Continuous centrifugal separation of selectively precipitated α-lactalbumin

Journal Pre-proof Continuous centrifugal separation of selectively precipitated α-lactalbumin Nicole Haller, Ulrich Kulozik PII:

S0958-6946(19)30203-1

DOI:

https://doi.org/10.1016/j.idairyj.2019.104566

Reference:

INDA 104566

To appear in:

International Dairy Journal

Received Date: 24 February 2019 Revised Date:

8 September 2019

Accepted Date: 9 September 2019

Please cite this article as: Haller, N., Kulozik, U., Continuous centrifugal separation of selectively precipitated α-lactalbumin, International Dairy Journal, https://doi.org/10.1016/j.idairyj.2019.104566. 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

Continuous centrifugal separation of selectively precipitated α-lactalbumin

2 3 4 5 6 7

Nicole Haller*, Ulrich Kulozik

8 9 10 11 12 13

Chair of Food and Bioprocess Engineering, Technical University of Munich, Weihestephaner

14

Berg 1, 85354 Freising, Germany

15 16 17 18 19 20

*Corresponding author. Tel.: +49 8161 713536

21

E-mail address: [email protected] (N. Haller)

22

23

___________________________________________________________________________

24

ABSTRACT

25 26

The whey proteins α-lactalbumin (α-la) and β-lactoglobulin (β-lg) are valuable proteins for

27

various food, nutraceutical, and medical applications. Their fractionation is, however, still a

28

challenging task on industrial scale. This study describes a novel separation approach using

29

selective thermal precipitation of α-la under acidic conditions, followed by high-throughput

30

continuous centrifugation. The obtained α-la enriched sediment presented pseudoplastic flow

31

behaviour with strong dependency on dry matter. As the sediment still contained residues of

32

soluble β-lg and buffer salts, two steps of reslurry washing were performed. Along the three

33

centrifugation steps, a slight susceptibility to particle fragmentation was recognised, however,

34

separation efficiency was rarely impaired by it. Finally, two high purity protein fractions were

35

obtained, the supernatant contained 99.7% pure and native β-lg, while the washed α-la

36

sediment fraction was 99.4% free of β-lg.

37

___________________________________________________________________________

1

38

1.

Introduction

39 40

The whey proteins α-lactalbumin (α-la; 14.2 kDa) and β-lactoglobulin (β-lg; 18.0 kDa)

41

are small globular proteins of exceptional biological value and excellent functionality. Their

42

specific properties and potential application areas are summarised in numerous reviews

43

(Chatterton, Smithers, Roupas, & Brodkorb, 2006; De Wit, 1990; Marshall, 2004; Smithers,

44

2008) and are briefly outlined in the following.

45

On the one hand, formulae enriched with α-la were shown to be superior to classical

46

infants’ formula (Heine, Radke, Wutzke, Peters, & Kundt, 1996). This can be attributed, on

47

the one hand, to the high similarity to the amino acid profile of mothers’ milk, and on the

48

other hand to the absence of β-lg in human milk (Heine, Klein, & Reeds, 1991; Jost, Maire,

49

Maynard, & Secretin, 1999; Lien, Alvarez, Menéndez, Riera, & Alvarez, 2004; Trabulsi et al.,

50

2011). On the other hand, β-lg comprises a high amount of branched chain amino acids (Etzel,

51

2004) and is appreciated for its technofunctional properties, i.e., gelation, emulsification, and

52

foaming characteristics (Das & Kinsella, 1989; Dombrowski, Gschwendtner, & Kulozik,

53

2017; Mulvihill & Kinsella, 1988). Moreover, studies have attested antimicrobial, antiviral,

54

anti-tumour properties to these proteins (Håkansson et al., 2000; Ng et al., 2015; Pellegrini,

55

Thomas, Bramaz, Hunziker, & Fellenberg, 1999; Rammer et al., 2010).

56

In nature, α-la and β-lg do not appear in an isolated form, but in a complex aqueous

57

protein-carbohydrate system, generally known as whey. There is still a lack of an efficient

58

process for fractionation of these two proteins that provides high yield, completely

59

undenatured products, and is suitable for scale-up. Several approaches for isolation of the

60

single protein fractions are described in literature and are briefly reviewed in the following.

61

Primarily, ion-exchange chromatography generally achieves high purities for the

62

isolated fractions; however, it presents major drawbacks regarding throughput and price

63

efficiency (Santos, Teixeira, & Rodrigues, 2012). Another approach is two-phase partition, 2

64

which often comprises one aqueous and one organic solvent phase. Thereby, the affinity of α-

65

la to some organic solvents is exploited, while β-lg mainly stays in the aqueous phase

66

(Kalaivani & Regupathi, 2013). A recent publication presents an environmental-friendly

67

aqueous two-phase flotation approach using a PEG/trisodium citrate system with subsequent

68

dialysis for isolation of α-la (Jiang et al., 2018). The achieved purities were promisingly high;

69

however, this process requires a proof of concept at industrially relevant scale. In another

70

study, the aqueous two-phase flotation was shown to be superior to aqueous two-phase

71

extraction for the fractionation of antioxidant peptides from whey hydrolysate (Jiang et al.,

72

2019). Alternatively, the recovery of β-lg by NaCl salting-out at low pH values, followed by

73

centrifugation and diafiltration for salt removal was demonstrated by Maté and Krochta

74

(1994). The purity of 95% was satisfactory, but the recovery rate was only 65%, and also the

75

denaturation of the proteins might be severe with this procedure. Recent investigations show

76

that casein exhibits chaperone like activities under high hydrostatic pressure and enhances the

77

aggregation of β-lg. This effect may be used to deliver α-la-enriched fractions (Marciniak et

78

al., 2019).

79

Another emerging application in whey protein fractionation is high gradient magnetic

80

separation (Chen, Guo, Guan, & Liu, 2007; Meyer, Berensmeier, & Franzreb, 2007). Here,

81

magnetic beads are loaded with the target protein, magnetically separated, and after a washing

82

step, target protein can be recovered from the particles. This method appears to have high

83

selectivity and might be competitive to other chromatographic solutions. However, this

84

application is limited to lactoferrin, which incorporates ferrous ions. Further to that, also the

85

separation based on membrane filtration of the native proteins seems to be not appropriate as

86

the molecular weights of α-la and β-lg present too little difference (Heidebrecht & Kulozik,

87

2019). However, the separation of a fraction containing both proteins, α-la and β-lg, from the

88

other minor whey proteins comprising higher molecular weights can be successfully

89

conducted by ultrafiltration (Bhattacharjee, Ghosh, Datta, & Bhattacharjee, 2006). 3

90

In contrast to the above described separation methods, in this study a combination of

91

selective aggregation of one candidate protein with a subsequent centrifugal separation is

92

suggested. A reversible precipitation mechanism via hydrophobic interactions for α-la is

93

known. Based on this knowledge, a process can be designed, that enables the reception of the

94

α-la and β-lg fraction in native state.

95

The basic research for the selective aggregation of α-la using acidic conditions close to

96

its isoelectric point (i.e., pH 4.6) were conducted by Bramaud, Aimar, and Daufin, (1995;

97

1997a), Maubois, Pierre, Fauquant, and Piot (1987), Pearce (1983; 1987) and Pierre and

98

Fauquant (1986). The model derived describes the equilibrium between the calcium-bound

99

holo- and the calcium-free apo-form of α-la as the central aspect of the selective precipitation

100

reaction. This conversion into the apo-form under acidic conditions on the one hand reduces

101

the heat stability of the protein (Hendrix et al., 2000), but on the other hand it increases its

102

hydrophobicity. By the addition of a calcium sequestrant, a further improvement of this

103

precipitation method was realised. Thus, shifting of the equilibrium towards the apo-α-la is

104

promoted, as demonstrated by various authors (Alomirah & Alli, 2004; Bramaud, Aimar, &

105

Daufin 1997b; Lucena, Alvarez, Menéndez, Riera, & Alvarez, 2006; Tolkach & Kulozik,

106

2008; Toro-Sierra, Tolkach & Kulozik, 2013). Independent of the choice of α-la

107

destabilisation strategies, the secondary structure stays widely intact and allows subsequent

108

molecule dissociation and later refolding under adequate conditions (Permyakov & Berliner,

109

2000; Toro-Sierra et al., 2013).

110

As the precipitation mechanism of α-la and its influencing factors are well understood,

111

the need for an industrially adequate separation method for large scale applications arises.

112

Studies describing the removal of the precipitated α-la fraction by membrane filtration,

113

however, resulted in low permeability due to fouling issues (Gésan-Guiziou, Daufin, Timmer,

114

Allersma, & Van der Horst, 1999; Toro-Sierra et al., 2013). Centrifugal separation of the

115

precipitates has been successfully conducted in discontinuous mode (Fernández, Menéndez, 4

116

& Riera, 2012; Lucena et al., 2006), and a promising approach was demonstrated by means of

117

a continuously working laboratory scale decanter (Haller & Kulozik, 2019). The advantage of

118

continuous centrifugation as separation method is its simplicity and speed, also enabling the

119

removal of small precipitated aggregates under high acceleration forces. The challenges related to centrifugal separation of α-la aggregates were demonstrated

120 121

in our previous study (Haller & Kulozik, 2019). The α-la precipitates were characterised as

122

fragile and small particles of low density. The sediment formed had a sludgy to pasty

123

appearance and exhibited a high adhesion tendency. Moreover, it showed poor drainage

124

capacity despite high compressibility. In this study, we wanted to evaluate the suitability of a co-current decanter centrifuge

125 126

for large-scale separation of precipitated α-la. Hence, one key question was whether a

127

continuous sediment discharge could be achieved with this type of decanter, thus resulting in

128

two high purity fractions, i.e., a high dry matter α-la sediment and a well clarified β-lg

129

supernatant. The hypothesis was that a decanter characterised by a bowl construction without

130

drying zone, i.e., with a fully submerged separation room, should allow for better sediment

131

transportation characteristics. Furthermore, we expected higher sediment dewatering by

132

consolidation as could be achieved by drainage on a drying zone in the lab-scale device. The

133

second key question therefore arose as to whether the fragile precipitates would maintain their

134

physical integrity along the separation process, despite high mechanical stress imposed by

135

high g-forces, compressive yield stress and shearing through the scroll. Thus, it was critical to

136

investigate the aggregate stability of the α-la precipitates during the washing process, which is

137

important for the realisation of a high yield of the α-la fraction and a high purity of soluble

138

native β-lg in the supernatant.

139 140

2.

Material and methods

141 5

142

2.1.

Preparation of α-la aggregate suspensions

143 144

Whey protein concentrate (WPC; WPC80 Milei GmbH, Leutkirchen, Germany) with a

145

protein content of 80.0% (Vario MAX cube, Elementar Analysesysteme, Langenselbold,

146

Germany) and whey protein isolate (WPI; Davisco, Le Seur, Minnesota, USA) with a protein

147

content of 94.2%, as related to the dry matter (CEM, Kamp-Lintfort, Germany), served as

148

basis for the preparation of α-la aggregate suspensions. WPC80, made from fresh sweet whey,

149

had a protein composition of 21% α-la, 34% β-lg A, 32% β-lg B, and 13% minor proteins,

150

with 5% lactose and, on a WPC powder basis, 2.4 mg g-1 sodium and 1.3 mg g-1 calcium. The

151

WPI had a protein composition of 18% α-la, 44% β-lg A, 30% β-lg B, and 8% minor proteins

152

with 6.0 mg g-1 sodium, 1.2 mg g-1 calcium, and no lactose.

153

The different whey protein powders (WPC80 or WPI) were dissolved in deionised

154

water (DIW) to a final protein concentration of 150 g L-1 using a powder mixer system

155

(Fristam Pumpen KG, Hamburg, Germany). The end volume of WPC batch was 750 L, and

156

250 L for WPI. Gentle agitation for 12 h at 4 °C ensured complete hydration of the proteins.

157

Prior to further processing, the whey protein solutions were tempered to 20 °C. Then, the pH

158

value was adjusted to 3.4 using trisodium citrate dihydrate and citric acid monohydrate

159

(Bernd Kraft GmbH, Duisburg, Germany), whereby a final citrate content of 60 g L-1 was

160

striven for (Toro-Sierra et al., 2013). Selective formation of α-la aggregates was achieved by

161

subsequent heat treatment by indirect heating via the heating jacket of a 500 L tank. Thereby,

162

a constant heating rate of 0.5 K min-1 was applied and strictly regulated. Steady stirring of the

163

whey protein solutions at 150 rpm ensured even heating. Upon reaching of 50.0 °C ± 0.5 °C, a

164

holding period of 120 min was followed by a fast cooling phase down to a temperature of

165

20.0 °C. For monitoring of the progress of α-la aggregation, whey protein solution samples

166

were taken every 15 min and analysed using RP-HPLC (Section 2.3).

167 6

168

2.2.

Centrifugal separation

169 170

Separation of α-la aggregate suspensions was conducted with the industrial-scale

171

decanter centrifuge Sedicanter® S3E-3 (Flottweg SE, Vilsbiburg, Germany), which is

172

illustrated in Fig. 1. It is a horizontal bowl centrifuge with an integrated conveyor screw,

173

wherefore continuous separation of solid and liquid phases is enabled. The basic working

174

principle is as follows: the aggregate suspension is pumped through an inlaying pipe to the

175

end of the double conical bowl, where it enters the separation zone via the feed port. The

176

conveyor scroll transports the sediment cake towards the junction of the two cones, where the

177

radius is maximal, resulting in high compressive forces on the solids. The sediment is pressed

178

underneath the weir disc and is actively discharged by a second scroll at the steep cone. The

179

liquid phase is collected by an impeller disc and is discharged under pressure as the so-called

180

centrate.

181

The precipitate suspension made from WPC80, which was produced as described in

182

section 2.1, was diluted with DIW at 20 °C or 50 °C to 400 L with dry matter content of 20%

183

(w/w), 800 L with 10% (w/w), and 400 L with 6% (w/w), respectively. The precipitate

184

suspension from WPI was diluted to a volume of around 800 L with a dry matter content of

185

6% (w/w).

186

Suspensions were kept under agitation at 100 rpm to ensure a uniform temperature and

187

to avoid sedimentation of the aggregates in the jacketed feed tank. If needed, the centrifuge

188

was pre-warmed with hot DIW to adapt the bowl temperature to the suspension. Specific

189

constructional details and the used settings are listed in Table 1. Weir diameter and

190

differential speed were regulated inline to obtain best separation results. The suspension was

191

pumped from the jacketed vessel to the centrifuge using a NEMO® Progressing Cavity Pump

192

(Netzsch Pumpen & Systeme GmbH, Waldkraiburg, Germany). Sampling was performed

193

after reaching a steady state of separation performance, representing 3–4 bowl throughputs. 7

After initial separation, the sediment still contained soluble β-lg and citrate. To

194 195

maximise the purity of the α-la fraction, reslurry washing was performed. This was done by

196

suspending of the sediment in DIW at 50 °C, which was acidified with 3 N HCl to pH 3.4,

197

corresponding to a wash ratio of 4 (wash liquid is four times the volume of the residual

198

humidity in the cake). Reslurried suspensions were separated again into sediment and liquid

199

phase using an acceleration of 10,000 × g. In total, two washing steps of sediment were

200

performed. The inlet dry matter contents were 8.7% and 7.9% for the first and second wash,

201

respectively. For each separation step, samples were taken simultaneously from the inlet, the

202

centrate and the sediment.

203 204

2.3.

Analysis of precipitate suspensions

205 206

To investigate the progress in aggregate formation during heat treatment, taken

207

samples were immediately cooled down in an ice water bath to stop further reaction. Firstly,

208

the sample preparation comprised a pre-dilution with DIW to a protein content around 20 mg

209

mL-1. From each sample, an aliquot of 200 µL was directly mixed with 800 µL guanidine

210

buffer (6 N guanidine buffer containing 19.5 mM DTT) to determine the total protein content.

211

The rest of the sample was centrifuged at 6000 × g for 15 min in a laboratory centrifuge

212

(Multifuge 1S-R, Heraeus Holding GmbH, Hanau, Germany). The supernatant was then

213

filtered through a 0.45 µm syringe filter to remove remaining precipitates, and again an

214

aliquot of 200 µL was diluted in the guanidine buffer and forwarded to RP-HPLC analysis.

215

Incubation time for complete dissolution of proteins was at least 30 min. In Dumpler,

216

Wohlschläger, and Kulozik (2017), a constant value for the whey protein content was proven

217

for incubation times up to 72 h. In this study all samples were generally analysed within 24 h

218

(maximum 48 h) after preparation; samples were stored chilled until analysis started.

8

An Agilent 1100 Series chromatograph (Agilent Technologies, Waldbronn, Germany)

219 220

was used, equipped with a C18 analytical silica-based column (Agilent Zorbax 300SB-C18,

221

4.6 × 150 mm, 5 µm) as described by Dumpler et al. (2017). Solvent A consisted of 0.1%

222

trifluoroacetic acid (TFA), 10% acetonitrile in HPLC grade water. Solvent B was 0.07% TFA,

223

90% acetonitrile in HPLC grade water. The injection volume was 20 µL, the flow rate was set

224

to 1.2 mL min-1, while the column temperature was constantly at 40 °C. Protein detection was

225

done with a UV Detector at 226 nm. The yield of the precipitated fraction XP for each protein x (x = α-la or β-lg) was

226 227

calculated according to Eq. 1. Hereby, the difference of total concentration cx,total of each

228

protein x in the suspension to the concentration of this protein in the supernatant cx,supernatant is

229

used.

230

=

,





,

,

(Eq. 1)

Furthermore, particle size distribution of the precipitates was measured by laser

231 232

diffraction using the Malvern Mastersizer 2000 equipped with the Malvern Hydro 2000S

233

sample dispersion unit (Malvern Instruments GmbH, Herrenberg, Germany). Refractive

234

indexes for water (dispersant) and protein were set to 1.33 and 1.41, respectively, according to

235

manufacturer’s protocol. The dry matter content of the suspensions was determined using differential weighing

236 237

(CEM Smart Turbo CEM GmbH, Kamp-Lintfort, Germany).

238 239

2.4.

Analysis of centrates

240 241

For evaluation of the separation success achieved by centrifugation, the separation

242

efficiency ηsep was calculated, based on RP-HPLC data (Section 2.3). The ηsep of precipitated

243

α-la from the centrate was then calculated using concentration of α-la in centrate cα-la, centr, 9

244

total obtained volume of centrate Vcentr, as well as α-la concentration cα-la, inlet and total volume

245

Vinlet of inlet suspension.

246

-

= 1−

,!

∙ #!

,$

∙ #$

∙ 100%

(Eq. 2)

247

Furthermore, quantification of citrate content was done via HPLC as described by

248

Schmitz-Schug (2014). Proteins in the samples were removed by the addition of 50 µL of

249

60% perchloric acid to 1 mL of sample and subsequent filtration through a 0.45 µm syringe

250

filter. Samples were diluted with DIW to fit the calibration range.

251 252

2.5.

Analysis of the sediments

253 254 255 256

Success in removal of soluble β-lg and citrate from the sediment were calculated as given in Eq. 3 and 4, and expressed as ηsep (β-lg) and ηsep (citrate): '- ( =

)

*,!

∙ #!

)

*,$

∙ #$

∙ 100%

(Eq. 3)

257

where cβ-lg, centr and cβ-lg, inlet are concentration of soluble β-lg in centrate and the inlet

258

suspension, respectively, and:

259

+,-./ .01 =

!$

∙ #!

,! !$

,$

∙ #$

∙ 100%

(Eq. 4)

260

where ccitrate, centr and ccitrate, inlet are citrate content in the centrate and the inlet suspension,

261

respectively.

262

Furthermore, flow properties of the sediment in dependence of dry matter content

263

(Section 2.3) were investigated by rheological measurements. For this, a cone-plate geometry

264

(diameter = 50 mm, angle = 2°, gap width = 0.213 mm) and the rheometer MCR302 (Anton

265

Paar, Graz, Austria) equipped with the software Rheoplus/32 v3.61 were used. Shear rate was

266

linearly increased from 1–100 s-1 within 3 min, shear rate of 100 s-1 was held for another

267

minute, followed by a 3 min downwards ramp to 1 s-1. The experimental data of the

10

268

descending ramp was fitted using the Herschel-Bulkley model (Eq. 5) with τ as shear stress, τ0

269

as yield shear stress, k as the consistency index, γ as the shear rate, and n as the flow index.

270

2 = 23 + 56 7

If n < 1 the fluid is shear-thinning, for n > 1 the fluid is shear-thickening, and if n = 1

271 272

(Eq. 5)

and τ0 = 0 it presents Newtonian flow behaviour.

273 274

2.6.

Overall process evaluation

275

For overall evaluation of process success after the washing steps, purity and yield of

276 277

the protein fractions x (α-la or β-lg) were determined based on concentrations determined

278

using RP-HPLC analysis according to Eq. 6 and Eq. 7, respectively. The yield of α-la refers to

279

recovered amount of this protein in the sediment after second washing step, whereas the yield

280

of β-lg refers to the sum of recovered amounts in all three supernatants.

281

89/-.: +;1 =

282

=-0 > +;1 =

< ) * , ,

? ∙ #

?

∙ #

∙ 100%

(Eq. 6)

∙ 100%

(Eq. 7)

The index start refers to the initial precipitate suspension, while the index end refers to

283 284

last fraction obtained, i.e., the sediment after the second washing step for α-la, as well as the

285

centrate of the initial separation step for or β-lg.

286 287

2.7.

Statistical analysis

288 289

In total, two batches of each sample condition were prepared, and each sample was

290

analysed in duplicate. Thus, a four-fold determination was carried out. Data points shown in

291

figures and tables represent the mean values, while the error bars give the standard deviations

292

(n = 4), which were calculated with Microsoft Excel (Microsoft Corporation, Redmond, 11

293

USA). The 95% confidence intervals were calculated by use of Student’s t-distribution. Data

294

were evaluated and plotted using OriginPro 2018b (OriginLab Corporation, Northampton,

295

USA). Additionally, a principal component analysis (PCA) was carried out to evaluate which

296 297

variables convey the most variation in the dataset, i.e., which influencing factors contribute

298

the most in the overall separation. Based on the correlation matrix and the scree diagram a

299

reduction to three PC (with eigenvalue > 1) was conducted, which explained 89.1% of the

300

variance in the data. The resulting PCA biplot is a combination of the loading plot (arrows)

301

and the score plot (dots).

302 303

3.

Results and discussion

304 305

In the following section, results are described for precipitation procedure, centrifugal

306

separation, including initial separation and the consecutive washing steps, as well as analysis

307

of rheological properties. The whole process was conducted using two different whey starting

308

materials, WPC80 and WPI, respectively. The two systems were chosen to compare an

309

“ideal” protein system (WPI) with a more realistic protein system (WPC80), which comprises

310

traces of impurities like lactose, fat, and minerals. In industrial processing, most likely raw

311

whey will be chosen as starting material. Depending on the intensity of pretreatment,

312

including concentration and purification steps, this material may present comparable

313

characteristics as the ones described here. To maintain comparability and to standardise the

314

starting material, we used the above-mentioned whey powders.

315 316

3.1.

Precipitation of α-lactalbumin

317

12

318

For assessing the progress of aggregate formation during heat treatment and for

319

determination of final precipitation yield, samples were regularly taken and analysed by

320

means of RP-HPLC. Fig. 2 depicts the precipitation progress for the fractions XP of α-la and

321

β-lg, respectively. Investigation of precipitation was performed using WPI (Fig. 2a) and

322

WPC80 (Fig. 2b) as starting material. The curves are presented in dependence of time and

323

temperature, starting with the time point, where the desired environmental conditions were

324

established. As the heating-up phase to reach the target temperature of 50 °C took one third of

325

the total heating time, the aggregation progress in this phase is also depicted. It is marked with

326

grey background.

327

As can be seen in Fig. 2a, the main part of α-la precipitation occurred already during

328

the heating-up phase, although temperatures were still far below the target temperature.

329

According to Hendrix, Griko, and Privalov (2000), the transition temperature of holo- to apo-

330

α-la is between 22–34 °C, depending on the environmental conditions. This implies that

331

unfolding of apo-α-la, which is followed by its precipitation, even starts at room temperature.

332

During the heating-up phase, XP already reached 0.88, at the end of the holding time a nearly

333

complete precipitation yield of 0.98 was achieved. Regarding the precipitation kinetic of α-la,

334

the obtained asymptotical shape of the curve was already described by Bramaud et al. (1995)

335

in dependence of denaturation time under acidic conditions.

336

The free content of β-lg initially decreased slightly due to entrapment in aggregates. In

337

the course of the heating time, however, some enclosed β-lg was depleted from the

338

precipitates, which happens presumably through a rearrangement of the α-la precipitated due

339

to stirring. At the end of the heating, an amount of 6% β-lg stayed within the aqueous phase

340

inside the α-la precipitates. A formation of stable α-la-β-lg aggregates seems to be

341

improbable, because the exposition of the free thiol group of β-lg usually starts at

342

temperatures of 60 °C, upon unfolding (De Wit, 2009). Furthermore, in an acidic environment

343

the reactivity of sulfhydryl groups is reduced (Guyomarc’h et al., 2015; Park & Lund, 1984), 13

344

and overall denaturation temperature of β-lg is higher (Dissanayake, Ramchandran, Piyadasa,

345

& Vasiljevic, 2013). The precipitation curve of WPC80 in Fig. 2b presented a similar parabolic shape,

346 347

obtaining a final XP 0.92, with 5% of β-lg remaining in the aggregates after 180 min heating

348

time. However, the curve suggests that the equilibrium is not necessarily reached after this

349

time. The root cause is probably a deceleration of the precipitation due to hindrance effects in

350

the aggregate rearrangement by lactose and fat, which are present in WPC, but not in WPI.

351 352

3.2.

Optimisation of centrifugal conditions

353 354

After precipitation, suspensions obtained from WPC80 were diluted to dry matter

355

contents of 20%, 11%, and 6% (w/w), using cold or pre-warmed DIW at 20 °C or 50 °C,

356

respectively. In the following, the influences of inlet concentration, suspension temperature,

357

and applied g-force were investigated to optimise centrifugal separation of α-la aggregates.

358

The results for the separation efficiencies of centrate are displayed in Fig. 3a, whereas the

359

sediment dry matters obtained are presented in Fig. 3b, as a function of the applied centrifugal

360

acceleration for all suspensions.

361

Relating to Fig. 3a, it can be clearly seen that the separation efficiency increased upon

362

reduction of the inlet dry matter for 50 °C. If a suspension contains less precipitates, the

363

particle-particle interactions are reduced during sedimentation, thus enabling higher median

364

settling velocities. Moreover, the residence time of particles settled on the cake is longer, as

365

there is a lower number of particles coming after. This is accompanied by a longer phase for

366

sediment consolidation, thus enabling a better dewatering of the sediment.

367

Referring to the impact of temperature, the increase in temperature from 20 °C to 50

368

°C had a tremendous effect on separation. On the one hand, this is due to the reduction of the

369

viscosity of the continuous phase, which reduces the uplifting drag force for the particles 14

370

while sedimentation. On the other hand, the higher temperature also might influence the

371

aggregate stabilisation. The apo-α-la form is characterised as the calcium-free variant of this

372

protein, which presents a less dense conformation of the tertiary structure (Hirose, 1993).

373

Moreover, it was demonstrated that the apo-form leads to partly unfolding, thus exposing

374

inner patches of the protein resulting in higher hydrophobicity (Stănciuc, Râpeanu, Bahrim, &

375

Aprodu, 2012). As the system attempts to minimise thermodynamically unfavourable states,

376

non-polar groups establish strong interactions, the hydrophobic interactions. Characteristic for

377

hydrophobic interactions is their tendency to get stronger upon moderate temperature increase

378

up to 60 °C (Bryant & McClements, 1998). The intensified interaction between the protein

379

molecules might, thus, also lead to an intensified exclusion of the polar water molecules,

380

which consequently increases the dewatering capability of the sediment.

381

Regarding the influence of the applied g-force, a nearly linear trend was observed as a

382

function of sediment dry matter (Fig. 3b). A higher acceleration leads to better dewatering of

383

the sludge-like sediments by compression. Thus, the sediment pores are compressed, resulting

384

in drainage of the aqueous phase in direction to the top of the sediment layer. As can also be

385

seen from Fig. 3a, the maximum in separation efficiency ηsep was already achieved at 6500 ×

386

g, and not at the highest applied acceleration. Regarding Stokes’ law, an enhancement of

387

centrifugal acceleration results in a proportional increase in sedimentation velocity,

388

correlating with the separation efficiency. This effect obviously applied up to 6500 × g. A

389

further increase to 10,000 × g, however, showed no enhancement in separation efficiency for

390

all suspensions. As stated by Stahl (2004), disintegrative forces and shear stress on particles

391

are highest when entering the bowl via the feed port. Transported under laminar flow

392

conditions through the feed pipe, the particles are rapidly accelerated by the rotational forces

393

generated by the bowl. Probably, the loosely bound molecules at the outer parts of the

394

aggregates were mainly affected by the fragmentation, whereas the precipitate kernels were

395

more resistant to destruction by shear (Byrne & Fitzpatrick, 2002). Some of these fragments 15

396

were much smaller than the foregoing aggregates so that they were not able to sediment

397

anymore in the applied centrifugal field, and consequently stayed in the supernatant. As the

398

precipitate kernels were not destroyed by acceleration forces, they were able to settle

399

independently from g-force at accelerations above 6500 × g.

400 401

3.3.

Investigation of flow properties of the α-la sediment

402 403

The transportability of a sediment is distinctly affected by its flow behaviour, which

404

depends beside others on the total number of potentially interacting particles. Thus,

405

rheological experiments were performed with α-la aggregate suspension containing dry matter

406

contents between 5–28%, as described in Section 2.5. Experimental results from WPI and

407

WPC80 were fitted according to Eq. 5. The results for the received flow index n (Fig. 4a) and

408

the consistency index k (Fig. 4b) are presented in dependency of dry matter concentration.

409

In Fig. 4a, for concentrations up to 10% (w/w), a Newtonian flow behaviour was

410

observed (n = 1). With increasing concentration, the suspension presented a tendency to shear

411

thinning behaviour (n < 1), which got more prominent at concentrations higher than 25% dry

412

matter. The change to a pseudoplastic flow behaviour at high particle concentrations was

413

additionally reflected by the consistency factor, which showed also a slope dived in three

414

distinct parts.

415

Many other workers investigated the flow properties of native WPC in dependence of

416

protein concentration (Herceg & Lelas, 2005; Hermanson, 1975; Patocka, Cervenkova,

417

Narine, & Jelen, 2006; Tang, Munro, & McCarthy, 1993). All agree on Newtonian behaviour

418

up to 10% (w/w), which matches the results presented here. In the range between 10% < w <

419

20% the flow behaviour was demonstrated to present pseudoplastic flow behaviour at shear

420

rates below 50 s-1. In contrast, at higher shear rates (> 50 s-1), the rheological behaviour

421

appears to be Newtonian (Tang et al., 1993). In this study, the applied shear rate was 100 s-1; 16

422

however, a pseudoplastic behaviour is clearly visible in this concentration range. This

423

observation is considered to be caused by the presence of the α-la aggregates, which provide

424

structural elements in the suspension compared with native whey proteins. For concentrations

425

over 20% a pseudoplastic rheological behaviour is described in literature (Alizadehfard &

426

Wiley,1996), which is also seen in this study with the α-la aggregate suspension. With an

427

increase in the protein content, the respective volume fraction of particles rises as well;

428

consequently, the distance between α-la aggregates is diminished. Additionally, the net charge

429

of α-la is supposed to be on a low level, as the environmental pH is close to its isoelectric

430

point. Thus, the attractive Van der Waals forces may favour an entanglement of the apo-α-la

431

molecules.

432

Comparing the curves from WPI and WPC80, the same trending of the curves can be

433

observed. The values for WPI were comparable with the ones for WPC80 in the Newtonian

434

flow region, but were slightly higher in pseudoplastic region. However, the distinction

435

appears not to be significant. As stated before, the main difference of the two systems is the

436

lactose content in WPC80. Thus, sediments of WPI and WPC80 containing the same value of

437

dry matter might differ in their relative protein concentration. However, lactose generally

438

presents high solubility in water and preferentially remains in the liquid phase, i.e., the

439

supernatant. As only low amounts of lactose stay in the sediment, the influence in diminishing

440

the relative protein content is obviously minor. Furthermore, the non-Newtonian flow

441

behaviour is generally originated from large molecules, e.g., proteins. Thus, it is not

442

surprising that the lactose does not affect the flow behaviour, and the curves from the both

443

systems look similar (Morison & Mackay, 2001).

444

With these data, it was shown that the flow behaviours of WPC80 and WPI have high

445

similarity. Thus, the investigation of different centrifugal conditions using WPC80 appears to

446

be appropriately and transferable to the WPI system. Moreover, a strong dependency on dry

447

matter content was demonstrated. This development of flow behaviour might, on the one 17

448

hand, promote the sediment accumulation on the lowest point of the bowl, thus further

449

improving dewatering. On the other hand, the intense particle entanglement might lead to

450

particle disintegration, when shear forces perpendicular to flow direction are applied.

451 452

3.4.

Separation performance and aggregate stability along the separation process

453 454

Based on the observations in section 3.2 and 3.3, the whole separation process of α-la

455

including two washing steps was performed. In this study, the focus was on evaluating the

456

separation efficiency of all three centrifugation steps and on the observation of aggregate

457

stability. For the further experiments WPI was used as starting material, because this system

458

provides a lower level of impurities (lactose, fat, ash) and allows investigation of the specific

459

protein interactions and behaviour, nearly without unpredictable cross-reactions from other

460

components. The separation was conducted using the aggregate suspension at 50 °C with an

461

inlet dry matter of 6% (w/w) at centrifugal acceleration of 6500 × g. A two-step reslurry

462

washing procedure was chosen to remove other low molecular solutes and soluble proteins

463

like β-lg from the aqueous phase filling the sediment pores. Table 2 presents the results of the

464

separation efficiencies (Eq. 2–4) for initial separation as well as the two subsequent washing

465

steps and the final yields (Eq. 6) and purities (Eq. 7) for the respective protein fractions. In

466

addition to that, the respective chromatograms for the WPI inlet (Fig. 5a), the centrate after

467

the initial separation (Fig. 5b), and the twice washed sediment (Fig. 5c) are presented.

468

After the initial separation step, the centrate consisted of over 99.7% β-lg, mainly in

469

the native state. This means that the α-la precipitates were (almost) completely removed from

470

the supernatant, as can also be seen in Fig. 5b. For the first and second washing step, ηsep for

471

α-la reached 95.0% and 92.6%, respectively. The reduction of ηsep presumably is an

18

472

indication for the fragmentation of precipitated α-la, which led to an increased impurity of the

473

supernatant. In total, 75% of α-la aggregates were recovered throughout the process.

474

Within the initial separation step, nearly 87% of soluble β-lg migrated into the

475

centrate. Vice versa, this means that around 13% of soluble β-lg remained in the sediment

476

fraction, which was enclosed in the aqueous phase of the particle’s pores. After the second

477

washing step, an almost complete absence of soluble β-lg, with only 0.3% remaining in the

478

sediment, was achieved. This refers to a purity of 99.7% of the α-la sediment fraction, which

479

was calculated based on the chromatogram shown in Fig. 5c. Also, in laboratory scale, two

480

washing steps using acidified water at 55 °C as washing medium were shown to be suitable

481

for nearly complete depletion of β-lg (Fernández et al., 2012; Lucena et al., 2006).

482

The removal of citrate, which was initially added to act as calcium sequestrant, was,

483

however, not as successful as the washing out of β-lg. After the initial separation, the

484

sediment comprised 28.3% citrate, which could be reduced to 22.6% and 21.5% with the

485

following washing steps. The reason is that the formed tricalcium citrate shows poor

486

solubility (1 g L-1 at 25 °C) compared with trisodium citrate (425 g L-1 at 25 °C), and

487

therefore, tends to be present in solid state. Due to its high density, it is likely that it is

488

obtained in the sediment. The conclusion from this is that the calcium citrate should be

489

removed by a subsequent ultrafiltration step after the resolubilisation of the α-la molecules.

490

To prove the assumption of particle fragmentation, particle size measurements were

491

conducted. Fig. 6 depicts the particle size sum distribution Q3(x) for the inlets of the initial

492

separation as well as for both washing steps. As can be clearly seen, aggregate sizes were

493

reduced with each separation step. The median particle size of the initial suspension was 10.3

494

µm, the suspension entering the first washing step only had a median size of 0.6 µm, and for

495

the second washing step the particles were smaller than 0.4 µm, respectively. The particle size

496

reduction of the α-la aggregates by 94% in the initial separation step may therefore be

497

explained by fragmentation into the primary aggregates or small fragments. According to the 19

498

model of floc disintegration by Kleine and Stahl (1989), fragmentation generally occurs when

499

the applied acceleration force FC exceeds the individual linkage forces FA. FC is hereby

500

proportional to the product of the density difference ∆ρ, gravitational acceleration g,

501

acceleration factor c, and the cubic of particle size x. In concrete, this means that especially

502

big aggregates are subjected to fragmentation as the size is a cubic factor in Eq. 8.

503

@A ∝ ∆D ( , ; E

(Eq. 8)

The linkage force FA of each particle within a floc arises from the charges on the

504 505

surface of the particle. Here, FA is proportional to a surface specific adhesion force f and the

506

corresponding particle surface area.

507

@F ∝ G ; H

(Eq. 9)

However, also the interparticular shear forces may not be neglected, especially at high

508 509

particle concentrations. Such high concentrations can be found towards the end of their

510

settling way in the centrifuge (region of hindered settling), and of course in the sediment itself

511

(Scales, Johnson, Healy, & Kapur, 1998). Particularly, for soft particles, as the α-la

512

aggregates, the compressive yield stress may be considered as an additional mechanism for

513

particle disruption (Haller & Kulozik, 2019). The particles experience compressive forces due

514

to the pressure from the weight of the upper layers of the sediment, which is multiplied by the

515

centrifugal acceleration number (Green, Eberl, & Landman, 1996). Moreover, the upward

516

drag force of liquid, which is squeezed out of the particle pores generates a counterflow and

517

induces shear forces. Additional shearing in decanter centrifuges is executed by the scroll in

518

horizontal direction. Indeed, fragmentation can be observed in this approach due to various

519

reasons. However, the high acceleration possibilities of this centrifuge still enable adequate

520

separation efficiencies.

521 522

3.5.

Sediment analysis along the centrifugation steps

523 20

524

Within the three separation steps, the α-la sediment was purified from soluble β-lg and

525

citrate, simultaneously changing its composition and visible appearance. In the first separation

526

step, irregularly shaped lumps of sedimented aggregates were observed falling out from the

527

sediment discharge port (Fig. 7a). After the first wash of the sediment, the texture of the

528

separated sediment got visibly smoother (Fig. 7b) until creamy long filaments were obtained

529

after the second wash (Fig. 7c). As listed in Table 3, the total sediment dry matter decreased

530

slightly from 30.5% in initial separation to 26.6% and 26.3% within the washing steps.

531

Additionally, the protein to dry matter ratio increased from 84.8% up to 88.5% and finally to

532

90.0%, respectively. The remaining 10% of dry matter within the α-la sediment can

533

presumably be ascribed to the residues of enclosed solid calcium citrate, as described

534

previously.

535

The reduction of overall achievable dry matter along the washing steps is probably

536

related to the particle breakage. During the production of the precipitate suspension, the apo-

537

α-la molecules formed oligomers due to attractive forces of their hydrophobic regions.

538

Through the long heating duration and the gentle stirring, strong and particulate aggregates

539

arranged (Byrne & Fitzpatrick, 2002). These aggregates were thermodynamically stable as the

540

non-polar amino acid residues conglomerated, thus excluding the polar water molecules

541

(Byrant & McClements, 1998). From a superior view, this equals a system that has a low

542

amount of interior bound water. Upon breakage of the precipitates, an increase of the surface

543

to volume ratio of the aggregates occurs. Thus, a larger surface area is available for hydration,

544

as proteins generally have a high solubility in water.

545

The observed change of the sediment texture from clumpy to filamentous shall be

546

discussed in the following. The arrangement of single protein molecules to particulate

547

aggregates is a highly complex process, subjected to collision mechanisms, e.g. as described

548

in Smoluchowski (1918). In this regard, it appears to be improbable that particle fragments re-

549

arrange into globular aggregates during the reslurry washing steps. As the interior 21

550

hydrophobic regions of the apo-α-la are still exposed and “reactive”, it is likely that they will

551

resume the attractive interactions with hydrophobic patches from other apo-α-la molecules.

552

There is a high probability that the molecules will adopt new structures, which are easier to

553

achieve than particulates, e.g., filaments (Pelegrine & Gasparetto, 2005). As the

554

environmental conditions are close to the IEP of α-la, the low level of electrostatic repulsion,

555

on the on hand, and the increased importance of the van-der-Waals forces, on the other hand,

556

might also act in a supportive manner in this case.

557 558

3.6.

Overall evaluation

559

For overall evaluation of the data, a PCA was performed that aimed at emphasising the

560 561

main influencing factors and also at discovering correlations of the separation variables. The

562

PCA is a statistical tool that reduces the number of dimensionalities, thus transforming a large

563

set of variables into a smaller one that still contains most of the information. The received

564

biplot comprises a loading diagram (blue arrows) and a score plot (black dots), and is given in

565

Fig. 8.

566

The PC1 is given on the x-axis, the PC 2 on the y-axis, with percentages of 42.0% and

567

29.4% of the data variance, respectively. In this standardised diagram, the length of the

568

loading vectors, the arrows, is a direct measurement on how big their influence is on the PCs

569

(–1 = maximal negative influence, 0 = no influence, +1 = maximal influence). Here, it is

570

obvious that nearly all characteristics have a severe influence on the PCs, except the d50,3,

571

which is the median particle size in suspension, that has only little influence, followed by the

572

applied g-force, which is the second shortest vector.

573

Moreover, the angle between a loading vector and the x- or y-axis also represent the

574

degree to which PC1 or PC2 are influenced by this vector. This relationship additionally

575

provokes that the correlation of one variable with another is described by the angle between 22

576

them (0° < 90° positive correlation, 90° independent, 90° > 180° negative correlation).

577

Looking on the given vectors, a negative correlation with an angle close to 180° can be seen

578

for particle size and g-force. Basically, this means that a higher g-force is able to compensate

579

a reduction in particle size to keep the acceleration force (or separation result) constant, i.e.,

580

particle size and g-force are indirectly proportional. This relationship was already given in Eq.

581

8, which described the proportionality of the acceleration forces. Thus, the loading diagram

582

emphasises which factors correlate either positively or negatively with the separation

583

efficiency.

584

A strong positive correlation with the separation efficiency is seen for the influencing

585

factor temperature, but also for residence time and g-force. Basically, one could classify these

586

variables as process-related influencing factors. A negative correlation for separation

587

efficiency is seen for the arrows for particle size described as d50,3, lactose, and dry matter

588

content, as the intermediate angle is over 90°. These three variables can be generalised as

589

suspension-related influencing factors. These relationships suggest that process-related

590

variables, i.e., higher temperature, extended residence time, beneficially influence the

591

separation efficiency. While suspension-related variables, i.e., higher lactose content, more

592

inlet dry matter, impede the separation efficiency. It seems astonishing that the d50,3 is

593

assigned to negatively correlate to the separation efficiency. However, as stated already

594

above, due to the short absolute length of this arrow, its influence is minor. Apart from that,

595

bigger particles in the inlet suspension are more prone to disintegration during centrifugation

596

than already fragmented particles, as discussed in section 3.4. A higher occurrence of

597

fragmentation during a centrifugation step might thereby impair the overall separation

598

efficiency of this step.

599

Regarding the data points, two distinct clusters are visible, one comprising all values

600

from separations performed at 20 °C, and one with the data points from 50 °C. Within the 50

601

°C separation temperature cluster, a smaller one is highlighted that comprises the data points 23

602

from all the washing steps. Interestingly, no different behaviours of WPI (stars) and WPC80

603

(dots) are observed as the data points lie close to each other and are not separated. This means

604

that the aggregates of both starting material follow the same dependencies in separation and

605

the differences due to lactose content are negligible. From this diagram, one can conclude that

606

the temperature is one of the main influencing factors.

607 608

4.

Conclusion

609 610

In this study a successful process for the separation of the two major whey proteins, α-

611

la and β-lg, was demonstrated that has high suitability for industrial utilisation. A lean process

612

design is described that comprises high throughput unit operations, where further scale-up

613

appears to be simple. Moreover, we achieved two high purity fractions, with both proteins in

614

native state, with respectable yields. Further to that, for this approach only components with

615

the GRAS (generally regarded as safe) status were used. Due to these advantages, we consider

616

this process to be superior to many others suggested in literature.

617

Nonetheless, the major challenges of this fractionation method shall be outlined,

618

which needed special attention in this study. The selective thermal precipitation of α-la under

619

acidic conditions is a well described, state-of-the-art pre-treatment for whey proteins.

620

However, the generated α-la precipitates are small, fragile particles that are prone to shear

621

forces. Indeed, analysis presented a significant particle size reduction from initially 10 µm to

622

0.4 µm after three consecutive centrifugation steps. Nonetheless, separation efficiencies just

623

declined from 99.7% to 92.6% over the whole process. Moreover, a pseudoplastic flow

624

behaviour of the α-la sediment was demonstrated being strongly dependent on dry matter.

625

Generally, such characteristics might cause discharge issues due to sediment backflow to the

626

lowest point of the bowl or adhesion to the scroll flights. However, in this study a constant

627

discharge with highly dewatered sediment was observed. In summary, the suitability of the 24

628

herein used decanter centrifuge for this application was proven, achieving purities for α-la and

629

β-lg fraction of over 99%.

630

However, there is further potential for improvement of this process, thus making it

631

even more time- and cost effective. On the one hand, the time span chosen for powder

632

hydration was generously long. Re-hydration of proteins might be completed after several

633

minutes, depending on the powder used. Alternatively, also liquid raw whey can be

634

considered as starting material, thus eliminating the whole re-hydration step. On the other

635

hand, the heating process has potential for improvement by using steeper heating rates in a

636

tubular counter-current heat exchanger. Potentially, also the heat holding phase could be

637

reduced by applying higher temperatures; however, the thermal stabilities of the other proteins

638

need to be respected.

639

In the study, two different starting materials, WPI and WPC80 were used. Generally,

640

both systems showed promising results and had comparable values in rheological analysis.

641

However, separation efficiencies were higher using the purer WPI system. Presumably the α-

642

la precipitates originated from WPC80 have lower stability while centrifugation, resulting in

643

hardly separable particle fragments. As the main difference of the composition of the two

644

whey powders is the content of lactose and fat, we assume that these components are

645

responsible for impairing the aggregate strength in WPC80 batches.

646 647

Acknowledgement

648 649

Flottweg SE is gratefully acknowledged for providing the Sedicanter® centrifuge.

650

Especially we want to thank Sebastian Gillig and Stefan Ecker from Flottweg for technical

651

support. We further thank Annette Brümmer-Rolf and Rupert Wirnharter for experimental

652

support.

25

653

This IGF Project of the FEI was supported via AiF (AiF 18126 N) within the

654

programme for promoting the Industrial Collective Research (IGF) of the German Ministry of

655

Economic Affairs and Energy (BMWi), based on a resolution of the German Parliament.

656 657

References

658 659 660

Alizadehfard, M. R., & Wiley, D. E. (1996). Non-Newtonian behaviour of whey protein solutions. Journal of Dairy Research, 63, 315–320.

661

Alomirah, H. F., & Alli, I. (2004). Separation and characterisation of β-lactoglobulin and α-

662

lactalbumin from whey and whey protein preparations. International Dairy Journal,

663

14, 411–419.

664

Bhattacharjee, S., Ghosh, S., Datta, S., & Bhattacharjee, C. (2006). Studies on ultrafiltration

665

of casein whey using a rotating disk module: effects of pH and membrane disk

666

rotation. Desalination, 195, 95–108.

667

Bramaud, C., Aimar, P., & Daufin, G. (1995). Thermal isoelectric precipitation of α‐

668

lactalbumin from a whey protein concentrate: Influence of protein–calcium

669

complexation. Biotechnology and Bioengineering, 47, 121–130.

670 671 672

Bramaud, C., Aimar, P., & Daufin, G. (1997a). Optimisation of a whey protein fractionation process based on the selective precipitation of α-lactalbumin. Lait, 77, 411–423. Bramaud, C., Aimar, P., & Daufin, G. (1997b). Whey protein fractionation: Isoelectric

673

precipitation of α-lactalbumin under gentle heat treatment. Biotechnology and

674

Bioengineering, 56, 391–397.

675

Bryant, C. M., & McClements, D. J. (1998). Molecular basis of protein functionality with

676

special consideration of cold-set gels derived from heat-denatured whey. Trends in

677

Food Science & Technology, 9, 143–151.

26

678

Byrne, E. P., & Fitzpatrick, J. J. (2002). Investigation of how agitation during precipitation,

679

and subsequent processing affects the particle size distribution and separation of α-

680

lactalbumin enriched whey protein precipitates. Biochemical Engineering Journal, 10,

681

17–25.

682

Chatterton, D. E., Smithers, G., Roupas, P., & Brodkorb, A. (2006). Bioactivity of β-

683

lactoglobulin and α-lactalbumin—Technological implications for processing.

684

International Dairy Journal, 16, 1229–1240.

685 686 687 688 689 690 691 692 693

Chen, L., Guo, C., Guan, Y., & Liu, H. (2007). Isolation of lactoferrin from acid whey by magnetic affinity separation. Separation and Purification Technology, 56, 168–174. Das, K. P., & Kinsella, J. E. (1989). pH dependent emulsifying properties of β-lactoglobulin. Journal of Dispersion Science and Technology, 10, 77–102. De Wit, J. N. (1990). Thermal stability and functionality of whey proteins. Journal of Dairy Science, 73, 3602–3612. De Wit, J. N. (2009). Thermal behaviour of bovine β-lactoglobulin at temperatures up to 150 C. A review. Trends in Food Science & Technology, 20, 27–34. Dissanayake, M., Ramchandran, L., Piyadasa, C., & Vasiljevic, T. (2013). Influence of heat

694

and pH on structure and conformation of whey proteins. International Dairy Journal,

695

28, 56–61.

696

Dombrowski, J., Gschwendtner, M., & Kulozik, U. (2017). Evaluation of structural

697

characteristics determining surface and foaming properties of β-lactoglobulin

698

aggregates. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 516,

699

286–295.

700

Dumpler, J., Wohlschläger, H., & Kulozik, U. (2017). Dissociation and coagulation of caseins

701

and whey proteins in concentrated skim milk heated by direct steam injection. Dairy

702

Science & Technology, 96, 807–826.

27

703 704 705

Etzel, M. R. (2004). Manufacture and use of dairy protein fractions. Journal of Nutrition, 134, 996S–1002S. Fernández, A., Menéndez, V., & Riera, F. A. (2012). α-Lactalbumin solubilisation from a

706

precipitated whey protein concentrates fraction: pH and calcium concentration effects.

707

International Journal of Food Science & Technology, 47, 467–474.

708

Gésan-Guiziou, G., Daufin, G., Timmer, M., Allersma, D., & Van der Horst, C. (1999).

709

Process steps for the preparation of purified fractions of α-lactalbumin and β-

710

lactoglobulin from whey protein concentrates. Journal of Dairy Research, 66, 225–

711

236.

712

Guyomarc’h, F., Famelart, M. H., Henry, G., Gulzar, M., Leonil, J., Hamon, P., et al. (2015).

713

Current ways to modify the structure of whey proteins for specific functionalities—a

714

review. Dairy Science & Technology, 95, 795–814.

715

Green, M. D., Eberl, M., & Landman, K. A. (1996). Compressive yield stress of flocculated

716

suspensions: Determination via experiment. American Institute of Chemical Engineers

717

Journal, 42, 2308–2318.

718

Håkansson, A., Svensson, M., Mossberg, A., Sabharwal, H., Linse, S., Lazou, I., et al. (2000).

719

A folding variant of α‐lactalbumin with bactericidal activity against Streptococcus

720

pneumoniae. Molecular Microbiology, 35, 589–600.

721

Haller, N., & Kulozik, U. (2019). Separation of whey protein aggregates by means of

722

continuous centrifugation. Food and Bioprocess Technology, 12, 1052–1067.

723

Heidebrecht, H. J., & Kulozik, U. (2019). Fractionation of casein micelles and minor proteins

724

by microfiltration in diafiltration mode. Study of the transmission and yield of the

725

immunoglobulins IgG, IgA and IgM. International Dairy Journal, 93, 1–10.

726

Heine, W. E., Klein, P. D., & Reeds, P. J. (1991). The importance of α-lactalbumin in infant

727

nutrition. Journal of Nutrition, 121, 277–283.

28

728

Heine, W., Radke, M., Wutzke, K. D., Peters, E., & Kundt, G. (1996). α-Lactalbumin‐

729

enriched low‐protein infant formulas: A comparison to breast milk feeding. Acta

730

Paediatrica, 85, 1024–1028.

731

Hendrix, T., Griko, Y. V., & Privalov, P. L. (2000). A calorimetric study of the influence of

732

calcium on the stability of bovine α-lactalbumin. Biophysical Chemistry, 84, 27–34.

733

Herceg, Z., & Lelas, V. (2005). The influence of temperature and solid matter content on the

734

viscosity of whey protein concentrates and skim milk powder before and after

735

tribomechanical treatment. Journal of Food Engineering, 66, 433–438.

736 737 738 739 740

Hermansson, A. M. (1975). Functional properties of proteins for foods‐flow properties. Journal of Texture Studies, 5, 425–439. Hirose, M. (1993). Molten globule state of food proteins. Trends in Food Science & Technology, 4, 48–51. Jiang, B., Wang, L., Na, J., Zhang, X., Yuan, Y., Liu, C., et al. (2018). Environmentally-

741

friendly strategy for separation of α-lactalbumin from whey by aqueous two phase

742

flotation. Arabian Journal of Chemistry, 2018, Article 13.

743

Jiang, B., Na, J., Wang, L., Li, D., Liu, C., & Feng, Z. (2019). Separation and enrichment of

744

antioxidant peptides from whey protein isolate hydrolysate by aqueous two-phase

745

extraction and aqueous two-phase flotation. Foods, 8, Article 34.

746

Jost, R., Maire, J. C., Maynard, F., & Secretin, M. C. (1999). Aspects of whey protein usage

747

in infant nutrition, a brief review. International Journal of Food Science &

748

Technology, 34, 533–542.

749

Kalaivani, S., & Regupathi, I. (2013). Partitioning studies of α-lactalbumin in environmental

750

friendly poly (ethylene glycol)—citrate salt aqueous two phase systems. Bioprocess

751

and Biosystems Engineering, 36, 1475–1483.

752 753

Kleine, U., & Stahl, W. (1989). Model concept of floc disintegration in centrifugal fields. Chemical Engineering & Technology, 12, 200–204. 29

754

Lien, E. L., Davis, A. M., Euler, A. R., & Group, M. S. (2004). Growth and safety in term

755

infants fed reduced-protein formula with added bovine α-lactalbumin. Journal of

756

Pediatric Gastroenterology and Nutrition, 38, 170–176.

757

Lucena, E. M., Alvarez, S., Menéndez, C., Riera, F. A., & Alvarez, R. (2006). β-

758

Lactoglobulin removal from whey protein concentrates. Separation and Purification

759

Technology, 52, 310–316.

760

Marciniak, A., Suwal, S., Brisson, G., Britten, M., Pouliot, Y., & Doyen, A. (2019).

761

Evaluation of casein as a binding ligand protein for purification of α-lactalbumin from

762

β-lactoglobulin under high hydrostatic pressure. Food Chemistry, 275, 193–196.

763 764 765 766 767 768 769

Marshall, K. (2004). Therapeutic applications of whey protein. Alternative Medicine Review, 9, 136–156. Maté, J. I., & Krochta, J. M. (1994). β-Lactoglobulin separation from whey protein isolate on a large scale. Journal of Food Science, 59, 1111–1114. Maubois, J. L., Pierre, A., Fauquant, J., & Piot, M. (1987). Industrial fractionation of main whey proteins. International Dairy Federation Bulletin, 212, 154–159. Meyer, A., Berensmeier, S., & Franzreb, M. (2007). Direct capture of lactoferrin from whey

770

using magnetic micro-ion exchangers in combination with high-gradient magnetic

771

separation. Reactive and Functional Polymers, 67, 1577–1588.

772 773

Morison, K. R., & Mackay, F. M. (2001). Viscosity of lactose and whey protein solutions. International Journal of Food Properties, 4, 441–454.

774

Mulvihill, D. M., & Kinsella, J. E. (1988). Gelation of β-lactoglobulin: Effects of sodium

775

chloride and calcium chloride on the rheological and structural properties of gels.

776

Journal of Food Science, 53, 231–236.

777

Ng, T. B., Cheung, R. C. F., Wong, J. H., Wang, Y., Ip, D. T. M., Wan, D. C. C., et al. (2015).

778

Antiviral activities of whey proteins. Applied Microbiology and Biotechnology, 99,

779

6997–7008. 30

780 781 782

Park, K. H., & Lund, D. B. (1984). Calorimetric study of thermal denaturation of βlactoglobulin. Journal of Dairy Science, 67, 1699-1706. Patocka, G., Cervenkova, R., Narine, S., & Jelen, P. (2006). Rheological behaviour of dairy

783

products as affected by soluble whey protein isolate. International Dairy Journal, 16,

784

399–405.

785 786 787 788

Pearce, R. J. (1983). Thermal separation of β-lactoglobulin and α-lactalbumin in bovine cheddar cheese whey. Australian Journal of Dairy Technology, 38, 144–149. Pearce, R. J. (1987). Fractionation of whey proteins. Australian Journal of Dairy Technology, 42, 75–78.

789

Pellegrini, A., Thomas, U., Bramaz, N., Hunziker, P., & Fellenberg, R. von (1999). Isolation

790

and identification of three bactericidal domains in the bovine α-lactalbumin molecule.

791

Biochimica et Biophysica Acta -General Subjects, 1426, 439–448.

792

Pelegrine, D. H. G., & Gasparetto, C. A. (2005). Whey proteins solubility as function of

793

temperature and pH. LWT-Food Science and Technology, 38, 77–80.

794

Permyakov, E. A., & Berliner, L. J. (2000). α-Lactalbumin: Structure and function.

795 796 797 798

Federation of European Biochemical Societies Letters, 473, 269–274. Pierre, A., & Fauquant, J. (1986). Principes pour un procédé industriel de fractionnement des protéines du lactosérum. Lait, 66, 405–419. Rammer, P., Groth-Pedersen, L., Kirkegaard, T., Daugaard, M., Rytter, A., Szyniarowski, P.,

799

et al. (2010). BAMLET activates a lysosomal cell death program in cancer cells.

800

Molecular Cancer Therapeutics, 9, 24–32.

801

Santos, M. J., Teixeira, J. A., & Rodrigues, L. R. (2012). Fractionation of the major whey

802

proteins and isolation of β-lactoglobulin variants by anion exchange chromatography.

803

Separation and Purification Technology, 90, 133–139.

31

804

Scales, P. J., Johnson, S. B., Healy, T. W., & Kapur, P. C. (1998). Shear yield stress of

805

partially flocculated colloidal suspensions. American Institute of Chemical Engineers

806

Journal, 44, 538–544.

807

Schmitz-Schug, I. (2014). Improving the nutritional quality of dairy powders – analysing and

808

modelling lysine loss during spray drying as influenced by drying kinetics, thermal

809

stress, physical state and molecular mobility. PhD thesis. Technical University

810

Munich, Freising, Germany.

811 812 813 814 815 816 817 818 819

Smithers, G. W. (2008). Whey and whey proteins—from ‘gutter-to-gold’. International Dairy Journal, 18, 695–704. Smoluchowski, M. V. (1918). Versuch einer mathematischen Theorie der Koagulationskinetik kolloider Lösungen. Zeitschrift für Physikalische Chemie, 92, 129–168. Stahl, W. H. (2004). Industrie-Zentrifugen, Fest-Flüssig-Trennung, Band II, Maschinen-& Verfahrenstechnik (1st edn.). Männedorf, Switzerland: DrM Press Stănciuc, N., Râpeanu, G., Bahrim, G., & Aprodu, I. (2012). pH and heat-induced structural changes of bovine apo-α-lactalbumin. Food Chemistry, 131, 956–963. Tang, Q., Munro, P., & McCarthy, O. (1993). Rheology of whey protein concentrate solutions

820

as a function of concentration, temperature, pH and salt concentration. Journal of

821

Dairy Research, 60, 349–361.

822

Tolkach, A., & Kulozik, U. (2008). Kinetic modeling of the thermal denaturation of bovine α-

823

lactalbumin in an acidic pH-region and in the presence of a calcium complexing agent.

824

Chemie Ingenieur Technik, 80, 1165–1173.

825

Toro-Sierra, J., Tolkach, A., & Kulozik, U. (2013). Fractionation of α-lactalbumin and β-

826

lactoglobulin from whey protein isolate using selective thermal aggregation, an

827

optimized membrane separation procedure and resolubilization techniques at pilot

828

plant Scale. Food and Bioprocess Technology, 6, 1032–1043.

32

829

Trabulsi, J., Capeding, R., Lebumfacil, J., Ramanujam, K., Feng, P., McSweeney, S., et al.

830

(2011). Effect of an α-lactalbumin-enriched infant formula with lower protein on

831

growth. European Journal of Clinical Nutrition, 65, 167–174.

33

1

Figure legends

2 3

Fig. 1. Schematic drawing of the Sedicanter® (modified image, permission for use given by

4

Flottweg, Vilsbiburg, Germany)

5 6

Fig. 2. Precipitation yield, Xp, for α-la ( ) and β-lg ( ) as a function of time and temperature

7

using WPI (a) or WPC80 (b) as starting material.

8 9

Fig. 3. Separation efficiency of α-la precipitates (a) and dry matter of obtained sediments (b)

10

at 20 °C at 11% inlet dry matter ( ) and at 50 °C at 6% ( ), 11% ( ) and 20% ( ) inlet dry

11

matter (w/w).

12 13

Fig. 4. Flow index n (a) and consistency factor k (b) of precipitated α-la in dependency of dry

14

matter fraction of sediment for the two starting materials WPC80 ( ) and WPI ( ).

15 16

Fig. 5. Chromatograms from initial inlet (a), centrate after initial separation (b), and sediment

17

after second washing step (c), using WPI as starting material.

18 19

Fig. 6. Particle sizes of α-la precipitate suspension of initial separation ( ), first wash (),

20

and second wash ( ) displayed as sum distribution, using WPI as starting material.

21 22

Fig. 7. Discharged α-la sediment after initial centrifugal separation (a), after the first washing

23

step (b) and after the second washing step (c), using WPI as starting material.

24 25

Fig. 8. PCA biplot presenting the loading plot of all analysed variables (blue arrows) and the

26

score plot ( , WPC80;

, WPI), which reflect the measured data points.

Table 1 Machine conditions and decanter settings used in this study. Parameter

Value

Bowl diameter (mm)

300 (max.)

Clarification length (mm)

560

Bowl speed (min-1)

5531–7735

Centrifugal acceleration (× g)

5000–10,000

Differential speed (rpm)

5.4–6.4

Weir height (mm)

126–132

Feed rate (L h-1)

150

Table 2 Summary of achieved separation efficiencies along the separation and washing steps, with additional yield and purity of target protein fractions. a Parameter

Centrate

Sediment

Precipitated α-la (%)

Soluble β-lg (%)

Citrate (%)

ηSep (initial separation)

99.7 ± 0.1

86.7 ± 0.9

71.7 ± 2.4

ηSep (wash 1)

95.0 ± 0.2

85.5 ± 5.4

20.1 ± 4.2

ηSep (wash 2)

92.6 ± 1.0

68.3 ± 10.0

4.8 ± 3.4

Final yield

75.2

83.5

n.a.

Final purity

99.4*

99.7**

n.a.

a

Asterisks indicate (*) fraction after resolubilisation (pH adjustment to 5.1 and removal of

denatured β-lg) and (**) supernatant of initial separation step; n.a., not applicable.

Table 3 Results obtained for respective sediments along the three separation steps. Separation step

Dry matter content (%, w/w)

Protein to dry matter ratio (%, w/w)

Initial separation

30.5 ± 0.7

84.8 ± 0.1

Wash 1

26.6 ± 0.1

88.5 ± 1.5

Wash 2

26.3 ± 0.4

90.0 ± 2.5

Figure 1

Figure 2

, (g)

Figure 3

(g)

,

Figure 4

,

Figure 5

Figure 6

Figure 7

Figure 8