OE

RAC

a

ARRA

KSTEBDR

1

oSettNci

pttoodlebcit

0h

J. of Supercritical Fluids 74 (2013) 105– 114

Contents lists available at SciVerse ScienceDirect

The Journal of Supercritical Fluids

jou rn al h om epage: www.elsev ier .com/ locate /supf lu

ptimization of the supercritical fluid extraction of triterpenic acids fromucalyptus globulus bark using experimental design

ui M.A. Domingues, Marcelo M.R. de Melo, Eduardo L.G. Oliveira, Carlos P. Neto,rmando J.D. Silvestre, Carlos M. Silva ∗

ICECO, Department of Chemistry, University of Aveiro, Aveiro 3810-193, Portugal

r t i c l e i n f o

rticle history:eceived 30 June 2012eceived in revised form 3 December 2012ccepted 4 December 2012

a b s t r a c t

The supercritical CO2 extraction of E. globulus deciduous bark was carried out at different temperatures(40–60 ◦C), pressures (100–200 bar), and ethanol contents (0.0–5.0 wt. %) to study triterpenic acids (TTAs)recovery. A factorial design of experiments and response surface methodology were implemented toanalyze the influence of these variables upon extraction and perform its optimization. The best conditionswere 200 bar, 40 ◦C and 5% ethanol, for which the statistically validated regression models provided:

eywords:upercritical fluid extractionriterpenic acidsucalyptus globulusarkesign of experiments

extraction yield of 1.2% (wt.), TTAs concentration of 50%, which corresponds to TTAs yield of 5.1 g/kg ofbark and a recovery of 79.2% in comparison to the Soxhlet value. The trends of the free and acetylated TTAswere very different, due to their distinct CO2-philic character caused by dissimilar polarities: the acetylderivatives approached a plateau near 200 bar and 5% ethanol, while the free TTAs extraction alwaysincreased in the range of conditions studied.

esponse surface method

. Introduction

During the last decade the interest in biosourced productsbtained from biomass, particularly biomass wastes, has increased.uch renewable resources can be transformed into chemicals,nergy, transportation fuels, and materials, being on the basis ofhe concept of biorefinery [1,2]. In the long term it is expectedhat many crops will be grown to supply future biorefineries.onetheless, presently, the development of this concept is largelyoncentrated in the exploitation of biomass by-products from exist-ng and well established agro-forest industries.

Eucalyptus spp. are extensively used for the pulp and paperroduction with Eucalyptus grandis and Eucalyptus urograndis ashe preferred species in South America (3.75 million ha of planta-ion), and Eucalyptus globulus in Iberian countries (1.29 million haf plantation [3]). Pulp and paper mills are particularly interestedn the integrated exploitation of plants biomass, since they pro-uce considerable amounts of residues, such as bark and general

ogging wastes (e.g., leaves, branches, fruits, sawdust), which areither left in the forest for soil nutrition or ultimately burned in theiomass boiler for energy production [4]. Some of these residues

ontain high value components that can be extracted before burn-ng, i.e. without affecting the current outputs (pulp and power) ofhe existent mills. The exploitation of valuable extractives, such as

∗ Corresponding author. Tel.: +351 234 401549; fax: +351 234 370084.E-mail address: carlos.manuel@ua.pt (C.M. Silva).

896-8446/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.supflu.2012.12.005

© 2012 Elsevier B.V. All rights reserved.

phytosterols, namely ˇ-sitosterol [5,6], lignans [7], and betulin [8],from by-products (e.g., bark, wood knots, pulping liquors) of woodindustrial processing are examples of this strategy implemented insome pulp industries.

In the last few years, we have been investigating the potentialof Eucalyptus spp. residues in order to extend the concept “fromwaste to value” to pulp and paper mills [4,9–16]. In these stud-ies it is reported that the outer barks of several Eucalyptus speciescontain high amounts of triterpenoids (5.2–24.6 g/kg of bark,depending on the species), along with monoterpenes and sesquiter-penes, followed by smaller amounts of fatty acids, fatty alcohols,and aromatic compounds [4,9–11]. The triterpenoids fractionis mainly composed of triterpenic acids (TTA), namely, betu-lonic, betulinic, 3-acetylbetulinic, ursolic, 3-acetylursolic, oleanolicand 3-acetyloleanolic acids (Fig. 1). These TTAs exhibit a widerange of biological activities being recognized as promising com-pounds for the development of new multi-targeting bioactiveagents [17–20]. One may cite, for instance, antitumor [21], anti-HIV [22], antibacterial [23] and anti-inflammatory [24] activities.Hence, their extraction may justify a “high-value low-volume”output approach in pulp mill integrated biorefineries. In thiswork, the supercritical fluid extraction (SFE) of triterpenic acidsfrom E. globulus is focused due to their large abundance inthis species particularly the economically interesting ursolic acid

[9,10].

It is vital for future biorefineries to be based on greentechnologies, where novel solvents like supercritical carbon diox-ide (SC-CO2) and ionic liquids are highly desirable [25–27]. In

106 R.M.A. Domingues et al. / J. of Supercritical Fluids 74 (2013) 105– 114

ids ide

pttsty

tborTwiyCmSwet

pcpeipirt

2

2

a8it2umfm

2

w

Fig. 1. Main triterpenic ac

articular, SC-CO2 has been successfully employed in the extrac-ion of numerous vegetable matrices [15,16,28–35] mainly dueo its favorable transport properties [36–40] and tunable solventtrength, allied to low price and non-toxicity. Several reviews onhe SFE modeling and applications have been published over theears [34,35,41,42].

The SC-CO2 has been successfully applied in the extraction ofriterpenes from several vegetable raw materials as, for instance,etulin from the bark of Betula platyphylla [25] or ursolic andleanolic acids from the seeds of Plantago major L. [26]. We haveecently published [14] some preliminary results on the SFE ofTAs from E. globulus deciduous bark with pure and modified CO2,here it is demonstrated the chief influence of pressure and the

mportant role played by ethanol (cosolvent) upon the extractionields. Regarding the same species, a detailed study using pure SC-O2 has been also published, where the influence of solubility isodeled and specifically focused [15]. The kinetic behavior of the

FE of E. globulus has been also addressed in a different paper [16],here extraction curves were measured and modeled using simple

xpressions from the literature in order to devise the relevant massransfer limitations of the process.

These results induced us to carry out a detailed experimentalrogramme with the objective of evaluating the individual andombined effects of the SFE operating conditions (pressure, tem-erature, and cosolvent content) upon the yield and quality of thextracts of E. globulus deciduous bark. Such quality was measuredn terms of the concentration of TTAs in the final extracts. For thisurpose, a design of experiments (DOE) was followed in order to

dentify the main factors and interactions influencing the measuredesponses. The response surface methodology (RSM) was adoptedo establish validated response functions.

. Materials and methods

.1. Bark samples

Deciduous bark of E. globulus was randomly harvested from 20-year-old clone plantation cultivated in Eixo (40◦37′13.56′′N,◦34′08.43′′W), region of Aveiro, Portugal. The bark was then dried

n an oven at 40 ◦C during approximately 72 h, reaching final mois-ure content between 2 and 5%, milled to granulometry lower than

mm prior to extraction, and stored in hermetically sealed bagsntil use. Deciduous bark was selected as substrate, since it isostly outer bark (which favors the TTAs concentration) and avoids

elling a large number of trees to ensure a continuous supply of rawaterial of controlled origin for a long-term study.

.2. Chemicals

Nonacosan-1-ol (98% purity) and ˇ-sitosterol (99% purity)ere purchased from Fluka Chemie (Madrid, Spain); ursolic

ntified in E. globulus bark.

acid (98% purity), betulinic acid (98% purity), and oleanolicacid (98% purity) were purchased from Aktin Chemicals(Chengdu, China); betulonic acid (95% purity) was purchasedfrom CHEMOS GmbH (Regenstauf, Germany); palmitic acid(99% purity), dichloromethane (99% purity), pyridine (99%purity), bis(trimethylsilyl)trifluoroacetamide (99% purity),trimethylchlorosilane (99% purity), and tetracosane (99% purity)were supplied by Sigma Chemical Co. (Madrid, Spain). Carbondioxide was supplied with a purity of 99.95% from Praxair (Porto,Portugal).

2.3. Soxhlet extraction

Samples of E. globulus deciduous bark (around 25 g) wereSoxhlet-extracted with dichloromethane for 7 h. The solvent wasevaporated to dryness, the extracts were weighed, and the resultswere expressed in percent of dry bark. The results from Soxhletextraction were used as reference to measure the efficiency of thesupercritical extractions. Dichloromethane was chosen because itis a fairly specific solvent for lipophilic extractives.

2.4. Supercritical fluid extractions

Supercritical extraction experiments were performed in anapparatus purchased from Applied Separations (USA) (seeFig. S1 in Supplementary Data). In this system, the liquid CO2 takenfrom a cylinder is compressed to the desired extraction pressureby means of a cooled liquid pump after which it goes through aCoriolis mass flow meter used to measure the CO2 flow rate. Theliquid stream is then heated to the operating temperature in avessel placed before the extractor column. The solvent in the super-critical state then flows through the extractor where the samplewas previously loaded. Afterwards, the effluent from the extrac-tor is depressurized in a heated backpressure regulator valve andbubbled in ethanol to capture the extract for subsequent GC–MSanalysis. The spent CO2 is vented to the atmosphere. The additionof cosolvent to the system is accomplished by a liquid pump (LabAl-liance Model 1500) coupled to the gas line between the mass flowmeter and the heating vessel placed before the extractor column, inorder to mix both CO2 and cosolvent before feeding the extractionvessel. The cosolvent flow rate is controlled by the liquid pump.

In each run, 70–80 g of milled bark was introduced in the extrac-tion vessel. A constant CO2 mass flow rate of 6 g min−1 was used inall extractions during 6 h, totalizing 2.2 kg of spent carbon dioxide.Table 1 lists the remaining experimental conditions: temperature,

T, pressure, P, and cosolvent (ethanol:EtOH) added to CO2. The sol-vent was evaporated first in a rotary evaporator, then in a nitrogenstream and finally in a vacuum oven at 60 ◦C for 6 h. The extractswere weighed and the results expressed in percentage of dry bark.

R.M.A. Domingues et al. / J. of Supercri

Table 1Codification and levels of the three independent variables considered for the SFEfactorial design of experiments.

Variable Level correspondence

Low (−1) Medium (0) High (+1)

Pressure (XP) 100 bar 150 bar 200 bar

2

mdtmir

(wrvsTat3cs

aa

X

wrkma

Y

wlc

u(aiocu

arr3rtm4

Temperature (XT) 40 ◦C 50 ◦C 60 ◦CEthanol content (XEtOH) 0.0% 2.5% 5.0%

.5. Design of experiments and response surface methodology

RSM is a useful used tool for analyzing the relationships betweeneasured responses (dependent variables) and factors (indepen-

ent variables), allowing to minimize experimentation and leadingo correlations that can be used for optimization purposes [43],

ainly when rigorous modeling is difficult to apply. In recent yearst is being adopted to the SFE of solutes from distinct matrixes (e.g.,eferences [12,44,45]).

In this work, it was studied the influence of temperature40–60 ◦C), pressure (100–200 bar) and cosolvent addition (0–5%t.) upon the extraction yield, triterpenic acids recovery, and their

espective concentrations in extracts. The remaining independentariables (e.g., solvent flow rate, extraction time, and bark particleize) were kept constant during the experimental procedures. Inable 1, the levels of the independent variables under investigationre listed. In the whole, 26 experiments have been carried out inhis essay. Nonetheless, in order to perform with advantage a full3 factorial design of experiments, 6 results from a previous publi-ation [15] have been also included to complement and enrich thetatistical analysis of our work.

The three levels defined in Table 1 for each factor were codifiedccording to Eq. (1), so that their ranges of variation lie between −1nd 1:

k = xk − x0

�xk(1)

here Xk is the coded value of the independent variable xk, x0 is itseal value at the center point, and �xk is the step change in variable. In a system involving three significant independent variables, theathematical relationship of each response on these variables is

pproximated by the general quadratic polynomial:

= ˇ0 + ˇ1XP + ˇ2XT + ˇ3XEtOH + ˇ11X2P + ˇ22X2

T + ˇ33X2EtOH

+ ˇ12XPXT + ˇ13XPXEtOH + ˇ23XT XEtOH (2)

here Y is the response, ˇ0is a constant, ˇ1, ˇ2 and ˇ3 are theinear coefficients, ˇ12, ˇ13 and ˇ23 are the interaction or crossedoefficients, and ˇ11, ˇ22and ˇ33 are the quadratic coefficients.

STATISTICA software (version 5.1, StatSoft Inc., Tulsa, USA) wassed for statistical treatment of the results. Analysis of varianceANOVA) was employed to assess the statistically significant factorsnd interactions using Fisher’s test and its associated probabil-ty p(F), while t-tests were performed to judge the significancef the estimated coefficients in each model. The determinationoefficients, R2, and their adjusted values, R2

adj, were used to eval-ate the goodness of fit of the regression models.

In order to check the reproducibility of our experiments and thepplicability of the models for subsequent predictions, additionaluns were performed: (i) replicates at [200 bar, 60 ◦C, 0.0% EtOH;uns 32 and 33 in Table 2] and [200 bar, 60 ◦C, 5.0% EtOH; runs4 and 35 in Table 2] for the reproducibility assessment, and one

◦

un at [170 bar, 50 C, 2.5% EtOH, run 31, see Table 2] to constructhe validation set of the model. This set includes also three experi-

ents retrieved from previous work [15], at [140 bar, 0% EtOH, and0/50/60 ◦C; runs 28–30 in Table 2].

tical Fluids 74 (2013) 105– 114 107

2.6. GC–MS analyses

Before each GC–MS analysis, nearly 20 mg of dried sample wereconverted into trimethylsilyl (TMS) derivatives according to the lit-erature [11]. GC–MS analyses were performed using a Trace GasChromatograph 2000 Series equipped with a Thermo ScientificDSQ II mass spectrometer, using helium as carrier gas (35 cm s−1),equipped with a DB-1 J&W capillary column (30 m × 0.32 mm i.d.,0.25 �m film thickness). The chromatographic conditions were asfollows: initial temperature: 80 ◦C for 5 min; temperature rate of4 ◦C min−1 up to 260 ◦C and 2 ◦C min−1 till the final temperatureof 285 ◦C; maintained at 285 ◦C for 10 min; injector temperature:250 ◦C; transfer-line temperature: 290 ◦C; split ratio: 1:50. The MSwas operated in the electron impact mode with electron impactenergy of 70 eV and data collected at a rate of 1 scan s−1 over arange of m/z 33–700. The ion source was maintained at 250 ◦C.

For quantitative analysis, the GC–MS instrument was cali-brated with pure reference compounds, representative of themajor lipophilic extractives components (namely, palmitic acid,nonacosan-1-ol, ˇ-sitosterol, betulinic acid, ursolic acid and,oleanolic acid), relative to tetracosane, the internal standard used.The respective multiplication factors needed to obtain correctquantification of the peak areas were calculated as an average ofsix GC–MS runs. Compounds were identified, as TMS derivatives,by comparing their mass spectra with the GC–MS spectral library(Wiley, NIST Mass Spectral Library 1999), with data from the lit-erature [11,46,47] and, in some cases, by injection of standards.Each sample was injected in triplicate. The results presented arethe average of the concordant values obtained for each sample (<5%variation between injections of the same sample).

3. Results and discussion

The main objective of this work is to study the SC-CO2 extrac-tion of E. globulus deciduous bark maximizing TTAs yield andtheir concentration in extracts. Taking into account the struc-tures of the TTAs (Fig. 1), particularly the different polaritiesbetween free TTAs (ursolic, oleanolic, betulinic, and betulonicacids) and acetylated TTAs (3-acetylursolic, 3-acetyloleanolic acidsand 3-acetylbetulinic), they exhibit distinct solubilities in SC-CO2.Accordingly, five responses of interest were studied in parallelbased on the 33 full factorial DOE, specifically total extraction yield,TTAs yield, their respective free and acetylated yields, and TTAsconcentration.

The results of the 35 SFE runs considered in this work are col-lected in Table 2, along with the three independent variables incoded form. Globally, there are 21 new experiments for the DOE, 4for reproducibility test, and 1 for the validation of the models; theremaining 6 results were taken from de Melo et al. [15].

Within the experimental space considered, the total extractionyield ranged from 0.04% (wt.) in run 7 [100 bar, 60 ◦C, 0% EtOH] to1.2% (wt.) in run 21 [200 bar, 40 ◦C, 5% EtOH]. Regarding the extrac-tion of TTAs, experimental runs led to a minimum of 0.05 g/kg ofbark and a maximum of 4.95 g/kg of bark, which correspond to 0.7and 76.8% of the reference value measured by conventional Soxhlet,respectively. The TTAs concentration in the extracts ranged from5.7% in runs 4 [100 bar, 50 ◦C, 0% EtOH] and 16 [100 bar, 60 ◦C, 2.5%EtOH] to 43.0% and 43.1% obtained in runs 27 [200 bar, 60 ◦C, 5%EtOH] and 24 [200 bar, 50 ◦C, 5% EtOH], respectively.

In order to investigate the main factors and interactions influ-encing the five responses of interest, and to evaluate their trends

upon variation of the factors within the ranges considered, poly-nomial regressions were fitted following the RSM. The regressioncoefficients obtained for each response using coded variables, arelisted in Table 3. The statistically significant coefficients at 95%

108 R.M.A. Domingues et al. / J. of Supercritical Fluids 74 (2013) 105– 114

Table 2Results of the SFE of E. globulus bark samples considered in this work.

Exp Objectivea XP XT XEtOH Extraction yield (%, wt.) Yield TTA (mg/kgbark) TTA concent. (%, wt.)

Free Acetylated Total

1b DOE −1 −1 −1 0.27 67.1 368.3 435.4 16.122 DOE 0 −1 −1 0.46 339.4 764.5 1103.9 23.893b DOE 1 −1 −1 0.57 403.0 1437.3 1840.3 32.044b DOE −1 0 −1 0.09 22.6 28.4 51.0 5.675 DOE 0 0 −1 0.46 353.5 729.9 1083.4 23.626b DOE 1 0 −1 0.53 690.6 1300.9 1991.5 37.617b DOE −1 1 −1 0.04 17.7 27.8 45.6 11.398 DOE 0 1 −1 0.33 169.1 425.6 594.7 18.029b DOE 1 1 −1 0.77 495.1 1044.7 1539.7 20.00

10 DOE −1 −1 0 0.50 535.3 1158.6 1693.9 33.9211 DOE 0 −1 0 0.81 1252.8 2080.0 3332.8 41.3512 DOE 1 −1 0 0.86 1339.2 1987.4 3326.6 38.8413 DOE −1 0 0 0.34 261.0 630.9 891.9 26.1214 DOE 0 0 0 0.66 894.2 1461.8 2356.0 35.6715 DOE 1 0 0 0.83 1330.6 2082.2 3412.8 40.9716 DOE −1 1 0 0.14 26.4 51.2 77.6 5.7117 DOE 0 1 0 0.71 738.4 1390.5 2128.9 29.8518 DOE 1 1 0 0.92 1083.0 1922.7 3005.7 32.5619 DOE −1 −1 1 0.70 1164.5 1715.7 2880.3 41.1520 DOE 0 −1 1 0.91 1783.9 1785.9 3569.8 39.1421 DOE 1 −1 1 1.23 2744.2 2205.1 4949.4 40.2422 DOE −1 0 1 0.56 747.8 1060.1 1807.8 32.2823 DOE 0 0 1 0.92 1412.4 1599.1 3011.6 32.6624 DOE 1 0 1 1.03 2285.5 2156.2 4441.7 43.1225 DOE −1 1 1 0.14 136.2 200.0 336.2 24.0226 DOE 0 1 1 0.79 1207.3 1436.0 2643.3 33.3727 DOE 1 1 1 1.00 2227.5 2114.8 4342.3 43.0028b Valid. −0.2 −1 −1 0.43 296.7 657.9 954.6 22.229b Valid. −0.2 0 −1 0.41 286.7 599.9 886.6 21.630b Valid. −0.2 1 −1 0.26 176.8 325.0 501.8 19.331 Valid. 0.4 0 0 0.68 1040.0 1368.4 2408.4 35.432 Reprod. 1 1 −1 0.74 498.6 998.2 1496.8 20.133 Reprod. 1 1 −1 0.76 520.1 1101.6 1621.7 21.234 Reprod. 1 1 1 0.94 2000.5 1811.2 3811.7 40.5

t.

cq(cb

e

tf

TRm

35 Reprod. 1 1 1 1.01

a DOE = design of experiments; Valid. = validation set; Reprod. = reproducibility seb Retrieved from [15].

onfidence level are highlighted in bold. Concerning the fitting ade-uacy, the obtained models represented well the experimental dataR2 in the range of 0.857–0.983). The regression for the TTAs con-entration was the only which led to determination coefficients

2 2

elow 0.9, namely R = 0.857 and Radj = 0.782. If compared within

ach response, the adjusted determination coefficients (R2adj), that

ake into account both R2 and the degrees of freedom, did not dif-er notably from their corresponding R2. It is known that when

able 3egression coefficients of the quadratic model given by Eq. (2), their individual significaodel (FM). (Values in bold represent significant coefficients.)

Regressedcoefficients of Eq. (2)

Total extraction yield TTAs extraction yield

Free TTAs

FM p FM P

ˇ0 0.688148 <0.001 865.6926 <0.001

ˇ1 0.275556 <0.001 534.45 <0.001

ˇ2 −0.08167 <0.001 −196.039 <0.001

ˇ3 0.208889 <0.001 619.5111 <0.001

ˇ12 0.098333 <0.001 75.48333 0.047

ˇ13 0.0325 0.140 310.6167 <0.001

ˇ23 −0.0625 0.008 −166.167 <0.001

ˇ11 −0.08778 0.009 −40.2611 0.430

ˇ22 0.017222 0.569 −14.7944 0.770

ˇ33 −0.04111 0.184 74.75556 0.152

R2 0.965 0.983

R2adj 0.947 0.974

2103.5 1978.2 4081.7 40.4

R2and R2adj differ considerably the model is prone to include non-

significant terms [43].The coefficients that were not significant (t-test, 17 degrees of

freedom, p > 0.05) were removed from the general polynomial, Eq.

(2). Then the final models were refitted to data and converted touncoded variables, namely, extraction pressure (in bar), temper-ature (in ◦C), and content of ethanol added to SC-CO2 (%, wt.).The reduced experimental equations are shown in Table 4. In all

nce at 95% confidence level, and respective determination coefficients for the full

TTAs concentration

Acetylated TTAs Total

FM p FM p FM p

1486.411 <0.001 2352.211 <0.001 33.93259 <0.001611.6833 <0.001 1146.128 <0.001 7.333333 <0.001

−271.639 <0.001 −467.794 <0.001 −4.93167 0.001452.5278 <0.001 1072.039 <0.001 7.812222 <0.001201.3333 0.003 276.8 0.004 2.875833 0.072

11.825 0.844 322.4667 0.001 −2.29667 0.144−73.6583 0.230 −239.733 0.011 0.208333 0.891

−103.017 0.235 −143.45 0.242 −1.68778 0.4370.95 0.991 −13.4833 0.911 −1.71278 0.430

−285.017 0.003 −210.517 0.093 −2.92444 0.186

0.947 0.973 0.8570.919 0.959 0.782

R.M.A. Domingues et al. / J. of Supercritical Fluids 74 (2013) 105– 114 109

Table 4Reduced experimental models (RM) fitted to the responses of the 27 DOE runs of Table 2.

Response Reduced experimental models Equation R2 R2adj

Extraction yield(%, wt.)

Y1 = 0.43343 − 0.031412T + 0.0062111P + 0.20826EtOH − 0.000035111EtOH2 + 0.000019667P −0.0025000TEtOH

(3) 0.956 0.943

Free TTAs(mg/kgbark)

Y2 = 869.43 − 25.632T − 3.0717P + 207.40CoSolv − 6.6467TEtOH + 2.4849PEtOH (4) 0.980 0.974

Acetylated TTAs(mg/kgbark)

Y3 = 3224.0 − 87.564T − 7.8997P − 45.603EtOH2 + 409.02EtOH + 0.40267PT (5) 0.938 0.923

Total TTAs(mg/kgbark)

Y4 = 3856.5 − 105.85T − 11.207P + 521.54EtOH + 0.5536PT − 9.5893TEtOH + 2.5797PEtOH (6) 0.966 0.955

TTAs concentration(mg/kgbark)

Y5 = 24.562 − 0.49317T + 0.14667P + 3.1249EtOH (7) 0.780 0.751

F (a) tote .

mcoiasirsbmpits(a(ce

3

ymgtTw

% EtOH) (p < 0.001, as well as the interactions P–T (p < 0.001) andT–%EtOH (p < 0.008), affected the total extraction yield significantly(see Table 3 or the reduced model (Eq. (3)) in Table 4).

ig. 2. Response surfaces plotting the effects of pressure and temperature over:

xperimental data, and surfaces are given by Eqs. (3) and (6) (Table 4), respectively

odels the individual factors (pressure, temperature and ethanolontent) revealed to be statistically significant. From the signalf the coefficients, it may be figured out the positive or negativempact of these factors upon the responses. In all cases, pressurend ethanol addition favored the results, while temperature waseen to unfavorably affect all responses. It is well known that ansobaric increase of temperature decreases the SC-CO2 density, thuseducing its solvent power. On other hand, the vapor pressure ofolutes increases with temperature, which increments their solu-ilities in the supercritical solvent. Taking into account that theeasured extraction yields evidence no enhancement with tem-

erature (at the same pressure), it is possible to conclude thatn the range of our experimental conditions the density reduc-ion prevails over vapor pressure increase. This behavior can beeen in Fig. 2, where the total extraction (Fig. 2a) and the TTAsFig. 2b) yields are plotted as functions of P and T for 5% EtOHddition. However, this trend is more pronounced at low pressure100 bar) due to the large compressibility of CO2 near the criti-al point. At 200 bar this effect is much more attenuated but notliminated.

.1. Total extraction yield

Fig. 3 shows the fitted response surface for the total extractionield (Eq. (3) in Table 4) at 40 ◦C, where dots represent experi-ental data. One may conclude that the regression model is in

ood agreement with the lab results, which is also confirmed byhe determination coefficients found (R2 = 0.956 and R2

adj = 0.943).he unbiased distribution of the points near the diagonal in Fig. 4,here predicted yields are plotted against the measured values,

al extraction yield, and (b) triterpenic acids yield, for 5% EtOH content. Dots are

also confirms this assertion. This trend was also observed in theother responses studied. All the independent variables (P, T, and

Fig. 3. Response surface showing the effect of pressure and ethanol content on thetotal extraction yield at 40 ◦C. Dots are experimental data, and surface is given byEq. (3) (Table 4).

110 R.M.A. Domingues et al. / J. of Supercri

Fig. 4. Measured versus predicted total extraction yield (%, wt).

Fig. 5. Response surfaces at 40 ◦C showing the effects of pressure and ethanol content onand (c) individual contribution of free TTAs. Dots are experimental data, and surfaces are

tical Fluids 74 (2013) 105– 114

As previously referred, both P and % EtOH have a positive effectupon the extraction yield. By comparing the results measured at200 bar and 0% EtOH (0.57%, wt.) and 100 bar and 5% EtOH (0.76%,wt.), one conclude that 5% EtOH leads to an extraction enhancementhigher than that achieved by a 100 bar increment. This behavioremphasizes the chief role played by cosolvent due to the polaritymodification imparted to the supercritical solvent.

Despite the apparent benefit of cosolvent upon the extractionyield, it may increase the solubility of both valuable and nonva-luable compounds. For this reason, detailed responses regardingTTAs extraction and their concentration will be analyzed in the nextsections.

3.2. Triterpenic acids extraction yield

The fitted surfaces and experimental data of the overall TTAs

extraction yield (Eq. (6) in Table 4) and their individual subgroups– free TTAs (Eq. (4) in Table 4) and acetylated TTAs (Eq. (5) inTable 4) – are illustrated in Fig. 5a–c, respectively, for the constanttemperature of 40 ◦C.

the (a) total TTAs extraction yield, (b) individual contribution of acetylated TTAs, given by Eqs. (6), (5) and (4) (Table 4), respectively.

R.M.A. Domingues et al. / J. of Supercritical Fluids 74 (2013) 105– 114 111

F iterpe ◦ ◦

e

tosttbm2d[7t

(acipsaSniTCrt

fatrott

3

ea

ig. 6. Response surface showing the effect of pressure and ethanol content on trxperimental data, and surface is given by Eq. (7) (Table 4).

The overall TTAs extraction yield clearly depends not only onhe ethanol content (p < 0.001) and pressure (p < 0.001) but alson their combined effect (p < 0.001), in a very parallel way to thateen above for total extraction yield. It should be pointed outhat experiments without cosolvent addition moderately uptakehis family of compounds comparatively to the removal achievedy Soxhlet extraction with dichloromethane. For instance, theaximum and minimum values obtained without cosolvent are

.00 [200 bar, 50 ◦C] and 0.05 g/kg [100 bar, 60 ◦C], while the intro-uction of ethanol raises them to 4.95 [200 bar, 40 ◦C] and 0.34100 bar, 60 ◦C], respectively; these values correspond to 30.9, 0.7,6.9 and 5.2% of the TTAs yield registered by Soxhlet extrac-ion.

With regard to the individual contributions of the free TTAsFig. 5b) and acetyl derivatives (Fig. 5c) to the whole triterpeniccids family at 40 ◦C, considerable differences between their trendsan be perceived: while the extraction of acetylated moleculesncrease with pressure and cosolvent content until reaching alateau near 200 bar and 5% EtOH, the free TTAs removal does nottabilize within the range of conditions studied. Acetylated TTAsre extracted in large extent with pure CO2, attaining 65.6% of theoxhlet reference value at 200 bar and 40 ◦C, because of their higheronpolar character and thus higher affinity to SC-CO2. This behav-

or is confirmed by their contributions to the overall extraction ofTAs: there are 78% of acetylated acids at 100 bar and 85% at 200 bar.onsequently, without the aid of EtOH, the global TTAs extractionemains reliant on the acetylated TTAs contribution, as the yield ofhe free acids is too low to play a relevant role.

On the other hand, the increased polarity imparted by ethanolavors molecular interactions between the supercritical solventnd the more polar triterpenoids, increasing the removal of freeriterpenic acids (2.74 g/kg of bark for 200 bar, 40 ◦C, 5% EtOH, cor-esponding to 62.8% of the Soxhlet extraction value) and thus theverall TTAs extraction yield. This subgroup is the most abundant ofhe TTAs existent in the bark and represents 67.7% of their respec-ive Soxhlet content.

.3. Triterpenic acids concentration in extracts

Fig. 6a and b plots the experimental TTAs concentration in thextracts along with the regressed surfaces (Eq. (7) in Table 4) at 40nd 60 ◦C, respectively.

nic acids concentration in the extracts for: (a) T= 40 C, and (b) T= 60 C. Dots are

Despite the lower correlation coefficients achieved (Table 4),the fitted model reasonably describe the experimental resultsand trends. The maximum concentration predicted by the modelrounds 50% (wt.) and it is achieved at 200 bar, 40 ◦C, 5% EtOH(Fig. 6a). This result is overestimated since the experimental con-centration is 43.0% (wt.). On the other hand, the minimum at 40 ◦Ctotals 16% (wt.), being obtained at 100 bar without ethanol addi-tion. The pressure range of 100 to 150 bar, and 0% EtOH correspondto the poor extraction conditions in terms of purity: between 15and 25% only.

The aforepresented results revealed to be possible to boost TTAsextraction using the combination of both pressure and ethanoladdition, as they expose that the same conditions that favor theincrease of the total extraction yield also lead to an enhancementof the TTA concentration in the extracts.

3.4. Reproducibility and validation tests

As formerly stated, some runs were carried out to test theexperimental reproducibility and assess the validation of regressedmodels. In this respect, two replicates of runs 9 and 27 wereconsidered (runs 32–35), while runs 28–31 were used for vali-dation purposes. The individual responses are compiled in Table 5,together with the calculated averages (y) and standard deviations(SD) for the reproducibility set, and the average absolute relativedeviations (AARDs) for the validation set.

With regard to the reproducibility tests, the results show that,in general, the standard deviations are within 2–8% of the aver-ages, indicating that the sample-to-sample reproducibility is veryacceptable considering that natural raw materials are involved inthe experiments.

With regard to the models validation, it is possible to observethat they provide reliable estimates of the measured data. Note thatthe validation set is included in the experimental space but wasestablished for conditions not considered in the design of exper-iments (e.g., 140 bar, 170 bar, 2.5% EtOH). The AARDs found were15.1, 11.7, 21.2, 11.5 and 11.8% for the global, free TTAs, acetylatedTTAs, total TTAs extraction yields and TTAs concentration in the

extracts, respectively. In the whole, the average deviation of 14.3%found confirms the acceptable predictive capability of the regressedequations, particularly if we take into account that they are strictreduced models (Table 4) and because of the aforementioned

112R

.M.A

. D

omingues

et al.

/ J.

of Supercritical

Fluids 74 (2013) 105– 114

Table 5Calculated results for the reproducibility and validation tests.

Run P (bar) T (◦C) EtOH (%, wt.) Total extraction yield (%, wt.) TTAs yield (mg/kgbark) TTAs concentration (%, wt.)

Free TTAs Acetylated TTAs TotalPredict. (Eq. (3)) Exp Predict. (Eq. (4)) Exp Predict. (Eq. (5)) Exp Predict. (Eq. (6)) Exp Predict. (Eq. (7)) Exp

Validation set28a 140 40 0 0.44 0.43 259.5 296.7 870.4 657.9 1153.7 954.6 25.4 22.229a 140 50 0 0.40 0.41 214.6 286.7 558.5 599.9 870.2 886.6 20.4 21.630a 140 60 0 0.37 0.26 169.6 176.8 246.6 325.0 586.8 501.8 15.5 19.331 170 50 2.5 0.77 0.68 1092.6 1040.0 1663.1 1368.4 2566.0 2408.4 32.7 35.4

Average absolute relative deviations AARD = 15.1% AARD = 11.7% AARD = 21.2% AARD = 11.5% AARD = 11.8%

Reproducibility set9a 200 60 0 0.77 583.3 1159.5 1742.8 22.6

32 200 60 0 0.74 498.6 998.2 1496.8 20.133 200 60 0 0.76 520.1 1101.6 1621.7 21.2

y = 0.76 y = 534.0 y = 1086.4 y = 1620.4 y = 21.3SD = 0.02 SD = 44.0 SD = 81.7 SD = 123.0 SD = 1.3

27 200 60 5 1.01 2227.5 2114.8 4242.3 43.134 200 60 5 0.94 2000.5 1811.2 3811.7 40.535 200 60 5 1.01 2103.5 1978.2 4081.7 40.4

y= 0.98 y= 2110.5 y= 1968.1 y= 4045.2 y= 41.3SD = 0.05 SD = 113.6 SD = 152.1 SD = 217.6 SD = 1.5

a Retrieved from [15]; RM = reduced model; Exp = experimental value; SD = standard deviation; y = average; AARD = average absolute relative deviations.

percri

nane

4

l((wevt(oeTwattS

A

2ftpC

A

f2

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

R.M.A. Domingues et al. / J. of Su

ature of the biomass. In numerous works in the literature, authorsccomplish the calculations with full models, despite containingon-significant terms, which obviously improve the fittings andstimates.

. Conclusions

In this work, the supercritical CO2 extraction of E. globu-us deciduous bark was carried out at different temperatures40–60 ◦C), pressures (100–200 bar), and cosolvent (EtOH) contents0.0–5.0 wt. %) in order to select the optimum conditions that,ithin the ranges considered, maximize triterpenic acids recov-

ry and their concentration in extracts. The optimized independentariables were: 200 bar, 40 ◦C and 5% EtOH. Under these condi-ions, the regressed models provided an extraction yield of 1.2%wt.), TTAs concentration of 50%, which corresponds to TTAs yieldf 5.1 g/kg of bark and a recovery of 79.2% in comparison to ref-rence Soxhlet extraction. The behavior of the free and acetylatedTAs extraction was significantly different. The acetyl derivativesere extensively removed and tended to a plateau near 200 bar

nd 5% EtOH, while the free acids extraction always increased inhe range of conditions studied. These results may be attributed tohe inferior polarity of acetylated TTAs, and thus higher affinity toC-CO2, when compared to the unesterified triterpenoids.

cknowledgments

We acknowledge the 7th Framework Programme FP7/2007–013 for funding project AFORE: Forest Biorefineries: Added-valuerom chemicals and polymers by new integrated separation, frac-ionation and upgrading technologies (CP-IP 228589-2), the BIIPProject (QREN 11551) for awarding a grant to R.M.A.D., and toICECO (Pest-C/CTM/LA0011/2011).

ppendix A. Supplementary data

Supplementary data associated with this article can beound, in the online version, at http://dx.doi.org/10.1016/j.supflu.012.12.005.

eferences

[1] S. Fernando, S. Adhikari, C. Chandrapal, N. Murali, Biorefineries: current status,challenges, and future direction, Energy & Fuels 20 (2006) 1727–1737.

[2] J.E. Holladay, J.F., White, J.J., Bozell, D. Johnson. Top value-added chemicals frombiomass: vol II—results of screening for potential candidates from biorefinerylignin, 2007. Available from: http://www.pnl.gov/main/publications/external/technical reports/PNNL-16983.pdf

[3] G.I. Trabado, D. Wilstermann, Eucalyptus universalis, Global Cultivated Euca-lypt Forest Map 200 (2008) 8, Available from: http://www.git-forestry.com/

[4] R.M.A. Domingues, G.D.A. Sousa, C.S.R. Freire, A.J.D. Silvestre, C.P. Neto,Eucalyptus globulus biomass residues from pulping industry as a source ofhigh value triterpenic compounds, Industrial Crops and Products 31 (2010)65–70.

[5] P. Fernandes, J.M.S. Cabral, Phytosterols: applications and recovery methods,Bioresource Technology 98 (2007) 2335–2350.

[6] A. Hamunen. Process for the purification of �-sitosterol isolated from theunsaponifiables in crude soap from the sulphate cellulose process, 1983. USPatent 4422974.

[7] S.P. Pietarinen, S.M. Willfor, M.O. Ahotupa, J.E. Hemming, B.R. Holmbom, Knot-wood and bark extracts: strong antioxidants from waste materials, Journal ofWood Science 52 (2006) 436–444.

[8] P.A. Krasutsky, Birch bark research and development, Natural Product Reports23 (2006) 919–942.

[9] R.M.A. Domingues, D.J.S. Patinha, G.D.A. Sousa, J.J. Villaverde, C.M. Silva, C.S.R.Freire, A.J.D. Silvestre, C.P. Neto, Eucalyptus biomass residues from agro-forestand pulping industries as sources of high-value triterpenic compounds, Cellu-

lose Chemistry and Technology 45 (2011) 475–481.

10] R.M.A. Domingues, G.D.A. Sousa, C.M. Silva, C.S.R. Freire, A.J.D. Silvestre, C.P.Neto, High value triterpenic compounds from the outer barks of several Euca-lyptus species cultivated in Brazil and in Portugal, Industrial Crops and Products33 (2011) 158–164.

[

[

tical Fluids 74 (2013) 105– 114 113

11] C.S.R. Freire, A.J.D. Silvestre, C.P. Neto, J.A.S. Cavaleiro, Lipophilic extractivesof the inner and outer barks of Eucalyptus globulus, Holzforschung 56 (2002)372–379.

12] S.A.O. Santos, J.J. Villaverde, C.M. Silva, C.P. Neto, A.J.D. Silvestre, Supercriticalfluid extraction of phenolic compounds from Eucalyptus globulus Labill bark,Journal of Supercritical Fluids 71 (2012) 71–79.

13] D.J.S. Patinha, R.M.A. Domingues, J.J. Villaverde, A.M.S. Silva, C.M. Silva, C.S.R.Freire, C.P. Neto, A.J.D. Silvestre, Lipophilic extractives from the bark ofEucalyptus grandis x globulus, a rich source of methyl morolate: selective extrac-tion with supercritical CO2, Industrial Crops and Products 43 (2013) 340–348.

14] R.M.A. Domingues, E.L.G. Oliveira, C.S.R. Freire, R.M. Couto, P.C. Simoes, C.P.Neto, A.J.D. Silvestre, C.M. Silva, Supercritical fluid extraction of Eucalyptusglobulus bark-A promising approach for triterpenoid production, InternationalJournal of Molecular Sciences 13 (2012) 7648–7662.

15] M.M.R. de Melo, E.L.G. Oliveira, A.J.D. Silvestre, C.M. Silva, Supercritical fluidextraction of triterpenic acids from Eucalyptus globulus bark, Journal of Super-critical Fluids 70 (2012) 137–145.

16] R.M.A. Domingues, M.M.R. de Melo, C.P. Neto, A.J.D. Silvestre, C.M. Silva,Measurement and modeling of supercritical fluid extraction curves of Euca-lyptus globulus bark: influence of the operating conditions upon yieldsand extract composition, The Journal of Supercritical Fluids 72 (2012)176–185.

17] M.N. Laszczyk, Pentacyclic triterpenes of the lupane, oleanane and ursane groupas tools in cancer therapy, Planta Medica 75 (2009) 1549–1560.

18] P. Dzubak, M. Hajduch, D. Vydra, A. Hustova, M. Kvasnica, D. Biedermann,L. Markova, M. Urban, J. Sarek, Pharmacological activities of natural triter-penoids and their therapeutic implications, Natural Product Reports 23 (2006)394–411.

19] N. Sultana, A. Ata, Oleanolic acid and related derivatives as medicinally impor-tant compounds, Journal of Enzyme Inhibition and Medicinal Chemistry 23(2008) 739–756.

20] P. Yogeeswari, D. Sriram, Betulinic acid and its derivatives: a review on theirbiological properties, Current Medicinal Chemistry 12 (2005) 657–666.

21] J. Li, W.J. Guo, Q.Y. Yang, Effects of ursolic acid and oleanolic acid on humancolon carcinoma cell line HCT15, World Journal of Gastroenterology 8 (2002)493–495.

22] F. Mengoni, M. Lichtner, L. Battinelli, M. Marzi, C.M. Mastroianni, V. Vullo,G. Mazzanti, In vitro anti-HIV activity of oleanolic acid on infected humanmononuclear cells, Planta Medica 68 (2002) 111–114.

23] S. Fontanay, M. Grare, J. Mayer, C. Finance, R.E. Duval, Ursolic, oleanolicand betulinic acids: antibacterial spectra and selectivity indexes, Journal ofEthnopharmacology 120 (2008) 272–276.

24] G.B. Singh, S. Singh, S. Bani, B.D. Gupta, S.K. Banerjee, Anti-inflammatory activityof oleanolic acid in rats and mice, Journal of Pharmacy and Pharmacology 44(1992) 456–458.

25] J.H. Clark, S.J. Tavener, Alternative solvents: shades of green, Organic ProcessResearch & Development 11 (2007) 149–155.

26] F.E.I. Deswarte, J.H. Clark, J.J.E. Hardy, Extraction of high-value chemicals fromwheat straw by supercritical carbon dioxide, Abstracts of Papers of the Ameri-can Chemical Society 230 (2005) U61–U62.

27] C. Eckert, C. Liotta, A. Ragauskas, J. Hallett, C. Kitchens, E. Hill, L. Draucker, Tun-able solvents for fine chemicals from the biorefinery, Green Chemistry 9 (2007)545–548.

28] V. Castola, B. Marongiu, A. Bighelli, C. Floris, A. Laï, J. Casanova, Extractives ofcork (Quercus suber L.): chemical composition of dichloromethane and super-critical CO2 extracts, Industrial Crops and Products 21 (2005) 65–69.

29] A. Felföldi-Gáva, S. Szarka, B. Simándi, B. Blazics, B. Simon, A. Kéry, Supercriticalfluid extraction of Alnus glutinosa L. Gaertn, The Journal of Supercritical Fluids61 (2012) 55–61.

30] C.P. Passos, R.M. Silva, F.A. Da Silva, M.A. Coimbra, C.M. Silva, Enhancementof the supercritical fluid extraction of grape seed oil by using enzymaticallypre-treated seed, The Journal of Supercritical Fluids 48 (2009) 225–229.

31] C.P. Passos, R.M. Silva, F.A. Da Silva, M.A. Coimbra, C.M. Silva, Supercritical fluidextraction of grape seed (Vitis vinifera L.) oil. Effect of the operating conditionsupon oil composition and antioxidant capacity, Chemical Engineering Journal160 (2010) 634–640.

32] J. Shi, C. Yi, S.J. Xue, Y. Jiang, Y. Ma, D. Li, Effects of modifiers on the profileof lycopene extracted from tomato skins by supercritical CO2, Journal of FoodEngineering 93 (2009) 431–436.

33] B. Damjanovic, D. Skala, J. Baras, D. Petrovic-Djakov, Isolation of essential oiland supercritical carbon dioxide extract of Juniperus communis L. fruits fromMontenegro, Flavour and Fragrance Journal 21 (2006) 875–880.

34] H. Sovova, R.P. Stateva, Supercritical fluid extraction from vegetable materials,Reviews in Chemical Engineering 27 (2011) 79–156.

35] E. Reverchon, I. De Marco, Supercritical fluid extraction and fractionation ofnatural matter, The Journal of Supercritical Fluids 38 (2006) 146–166.

36] A.L. Magalhaes, S.P. Cardoso, B.R. Figueiredo, F.A. Da Silva, C.M. Silva, Revisitingthe Liu–Silva–Macedo model for tracer diffusion coefficients of supercritical,liquid, and gaseous systems, Industrial & Engineering Chemistry Research 49(2010) 7697–7700.

37] A.L. Magalhães, F.A. Da Silva, C.M. Silva, New models for tracer diffusioncoefficients of hard sphere and real systems: application to gases, liquids andsupercritical fluids, The Journal of Supercritical Fluids 55 (2011) 898–923.

38] A.L. Magalhães, F.A. Da Silva, C.M. Silva, Tracer diffusion coefficients of polarsystems, Chemical Engineering Science 73 (2012) 151–168.

1 percri

[

[

[

[

[

[

[

[

14 R.M.A. Domingues et al. / J. of Su

39] H. Liu, C.M. Silva, E.A. Macedo, New equations for tracer diffusion coefficientsof solutes in supercritical and liquid solvents based on the Lennard-Jones fluid model, Industrial & Engineering Chemistry Research 36 (1997)246–252.

40] A.L. Magalhaes, F.A. Da Silva, C.M. Silva. Free-volume model for the diffu-sion coefficients of solutes at infinite dilution in supercritical CO2 and liquidH2O, The Journal of Supercritical Fluids, http://dx.doi.org/10.1016/j.supflu.2012.12.004, in press.

41] E.L.G. Oliveira, A.J.D. Silvestre, C.M. Silva, Review of kinetic models for super-

critical fluid extraction, Chemical Engineering Research and Design 89 (2011)1104–1117.

42] M.J.H. Akanda, M.Z.I. Sarker, S. Ferdosh, M.Y.A. Manap, N.N.N. Ab Rahman, M.O.Ab Kadir, Applications of supercritical fluid extraction (SFE) of palm oil and oilfrom natural sources, Molecules 17 (2012) 1764–1794.

[

tical Fluids 74 (2013) 105– 114

43] D.C. Montgomery, Design and Analysis of Experiments, fifth ed., Wiley, NewYork, 2000.

44] S.G. Özkal, M.E. Yener, L. Bayındırlı, Response surfaces of apricot kernel oil yieldin supercritical carbon dioxide, LWT – Food Science and Technology 38 (2005)611–616.

45] M.G. Bernardo-Gil, R. Roque, L.B. Roseiro, L.C. Duarte, F. Gírio, P. Esteves,Supercritical extraction of carob kibbles (Ceratonia siliqua L.), The Journal ofSupercritical Fluids 59 (2011) 36–42.

46] A.N. Assimopoulou, V.P. Papageorgiou, GC–MS analysis of penta- and tetra-

cyclic triterpenes from resins of Pistacia species. Part II. Pistacia terebinthus var.Chia, Biomedical Chromatography 19 (2005) 586–605.

47] H. Budzikiewicz, J.M. Wilson, C. Djerassi, Mass spectrometry in structural andstereochemical problems. 32 pentacyclic triterpenes, Journal of the AmericanChemical Society 85 (1963) 3688–3699.