http://informahealthcare.com/mncISSN: 0265-2048 (print), 1464-5246 (electronic)

J Microencapsul, Early Online: 1–11! 2014 Informa UK Ltd. DOI: 10.3109/02652048.2014.932028

RESEARCH ARTICLE

Microencapsulation of krill oil using complex coacervation

Sarya Aziz1, Jagpreet Gill1, Pierre Dutilleul2, Ronald Neufeld3, and Selim Kermasha1

1Department of Food Science and Agricultural Chemistry, McGill University, Ste-Anne de Bellevue, QC, Canada, 2Department of Plant Science, McGill

University, Ste-Anne de Bellevue, QC, Canada, and 3Department of Chemical Engineering, Queen’s University, Kingston, ON, Canada

Abstract

The research work was aimed at the development of a process to yield gelatin-gum Arabicmultinuclear microcapsules of krill oil (KO), via complex coacervation. On the basis of theexperimental results of the screening trials, a three-level-by-three-factor Box–Behnken designwas used to evaluate the effects of the ratio of the core material to the wall (RCW; x1), thestirring speed (SP; x2) and the pH (x3) on the encapsulation efficiency (EE). The experimentalfindings indicated that x3 has the most significant linear and quadratic effects on the EE of KOand a bilinear effect with x1, whereas x2 did not have any significant effect. The optimalconditions for a 92% of EE were: 1.75:1 for RCW, 3.8 for pH and 3 for SP. The microcapsules,formed by complex coacervation and without any cross-linking agent, were multinucleated,circular in shape and had sufficient stability to maintain their structure.

Keywords

Beef gelatin, Box–Behnken, gum Arabic,multinucleated capsules

History

Received 25 November 2013Revised 17 May 2014Accepted 3 June 2014Published online 21 August 2014

Introduction

Krill oil (KO) offers a new abundant source of n-3-polyunsatur-ated fatty acids (n-3-PUFAs) on the market (Massrieh, 2008), inparticular, eicosapentaenoic acid (EPA, C20:5 n-3) and docosa-hexaenoic acid (DHA, C22:6 n-3) which are widely recognised fortheir nutritional and health benefits (Kidd, 2007). As compared toother marine oils, KO contains up to 40% of phospholipids anddiverse naturally occurring antioxidants mainly astaxanthin,which confers to the oil its characteristic orange colour(Deutsch, 2007; Massrieh, 2008). However, its incorporation infoods products is limited because of its low solubility in thehydrophilic media (Liu et al., 2010) and its oxidative instability(Bustos et al., 2003). Microencapsulation is considered aneffective method for the oxidative stabilisation of edible oils(Bustos et al., 2003), and is used for the protection and thedelivery of functional lipids in food applications (Champagne andFustier, 2007). Complex coacervation involves the electrostaticattraction between two biopolymers of opposing charges(Liu et al., 2010). As compared to other technologies, thecomplex coacervation has been successfully commercialised,since it offers several advantages including higher payload, tracesof surface oil and a relatively thick outer shell (Barrow et al.,2007). The microcapsules, obtained by coacervation, can bedivided into mononuclear, which are formed when a given oil isencapsulated by coacervates, and multinuclear that are formed bythe aggregation of multiple mononuclear ones (Dong et al., 2007).Spherical multinuclear microcapsules have been found to possessbetter controlled-release characteristics than their mononuclearcounterparts (Dong et al., 2007, 2011). The literature (Prata et al.,2008; Liu et al., 2010; Dong et al., 2011; Qv et al., 2011) reportedthat the gelatin (GE)–gum Arabic (GA) is the most investigated

system in the formation of complex coacervation, where its useas a delivery matrix has many advantages including its abun-dance and its biodegradability (Liu et al., 2010). Because of thedietary and religious customs limitations, which may preventthe use of porcine gelatin, beef-hide gelatin was used (Karim andBhat, 2008).

In order to investigate the individual, combined and cumula-tive effects of various factors, response surface methodology(RSM) is applied as an effective statistical technique because ofits advantages, mainly by the limited number of experimental runsthat it requires (Myers et al., 2009). Box–Behnken designs(BBDs) are a class of rotatable or nearly rotatable second orderdesigns based on three-level incomplete factorial designs; BBDhas been demonstrated to be slightly more efficient than thecentral composite design, but much more efficient than the three-level full factorial designs. Another advantage of the BBD is thatit does not contain combinations for which all factors aresimultaneously at their highest or lowest levels. It has no points onthe vertices of the cube and this could be useful in avoidingexperiments under extreme conditions, for which unsatisfactoryresults might occur (Ferreira et al., 2007).

The aim of the present work was to investigate the effectsof the most relevant process parameters on the morphology andthe EE of GE–GA microcapsules, via complex coacervation ofkrill oil (KO), using the BBD as statistical model. In addition,the physicochemical properties of the optimised microcapsules inthe absence of a cross-linker were also studied.

Materials and methods

Materials

Beef-hide gelatin Kosher-certified (GE, Type B, 250 ± 10 Bloom,12.0% moisture) was obtained from Vyse Gelatin Company(Schiller Park, IL). Gum Arabic (GA) was purchased from ACPChemicals Inc. (Montreal, QC). High-potency KO, extracted fromEuphausia superba, was generously obtained from Enzymotec

Address for correspondence: Selim Kermasha, Department of FoodScience and Agricultural Chemistry, McGill University, 21,111Lakeshore, Ste-Anne de Bellevue, QC, Canada H9X 3V9. Tel: +1 514398 7922. Fax: +1 514 398 7977. E-mail: selim.kermasha@mcgill.ca

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

Ltd (Morristown, NJ). Ammonium hydroxide, ethanol, glacialacetic acid and organic solvents of high-performance liquidchromatography (HPLC) grade were purchased from FisherScientific (Fair Lawn, NJ).

Optimisation of krill oil microencapsulation process

Preparation of microcapsules

The coacervated particles were produced according to a modifi-cation of the methods of Dong et al. (2007) and Liu et al. (2010).Solutions of GE powder (1% w/v, 45 ± 3 �C) were prepared by itsdispersion in 50 ml of deionised water (Milli-Q, MilliporeCorporation; Billerica, MA) under constant agitation at a scaleof 3 over 10 for 10 min, using an electrically powered overheadstirrer (Caframo Stirrer; Wiarton, ON). Solutions of GA (1% w/v,45 ± 3 �C) were also prepared in a similar manner. Themicroencapsulation was carried out in a double-jacketed reactorlinked to a circulator-water bath (ThermoScientific; Neslab EX-7,Newington, NH) to maintain a constant temperature of 45 ± 3 �C.All operation steps were prepared at 45 �C, unless otherwiseindicated. An amount of 2 g of KO was emulsified into 50 ml ofGE dispersion (20 500 rpm, 3 min), using a PowerGen 125 high-shear homogenizer (KIA; Wilmington, NC). The homogenisationtook place at room temperature, using the pre-heated mixture andthe emulsion was placed back into the double-jacketed reactor tomaintain the temperature. The GA solution (50 ml) was thenadded dropwise, using a Pasteur pipette (Wheaton; Millville, NJ)to the GE-stabilised emulsion for a final volume of 100 ml (pH5.37). The emulsion was then stirred for an additional 5 min,followed by an acidification to pH 4.0 by the dropwise addition of10% (v/v) acetic acid to induce the complex coacervation. Themixture was allowed to cool slowly to room temperature over a2-h period, under constant mechanical stirring, at a scale of 3 over10. The particle suspension was slowly cooled further, withcontinuous stirring, to 10 �C, using an ice-bath. The stirring wasthen halted so to allow the phase separation of an upper aqueous-rich phase and a lower coacervate rich one with the entrappedoils layer. The upper aqueous-rich phase was removed and thecoacervate rich one was recovered and stored at 4 �C for furtheranalyses.

Screening single factor experiments

In order to identify the parameters that potentially have the mostsignificant effects on the EE of KO, initial screening (singlefactor) trials were performed. The investigated parameters,include the homogenisation rate (11 500, 14 500, 20 500 and30 000 rpm), the ratio of the core material to the wall (RCW) of1:1, 1.5:1, 2:1, 2.5:1 and 3:1, the concentration of wall materials(CWM) of 0.5%, 1%, 1.5% and 2% w/v, the pH of 3.4, 3.7, 4.0 and4.3 and the stirring speed (SP) of 2, 3, 4 and 5, over a scale of 10.The response was the %EE of krill oil.

Determination of the encapsulation efficiency of KO

Determination of the surface oil. The determination of thesurface oil on the wet capsules was carried out according to themethod of Drusch et al. (2006), where an amount of wetmicrocapsules was filtered (Whatman #41) to yield 500 mg ofdrained ones. Fifteen millilitres of petroleum ether (B.P. 36–60 �C) was added and the surface oil was extracted at roomtemperature from the samples by shaking (90 rpm, 15 min) themixture, using an orbital incubator shaker (New BrunswickScientific Co., Inc., Edison, NJ). The upper organic phase wasrecovered and concentrated, using a Thermo Savant AutomaticEnvironmental Speedvac System (Model AES1010; Thermo

Scientific, Fair Lawn, NJ). The amount of surface oil wasdetermined gravimetrically.

Determination of total oil content. Total oil content, includingencapsulated and surface, of the wet capsules was determinedaccording to a modification of the methods of Drusch et al. (2006)and Liu et al. (2010). A defined amount of the wet microcapsuleswas filtered (Whatman #41) to yield 500 mg of drained ones,which were then suspended into 3 ml of deionised water andplaced in the orbital incubator shaker (300 rpm, 15 min, 65 �C).One millilitre of 14.8 N ammonium hydroxide solution was addedto the mixture and the shaking was monitored for an additional15 min; the mixture was then cooled to room temperature.

Four millilitres of ethanol, petroleum ether and hexane weresuccessively added and the samples were agitated in the orbitalincubator shaker (300 rpm, 2 min, 25 �C). This mixture was thencentrifuged (IEC Central CL2; International Equipment Co.,Needham Heights, MA) at 1000 rpm. The extraction was repeatedwith the use of 2.5 ml of each solvent. The upper layer wasrecovered and the combined organic phases were concentrated,using the Thermo Savant Automatic Environmental SpeedvacSystem. Total oil content was determined gravimetrically.

Determination of encapsulation efficiency. Percent encapsulationefficiency (EE) was determined as:.

%EE ¼ Total oil� Surface oilð Þ=Total oil� 100½ � ð1Þ

For each batch, the determination of EE was performed intriplicate.

Experimental design

On the basis of the experimental findings of single factor trials, a‘‘three-level-by-three-factor’’ Box–Behnken was used to optimisethe %EE of KO. The treatment combinations were constructedfrom 12 factorial points and 5 central points. Duplicate reactionswere carried for all designed points except the central one,resulting in 29 treatments. All reactions were carried in arandomised order to minimise variability in the observedresponses that could result from extraneous factors.

The variables as well as their coded and uncoded values arepresented in Table 1. Table 2 shows the experimental data for thethree-level-by-three-factor response surface analysis. The inves-tigated response Y was the percent of EE. A quadratic polynomialregression model was used to predict the Y variable. The responsesurface model proposed for Y was:

Y ¼ �0 þ ��ixi þ ��iix2i þ ���ijxixj ð2Þ

where �0, �i, �ii and bij are the intercept and the linear, quadraticand interaction regression coefficient terms, respectively, and xi

and xj are the independent variables (Myers et al., 2009). The datawere analyzed using the Design Expert program (State-Ease Inc.,Version 7.1.4, Statistics Made Easy, Minneapolis, MN). TheRSREG and GLM (with ss3 option) procedures of the StatisticalAnalytical System (SAS 9.2, SAS Institute, Inc., Cary, NC) wereused for canonical analysis.

Table 1. Process variables and their levels, used in the Box–Behnkendesign.

Variables NameCoded levels

�1 0 +1

x1 Ratio of core material to wall 1.25:1 1.50:1 1.75:1x2 Stirring speed (over a scale of 10) 2 3 4x3 pH 3.8 4.0 4.2

2 S. Aziz et al. J Microencapsul, Early Online: 1–11

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

Physicochemical properties of optimised KOmicrocapsules

Using the optimised conditions, a portion of the microcapsuleswas lyophilised (Labconco Co.; Model 79480, Kansas City, MO)and stored at 4 �C. The morphology of the optimised wet andlyophilised microcapsules was monitored on a Hemacytometer(Hausser Scientific, Horsham, PA), using a stereomicroscope(Model SZX12, Olympus America Inc.; Melville, NY) equippedwith a QICAM 12-bit camera (QImaging Corporation; Surrey,BC). The particle diameters and the size distribution weremeasured, using a Malvern Mastersizer 2000 equipped with a wetdispersion accessory, Hydro 2000S (Malvern Instruments Ltd,Worcestershire, UK). The water activity (aw) was measured at25 �C by a Novosina AW SPRINT TH-500 system (Axair Ltd.;Pfaffikon, Switzerland), using the humidity reference points. Themoisture content was determined using a vacuum oven, accordingto the AOAC (1990) method. All measurements were carried intriplicate.

Results and discussion

Optimisation of the krill oil microencapsulation process

Screening single factor trials

In order to identify the parameters that potentially have the mostsignificant effects on the encapsulation efficiency (%EE) of KOand to determine the appropriate levels of single factors, initialscreening trials were performed.

Effect of the homogenisation rate. The effect of the homogen-isation rates on the morphology and stability of GE–KOemulsions was investigated. The experimental results indicatethat with homogenisation rates of 11 500 and 14 500 rpm, free oilwas prevalent. Similar findings were reported by Liu et al. (2010)for the encapsulation of flaxseed oil (FSO) within GE–GA matrixvia complex coacervation, at low homogenisation rates. As thehomogenisation rate increased, smaller but more homogenousdroplets were obtained. These findings could be the results of thehigher mechanical energy and droplet surface area in the system(Liu et al., 2010). Although the size of the emulsion droplets wasthe smallest at the homogenisation rate of 30 000 rpm, the highermechanical energy resulted in the creation of a foam layer.Lemetter et al. (2009) reported that above a certain limit, air canbe incorporated in the system, thus creating a foam layer, in whichthe capsules are no longer subjected to shear; in order to preventfoaming formation, a maximum level of shear is required.

Figure 1 demonstrates the GE–GA coacervates, monitored by astereomicroscope, as a function of the homogenisation rate; underthe experimental process parameters, the formed microcapsulesconsisted of mononuclear capsules surrounding a single large oildroplet (Figure 1). In addition, Figure 1 shows that thehomogenisation rate affected the shape and stability of themicrocapsules. At a homogenisation rate of 11 500 (Figure 1a)and 14 500 (Figure 1b) rpm, the microcapsules were irregular inshapes, which could be a consequence of the heterogeneous sizeof the emulsion droplets. These experimental findings (Figure 1)could be explained by the Stokes law of oil, which depicts that thedroplet size of the emulsion could affect its stability (Druschet al., 2012). At a homogenisation rate of 20 500 rpm, themicrocapsules were circular in shape and consisted of mono-nuclear capsules; however, at a higher homogenisation rate, theaggregation of the capsules was obtained as a result of therupturing of the capsules membrane. Liu et al. (2010) reportedthat the possible exposure of the entrapped core could allow forincreased hydrophobic interactions and clustering of neighbour-ing capsules.

At the homogenisation rate of 20 500 rpm (Figure 1c), the %EEof wet microcapsules was 69.2 ± 0.27. However, other ratesresulted by their breaking and by the release of their content,which made it difficult to estimate their %EE; this breakdowncould be the result of the instability of the emulsion (Drusch et al.,2012). Based on these findings, the homogenisation rate of20 500 rpm was selected for further investigation.

Effect of RCW on the morphology of the microcapsules. Figure 2represents the GE–GA coacervates, monitored by a stereomicro-scope, as a function of the ratio of the core material to the wall(RCW). The experimental findings indicated that the RCW had aninfluence on the internal structure of the microcapsules. With aRCW lower than 2:1, the microcapsules were multinuclear withmultiple small oil droplets, entrapped within the GE–GA capsulesand they were circular in shape (Figure 2a and b). With a RCW of2:1 and higher (Figure 2c–e), the microcapsules were mono-nucleated, with a single oil droplet core entrapped in thehydrocolloid shell. The microcapsules, with a RCW of 2:1 and2.5:1, were circular in shape, whereas those with a RCW of 3:1were irregular. Dong et al. (2007) indicated that with increasingRCW, the morphology of microcapsules encapsulating pepper-mint oil changed from spherical to irregular (Figure 2e). Inaddition, the size of the microcapsules increased with increasingRCW, both in mononuclear (Figure 2a and b) and multinuclearcapsules (Figure 2c–e). These findings are in agreement withthose of Dong et al. (2007), who reported that when the RCW wasincreased from 1:1 to 2:1, the mean particle size increased from53 to 73 mm, and continued to increase at RCWs of 3:1 and 4:1.

Table 2. Box–Behnken second order design, experimental data for 3levels–3-factors response surface analysis.

RCWa Stirring speedb pH EE (%)c

Run x1 x2 x3 y

1 1.25d (�1)e 2d (�1)e 4.0d (0)e 74.432 1.25 (�1) 2 (�1) 4.0 (0) 78.663 1.25 (�1) 4 (1) 4.0 (0) 76.764 1.25 (�1) 4 (1) 4.0 (0) 72.165 1.75 (1) 2 (�1) 4.0 (0) 90.346 1.75 (1) 2 (�1) 4.0 (0) 90.937 1.75 (1) 4 (1) 4.0 (0) 84.648 1.75 (1) 4 (1) 4.0 (0) 76.909 1.50 (0) 2 (�1) 3.8 (�1) 81.80

10 1.50 (0) 2 (�1) 3.8 (�1) 74.4411 1.50 (0) 2 (�1) 4.2 (1) 64.4512 1.50 (0) 2 (�1) 4.2 (1) 47.3713 1.50 (0) 4 (1) 3.8 (�1) 75.3014 1.50 (0) 4 (1) 3.8 (�1) 96.9215 1.50 (0) 4 (1) 4.2 (1) 45.1616 1.50 (0) 4 (1) 4.2 (1) 51.5617 1.25 (�1) 3 (0) 3.8 (�1) 85.9218 1.25 (�1) 3 (0) 3.8 (�1) 64.5019 1.75 (1) 3 (0) 3.8 (�1) 88.8420 1.75 (1) 3 (0) 3.8 (�1) 92.2621 1.25 (�1) 3 (0) 4.2 (1) 78.0322 1.25 (�1) 3 (0) 4.2 (1) 79.6923 1.75 (1) 3 (0) 4.2 (1) 51.4924 1.75 (1) 3 (0) 4.2 (1) 54.2925 1.50 (0) 3 (0) 4.0 (0) 85.6126 1.50 (0) 3 (0) 4.0 (0) 83.9627 1.50 (0) 3 (0) 4.0 (0) 82.1328 1.50 (0) 3 (0) 4.0 (0) 78.7729 1.50 (0) 3 (0) 4.0 (0) 81.70

aRatio of core material to wall.bStirring speed over a scale of 10.cEncapsulation efficiency (EE) in percent was calculated as the difference

between the total oil minus the surface oil divided by total oil,multiplied by 100.

dActual experimental amounts.eActual experimental amounts in coded values.

DOI: 10.3109/02652048.2014.932028 Microencapsulation of krill oil 3

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

On the other hand, the results (Figure 2) show that themultinuclear microcapsules were larger in size, as compared totheir mononuclear counterparts. In comparison with the small oil-containing microcapsules, the large- and medium-sized micro-capsules could be more suitable for controlled core releaseproperties (Lee and Rosenberg, 1999).

The KO %EE of wet microcapsules reached 69.2 ± 0.27 and84.0 ± 0.48 with a RCW of 2:1 and 1.5:1, respectively. Themultinuclear capsules showed better stability and less suscepti-bility to breakage; this could explain the higher EE obtained withthe multinucleated ones. The literature (Yeo et al., 2005; Donget al., 2007, 2011; Liu et al., 2010) indicated that the multinuclearmicrocapsules are generally recognised as having better con-trolled release properties than the mononuclear designs, and theycan release the core material slowly, even if the wall of themicrocapsules is completely destroyed. In contrast, the mono-nuclear microcapsules with reservoir type structure can release allof the core material very quickly even when the microcapsule wallis only partly destroyed (Dong et al., 2007).

In the absence of cross-linking agents, it was necessary toinvestigate whether the microcapsules would keep their integrityafter their lyophilisation. The results show that the multinuclearmicrocapsules maintained more adequately their structure ascompared to their mononuclear counterparts. Likewise, Alvimand Grosso (2010) indicated that the lyophilisation processmaintained the wall integrity of the multinucleated capsules of amixture of paprika oleoresin and soybean oil, even without across-linker. The lyophilisation of the mononucleated capsulesresulted by their breakdown and the release of their content; thesefindings are in agreement with those of Alvim and Grosso (2010)who reported that the lyophilisation can cause the disintegrationof the capsule wall, which may affect the size of the particle andconsequently, the core release properties. The literature (Prataet al., 2008; Liu et al., 2010; Dong et al., 2011; Qv et al., 2011),the multinuclear microcapsules are believed to have better release

core properties than their mononuclear counterparts. Based on theexperimental findings, a ratio of core material to wall (RCW) of1:5:1 was selected for further studies.

Effect of CWM on the morphology of the microcapsules. Theeffect of the concentration of wall materials (CWM) on themorphology of the microcapsules was investigated (Figure 3). Theresults show that less free coacervates were formed with CWMsof 0.5% and 1% as compared to that with higher ones. Thesefindings are in agreement with those of Yeo et al. (2005), whoreported that reducing the concentrations of GE and GA solutionsdecreased the frequency of the free coacervates. Figure 3(a) showsthat with a CWM of 0.5%, the microcapsules were multinuclear,irregular in shape and large in size. Yeo et al. (2005) indicatedthat the microcapsules, prepared with the most diluted polymersolution (0.5%), formed a significant amount of aggregates, forwhich the size could not be reliably measured. As the CWMincreased from 0.5% to 2%, we found that the microcapsulesbecame smaller. Similarly, Liu et al. (2010) reported that the sizeof aggregated capsule clusters decreased substantially as thebiopolymer level was raised from 1% to 2%. In addition, with aCWM 41%, the capsules were mononuclear. The experimentalfindings (Figure 3) show that the CWM could influence theinternal structure as well as the morphology of the microcapsules.Likewise, the CWM was found to affect the number of coreaggregates in a microcapsule when the flavour oil wasencapsulated using complex coacervation method (Yeo et al.,2005). In summary, the use of a CWM of 1% resulted in theformation of spherical multinucleated capsules and such concen-tration was selected for further studies.

Effect of pH on the properties of coacervates. The effect of pHon the properties of coacervates, which are formed by theinteraction of two biopolymers with opposite charge, was

Figure 1. Gelatin–gum Arabic coacervates observed by stereomicroscope as a function of ratio of homogenisation rate of (a) 11 500 rpm, (b)14 500 rpm, (c) 20 500 rpm and (d) 30 000 rpm. Coarcervates were formed at a constant ratio of core material to wall of 2:1, concentration of wallmaterials of 1%, pH of 4 and a stirring speed of 3 out of a scale of 10.

4 S. Aziz et al. J Microencapsul, Early Online: 1–11

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

investigated. At a pH 54.0, the experimental results (Figure 4aand b) show that the formed microcapsules were of irregularshapes. Similarly, Dong et al. (2007) indicated that when the pHvalue was53.7, both the quantity and the viscosity of coacervatesstarted to decrease, where the spherical multinuclear becameirregular at pH 3.4. In addition, at pH 4.0 (Figure 4c), thespherical multinuclear capsules were formed. These resultscould be explained by the fact that, as the pH decreased,the GE tended to take more positive charges and interact withGA in order to form coacervates of high viscosity. Consequently,the mononuclear microcapsules were easier to collide witheach other and accumulated into larger size multinuclear micro-capsules (Figure 4c). Since the spherical multinuclear capsulesare of interest, the pH of 4.0 was considered for furtherinvestigation.

Effect of SP on the morphology of microcapsules. Figure 5demonstrates the effect of stirring speed (SP) on the morphologyof microcapsules. With a SP of 2 out of 10, microcapsulesof larger size and of high oil content were formed. Similarly,

Dong et al. (2007) reported that with lower SP, a lot of emulsiondroplets floated upward and accumulated into multinuclearmicrocapsules with larger size and higher oil content. Figure 5shows that as the SP increased, the size of microcapsulesdecreased and the proportion of mononuclear capsules increased.Lemetter et al. (2009) indicated that as the rotation speedincreased, the mean diameter of a capsule decreased; in addition,with higher SP, many more mononucleated capsules were presentin the dispersion and the remaining multinucleated ones werequite small. The higher the turbulence level, due to the higher SP,the shorter the contact time between the two oil droplets and theless likely they will merge together to form a spherical structureenclosing two oil droplets (Lemetter et al., 2009); this couldexplain the increase in the proportion of mononuclear capsules inFigure 5(c) and (d). The %EE of KO was 84.0 ± 0.48%,75.9 ± 0.74% and 65.0 ± 9.45% for SP values of 3, 4 and 5,respectively. The decrease in %EE could be explained by theincrease in the proportion of mononuclear capsules following theincrease in SP, as these then tend to be more vulnerable tobreakage (Yeo et al., 2005; Dong et al., 2011).

Figure 2. Gelatin–gum Arabic coacervates observed by stereomicroscope as a function of the ratio of the core material to the wall of (a) 1:1, (b) 1.5:1,(c) 2:1, (d) 2.5:1 and (e) 3:1. Coacervates were formed at a constant homogenisation rate of 20 500 rpm, concentration of wall materials of 1%, pH of 4and a stirring speed of 3 out of a scale of 10.

DOI: 10.3109/02652048.2014.932028 Microencapsulation of krill oil 5

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

Figure 4. Gelatin–gum Arabic coacervates observed by stereomicroscope as a function of pH of (a) 3.4, (b) 3.7, (c) 4.0 and (d) 4.3. Coarcervates wereformed at a constant ratio of core material to wall of 1.5:1, concentration of wall materials of 1%, homogenisation rate of 20 500 rpm and a stirringspeed of 3 out of a scale of 10.

Figure 3. Gelatin–gum Arabic coacervates observed by stereomicroscope as a function of the concentration of wall materials of (a) 0.5%, (b) 1.0%, (c)1.5% and (d) 2.0%. Coarcervates were formed at a constant ratio of core material to wall of 1.5:1, pH of 4, homogenisation rate of 20 500 rpm and astirring speed of 3 out of a scale of 10.

6 S. Aziz et al. J Microencapsul, Early Online: 1–11

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

Box–Behnken design

The appropriate ranges of values for factors, obtained from thesingle factor trials, were as follows: 1:25–1.75 for RCW, 2–4 forSP and 3.8–4.2 for pH. On that basis and taking into considerationthat the percentage of encapsulation efficiency (Y) is the responseto RCW (x1), stirring speed (x2) and pH value (x3), theexperiments were arranged according to a Box–Behnken design(BBD). The levels of the three potential explanatory variables arelisted in Table 1.

Model fitting and ANOVA. Throughout this study, the BBD wasused for the optimisation of the %EE of KO, using the complexcoacervation. The actual and coded values of the factors, togetherwith the corresponding %EE experimental data, are reported inTable 2. The response surface analysis (Table 3) shows that the%EE of KO can be suitably described with the following quadraticpolynomial model:

Y ¼ 81:17þ 1:22x1 � 1:44x2 � 11:75x3 þ 3:34x21 � 1:95x1x2

� 10:33x1x3 � 3:91x22 � 3:89x2x3 � 10:13x2

3 ð3Þ

where, Y¼%EE, and x1, x2 and x3 are the coded values of RCW,SP and pH value, respectively.

The response surface analysis (Table 3) for the optimisation ofthe %EE of KO showed that the predicted model was adequate(p¼ 0.0002). The lack-of-fit F value of 3.06 was non-significant,with an associated p value much40.05. These results imply thatthe model could well represent the response, within the ranges ofvalues selected for the reaction parameters.

Effects of variables on the EE of KO. Among the threeindependent variables, the pH value (x3) had significant linearand quadratic effects on the encapsulation efficiency (EE) of krilloil (KO) (Table 3), with p value of 50.0001 and 0.0036,respectively. On the other hand, RCW showed a significant

bilinear effect when it was combined with pH value (x1*x3;p¼ 0.0018). SP (x2) was found to have no significant linear,bilinear or quadratic effect on the EE of KO. In addition, no othercombined effects were found to be statistically significant.

Figure 6 shows the effects of RCW, pH and SP values, takentwo by two, on the EE of KO. In the two-dimensional responsesurfaces (Figure 6a–c), the factors that have not been displayedwere fixed at their central value. The statistical analysis (Table 3)indicates that the RCW has an important bilinear effect with pHon the EE of KO. As the RCW increased (Figure 6b), the EE

Figure 5. Gelatin–gum Arabic coacervates observed by stereomicroscope as a function of stirring speed of (a) 2, (b) 3, (c) 4 and (d) 5. Coarcervateswere formed at a constant ratio of core to wall ratio of 1.5:1, concentration of wall materials of 1%, homogenisation rate of 20 500 rpm and pH of 4.

Table 3. Analysis of variance of the regression parametersa.

Source

Numberof degree

of freedom

Type IIISum ofsquares

Meansquare F value p Value

x1 1 23.86 23.86 0.37 0.5505c

x2 1 33.12 33.12 0.51 0.4826c

x3 1 2207.59 2207.59 34.18 50.0001b

x1*x1 1 77.77 77.77 1.20 0.2862c

x1*x2 1 30.26 30.26 0.47 0.5019c

x1*x3 1 853.26 853.26 13.21 0.0018b

x2*x2 1 106.30 106.30 1.65 0.2150c

x2*x3 1 120.75 120.75 1.87 0.1875c

x3*x3 1 714.42 714.42 11.06 0.0036b

Model 9 4207.36 467.48 7.24 0.0002b

Error 19 1227.17 64.59 – –Lack of fit 3 447.44 149.15 3.06 0.0584c

Pure error 16 779.72 48.73 – –Total 28 5434.53 – – –

aThe encapsulation efficiency of krill oil has been investigated using threeindependent variables; ratio of core material to wall x1 (1.25:1–1.75:1),stirring speed x2 (2–4 over a scale of 10) and pH x3 (3.8–4.2) and it wascalculated as the difference between the total oil minus the surface oildivided by the total oil, multiplied by 100.

bSignificant at50.05 level.cNot significant at40.05 level.

DOI: 10.3109/02652048.2014.932028 Microencapsulation of krill oil 7

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

increased linearly over all the investigated ranges. Dong et al.(2007) reported that as the RCW increased, the formed coacer-vates were enough to encapsulate more core material.

On the other hand, as the pH increased, the response wasincreased linearly and up to a certain limit, followed by linear andquadratic decreases. These findings could be explained by the factthat with the increase in pH, GE took less positive charges andinteracted less strongly with the GA, decreasing hence the EE ofKO (Dong et al., 2007). In addition, the results (Table 3)show that the pH had significant linear and quadratic effects onthe EE of KO. Similarly, Qv et al. (2011) explored theoptimisation of process conditions for the encapsulation oflutein, using Box–Behnken as the statistical model; the authorsreported that pH had a significant linear and quadratic effectson the EE of lutein, since it controls the balance of macromol-ecules charges and therefore influencing the intensity of theelectrostatic interactions driving the formation of complexesbetween the two biopolymers.

The results (Table 3) show that SP had no effect on the EE.These results are in agreement with those of Dong et al. (2007),who reported that when the SP was increased, from 200 to400 rpm, the yield of microcapsules remained constant. Inaddition, Lemetter et al. (2009) indicated that during the shellgelation, the SP was critical to avoid coalescence that couldaffect the morphology and the size of the microcapsules ratherthan their EE.

Canonical analysis and the presence of a saddle point. Toexamine more closely the overall shape of the response surfacesand to characterise the nature of the stationary point, canonicalanalysis was performed on the matrices of the second derivativesof the predicted quadratic polynomial surfaces. The secondderivatives were calculated to determine whether the firstderivatives have been vanished at a maximum, minimum orsaddle point.

Canonical analysis is a mathematical method that decom-poses square matrices, such as matrices of second derivatives,into so-called ‘‘eigenvalues’’ and ‘‘eigenvectors’’. When alleigenvalues are negative, the stationary point is a maximum;when they are all positive, it is a minimum; and a mixture ofnegative and positive eigenvalues corresponds to a saddle point.Depending on the direction, the response may increase ordecrease when moving away from a saddle point (Carlson andCarlson, 2005).

The response surfaces showing the effects of RCW, SP andpH value, taken two by two, on the EE of KO are displayedin Figure 6. When RCW was combined with the SP and the pH,the response increased, or decreased from the centre of thesurface, depending on the direction (Figure 6a and b).Such saddle-shaped surfaces reflect the opposite effect of RCWon EE, as compared to SP and pH; this can also be seen fromthe mixed positive and negative signs of estimated coefficientsin Equation (3).

Although the maximisation of the EE of KO was of theobjective, the stationary point was a saddle point, which doesnot correspond to a unique optimum for the response (Bas andBoyacı, 2007). Hence, using RCW (x1), SP (x2) and pH (x3) asindependent variables, a maximum EE could not be obtained.Nevertheless, the canonical analysis revealed that a maximumpoint could be obtained when the SP (x2) and RCW (x3) areused as independent variables, that is, by excluding RCW (x1).Indeed, the response, once averaged over the experimentalvalues of RCW considered in the initial three-factor-by-three-level BDD, is a concave surface, with a maximum point(Figure 6c).

Response surface reduced quadratic model

In order to explore the response surface methodology and toovercome a saddle point, the response surface quadratic model

1.251.38

1.501.63

1.75

2.00 2.50

3.00 3.50

4.00

72

76.75

81.5

86.25

91

(a) (b)

(c)

Enc

apsu

latio

n ef

fici

ency

(%

)

1.251.38

1.501.63

1.75

3.80 3.90

4.00 4.10

4.20

51

62.75

74.5

86.25

98

Enc

apsu

latio

n ef

fici

ency

(%

)2.00

2.503.00

3.504.00

3.80 3.90

4.00 4.10

4.20

45

58

71

84

97 E

ncap

sula

tion

effi

cien

cy (

%)

Figure 6. Response surface and contour plots showing the effects of stirring speed versus ratio of core material to wall (RCW), pH versus RCW andstirring speed versus pH.

8 S. Aziz et al. J Microencapsul, Early Online: 1–11

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

was reduced by removing the non-significant variables,including the linear (x2), bilinear (x1*x2; x2*x3) and quadratic(x2

2) effects of stirring speed (SP). The analysis of variancefor the response surface reduced quadratic model is displayedin Table 4.

Model fitting and ANOVA. The analyses of variance and error(Table 4), performed for the optimisation of the encapsulationefficiency (EE) of KO, showed that the predicted model wasadequate (p50.0001). The lack-of-fit F value of 2.16 was non-significant, with an associated p value much40.05. These resultscould imply that the model represents well the relationshipsbetween the response and the reactions parameters, within theselected ranges. As indicated previously, the pH value (x3) had themost significant linear and quadratic effects on the EE of KO,with a p value of50.0001 and 0.0032, respectively. On the otherhand, RCW showed a significant bilinear effect, when it wascombined with the pH value (x1*x3; p¼ 0.0016). The equation of

predicted model, using coded values for the two-factor-by-three-level BBD, as follows:

Y ¼ 80:82þ 1:22x1 � 11:75x3 � 10:33x1x3 � 10:07x23 ð4Þ

where, Y¼EE (%), and x1 and x3 are the coded values of RCWand pH value, respectively.

Canonical analysis. Canonical analysis was performed on thereduced quadratic predicted model. The response surface,illustrating the individual effects of RCW (x1) and pH (x3) andtheir combined effects on the EE of KO, is displayed in Figure 7.The response, once averaged over the experimental values of SP,considered in the initial two-factor-by-three-level BDD, is aconcave surface with a maximum point.

Model validation. The optimal conditions for the EE of KOwere 1.75:1 for RCW, 3 out of scale of 10 for SP and 3.8 for pH.

Figure 8. Particle size distribution of the krilloil microcapsules prepared under the optimalconditions.

Table 4. ANOVA for response surface reduced quadratic modela.

SourceNumber of degree of

freedomType III Sum of

squares Mean square F value p Value

Model 4 3811.85 952.96 14.09 50.0001b

x1 1 23.86 23.86 0.35 0.5580c

x3 1 2207.59 2207.59 32.65 50.0001b

x1*x3 1 853.26 853.26 12.62 0.0016b

x3*x3 1 727.14 727.14 10.75 0.0032b

Error 24 1622.68 67.61 – –Lack of fit 8 842.95 105.37 2.16 0.0901c

Pure error 16 779.72 48.73 – –Total 28 5434.53 – – –

aThe encapsulation efficiency of krill oil has been investigated using two independent variables; ratio of core material towall x1 (1.25:1–1.75:1) and pH x3 (3.8–4.2) and it was calculated as the difference between the total oil minus the surfaceoil divided by the total oil, multiplied by 100.

bSignificant at50.05 level.cNot significant at40.05 level.

1.25 1.38 1.50 1.63 1.753.80

3.90

4.00

4.10

4.20

Ratio of core material to wall (x1)

pH v

alue

(x 3

)

57.2567

64.6146

71.9725

71.9725

79.3304

79.3304

86.6883

2 2

2 2

5

Prediction 94.0463

1.251.38

1.501.63

1.75

3.80 3.90

4.00 4.10

4.20

49

60.5

72

83.5

95

(a) (b)

Enc

apsu

latio

n ef

fici

ency

(%

)

Figure 7. Response surface and contour plot showing the effect of pH versus ratio of core material to wall (RCW).

DOI: 10.3109/02652048.2014.932028 Microencapsulation of krill oil 9

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

Under these conditions, the value of EE predicted by the modelwas 94%. Accordingly, the validation of the proposed model wasconducted and carried out with three repetitions. The mean of92% for EE, with a standard deviation of 0.16%, supports a strongcorrelation between the experimental results and the statisticalpredictions of the model.

Physicochemical properties of the optimised KOmicrocapsules

Using three measurements, the particle size distribution of theoptimised multinuclear microcapsules (Figure 8) exhibited anormal narrow one with the same curve shape, suggesting hence agood reproducibility. In addition, the size band values, the meanvolume and the standard deviation error bar are listed in Table 5.The results also show that the volume median diameter d (0.5)was 285.207mm, which suggests that 50% of the distribution waseither higher or lower than this value. In addition, 10% of themicrocapsules had a volume diameter d (0.1),5197.443 mm while90% of them had a volume diameter d (0.9) of 412.515mm. Thesefindings indicate that the microcapsules, obtained under theoptimum conditions, had a good homogeneity.

The lyophilisation of the GE–GA, with the entrapped KO,resulted by a free-flowing red–orange powder, with low moisturecontent of 0.19 ± 0.005% and a aw of 0.197 ± 0.00. Liu et al.(2010) reported similar results for the lyophilised microcapsules,with entrapped FSO of low moisture content of 3.17 ± 0.08% andwith aw of 0.18 ± 0.00. Qv et al. (2011) indicated that themicrocapsules of encapsulated lutein had a water content of3.12%; these authors suggested that the low moisture content andthe aw are beneficial for the prevention of food products againstoxidation. Klaypradit and Huang (2008) reported that themaximum moisture specification for most dried food products is3–4%, with a aw close to 0.3.

The KO EE reached 92 ± 0.16%, with low amounts of oilextracted from the surface. The %EE of the designed capsules arecomparable to others in literature (Yeo et al., 2005; Alvim andGrosso, 2010; Liu et al., 2010; Qv et al., 2011), when complexcoacervation was used as method of encapsulation.

To investigate whether the microcapsules will keep theirintegrity after drying, the optimised microcapsules were subjectedto lyophilisation. Figure 9 demonstrates a stereomicroscope viewof KO microcapsules, prepared under the optimal conditions,before and after lyophilisation. Figure 9(b) shows that thecapsules maintained their wall integrity after lyophilisationdespite the absence of cross-linking agents. These findings arein agreement with those reported in literature for multinucleatedmicrocapsules (Alvim and Grosso, 2010; Liu et al., 2010).

Conclusion

Using complex coacervation, the Box–Behnken design wassuccessfully used for the optimisation of the EE of KO. The

Figure 9. Stereomicroscope view of krill oil microcapsules prepared under the optimal conditions obtained: (a) before lyophilisation and (b) afterlyophilisation.

Table 5. Size band values and mean volume of krill oil microcapsules.

Size bandvalues (mm)a

Meanvolume (%)b

Standard deviationerror bar

104.7130.02 0.02

120.2260.25 0.11

138.0381.61 0.36

158.4894.39 0.41

181.9708.75 0.17

208.93013.47 0.54

239.88316.95 1.07

275.42317.63 1.14

316.22815.30 0.66

363.07811.06 0.20

416.8696.48 0.70

478.6302.99 0.80

549.5410.91 0.62

630.9570.17 0.18

724.4360.01 0.02

831.764

aSize band value (mm), measured by the Malvern Mastersizer 2000.bIs defined as the relative percent of volume of particles within the range

of two subsequent size band values.

10 S. Aziz et al. J Microencapsul, Early Online: 1–11

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.

chromogenic red–orange colour of KO, conferred by astaxanthin,facilitated the stereomicroscopic visualisation of the entrapped oilwithout the need of a lipid-soluble dye. The EE and the stabilityof the microcapsules were influenced by their internal structure,which can be obtained by adjusting the RCW and the SP. Theverification experiment, carried out under the optimal conditions,confirmed the validity of the prediction process. This encapsu-lation design could contribute to the use of krill oil as anutraceutical product in food systems.

Acknowledgements

The authors would like to thank Dr Jaqueline C. Bede for providingaccess to the stereomicroscope facility.

Declaration of interest

The authors report no conflicts of interest. The authors alone areresponsible for the content and writing of this article. This research wassupported by a Discovery Grant from the Natural Sciences andEngineering Research Council of Canada (NSERC). Sarya Aziz was therecipient of a graduate student fellowship, awarded by the FondsQuebecois de la Recherche sur la Nature et les Technologies (FQRNT).

References

Alvim ID, Grosso CRF. Microparticles obtained by complex coacerva-tion: Influence of the type of reticulation and the drying process on therelease of the core material. Cienc Tecnol Aliment, 2010;30:1069–76.

AOAC. 1990. Moisture in animal feed. Official methods of analysis934.01, 14th edn. Washington, DC: Association of Official AnalyticalChemists.

Barrow J, Nolan C, Jin Y. Stabilization of highly unsaturated fatty acidsand delivery into foods. Lipid Technol, 2007;19:108–11.

Bas D, Boyaci _IH. Modeling and optimization I: Usability of responsesurface methodology. J Food Eng, 2007;78:836–45.

Bustos R, Romo L, Yanez K, Diaz G, Romo C. Oxidative stability ofcarotenoid pigments and polyunsaturated fatty acids in microparticu-late diets containing krill oil for nutrition of marine fish larvae. J FoodEng, 2003;56:289–93.

Carlson R, Carlson JE. Canonical analysis of response surfaces: Avaluable tool for process development. Org Proc Res Dev, 2005;9:321–30.

Champagne CP, Fustier P. Microencapsulation for the improved deliveryof bioactive compounds into foods. Curr Opin Biotechnol, 2007;18:184–90.

Deutsch L. Evaluation of the effect of Neptune krill oil on chronicinflammation and arthritic symptoms. J Am Coll Nutr, 2007;26:39–48.

Dong Z, Ma Y, Hayat K, Jia C, Xia S, Zhang X. Morphology and releaseprofile of microcapsules encapsulating peppermint oil by complexcoacervation. J Food Eng, 2011;104:455–60.

Dong ZJ, Toure A, Jia CS, Zhang XM, Xu SY. Effect of processingparameters on the formation of spherical multinuclear microcapsulesencapsulating peppermint oil by coacervation. J Microencapsul, 2007;24:634–46.

Drusch S, Regier M, Bruhn M. 2012. Recent advances in themicroencapsulation of oils high in polyunsaturated fatty acids. In:Mcelhatton A, Sobral PJDA, eds. Novel technologies in food science –Their impact on products, consumer trends and environment.New York, NY: Dordrecht, Heidelberg, London: Springer, pp. 159–81.

Drusch S, Serfert Y, Schwarz K. Microencapsulation of fish oil with n-octenylsuccinate-derivatised starch: Flow properties and oxidativestability. Eur J Lipid Sci Technol, 2006;108:501–12.

Ferreira SLC, Bruns RE, Ferreira HS, Matos GD, David JM, BrandaoGC, da Silva EGP, Portugal LA, dos Reis PS, Souza AS, dos SantosWNL. Box-Behnken design: An alternative for the optimization ofanalytical methods. Anal Chim Acta, 2007;597:179–86.

Karim AA, Bhat R. Gelatin alternatives for the food industry: Recentdevelopments, challenges and prospects. Trends Food Sci Technol,2008;19:644–56.

Kidd PM. Omega-3 DHA and EPA for cognition, behavior, and mood:Clinical findings and structural-functional synergies with cell mem-brane phospholipids. Altern Med Rev, 2007;12:207–27.

Klaypradit W, Huang Y-W. Fish oil encapsulation with chitosan usingultrasonic atomizer. Food Sci Technol, 2008;41:1133–9.

Lee SJ, Rosenberg M. Preparation and properties of glutaraldehyde cross-linked whey protein-based microcapsules containing theophylline.J Control Release, 1999;61:123–36.

Lemetter CYG, Meeuse FM, Zuidam NJ. Control of the morphology andthe size of complex coacervate microcapsules during scale-up. Am InstChem Eng, 2009;55:1487–96.

Liu S, Low N, Nickerson M. Entrapment of flaxseed oil within gelatin-gum Arabic capsules. J Am Oil Chem Soc, 2010;87:809–15.

Massrieh W. Health benefits of omega 3 fatty acids from Neptune krill oil.Lipid Technol, 2008;20:108–11.

Myers RH, Montgomery DC, Anderson-Cook CM. 2009. The analysis ofsecond-order response surfaces. Response surface methodology:Process and product optimization using designed experiments.Hoboken, NJ: John Wiley & Sons, Inc., pp. 219–65.

Prata AS, Zanin MHA, Re MI, Grosso CRF. Release properties ofchemical and enzymatic crosslinked gelatin-gum Arabic microparticlescontaining a fluorescent probe plus vetiver essential oil. Colloids SurfB, 2008;67:171–8.

Qv X-Y, Zeng Z-P, Jiang J-G. Preparation of lutein microencapsulation bycomplex coacervation method and its physicochemical properties andstability. Food Hydrocoll, 2011;25:1596–603.

Yeo Y, Bellas E, Firestone W, Langer R, Kohane DS. Complexcoacervates for thermally sensitive controlled release of flavorcompounds. J Agric Food Chem, 2005;53:7518–25.

DOI: 10.3109/02652048.2014.932028 Microencapsulation of krill oil 11

Jour

nal o

f M

icro

enca

psul

atio

n D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

cgill

Uni

vers

ity o

n 09

/16/

14Fo

r pe

rson

al u

se o

nly.