

26

Pearson coefficient of correlation determined by linear regression; insignificant correlations

27

are crossed out red (P < 0.05); bold face: parameters remaining for the reduced data set

M AN U

SC

24

28

Figure 4: Scree plot of all factors derived from the dependent parameters; circle: observed

30

eigenvalues, square: eigenvalues from Horn's parallel analysis of uncorrelated random data;

31

Horn criterion: observed eigenvalue > random eigenvalue, Kaiser criterion: observed

32

eigenvalue > 1

EP

33

TE D

29

Figure 5: Correlation of the first and second factor (a) from multiple factor analysis (MFA) of

35

the reduced data set, dashed vector: supplementary parameters, parentheses: proportion of the

38

AC C

34

39

clustering at 3 levels (inertia criterion)

36 37

variance explained by each factor; (b) Individual factor map of all trials, A–Z: capital letters referring to sample codes in Table 1, *asterisked capital letters referring to sonicated samples, supp.: supplementary data from previous study (Nöbel et al., 2016); subsequent hierarchical

2

ED

M AN

AC C

EP

TE

D SC

M AN U R

EP TE D

SC

M AN U

R

TE D

M AN U

ED

M AN

ACCEPTED MANUSCRIPT

Institute of Food Science and Biotechnology

University of Hohenheim (150e) | 70593 Stuttgart

Soft Matter Science and Dairy Technology

Editorial board of the Journal of Food Engineering

RI PT

Prof. Dr.-Ing. habil. Jörg Hinrichs

per pro Stefan Nöbel T +49 711 459 24208 F E

+49 711 459 23617 [email protected]

M AN U

Dear Dr. Evans, Dear editorial board of the Journal of Food Engineering,

SC

14.07.2016

Please find below the revised highlights of the manuscript: "Sonication induced particle formation in yogurt: Influence of the dry matter content on the physical properties", by Stefan Nöbel, Kristin Protte, Adrian Körzendörfer, Bernd Hitzmann, and Jörg Hinrichs. Particle formation due to sonication during yogurt fermentation is investigated

•

Increasing the dry matter content and applying sonication favors large particles

•

Rheological properties and particle sizes of stirred yogurt are mainly related to the dry matter content

Additional particle formation due to sonication is only observed below 14.2% dry matter

Sincerely yours,

AC C

Jörg Hinrichs

EP

•

TE D

•

1I1 UNIVERSITY OF HOHENHEIM Garbenstr. 21 70599 Stuttgart Germany www.uni-hohenheim.de/english

INFORMATION https://www.uni-hohenheim.de/institution/ fg-lebensmittel-tierischer-herkunft

DIRECTIONS Tram U3, Plieningen (Universität Hohenheim) Bus 65, 70, 73, 74, 75, 76, 79