International Journal of Food Science and Technology (1989) 24, 17-27

Protein extraction from lupin seeds: a mathematical

model
J . M. AGUILERA* AND HILDA D. GARCIA
Summary
The effect of particle size and microstructural modification on protein extraction from
defatted lupin was studied. Particles of different sizes (128-1425 p,m) from untreated
(control), flaked and exploded samples were extracted at pH 8.0-8.5 and different
temperatures. A three-parameter kinetic model as well as a diffusional model were
fitted to data. The yield of protein from the surface of particles, and the total extracted
protein yield increased as particle size decreased, while the first order rate constant
increased slightly or remained constant. Diffusion coefficients varied between 0.5 and
4 . 5 ~ m2 s-l and were higher in exploded and flaked material than in the control.
The energy of activation for diffusion was 43 kJ mol-'. Differences in microstructure
were studied by scanning electron microscopy.
Keywords
Diffusion, extraction modelling, Lupinus albus cv. Multolupa, microstructure, protein
extraction dynamics.
Introduction
Vegetable protein isolates are refined food ingredients (over 90% protein) manufac-
tured by solubilization and extraction of storage protein from ground cotyledon par-
ticles at alkaline pH, followed by isoelectric precipitation, concentration and spray
drying.
1 entry of the solvent into the particle;
2 redistribution of solvent in cell compartments and expansion of the solid matrix;
3 solubilization andlor degradation of components;
4 transport of the solute to the exterior of the particle; and,
5 migration of the extracted solute from the surface of the particle into the bulk
The rate of protein extraction is usually controlled by processes occurring in the interior
of the particle, rather than by external factors (Schwartzberg & Chao, 1982), which
makes microstructural modifications important. In practice size reduction is used to
obtain high extraction rates (Coats & Wingard, 1950; Clo & Voilley, 1983); but this
results in high energy expenditure during milling, and problems in downstream separa-
tion of the protein solution from a fine residue.
Authors' address: Department of Chemical Engineering, Universidad Catolica de Chile, PO Box 6177,
Santiago, Chile.
Torrespondent.
Extraction is a complex process involving:
solution.
18
J . M. Aguilera and Hilda D. Garcia
A basic understanding of the effect of particle size and microstructure is desirable
for designing efficient processes. Several models for solute extraction from particles
have been proposed (Schwartzberg, 1975; Spiro & Siddique, 1981; Spiro & Selwood,
1984), but microstructural aspects have only been related recently to mechanisms of
extraction (Aguilera et al . , 1987).
Lupin, a legume with overall chemical composition of seed similar to that of soy-
bean, is now being processed to remove toxic alkaloids and produce improved protein
products and oil (Aguilera & Trier, 1978; Aguilera, Gerngross & Lusas, 1983; Hill,
1986). The object of this work was to study and model the effect of particle size and
microsctructural modification on the rate of protein extraction.
Materials and methods
Raw material
Sweet lupin seeds (Lupinus albus cv. Multolupa), obtained from Campex Baer
(Temuco Chile), were harvested in 1985. The alkaloid content was approximately
0.02-0.03%. Whole seeds were ground into relatively large pieces or grits in a Bauer
disk mill (Bauer Bros, Springfield, OH U. S. A. ) and dehulled in a zigzag air classifier
(I NTEC Santiago Chile). Dehulled material was defatted with hexane (1 : 4, solid :
solvent ratio) in a soxhlet extractor and solvent removed at 3TC, giving a final oil con-
tent of 1.7%.
Pretreatments
Three samples were prepared for study of the effect of microstructure on extraction:
Flaked material was made by compression of defatted grits in a flaking mill (INTEC) to
a mean thickness of 500 pm. Exploded particles were first tempered with a water spray
to 26% moisture content and equilibrated for 24 h. Explosion was achieved by depres-
surization from 3.6 MPa to atmospheric pressure in a hydrogenation reactor (Autoclave
Engineering Erie PA U. S. A. ) in less than 1 s. The control sample was the defatted
material.
All samples were then ground in a laboratory hammer mill (Adam Baurniiller
Marktredwitch F.R.G.), and classified in U.S. standard sieves, on a shaker, into the
following fractions which passed (+), but were retained (-) on, screens identified by
their mesh size. The average particle size (km) is shown in parenthesis: +10 -20 (1425);
+20 -30 (725); +30 -40 (513); +40 -60 (338); +60 -100 (200); +lo0 -140 (128).
Protein extraction
Protein extraction was carried out at 25C on 18-g samples (7% moisture) by adding
distilled water (400 g) to give a solid : water ratio 1 : 25, in a 1 litre double-wall cylindrical
glass reactor with funnel-shaped bottom. Sampling was done at predetermined intervals
through a stopcock valve at the bottom, and fines were retained in a metal screen. Agi-
tation was provided by a stirrer. Temperature was controlled by circulating water from a
thermostatted bath through the jacket of the glass reactor. pH was maintained at
8.0-8.5 by addition of NaOH (0.1 M). The protein content of the solution was moni-
tored by the Biuret method (Cooper, 1977). Protein was extracted from control ma-
terial at 25,35,45 and 55"C, to determine an activation energy for the process.
Degree of hydration of particles
Particles of different sizes were soaked in water for 1 h at 20C and the amount of
Modelling extraction of lupin protein 19
water absorbed was determined by oven drying (20C, 1 h). Increase in particle size
due to hydration was determined by measurement under a stereomicroscope (Wild M3,
Heerbrugg F.R.G.).
Kinetic model
two extraction phases:
1 instantaneous removal of superficial protein; and,
2 time-dependent extraction of intracellular, soluble protein.
It was assumed that the washed superficial protein was contained in a one-cell outer
layer, representing the external broken cells of a spherical particle. For extraction of the
intracellular, soluble protein the following first order kinetic expression was proposed
(Aguilera et al., 1987):
The mathematical model applied to actual batch extractions needed to consider only
where C(t ) is the concentration of protein in the extract at any time t, C, is the concent-
ration after initial washing (at time 0), C, is a final concentration (hypothetical) in the
extract, and k is a rate constant. To determine these three parameters (Cffi, C, and k ) ,
data were fitted using the least squares method and solved by the multiple-variable
Newton algorithm in a VAX 8600 computer.
Estimation of a diffusion coefjicient
Effective coefficients are usually used in extraction to characterize mass transfer
inside a particle. Thus, a diffusion model was postulated to study the effect of treatment
and particle size on diffusivity. Microscopic observation of particles before and after
hydration showed an average 30% increase in size. Hence, the actual particle was
assumed to be a hydrated sphere with a diameter 1.3 times the dry particle size, minus
30 Krn corresponding to twice the half-thickness of the outer cell layer depleted by
washing. The model was solved by superposition of solutions to Fick's second law for
two cases:
1 constant surface concentration and initial uniform concentration; and
2 variable surface concentration and initial concentration inside the sphere equal to
Overall boundary conditions were:
1 Uniform initial concentration inside the sphere:
zero (Crank, 1975).
Ci n(r, 0) = COfor r < R
where R is the radius of the sphere and Ci, refers to protein concentration inside the
sphere.
2 Concentration at the particle-solution interface is equal to the instantaneous con-
centration in the bulk solution [ C( t ) ] :
c~,(R, t ) = C,-(C,-C,)e-k'
3 Finite concentration at the centre of the sphere:
Cin(O, t ) = finite.
20 1. M. Aguilera and Hilda D. Garcia
A diffusion coefficient was calculated for each time t from the following relationship:
where Vsol is the volume of the protein solution and A is the area available for mass
transfer, which depends on particle size. An overall process diffusion coefficient was
determined by averaging the individual values.
Microscopy
Samples of raw, control, flaked and exploded lupin particles were dried, mounted
on aluminium stubs, coated with platinum and observed using a scanning electron mic-
roscope (J EOL JSM-25 SII Tokyo J apan) at an accelerating voltage of 10 kV.
Results and discussion
Preliminary batch extractions showed that stirring speeds above 450 rpm had little effect
on the protein concentration in solution, confirming that the overall extraction rate was
not controlled by the external resistance. A stirrer speed of 1200 rpm was selected so
that particles were kept in suspension, thus facilitating sampling from the bottom part
of the reactor.
Dehulled, defatted lupin contained 52.4% protein and 1.7% lipids, on dry weight
basis. A protein balance, considering that 95% of the protein was soluble at the pH of
the experiment (Fan & Sosulski, 1974), gave a-maximum theoretical concentration in
solution of 21 mg protein ml-', which was slightly lower than the 23-24 mg ml-'
obtained for the smallest particle size after 2 h of extraction. The difference may arise
from differences in analytical methods for protein determination (Kjeldhal for total
protein and Biuret for extracts).
Extraction is a complex process which initially involves counterdiffusion of water
and protein, and changes in particle size due to hydration. After 2 h of extraction, the
average amount of water absorbed by the particles was 5.49 g 8-l dry matter, and the
average wet diameter was about 30% larger than the diameter of the dry particle.
Results from extraction experiments at 25C for control, flaked and exploded mate-
rials are shown in Figs l a, l b and lc, respectively. Inspection of Fig. 1 reveals that as
particle size increased:
1 the amount of washed protein (e.g. at time 1 min) decreased;
2 the overall extraction rate was hindered;
3 the extent of protein extraction after 2 h was reduced; and,
4 almost 90% of the protein extracted after 2 h was in solution in the first 20 min.
Ruiz & Hove (1976) reported that extraction yields of lupin protein increased from
55 to 96% as particle size decreased from 846 pm to less than 150 pm, which coincides
with the data for control. A similar effect of particle size has been found by Voilley &
Simatos (1980) for coffee and Aguilera et al. (1987) for red peppers. Russel & Tsao
(1982), however, studied extraction of zein for times up to 16 h, showing that over long
periods, various particles gave the same extent of extraction; this may also be the case
here, since after 2 h protein was still being extruded. This suggests that for large particle
sizes (e.g. > 200 pm for control) there is a third extraction period characterized by very
slow rate; this is not important in a practical extraction process.
Modelling extraction of lupin protein
( a ) Control
25
-
-
!-
: 20
E
1
5 I 5
e
8 10
5
5 5
e
0
+
+
c
+
a
Time (min)
21
Flaked
Time (min)
Exploded
L
20
E
.s I 5
g 10
-
c
e
+
a,
0
-: + 5
e
a
0
0 20 40 60 80 100 I20
Time (min)
Figure 1. Relationship between concentration of dissolved protein and time for the extraction
of defatted lupin seeds of different particle size. Solid: water ratio, 1: 25; temperature, 25C. (a)
Control; (b) flaked; and (c) exploded. (0) 1425 pm; (0) 725 pm; (M) 513 pm; ( 0) 338 Fm;
(A) 200 pm; ( A) 128 pm. Solid lines represent model predictions.
22
J. M. Aguilera and Hilda D. Garcia
Modelling extraction of lupin protein
23
Major physical differences between various particles were the diffusion length and
the microstructure. Fig. 2 shows microstructural characteristics of raw lupin, control,
flaked and exploded samples. Intact raw lupin had polygonal cells, 30-35 pm in size,
with thick walls completely surrounding the cytoplasm where protein bodies are
situated (Fig. 2a). This ordered cellular arrangement was well preserved in the control
material after extraction of oil (Fig. 2b). Flaking induced compaction of the cytoplasm,
and breakage and separation of cell walls (Fig. 2c). The main effect of explosion was
cracking of particles and loss of the cellular microstructure (Fig. 2d). Microstructural
models and outstanding features of the three types of particle are shown diagrammati-
cally in Fig. 3, Outer broken cells were present in all three models, and porosity
appeared to be significant only in flaked and exploded particles. Thus, differencesin the
pattern of extraction between the control and pretreated samples shown in Fig. 1 could
be explained by a possible change in the mechanism of protein diffusion from the
interior of the particle, due to microstructural modifications.
Control Fl aked Exploded
Figure 3. Microstructural models and outstanding features of control, flaked and exploded
defatted lupin particles: d,, particle size; d,, Characteristic dimension. For large flakes
(dp =1425 and 725 pm) the characteristic dimension (d,) of the flake is the thickness (c, 500 km).
Large arrows in flaked and exploded particles represent the direction of the applied forces dur-
ing pretreatment.
Parameters C,, C, and k of the kinetic model and diffusion coefficients are pre-
sented in Table 1, and were used to plot the lines in Fig. 1. Protein concentration from
exploded samples tended to be lower than predicted by the model, possibly due to small
protein losses during water tempering before explosion and extraction. Using the three
constants, the overall quality of the fit of the model to the experimental data is good.
The removal of surface protein was an important step in the extraction mechanism,
particularly as the specific surface (surface area/mass) of the particles increased. C,
accounted for 1% of the total protein extracted in the largest exploded particle and for
82% in the case of the smallest flake. The relationship between particle size and C, is
shown in Fig. 4. A hypothetical concentration, obtained by emptying of protein from
the outer single cell layer of a sphere, was calculated from the protein content in a shell
with a thickness of 30 pm. This is plotted as the line in Fig. 4 for comparison. C, of con-
trol particles, which most closely resemble a homogeneous, spherical body, followed
the theoretical prediction well.
h
)

P

T
a
b
l
e

1
.

K
i
n
e
t
i
c

p
a
r
a
m
e
t
e
r
s

a
n
d

d
i
f
f
u
s
i
o
n

c
o
e
f
f
i
c
i
e
n
t
s

f
o
r

p
r
o
t
e
i
n

e
x
t
r
a
c
t
i
o
n

f
r
o
m

d
e
f
a
t
t
e
d

l
u
p
i
n

p
a
r
t
i
c
l
e
s

M
e
a
n

p
a
r
t
i
c
l
e

s
i
z
e

(
w
)

C
o
n
t
r
o
l

H
y
d
r
a
t
e
d

r
a
d
i
u
s

k

C
W

C
,
D
x
1
0
l
2

(
p
m
)

(
m
i
n
-
'
)

(
m
g

m
1
-
I
)

(
m
g

m
1
-
I
)

(
m
2

s
-
'
)

F
l
a
k
e
d

E
x
p
l
o
d
e
d

k

C
W

C
,

D
x
1
0
l
2

(
m
i
n
-
'
)

(
m
g

m
1
-
I
)

(
m
g

m
l
-
'
)

(
m
2
s
-
'
)

1
4
2
5

7
2
5

5
1
3

3
3
8

2
0
0

1
2
8

9
2
7

0
.
0
6
5

0
.
7
6

1
2
.
6
0

1
.
5

4
7
2

0
.
1
1
7

2
.
9
3

1
5
.
1
5

0
.
9

3
3
3

0
.
0
9
0

8
.
2
1

1
9
.
1
4

0
.
6

2
2
0

0
.
1
1
3

1
0
.
9
6

2
0
.
4
6

0
.
5

1
3
0

0
.
1
4
1

1
3
.
8
6

2
3
.
0
1

0
.
5

8
3

0
.
1
3
9

1
8
.
5
0

2
3
.
1
6

-

0
.
0
6
2

2
.
8
9

2
1
.
2
0

4
.
5

0
.
0
8
5

6
.
4
6

2
1
.
3
0

1
.
7

0
.
1
3
6

7
.
7
7

2
1
.
8
0

1
.
5

0
.
1
3
4

1
2
.
1
0

2
8
.
9
2

1
.
3

0
.
1
3
5

1
5
.
4
0

2
4
.
2
7

1
.
1

0
,
1
0
5

2
0
.
3
2

2
4
.
6
4

-

k

c
w

C
,

D
x
l
O
'
*

(
m
i
n
-
I
)

(
m
g

m
1
-
I
)

(
m
g

m
1
-
I
)

(
m
2
s
-
'
)

0
.
0
6
8

0
.
2
2

1
8
.
3
0

3
.
5

0
.
1
7
9

2
.
2
7

1
9
.
6
0

2
.
5

0
.
2
4
5

2
.
1
5

2
1
.
6
1

2
.
0

0
.
2
6
0

6
.
9
9

2
1
.
9
5

1
.
1

0
.
2
6
5

1
3
.
5
6

2
2
.
0
7

0
.
6

0
.
2
4
7

1
5
.
7
8

2
2
.
4
4

0
.
4

S
o
l
i
d
:

w
a
t
e
r

r
a
t
i
o

=

1
:
2
5
.

T
e
m
p
e
r
a
t
u
r
e

=

2
5

C
.

k

=

f
i
r
s
t

o
r
d
e
r

r
a
t
e

c
o
n
s
t
a
n
t
.

C
,
=

p
r
o
t
e
i
n

c
o
n
c
e
n
t
r
a
t
i
o
n

d
u
e

t
o

w
a
s
h
i
n
g
.

C
,

=

f
i
n
a
l

p
r
o
t
e
i
n

c
o
n
c
e
n
t
r
a
t
i
o
n
.

D

=

d
i
f
f
u
s
i
o
n

c
o
e
f
f
i
c
i
e
n
t
.

a

;?
:
A

Q

e

g
.

Modelling extraction of lupin protein
27.0
- 26.5-
Q
' 26.0-
u
c
-
25.5
25
-
-
Radius ( pm)
Figure 4. Relationship between particle size and concentration of protein as a result of the
removal of surface protein in the extraction of defatted lupin seeds. Solid :water ratio, 1: 25;
temperature, 25C. Solid line represents the yield of protein from a hypothetical spherical par-
ticle with protein contained at the surface in a shell of 30 pm thickness. (0) Control; ( . )
flaked; (0) exploded.
For small particles, C, values were similar for all treatments; however, large differ-
ences were apparent for the two largest sizes. This was probably due to better avail-
ability, via pores and broken cells, for extraction of the protein entrapped in internal
cells in particles that were flaked or exploded. This effect was less important as particle
size became of the order of several cells and alternative free pathways (e.g. microcracks
due to fine milling) were available in the control sample.
The first order rate constant showed little variation with particle size for the smaller
particles, suggesting a single predominant mechanism dependent on microstructure.
Exploded particles of size < 725 pm had k values about twice as large as those of the
other two treatments, indicating a higher extraction rate due to facilitated transport
through the cracked microstructure. Lower and similar k values for the largest particle
size (1425 pm = 1.4 mm) for all treatments might be explained by a change from a
microstructure-related limiting mechanism to that of macrostructural effects.
27.5 r
25.0 I I I I I I I I I
3.00 3.05 3.10 3.15 3.20 3.25 3.30 3.35 3.40
lo3 T?K)
Figure 5. Arrhenius plot for effect of temperature on the diffusion coefficient of the extraction
of protein from defatted lupin seeds. Control material, particle size 1425 pm.
26
J . M. Aguilera and Hilda D. Garcia
Diffusion coefficients were higher in treated samples than in control (Table l), and
decreased as particle size was reduced. Values varied between 4 . 5 ~ 1 0 - ' ~ and
0 . 4 ~ lo-'' m2 s-l, which are one to two orders of magnitude lower than those for bulk
diffusion of proteins in water at infinite dilution (Schwartzberg & Chao, 1982). Russell
& Tsao (1982) reported 1000-fold reduction of the free solution diffusion coefficient due
to adsorption and steric hindrance, but their analysis also included late stages of protein
removal to complete extraction. They also reported similar effects of particle size on dif-
fusivity ; the explanation suggested by these authors was that, as grinding proceeded,
harder and denser material was generated.
Diffusivities were calculated from the model and correlated with temperature,
according to an Arrhenius expression ( r 2 = 0.994) which is represented in Fig. 5. The
calculated activation energy was 42.7 kJ mol-', which is in the upper limit of typical val-
ues for diffusional processes, and slightly higher than that reported by Spiro & Selwood
(1984) for coffee infusion.
In conclusion, changes in extraction behaviour of particles that underwent different
treatments to facilitate protein removal correlated well with observed microstructural
modifications and quantitative information generated from extraction curves.
Acknowledgments
This work was funded by the Office of Research (DIUC), Universidad Catolica de Chile
through grant 36/85. Assistance of Mr J . Delgado in model-fitting and Mrs Ana Maria
Mujica in microscopy work is appreciated.
References
Aguilera, J.M. &Trier, A. (1978). The revival of the
lupin. Food Technology, 32, (8), 70-76.
Aguilera, J .M., Gerngross, M. F. & Lusas, E.W.
(1983). Aqueous processing of lupin seed. Journal
of Food Technology, 18,327-333.
Aguilerd, J. M. , Escobar, G.A., del Valle, J. M. &
San Martin, R. (1987). Ethanol extraction of red
peppcrs: kinetic studies and microstructure. Inter-
national Journal of Food Science and Technology,
Clo, G. & Voilley, A. (1983). Evaluation of ground
coffee particles for optimum extraction.
Lebensmittel Wissenschaft und Technologie, 16,
39-42.
Coats, H.B. & Wingard, M.R. (1950). Solvent
extraction. 111. The effect of particle size on
extraction rate. Journal of the American Oil
Chemists' Society, 27,93-96.
Cooper, G.T. (1977). The Tools of Biochemistry.
P. 51. New York: J ohn Wiley.
Crank, J . (1975). The Mathematics of Diffusion.
Pp. 90-93. London: Oxford University Press.
Fan, T.Y. & Sosulski, F. W. (1974). Dispersibility
and isolation of proteins from legumes. Canadian
Institute of Food Science and Technology Journal,
22,225-232.
7,256-259.
Hill, G.D. (1986). Recent developmentsin theuseof
lupins for animal and human nutrition. Proceed-
ings of the Fourth International Lupin Conference,
August 15-22, 1986, Geraldton, Western
Australia. Pp. 40-63. South Perth: The Depart-
ment of Agriculture.
Ruiz, L.P. & Hove, L. E. (1976). Conditions affect-
ing production of a protein isolate from lupin secd
kernels. Journal of the Science of Food and
Agriculture, 27,667-674.
Russell, M. H. & Tsao, G.T. (1982). Protein removal
from corn endosperm by solvent extraction.
American Institute of Chemical Engineers Sym-
posium Series, No. 218,78,83-89.
Schwartzberg, H.G. (1975). A mathematical
analysis of solubilization kinetics and diffusion in
foods. Journal of Food Science, 40,211-213.
Schwartzberg, H.G. & Chao, R.Y. (1982). Solute
diffusivities in leaching processes. Food Techno-
Spiro, M. & Siddique, S. (1981). Kinetics and
equilibria of tea infusion: Kinetics of extraction of
theaflavins, thearubigins and caffeine from Koon-
song broken Pekoe. Journal of the Science of Food
and Agriculture, 32, 1135-1139.
logy, 36,76-83.
Modelling extraction of lupin protein 27
Spiro, M. & Selwood, R. M. (1984). The kinetics and
mechanism of caffeine infusion from coffee: the
effect of particle size. Journal of the Science of
Food and Agriculture, 35,915-924.
Voil\ey, A. & Sirnatos, D. (1980). Modeling and
solubilization process during coffee brewing. Jour-
nalof Food Processing Engineering, 3, 185-198.
(Received 6 February, accepted 14 October 1988)