Universidade de Aveiro Departamento de Química

2019

Vanessa Ariana Utilização de solventes alternativos para a extração Azevedo Vieira de compostos bioativos de Juglans regia L.

Using alternative solvents for the extraction of bioactive compounds from Juglans regia L.

Universidade de Aveiro Departamento de Química

2019

Vanessa Ariana Utilização de solventes alternativos para a extração Azevedo Vieira de compostos bioativos de Juglans regia L.

Using alternative solvents for the extraction of bioactive compounds from Juglans regia L. Tese apresentada à Universidade de Aveiro para cumprimento dos requisitos necessários à obtenção do grau de Doutor em Engenharia Química, realizada sob a orientação científica do Professor Doutor João Manuel da Costa e Araújo Pereira Coutinho, Professor Catedrático do Departamento de Química da Universidade de Aveiro, da Professora Doutora Maria Olga de Amorim e Sá Ferreira, Professora Adjunta da Escola Superior de Tecnologia e Gestão do Instituto Politécnico de Bragança, e Professora Doutora Isabel Cristina Fernandes Rodrigues Ferreira, Professora Coordenadora-Principal da Escola Superior Agrária do Instituto Politécnico de Bragança.

Apoio financeiro do POCTI no âmbito O doutoramento agradece o apoio financeiro da FCT e do FSE no âmbito do III do III Quadro Comunitário de Apoio. Quadro Comunitário de Apoio (SFRH/BD/108487/2015). Este trabalho foi desenvolvido no âmbito dos projetos: CICECO-Instituto de Materiais de Aveiro, Refª. FCT UID/CTM/50011/2019, financiado por fundos nacionais através da FCT/MCTES; CIMO, Refª. UID/AGR/00690/2019, financiado por fundos nacionais e europeus; Laboratório Associado LSRE-LCM - UID/EQU/50020/2019 - financiado por fundos nacionais através da FCT/MCTES (PIDDAC); Project AllNat, Refº. POCI-01-0145-FEDER-030463 (PTDC/EQU-EPQ/30463/2017), financiado por fundos FEDER através do COMPETE2020 – Programa Operacional Competitividade e Internacionalização (POCI), Portugal 2020 e por fundos nacionais através da FCT/MCTES.

Ao João

o júri

Presidente Professor Doutor João de Lemos Pinto Professor Catedrático da Universidade de Aveiro

Doutora Maria Arminda Costa Alves Professora Catedrática da Universidade do Porto

Doutor Jesus Simal-Gandra Professor Catedrático da Universidade de Vigo

Doutor Simão Pedro de Almeida Pinho Professor Coordenador do Instituto Politécnico de Bragança

Doutora Mara Guadalupe Freire Martins Investigadora Coordenadora em regime laboral da Universidade de Aveiro

Doutora Maria Olga de Amorim e Sá Professora Adjunta do Instituto Politécnico de Bragança

agradecimentos Primeiramente, dirijo os meus agradecimentos aos meus orientadores. Obrigada pela oportunidade que me deram, pela vossa compreensão, apoio, tempo e paciência. Prof. João Coutinho, muito obrigada por me ensinar a olhar para tudo com outros olhos, muitas das vezes a fazer-me sentir “enorme” mas sempre mostrando que o caminho é em frente. Prof. Olga, obrigada pelo carinho, tempo dedicado, palavras de encorajamento e, sobretudo, por abraçar esta aventura. Prof. Isabel, obrigada por TUDO. Obrigada por estar sempre ao meu lado, por essa objetividade que torna tudo simples e atingível, por alimentar o amor à ciência. Gostaria também de agradecer a todos que os que colaboraram direta ou indiretamente neste percurso, pelo auxílio e partilha de conhecimento, contribuindo para o desenvolvimento destes trabalhos que tenho a oportunidade de apresentar. Em particular, quero agradecer à Dra. Lillian Barros pelo ser especial que é, por seguir cada passo dado e ser uma fonte de soluções. Também quero agradecer generosidade dos “parceiros” Dr. Ricardo C. Calhelha e Dr. Miguel A. Prieto. Um bem-haja a todos! Naturalmente, um carinhoso agradecimento terá que ser dirigido aos meus colegas e amigos dos grupos de investigação PATh, BioChemCore e LSRE- CIMO. Obrigado por serem bons companheiros! Um obrigado especial a todos os amigos, por me encorajarem, por festejarem vitórias... por estarem sempre lá. Obrigada Josy, Cristina, Sónia, Rita, Carlitos e a toda a equipa de ACSP dos SAMS – Centro Sul e Ilhas. Para terminar, quero agradecer à minha família. Aos meus pais por fazerem de mim quem sou. Ao meu irmão por todo o amor e cumplicidade. Aos meus segundos pais. Ao João.

palavras-chave Juglans regia L., compostos fenólicos, extratos bioativos, solventes alternativos

resumo As plantas e os seus resíduos resultantes das indústrias agro-florestal e alimentar são uma potencial fonte de compostos bioativos. Neste contexto, revela-se de grande importância a caracterização química e bioquímica dos extratos de diversas matrizes vegetais avaliando a sua bioatividade em várias dimensões (antioxidante, anti-inflamatória, antimicrobiana, antitumoral, entre outras), bem como a obtenção desses extratos através de processos mais eficientes, sustentáveis e seguros. Em particular, os compostos fenólicos representam a fração de metabolitos secundários de plantas mais amplamente extraída e reconhecida pela sua abundância e bioatividade. Nesta tese, estudou-se o perfil fitoquímico e a bioatividade de duas matrizes vegetais que se destacam como fontes naturais deste tipo de compostos: as folhas da nogueira e os mesocarpos da noz. Assim, após a caracterização química e bioativa da biomassa, otimizou-se o processo de extração de compostos fenólicos, aplicando um planeamento experimental e a metodologia de superfície de resposta, e utilizando etanol como solvente numa primeira fase. Depois, estudaram-se solventes alternativos aos solventes orgânicos voláteis, tais como alcanodióis e solventes eutécticos baseados em cloreto de colina e ácidos carboxílicos ou alcanodióis. Neste último caso, os extratos inseridos no solvente poderão ser utilizados como componentes de formulações com potencial aplicação nas indústrias alimentar, farmacêutica ou cosmética. Finalmente, a escolha dos solventes de extração e purificação de compostos fenólicos é realizada, em geral, de forma empírica, existindo pouca informação sobre a sua solubilidade em solventes orgânicos comuns. Neste sentido, considerou-se oportuno estudar a solubilidade de três ácidos aromáticosde enorme relevância no contexto das biorrefinarias. A modelação termodinâmica do equilíbrio sólido-líquido destas misturas foi efetuada com o modelo NRTL- SAC (Nonrandom Two-Liquid Segment Activity Coefficient) e o modelo de solvatação de Abraham, cujo carácter semi-preditivo permitirá simplificar a seleção preliminar de solventes para um determinado processo.

keywords Juglans regia L., phenolic compounds, bioactive extracts, alternative solvents

abstract Plants and its residues resulting from the agroforest and food industries are potential sources of bioactive compounds. In this context, it is of great importance to carry out the chemical and biochemical characterization of the extracts of various plant materials evaluating their bioactivity in different dimensions (antioxidant, anti-inflammatory, antimicrobial, anti-tumour, among others), as well as to obtain those extracts by designing more efficient, sustainable and safer processes. In particular, phenolic compounds represent the fraction of secondary metabolites of plants most widely extracted and recognized for their abundance and bioactivity. In this thesis, the phytochemical profile and bioactivity of two plant matrices that stand out as natural sources of this type of compounds were studied: the leaves of walnut trees and walnut green husks. Thus, after the chemical and bioactive characterization of the biomass, the extraction process of phenolic compounds was optimized, applying an experimental design assisted by response surface methodology, using ethanol as solvent in a first stage. After, alternative solvents to volatile organic solvents were studied, such as alkanediols and eutectic solvents based on choline chloride and carboxylic acids or alkanediols. In the latter case, the extracts incorporated in the solvent may be used as components of formulations with potential application in the food, pharmaceutical or cosmetic industries. Finally, the choice of solvents for the extraction and purification of phenolic compounds is generally performed empirically, as scarce information on their solubility in common organic solvents is available. Therefore, it was considered appropriate to study the solubility of three arotimatic acids of great relevance in the biorefinery context. The thermodynamic modelling of the solid-liquid equilibrium of these mixtures was performed using the NRTL-SAC model (Nonrandom Two-Liquid Segment Activity Coefficient) and the Abraham's solvation model, whose semi-predictive character will simplify the preliminary selection of solvents for a given process.

List of Abbreviations

([M – H]–) Pseudomolecular ion 2D Two-dimensional 3D Three-dimensional 769-P Human kidney carcinoma cell line A1 Ethanol extract A10 Water extract A2 1,2-Ethanediol extract A3 1,2-Propanediol extract A4 1,3-Propanediol extract A-498 Human kidney carcinoma cell line A5 1,3-Butanediol extract A6 1,2-Pentanediol extract A7 1,5-Pentanediol extract A8 1,2-Hexanediol extract A9 Glycerol extract AAPH 2,2'-Azobis(2-amidinopropane)dihydrochloride ABTS 2,2'-Aazino-bis(3-ethylbenzothiazoline-6-sulphonic acid) ARD Average relative deviations AUC Area under the curve B1 Ethanol B10 Water B2 1,2-Ethanediol B3 1,2-Propanediol B4 1,3-Propanediol B5 1,3-Butanediol B6 1,2-Pentanediol B7 1,5-Pentanediol B8 1,2-Hexanediol B9 Glycerol BA Butyric acid Bt Betaine CA Carboxylic acids Caco-2 Human colorectal carcinoma cell line CC Choline chloride CCCD Circumscribed central composite design CCDC Cambridge crystallographic data centre

i CFU Colony forming unit CRL-2522 Human skin fibroblast cell line DAD Diode -array detector DES Deep eutectic solvents DMEM Dulbecco's modified eagle medium DMF Dimethylformamide DMSO Dimethyl sulfoxide DPPH 2,2-Diphenyl-1-picrylhydrazyl DSC Differential scanning calorimetry dw Dry weight

EC50 The half maximal effective concentration ESI/MSn Electrospray ionization mass spectrometer F Flourescein FAO Food and agriculture organization FDA Food and drug and administration GC Gas chromatography

GI50 The half maximal inhibition of the cell proliferation HBA Hydrogen bond acceptor

HbA1C Glycated haemoglobin HBD Hydrogen bond donor HCT15 Human colorectal carcinoma cell line HEA Heat assisted extraction HeLa Human cervical carcinoma cell line HeLa Cervical carcinoma human cell line HepG2 Human hepatocellular carcinoma cell line HL-60 Adult myeloid leukaemia cell line HPLC High-performance liquid chromatography

IC50 The half maximal inhibitory concentration INT p-iodonitrotetrazolium chloride LC Liquid chromatography LFER Linear free energy relationship LPS Lipopolysaccharide MAE Microwave assisted extraction MBC Minimum bactericidal concentration MCF-7 Human breast adenocarcinoma cell line MDA-TBA Malondialdehyde-thiobarbituric acid ME Maceration extraction MEM Minimum essential medium

ii MHB Mueller-Hinton broth MIC Minimum inhibitory concentration MRSA Methicillin-resistant Staphylococcus aureus MS2 Fragmentation pattern n.p. Not presented n.q. Not quantified n.t. Not tested NaCl Sodium chloride NADES Natural deep eutectic solvents NCI-H460 Non-small lung cancer human cell line NE Northeast NED N-(1-naphthyl)ethylenediamine dihydrochloride NOC-5 Nitric oxide amine-complex donor NRTL Nonrandom two-liquid ns Non-significant

- O2 Superoxide ORAC Oxygen radical absorbance assay OxHLIA Oxidative haemolysis inhibition assay PBS Phosphate buffered saline PC-3 Human prostate cancer cell line PDA Photodiode array detector PLP2 Primary liver porcine cells PPA Phenylpropionic acid PPD 1,3-Propanediol RSA Reference solvent approach RSM Response surface methodology SAC Segment activity coefficient SD Standard deviations SLE Solid-liquid equilibria SPE Solid-phase extraction SRB Sulforhodamine B TBARS Thiobarbituric acid reactive substances TSB Tryptic soy broth UAE Ultrasound assisted extraction UFLC Ultra fast liquid chromatography USA United States of America UV Ultraviolet VOC Volatile organic compounds

iii W Water

iv List of symbols

A Overall solute hydrogen bond acidity A Antioxidant concentration a Shape parameter related to the slope that can produce potential profiles a, b, e, l, s, v Abraham coefficients B Overall solute hydrogen bond basicity

b0 Constant coefficient

bi Coefficient of linear effect

bii Coefficients of quadratic effect

bij Coefficient of interaction effect

CH0 Absorbance of the complete haemolysis at 0 minutes E Solute excess molar refractivity

Ht50 Haemolytic time j Reference solvent L Logarithm of the gas-to-hexadecane partition coefficient at 298 K m Slope of a linear correlation n Number of variables N Number of the experimental data points NP Number of data points p p-value P Intact erythrocyte population 푃̄ Protected substrate

Pm Averaged maximum protected substrate

푃푆 Ratio between the molar solubilities in a given solvent R Universal gas constant R Responses along an arbitrary time series Rt Retention time S Solute dipolarity/polarizability S Solvent composition, ethanol proportion S Solvent composition, water amount in DES S Solute dipolarity/polarizability S/L Solid-liquid ratio

S0 Initial substrate in the reaction

S0 Absorbance of the sample at time 0 푒푥푝 푆푆 Experimental solubility in molar solubilities

푆푠 Solubility in an organic solvent Ss Solubility of the solute

v 푆푤 Solubility in water

St Absorbance of the sample at time t t Time T Temperature

TE Temperature depression

TE, ideal Temperature depression of an ideal mixture

Tm Melting temperature

TOperating Desirable temperature for a purpose V Solute’s McGowan characteristic molecular volume

Vm Average amount of protected molecules x Mole fraction composition of a given compound X Independent variable 푒푥푝 푥푆 Experimental solubility in mole fraction

푥푆 Solubility of a solute in a pure solvent X, Y+, Y-, Z Conceptual segment parameters: hydrophobic, hydrophilic, polar attractive, polar repulsive, respectively Y Dependent variable Y Equation describing the effect of independent variables for each RSM response α Probability of rejecting the null hypothesis when the null hypothesis is true α, β, γ Crystal X-ray diffraction angles γ activity coefficient

훾푆 Activity coefficient of a solute in a solvent

ΔmCp Difference between the molar heat capacity of compound i in the liquid and solid phases

ΔmH Melting enthalpy Δt Delay of time Δt Time interval ∆T Temperature range λ Wavelength

λmax Maximum absorbance wavelength

vi List of Tables

Table 1.1. Overview about the extraction studies of bioactive compounds from J. regia husks and leaves...... 14

Table 1.2. Overview about the use of glycerol and alkanediols for extraction purposes...... 18

Table 2.1. Bibliographic summary of the content of 3-O-caffeoylquinic acid (P1), quercetin 3-O-glucoside (P2) and quercetin O-pentoside (P3) from different source materials using different extraction techniques and conditions...... 37

Table 2.2. Results of the response surface experimental plan for the ME and MAE optimization process of independent variables of time (t), temperature (T), and solvent proportion (S). Response criteria comprise the following: 1) % yield of extraction; 2) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3); and 3) spectrophotometric quantification of total phenolics (PHE) and flavonols (FLAV)...... 44

Table 2.3. Estimated coefficient values obtained from the Box polynomial model, parametric intervals and numerical statistical criteria for each parametric response criteria of the extractions systems tested (ME and MAE). Response criteria comprise the following: 1) % yield of extraction; 2) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3); and 3) spectrophotometric quantification of total phenolics (PHE) and flavonols (FLAV)...... 46

Table 2.4. Operating optimal conditions of the variables involved that maximize the response values for RSM using a CCCD for each of the extracting techniques assessed (ME and MAE), for the three individual response value formats of each compound and for all the compounds. Response criteria comprise the following: 1) % yield of extraction; 2) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3); and 3) spectrophotometric quantification of total phenolics (PHE) and flavonols (FLAV).

vii The optimum values underlined are the results found experimentally, meanwhile the others are those found by the mathematical model...... 52

Table 2.5. Retention time (Rt), wavelengths of maximum absorption in the visible region

(max), mass spectral data, tentative identification, and quantification estimation (mean ± SD) of the tentatively identified phenolic compounds in green and yellow leaves extracts of J. regia...... 69

Table 2.6. Antioxidant, anti-inflammatory, and cytotoxic activities of J. regia leaves extracts and positive controls (mean ± SD)...... 74

Table 2.7. Antimicrobial activity (MIC and MBC mean values in mg/mL) of J. regia leaves extracts and positive controls...... 79

Table 2.8. Retention time (Rt), wavelengths of maximum absorption in the visible region

(max), mass spectral data, tentative identification, and quantification estimation (mean ± SD, n = 6) of the tentatively identified compounds in green husks extracts of J. regia...... 87

Table 2.9. Antioxidant activity, NO formation inhibition and cytotoxicity of J. regia green husks (mean ± SD, n = 9)...... 92

Table 2.10. Antibacterial properties of J. regia green husks against Gram-positive and Gram-negative bacteria (mean, n = 6)...... 97

Table 3.1. Composition, proportions, properties and responses of the set of DES used in the initial screening for the selection of the HBD...... 109

Table 3.2. Results of the response surface experimental plan for the BA and PPA optimization process of independent variables of time (t), temperature (T), and water proportion (S). Response criteria comprise the HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (TOTAL, P1+P2+P3)...... 113

Table 3.3. Estimated coefficient values obtained from the Box polynomial model, parametric intervals and numerical statistical criteria for each parametric response criteria of the extractions systems tested (BA and PPA). Response criteria comprise the HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside

viii content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3)...... 114

Table 3.4. Operating optimal conditions of the variables involved that maximize the response values for RSM using a CCCD for each of the extracting techniques assessed (BA and PPA), for the three individual response value formats of each compound and for all the compounds. Response criteria comprise the following: 1) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O- pentoside content (P3) and total HPLC content (TOTAL, P1+P2+P3)...... 117

Table 3.5. Quantification of the main phenolic compounds present in J. regia leaves (mean ± SD): 3-O-caffeoylquinic acid (P1), trans 3-p-coumaroylquinic acid (P2), quercetin 3-O-glucoside (P3), quercetin O-pentoside (P4) and total HPLC content (PT, P1 +P2 + P3 + P4)...... 136

Table 3.6. Bioactive properties of J. regia leaves. Cytotoxicity in human tumour cell lines

HeLa and non-tumour liver primary cells PLP2 (GI50). Anti-inflammatory activity by the inhibition of the NO production (EC50). Antioxidant activity by the ORAC assay (EC50). Quantification for extracts and solvents (mean ± SD), except for ORAC results as they are presented as mean ± CI...... 139

Table 4.1. Mass purity (%) and source of the organic compounds used in this work. 156

Table 4.2. Experimental solubility (g of solute/100 g of solvent) in water and organic solvents at 298.2 K and 313.2 K.a,b ...... 162

Table 4.3. The melting temperatures and enthalpies of the studied acids from the literature and measured in this work...... 166

Table 4.4: NRTL-SAC optimized parameters, reference solvent, outlier solvent and ARD (%) for each acid using water and six organic solvents in the fitting...... 169

Table 4.5. Estimated solute parameters in the Abraham solvation model, outlier solvent and ARD (%) for each acid using water and seven organic solvents in the correlation database...... 173

ix Table A1. Coded and natural values of the optimization parameters used in the response surface analysis applied in ME and MAE. Three independent variables, extraction time

(X1: t), temperature or power (X2: T); and solvent proportion (X3: S) are combined in five level experimental design of 14 independent variable combinations and 6 replicates on the centre of the experimental domain (20 data points)...... 234

Table B1. Coded and natural values of the optimization parameters used in the response surface analysis applied in BA and PPA. Three independent variables, extraction time (X1: t), temperature or power (X2: T); and water proportion (X3: S) are combined in five level experimental design of 14 independent variable combinations and 6 replicates on the center of the experimental domain (20 data points)...... 239

Table B2. Results of the response surface experimental plan for the eutectic mixtures with betaine and choline chloride. Optimization process of the independent variables: betaine or choline chloride (x), 1,3-propanediol (y) and water (w); expressed as mole fractions. Response criteria (dependent variables) were the results obtained by HPLC- DAD analysis for the contents of 3-O-caffeoylquinic acid (P1), trans 3-p-coumaroylquinic acid (P2), quercetin 3-O-glucoside (P3), quercetin O-glucoside (P4) and total HPLC content (Total, P1 + P2 + P3 + P4)...... 241

Table C1. Overview of the literature solubility data for the syringic acid, vanillic acid and veratric acid for the binary systems addressed in this work...... 245

Table C2. Densities used to convert the units of the solubilities data simulated by the Abraham Solvation Model to molar solubilities...... 246

Table C3. Predicted solute parameters in the Abraham solvation model, outlier solvent and ARD (%) ...... 258

x List of Figures

Figure 1.1. Chemical structures of benzoic and cinnamic acids (part A) and hydroxycinnamic acids (part B)...... 9

Figure 1.2. Generic structure of a flavonoid (part A). Chemical structure of a flavonol and main flavonol aglycones (part B)...... 9

Figure 1.3. Examples of cyclic diarylheptanoids...... 11

Figure 1.4. General structure of naphthalene derivatives and examples of natural naphthoquinone (orange) and tetralone (blue) derivatives (juglone and regiolone, respectively) present in J. regia samples (part A). Example of a natural naphthalene derivative (green) and its structurally similar naphthoquinone and tetralone derivatives (part B)...... 12

Figure 1.5. Representation of the solid-liquid equilibria of a eutectic mixture (blue) and a deep eutectic mixture (green). Adapted from Martins et al. (2018)...... 23

Figure 1.6. Distribution of the publications related to the topic “deep eutectic solvent”. Data retrieved from Scopus (May 28, 2019)...... 24

Figure 1.7. Examples of starting materials used to form DES...... 26

Figure 2.1. Response surface of each of the variables involved for ME and MAE systems for the HPLC quantification of 3-O-caffeoylquinic acid content (P1) and quercetin 3-O- glucoside content (P2) and for the spectrophotometric determination of the total phenolics (PHE). The other response determinations of quercetin O-pentoside content (P3) and total flavonols (FLAV) were excluded and presented in Figure A3 (Appendix A) due to their similarity with quercetin 3-O-glucoside content (P2) and total phenolics (PHE), respectively. Each of the net surfaces represents the theoretical three- dimensional response surface predicted with the second order polynomial of Eqs. [2.3], [2.4] and [2.7] in ME system and [2.10], [2.11] and [2.14] for MAE process. The statistical design and results are described in Table 2.2. Estimated parametric values are shown in Table 2.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 2.4)...... 51

xi Figure 2.2. Response surfaces of the total compounds content determined by HPLC (P1+P2+P3) and extraction yield from the developed equations for the ME and MAE systems optimization. Part A: The graphical 3D analysis is presented as a function of each of the variables involved. Each of the net surfaces represents the theoretical three- dimensional response surface predicted with the second order polynomial of Eqs. [2.6], [2.2], [2.13] and [2.9], respectively. The statistical design and results are described in Table 2.2. Estimated parametric values are shown in Table 2.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 2.4). Part B: To illustrate the goodness of fit, two basic graphical statistic criteria are used: the ability to simulate the changes of the response between the predicted and observed data; and the residual distribution as a function of each of the variables. Note all the differences in the axes scales...... 53

Figure 2.3. Final summary of the effects of all variables assessed for ME and MAE systems. Part A: Shows the individual 2D responses of all studied responses as a function of all the variables assessed. The variables in each of the 2D graphs were positioned at the optimal values of the others (Table 2.4). The dots () presented alongside each line highlight the location of the optimum value. Lines and dots are generated by the theoretical second order polynomial models of Eqs. [2.2]-[2.7] for ME and [2.9]-[2.15] for MAE. Part B: Shows the dose response of S/L ratio at the optimal values of the other three variables (Table 2.4). Points () represent the obtained experimental results, meanwhile the line shows the predicted pattern by simple linear relation. The limit value (~140 g/L) shows the maximum achievable experimental concentration until the sample cannot be physically stirred at laboratory scale...... 54

Figure 2.4. HPLC phenolic profile of J. regia L. leaves. Graphical representation of walnut green sample, recorded at 280 nm (A) and 370 nm (B). The peaks identification and quantification are presented in Table 2.5...... 70

Figure 2.5. Kinetic profile of the action of J. regia green leaves extract during the erythrocyte heamolysis (OxHLIA assay)...... 75

Figure 2.6. HPLC phenolic profile of J. regia L. green husks at 280 nm (A) and 370 nm (B). The peaks identification and quantification are presented in Table 2.8...... 88

xii Figure 2.7. Kinetic behavior of J. regia green husks extract during the erythrocyte heamolysis (OxHLIA assay)...... 92

Figure 3.1. Extraction yield (mg/g dw) of 3-O-caffeoylquinic acid (P1), quercetin 3-O- glucose (P2) and quercetin O-pentoside (P3) obtained in the initial screening of the choline chloride based DES described in Table 3.1, with the extraction conditions: 1 hr extraction time, 50 °C, 600 rpm and 20 % (w/w) of water (solvents with * only contain 5 % (w/w) of water)...... 110

Figure 3.2. Response surfaces of the total compounds content determined by HPLC (P1+P2+P3) for the BA and PPA systems optimization. Part A: Shows the joint graphical 3D analysis as a function of each of the variables involved. Each of the net surfaces represents the theoretical three-dimensional response surface predicted with the second order polynomial of Eqs. [3.4] and [3.8], respectively. The statistical design and results are described in Table 3.2. Estimated parametric values are shown in Table 3.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 3.4). Part B: To illustrate the goodness of fit, two basic graphical statistic criteria are used. The first one, the ability to simulate the changes of the response between the predicted and observed data; and the second one, the residual distribution as a function of each of the variables...... 119

Figure 3.3. Joint graphical 3D analysis as a function of each the variables involved for BA and PPA systems for the HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2) and quercetin O-pentoside content (P3)...... 120

Figure 3.4. Summary of the effects of all variables assessed for BA and PPA systems, showing the individual 2D responses of all studied responses as a function of all the variables assessed. The variables in each of the 2D graphs were positioned at the optimal values of the others (Table 3.4). The dots () presented alongside each line highlight the location of the optimum value. Lines and dots are generated by the theoretical second order polynomial models of Eqs. [3.1]-[3.4] for CC:BA and [3.5]-[3.8] for CC:PPA...... 122

Figure 3.5. Chromatographic records at 280 nm (part A) used for the quantification of phenolic acids (P1: 3-O-caffeoylquinic acid; P2: trans 3-p-coumaroylquinic acid), and, at

xiii 370 nm (part B) for the quantification of quercetin derivatives (P3: quercetin 3-O- glucoside; P4:quercetin O-pentoside) present in J. regia L. leaves samples. Representation of conventional solvents extracts (water and ethanol + water) solvents and graphical comparison with the extract obtained applying 1,3-propanediol aqueous solution during the screening process...... 137

Figure 3.6. Parts A and B show the application of the antioxidant responses of the ORAC assay based in the kinetic dose-response effect of the solvents & extract (mg F dw/mL of extraction) and solvents (%), respectively. The dots () are the experimental data series and the lines the fittings. Part C shows the numerical values of the rate parameter of the response as assessment criteria. The numerical notation 1 to 10 represents the solvents (ethanol, 1,2-ethanediol, 1,2-propanediol, 1,3-propanediol, 1,3-butanediol, 1,2-pentanediol, 1,5-pentanediol, 1,2-hexanediol, glycerol and water), respectively. 143

Figure 4.1. Chemical structures of: (a) syringic acid; (b) vanillic acid and (c) veratric acid...... 154

Figure 4.2. Comparison of the experimental solubility data obtained in this work with literature (Bowen et al., 2013; Li et al., 2013b; Noubigh et al., 2008b, 2008a; Noubigh and Abderrabba, 2016; Noubigh and Akermi, 2017; Queimada et al., 2009; Zhang et al., 2016b) data: (a) syringic acid + water; (b) syringic acid + ethanol; (c) vanillic acid + water; (d) vanillic acid + ethanol; (e) veratric acid + water; (f) veratric acid + ethanol...... 164

Figure 4.3. Comparison between experimental and calculated solubility for correlation: a) NRTL-SAC; b) NRTL-SAC + RSA...... 170

Figure 4.4. Comparison between experimental and calculated solubility data for prediction: a) NRTL-SAC; b) NRTL-SAC + RSA...... 171

Figure 4.5. Comparison between experimental and calculated solubility in the organic solvents used in the estimation of the Abraham solute parameters...... 173

Figure 4.6. Comparison between experimental and calculated solubility data in the predicted set using the Abraham solvation model...... 174

xiv Figure A1. Diagram of the different steps carried out in order to obtain an optimal phenolic extract from Juglans regia L. using the following quantification responses: total phenolics and flavonols, quercetin derivatives, 3-O-caffeoylquinic acid and yield. .... 235

Figure A2. The phenolic profile of J. regia leaves. The main compounds present in the matrix were one phenolic acid, 3-O-caffeoylquinic acid (P1) and two flavonols, quercetin 3-O-glucoside (P2) and quercetin O-pentoside (P3). This profile is according to the previous reported by Santos and co-workers (Santos et al., 2013)...... 236

Figure A3. Graphical representation of the 3D analysis as a function of each the variables involved for ME and MAE systems for the response determinations of quercetin O- pentoside content (P3) and total flavonols (FLAV). Each of the net surfaces represents the theoretical three-dimensional response surface predicted with the second order polynomial of Eqs. [2.4], [2.8], [2.11] and [2.15], respectively. The statistical design and results are described in Table 2.2. Estimated parametric values are shown in Table 2.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 2.4)...... 237

Figure B1. Diagram of the different steps carried out to obtain an optimal phenolic extract from Juglans regia L. using the following quantification responses: total phenolics and flavonols, quercetin derivatives and 3-O-caffeoylquinic...... 240

Figure B2. Schematic representation of the existing liquid phase equilibria (at 25 °C) of the three components used to design mixtures (solvents) to extract phenolic compounds from Juglans regia leaves...... 242

Figure B3. Chromatographic records at 280 nm (parts A) used for the quantification of phenolic acids, and, at 370 nm (parts B) for the quantification of quercetin derivatives present in J. regia L. leaves samples. Comparison between the optimized solvent extracts (1,3-propanediol + water (PPD:W), betaine + 1,3-propanediol + water (Bt:PPD:W) and choline chloride+ 1,3-propanediol + water (CC:PPD:W). Main phenolics identified and quantified: 3-O-caffeoylquinic acid (P1), trans 3-p-coumaroylquinic acid (P2), quercetin 3-O-glucoside (P3) and quercetin O-pentoside (P4)...... 243

xv Figure C1. Comparison of the experimental solubility data of syringic acid in aqueous systems available in literature (Noubigh et al., 2008b; Queimada et al., 2009) and obtained in this work...... 247

Figure C2. Comparison of the experimental solubility data of syringic acid available in literature (Noubigh and Akermi, 2017) and obtained in this work for the following organic solvents: a) ethyl acetate; b) methanol; c) ethanol; d) 2-propanol...... 248

Figure C3. Comparison of the experimental solubility data of vanillic acid available in literature (Noubigh et al., 2008a; Noubigh and Abderrabba, 2016; Zhang et al., 2016b) and obtained in this work for the following solvents: a) water; b) methanol; c) ethanol; d) 2-propanol; e) 1-propanol; f) ethyl acetate ...... 249

Figure C4. Comparison of the experimental solubility data of veratric acid available in literature (Bowen et al., 2013; Li et al., 2013b) and obtained in this work for the following solvents: a) ethanol; b) 2-propanol; c) 1-propanol; d) ethyl acetate; e) 2-butanone (f) methanol (g) water...... 251

Figure C5. Illustrative DSC fusion thermograms of: a) syringic acid; b) vanillic acid; c) veratric acid...... 252

Figure C6. Exemplificative thermograms of successive runs for: a) syringic acid; b) vanillic acid; c) veratric acid...... 253

Figure C7. Screenshot of the visual method measurement for vanillic acid when a change in the solid coloration (possible degradation) is observed...... 254

Figure C8. Comparison of the experimental X–ray powder diffraction patterns of recrystallized syringic acid from solutions with DMF, acetonitrile and 2-butanone with the powder diffraction pattern of syringic acid from the manufacturer and the simulated powder diffraction pattern reported by. Thipparaboina et al. (2016)...... 255

Figure C9. Comparison of the experimental X–ray powder diffraction patterns of recrystallized vanillic acid (from solutions with DMF, acetonitrile, 2-butanone, methanol and water) with the powder diffraction pattern of vanillic acid from the manufacturer and the simulated diffraction powder pattern reported by Kozlevčar et al. (2006). .. 256

xvi Figure C10. Comparison of the experimental X–ray powder diffraction patterns of recrystallized veratric acid (from solutions with DMF, acetonitrile, 2-butanone and methanol) with the powder diffraction pattern of veratric acid from the manufacturer and the simulated powder diffraction pattern reported by Pinkus et al. (2002)...... 257

xvii xviii Contents

1. General introduction ...... 1 1.1. General context ...... 3 1.1.1. The plant ...... 4

1.1.1.1. Husks ...... 5 1.1.1.2. Leaves ...... 6 1.1.1.3. Bioactive molecules in J. regia husks and leaves ...... 7 Phenolic compounds ...... 8 Naphthalene derivatives ...... 11 1.1.2. Solvents used in the extraction process ...... 13

1.1.2.1. Conventional solvents ...... 13 1.1.2.2. Glycerol and alkanediols ...... 17 1.1.2.3. Deep eutectic solvents ...... 22 1.2. Scope and objectives ...... 29 2. Characterization of the plant material ...... 31 2.1. Optimization and comparison of maceration and microwave extraction systems for the production of phenolic compounds from Juglans regia L. for the valorization of walnut leaves ...... 33 2.1.1. Introduction ...... 34

2.1.2. Materials and methods ...... 39

2.1.2.1. Standards and reagents ...... 39 2.1.2.2. Plant material ...... 39 2.1.2.3. Extraction systems for walnut leaves ...... 39 Maceration extraction ...... 39 Microwave-assisted extraction ...... 40 2.1.2.4. Quantification of the main compounds ...... 40 Chromatographic analysis of the main phenolic compounds ...... 40 Determination of total phenolics and flavonols by UV-Vis spectroscopy ...... 41 2.1.2.5. Experimental design, model analysis and statistical evaluation ...... 41 Experimental design ...... 41

xix Mathematical model ...... 42 Procedure to optimize the variables to a maximum response ...... 42 2.1.2.6. Fitting procedures and statistical analysis...... 42 2.1.3. Results and discussion ...... 43

2.1.3.1. Response criteria for the RSM analysis and statistical verification .... 43 2.1.3.2. Theoretical response surface models ...... 47 2.1.3.3. Effect of extraction variables on the target responses ...... 49 2.1.4. Conclusions ...... 58

2.2. Phenolic profile, antioxidant and antibacterial properties of Juglans regia L. (walnut) leaves from the Northeast of Portugal ...... 59 2.2.1. Introduction ...... 60

2.2.2. Material and methods ...... 62

2.2.2.1. Plant material and extract preparation ...... 62 2.2.2.2. Phenolic compounds characterization ...... 62 2.2.2.3. In vitro antioxidant assays ...... 63 2.2.2.4. Anti-inflammatory activity ...... 64 2.2.2.5. Cytotoxic assays ...... 65 2.2.2.6. Antibacterial activity ...... 66 2.2.2.7. Statistical analysis ...... 67 2.2.3. Results and discussion ...... 67

2.2.3.1. Composition in phenolic compounds ...... 67 2.2.3.2. Bioactivity of the hydroethanolic extracts ...... 73 2.2.4. Conclusions ...... 79

2.3. Phytochemical profile of Juglans regia L.: valorization of walnut residues through their bioactive potential ...... 80 2.3.1. Introduction ...... 81

2.3.2. Material and methods ...... 82

2.3.2.1. Plant material and extract preparation ...... 82 2.3.2.2. Phytochemical characterization ...... 83 2.3.2.3. In vitro antioxidant assays ...... 83

xx 2.3.2.4. Anti-inflammatory activity ...... 84 2.3.2.5. Cytotoxicity assays ...... 84 2.3.2.6. Antibacterial activity ...... 85 2.3.3. Results and discussion ...... 85

2.3.3.1. Phytochemical composition of walnut husks ...... 85 2.3.3.2. Bioactivity of the hydroethanolic extracts ...... 91 2.3.4. Conclusions ...... 97

3. Extractions with alternative solvents ...... 99 3.1. Enhanced extraction of phenolic compounds using choline chloride based deep eutectic solvents from Juglans regia L...... 101 3.1.1. Introduction ...... 102

3.1.2. Material and methods ...... 103

3.1.2.1. Standards and reagents ...... 103 3.1.2.2. Plant material ...... 104 3.1.2.3. Screening analysis of DES ...... 104 Preparation of DES...... 104 Conditions for the screening extraction by heat-assisted extraction ...... 104 3.1.2.4. Chromatographic analysis of the main phenolic compounds ...... 105 3.1.2.5. HAE optimization by RSM of the selected DES, experimental design, model analysis and statistical evaluation ...... 105 Experimental design ...... 105 Mathematical model ...... 106 Procedure to optimize the variables to a maximum response ...... 106 Fitting procedures and statistical analysis...... 107 3.1.3. Results and discussion ...... 107

3.1.3.1. DES screening to investigate the effect of CA molecular structure on the extraction of phytochemical compounds ...... 107 3.1.3.2. Response criteria for the RSM analysis and statistical verification .. 112 Theoretical response surface models predicted ...... 114 Overall effects of the extraction variables on the target responses ...... 116

xxi Numerical individual and global optimal conditions that maximize the extraction, statistical analysis and experimental verification of predictive models ...... 121 3.1.3.3. Dose-response analysis of the solid-to-liquid effect at the optimum conditions ...... 123 3.1.3.4. Comparison with other responses reported using non-DES ...... 123 3.1.4. Conclusions ...... 124

3.2. Insights on the extraction performance of alkanediols and glycerol: using Juglans regia L. leaves as a source of bioactive compounds ...... 126 3.2.1. Introduction ...... 127

3.2.2. Material and methods ...... 129

3.2.2.1. Plant material and extract preparation ...... 129 3.2.2.2. Plant material ...... 129 3.2.2.3. Chromatographic analysis of the main phenolic compounds l ...... 129 3.2.2.4. Extraction methodology ...... 130 Preparation of solvents ...... 130 Extraction technique ...... 130 3.2.2.5. Cytotoxicity assays ...... 130 3.2.2.6. Anti-inflammatory activity ...... 131 3.2.2.7. Oxygen radical absorbance assay (ORAC) ...... 131 3.2.2.8. Optimization of the solvent composition (eutectic mixtures) ...... 133 Experimental design ...... 133 Mathematic model for the triangular mixture design ...... 134 3.2.2.9. Statistical analysis ...... 135 3.2.3. Results and discussion ...... 135

3.2.3.1. Preliminary screening using alcohols ...... 135 Extraction yields ...... 135 Cytotoxicity of extracts and solvents ...... 137 Antioxidant and anti-inflammatory activity of extracts and solvents ...... 141 3.2.3.2. Eutectic mixtures based on 1,3-propanediol ...... 144 Optimization of the extraction yield of phenolic compounds ...... 144

xxii Cytotoxicity of the optimum mixtures (extracts and solvents) ...... 146 Anti-inflammatory and antioxidant activity of extracts and solvents ...... 148 3.2.4. Conclusions ...... 149

4. Solubility of target molecules ...... 151 4.1. Solvent and temperature effects on the solubility of syringic, vanillic or veratric acids: experimental, modeling and solid phase studies ...... 153 4.1.1. Introduction ...... 154

4.1.2. Experimental ...... 156

4.1.2.1. Chemicals ...... 156 4.1.2.2. Solubility experiments ...... 157 4.1.2.3. Melting properties ...... 158 4.1.2.4. Solid-phase studies ...... 158 Samples ...... 158 Powder and single X-ray diffraction ...... 158 4.1.3. Thermodynamic modeling ...... 159

4.1.3.1. The NRTL-SAC model ...... 159 4.1.3.2. The Abraham solvation model ...... 160 4.1.4. Results and discussion ...... 162

4.1.4.1. Experimental solubilities ...... 162 4.1.4.2. Data analysis ...... 163 4.1.4.3. Melting properties ...... 165 4.1.4.4. Solid phase studies ...... 167 4.1.4.5. Thermodynamic Modelling ...... 168 NRTL-SAC and NRTL-SAC + RSA ...... 168 4.1.4.6. Abraham solvation model ...... 171 4.1.5. Conclusions ...... 176

5. Conclusions and future work ...... 179 6. List of publications ...... 185 7. Bibliography ...... 189 8. Appendix ...... 231

xxiii Appendix A: Supporting information of Chapter 2 ...... 233 Appendix B: Supporting information of Chapter 3 ...... 239 Appendix C: Supporting information of Chapter 4 ...... 245

xxiv 1. General introduction

General introduction CHAPTER 1

1.1. General context The International Energy Agency defines biorefinery as “the sustainable processing of biomass into a spectrum of marketable products (food, feed, materials and chemicals) and energy (fuels, power and heat)” (Sonnenberg et al., 2007). In this context, the full exploitation of biomass residues generated from the food and agroforest processing industries may be accomplished by introducing an additional step for the recovery of added-value bioactive molecules, as they could be a rich source of natural antioxidants such as phenolic compounds (Balasundram et al., 2006) with potential application as additives in different areas (Carocho et al., 2015; Ribeiro et al., 2015).

From the characterization of the chemical and biological properties of more than 300 plant species from Northeast (NE) Portugal (www.esa.ipb.pt/biochemcore/), the leaves of walnut trees (Juglans regia L.) and the green husks of walnuts stand out as a rich source of bioactive compounds. In particular, the leaves present a well-studied phenolic profile, composed mostly of phenolic acids and flavonoids.

Conventionally, volatile organic compounds (VOC) are common extraction solvents, though some disadvantages related to their flammability and toxicity can be pointed out. More recently, natural deep eutectic solvents have been proposed in the literature as alternative solvents to extract phytochemicals, presenting some advantages. In principle, depending on their components, less toxic and biocompatible solvents could be designed for a given application, while obtaining higher extraction yields of the target compounds. To support the selection of the solvents, the knowledge of the solid-liquid equilibria of these systems is crucial for the design of their separation and purification processes. However, even basic information regarding the melting properties of pure phenolic compounds or their solubility in conventional organic solvents is scarce which difficults the development of thermodynamic models which are essential tools for solvent screening.

In the following section, a brief introduction about the biomass studied and their main families of phytochemicals will be given. The main conventional solvents used in the extraction of bioactive compounds will be also described, followed by a short review on the application of glycerol, alkanediols and deep eutectic solvents in this area.

3 General introduction CHAPTER 1

1.1.1. The plant Juglans regia L. is a well-known plant especially by the consumption of its fruit, being widespread over the world, and the most cultivated nut tree (Martínez et al., 2010). Walnuts are interesting fruits in our diet since they are a source of many essential nutrients and bioactive compounds supporting its classification as functional food (Tapia et al., 2013), being listed as a priority plant for the human nutrition by the Food and Agriculture Organization (FAO) (Kasmi et al., 2013). Besides the walnut nutritional value, other vegetative parts from the walnut tree have been explored as natural sources of bioactive compounds with preventing and therapeutic benefits, including the roots, flowers, leaves, bark, green husks and shells (Delaviz et al., 2017; Panth et al., 2016). In the context of their use in traditional medicine, many applications have been collected (Carvalho and Morales, 2010; Delaviz et al., 2017; Panth et al., 2016; Ribeiro et al., 2015; Taha and Al-wadaan, 2011) and several scientific studies have reported the antioxidant, anti-inflammatory, anti- diabetic, anti-proliferative, antimicrobial and hepatoprotective potentials of different vegetative parts of J. regia (Jahanban-Esfahlan et al., 2019; Panth et al., 2016). The phytochemical characterization of the J. regia plant was also reported, varying the main compounds according to the plant cultivar, the plant material or the solvent used (among other factors). Phenylpropanoids, flavonoids, diarylheptanoids, naphthoquinones, tetralone derivatives, steroids, tannins, terpenes and terpenoids were the most common families of compounds identified in the J. regia extracts (Delaviz et al., 2017; Panth et al., 2016; Taha and Al-wadaan, 2011).

Different by-products are obtained from the industrial activity related to the walnuts (fruits) production, especially the walnut husks. A few studies have proposed the valorisation of this by-product, either by exploring the bioactive potential of the extracts (Akbari et al., 2012; Bagheri et al., 2012; Beiki et al., 2018; Fernández-Agulló et al., 2013; Oliveira et al., 2008; Sharma et al., 2013; Zhang et al., 2014) and isolated compounds (Alshatwi et al., 2012; Li et al., 2013a; Yang et al., 2019), or by directly using the raw material as an ingredient to enhance the quality of a final product (Salejda et al., 2016). The leaves of walnut tree are also recognised to be a rich source of phytochemicals with biological activity (Carvalho and Morales, 2010; Jahanban-Esfahlan et al., 2019; Panth et al., 2016;

4 General introduction CHAPTER 1

Ribeiro et al., 2015; Santos et al., 2013). Thus, both walnut husks and leaves constitute interesting matrices to be explored, aiming to promote their potential in the pharmaceutical, cosmetic and food areas, as detailed below.

1.1.1.1. Husks Juglans regia L. green husks constitute the main agroforest waste related to the walnut production (Ciniglia et al., 2012). Some studies can be found about the biological activity of the extracts of this by-product. In particular, the antioxidant activity was often used as an indicator of its biological activity (Bagheri et al., 2012; Carvalho et al., 2010; Chatrabnous et al., 2018; Fernández-Agulló et al., 2013; Soto-Maldonado et al., 2019; Zhang, 2015). The antimicrobial properties of the extracts of walnut husks were also studied, showing activity against different microorganisms such as bacteria (Gram-positive and Gram-negative) and fungi (Fernández-Agulló et al., 2013; Keskin et al., 2012; Oliveira et al., 2008; Sharma et al., 2013; Zhang et al., 2014). However, data concerning the cytotoxic potential of this raw material are scarcer. Even so, the walnut green husks have shown satisfactory results against different human tumour cell lines, including renal (A-498 and 769-P), prostate (PC- 3) and hepatic (HepG2) carcinomas (Alshatwi et al., 2012; Carvalho et al., 2010; Soto- Maldonado et al., 2019). Furthermore, recently, five isolated compounds (juglanins F, G, H, I and J) from walnut green husks of J. regia showed cytotoxic activity against HepG2 cells (Yang et al., 2019). However, it is not the first time that isolated compounds from green husks belonging to Juglans spp. presented antitumor activity, reinforcing the importance of this by-product as a source of natural active molecules with pharmaceutical interest (Li et al., 2008b; Zhou et al., 2015). In fact, isolated diarylheptanoids, tetralone derivatives and naphthalene derivatives from this genus showed cytotoxic potential against different human tumour cell lines, namely HepG2 (Li et al., 2008b; Yang et al., 2019; Zhou et al., 2015), HeLa (cervical carcinoma) (Li et al., 2013a, 2008b), HL-60 (Li et al., 2008b; Yang et al., 2019), among others less frequently studied (A549, MCF7, BEL-7402, COLO205, BGC- 823, SK-OV-3) (Li et al., 2013a). Also, using cell-based assays, some authors have reported the bioactivity of the extract namely the anti-platelet aggregation activity (Meshkini and Tahmasbi, 2017; Rywaniak et al., 2015), while Wang et al. (2019) reported the lipid- lowering potential through an animal approach (high fat diet-induced obesity rats).

5 General introduction CHAPTER 1

In addition, phenolic compounds, which are widely recognised as bioactive molecules, also have been isolated and identified. Overall, juglone, phenolic acids (mainly hydroxycinnamic acids) and flavonoids (mainly flavonols) are the most common compounds found in the green husks from J. regia L. (Akbari et al., 2012; Chatrabnous et al., 2018; Cosmulescu et al., 2010; Gawlik-Dziki et al., 2014; Stampar et al., 2006). Nevertheless, further contributions for the complete identification and quantification of the phytochemicals present in the green husks of J. regia are still necessary.

1.1.1.2. Leaves The antioxidant activity has been the most studied bioactivity dimension of the extracts of walnut leaves (Carvalho et al., 2010; Gawlik-Dziki et al., 2014; Miliauskas et al., 2004; Pereira et al., 2007; Santos et al., 2013; Wojdylo et al., 2007; Zhao et al., 2014). Also, the anti-diabetic potential of the extracts has been also explored, to support the ethnobotanical use of walnut leaves for that purpose (Abdoli et al., 2017; Asgary et al., 2008; Hosseini et al., 2014; Javidanpour et al., 2012; Mohammadi et al., 2012; Mollica et al., 2017; Pitschmann et al., 2014; Rabiei et al., 2018; Teimori et al., 2009). Furthermore, the studies carried out to evaluate the antimicrobial activity of these extracts have shown potential against Gram positive (including Mycobacterium tuberculosis) and Gram negative bacteria and fungi (Cruz-Vega et al., 2008; Gîrzu et al., 1998; Pereira et al., 2007; Qa’dan et al., 2005; Rather et al., 2012; Sharafati-Chaleshtori et al., 2011). Moreover, relevant results were obtained regarding the cytotoxic potential of J. regia leaves against human tumour cell cultures such as renal (A-498 and 769-P), colorectal (Caco-2 and HCT15), hepatic (HepG2), cervical (HeLa), breast (MCF-7), and non-small lung cancer (NCI-H460) carcinoma (Carvalho et al., 2010; Santos et al., 2013). In addition, the anti-dyslipidemic (Hosseini et al., 2014; Mohammadi et al., 2012; Mollica et al., 2017), anti-inflammatory (Erdemoglu et al., 2003; Hosseinzadeh et al., 2011), analgesic (Erdemoglu et al., 2003; Gîrzu et al., 1998; Hosseinzadeh et al., 2011), neuroprotective (Orhan et al., 2011) and vasorelaxant (Perusquía et al., 1995) potentials of the walnut leaves extracts have also been reported. Other studies also highlighted the potential of walnut leaves extracts to preserve and protect the heathy stage of different organs including kidney, liver, pancreas and stomach

6 General introduction CHAPTER 1 using animal models (rats) (Eidi et al., 2013; Erdemoglu et al., 2003; Mohammadi et al., 2012; Mollica et al., 2017).

The findings about the bioactivity of the plant material encouraged its chemical characterization. Reviewing the contribution of several authors, the phytochemical profile of J. regia leaves contains diverse secondary metabolites such as flavonoids, phenolic acids, hydrolysable tannins, diarylheptanoids, phenylpropanoids, terpenes, apocarotenoids, steroids, anthraquinones, tetralone derivatives, and juglone derivatives (Cosmulescu et al., 2014a, 2014b, 2011; Pitschmann et al., 2014; Schwindl et al., 2019; Schwindl and Kraus, 2017). Meanwhile, the most abundant compounds found in leaves were flavonoids (mainly flavonols) and phenolic acids (mainly hydroxycinnamic acids) (Gawlik-Dziki et al., 2014; Mollica et al., 2017; Pereira et al., 2007; Pitschmann et al., 2014; Santos et al., 2013; Wojdylo et al., 2007; Zhao et al., 2014). Less frequently isolated compounds were organic acids, terpenes, steroids, anthraquinones, tocopherols, megastigmanes and proanthocyanidins (Forino et al., 2016; Rather et al., 2012; Santos et al., 2013; Schwindl et al., 2019; Schwindl and Kraus, 2017; Verma et al., 2013).

In addition, the bioactivity of some of the isolated compounds from walnut leaves was studied such as the anti-bacterial (terpenes), cytotoxic (flavonols, tetralone derivatives, steroids) and sedative (juglone) activities (Gîrzu et al., 1998; Rather et al., 2012; Salimi et al., 2014). Therefore, the current research on the walnut leaves bioactivity and their abundancy in phytochemicals supports the development of applications in the food, cosmetic and pharmaceutical areas.

1.1.1.3. Bioactive molecules in J. regia husks and leaves During the past decades, the study of the bioactive potential of natural matrices has been accompanied by extensive research aiming to identify and quantify the main phytochemical constituents that confer those biological properties. Consequently, the study of natural products contributed for the development of separation, purification and identification techniques, being used as a source of more refined products such as new drugs, insecticides or herbicides (Rostagno and Prado, 2013). Thus, not only plant extracts but also their individual components have gained much attention for both academia and industry. In this

7 General introduction CHAPTER 1 section, the most abundant families of compounds present in J. regia leaves and husks (phenolic acids, flavonols, diarylheptanoids and naphthalene derivatives) will be briefly presented.

Phenolic compounds Among the huge number of phenolic compounds in nature, phenolic acids and flavonoids were the major sub-groups of phenolics found in Juglans regia L., especially in their leaves and green husks. Diarylheptanoids have been also identified but less frequently.

Phenolic acids can be divided in two groups: hydroxycinnamic acids and hydroxybenzoic acids, which are derived from the non-phenolic compounds cinnamic acid and benzoic acid, respectively (Figure 1.1, part A) (Heleno et al., 2015). In particular, hydroxycinnamic acids are abundant in J. regia leaves and husks, being the chemical structure of the most common presented in Figure 1.1 (part B). They result from the hydroxylation of the aromatic ring of cinnamic acid, which can assume different structures, comprising a three-carbon side chain (C6–C3) skeleton: coumaric, caffeic, ferulic and sinapic acids and their derivatives (usually by esterification with quinic acid). Moreover, hydroxycinnamic acids have been reported to present diverse biological functions including antioxidant, anti-inflammatory, antimicrobial, anti-collagenase, anti-melanogenic, anti-proliferative activities, suggesting that this group of natural occurring compounds have potential in the pharmaceutical and cosmetic sectors (e.g. skin care formulations) (Heleno et al., 2015; Taofiq et al., 2017). The biological activity of phenolic acids can be related to their chemical structure. In the special case of the antioxidant properties, their aromatic scaffold with hydroxyl groups attached to the ring structure confer these molecules a character of reducing agents, hydrogen donators, singlet oxygen quenchers, superoxide radical scavengers and metal chelators over hydroxyl and peroxyl radicals, superoxide anions and peroxynitrites, being the antioxidant activity as higher as the increasing number of hydroxyl substituents in the aromatic ring (Rice-Evans et al., 1996; Terpinc et al., 2011).

8 General introduction CHAPTER 1

A B 6 4 5 R1 R2 R3 Identification Benzoic acid 1 3 2 H H H p-Coumaric acid

OH H H Caffeic acid Quinic acid

OCH3 H H Ferulic acid

R2 OCH3 OCH3 H Sinapic acid

Cinnamic acid R3 H H Quinic acid Coumaroylquinic acid

OH H Quinic acid Caffeoylquinic acid

OCH3 H Quinic acid Feruloylquinic acid R1

Figure 1.1. Chemical structures of benzoic and cinnamic acids (part A) and hydroxycinnamic acids (part B).

Flavonols are natural occurring molecules widely spread in plants. They belong to a major group of molecules: flavonoids (Figure 1.2, part A). Flavonoids are polyphenol compounds, structurally composed by two aromatic rings (A and B) and another heterocyclic one (C), generally viewed as C6-C3-C6 skeletons, as depicted in part A of Figure 1.2. Flavonols are hydroxylated at positions 3 (C ring), 5 and 7 (A ring), and have a 2,3-double boud (C ring) and a 4-carbonyl group (C ring) in the general flavonoid skeleton. The main flavonol aglycones are kaempferol, quercetin, myricetin and isorhamnetin, depending on the functional groups found in the B ring (Figure 1.2, part B). This sub-group of flavonoids (along with hydroxycinnamic acids) were the main phenolic compounds present in J. regia leaves and husks samples.

A B R3’

R3’ R4’ R5’ Identification 3’ R4’ 2’ 4’ H OH H Kaempferol 1 B 8 1’ OH OH H Quercetin 2 5’ 7 R5’ OH OH OH Myricetin A C 6’

6 3 OCH3 OH H Isorhamnetin 5 4

Figure 1.2. Generic structure of a flavonoid (part A). Chemical structure of a flavonol and main flavonol aglycones (part B).

Rice-Evans et al. (1996) showed that the antioxidant potential of individual flavonols is significantly affected by structural changes such as the hydroxylation of the B ring as well

9 General introduction CHAPTER 1 as by the glycosylation of the C ring, being flavonol aglycones the most active ones (e.g. kaempferol < quercetin-3-O-rutinoside < myricetin < quercetin) (Rice-Evans et al., 1996). Moreover, some studies using cell or animal models also have highlighted the potential of flavonols namely as antioxidant (Chen and Chen, 2013; Kashyap et al., 2017; Li and Ding, 2012; Suganthy et al., 2016), anti-inflammatory (Chen and Chen, 2013; Devi et al., 2015a; Kashyap et al., 2017; Li and Ding, 2012), antidiabetic (Li and Ding, 2012; Shi et al., 2019) and antiproliferative (Chen and Chen, 2013; Devi et al., 2015b; Kashyap et al., 2017; Shi et al., 2019; Suganthy et al., 2016) agents. In addition, the human trials using flavonols have shown important conclusions regarding their advantages in tumour related treatments, as far as clinical studies showed that they not only enhance the action of chemotherapeutic drugs but also due to their ability to reduce the ineluctable side effects (Conklin, 2000; Kashyap et al., 2017; Suganthy et al., 2016).

Another group of natural occurring compounds in Juglans ssp. is a series of molecules with a 1,7-diphenylheptane skeleton (C6-C7-C6), so-called diarylheptanoids or diphenylheptanoids and, as other phenolics, they are also spread across different species of plants (Alberti et al., 2019). This group of molecules can be sub-divided in two major groups depending on their cyclic (diarylether- or biaryl-types: connection of the aromatic rings form [7.1]-meta,para- or [7.0]-meta,meta-cyclophanes, respectively) or non-cyclic (linear) molecular configuration. So far, more than 400 diarylheptanoids were identified in nature, being the macrocyclic ones the most typical in Juglans spp. (Alberti et al., 2019). According to the review made by Bi et al. (2016), linear diarylheptanoids occur mainly in the roots of walnut trees, while the cyclic ones are the most spread (steam barks, husks, leaves and roots), being juglanin B the most common on Juglans spp.. In this regard, some diarylheptanoids, isolated from Juglans ssp., showed potential against the proliferation of tumour cell cultures such as HL-60 (Li et al., 2008b; Yang et al., 2019), HepG2 and HeLa (Li et al., 2008b). In Figure 1.3, the main natural cyclic diarylheptanoids are presented.

10 General introduction CHAPTER 1

7 11 R1 R2 R3 R4 Identification 4 5 8 10 9 OCH3 OH H OCH3 Juglanin A 3 6 R3 12 18 19 OCH3 OH H OH Myricatomentogenin 2 1 17 14 OCH H H OH Galleon R4 3 13

16 15 OH H H OCH3 Juglanin C

R1 R2 OH H OH OH Pterocarine

Rhoiptelol Juglanin B

Figure 1.3. Examples of cyclic diarylheptanoids.

Naphthalene derivatives Natural naphthalene derivatives are another class of compounds typically found in Juglans spp. samples, namely in the leaves and husks. They can occur with a naphthalene scaffold (usually coupled to substituents such as sugars and hydroxyl groups), but they are also precursors of two other frequent sub-class of compounds: naphthoquinones and tetralones that can be also glycosylated, hydroxylated, etc. (Bi et al., 2016; Wang et al., 2017). Naphthalene derivatives can lead to structurally related compounds such as the pair juglone (5-hydroxynaphthalene-1,4-dione) and regiolone ((4R)-4,8-dihydroxy-3,4-dihydro- 2H-naphthalen-1-one) depicted in part A of Figure 1.4 as well as the similar naphthalene, naphthoquinone and tetralone derivatives presented in part B (Bi et al., 2016; Talapatra et al., 1988). Isolated natural naphthalene derivatives (including tetralones and naphthoquinones) have shown cytotoxic potential against the HepG2 (Li et al., 2008b; Zhou et al., 2015), HeLa (Li et al., 2013a, 2008b), HL-60 (Li et al., 2008b; Soto-Maldonado et al., 2019), A549, MCF7, BEL-7402, COLO205, BGC-823 and SK-OV-3 (Li et al., 2013a) cell lines.

11 General introduction CHAPTER 1

Nevertheless, the glycosylated forms of these group of molecules have shown to be less active than the corresponding aglycones (Zhou et al., 2015).

Figure 1.4. General structure of naphthalene derivatives and examples of natural naphthoquinone (orange) and tetralone (blue) derivatives (juglone and regiolone, respectively) present in J. regia samples (part A). Example of a natural naphthalene derivative (green) and its structurally similar naphthoquinone and tetralone derivatives (part B).

Beyond the interest on those natural compounds, much attention has been applied to the development of new synthetic molecules with a naphthalene (derivative) skeleton. One of the examples, is the use of tetralone based structures to design new pharmaceuticals (Carro et al., 2014; Manvar et al., 2015), as well as mediator or starting materials for the synthesis of new compounds with biological activity and other interesting properties such as steroids, prostaglandin analogues, dyes and herbicides (Karmakar and Mal, 2011; Silveira et al., 2004; Zhang et al., 2018a; Zhu and Koltunov, 2016). The same for the synthetic molecules with a naphthalene core, from which several drugs were developed such as propranolol (antihypertensive) and naproxen (anti-inflammatory), among others, as recently reviewed by Makar et al. (2019). Thus, the abundancy on naphthalene derivatives on J. regia samples could also contribute to the biological effects found in their extracts.

12 General introduction CHAPTER 1

Overall, the bioactive compounds found in the extracts of J. regia leaves and husks can be potentially used as ingredients in different sectors (Chatrabnous et al., 2018; Jahanban- Esfahlan et al., 2019; Ribeiro et al., 2015).

1.1.2. Solvents used in the extraction process Solvents are widely applied at laboratorial and industrial scales, constituting an important factor in the design of different reaction and separation processes. The extraction process, in particular, is the main step for the recovery and isolation of valuable fractions from natural resources. The proper choice of a solvent and extraction technique are crucial to design an efficient process (Rostagno and Prado, 2013; Zhang et al., 2016a). Conventionally, organic solvents such as light alcohols (methanol and ethanol), acetone, diethyl ether, ethyl acetate (Kerton, 2009; Stalikas, 2007), acetonitrile or even chlorinated solvents (e.g. chloroform, carbon tetrachloride and chlorobenzene) (Tiwari, 2015) are used, being a significant fraction of the consumables in analytical laboratories and in different industrial sectors (Dai et al., 2013b; Pena-Pereira et al., 2015). Despite their extraction efficiency and wide application, conventional volatile organic solvents present some drawbacks such as flammability, toxicity, environmental ubiquity due to their volatility, etc.. Therefore, the design of more efficient, sustainable, and safer extraction processes is crucial in several research and industrial areas (Chemat et al., 2012).

In the following sections, a review of the main solvents and techniques applied in the extraction of bioactive compounds from J. regia leaves and husks is presented, and after a brief introduction about the use of glycerol, alkanediols and deep eutectic solvents for the extraction of bioactive compounds from plant material is given.

1.1.2.1. Conventional solvents The selection of a solvent depends on the nature of the target molecules and natural matrices. In Table 1.1, a summary of the preferred solvents for the extraction of bioactive compounds from walnut leaves and husks is presented. The information is organized according to the solvent, extraction technique, family of compounds identified, geographical origin and main biological activities for each plant material (leaves or husks).

13 General introduction CHAPTER 1

Table 1.1. Overview about the extraction studies of bioactive compounds from J. regia husks and leaves.

Solvent Technique Main compounds Bioactivities Origin References

Husks

Acetone + water Reflux Sterols Antioxidant Iran (Meshkini and Tahmasbi, 2017) (70 %, v/v) Phenolic acids Anti-platelet aggregation α-Tocopherol Juglone Others Acetone + water Stirring Phenolic acids Antioxidant Poland (Rywaniak et al., 2015) (70 %, v/v) Flavonols Anti-platelet aggregation Tannins Acetone + water UAE Juglone Iran (Beiki et al., 2018) (70 %, v/v) Diethyl ether n.p. Fatty acids n.p. Iran (Chatrabnous et al., 2018) Ethanol Soxhlet Phenolic acids Antioxidant4 Iran (Chatrabnous et al., 2018) Ethanol + water Maceration Diarylheptanoids Cytotoxicity2 China (Yang et al., 2019) (95 %) Methanol HAE Juglone Antioxidant Chile (Soto-Maldonado et al., 2019) Ethanol Anti-proliferative Methanol Reflux Diarylheptanoids n.p. China (Liu et al., 2008) Tetralone derivatives Methanol + 1 % BHT UAE Phenolic acids n.p. Romania (Cosmulescu et al., 2010) Myricetin Juglone Methanol + water Maceration Phenolic acids Antioxidant Iran (Akbari et al., 2012) (50 %) Flavonoids Methanol + water UAE Diarylheptanoids n.p. China (Li et al., 2008a) (80 %, v/v) Sclerone Ethanol + water n.p. Juglanone A and B Cytotoxicity2 China (Li et al., 2013a) (95 %)

Husks and leaves

Ethanol + water Maceration Phenolic acids Antioxidant Poland (Gawlik-Dziki et al., 2014) (70 %, v/v) Flavonols Methanol + 1 % BHT UAE Juglone n.p. Romania (Cosmulescu et al., 2011) Water Decoction Volatile compounds3 Antimicrobial Turkey (Keskin et al., 2012)

Leaves

Ethanol + water MAE Phenolic acids Antioxidant China (Zhao et al., 2014) (60 %, v/v) Flavonoids Cytotoxicity5

14 General introduction CHAPTER 1

Esculetin Ethanol + water Maceration Phenolic acids Antioxidant Poland (Gawlik-Dziki et al., 2014) (70 %, v/v) Flavonols Ethanol + water Maceration Saccharides n.p. Portugal (Santos et al., 2013) (80 %, v/v) Metaphosphoric acid + water Maceration Organic acids n.p. Portugal (Santos et al., 2013) (4 %) Methanol Maceration Phenolic compounds Antioxidant Portugal (Santos et al., 2013) Water Decoction Antitumor Hepatotoxicity Methanol n.p. Phenolic acids n.p. Portugal (Amaral et al., 2004) Chlorophorm Flavonols Acidified water (pH 2, HCl) Methanol Accelerated Solvent Phenolic acids Hypoglycemic Austria (Pitschmann et al., 2014) Extraction Flavonols Trihydroxynaphthalene-hexoside Methanol + 1 % BHT UAE Juglone n.p. Romania (Cosmulescu et al., 2011) Methanol + 1 % BHT UAE Phenolic compounds n.p. Romania (Cosmulescu et al., 2014b) Juglone Methanol + 1 % BHT UAE Phenolic compounds n.p. Romania (Cosmulescu et al., 2014a) Juglone Methanol + water Soxhlet Phenolic acids α-amylase inhibitory activity Italy (Mollica et al., 2017) (60 %) Decoction Flavonoids Antidiabetic MAE Acute oral toxicity Methanol + water UAE Megastigmanes n.p. Serbia (Schwindl and Kraus, 2017) (70 %, v/v) Tetralone derivatives Phenylpropanoids Juglone glycosides Methanol + water n.p. Diarylheptanoids n.p. Germany (Schwindl et al., 2019) (70 %, w/w) Ascorbic acid derivatives Juglone derivates Anthraquinones Flavonoids Proanthocyanidins Terpenes Methanol + water UAE Phenolic acids Antioxidant Poland (Wojdylo et al., 2007) (80 %) Flavonoids n.p. Methanol and hexane Vortex Tocopherols n.p. Portugal (Santos et al., 2013) Water Decoction Phenolic acids Antioxidant Portugal (Pereira et al., 2007) Flavonols Antibacterial Water Hydrodistillation Terpenes Antioxidant India (Rather et al., 2012) Antibacterial Water Hydrodistillation Terpenes n.p. India (Verma et al., 2013) n.p. not presented; HEA: heat assisted extraction; UAE: ultrassond assisted extraction; MAE: microwave assisted extraction.

15 General introduction CHAPTER 1

1- The antioxidant activity was assessed for an methanolic extract; 2- The anti-tumor activity was assessed for the isolated compounds; 3- Volatile compounds by GC-MS, ethylene oxide was the main compounds (83.67 %); 4- Optimized solvent composition, percentages for the correspondent technique of extraction; 5- Cytotoxic potential of the purified flavonoid extract, but no effect was observed.

16 General introduction CHAPTER 1 As can be seen, methanol and ethanol (and their mixtures with water) were the most common solvents used, as frequently done in the extraction of added-value compounds (such as phenolics) from several natural sources (Rostagno and Prado, 2013). Nevertheless, ethanol is replacing methanol since they often share advantages (high extraction yields, low boiling point), being ethanol a more benign choice due to its lower toxicity (Prat et al., 2016). Furthermore, ethanol can be considered a sustainable solvent when produced from renewable sources, traditionally through the conversion of accessible sugars from sugarcane and corn (Kerton, 2009; Tatsis and Connor, 2016).

1.1.2.2. Glycerol and alkanediols Nowadays, the transformation of biomass in added-value chemicals and fuels is a hot topic within the scientific and industrial communities (Yfanti and Lemonidou, 2019). It entails the production of biofuels, in which bioethanol currently represents the major biofuel in use (Tatsis and Connor, 2016). On the other hand, biodiesel also represents an important product from the biomass conversion to fuels, searching for an alternative and sustainable strategy to petrodiesel as a source of energy. Nevertheless, during the biodiesel production chain, by-products are formed, namely glycerol. In this context, for the energy sector, glycerol is considered as the major bottleneck of the process since the transesterification reaction produces biodiesel and glycerol at volumetric ratio about 10:1 (Monteiro et al., 2018). In return, glycerol can be used as starting material for the production of 1,2-propanediol and 1,3-propanediol (e.g. by hydrogenolysis), among other chemicals (e.g. lactic acid, glyceric acid, other propanol isomers) (Sun et al., 2016). This process has high commercial value since both C3-diols present a wide range of applications (e.g. solvents, cosmetics and food additives) (European Commission, 2006; Lavaud et al., 2016; Martins et al., 2019). In this regard, biofuel by- products and their derivatives are atractive alternatives for several processes, including as low-cost renewable solvents for extraction purposes (Monteiro et al., 2018). Table 1.2 summarizes the few studies reporting the use of glycerol and propanediols as solvents in the extraction of bioactive ingredients from plant material. In general, they are preferably used in aqueous mixtures, but mixtures with ethanol were also applied (Baranauskaitė et al., 2016; Lahucky et al., 2010; Moraes et al., 2016; Philippi et al., 2016).

17 General introduction CHAPTER 1

Table 1.2. Overview about the use of glycerol and alkanediols for extraction purposes.

Plant material Origin Solvent Technique Main compounds Bioactivities References

Glycerol-based solvents

Artemisia arborescens L. Greece Glycerol + water Stirring + heating Total phenolic content2 Antioxidant (Shehata et al., 2015) Artemisia inculta Delile (90 %, w/v)1 Phenolic acids Ferric reducing power (aerial parts) Flavonoids DPPH Others Hypericum perforatum L. (aerial Germany Glycerol + water Stirring + heating Total phenolic content2 Antioxidant (Karakashov et al., 2015) parts) (10 %, w/v)1 Phenolic acids Ferric reducing power Flavonols Olea europaea L. Greece Glycerol + water Stirring + heating Total phenolic content2 n.p. (Apostolakis et al., 2014) (leaves) (9.3 %, w/v)1 Flavonoids Others Origanum onites L. Lithuania Glycerol + ethanol UAE Rosmarinic acid n.p. (Baranauskaitė et al., 2016) Origanum vulgare spp. hirtum Turkey (80–100 %, v/v) Heat-reflux Others Origanum vulgare L. Ethanol + water Stirring (herbs) (30-96 %, v/v) Maceration Methanol Percolation Oryza sativa L. China Glycerol + water (19.47 %)1 Shaking + heating Total phenolic content2 n.p. (Huang et al., 2018) (rice bran) Phenolic acids Solanum melongena L. Greece Glycerol + water UAE Total phenolic content2 Antioxidant (Philippi et al., 2016) (peels) (90 %, w/v) Phenolic acids Ferric reducing power Ethanol + water Flavonoids DPPH (40 %, v/v)

Propylene glycol-based solvents

Calendula officinalis L. Brazil Propylene glycol + ethanol Shaking Phenolic acids n.p. (Moraes et al., 2016) (flowers) + water Flavonoids (0.4 + 0.4 + 0.2, v/v/v) 3 Mellissa officinalis L. Slovakia Ethanol + propylene glycol6 Commercial Total phenolic content2 Antioxidant (Lahucky et al., 2010) Origanum vulgaris L. extracts ABTS Salvia officinalis L. TBARS (herbs) Malus sylvestris (L.) Mill. Serbia Ethanol + water (70 %, v/v) Maceration Total phenolic content2 Antioxidant (Stojiljković et al., 2016) (wild fruit) Propylene glycol + water Percolation Total flavonoids4 DPPH (45 %, w/w) Soxhlet Ultrasonic Total tannins5 Ferric reducing power Propylene glycol + water (8 Inhibition of linoleic acid %, w/w) oxidation water

18 General introduction CHAPTER 1

Origanum onites L. Lithuania Propylene glycol + ethanol UAE Rosmarinic acid n.p. (Baranauskaitė et al., 2016) Origanum vulgare spp. hirtum Turkey (70–9 0 %, v/v) Heat-reflux Others Origanum vulgare L. Ethanol + water Stirring (herbs) (30-96 %, v/v) Maceration Methanol Percolation Pandanus amaryllifolius Roxb. Thailand Propylene glycol Maceration n.p. Antioxidant (Jimtaisong and Krisdaphong, (leaf and root) Ethanol (95 %) DPPH 2013) Propylene glycol + Linoleic acid emulsion– ethanol (1:4 and 1:1, v/v) thiocyanate method Stevia rebaudiana Bert. Poland Propylene glycol + water Stirring Total phenolic content2 Antioxidant (Gaweł-Bȩben et al., 2015) (leaf) (4:1) Phenolic acids DPPH Ethanol (96 %) Flavonoids ABTS Ferric reducing power Cytotoxicity7 CRL-2522 Terminalia chebula Retz Thailand Ethanol + water (76.4 %, Reflux Total phenolic content2 Antioxidant (Tubtimdee and Shotipruk, (dried fruits) v/v) 1 Gallic acid ABTS 2011) Propylene glycol + water Ellagic acid (36 %, v/v) 1 Terminalia cheubula Retz Thailand Ethanol + water (30, 50, 70 Maceration Total phenolic content2 Antioxidant (Chulasiri et al., 2011) (fruits) and 100 %) DPPH Propylene glycol + water H2O2 inhibition (30, 50, 70 and 100 %) AAPH induced haemolysis ABTS Photochemiluminescence Vitis vinifera L. Italy Propylene glycol + ethanol Stirring Total phenolic content2 Antioxidant (Manconi et al., 2017) (pomace) (1:1 and 1:3, Gallic acid DPPH v/v) Flavonoids AAPH induced haemolysis

Ethylene glycol-based solvents

Morus alba L. Korea Ethylene glycol + water (25, Heat extraction Total phenolic content8 Antioxidant (Kim et al., 2007) (leaf) 42 and 58%) DPPH Acetone + water (47 and 57 %) Acetone + methanol (27 %)

19 General introduction CHAPTER 1

Regarding the use of propylene glycol, in some cases, higher extraction yields were achieved by applying aqueous mixtures of this solvent to recover target compounds compared to the use of water and/or ethanol. That was the case of the total HPLC phenolic content obtained from Stevia rebaudiana leaves (Gaweł-Bȩben et al., 2015) and the total phenolic contents determined in Terminalia chebula fruits (Chulasiri et al., 2011) and Malus sylvestris wild fruit (Stojiljković et al., 2016). Moreover, the pure solvent also extracted a higher amount of phenolic compounds from Pandanus amaryllifolius leaves (Jimtaisong and Krisdaphong, 2013). On the other hand, the aqueous propylene glycol extract of Terminalia chebula fruits contained lower amounts of gallic and ellagic acids comparing to the hydroethanolic extract (Tubtimdee and Shotipruk, 2011).

Concerning the use of glycerol, their aqueous mixtures seem to perform better than pure water as extraction solvent (Apostolakis et al., 2014; Karakashov et al., 2015). Higher total phenolic contents were obtained, using small amounts of glycerol in water, from Olea europaea leaves (9.3 %, w/v) (Apostolakis et al., 2014), Hypericum perforatum dried aerial parts (10 %, w/v) (Karakashov et al., 2015) and Oryza sativa bran (20 %, v/v) (Huang et al., 2018). On the other hand, the increasing concentration of glycerol (10 to 90 %, w/v) lead to a higher total phenolic content when applied to two Artemisia species (Shehata et al., 2015). Philippi et al. (2016) compared the extraction yields of aqueous solutions of glycerol or ethanol, obtaining similar total phenolic contents in the extracts of Solanum melongena (Philippi et al., 2016). Finally, a higher total phenolic content was obtained from Morus alba L. when using aqueous solutions of ethylene glycol, compared to acetone + water/methanol mixtures (Kim et al., 2007).

Besides the aqueous solutions of alcohols, other combinations were also studied. Binary mixtures of ethanol + propylene glycol or glycerol were more effective than ethanol for the extraction of rosmarinic acid, carvacrol, oleanolic acid and ursolic acid from three Origanun species by HAE (Baranauskaitė et al., 2016), while Moraes et al. (2016) found higher amounts of isorhamnetin-3-O-rutinoside in Calendula officinalis L. flowers using ethanol + propylene glycol + water equimolar solvent than using pure solvents or binary mixtures. Regarding the biological activity of the extracts obtained using glycerol or short chain alkanediols as extraction solvents, most studies evaluated the antioxidant potential of the

20 General introduction CHAPTER 1 different plant liquid extracts (Table 1.2). In general, the DPPH (2,2-diphenyl-1- picrylhydrazyl) assay was applied. Chulasiri et al. (2011) reported that the propylene glycol extracts from Terminalia cheubula fruits showed higher antioxidant potential than the ethanolic ones. With Stevia rebaudiana leaves, a similar result was achieved, being the glycolic-aqueous extract more active than the aqueous and ethanolic ones (Gaweł-Bȩben et al., 2015). Stojiljković et al. (2016) found similar antioxidant activity between the aqueous ethanol (70 %) and aqueous propylene glycol (45 %) extracts of Malus sylvestris wild fruits. Regarding the use of glycerol + water mixtures, Philippi et al. (2016) obtained a lower antioxidant activity of Solanum melongena peels extract compared to the hydroethanolic one. Concerning the antioxidant potential of the extracts obtained with ethylene glycol, it was found that those extracts of Morus alba leaves presented higher activity with 58 % of ethylene glycol when compared to extracts containing lower ethylene glycol concentrations or even acetone-based solvents (Kim et al., 2007).

Considering the reducing power assay, Karakashov et al. (2015) confirmed the higher antioxidant potential of Hypericum perforatum dried aerial parts obtained using aqueous glycerol (10 %) extracts than the ones found using water. On the other hand, Solanum melongena peels aqueous glycerol extract showed a lower reducing power than the conventional hydroethanolic extract (Philippi et al., 2016).

Regarding the assays using aqueous propylene glycol, a higher antioxidant activity of the Terminalia chebula Retzius fruits extracts was found by the ABTS assay when compared to the hydroethanolic extract (Tubtimdee and Shotipruk, 2011).

Finally, the evaluation of the cytotoxicity of the liquid extracts obtained with these solvents was seldom carried out. A comparative study about the cytotoxic potential of Stevia rebaudiana leaves extracts and the selected solvents (water, ethanol and aqueous propylene glycol) was performed by Gaweł-Bȩben et al. (2015) using skin fibroblasts cell as a model to evaluate the extracts potential as food and cosmetic ingredients. Generally, the extracts have shown higher cytotoxic potential than the solvents for the same solvent concentration. Furthermore, only the aqueous ethanolic and propylene glycol extracts (between 2 % and 5 %) promoted a significant decrease on the skin fibroblasts viability.

21 General introduction CHAPTER 1

Thus, the authors concluded that the potential of those extracts for cosmetic or food formulations are dependent of further studies to establish an appropriate dose, adapted to the desirable product.

Overall, the advantages of using alkanediols and glycerol for extraction processes goes beyond the industrial and environmental benefits previously discussed (integrated green recovery of phenolic compounds) (Huang et al., 2018). Additionally, the study of the direct incorporation of the resulting extracts in cosmetic, pharmaceutical and nutraceutical formulations is being studied (Huang et al., 2018; Lavaud et al., 2016; Philippi et al., 2016), since the solvents are used as ingredients in such products (European Commission, 2006; Martins et al., 2019), thus, leading to processes with less steps (e.g. purification, evaporation of the solvent) (Monteiro et al., 2018) and wastes (Chemat et al., 2012; Huang et al., 2018). Besides their use as pure solvents or in mixtures with ethanol or water, glycerol and alkanediols are also being studied as components of a new generation of green solvents, called deep eutectic solvents (DES). They will be introduced in the next section.

1.1.2.3. Deep eutectic solvents Deep eutectic solvents (DES) are an emerging class of solvents introduced by Abbott et al. (2003), describing eutectic mixtures of amides and quaternary ammonium salts with interesting solvent properties. According to the authors research group (Abbott et al., 2004), they are prepared by mixing two (or more) starting materials (hydrogen bond acceptor - HBA and hydrogen bond donor - HBD), suggesting that the hydrogen bonding between both compounds induces a significant decrease of the melting point (∆T) of the mixtures compared to the melting point of the individual components (Figure 1.5).

An eutectic point is an isobaric invariant of the system, and represents the composition and the minimum melting temperature along the two intersecting melting curves (Gamsjäger et al., 2008). The solid-liquid equilibria can be described by the following equation (Eq. 1.1) from classical thermodynamics, considering pure solid phases and neglecting the temperature influence on the heat capacities(Prausnitz et al., 1999):

훥푚퐻 1 1 훥푚퐶푃 푇푚 푇푚 ln(푥푖훾푖) = ( − ) + ( − ln − 1) [1.1] 푅 푇푚 푇 푅 푇 푇

22 General introduction CHAPTER 1

where γi is activity coefficient of compound i at a certain liquid mole fraction composition xi, T is the absolute temperature, Tm and ΔmH are the melting temperature and enthalpy of the pure compound, respectively, R is the universal gas constant, and ΔmCp is the difference between the molar heat

capacity of compound i in the THBA x = 0 liquid and solid phases (Martins

THBD et al., 2018). Furthermore, x = 1 when the equilibrium

temperature and the melting Temperature temperature of the pure TE, ideal

∆T compound are not far each TOperating other, the last term of the TE equation can be neglected x x when it assumes a small value 1 2 Mole fraction Figure 1.5. Representation of the solid-liquid equilibria of comparing to melting enthalpy a eutectic mixture (blue) and a deep eutectic mixture term. Therefore, for ideal (green). Adapted from Martins et al. (2018). systems (γ = 1), it is possible to plot the melting curves of mixtures using only the melting properties of the compounds (blue line of Figure 1.5) by modifying the xi composition.

The DES concept was discussed by Martins et al. (2018) in a review, because different definitions could be found for the same concept. According to the authors, the main reason could be attributed to the limited number of studies devoted to understand the DES nature, which should be distinct of eutectic mixtures (Figure 1.5). Thus, some adjustments to the initial definition were proposed namely:

1. In the case of a DES, a temperature depression is observed (TE) compared to the ideal

behaviour of the mixture (TE, ideal); 2. A DES can be any composition of the liquid-phase region at a desirable temperature

(TOperating) as suggested by the region between x1 and x2 mole fractions, and not at a specific stoichiometric proportion.

23 General introduction CHAPTER 1

700

600 Chemistry Chemical Engineering

500 Materials Science Physics and Astronomy Environmental Science 400 * Energy

Biochemistry,Chemistry Genetics and Molecular BiologyChemical Engineering 300 EngineeringMaterials Science Physics and Astronomy Environmental Science Energy Number of of Number publications Agricultural and Biological Sciences Biochemistry, Genetics and Molecular Biology Engineering Pharmacology, Toxicology and Pharmaceutics 200 Agricultural and Biological Sciences Pharmacology, Toxicology and Pharmaceutics Medicine Medicine Immunology and Microbiolog ImmunologyMultidisciplinary and Microbiolog Computer Science Others 100 Multidisciplinary Computer Science

0 Others 2004 2009 2014 2019 Figure 1.6. Distribution of the publications related to the topic “deep eutectic solvent”. Data retrieved from Scopus (May 28, 2019).

Since DES were introduced by Abbott et al. (2003), the number of publications about this topic has significantly increased, being most of them in the areas of chemistry, chemical engineering and materials science (Figure 1.6). Until the end of May 2019, the number of papers available on Scopus for the keywords “deep eutectic solvent” was about 2633 documents, distributed by several subject areas, including 126 reviews.

The search of new eco-friendly solvents constitutes one of the most attractive topics of green chemistry and green analytical chemistry (Płotka-Wasylka et al., 2017). In this regard, during the past decade, DES have been presented as a new generation of solvents with some advantages when compared to conventional VOC, as well as over other alternative solvents such as supercritical fluids and ionic liquids (Huang et al., 2019; Ruesgas-Ramón et al., 2017), depending on the final application.

As DES are the result of combining (at least) two components, a large number of mixtures can be envisioned. Several combinations of HBA (mainly choline chloride and betaine) and HBD (sugars, polyols, alcohols, acids, amides, etc.) have been proposed (Figure 1.7). This division between HBA and HBD should not be seen in a strict way, as most compounds can be both hydrogen bond donors and acceptors and the interactions in these eutectic systems are quite complex. Consequently, different solvent properties can be tuned (range of liquid state, viscosity, polarity and solubilization capacity), conferring a character of

24 General introduction CHAPTER 1 designer solvents to these new generation of liquids (Fernández et al., 2018b). In most cases, where DES were used as extraction solvents, a given amount of water was also added to decrease the viscosity of the mixtures, being an important variable when optimizing the extraction yields. The selection of benign starting materials, such as many of the examples shown in Figure 1.7, has been also used as an argument to support their biodegrability, biocompability and low toxicity (Fernández et al., 2018a; Florindo et al., 2019). Also, they are easy to prepare (usually by heating, grinding or evaporation procedures) and, in many cases, can be formed by low cost compounds (Płotka-Wasylka et al., 2017; Yang et al., 2019).

They present attractive physicochemical properties such as their low vapor pressure (García et al., 2016). On the other hand, for some applications, it’s necessary to remove the extracted compounds from the solvent. The challenge of removing (and recycling) the solvent from extracts motivated different approaches such as the use of microporous resins, supercritical carbon dioxide, solid-phase extraction (SPE) and the use of anti- solvents (Huang et al., 2019; Ruesgas-Ramón et al., 2017). Nevertheless, the search for solvents compatible with the final product is being studied, aiming to obtain liquid extracts that can be directly used in a final application (Ruesgas-Ramón et al., 2017). Furthermore, some studies have shown the higher thermal stability of biomolecules in this type of solvents (Dai et al., 2016; Fernández et al., 2018b; Zainal-Abidin et al., 2017). An interesting application was reported by Dai et al. (2014) that studied the stability of both pure natural colourant (carthamin, a phenolic compound) and extracts from safflower against heating, light, storage time, ambient conditions in sunlight, and water content variations, showing greater stability in sugar-based NADES when compared to water and 40 % ethanol.

25 General introduction CHAPTER 1

HBD HBA HBD DES Sugars Alcohols Acids HBA Ethyleneglycol

Glucose 1,2-Propanediol Choline Acetic acid Propionic acid chloride 1,4-Butanediol

Fructose Glycerol

Betaine Polyols / Sugar alcohols Oxalic acid Lactic acid

Sucrose Xylitol Proline Malic acid Sorbitol

Levulinic acid

Acetamide Urea Tetraethylammonium Others chloride Phenylpropionic acid

Figure 1.7. Examples of starting materials used to form DES.

26 General introduction CHAPTER 1

As recently reviewed by Benvenutti et al. (2019), Choi and Verpoorte(2019), Fernández et al. (2018a), Huang et al. (2019), Jablonský et al. (2018), Liu et al. (2018b), Ruesgas-Ramón et al. (2017) and Vanda et al. (2018), DES have become an interesting alternative for the extraction of bioactive components from natural sources, including phenolic compounds. Nevertheless, the choice of the best DES is still a challenge. An important contribution is given by Duan et al. (2016). The authors compared the extraction capacity of a series of DES, including 43 combinations between choline chloride/betaine/L-proline and amides, carboxylic acid, alcohols and sugars to extract alkaloids, flavonoids, saponins, anthraquinones, and phenolic acids from different natural sources. In summary, the authors reported the following pairs of phytochemical vs DES according to best extraction yields: alkaloids and carboxylic acids based-DES; saponins and amide based-DES; phenolic acids and choline chloride based-DES; flavonoids and L-proline based-DES. All these examples outperformed the methanol capacity; nevertheless, none of the prepared solvents performed better than methanol to extract anthraquinones, probably due to the lower polarity of this class of natural compounds. In another study, Zhuang et al. (2017) verified that acidic DES (levulinic acid) lead to higher extraction yields of flavonoids from Platycladus orientalis (L.) Franco dried branches and leaves, with a broader extraction capacity than sugar, alcohol and amide based-DES. On the other hand, the ionic character of some HBA’s (e.g. choline chloride and betaine) may also contribute to the solubilization and extraction of compounds with a wider polarity range (Huang et al., 2019).

The path for a sustainable, eco-friendly and cost-effective process during the extraction of added-value compounds from natural matrices has also been carried out by studying alternative mechanisms of extraction. In fact, by applying techniques such as microwave assisted extraction (MAE) and ultrassond assisted extraction (UAE), researchers were capable to improve extraction yields accompanied with a simultaneous reduction of the extraction of time, not only by using conventional solvents but also DES (Cunha and Fernandes, 2018). Independently of the selected technique (and optimized conditions of extraction as well), the extraction yields of eutectic and conventional solvents must be considered as well as the bioactivity of the extracts (Murador et al., 2019; Zainal-Abidin et al., 2017). In fact, as important as evaluating the toxicity of the solvent, it is also necessary

27 General introduction CHAPTER 1 to consider the toxicity of the final product, extract plus solvent. (Murador et al., 2019). So far, most literature data are about the antioxidant potential of DES extracts (Athanasiadis et al., 2017; Bakirtzi et al., 2016; Cao et al., 2018; Jancheva et al., 2017; Jeong et al., 2018; Kottaras et al., 2017; Nam et al., 2015; Ozturk et al., 2018; Peng et al., 2018; Radošević et al., 2016; Rajan et al., 2015) and the toxicologic profile of the pure solvents (Castro et al., 2018; De Morais et al., 2015; Hayyan et al., 2013; Huang et al., 2017; Macário et al., 2019, 2018b, 2018a; Mbous et al., 2017b; Paiva et al., 2014; Radošević et al., 2016, 2015, 2018; Wen et al., 2015; Zhao et al., 2015). The bioactivity of the liquid extract was less explored in terms of cytotoxicity and other biological activities besides the antioxidant potential (Lavaud et al., 2016; Murador et al., 2019; Radošević et al., 2016).

DES can be used as selective solvents, not only for the enrichment of target compounds, but also to check their affinity to extract toxic molecules. The study made by Liu et al. (2018a) showed the possibility to obtain high-quality Gingko biloba L. leaves preparations. The authors found significantly lower amounts of ginkgolic acids which present toxicity, but higher phenolic contents (chlorogenic acid, rutin and quercetin) using several DES instead of methanol. Therefore, the tunable properties of DES showed advantages over conventional solvents (methanol). Moreover, they can also be used as an alternative in analytical chemistry. In this context, choline chloride : glycerol (1:2) and choline chloride : urea (1:2) (both with concentrations of water up to 40 %, w/w) were proposed for the extraction and determination of the mycotoxin Ochratoxin A from wheat and derived samples suggesting a new green strategy in the area of food quality control (Piemontese et al., 2017).

To conclude, fundamental studies exploring different types of eutectic mixtures, with systematic variations in terms of the molecular structures of the components (e.g. functional groups, alkyl chain length, etc.) should be carried out, to evaluate the extraction of different classes of compounds from a given plant material. On the other hand, the selection of solvents should also consider the final application of the extract.

28 General introduction CHAPTER 1

1.2. Scope and objectives Natural matrices play an important role in the food, pharmaceutical and cosmetic industries due to the valuable bioactive compounds present in their extracts (Martins et al., 2011). In particular, phenolic compounds stand out as relevant and abundant natural antioxidants and chemopreventive agents (Carocho and Ferreira, 2013).

One of the main aims of this work is to characterize the phytochemical profile and bioactivity of the extracts from the leaves and green husks of Juglans regia L. using ethanol and water as solvents. First, in Chapter 2, the process of extracting phenolic compounds from the leaves was optimized using the ethanol + water mixed solvent and two techniques (maceration, also referred as heat assisted extraction in later chapters, and microwave assisted extractions). Temperature, time, solvent composition and solid-liquid ratio were the optimized extraction parameters. After, the phytochemical profile and the bioactivity of the extracts of leaves in two phenological stages were studied, under the previously optimized extraction conditions (Section 2.2.). In addition, the green husks of walnuts were also characterized (phytochemical composition and bioactivities) due to their relevance as a by-product of walnuts production and thus contributing to their valorization (Section 2.3.).

The search for alternative solvents that are more efficient, but also sustainable and biocompatible, was the purpose of the work developed in Chapter 3. During this chapter, walnut leaves were studied as the source of phenolic compounds and different solvents were explored, aiming to understand the main factors affecting the extraction process. Firstly, a preliminary screening of choline chloride and carboxylic acids based eutectic solvents was carried out (Section 3.1.). The extraction conditions (water content, temperature, time and solid-liquid ratio) using the best DES were optimized. In a second stage, glycerol and a series of alkanediols were studied as they are liquid at room temperature (Section 3.2.). Again, the best candidate was further studied, by forming eutectic mixtures with betaine or choline chloride. The composition of the ternary mixture (alcohol + choline chloride or betaine + water) was optimized in order to obtain the highest extraction of phenolic compounds. The bioactivity of both extracts and solvents was evaluated.

29 General introduction CHAPTER 1

Finally, experimental and thermodynamic modelling studies about the solubility of different classes of phenolic compounds are fundamental to support the initial screening of solvents, hopefully decreasing the experimental workload. That information is however scarce. Therefore, in Chapter 4, the solubility of three aromatic acids (vanillic, syringic and veratric acis) were studied due to their relevance as phytochemicals. Two semi-predictive thermodynamic models were also applied. Finally, the main conclusions of the developed work are addressed in Chapter 5, that also includes the envisioned future work.

30

2. Characterization of the plant material

Characterization of the plant material CHAPTER 2

2.1. Optimization and comparison of maceration and microwave extraction systems for the production of phenolic compounds from Juglans regia L. for the valorization of walnut leaves Vieira, V., Prieto, M.A., Barros, L., Coutinho, J.A.P, Ferreira, O., Ferreira, I.C.F.R., 2017. Ind. Crops. Prod. 107, 341-352.1

Abstract

This study compares maceration (ME) and microwave assisted (MAE) extractions aiming to maximize the extraction of valuable compounds from Juglans regia L. leaves. An experimental design, assisted by response surface methodology, was developed to optimize the variables of the ME and MAE systems (time, temperature and ethanol-water proportion). The responses used as criteria were the quantification by HPLC-DAD (3-O- caffeoylquinic acid, quercetin 3-O-glucoside and quercetin O-pentoside), the extraction yield and the spectrophotometric results of total phenolics and flavonols. The global optimum conditions were: 115.6 min, 61.3 °C and 50.4 % of ethanol for ME; and 3.0 min, 107.5 °C and 67.9 % of ethanol for MAE. The solid/liquid ratio effect was studied at the optimal values showing a decreasing linear relation until 140 g/L for all criteria assessed. This study contributes to the valorization of walnut leaves as source of valuable compounds (e.g., quercetin) to be used as ingredients for the development of functional foods.

Keywords: Phenolic compounds; Microwave-assisted extraction; Maceration; Functional food ingredients; Juglans regia L.

1 My contribution to this work was the measurement of all experimental data, the interpretation of the RSM results and writing of the manuscript.

33 Characterization of the plant material CHAPTER 2

2.1.1. Introduction There have been an increasing number of studies on extraction methodologies of secondary metabolites from plants, namely phenolic compounds, due to their bioactivity and manifest effects on oxidative processes related to several health diseases (Carocho and Ferreira, 2013). Some of these bioactive compounds are used as nutraceuticals due to their benefits in the prevention or treatment of health conditions (Bravo and Mateos, 2008). Moreover, they have been widely used in pharmaceuticals, functional foods and natural cosmetics (Martins et al., 2011). Juglans regia L. (walnut) leaves have been applied in traditional medicine for topical and enteric use with different purposes (Carvalho and Morales, 2010; Guarrera, 2005) and several studies have reported the in vitro bioactivity of its extracts including antioxidant (Almeida et al., 2008; Carvalho et al., 2010; Pereira et al., 2007; Santos et al., 2013), antitumor (Santos et al., 2013), anti-inflammatory, and antiproliferative (Carvalho et al., 2010) properties. To demonstrate the bioactivity of the extracts, some in vivo models were also applied. Erdemoglu and co-workers (Erdemoglu et al., 2003) studied the oral administration of walnut leaves extracts with successful results on the anti-inflammatory activity using mice models inducing an antinociceptive action. The antidiabetic activity induced by walnut leaves extracts through rats models were also described in previous reports (Asgary et al., 2008; Mohammadi et al., 2012) inducing the reduction of blood sugar levels as well as the decrease of HbA1C fraction. In addition, some of these studies refer the absence of cytotoxicity in porcine liver cells (Santos et al., 2013) or gastric damage in mice (Erdemoglu et al., 2003). Previous studies (Santos et al., 2013) indicated that the alcoholic extracts of walnut leaves have a promising bioactivity and their chemical characterization showed that they are a rich source of phenolic compounds, being 3-O-caffeoylquinic acid, quercetin 3-O-glucoside and quercetin O-pentoside the main molecules present in J. regia leaves. Examples of the individual importance of these compounds can be found. 3-O-caffeoylquinic acid acts as a

- scavenger on O2 capture, exerting an inhibitory effect against oxidation of methyl linoleate (Nakatani et al., 2000). Quercetin aglycones (namely 3-O-glucoside) are found to play a critical role in the inhibition of human neutrophil elastase (so, immunomodulatory action

34 Characterization of the plant material CHAPTER 2 in inflammatory responses) (Ryu et al., 2016) as well as in the cytotoxic protection against

H2O2, leading to a decrease in the generation of reactive oxygen species (Shokoohinia et al., 2015). Thus, extracts rich in these compounds could potentially be included in several industries due to their versatility (e.g., as ingredients in functional foods). Generally, 3-O-caffeoylquinic acid is found in small amounts in natural matrices: 0.015 mg/g dw in Vaccinium floribundum Kunth berries (Prencipe et al., 2014); 0.76 and 0.89 mg/g dw in Lonicera japonica Thunb. flowers (Li et al., 2014); 6.41 mg/g dw in J. regia leaves (Santos et al., 2013), being the highest content (17.2 mg/g dw) reported by Orqueda and co-workers (Orqueda et al., 2017) on Solanum betaceum skin extract. On the other hand, quercetin 3-O-glucoside is frequently present in plant material in higher quantities: 33.40 mg/g dw in Securigera securidaca flowers (Ibrahim et al., 2015), 36.7 mg/g dw in Dasiphora fruticosa (L.) Rydb. (Tomczyk et al., 2012), 0.218 mg/g dw in Vaccinium corymbosum L. (Zorenc et al., 2017); 0.494 mg/g dw in Salvia fruticosa Mill. Leaves (Sarrou et al., 2016). However, the presence of quercetin O-pentoside is not so recurrent: 0.0133 mg/g dw on Rosa canina hips (Cunja et al., 2015), 5.04 mg/g dw in J. regia leaves (Santos et al., 2013), 0.101 mg/g fw in V. floribundum berries (Prencipe et al., 2014). More information can be found in Table 2.1. The solid-liquid extraction of these compounds is usually carried out using organic solvents such as methanol, ethanol, acetone and ethyl acetate (Kerton, 2009). In particular, the most common solvents used by researchers are aqueous methanol and ethanol (Bravo and Mateos, 2008). The Food and Drug Administration (FDA) favours ethanol in the pharmaceutical and food industries due to its lower toxicity. In addition, this alcohol can be obtained from renewable sources and is safe for human consumption (Kerton, 2009). Conventional systems such as Soxhlet and maceration (ME) require considerable amounts of solvents and are time-consuming processes. Thus, modern extraction techniques are being applied in the extraction of valuable compounds from natural sources. Microwave assisted extraction (MAE) is one of the most employed alternative extraction methods, commonly using mixtures of ethanol and water as solvent. Among the advantages reported for MAE are its lower extraction times and solvent consumption, when compared to conventional methods (Destandau et al., 2013; Routray and Orsat, 2012). Independently of

35 Characterization of the plant material CHAPTER 2 the system applied on the extraction of target compounds from matrices, there are several process variables that require individual consideration due to the intrinsic nature and stability of these target compounds. Therefore, it is essential to identify the main variables involved, prior to the optimization process to maximize their responses, using the minimum time, energy and solvent consumption and designing the most cost-effective and profitable extraction system (Dai and Mumper, 2010). The most frequent form to carry out an optimization is by independently measuring the influence of each variable fixing all others. Nevertheless, the application of mathematical models such as the response surface methodology (RSM) is gaining visibility in the scientific community. The RSM design allows to optimize all the variables simultaneously and to predict the most efficient conditions. This is achieved by using second order polynomial models with interactions, that are able to describe and maximize the response criteria selected, based on the experimental range tested (Bezerra et al., 2008; Ferreira et al., 2007; Kalil and Maugeri, 2000). Therefore, this work aims to evaluate the extraction of phenolic compounds from J. regia leaves by RSM using two extraction methods (ME and MAE), with ethanol:water mixtures as solvent, in order to obtain the following targets: 1) develop a process in a pre-industrial form for the extraction of 3-O-caffeoylquinic acid, quercetin 3-O-glucoside and quercetin O-pentoside from J. regia (walnut) leaves; and 2) optimize the primary variable conditions of the ME and MAE systems (time, temperature, ethanol-water proportion and solid-liquid ratio) and maximize the compounds extraction assisted by the statistical RSM technique, contributing to the understanding of the potential of J. regia leaves for industrial applications. In supplementary material (Figure A1, Appendix A), a summary diagram containing the different steps carried out to obtain the optimal phenolic extract is presented.

36 Characterization of the plant material CHAPTER 2

Table 2.1. Bibliographic summary of the content of 3-O-caffeoylquinic acid (P1), quercetin 3-O-glucoside (P2) and quercetin O-pentoside (P3) from different source materials using different extraction techniques and conditions.

Technique Plant used for the extraction Content Extraction conditions References Source material Plant part Solvent and proport. Temp. time Others mg/g dw -- % °C min --

A: 3-O-Caffeoylquinic Acid quantification (P1)

Vaccinium floribundum Kunth Berries 0.015 EtOAC 100 RT 60 -- (Prencipe et al., 2014) ME Ipomoea batatas L. Leaf, stem & flower 0.13-0.83 MeOH:H2O 80:20 RT >500 -- (Jeng et al., 2015)

MAE Lonicera japonica Thunb. Flowers 0.89 EtOH:H2O 50:50 40-70 3-12 -- (Li et al., 2014)

Solanum betaceum Cav. Skin, pulp & seeds 13.2-17.2 EtOH:H2O 95:5 RT 30 -- (Orqueda et al., 2017) UAE Solanum betaceum Cav. Yellow & purple cultivars 0.25-1.64 MeOH:H2O 75:25 RT 60 -- (Espin et al., 2016) Vaccinium corymbosum L. Bluecrop, jersey & earliblue 0.06-0.61 MeOH:CH2O2 97:3 4-100 60 -- (Zorenc et al., 2017)

HRE Lonicera japonica Thunb. Flowers 0.76 EtOH:H2O 50:50 50 60 -- (Li et al., 2014)

B: Quercetin Derivaives Quantification (P2 & P3)

Salvia fruticosa Mill. Leaves 0.03 MeOH:H2O 80:20 4 >500 -- (Sarrou et al., 2016) Aspalathus linearis (Burm. F.) R. Leaves 0.49 H O 100 65-85 5-10 -- (Santos et al., 2016) Dahlgren 2

Vitis vinifera L. Grape skin <1 CC:OA:H2O 75:25 RT >500 -- (Bubalo et al., 2016) Juglans regia L. Leaves 7.41 MeOH 100 RT 120 -- (Santos et al., 2013) V. floribundum Berries 0.10 a EtOAC 100 RT 60 -- (Prencipe et al., 2014) Securigera securidaca L. Flowers 33.40 EtOH:H2O 90:10 4 -- -- (Ibrahim et al., 2015) ME Rubus plicatus Weihe & Nees Fruits 0.04 MeOH:CH2O2:H2O 60:3:37 ------(Kaume et al., 2012) (Bonaccorsi et al., Allium cepa L. Red & gold onions 0.001-0.025 a MeOH 100 RT >500 -- 2008) Juglans regia L. Leaves 7.64 H2O 100 100 RT 10 -- (Santos et al., 2013) (Bonaccorsi et al., Allium ascalonicum Hort. French & Italian shallot 0.006-0.015 a MeOH 100 RT >500 -- 2008) Grape berry skin Vitis vinifera L. 0.006-0.080 MeOH:H2O 80:20 RT 120 -- (Ferreira et al., 2016) White, red & black cultivars

Vitis vinifera L. Grape skin <1 CC:OA:H O 75:25 65 50 -- (Bubalo et al., 2016) MAE 2 Prunus laurocerasus L. Leaves 1.75 b MeOH 100 -- 10-30 300-600 W (Karabegović et al., 2013)

37 Characterization of the plant material CHAPTER 2

b Potentilla thuringiaca Bernh. Herb 33.8 MeOH:H2O 80:20 40 45 -- (Tomczyk et al., 2012) Vitis vinifera L. Grape skin <1 CC:OA:H2O 75:25 65 50 35 kHz (Bubalo et al., 2016) b Drymocallis rupestris (L.) Soják Herb 26.6 MeOH:H2O 80:20 40 45 -- (Tomczyk et al., 2012) Myrica rubra Sieb. et Zucc. Bayberry cultivars 2.13-8.76 EtOH:CH2O2:H2O 95:1:4 RT 30 60Hz,30W (Zhang et al., 2015) Dasiphora fruticosa L. Herb 36.7 MeOH:H2O 80:20 50 30 -- (Tomczyk et al., 2012) UAE Vaccinium corymbosum L. Bluecrop, jersey & earliblue 0.10-0.22 MeOH:CH2O2 97:3 4-100 60 -- (Zorenc et al., 2017) Vitis vinifera L. Grape pomace 0.475-0.609 EtOH:H2O 50:50 20 60 300W, 25Hz (Drosou et al., 2015) Allium cepa L. Red, yellow & chart. Onion 0.00-11.3 MeOH:CH2O2:H2O 50:5:45 RT 35 (Kwak et al., 2016) Potentilla nepalensis Hook Leaves 27.1 MeOH:H2O 80:20 50 30 -- (Tomczyk et al., 2012) Allium cepa L. Onion dry skin 0.09 MeOH:H2O 80:20 1 -- (Wiczkowski et al., 2014) Rosa canina L. Hips 0.0078-0.0133 MeOH:CH2O2 97:3 -- 60 -- (Cunja et al., 2015)

c (Shanmugam et al., Passiflora subpeltata Ortega Leaves 38.70 C3H6O 100 ------Soxhlet 2016)

(Fernández-Ponce et al., Mangifera indica L. Leaves 0.4-2.3 CO2:EtOH 80:20 55 180 10 MPa 2012) (Fernández-Ponce et al., Pressurized Mangifera indica L. Leaves 17.9-29.1 b EtOH 100 60-100 -- 4-20 MPa systems (SFE, 2015) b (Fernández-Ponce et al., Mangifera indica L. Leaves 13.1-20.1 CO2:EtOH:H2O 50:25:25 60-100 -- 12-20 MPa PLE, ESE, SWE, 2015) SAE) (Fernández-Ponce et al., Mangifera indica L. Leaves 1.3-4.2 H2O 100 100 180 4 MPa 2012) Mangifera indica L. By-products 5.95 C3H6O:H2O 80:20 40 -- 10 MPa (Meneses et al., 2015)

NOTATIONS: Supercritical Antisolvent Extraction (SAE); Enhanced Solvent Extraction (ESE); Pressurized Liquid Extraction (PLE); Supercritical Fluid Extraction (SFE); Subcritical Water Extraction (SWE); Heat Reflux Extraction (HRE); Microwave Assisted Extraction (MAE); Ultrasound Assisted Extraction (UAE); Maceration Extraction (ME); Methanol (MeOH); Ethyl acetate (EtOAC); Water (H2O); Choline Chloride: Oxalic Acid (1:1) a b c (CC:OA); Formic Acid (CH2O2); Ethanol (EtOH); Acetone (C3H6O); Room Temperature (RT); fresh weight ( ); dry extract ( ); mg/mL ( ).

38 Characterization of the plant material CHAPTER 2

2.1.2. Materials and methods

2.1.2.1. Standards and reagents HPLC-grade acetonitrile was from Fisher Scientific (Lisbon, Portugal). The phenolic compound standards (5-O-caffeoylquinic acid and quercetin 3-O-glucoside) were from Sigma-Aldrich (St. Louis, MO, USA). All the other chemicals were of analytical grade and purchased from common sources. Water was treated in a Milli-Q water purification system (Millipore, model A10, Billerica, MA, USA).

2.1.2.2. Plant material Juglans regia L. (walnut) dried leaves were purchased from Soria Natural, S.A., Spain. According to the distributor, the leaves were collected in Soria (Spain) in June 2014 and naturally dried in a room with controlled humidity. The samples were reduced to a fine powder (60 to 20 mesh) and stored in a desiccator protected from light for subsequent assays.

2.1.2.3. Extraction systems for walnut leaves

Maceration extraction Solid-liquid maceration extractions (ME) of walnut leaves were carried out using a commercial Carousel (Carousel 12 Plus Reaction Station™, Radleys Tech) able to both stir and maintain the temperature within ±0.5 °C, protected from light. The equipment is coupled to a refrigeration system, avoiding the loss of solvent. The powdered samples were extracted using different time (t, X1), temperature (T, X2) and ethanol proportion (S, X3) conditions that ranged as defined by the RSM design (Table A1, Appendix A) maintaining the solid/liquid ratio (S/L, X4) at 30 g/L. The S/L condition was further studied (5 to 140 g/L) after the optimization of X1, X2 and X3 for each of the responses assessed. Samples were stirred at 600 rpm and the condensation/refrigeration system was kept at 15 °C. The solvent (ethanol, 0 to 100 %, v/v) volume was fixed at 10 mL. After extraction, the mixture was centrifuged at 6000 rpm for 10 min at room temperature, the pellet was discarded and

39 Characterization of the plant material CHAPTER 2 the supernatant was carefully collected for further analysis. The dry weight (dw) obtained from each solution was evaluated to determine the extraction yield (g extract/g sample).

Microwave-assisted extraction The Microwave-assisted extraction (MAE) process was performed using a commercial microwave (Monowave 300, Anton Paar GmbH) in 30 mL closed vessels of high-precision glass. The solvent (ethanol 0 to 100 %, v/v) volume was fixed at 10 mL. The powdered samples were extracted using the different conditions (t, T and S), as described by the RSM design (Table A1, Appendix A) maintaining S/L at 30 g/L. After, S/L was studied between 5 and 140 g/L for all the responses assessed. During processing, samples were stirred at 600 rpm using a magnetic stirring bar and irradiated at 850 W. In previous studies, other authors (Pinela et al., 2016) have shown that the microwave power has nearly no effect on the extraction process. Afterwards, the mixture in the extraction vessel was quickly cooled in the processing chamber until 55 °C. The mixture was centrifuged at 6000 rpm for 10 min at room temperature, the pellet was discarded and the supernatant was carefully collected for further analysis. The dry weight (dw) obtained from each solution was evaluated to determine the extraction yield (g extract/g sample).

2.1.2.4. Quantification of the main compounds

Chromatographic analysis of the main phenolic compounds After the ME and MAE processes, the extract solutions were filtered through 0.22 µm disposable LC filter disks. The samples were analysed using a Shimadzu 20A series UFLC (Shimadzu Corporation, Kyoto, Japan) with a quaternary pump and a photodiode array detector (PDA) coupled to an LC solution software data-processing station. Separation was achieved using a Waters Spherisorb S3 ODS-2 C18, (3 μm, 4.6 mm × 150 mm) column operating at 35 °C. The mobile phase was a mixture of formic acid in water 0.1 % (A) and acetonitrile (B), and the established elution gradient was as follows: 15 % B for 5 min, 15 % B to 20 % B over 5 min, 20-25 % B over 10 min, 25-35 % B over 10 min, 35-50 % B for 10 min, and column re-equilibration (15 min), using a flow rate of 0.5 mL/min. Double online detection was carried with a diode array detector (DAD) operating at 280 and 370 nm as

40 Characterization of the plant material CHAPTER 2 preferred wavelengths and the target phenolic compounds were identified according to their UV spectra and retention time (Barros et al., 2013). For the quantitative analysis, a baseline to valley integration with baseline projection mode was used to calculate peak areas and external standards were used for quantification. The results were expressed in mg per g of dry weight.

Determination of total phenolics and flavonols by UV-Vis spectroscopy The extract solutions obtained by the ME and MAE systems were filtered through Whatman n°4 filters. The extracts were diluted to obtain solutions with 79.8 % v/v of ethanol. The extract solutions with different concentrations (0.25 mL) were mixed with ethanol:HCl (0.25 mL, HCl 0.1 % v/v in the ethanol 95 % solution) and HCl (4.55 mL, 2 % v/v). The tubes were vortex mixed for 15 s and allowed to stand for 15 min at room temperature. The absorbance was then measured at 360 and 280 nm (total flavonols and total phenolics, respectively) in a Pharmaspec Shimadzu UV-1700 spectrophotometer. Quercetin 3-O-glucoside and chlorogenic acid were used to obtain the standard curves and results were expressed as mg of quercetin 3-O-glucoside or chlorogenic acid equivalents (QE or CAE, respectively) per g of dry plant.

2.1.2.5. Experimental design, model analysis and statistical evaluation

Experimental design The study of the impact of all independent variables was carried using one-factor-at-a-time, to pick the most influents, and to determine the initial range of the processing variables.

Through the analysis of this experimental information (data not shown), X1 (t, min), X2 (T, in °C) and X3 (S, in %) were chosen as variables for the RSM design. Therefore, the combined effect of these three variables on the production of the three main phenolic compounds present in J. regia (maximizing responses individually or globally) was studied using central composite design with 5 levels of each factor. The experimental design is based on 20 independent combinations 6 of which are replicas at the central point of the experiment. The points were built around the centre point as a sphere. According to Box & Hunter (1957), the centre point is presumed to be close to the optimum position for the response,

41 Characterization of the plant material CHAPTER 2 thus it is repeated to maximize the prediction. To minimize the unpredictable effects in the observed responses, experimental runs were random. The mathematical expressions used to calculate the design distribution, code and decode the tested variables can be found all detailed in the supplemental section (Table A1, Appendix A). Once the optimal conditions

(X1, X2 and X3) were optimized, the study was advanced furthermore with the study of the

S/L condition (X4, in g/L).

Mathematical model The response surface models were fitted by means of least-squares calculation using the following second-order polynomial equation:

nnnn −1 Ybb=+++ XbXXbX 2 0 iiijijiii [2.1] iiji====1121 ji

In this equation, Y represents the dependent variable (response variable) to be modelled, the independent variables are Xi and Xj, b0 is the constant coefficient, bi the coefficient of linear effect, bij the coefficient of interaction effect, bii the coefficients of quadratic effect and n is the number of variables. Chromatographic and spectrophotometric analysis in the dependent variable responses were used to quantify the main phenolic compounds.

Procedure to optimize the variables to a maximum response A simplex method was used to maximize the model process by solving non-linear problems to optimize the quercetin 3-O-glucoside, quercetin O-pentoside, 3-O-caffeoylquinic acid, yield, total phenolics and flavonols extraction (Heleno et al., 2016; Pinela et al., 2016). To avoid unnatural conditions of the variables certain limitations were imposed (i.e., times lower than 0).

2.1.2.6. Fitting procedures and statistical analysis The experimental results statistical analysis and fitting was built according to the equations for the responses obtained using a Microsoft Excel spreadsheet in three phases:

1) Coefficients measurement was achieved using the nonlinear least-square (quasi- Newton) method provided by the macro Solver in Microsoft Excel (Kemmer and Keller,

42 Characterization of the plant material CHAPTER 2

2010), by minimization of the sum of quadratic differences between observed and model- predicted values.

2) Coefficients significance was obtained via ‘SolverAid’ (Levie, 2008) to determine the parametric confidence intervals. The terms that were not statistically significant (p-value > 0.05) were dropped to simplify the model.

3) Model reliability was confirmed by applying the following standards: a) the Fisher F-test (α = 0.05) was used to determine the consistency of the constructed models to describe the obtained data (Shi and Tsai, 2002); b) the ‘SolverStat’ macro was used to make assessment of parameter and model prediction uncertainties (Comuzzi et al., 2003); c) R² was determined to explain the proportion variability of the dependent variable obtained by the model.

2.1.3. Results and discussion

2.1.3.1. Response criteria for the RSM analysis and statistical verification The phenolic profile of J. regia leaves extracts is represented in Figure A2 (Appendix A). The main compounds present in the matrix were one phenolic acid, 3-O-caffeoylquinic acid (P1) and two flavonols, quercetin 3-O-glucoside (P2) and quercetin O-pentoside (P3). This profile agrees with the previous one reported by Santos and co-workers (Santos et al., 2013). These compounds were identified by comparison of their UV spectra and retention times with those obtained in previous findings, as well as with the commercial standards. Both quantification results (HPLC-PDA and UV-Vis) are represented in Table 2.2 for the different runs of the RSM design.

43 Characterization of the plant material CHAPTER 2

Table 2.2. Results of the response surface experimental plan for the ME and MAE optimization process of independent variables of time (t), temperature (T), and solvent proportion (S). Response criteria comprise the following: 1) % yield of extraction; 2) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3); and 3) spectrophotometric quantification of total phenolics (PHE) and flavonols (FLAV).

Experimental design Maceration responses Microwave responses

Maceration Microwave Coded values Residue HPLC content Others Residue HPLC content Others conditions conditions

X1: t X2: T X3: S X1: t X2: T X3: S Yield P1 P2 P3 TOTAL PHE FLAV Yield P1 P2 P3 TOTAL PHE FLAV X1 X2 X3 mg/g mg/g mg/g mg/g mg/g mg/g mg/g mg/g mg/g mg/g mg/g mg/g min °C % min w % % % dw dw dw dw dw dw dw dw dw dw dw dw

1 -1 -1 -1 54.3 42.2 20.3 11.5 85 20.3 22.53 5.0 6.7 7.2 19.0 72.3 32.6 24.86 5.9 8.5 9.0 23.5 60.2 26.4 2 1 -1 -1 125.7 42.2 20.3 36.5 85 20.3 23.66 5.6 6.8 7.2 19.7 75.0 33.6 25.65 6.3 7.6 7.1 21.0 66.8 29.4 3 -1 1 -1 54.3 77.9 20.3 11.5 155 20.3 25.29 5.8 7.7 8.1 21.6 90.6 40.0 38.60 6.0 8.2 8.5 22.6 125.4 50.4 4 1 1 -1 125.7 77.9 20.3 36.5 155 20.3 25.45 5.3 6.6 6.5 18.5 90.4 39.4 37.90 5.9 7.1 6.2 19.3 126.9 48.2 5 -1 -1 1 54.3 42.2 79.7 11.5 85 79.7 19.13 1.3 8.1 9.5 18.9 86.9 43.2 24.60 2.4 14.0 13.0 29.4 105.4 50.4 6 1 -1 1 125.7 42.2 79.7 36.5 85 79.7 20.84 1.3 7.6 9.3 18.2 95.7 47.4 25.41 2.5 14.3 13.4 30.2 114.6 54.0 7 -1 1 1 54.3 77.9 79.7 11.5 155 79.7 23.67 0.0 8.2 10.4 18.6 121.4 58.1 30.77 3.2 14.7 13.6 31.5 167.7 70.3 8 1 1 1 125.7 77.9 79.7 36.5 155 79.7 24.18 1.3 8.3 10.0 19.5 96.6 44.4 32.46 3.3 13.8 12.7 29.8 179.8 74.2 9 -1.68 0 0 30 60 50 3 120 50 25.85 2.1 12.6 11.9 26.7 98.9 44.3 28.80 3.8 14.3 12.8 30.9 121.5 51.2 10 1.68 0 0 150 60 50 45 120 50 26.77 2.2 13.6 12.6 28.5 110.3 47.5 31.99 3.6 14.1 12.5 30.2 128.7 54.4 11 0 -1.68 0 90 30 50 24 60 50 23.77 1.9 12.7 11.7 26.4 77.1 36.2 26.26 3.7 14.4 12.4 30.5 100.3 45.7 12 0 1.68 0 90 90 50 24 180 50 29.15 2.4 13.6 12.8 28.7 129.3 54.8 44.26 4.4 9.4 5.4 19.2 194.9 65.8 13 0 0 -1.68 90 60 0 24 120 0 19.71 0.6 5.2 4.8 10.6 58.6 23.8 27.32 6.4 6.4 6.1 18.9 74.4 30.1 14 0 0 1.68 90 60 79.7 24 120 100 11.11 0.3 3.9 3.7 8.0 49.4 25.6 17.65 1.2 11.8 13.3 26.3 65.3 34.3 15 0 0 0 90 60 50 24 120 50 25.98 2.4 13.7 13.1 29.2 113.2 50.2 30.28 3.6 14.1 12.9 30.7 129.4 53.6 16 0 0 0 90 60 50 24 120 50 25.65 2.2 13.5 12.7 28.4 111.8 48.9 31.47 3.5 14.1 12.1 29.6 119.0 51.0 17 0 0 0 90 60 50 24 120 50 26.25 2.1 12.9 12.1 27.1 106.6 46.4 34.41 3.5 14.2 13.1 30.9 124.2 52.3 18 0 0 0 90 60 50 24 120 50 25.98 2.3 13.4 12.9 28.6 104.0 45.2 32.81 3.5 14.4 13.5 31.4 124.0 53.4 19 0 0 0 90 60 50 24 120 50 25.58 2.2 13.1 12.6 27.9 107.2 46.4 30.76 3.5 14.1 13.2 30.8 123.7 52.9 20 0 0 0 90 60 50 24 120 50 26.57 2.4 13.5 13.0 29.0 108.9 47.5 31.29 3.4 14.4 13.4 31.3 127.2 54.2

44 Characterization of the plant material CHAPTER 2

Regarding HPLC-PDA results, the levels of the two flavonols ranged from 3.9 to 13.7 mg/g dw for P2 with the ME process (runs 14 and 15) and 6.4 to 14.7 mg/g dw using the MAE system (runs 13 and 7) and for P3 between 3.7 to 13.1 mg/g dw by employing the ME (runs 14 and 15) and 5.4 to 13.6 mg/g dw through the MAE (runs 12 and 7) techniques. The amounts of 3-O-caffeoylquinic acid (P1) ranged from 0.0 (not detected, under the LOQ) to 5.8 mg/g dw using the ME (runs 7 and 3) system and 1.2 to 6.3 mg/g dw applying the MAE process (runs 14 and 2).

The amounts of total phenolics and flavonols were also determined using UV-Vis measurements. Observing the total phenolic content, the levels ranged from 49.4 to 129.3 mg CAE/g dw (runs 14 and 12) and 23.8 to 58.1 mg QE/g dw (runs 13 and 7) using the ME system and 60.2 to 194.9 mg CAE/g dw (runs 1 and 12) and 26.4 to 74.2 mg QE/g dw (runs 1 and 8) applying the MAE process.

Concerning the extraction yields, both techniques are consistent: the highest yield was obtained in run 12 (29.15 % using ME and 44.26 % with MAE) and the lowest in run 14 (11.11 % using ME system and 17.65 % with MAE).

Overall, applying the RSM conditions (Table 2.2) to the MAE technique not only lead to higher amounts of P1, P2, P3, total phenolics and flavonols but also higher extraction yields. Accordingly, the sum of the three main compounds (P1+P2+P3) using the HPLC-PDA measurements was also superior for the MAE process (run 7: 31.5 mg/g dw) than the ME system (run 15: 29.2 mg/g dw). The quantification results obtained from the chromatographic and spectrophotometric methodologies were used as response criteria to optimize the ME and MAE conditions by RSM. The obtained parametric fitting values, confidence intervals and statistical information are presented in Table 2.3. All coefficients showed significant parametric intervals at the 95 % confidence level (α = 0.05) and the correlation coefficients were always higher than 0.87.

45 Characterization of the plant material CHAPTER 2

Table 2.3. Estimated coefficient values obtained from the Box polynomial model, parametric intervals and numerical statistical criteria for each parametric response criteria of the extractions systems tested (ME and MAE). Response criteria comprise the following: 1) % yield of extraction; 2) HPLC quantification of 3- O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3); and 3) spectrophotometric quantification of total phenolics (PHE) and flavonols (FLAV).

Residue HPLC Content Others

Parameters

Yield P1 P2 P3 Total PHE FLAV

Maceration extraction

Intercept b0 26.23 ±0.38 2.25 ±0.16 13.26 ±0.73 12.68 ±0.47 28.28 ±0.83 109.2 ±3.64 48.16 ±1.75 b 0.55 ±0.29 ns 0.17 ±0.12 0.12 ±0.02 0.30 ±0.11 6.29 ±1.94 2.17 ±1.07 Linear 1 b 1.50 ±0.38 ns 0.20 ±0.12 0.24 ±0.02 0.21 ±0.11 11.49 ±3.01 3.60 ±1.39 effect 2 b3 -1.95 ±0.36 -2.43 ±0.12 0.28 ±0.12 0.54 ±0.02 ns 4.15 ±3.01 2.98 ±1.31 Quadratic b11 ns ns -0.15 ±0.12 -0.10 ±0.02 -0.19 ±0.11 -3.50 ±1.36 -1.37 ±0.72 effect b22 ns 0.17 ±0.11 -0.39 ±0.12 -0.23 ±0.02 -0.25 ±0.11 ns ns b33 -3.66 ±0.31 0.91 ±0.10 -3.39 ±0.12 -3.18 ±0.02 -7.15 ±0.11 -17.88 ±2.90 -7.03 ±1.17 b ns -0.28 ±0.16 -0.20 ±0.12 -0.32 ±0.02 -0.22 ±0.11 -4.55 ±3.93 -1.85 ±2.09 Interactive 12 b ns ns ns ns 0.37 ±0.11 -2.30 ±2.20 ns effect 13 b23 ns 0.20 ±0.17 ns 0.052 ±0.02 ns ns ns

Statistics (R²) 0.9240 0.9662 0.8706 0.9211 0.9488 0.9021 0.8965

Microwave assisted extraction

Intercept b0 31.81 ±0.68 3.50 ±0.18 14.24 ±0.60 12.95 ±0.50 30.74 ±1.12 125.3 ±6.07 53.41 ±2.73 b 0.58 ±0.45 ns -0.21 ±0.19 -0.37 ±0.22 -0.57 ±0.54 3.04 ±2.71 1.05 ±0.97 Linear 1 b 5.09 ±0.45 0.18 ±0.12 -0.65 ±0.40 -0.99 ±0.39 -1.47 ±0.74 30.16 ±4.74 8.54 ±2.14 effect 2 b3 -2.20 ±0.45 -1.58 ±0.12 2.52 ±0.40 2.49 ±0.39 3.44 ±0.74 12.67 ±4.74 7.43 ±2.14 Quadratic b11 -0.34 ±0.33 0.19 ±0.11 -0.13 ±0.12 ns ns ns ns effect b22 1.38 ±0.44 0.30 ±0.11 -0.94 ±0.38 -1.41 ±0.38 -2.05 ±0.72 9.05 ±4.59 1.74 ±1.22 b33 -3.14 ±0.44 0.23 ±0.11 -1.92 ±0.38 -1.14 ±0.38 -2.84 ±0.72 -18.45 ±4.59 -6.57 ±2.07 b ns ns ns ns ns ns ns Interactive 12 b ns ns ns ns ns ns ns effect 13 b23 -1.60 ±0.59 0.26 ±0.15 ns ns ns ns ns

Statistics (R²) 0.9791 0.9614 0.9011 0.8903 0.9224 0.9588 0.8874

ns: not significant.

46 Characterization of the plant material CHAPTER 2

2.1.3.2. Theoretical response surface models It is essential to evaluate the precision of the RSM mathematical model on the ideal prediction of variances. Fitting the models to the selected responses is the key to elucidate this prediction; therefore, the models for each individual response were constructed fitting the second-order polynomial model of Eq. [2.1] to the experimental values in Table 2.3 using nonlinear least-squares estimations. Therefore, the resulting models for each assessed extraction technique are the following, for the ME system:

ME 2 [2.2] Yyield =26.2 + 0.55 t + 1.5 T − 2.0 S − 3.7 S

ME 22 [2.3] YSTStTTSP1 =−++−+2.32.40.170.910.280.20

ME 222 [2.4] YtTStTStTP2 =+++−−−−13.30.170.200.280.150.393.40.20

ME 222 [2.5] YtTStTStTTSP3 =+++−−−−+12.70.120.240.540.10.233.180.320.05

ME 222 [2.6] YtTtTStTtSPPP1++ 2 3 =++−−−−+28.3 0.300.210.190.257.10.220.37

ME 22 [2.7] YtTStStTtSPHE =+++−−−−109.2 6.311.54.23.517.94.62.3

ME 22 [2.8] YtTStStTFLAV =+++−−−48.22.23.63.01.47.01.9

And for the MAE process:

MAE 222 [2.9] YtTStTSTSyield =++−−+−−31.8 0.585.12.20.341.43.11.6

MAE 222 [2.10] YTStTSTSP1 =+−++++3.5 0.182.60.190.300.230.26

MAE 2 2 2 [2.11] YP2 =14.2 − 0.21 t − 0.65 T + 2.5 S − 0.13 t − 0.94 T − 1.9 S

MAE 22 [2.12] YP3 =13.0 − 0.37 t − 0.99 T + 2.5 S − 1.4 T − 1.1 S

MAE 22 [2.13] YtTSTSPPP1++ 2 3 =−−+−−30.7 0.57 1.5 3.4 2.12.8

MAE 22 [2.14] YtTSTSPHE =++++−125.3 3.0 30.2 12.7 9.118.5

MAE 22 [2.15] YFLAV =53.4 + 1.1 t + 8.5 T + 7.4 S + 1.7 T − 6.6 S

47 Characterization of the plant material CHAPTER 2

These equations translate the response patterns for individual measurement of phenolic compounds and Table 2.3. shows the complexity of possible sceneries. Not all the parameters of Eq.[2.1] were used for building the model since some coefficients were non- significant.

Regarding the linear effect on the ME process, the effect of t and T were significant in a positive mode in all cases except for 3-O-caffeoylquinic acid (no significant effects were found). The variable S played an important and significant role for the yield and 3-O- caffeoylquinic acid in a negative way, but positive for the quercetin derivatives, total phenolics and flavonols. However, no significant effect was found for the total HPLC content. Observing the quadratic effect, t exerts a negative significant role to quercetin derivatives, total HPLC content, total phenolics and flavonols, T also plays a negative significant effect for the quercetin derivatives and total HPLC content, but positive for the 3-O-caffeoylquinic acid. The variable S also played a positive role for 3-O-caffeoylquinic acid and negative to the remaining ones. On the other hand, the interactive effect just had a significant effect in some cases: the interactions between t & T caused a significant negative effect in all responses except for the yield where no significant interactions were found. The interaction between t & S were only significant for the total HPLC content (positive) and total phenolics (negative) responses. The variables interactions T & S caused significant positive effects on 3-O-caffeoylquinic acid and quercetin O-pentoside.

Regarding the MAE process, the linear effect had the most important and significant role for all the variables with exception of t on the 3-O-caffeoylquinic acid content (no significant effect was found). The effect of t was significant in a positive mode for the yield, total phenolic content and flavonols content and negative for the remaining HPLC measurements. The variable T showed a significant positive effect for yield, 3-O- caffeoylquinic acid, total phenolics and flavonols and a negative effect on quercetins derivatives and total HPLC content. The variable S played an important and significant role for the yield and 3-O-caffeoylquinic acid in a negative mode, however, positive to quercetin derivatives, total HPLC, total phenolics and flavonols. Regarding the quadratic effect, t exerts a negative significant role in yield and quercetin 3-O-glucoside, but positive on 3-O- caffeoylquinic acid response, T plays an important and significant role in all responses

48 Characterization of the plant material CHAPTER 2

(positive: yield, 3-O-caffeoylquinic acid, total phenolics and flavonols; negative: quercetin derivatives and total HPLC content), the variable S also exerts significant influence in all parameters of this study: positive to 3-O-caffeoylquinic acid and negative to the remaining ones. Observing the interactive effect, the interaction between T & S only presented a significant effect on the response of yield and 3-O-caffeoylquinic acid (negative and positive, respectively).

2.1.3.3. Effect of extraction variables on the target responses The patterns of extraction can also be depicted by graphical representation. Figure 2.1 represents a 3D graphical analysis of the results in mg/g dw of 3-O-caffeoylquinic acid and quercetin 3-O-glucoside by HPLC quantification and total phenolics by spectrophotometric determination obtained for both extraction systems. Figure 2.2 represents the results in mg/g dw of total HPLC and yield. They are divided in two parts (columns) for each tested technique. Each column is divided into two subsections (A and B). Subsection A illustrates a 3D response surface plots predicted with their respective second order polynomial equation described by Eq. [2.6], [2.2], [2.13] and [2.9]. In this subsection, a binary representation is shown, in which the excluded variable is positioned at its optimal experimental domain. Subsection B shows the capability to forecast the achieved results and the residual distribution as a function of the individual considered variables.

Observing all the plots in Figure 2.1 it is possible to verify that the amount of extracted material increases to an optimum value and, in most cases, decreases as a function of each assessed independent variable. Similar behaviour can be found in Figure 2.2 and Figure A3 (Appendix A). Consequently, in almost all combinations the optimum value in single points can be found, which allows computing the conditions that lead to an absolute maximum. Figure 2.3 simplifies the interpretation of the influence of the independent variables in a bi-dimensional representation of the optimal conditions of the RSM variables.

Part A of Figure 2.3 shows a two-dimension representation of the influence of the independent variables relative to each extraction method. The operating conditions that maximize the extraction conditions are represented with dots (), these singular conditions are represented in Table 2.4 to clarify not only the optimal independent

49 Characterization of the plant material CHAPTER 2 variables conditions but also the optimal values predicted. The experimental results are also presented in this table.

50 Characterization of the plant material CHAPTER 2

Figure 2.1. Response surface of each of the variables involved for ME and MAE systems for the HPLC quantification of 3-O-caffeoylquinic acid content (P1) and quercetin 3-O-glucoside content (P2) and for the spectrophotometric determination of the total phenolics (PHE). The other response determinations of quercetin O-pentoside content (P3) and total flavonols (FLAV) were excluded and presented in Figure A3 (Appendix A) due to their similarity with quercetin 3-O-glucoside content (P2) and total phenolics (PHE), respectively. Each of the net surfaces represents the theoretical three-dimensional response surface predicted with the second order polynomial of Eqs. [2.3], [2.4] and [2.7] in ME system and [2.10], [2.11] and [2.14] for MAE process. The statistical design and results are described in Table 2.2. Estimated parametric values are shown in Table 2.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 2.4).

51 Characterization of the plant material CHAPTER 2

Table 2.4. Operating optimal conditions of the variables involved that maximize the response values for RSM using a CCCD for each of the extracting techniques assessed (ME and MAE), for the three individual response value formats of each compound and for all the compounds. Response criteria comprise the following: 1) % yield of extraction; 2) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3); and 3) spectrophotometric quantification of total phenolics (PHE) and flavonols (FLAV). The optimum values underlined are the results found experimentally, meanwhile the others are those found by the mathematical model.

Optimal variable conditions Optimum response Criteria X1: t (min) X2: T (°C) X3: S (%) (mg/g dw)

Individual optimal variable conditions for 3-O-caffeoylquinic acid:

Maceration 150.0 29.9 0.0 10.7±1.45 11.0 Microwave 3.0 61.1 0.0 8.6±0.98 6.4

Individual optimal variable conditions for the quercetin 3-O-glucoside:

Maceration 112.5 62.93 50.6 13.2±1.37 13.8 Microwave 14.1 107.81 69.4 15.2±1.11 14.2

Individual optimal variable conditions for the quercetin O-pentoside:

Maceration 112.5 62.9 50.6 12.7±1.23 11.7 Microwave 3.0 107.7 82.5 15.1±0.95 14.0

Global optimal variable conditions for the total HPLC compounds:

Maceration 115.6 61.3 50.4 28.3±2.43 28.8 Microwave 3.0 107.5 67.9 32.9±3.16 29.3

Individual optimal variable conditions for the yield residue extraction:

Maceration 150.0 90.0 42.1 29.9±1.22 29.3 Microwave 34.6 180.0 26.8 46.4±4.15 44.3

Individual optimal variable conditions for the total phenolics:

Maceration 81.1 90.0 53.9 129.1±23.46 124.0 Microwave 45.0 180.0 60.2 209.1±16.21 194.9

Individual optimal variable conditions for the total flavonols:

Maceration 77.7 90.0 56.3 54.7±2.98 53.3 Microwave 45.0 180.0 66.8 66.5±6.56 66.0

52 Characterization of the plant material CHAPTER 2

MACERATION EXTRACTION MICROWAVE ASSISTED EXTRACTION

TOTAL HPLC YIELD TOTAL HPLC YIELD

Figure 2.2. Response surfaces of the total compounds content determined by HPLC (P1+P2+P3) and extraction yield from the developed equations for the ME and MAE systems optimization. Part A: The graphical 3D analysis is presented as a function of each of the variables involved. Each of the net surfaces represents the theoretical three-dimensional response surface predicted with the second order polynomial of Eqs. [2.6], [2.2], [2.13] and [2.9], respectively. The statistical design and results are described in Table 2.2. Estimated parametric values are shown in Table 2.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 2.4). Part B: To illustrate the goodness of fit, two basic graphical statistic criteria are used: the ability to simulate the changes of the response between the predicted and observed data; and the residual distribution as a function of each of the variables. Note all the differences in the axes scales.

53 Characterization of the plant material CHAPTER 2

Figure 2.3. Final summary of the effects of all variables assessed for ME and MAE systems. Part A: Shows the individual 2D responses of all studied responses as a function of all the variables assessed. The variables in each of the 2D graphs were positioned at the optimal values of the others (Table 2.4). The dots () presented alongside each line highlight the location of the optimum value. Lines and dots are generated by the theoretical second order polynomial models of Eqs. [2.2]-[2.7] for ME and [2.9]-[2.15] for MAE. Part B: Shows the dose response of S/L ratio at the optimal values of the other three variables (Table 2.4). Points () represent the obtained experimental results, meanwhile the line shows the predicted pattern by simple linear relation. The limit value (~140 g/L) shows the maximum achievable experimental concentration until the sample cannot be physically stirred at laboratory scale.

54 Characterization of the plant material CHAPTER 2

Regarding the particular case of the ME system (Figure 2.3), the intense curvature of the solvent observed on the total HPLC contents dictates its action: when raising the percentage of ethanol until 50 % the extraction content also increases (~ 28 mg/g dw – represented by ), followed afterwards by a decrease of this content. Higher amounts of 3-O-caffeoylquinic acid are obtained when using 0 % of ethanol (100 % water) and quercetin derivatives with 50 % of solvent. As quercetin derivatives are present in higher amounts comparatively to 3-O-caffeoylquinic acid, the optimal for total HPLC content is obtained in similar conditions of quercetin derivatives. Therefore, as can be seen in Table 2.4, the maximum global optimal variable conditions for the total compounds reached by the ME system were 115.6 min (X1), 61.3 °C (X2) and 50.4 % of ethanol (X3), predicting 28.3 mg/g dw (28.8 mg/g dw experimental). Since the total HPLC is the result of the sum of the obtained responses regarding the 3-O-caffeoylquinic acid (10.7 mg/g dw predicted, 11.0 mg/g dw experimental), quercetin 3-O-glucoside (13.2 mg/g predicted, 13.8 mg/g dw experimental) and quercetin O-pentoside (12.7 mg/g dw predicted, 11.7 mg/g dw experimental) amounts, the optimal variable conditions should be around the ones which are in higher quantities individually. Due to structural similarities, quercetin derivatives present close to optimal conditions of extraction with this technique (quercetin 3-O- glucoside: 112.5 min, 62.93 °C and 50 % of ethanol; quercetin O-pentoside: 112.5 min, 62.9 °C and 50.6 % of ethanol), thus they exert a strong influence on the final conditions to the optimal variable conditions on the total HPLC content. However, using the MAE process, factors beyond structural similarities are essential for the explanation of the obtained results. In this technique, the energy provided by microwaves is applied facilitating the partition of molecules present in the matrix into the solvent (Dai and Mumper, 2010). Consequently, the behaviour of the independent variables is not as obvious as those observed in the ME system and unexpected results were obtained when applying the MAE process. In fact, for quercetin derivatives extraction proximal conditions of T and S were observed (quercetin 3-O-glucoside: 107.81 °C and 69.4 % of ethanol; quercetin O- pentoside: 107.7 °C and 82.5 % of ethanol) as equal optimum responses (quercetin 3-O- glucoside: 15.2 and 14.2 mg/g dw, predicted and experimental, respectively; quercetin O- pentoside: 15.1 and 14.0 mg/g dw, predicted and experimental, respectively).

55 Characterization of the plant material CHAPTER 2

Nevertheless, a discrepancy was found in the time of extraction (quercetin 3-O-glucoside: 14.1 min; quercetin O-pentoside: 3.0 min). At first, it can be seen that times are shorter for the MAE than the ME techniques (as expected), and second, the kinetics of extraction of each compound (although they are structurally quite similar) are singular: quercetin O- pentoside shows a faster optimum value than quercetin 3-O-glucoside. As observed on the MAE process, optimal response on 3-O-caffeoylquinic acid was obtained with 0 % of ethanol, and increasing the temperature up to 61.1 °C (ME: 29.9 °C) decreases the time of extraction until 3.0 min, compared to 150.0 min for the ME system optimal conditions (predicted: 10.7 mg/g dw; experimental: 11.0 mg/g dw). Thus, comparing the predicted and experimental values obtained through HPLC maximized conditions (individual and sum of main compounds), it can be concluded that polynomial equations are well adjusted. Concerning the individual optimal variable conditions for the UV-Vis determinations, the maximum values for each extraction technique were achieved in similar conditions. In fact, for the ME system, the total phenolics (81.1 min, 90.0 °C, 53.9 % of ethanol) and total flavonols (77.7 min, 90.0 °C, 56.3 % of ethanol) conditions are close. The same was observed in the MAE process for the total phenolics (45.0 min, 180 °C, 60.2 % of ethanol) and total flavonols (45.0 min, 180.0 °C, 66.8 % of ethanol). The predicted and experimental values obtained by UV-Vis measurements were also satisfactory. Regarding the total phenolics, the predicted response using the ME system was 129.1 mg CAE/g dw, but the experimental result was 124.0 mg CAE/g dw, and for the MAE process the predicted value was 209.1 mg CAE/g dw, and the experimental 194.9 mg CAE/g dw. The same findings were observed in total flavonols predicted and experimental results, applying the ME technique (54.7 and 53.3 mg QE/ g dw, respectively) and the MAE process (66.5 and 66.0 mg QE/g dw, respectively). The particular optimal variable conditions in yield residue extraction were different for the ME and MAE systems. The optimal response for the ME system (29.94 % predicted and 29.3 % experimental) were obtained at 150.0 min, 90.0 °C and 42.1 % of ethanol. In contrast, the optimum values obtained for the MAE process (46.4 and 44.3 %, predicted and experimental) were reached at 34.6 min, 180.0 °C and 60.2 % of ethanol. Thus, the MAE process leads to higher extraction yields in a shorter time. This profile of extraction was also verified for the remaining measured parameters. In fact, in all cases,

56 Characterization of the plant material CHAPTER 2 the MAE process leads to higher extraction yields (mg/g dw, mg CAE/g dw or mg QE/g dw) with shorter extraction times.

Due to the accurate HPLC results, the studies on S/L were performed in optimal conditions predicted by the polynomial models obtained for each extraction technique using the total HPLC content as the response factor (Table 2.4). Preliminary results (data not shown) indicate that the experimental limit value of concentration was proximal to 140 g/L due to the conditioned experimental stirring for higher S/L values. Therefore, both processes of extraction were designed to verify the S/L behaviour between 5 g/L until 140 g/L. The dose response of S/L ratio is presented on Part B of Figure 2.3. The responses obtained through the ME and the MAE systems are consistent: the S/L ratio can be described by a simple linear relationship. All experimental points are distributed around the equation with only one independent variable, consequently the dose response is explained by the slope (m) of the linear relation. The S/L increase leads to a decrease in the extraction ability of the solvent; consequently, the extraction responses reach a maximum value at 5 g/L and a minimum at 140 g/L and the respective losses are justified by the m value. However, the decreases observed are slight (less than -0.4). The m values in 3-O-caffeoylquinic acid for the ME and MAE techniques were -0.0429 and -0.0169, meaning that for the ME system, the increase of 1 g/L on S/L implies the loss of 0.0429 mg/g dw and for the MAE process 0.0169 mg/g dw. Although the ME system leads to higher losses of 3-O-caffeoylquinic acid with the optimal proposed conditions, for the final S/L, it remains better than the MAE process. In fact, the ME system is more efficient to extract this phenolic acid (ME: ~ 11 to 5 mg/g dw; MAE: ~ 6 to 4 mg/g dw). However, the same tendency is not verified in the remaining responses. In fact, the parameter m assumes higher absolute values in the MAE process compared to those found in the ME system (quercetin 3-O-glucoside: -0.0291 and -0.0143; quercetin O-pentoside: -0.0163 and -0.0128; total HPLC: -0.0541 and -0.0146; yield: -0.0832 and -0.0597; total phenolics: -0.3630 and -0.2036; total flavonols: -0.0983 and -0.0545). The higher loss (expressed on mg/g) induced by the increase of S/L was verified for the total phenolics amounts obtained by the ME system (m = -0.3630). However, the amounts reached at 140 g/L using the MAE process (~ 140 mg/g dw) were higher than the optimal for the ME system (5 g/L: ~ 125 mg/g dw). In contrast, the lower

57 Characterization of the plant material CHAPTER 2 loss of mg/g made by the increase of S/L was shown on the quercetin O-pentoside by the ME system (m = -0.0128). Nevertheless, a similar behavior was found for the MAE process (m = -0.0163) and the extraction yield (mg/g dw) varies in the same magnitude for both extraction techniques (~ 14 to 10 mg/g dw).

2.1.4. Conclusions The combined effects of three independent variables were studied to maximize the intrinsic responses on the J. regia chemical composition. A 5-level full factorial Box- Behnken design of 16 combinations and 6 replicates at the centre of the experimental domain was successfully applied for the optimization of the ME and the MAE techniques by RSM. Polynomial responses were successfully designed and experimentally verified on 3-O-caffeoylquinic acid (P1), quercetin 3-O-glucoside (P2), quercetin O-pentoside (P3), total HPLC (sum of the most abundant compounds), total phenolics (PHE) and total flavonols (FLAV) extraction and yield maximizations. Under the global HPLC optimum conditions for the ME and the MAE systems (t = 115.6 min, T = 61.3 °C, S = 50.4 % of ethanol and t = 3.0 min, T = 107.5 °C, S = 67.9 % of ethanol, respectively), the values for P1, P2, P3, P1+P2+P3, PHE and FLAV were 10.7 and 8.6 mg/g dw, 13.2 and 15.2 mg/g dw, 12.7 and 15.1 mg/g dw, 28.3 and 32.9 mg/g dw, 29.9 and 46.4 %, 129.1 and 209.1 mg/g dw, 54.7 and 66.5 mg/g dw. For the S/L at the optimum conditions, the responses for all criteria assessed followed a decreasing linear relation until 140 g/L. Statistical validation was made by the high values of the adjusted coefficient of determination. This study contributed to the valorization of walnut leaves as a source of valuable compounds. Moreover, the use of the mathematical models allowed an optimization of the target responses aimed at valorising the obtained extracts with potential application as ingredients in functional foods.

58 Characterization of the plant material CHAPTER 2

2.2. Phenolic profile, antioxidant and antibacterial properties of Juglans regia L. (walnut) leaves from the Northeast of Portugal Vieira, V., Pereira, C., Pires, T.C.S.P., Calhelha, R.C., Alves, M.J., Ferreira, O., Barros, L., Ferreira, I.C.F.R., 2019. Ind. Crops. Prod. 134, 347-355.2

Abstract

Juglans regia L. (walnut tree) is a recognized source of bioactive compounds with potential health benefits. In this work, hydroethanolic extracts of J. regia leaves were obtained by heat assisted extraction from different Portuguese samples in two phenological stages (green and yellow leaves) aiming to assess the impact of seasonal variations. The samples were compared regarding their phenolic composition and bioactivity. Seventeen phenolic compounds were identified by liquid chromatography combined with a diode array detector and electrospray ionization mass spectrometer (LC-DAD-ESI/MSn): six phenolic acids, ten flavonoids and one tetralone derivative. The green leaves extracts presented a higher amount of total phenolic compounds (29.70 ± 0.03 mg/g extract) compared with the yellow leaves (23.26 ± 0.06 mg/g extract). In particular, yellow samples were richer in flavonoids (17.4 ± 0.2 mg/g extract; mainly quercetin-3-O-glucoside: 3.64 ± 0.01 mg/g extract), while the green ones presented higher phenolic acids content (16.7 ± 0.2 mg/g extract; mainly trans 3-p-coumaroylquinic acid: 6.9 ± 0.5 mg/g extract). Green leaves extract also presented higher antioxidant potential, achieving IC50 values around 32 ± 2 µg/mL and 26.8 ± 0.2 µg/mL for the oxidative haemolysis inhibition and the thiobarbituric acid reactive substances assays, respectively. Furthermore, only green leaves samples showed anti-inflammatory potential. The cytotoxic evaluations revealed similar anti- proliferative action of both extracts against the tumor cell lines tested. Also, an analogous anti-bacterial potential of the extracts was observed, with preferential action against Gram-positive clinical isolated bacteria, with lower minimum inhibitory concentration

2 My contribution to this work was the measurement of the experimental data, their interpretation and writing of the manuscript.

59 Characterization of the plant material CHAPTER 2

(MIC) values for Enterococcus faecalis and Listeria monocytogenes (MIC = 2.5 mg/mL). Therefore, the present study suggests the use of walnut leaves as a source of active ingredients without hepatotoxic effects to be used in different applications in the food or pharmaceutical areas.

Keywords: Juglans regia L. leaves, Phenolic compounds, Antioxidants, Anti-inflammatory, Antibacterial, Cytotoxicity.

2.2.1. Introduction Medicinal plants have been used for centuries by different civilizations as therapeutic agents due to their preventive and curative properties (Tasneem et al., 2019). Juglans regia L. is spread across the globe, being walnut leaves widely used by the folk medicine as antiseptic, anti-inflammatory, antidiabetic, anti-helminthic and antidiarrheic agents. Furthermore, they are also traditionally used for the treatment of hyperhidrosis, ulcers and dandruff (Carvalho and Morales, 2010; Cosmulescu et al., 2014b). Moreover, J. regia leaves extract is an inventoried ingredient for cosmetic products, with reported astringent, soothing, cleansing, abrasive, bulking, skin conditioning, and masking functions (European Commission, 2006).

Many studies reported the bioactivity of walnut leaves by several in vitro (cell cultures) and in vivo (rats/mice) assays regarding their antidiabetic (Asgary et al., 2008; Hosseini et al., 2014; Javidanpour et al., 2012; Mohammadi et al., 2012; Mollica et al., 2017; Pitschmann et al., 2014; Teimori et al., 2009), anti-inflammatory (Erdemoglu et al., 2003; Hosseinzadeh et al., 2011), anti-proliferative (Salimi et al., 2014; Santos et al., 2013), anti- septic/biofilm/bacterial (Dolatabadi et al., 2018; Pereira et al., 2007), hepatoprotective (Eidi et al., 2013), sedative and analgesic (Erdemoglu et al., 2003; Gîrzu et al., 1998; Hosseinzadeh et al., 2011), vasorelaxant (Perusquía et al., 1995), and neuroprotective (Orhan et al., 2011) properties. The study performed by Wang et al. (2015) evaluated the anti-hyperuricemia potential of J. regia leaves by determining the xanthine oxidase inhibitory activity. Moreover, the extensive evidences on the anti-diabetic activity of

60 Characterization of the plant material CHAPTER 2 walnut leaves encouraged Abdoli et al. (2017) to carry out a human trial using walnut leaves-based preparations in Type II diabetic patients, obtaining positive results for the reduction of the glycemic index and HbA1c (glycated hemoglobin) values.

The secondary metabolites of J. regia leaves have been related to its therapeutic effects, such compounds including phenolic acids, flavonoids, organic acids, tocopherols, triterpenic acids, terpenes, terpenoids, tetralone derivatives, megastigmane derivatives, hydroxy-1,4-naphthoquinone (juglone) derivatives, among others (Cosmulescu et al., 2011; Forino et al., 2016; Panth et al., 2016; Rather et al., 2012; Salimi et al., 2014; Santos et al., 2013; Schwindl and Kraus, 2017). Phenolic compounds are the main fraction found in the plant material. They are natural occurring antioxidants, largely recognized because of their health benefits due to their radical scavenging activity, reducing oxidative stress, which is the main cause of several disorders. In fact, these exogenous antioxidant source play an important role for the oxidative stress balance, providing additional antioxidant potential for living organisms, since there is an insufficient action of our endogenous antioxidant system (Carocho and Ferreira, 2013). In this context, the inclusion of walnut leaves preparations (food and supplements) in our diet could act as a preventive medicine with several health benefits (Pires et al., 2018).

The use of J. regia leaves goes back from centuries, and in Portugal, especially in the northeast region (Bragança) they are traditionally used as an external antiseptic, anti- inflammatory and antidiabetic agent (Carvalho and Morales, 2010). In this context, the study of walnut leaves from this region, characterizing their main constituents and potential bioactivities, will contribute to support their traditional medicinal uses with scientific evidences, valorizing this species and its origin. In this work, Portuguese walnut leaves in two phenological stages (green and yellow leaves) were evaluated aiming to assess the impact of seasonal variations on the phenolic compounds composition and the bioactive potential of the extracts obtained. The seasonal variation of the phytochemical profile of this plant material has been previously established by Cosmulescu et al. (2014b) using Romanian samples. Along that study, samples were collected during each fortnight from 1st of June to 15th of August 2012 and were only compared in terms of their phytochemical abundancy (ellagic acid, rutin, myricetin and juglone).

61 Characterization of the plant material CHAPTER 2

In order to achieve the aims proposed in this study, a hydroethanolic extract was evaluated regarding its antioxidant, cytotoxic and antibacterial activities, and the phenolic profile was obtained by LC-DAD-ESI/MSn.

2.2.2. Material and methods

2.2.2.1. Plant material and extract preparation Juglans regia L. (walnut) leaves were collected in Bragança, Northeast of Portugal, in May (green leaves) and November (yellow leaves) of 2018. The samples were dried until constant weight in an incubator at 35 °C. Then, the plant material was ground to approximately 40 mesh, and the homogeneous samples were stored in a desiccator protected from light.

The hydroethanolic extract was obtained by heat assisted extraction, using the previous optimized conditions for the extraction of phenolic compounds reported by the authors (Vieira et al., 2017): aqueous ethanolic solution (50.4 %, v/v; 30 g/mL) at 61.3 °C for 116 minutes. After filtration (Whatman no. 4 filter), the solvent was first evaporated at 40°C, under reduced pressure, in a rotary evaporator (Büchi R-210, Flawil, Switzerland) and the residual solvent was removed in a freeze drier (Telstar Cryodos-80, Terrassa, Barcelona).

2.2.2.2. Phenolic compounds characterization The dry extracts were re-suspended at a concentration of 10 mg/mL using aqueous ethanol (20 %, v/v) and filtered (0.2 µm disposable LC filter disk, 30 mm, nylon). Afterwards, the phenolic profile of walnut leaves was found by liquid chromatography with diode-array detector (280, 330, and 370 nm wavelengths) coupled to an electrospray ionization mass spectrometry operating in negative mode (Dionex Ultimate 3000 UPLC and Linear Ion Trap LTQ XL, Thermo Scientific, San Jose, CA, USA), as previously described by the authors (Bessada et al., 2016). The phenolic compounds were identified according to their chromatographic characteristics by comparison to those obtained with standard compounds and with literature. Calibration curves of appropriate standards were obtained

62 Characterization of the plant material CHAPTER 2 in the range 200-5 µg/mL, for the quantitative analysis. The results were expressed in mg per g of extract (mg/g).

2.2.2.3. In vitro antioxidant assays The lipid peroxidation inhibition in porcine (Sus scrofa) brain homogenates was evaluated through the thiobarbituric acid reactive substances (TBARS) assay. The ethanolic extracts were prepared at a concentration of 10 mg/mL in distilled water and further successive dilutions were performed to obtain the working concentrations ranging from 400 to 6.25 µg/mL. The decrease of the TBARS production was performed by measuring the color intensity of the malondialdehyde-thiobarbituric acid (MDA-TBA) at 532 nm. The methodology followed was previously reported by Barreira et al. (2013) with slight modifications. Briefly, dissected porcine brain was homogenized with Tris-HCl buffer (20 mM, pH 7.4) at a ratio of 1:2 (w/v). The homogenate was centrifuged (3000 rpm for 10 minutes) and an aliquot (100 µL) of the supernatant was incubated with the stock solutions of extract (200 µL), FeSO4 (10 mM; 100 µL) and ascorbic acid (0.1 mM; 100 µL) at 37 °C for 60 minutes. Then, the reaction was stopped by adding trichloroacetic acid (28 %, w/v, 500 µL), thiobarbituric acid (2 %, w/v, 380 µL), and heating the mixture at 80 °C for 20 minutes. After that, the mixtures were centrifuged (3000 rmp) for 10 minutes and the absorbances of the supernatants were measured. The inhibition ratio (%) was calculated using the following formula: [(A−B)/A] × 100 %, being A and B the absorbance of the control and the sample solution, respectively. The EC50 value defines the concentration providing 50 % of antioxidant activity, and was calculated by interpolation from the graph of TBARS formation inhibition percentage against sample concentration.

The anti-haemolytic activity of walnut leaves samples was evaluated by the oxidative haemolysis inhibition assay (OxHLIA). The adopted methodology was previously described by Takebayashi et al. (2012); however, slight modifications were implemented. Erythrocytes were obtained from sheep blood collection. Healthy animals were selected for the purpose and the full blood was centrifuged at 1000 g at 10 °C for 5 minutes. Then, plasma and buffy coats were discarded and erythrocytes were once washed with NaCl (150 nM) and three times with phosphate buffer solution (PBS, pH 7.4) (Evans et al., 2013), being

63 Characterization of the plant material CHAPTER 2 the erythrocyte fraction resuspended in PBS (2.8 %, v/v). The extract was dissolved in PBS (pH 7.4) to obtain a stock solution of 10 mg/mL, and successive solutions were prepared, ranging from 400 to 6.25 µg/mL. About 400 µL of each solution was transferred for a 48 well plate (duplicates). Water and PBS (pH 7.4) were used as controls of haemolysis (plasmolytic and isotonic media, respectively) and Trolox as positive control. Then, the suspension of erythrocytes (200 µL) was mixed with the previous solutions/controls and the 48-well plate was incubated for 10 minutes at 37.5 °C with agitation. After that, a solution of 2,2′-azobis(2-methylpropionamidine) dihydrochloride (AAPH, 160 mM in PBS, 200 µL) was added and the absorbance of the mixtures was measured at 690 nm (BioTek ELx800). The microplate was incubated under the same conditions and absorbances were measured every 10 minutes during approximately 400 minutes. The percentage of the erythrocyte population that remained intact (P) was calculated using the following equation:

P (%) = (St – CH0/S0 – CH0) × 100 [2.16] where St and S0 correspond to the absorbance of the sample at time t and 0 minutes, respectively, and CH0 is the absorbance of the complete haemolysis at 0 minutes. The results were, then, expressed as delayed time of haemolysis (Δt), which was calculated according to the following equation:

Δt (min) = Ht50 (sample) – Ht50 (control) [2.17] where Ht50 is the haemolytic time (minutes) obtained by graphical representation of the haemolysis curve of each sample concentration. Then, the Δt values were correlated to the antioxidant sample concentrations promoting the desirable Δt haemolysis delay

(Takebayashi et al., 2012). The results were expressed as IC50 values (µg/mL) at 60 min, meaning the extract concentration required to keep 50 % of the erythrocytes intact for 60 minutes.

2.2.2.4. Anti-inflammatory activity The walnut leaves extracts were re-dissolved in water at a concentration of 8 mg/mL and then diluted in the range of 400 to 6.25 µg/mL using supplemented DMEM (10 % heat-

64 Characterization of the plant material CHAPTER 2 inactivated fetal bovine serum, 2 mM glutamine, 10 % MEM non-essential amino acids solution, 100 U/mL penicillin and 100 µg/mL streptomycin). A mouse macrophage-like cell line RAW264.7 was used in this study and maintained at 37 °C under 5 % of CO2. Briefly, a suspension of cells (5 × 105 cells/mL) was allowed to sediment on the 96-well plate (1.5 × 105 cells/well) overnight. Then, cells were treated with the different concentrations of extract for 60 min. Thereafter, the cells were stimulated with LPS (1 µmg/mL, in supplemented DMEM) during 18 h. Besides the GRS, the kit contains sulfanilamide, N-(1- naphthyl)ethylenediamine dihydrochloride (NED) and nitrite solutions. The cell culture supernatant (100 µL) was transferred to the plate in duplicate and mixed with sulfanilamide and NED solutions, 5 and 10 minutes each, at room temperature. The nitrite produced was determined by measuring the optical density at 515 nm, using the ELx800 microplate reader (Bio-Tek Instruments, Inc; Winooski, VT, USA) following a methodology previously reported (Correa et al. 2015). The results were expressed in EC50 values (sample concentration providing 50 % of inhibition of NO production) and dexamethasone was used as a positive control, while in negative controls no LPS was added.

2.2.2.5. Cytotoxic assays The extracts were re-dissolved in water at 8 mg/mL concentration and further diluted in the range 400 to 6.25 µg/mL. The four human tumor cell lines assessed (MCF-7, breast adenocarcinoma; NCI-H460, non-small cell lung cancer; HeLa, cervical carcinoma; and HepG2, hepatocellular carcinoma) were maintained as adherent cell cultures in RPMI-1640 medium containing heat-inactivated fetal bovine serum (FBS, 10 %), glutamine (2 mM), MEM non-essential amino acids solution (10 %), and antibiotics (100 U/mL penicillin and

100 mg/mL streptomycin) at 37 °C, in a humidified air incubator with CO2 (5 %). For evaluation of the cytotoxicity in non-tumor cells, a cell culture designated as PLP2, was prepared from a freshly harvested porcine liver obtained from a local slaughterhouse, according to a procedure established by Abreu et al. (2011). The cell proliferation of each cell culture was monitored every two to three days using a phase contrast microscope and, before confluence was reached, the cells were subcultured in the supplemented RPMI- 1640 medium mentioned above. Then, the cells were distributed in 96-well-plates (190 µL) according to their appropriate density (1.0 × 104 cells/well for all cell cultures) and tested

65 Characterization of the plant material CHAPTER 2 with the solutions of extracts (10 µL) during 48 hours. Ellipticine was used as positive control. Hereinafter, the adherent cells were fixed by adding trichloroacetic acid (TCA, 10 %, 100 µL) and incubation at 4 °C for 60 minutes. After the incubation period, the TCA was removed and plates were washed (distilled water, three times) and dried. The sulforhodamine B solution (SRB, 0.1 % in 0.1 % acetic acid, 100 µL) was then added and allowed to bound at room temperature for 30 minutes. The excess of SRB was removed by washing the 96-well plates with acetic acid (1 %, three times) and allowed to dry. The bound SRB was solubilized by adding Tris buffer solution (10 mM, 200 µL) and the absorbances were measured at 540 nm using the microplate reader mentioned above. The results were expressed in GI50 values (concentration that inhibited 50 % of the cell proliferation), being the negative control constituted by each suspension of cells (Barros et al., 2013).

2.2.2.6. Antibacterial activity The antimicrobial potential of the walnut husks was assessed for bacterial strains obtained from patients hospitalized in different departments at the Northeastern local health unit (Bragança, Portugal) and Hospital Center of Trás-os-Montes and Alto Douro (Vila Real, Portugal). Five Gram-negative bacteria (Escherichia coli, isolated from urine; Proteus mirabilis, isolated from wound exudate; Klebsiella pneumoniae, isolated from urine; Pseudomonas aeruginosa, isolated from expectoration; and Morganella morganii, isolated from urine), three Gram-positive bacteria (Enterococcus faecalis, isolated from urine; Listeria monocytogenes, isolated from cerebrospinal fluid; methicillin-resistant Staphylococcus aureus (MRSA), isolated from expectoration) were tested. All these microorganisms were incubated at 37 °C in appropriate fresh medium for 24 hours before analysis to maintain the exponential growth phase. The minimum inhibitory concentration (MIC) determinations on all bacteria were performed using a colorimetric assay described by Pires et al. (2018).

The extracts were dissolved in a mixture of dimethyl sulfoxide (DMSO) + Mueller-Hinton Broth (MHB; E. coli, P. mirabilis, K. pneumoniae, P. aeruginosa, M. morganii and MRSA)/ Tryptic Soy Broth (TSB, E. faecalis and L. monocytogenes) (5 + 95 %, v/v) to give a final concentration of 100 mg/mL for the stock solution and successive dilutions were further

66 Characterization of the plant material CHAPTER 2 carried out ranging from 20 to 1.25 mg/mL. Then, 90 μL of each concentration and 10 μL of each inoculum (standardized at 1.5 × 108 Colony Forming Unit (CFU)/mL) were pipetted (duplicates) for a 96-well microplate. Three negative controls were prepared (1: MHB/ TSB; 2: extract; 3: medium + antibiotic + bacteria), whereas, the positive control was prepared with MHB/TSB and each inoculum. The Ampicillin and Imipenem antibiotics were used as negative control for all Gram-negative bacteria tested and Listeria monocytogenes, while Ampicillin and Vancomycin were selected for Enterococcus faecalis and MRSA. The microplates were covered and incubated at 37 °C for 24 hours. The MIC of samples was detected following addition of p-iodonitrotetrazolium chloride (INT) (0.2 mg/mL, 40 μL) and further incubation at 37 °C for 30 minutes. MIC was defined as the lowest concentration that inhibits the bacterial growth, determined visually by change of the coloration from yellow to pink (viable microorganism) as previously described by Kuete et al. (2011). To determine the minimum bactericidal concentration (MBC), 10 μL of liquid from each well that showed no change in colour was inoculated on Blood agar solid medium (7 % sheep blood) and incubated at 37 °C for 24 hours. The MIC was determined according to the lowest concentration that yielded no growth. Meanwhile, MBC was established as the lowest concentration required to kill bacteria.

2.2.2.7. Statistical analysis Triplicates of green and yellow samples were assayed and three repetitions of each methodology were performed, being the results expressed as mean values and standard deviations (SD). The significant differences between the two samples were established by applying a Student´s t-test, with p = 0.05 (SPSS v. 23.0 program).

2.2.3. Results and discussion

2.2.3.1. Composition in phenolic compounds The phenolic composition of the extracts of J. regia leaves was assessed for samples in two phenological stages (green and yellow samples, collected in May and November, respectively). The identified molecules and their quantification are presented in Table 2.5. In this context, phenolic compounds were tentatively identified according to their retention

67 Characterization of the plant material CHAPTER 2

– time (Rt), maximum absorbance wavelength (λmax), pseudomolecular ion ([M – H] ), and respective fragmentation pattern (MS2). For both samples, it was possible to identify seventeen phenolic compounds: six phenolic acids (p-hydroxycinnamic acid derivatives), ten flavonoids (flavonols) and one tetralone derivative (Figure 2.4).

Regarding the six phenolic acids identified in J. regia, peak 1 ([M – H]– at m/z 341), 6 ([M – H]– at m/z 355), and 7 ([M – H]– at m/z 325) were tentatively identified as caffeic, ferulic, and p-coumaric acid hexosides, based on the signal MS2 fragment at m/z 179 [caffeic acid – H]–, 193 [ferulic acid – H]–, and 163 [p-coumaric acid – H]–, respectively, revealing the loss of a hexosyl moiety (-162 u). The first two compounds were previously identified by Gawlik- Dziki et al. (2014), however, to the best of our knowledge, ferulic acid hexoside has not been previously identified in J. regia leaves, hence the free forms of these phenolic acids have also been previously identified (Wichtl and Anton, 1999). Peak 2 ([M – H]– at m/z 353) and peaks 3 and 4 ([M – H]– at m/z 337) were assigned as trans 3-O-caffeoylquinic acid and cis and trans 3-p-coumaroylquinic acid, respectively. These assumptions were taken into account, due to the hierarchical fragmentation pattern described by Clifford et al. (2003), but also by using literature data described by other authors, that identified these compounds in J. regia leaves (Amaral et al., 2004; Santos et al., 2013; Wichtl and Anton, 1999).

68 Characterization of the plant material CHAPTER 2

Table 2.5. Retention time (Rt), wavelengths of maximum absorption in the visible region (max), mass spectral data, tentative identification, and quantification estimation (mean ± SD) of the tentatively identified phenolic compounds in green and yellow leaves extracts of J. regia.

- 2 t-Students Rt λmax [M-H] MS Green leaves Yellow leaves Peak Tentative identification test (min) (nm) (m/z) (m/z) (mg/g extract) (mg/g extract) p-value 1 4.48 324 341 179(100) Caffeic acid hexosidea 1.80±0.08 0.44±0.01 <0.001 2 5.14 324 355 191(100),179(46),173(7),161(5),135(11) 3-O-Caffeoylquinic acida 3.5±0.4 0.191±0.005 <0.001 3 5.92 312 337 191(12),173(6),163(100),155(5),119(6) cis 3-p-Coumaroylquinic acida 2.34±0.08 0.99±0.01 <0.001 4 6.34 312 337 191(19),163(100),155(5),119(6) trans 3-p-Coumaroylquinic acida 6.9±0.5 2.87±0.09 <0.001 5 8.46 310 339 159(100),115(52) Dihydroxytetralone hexosideb,c nq nq - 6 9.28 320 355 193(100),175(20) Ferulic acid hexosided 0.73±0.06 0.46±0.01 <0.001 7 9.85 312 325 163(100) p-Coumaric acid hexosidea 1.43±0.03 0.91±0.03 <0.001 8 15.3 350 479 317(100) Myricetin-3-O-glucosided,f 0.97±0.05 0.87±0.01 0.010 9 17.87 350 463 317(100) Myricetin-O-rhamnosidef 0.78±0.01 0.83±0.03 0.038 10 18.73 353 463 301(100) Quercetin-3-O-glucosided,f 2.79±0.02 3.64±0.01 <0.001 11 19.2 353 463 301(100) Quercetin-O-hexosidef 2.28±0.2 2.75±0.03 0.005 12 20.75 350 433 301(100) Quercetin-O-pentosidef 0.67±0.02 0.857±0.005 <0.001 13 21.47 347 447 285(100) Kaempferol-3-O-glucosided 1.12±0.01 1.27±0.02 <0.001 14 21.8 351 433 301(100) Quercetin-O-pentosidef 1.43±0.02 3.21±0.03 <0.001 15 22.78 350 447 301(100) Quercetin-O-rhamnosidef 1.6±0.1 2.13±0.05 0.001 16 24.62 348 417 285(100) Kaempferol-O-pentosidef 0.721±0.002 1.03±0.04 <0.001 17 28.86 340 489 447(33),301(100) Acetylquercetin O-rhamnosided 0.627±0.005 0.81±0.01 <0.001

Total phenolic acids 16.7±0.2 5.9±0.1 <0.001 Total flavonoids 13.0±0.2 17.4±0.2 <0.001 Total phenolic compounds 29.70±0.03 23.26±0.06 <0.001 Calibration curves: Peaks 1 and 2: caffeic acid (y=388345x + 406369; r2 = 0.994; LOD=0.78 μg/mL; LOQ=1.97 μg/mL); peaks 3, 4 and 7: p-coumaric acid (y=301950x + 6966.7; r2 = 0.999; LOD=0.68 μg/mL; LOQ=1.61 μg/mL); peak 6: ferulic acid (y=63326x – 185462; r2 = 0.999 LOD=0.20 μg/mL; 1.01 μg/mL); peaks 8 to 17: quercetin 3-O-glucoside (y=34843x – 160173; r2 = 0.9998; LOD = 0.21 µg/mL; LOQ = 0.71 µg/mL). References applied for the tentative identification: a – Clifford et al. (2003), Clifford, Knight and Kuhnert (2005); b – Wang et al. (2017), c – (Jin-Hai et al., 2018); d- standard, DAD and MS fragmentation pattern; f – Santos et al. (2013).

69 Characterization of the plant material CHAPTER 2

mAU A 2200000

1800000

1400000

3 1000000 2 4 1 600000 5 7 200000 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Time (min)

mAU 600000 B

500000 10

400000

11 300000 14 13 200000 15

6 8 12 16 17 100000 9

0 0 5 10 15 20 25 30 35 40 45 50 55 60 Time (min)

Figure 2.4. HPLC phenolic profile of J. regia L. leaves. Graphical representation of walnut green sample, recorded at 280 nm (A) and 370 nm (B). The peaks identification and quantification are presented in Table 2.5.

70 Characterization of the plant material CHAPTER 2

Peak 5 ([M – H]– at m/z 353) belongs to a different group of phenolic molecules, the tetralone derivatives. This type of molecules was already reported in Juglans species. The

– fragmentation yielded the ion at m/z 159 ([M – H] - H2O – 180, loss of a hexosyl unit and a neutral loss of water) corresponding to the dihydroxytetralone. This molecule frequently occurs in branches, fruit and pericarps/husks of J. mandshurica samples (Bi et al., 2016; Gawlik-Dziki et al., 2014; Jin-Hai, Xiao-Wei, Guo-Dong, Wen-Ting, & Wei-Ming, 2018; Liu, Li, Koike, Zhang, & Nikaido, 2004; Wang et al., 2017; Zhou et al., 2015) and J. sigillata pericarps (Liu et al., 2010). To the best of our knowledge, it is the first time that the presence of dihydroxytetralone hexoside is reported as a phenolic constituent in J. regia leaves. Thus, a similar fragmentation pattern ([M – H]– at m/z 339, with a fragment at 159, UV-vis maximum absorption, 258, 313 nm) was previously found by Gawlik-Dziki et al. (2014) in J. regia leaves and pericarps. Nevertheless, the authors stated the molecule as an unknown phenolic compound (Gawlik-Dziki et al., 2014).

The last group of molecules were identified as flavonol glycoside derivatives, such as myricetin, quercetin and kaempferol derivatives. Compounds 8 (myricetin-3-O-glucoside), 10 (quercetin-3-O-glucoside), and 13 (kaempferol-3-O-glucoside) were positively identified by comparison with the corresponding commercial standards. The remaining compounds

2 (peaks 9, 11, 12, and 14-17) were identified as quercetin (λmax = 350 nm; MS fragment at

2 m/z 301), kaempferol (λmax = 348 nm; MS fragment at m/z 285), and myricetin (λmax = 350 nm, MS2 fragment at m/z 317), showing a MS2 fragmentation pattern with different losses of hexosyl (-162 u), pentosyl (-132 u), and rhamnosyl (-146 u) residues, being assigned as myricetin-O-rhamnoside (peak 9), quercetin-O-hexoside (peak 11), quercetin-O-pentoside (peaks 12 and 14), quercetin-O-rhamnoside (peak 15), and kaempferol-O-pentoside (peak 16). The majority of the identified compounds have been previously identified by Santos et al. (2013). Finally, peak 17 ([M – H]– at m/z 489) presented 42 u (acetyl moiety) higher then compound 15 (quercetin-O-rhamnoside), being therefore tentatively identified as acetylquercetin-O-rhamnoside. In fact, the occurrence of acetyl-quercetin derivatives (and kaempferol as well) in J. regia leaves has been also previously described by Gawlik-Dziki et al. (2014).

71 Characterization of the plant material CHAPTER 2

Regarding the general profile of the samples (Table 2.5), green leaves were mostly composed by phenolic acids (green samples: 16.7 ± 0.2 mg/g extract; yellow samples: 5.9 ± 0.1 mg/g extract), while the yellow ones by flavonoids (green samples: 13.0 ± 0.2 mg/g extract; yellow samples: 17.4 ± 0.2 mg/g extract). Overall, the green leaves of J. regia showed a significant higher amount of total phenolic compounds (29.70 ± 0.03 mg/g extract) in comparison to the yellow samples (23.26 ± 0.06 mg/g extract). As previous referred, phenolic acids found in walnut leaves were essentially hydroxycinnamic acids derivatives, being the trans 3-p-coumaroylquinic acid the most abundant phenolic acid for both samples (6.9 ± 0.5 and 2.87 ± 0.09 mg/g extract, for green and yellow leaves, respectively). The presence of hydroxycinnamic acid derivatives in J. regia leaves is common, regardless of the cultivar or their maturity stage, raising the importance of walnut leaves as a source of this type of bioactive compounds (Amaral et al., 2004; Gawlik-Dziki et al., 2014; Pereira et al., 2007; Pitschmann et al., 2014; Santos et al., 2013; Vieira et al., 2018, 2017; Wojdylo et al., 2007). Regarding the relative abundancy of 3-p-coumaroylquinic acid isomers as the main phenolic acids in walnut leaves samples, that is not frequent; in fact, caffeoylquinic acids have been found as the main phenolic acids. In this context, 3-O- caffeoylquinic acid was reported as the most abundant phenolic acid in several studies (Amaral et al., 2004; Pereira et al., 2007; Santos et al., 2013; Vieira et al., 2017; Vieira et al., 2018), while 5-O-caffeoylquinic acid was found in higher concentrations in Chinese samples (Zhao et al., 2014). On the other hand, ellagic acid was the main phenolic acid for Romanian samples (Pedro cultivar: 1.345 mg/g of plant dw), and caffeic acid for Polish samples (1.48 mg/g of plant dw) (Wojdylo et al., 2007). Despite the atypical relative abundancy of 3-p- coumaroylquinic acid isomers, it is relevant to mention that this compound frequently occurs in walnut leaves (Gawlik-Dziki et al., 2014; Pitschmann et al., 2014). Regarding the amounts found in Portuguese samples, Amaral et al. (2004) reported concentrations about 0.38 to 2.59 mg/g of plant dw and Pereira et al. (2007) from 4.69 to 5.99 mg/g of plant dw, depending on the J. regia cultivar, while Santos et al. (2013) obtained from 0.92 to 1.2 mg/g of plant dw. Finally, Zhao et al. (2014) quantified lower values, reaching 0.00672 mg/g of plant dw in Chinese walnut leaves.

72 Characterization of the plant material CHAPTER 2

In general, the amount of the individual flavonoids in J. regia leaves was significantly higher for the yellow samples. Quercetin-3-O-glucoside was the most abundant flavonoid for both samples, with yields of 3.64 ± 0.01 and 2.79 ± 0.02 mg/g of extract in the yellow and green leaves extracts, respectively. Quercetin-O-pentoside (peak 14) was the second main flavonoid in J. regia yellow leaves (3.21 ± 0.03 mg/g extract) and the third for the green samples (1.43 ± 0.02 mg/g extract). This tendency was also found by other authors (Santos et al., 2013; Vieira et al., 2017; Vieira et al., 2018). In this regard, Santos et al. (2013) found quercetin-3-O-glucoside concentrations around 2.37 to 2.56 mg/g of plant dw, and, 0.38 to 5.04 mg/g of plant dw for quercetin-O-pentoside for the methanolic extract and decoction, respectively. Higher amounts were found in Spanish hydroethanolic extracts by Vieira et al. (2017), achieving 13.8 to 14.2 mg/g of plant dw for quercetin-3-O-glucoside and 11.7 to 14.0 mg/g of plant dw, depending on the extraction methodology. Later, the authors increased the extraction yields, reaching 16.0 and 14.8 mg/g of plant dw for each quercetin derivative by applying choline chloride based-deep eutectic solvents. Moreover, quercetin-3-O-galactoside was reported to be the main flavonoid in Portuguese samples, and Amaral et al. (2004) obtained concentrations from 5.15 to 14.90 mg/g of plant dw, while Pereira et al. (2007) found 15.72 to 21.68 mg/g of plant dw, depending on the J. regia cultivar. The abundancy in flavonols, specially quercetin derivatives, is in good agreement with literature data (Amaral et al., 2004; Cosmulescu et al., 2014; Pereira et al., 2007; Pitschmann et al., 2014; Santos et al., 2013; Vieira et al., 2017; Vieira et al., 2018; Zhao et al., 2014). In addition, some walnut leaves also revealed the presence of other flavonoids, such as taxifolin (flavanonol), epicatechin, laricitrin, and procyanidin derivatives (Santos et al., 2013; Zhao et al., 2014).

2.2.3.2. Bioactivity of the hydroethanolic extracts The antioxidant potential for the two phenological stages of the J. regia leaves was assessed by two in vitro approaches: the inhibition of the lipid peroxidation by the TBARS assay and the anti-haemolytic activity through the OxHLIA assay. The obtained results are presented in Table 2.6.

73 Characterization of the plant material CHAPTER 2

Table 2.6. Antioxidant, anti-inflammatory, and cytotoxic activities of J. regia leaves extracts and positive controls (mean ± SD). J. regia t-Students test

Green leaves Yellow leaves p-value

Antioxidant activity (IC50 values µg/mL) OxHLIA, Δt =60 min 32±2 51±2 <0.001 TBARS formation inhibition 26.8±0.2 48.5 ±0.5 <0.001

Anti-inflammatory activity (EC50 values µg/mL) Nitric oxide (NO) production 319±13 >400 -

Cytotoxicity (GI50 μg/mL) Tumor cell lines MCF-7 (breast carcinoma) 268±12 305±5 0.001 NCI-H460 (non-small lung carcinoma) 288±5 328±5 <0.001 HeLa (cervical carcinoma) 280±8 312±2 0.002 HepG2 (hepatocelular carcinoma) 252±10 260±5 0.143 Non-tumor cells PLP2 (porcine liver primary culture) >400 >400 - IC50: extract concentration corresponding to a 50 % of antioxidant activity. Trolox IC50 values: 19.6 ± 0.6 µg/mL (OxHLIA); 5.8 ± 0.6 µg/mL µg/mL (TBARS inhibition). GI50 values correspond to the sample concentration responsible for 50 % inhibition of growth in a cell culture. GI50 values for Ellipticine (positive control): MCF-7: 1.21 ± 0.02 µg/mL; NCI-H460: 0.91 ± 0.11 µg/mL; HeLa: 1.03 ± 0.09 µg/mL; HepG2: 1.1 ± 0.09 µg/mL; PLP2: 2.29 ± 0.18 µg/mL. EC50 values correspond to the sample concentration responsible for 50 % inhibition of NO production. EC50 values for Dexamethasone (positive control): 16 ± 1 µg/mL.

The kinetic behavior of the OxHLIA was similar to the previous reported by Lockowandt et al. (2019). The heamolysis profile of the assay using the green leaves extract is depicted in Figure 2.5, representing the dose-dependent effect to maintain the erythrocyte integrity. As the graphical representation suggests, the heamolysis can be delayed by increasing the extract concentration.

The anti-haemolytic activity of walnut leaves was also evaluated by Carvalho et al. (2010), achieving IC50 values around 60 µg/mL after 3 hours of reaction (Δt ≈ 60 minutes). Regarding the IC50 values for the positive control (Trolox), they were lower than the ones obtained using the extracts (19.6 ± 0.6 and 5.8 ± 0.6 µg/mL, for OxHLIA and TBARS assays, respectively). That is, Trolox shows higher anti-haemolytic and anti-peroxidation activities than the natural extracts. However, Trolox is a pure antioxidant molecule while the extracts are composed by several compounds with different bioactive actions. The extracts showed a lower IC50 value for the TBARS assay compared with OxHLIA, being closer to the positive control in the latter case. In fact, the IC50 values of the extracts are 1.6 to 2.6 times higher than Trolox ([sample]/[Trolox]) in the OxHLIA assay, compared with the 4.2 to 8.4 times difference found for the TBARS assay.

74 Characterization of the plant material CHAPTER 2

100

75 PBS 1.25 µg/mL 2.5 µg/mL 50 P (%) P 5 µg/mL 10 µg/mL 20 µg/mL 25 40 µg/mL

0 150 200 250 300 350 400 Time (min)

Figure 2.5. Kinetic profile of the action of J. regia green leaves extract during the erythrocyte heamolysis (OxHLIA assay).

The green leaves extract showed the highest antioxidant potential for both TBARS and

OxHLIA assays, with lower IC50 values than the yellow leaves extract. Regarding the OxHLIA assay, the protection of a half of the erythrocyte population from the heamolysis was achieved by using 32 ± 2 and 51 ± 2 µg/mL concentrations, for green and yellow leaves extracts, respectively. The same tendency was observed for the TBARS assay, with slightly lower IC50 values: 26.8 ± 0.2 and 48.5 ± 0.5 µg/mL, for green and yellow samples extracts, respectively. The results of the TBARS assay for the green leaves extract is in good agreement with the previous report of J. regia leaves methanolic extract by Santos et al.

(2013), presenting a IC50 of 20.36 ± 0.82 µg/mL. Moreover, the antioxidant potential of the walnut leaf essential oil was also evaluated, reaching 50 % of TBARS inhibition formation with about 60 µg/mL (Rather et al., 2012).

The green leaves extract was the only sample showing inhibition of the NO production up to the maximum concentrations tested 400 µg/mL (Table 2.6). The concentration with the ability to inhibit 50 % of the nitric oxide production by the macrophages was found to be 319 ± 3 µg/mL. To the best of our knowledge, it is the first time that the anti-inflammatory

75 Characterization of the plant material CHAPTER 2 activity evaluation though an immune cells approach is made for this plant material. Nevertheless, the NO inhibition production by walnut leaves extracts was previously studied as antioxidant indicators. In this context, IC50 values of 1.95 ± 0.29 µg/mL were obtained by measuring the NO generated by the NOC-5 (nitric oxide amine-complex donor) decomposition. Moreover, Orhan et al. (2011) quantified the radical scavenging activity of a series of extracts obtained from different solvents (dichloromethane, ethyl acetate, acetone, methanol and water), and the leaf water extract at 2000 µg/mL was able to inhibit about 24 % of the NO production. On the other hand, some authors used mice models to perform the anti-inflammatory evaluations. The carrageenan-induced paw edema in mice assay was adopted by Erdemoglu et al. (2003), and showed potent anti-inflammatory activity for the ethanolic extract, achieving 67.9 % of inhibitory ratios, without any gastric lesions. Later, Hosseinzadeh et al. (2011) used the xylene-induced ear edema and the chronic inflammation (cotton-plate) test for the same purpose, achieving effective doses in lower concentrations for the ethanolic extract of J. regia leaves than the aqueous one. Overall, the data presented in this work demonstrate the anti-inflammatory potential of ethanolic-based extracts of walnut green leaves using a different method.

The results concerning the cytotoxic potential of the samples are also presented in Table 2.6. Regarding the results for the tumor cell cultures, the green leaves of J. regia showed a significant higher anti-proliferative potential against all the cell lines assessed. The most susceptible cell culture was the HepG2 (hepatocellular carcinoma) one, being a half of the cell proliferation inhibited by concentrations about 252 ± 10 to 260 ± 5 µg/mL (green and yellow samples, respectively), while the NCI-H460 (non-small lung carcinoma) cells were the less vulnerable ones (GI50 = 288 ± 5 and 328 ± 5 µg/mL, respectively, for the green and yellow samples). Interestingly, a sample previously studied from the same origin, did not present activity against these non-small lung cancer cell line (GI50 > 400 µg/mL) (Santos et al., 2013). This fact is easily justified by environmental changes, which contribute for phytochemical differences and then, slight variations in terms of bioactivities. Even so, these results are very close to the previously reported by the authors for methanolic and decoction extracts (Santos et al., 2013). The proliferative inhibition of human renal (A-498 and 769-P) and colon (Caco-2) tumor cell lines was prior evaluated by Carvalho et al. (2010),

76 Characterization of the plant material CHAPTER 2

achieving the walnut leaves methanolic extracts GI50 of 226, 352 and 229 µg/mL, respectively. Furthermore, the extracts studied by Santos et al. (2013) also presented anti- proliferative action against the human colon carcinoma (HCT-15) with GI50 of 215 to 258 µg/mL (methanol extract and decoction, correspondingly).

Finally, the tested extracts did not reveal cytotoxicity for the primary liver porcine cells (PLP2) up to 400 µg/mL. These results are according to the previous observations of Santos et al. (2013), for both methanolic and decoction extracts. Besides the evidences about absence of hepatotoxicity, a study performed by Eidi et al. (2013) showed a positive effect over the liver in an in vivo model. The authors observed a hepatoprotective effect in rats against the carbon tetrachloride (CCl4) induced oxidation damage in rats by an aqueous- ethanolic extract.

The walnut leaves hydroethanolic extracts were also studied regarding their antimicrobial properties. To this end, five Gram-negative (E. coli, K. pneumoniae, M. morganii, P. mirabilis, and P. aeruginosa) and three Gram-positive (E. fecalis, L monocytogenes, and MRSA) bacteria were selected. Data regarding the antibacterial activity of the extracts are presented in Table 2.7. The obtained results do not vary between samples, being the antibacterial potential of walnut leaves independent of the maturity stage. Furthermore, those results suggest a selective action against the Gram-positive strains. In fact, the achieved MIC values for the Gram-negative bacteria are about 20 mg/mL (E. coli and K. pneumoniae) or higher (M. morganii, P. mirabilis and P. aeruginosa), while the results for the Gram-positive ones are considerably lower (≥ 5 mg/mL). The lowest MIC value (highest antibacterial activity) was observed in E. faecalis and L. monocytogenes (MIC = 2.5 mg/mL). Comparatively, the methicillin-resistant S. aureus (MRSA) was inhibited with a higher concentration (MIC = 5 mg/mL). These findings allied to the fact that none of the extracts showed MBC up to 20 mg/mL, make J. regia leaves extracts good candidates for applications aiming to establish the balance of natural flora caused by Gram-positive bacteria proliferation. In this regard, the provided data of the present work, endorse the proposal of Qa’dan et al. (2005) for the use of J. regia leaves extracts for acne treatment.

77 Characterization of the plant material CHAPTER 2

In a previous study (also using bacteria from clinical isolates), Pereira et al. (2007) also reported the preferable Gram-positive susceptibility of J. regia leaves aqueous extracts from different Portuguese cultivars. The authors determined the MIC according to the agar streak dilution method based on radial diffusion, reporting concentrations about 0.1 to 1 mg/mL able to inhibit the S. aureus growth. The extracts were active against other Gram- positive bacteria (B. cereus and B. subtilis), but not for the studied Gram-negative ones (P. aeruginosa, E. coli and K. pneumonia). However, the study guided by Dolatabadi et al. (2018) using Iranian walnut leaves (aqueous and methanolic) extracts against clinical isolates of P. aeruginosa, revealed positive results. The authors used the microdilution method, determining the MIC and MBC of the samples by ELISA, achieving inhibitions with 4 and 8 mg/mL and bactericidal effect using 8 and 16 mg/mL (aqueous and methanolic extract, respectively). Iranian samples were also tested against other Gram-positive strains by Sharafati-Chaleshtori et al. (2011). The authors tested hydroethanolic extracts against face oral problematic bacteria (S. mutans, S. salivarius, S. sanguinis and A. viscosus), obtaining MIC values from 15.6 to 187.5 mg/mL and the MBC ranged from 31.25 to 250 mg/mL. The problematic bacteria M. turbeculosis was successfully inhibited by Mexican (walnut leaves) methanolic extract, with MIC values of 125 µg/mL. Finally, the bactericidal effect of Indian J. regia leaves essential oil was also an object of study by Rather et al. (2012), presenting lower MIC values against Gram-positive bacteria (B. subtilis, S. epidermidis and S. aureus) than the Gram-negative ones (E. coli, K. pneumonia, P. aeruginosa, P. vulgaris, S. typhi and S. dyssenteriae).

78 Characterization of the plant material CHAPTER 2

Table 2.7. Antimicrobial activity (MIC and MBC mean values in mg/mL) of J. regia leaves extracts and positive controls. Green Yellow Ampicillin Imipenem Vancomycin leaves leaves (20 mg/mL) (1 mg/mL) (1 mg/mL) MIC MBC MIC MBC MIC MBC MIC MBC MIC MBC Gram-negative bacteria Escherichia coli 20 >20 20 >20 <0.15 < 0.15 <0.0078 <0.0078 nt nt Klebsiella nt nt 20 >20 20 >20 10 20 <0.0078 <0.0078 pneumoniae Morganella nt nt >20 >20 >20 >20 20 >20 <0.0078 <0.0078 morganii Proteus nt nt >20 >20 >20 >20 <0.15 < 0.15 <0.0078 <0.0078 mirabilis Pseudomonas nt nt >20 >20 >20 >20 >20 >20 0.5 1 aeruginosa

Gram-positive bacteria Enterococcus 2.5 >20 2.5 >20 <0.15 <0.15 nt nt <0.0078 <0.0078 faecalis Listeria 2.5 >20 2.5 >20 <0.15 <0.15 nt nt nt nt monocytogenes MRSA* 5 >20 5 >20 <0.15 <0.15 nt nt <0.0078 <0.0078 MRSA- Methicillin Resistant Staphylococcus aureus; MIC: minimal inhibitory concentration; MBC: minimal bactericidal concentration; nt: not tested.

2.2.4. Conclusions The Portuguese walnut leaves proved to be a good source of hydroxycinnamic acid derivatives and flavonols, especially trans 3-p-coumaroylquinic acid and quercetin-3-O- glucoside. The former compounds were abundant in green samples, while the later ones were found in higher amounts in the yellow samples. The green leaves extract showed the highest antioxidant activity, being more capable to inhibit the lipid peroxidation as well as the erythrocyte heamolysis. Similarly, in the anti-inflammatory assay, only green samples presented anti-inflammatory potential. It was also verified a cytotoxic effect against tumor cell lines, but not for the non-tumor ones. Furthermore, both extracts were active against Gram-positive bacteria. Overall, this study shows the importance of walnut leaves as a source of bioactive molecules, providing antioxidant, anti-inflammatory, anti-proliferative and antibacterial properties, with potential to be used by different industries.

79 Characterization of the plant material CHAPTER 2

2.3. Phytochemical profile of Juglans regia L.: valorization of walnut residues through their bioactive potential Vieira, V., Pereira, C., Abreu, R.M.V., Calhelha, R.C., Alves, M.J., Coutinho, J.A.P., Ferreira, O., Barros, L., Ferreira, I.C.F.R., 2019. Food Chem., submited.3

Abstract

Juglans regia L. (walnut) green husks represent an important fraction of waste resulting from the walnut production. In this work, the phytochemical composition and bioactivity of the walnut green husks were studied. From the HPLC-DAD-ESI/MSn analysis, sixteen compounds were identified, showing the presence of one organic acid, two phenolic acids (hydroxycinnamic acids), six tetralone derivatives, three naphthalene derivatives and four flavonoids (quercetin derivatives). The bioactivity of the extract was studied considering the antioxidant, anti-inflammatory and cytotoxic potentials using cellular models, most of them applied for the first time to this walnut by-product. The antibacterial effects of the extract were also evaluated against Gram-negative and Gram-positive bacteria. The results obtained represent a stepping stone for the development of future applications, using walnut green husks as a source of added value compounds.

Keywords: Juglans regia L. husks, phenolic compounds, naphthalene derivatives, antioxidant, anti-inflammatory, antibacterial, cytotoxicity.

3 My contribution to this work was the measurement of the experimental data, their interpretation and writing of the manuscript.

80 Characterization of the plant material CHAPTER 2

2.3.1. Introduction Natural products are recognised sources of bioactive compounds that can find many applications as natural antioxidants, antimicrobial, anti-inflammatory, and natural colouring agents, among other functions (Carocho and Ferreira, 2013; Lockowandt et al., 2019; Pinela et al., 2019; Pires et al., 2018; Rostagno and Prado, 2013). Among them, by- products generated in the food and agricultural processing industries can be valorised as natural sources of antioxidants (Balasundram et al., 2006). That is the case of the green husks of walnut trees (Juglans regia L.), a common species in Portugal (Pereira et al., 2007) and the most widespread nut tree in the world (Martínez et al., 2010). They are part of the resulting waste from walnut (fruits) production and their extracts were already proposed as a natural source of dyeing and antimicrobial agents for cosmetic products (Beiki et al., 2018) or the reducing and stabilizing agents in the biosynthesis of gold nanoparticles (Izadiyan et al., 2018). In the food area, walnut green husks were studied as additives with functional properties in meat processing (Salejda et al., 2016) and their extracts were applied to preserve the quality of fresh walnuts during storage (Chatrabnous et al., 2018).

To assist the development of new applications, it is important to have a detailed characterization of the phytochemical profile of the extracts and, also, of their bioactivity. A few studies can be found for J. regia L. in which the phytochemical composition of green husks has been determined by HPLC coupled to a UV/DAD detector (Akbari et al., 2012; Chatrabnous et al., 2018; Cosmulescu et al., 2011, 2010; Liu et al., 2008; Soto-Maldonado et al., 2019; Stampar et al., 2006), reporting the amounts of juglone, phenolic acids, flavonoids and/or tetralone derivatives in the biomass. In another study, phenolic acids and tetralone derivatives were the main compounds identified in samples collected in China, using spectroscopic analysis (UV–Vis, 1H NMR, 13C NMR, 2D NMR, HR-ESI-MS) (Du et al., 2014). Other authors identified phenolic acids and flavonoids during the characterization performed this time by LC-MS (Gawlik-Dziki et al., 2014).

Regarding the bioactivity studies of the walnut green husks, remarkable results regarding its antioxidant activity have been reported, mainly because of their free-radical scavenger and metal chelator capacity (Akbari et al., 2012; Fernández-Agulló et al., 2013; Meshkini and Tahmasbi, 2017). However, only a few of studies (Bagheri et al., 2012; Carvalho et al.,

81 Characterization of the plant material CHAPTER 2

2010; Soto-Maldonado et al., 2019) provided cellular antioxidant activity models which are considered better representatives of the complexity of biological systems (Gupta, 2015). Cellular models have been also used to evaluate the antimicrobial and anti-tumour potentials in previous works (Alshatwi et al., 2012; Carvalho et al., 2010; Fernández-Agulló et al., 2013; Keskin et al., 2012; Oliveira et al., 2008; Sharma et al., 2013; Soto-Maldonado et al., 2019; Zhang et al., 2014).

The present work aims to further characterize the phytochemical composition of Juglans regia L. green husks by liquid chromatography combined with a diode array detector and electrospray ionization mass spectrometer (LC-DAD-ESI/MSn), and to extend the study of the bioactive potential of their hydroethanolic extracts using different in vitro cellular approaches. Therefore, the bioactivity of the extract was studied by evaluating the antioxidant (the thiobarbituric acid reactive substances and the oxidative haemolysis inhibition assays), anti-inflammatory (inhibition of the nitric oxide production by macrophages), cytotoxicity (tumor and non-tumor cell lines), and antibacterial (Gram- negative and Gram-positive bacteria) properties. To the best of our knowledge, it is the first time that the selected cell models are applied to evaluate the anti-inflammatory and cytotoxic activities of this extract.

2.3.2. Material and methods

2.3.2.1. Plant material and extract preparation Juglans regia L. green husks were collected in Bragança, Northeast of Portugal, during October of 2018. The samples were dried to a constant weight in an incubator at 35ºC. After, the plant material was ground (≈ 40 mesh), and the homogeneous sample was stored in a desiccator protected from light for subsequent assays.

The extract was obtained by stirring the raw material (1 g) in an aqueous ethanolic solution (80 % ethanol, v/v; 30 mL) at room temperature (25 °C) for 60 minutes. After filtration (Whatman no. 4), the extraction procedure was repeated once. Then, the solvent was recovered in order to obtain a dry extract: first, evaporation at 40 °C and reduced pressure

82 Characterization of the plant material CHAPTER 2

(rotary evaporator Büchi R-210, Flawil, Switzerland), and then freeze-drying (Telstar Cryodos-80, Terrassa, Barcelona).

2.3.2.2. Phytochemical characterization The dry extract was re-suspended at 10 mg/mL using aqueous ethanol (80% ethanol, v/v) and filtered through a 0.2 µm disposable LC filter disk (30 mm, nylon). The phytochemical characterization of walnut green husks was traced by liquid chromatography with diode- array detection (280, 330 and 370 nm wavelengths) coupled to electrospray mass ionization operating in negative mode (Dionex Ultimate 3000 UPLC and Linear Ion Trap LTQ

XL, Thermo Scientific, San Jose, CA, USA) by using a Waters Spherisorb S3 ODS-2 C18 (3 μm, 4.6 mm × 150 mm), as previously described by the authors (Bessada et al., 2016). The compounds were identified according to their retention time, UV-vis and mass spectra by comparison with those obtained using standard compounds, as well as with literature data. Calibration curves of appropriate standards (p-coumaric acid, α-tetralone, juglone and quercetin-3-O-glucoside, HPLC grade, Sigma-Aldrich) were obtained in the range 200 - 5 µg/mL, for the quantitative analysis. The results were expressed as milligrams of each compound per gram of extract (mg/g). Triplicates were analyzed with two independent injections.

2.3.2.3. In vitro antioxidant assays The antioxidant activity was evaluated by measuring the thiobarbituric acid reactive substances (TBARS) and by the oxidative haemolysis inhibition (OxHLIA) assay. In this context, the extract was diluted at a concentration of 10 mg/mL (distilled water and PBS solution, respectively for the TBARS and OxHLIA assays) and then, further dilutions were carried out (500 – 6.25 µg/mL).

The TBARS assay was performed by measuring the color intensity of the malondialdehyde- thiobarbituric acid (MDA-TBA) at 532 nm, following a methodology described by Barreira et al. (2013), and the results were expressed as IC50 values (sample concentration providing 50 % of antioxidant activity).

83 Characterization of the plant material CHAPTER 2

The OxHLIA assay was carried out by evaluating the delay effect of the extract on the erythrocyte haemolysis at 690 nm, as previously described by Lockowandt et al. (2019).

The results were presented as IC50 values (extract concentration that delayed the haemolysis time for 60 min, with 50 % of intact erythrocytes). Trolox was used as positive control for both antioxidant activity evaluations. Triplicates were used and three independent assays were performed.

2.3.2.4. Anti-inflammatory activity The walnut green husks extract was re-suspended in water at a concentration of 10 mg/mL and then diluted in the range of 500 to 7.8 µg/mL. To perform the assay, the nitric oxide production by a mouse macrophage-like cell line (RAW264.7) was followed, using the Griess Reagent System (GRS) kit. The inhibition of the NO production was performed according to the methodology described by Corrêa et al. (2015), with measurements at 515 nm (ELx800 microplate reader, Bio-Tek Instruments, Inc; Winooski, VT, USA). Results were expressed as EC50 values (sample concentration providing 50 % of inhibition of NO production) and dexamethasone was used as a positive control, while for the negative control, no lipopolysaccharide (LPS) was added. For the assay, triplicates were used in three independent assays.

2.3.2.5. Cytotoxicity assays The hydroethanolic extract of walnut green husks was re-suspended in water at 10 mg/mL and, then, submitted to further dilutions ranging from 500 to 7.8 µg/mL. The cytotoxic properties were assessed using four human tumor cell lines: MCF-7 (breast adenocarcinoma), NCI-H460 (non-small cell lung cancer), HeLa (cervical carcinoma), and HepG2 (hepatocellular carcinoma). A non-tumor cell line (porcine liver primary cells, PLP2) was also evaluated using a procedure described by Abreu et al. (2011). The Sulforhodamine B assay was performed according to a protocol described elsewhere (Barros et al., 2013). Ellipticine was used as a positive control, while the negative control was represented by each suspension of cells. The results were expressed in GI50 values (concentration that inhibited 50 % of the cell proliferation). Three independent assays were performed using triplicates.

84 Characterization of the plant material CHAPTER 2

2.3.2.6. Antibacterial activity The hydroethanolic extract of walnut green husks was dissolved in a mixture of dimethyl sulfoxide (DMSO) + Mueller-Hinton Broth (MHB)/ Tryptic Soy Broth (TSB) (5 + 95 %, v/v) to give a final concentration of 100 mg/mL for the stock solution and, then, successive dilutions were carried out ranging from 20 to 1.25 mg/mL. The antimicrobial potential was assessed using five Gram-negative bacteria (Escherichia coli, Proteus mirabilis, Klebsiella pneumoniae, Pseudomonas aeruginosa and Morganella morganii) and three Gram-positive bacteria (Enterococcus faecalis, Listeria monocytogenes, methicillin-resistant Staphylococcus aureus: MRSA). These strains were clinical isolates donated by hospitalized patients (Local Health Unit of Bragança and Hospital Centre of Trás-os-Montes and Alto- Douro, Vila Real, Northeast of Portugal) with multi-resistant profile, previously characterized by Alves et al. (2014). For each bacteria, the minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) were determined using a colorimetric assay as described by Kuete et al. (2011) and Pires et al. (2018). Duplicates were used and three independent assays were performed.

2.3.3. Results and discussion

2.3.3.1. Phytochemical composition of walnut husks The phytochemical composition of the hydroethanolic extracts of J. regia green husks was characterized for samples at the mature stage of the fruit (October 2018) and the main identified molecules are presented in Table 2.8 with their quantification. The tentative identification was carried out according to their retention time (Rt), maximum absorbance

– wavelength (λmax), pseudomolecular ion ([M – H] ), and the correspondent fragmentation pattern (MS2), and also using literature information. It was possible to identify sixteen compounds as presented in Figure 1. The extract was composed of three organic acids, of which two phenolic acids (p-hydroxycinnamic acid derivatives), six tetralone derivatives, three naphthalene derivatives, and four flavonoids (flavonols). The phytochemical profile of Juglans regia green husks is quite variable; however, our findings are in good agreement with the most abundant groups of compounds found in literature for Juglans spp. samples such as phenolic acids and flavonoids (Cosmulescu et al., 2010; Gawlik-Dziki et al., 2014;

85 Characterization of the plant material CHAPTER 2

Stampar et al., 2006), as well as tetralone and naphthalene derivatives (Du et al., 2014; Jin- Hai et al., 2018; Wang et al., 2017).

86 Characterization of the plant material CHAPTER 2

Table 2.8. Retention time (Rt), wavelengths of maximum absorption in the visible region (max), mass spectral data, tentative identification, and quantification estimation (mean ± SD, n = 6) of the tentatively identified compounds in green husks extracts of J. regia.

Rt λmax [M-H]- MS2 Walnut green husks Peak Tentative identification (min) (nm) (m/z) (m/z) (mg/g extract) 1 5.2 212 133 115(100),89(10),71(8) Malic acida n.q. 2 6.0 307 337 191(10),173(10),163(100),155(5),119(7) 3-p-Coumaroylquinic acidb 1.07±0.03 3 7.7 262,320 339 159(100),115(52) Dihydroxytetralone hexosidea,c 0.42±0.02 4 8.4 222,263 409 337(28),247(100),203(77),175(84),131(5) Trihydroxynaphthalene hexoside derivativec 4.2±0.2 5 10.9 263,320 177 159(100),131(15),115(5) Dihydroxytetralonea,c 7.1±0.4 6 11.9 310 337 191(5),173(100),163(5),119(3) 4-p-Coumaroylquinic acidb 1.356±0.001 7 13.0 223,264 193 175(100),157(5),147(3),131(3) Trihydroxytetralonea,c 1.174±0.005 8 15.7 260,320 491 331(92),271(100), 211(100),169(10) Dihydroxytetralone galloyl-hexoside isomer 1a,c 2.8±0.1 9 16.6 220,264 491 331(94),271(100),211(5),169(3) Dihydroxytetralone galloyl-hexoside isomer 2a,c 1.486±0.003 10 17.7 213,261,330 507 331(100),271(42),211(5),169(3) Trihydroxytetralone galloyl-hexosidea 1.23±0.04 11 18.5 350 463 301(100) Quercetin 3-O-glucosided 0.3255±0.0003 12 18.8 348 463 301(100) Quercetin O-hexosided 0.265±0.005 13 20.3 222,273 489 313(61),301(56),271(100),211(15),169(8) Trihydroxynaphthalene galloyl-hexosidec 4.7±0.3 14 21.4 350 433 301(100) Quercetin O-pentosided 0.3822±0.0003 15 22.4 350 447 301(100) Quercetin O-deoxyhexosidef 0.321±0.001 16 26.2 223,263 487 325(100),307(10) Jugnaphthalenoside Ac 0.76±0.02 Total phenolic acids 2.42±0.03 Total flavonoids 1.293±0.004 Total phenolic compounds 3.72±0.03 Total tetralone derivatives 14.2±0.2 Total naphthalene derivatives 9.68±0.04 Calibration curves: Peaks 2 and 6: p-coumaric acid (y=301950x + 6966.7; r2 = 0.999); peaks 3, 5, 7, 9, 10 and 11: α-tetralone (y=6173.3x + 58913; r2 = 0.9991; LOD = 3.1 µg/mL; LOQ = 9.5 µg/mL); peaks 4, 14: juglone (y=3754.1x+ 45119; r2 = 0.9992; LOD = 2.9 µg/mL; LOQ = 8.8 µg/mL); peaks 12, 13, 15 and 16: quercetin 3-O-glucoside (y=34843x – 160173; r2 = 0.9998; LOD = 0.21 µg/mL; LOQ = 0.71 µg/mL). References applied for the tentative identification: a – Wang et al. (2017); b –Clifford et al. (2003), Clifford, Knight and Kuhnert (2005); c – Jin-Hai et al. (2018) ; d – Santos et al. (2013); f – Gawlik-Dziki et al. (2014). n.q.: not quantified.

87 Characterization of the plant material CHAPTER 2

mAU A 1400000

1000000

1

5 6 2 600000 8 4 13 3 9 10 7

200000

0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 Time (min)

mAU 180000 B

11 140000

7 15 100000 12 14

16

60000

20000

0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 Time (min)

Figure 2.6. HPLC phenolic profile of J. regia L. green husks at 280 nm (A) and 370 nm (B). The peaks identification and quantification are presented in Table 2.8.

88 Characterization of the plant material CHAPTER 2

The first compound presented a fragment at m/z 133 ([M – H]–) and maximum UV absorption at 212 nm, additional MS2 fragments at m/z 115, 89 and 71 were found; therefore, compound 1 was tentatively identified as malic acid. The presence of this organic acid in Juglans spp., as well as the MS/MS fragmentation pattern, was previously reported by Wang et al. (2017) in J. mandshurica Maxim. branches.

The two phenolic acids identified in J. regia green husks, peaks 2 and 6 ([M – H]– at m/z 337), were assigned as 3- and 4-p-coumaroylquinic acid, respectively. The MS2 fragments at m/z 191, 173, 163, 155 and 119 are in accordance to the hierarchical fragmentation pattern described by Clifford et al. (2003), with base peak at m/z 163 for the 3-p- coumaroylquinic acid and m/z 173 for the 4-p-coumaroylquinic acid. These assumptions were also in agreement with the observations of Gawlik-Dziki et al. (2014) using the same plant material, reporting similar mass fragmentations and maximum UV-Vis absorption (307 and 310 nm, respectively). The amounts of both hydroxycinnamic acids were very similar (1.07 ± 0.03 mg/g dry extract and 1.356 ± 0.001 mg/g dry extract, for 3-p- coumaroylquinic and 4-p-coumaroylquinic acids, respectively), yielding a total amount of phenolic acids of 2.42 ± 0.03 mg/g dry extract.

Peaks 11 and 12 ([M – H]– at m/z 463), 14 ([M – H]– at m/z 433) and 15 ([M – H]– at m/z 447) yielded the same base peak at m/z 301. These compounds were identified as quercetin glycoside derivatives, more precisely: quercetin-3-O-glucoside (11), quercetin-O-hexoside (12), quercetin-O-pentoside (14) and quercetin-O-deoxyhexoside (15). The presence of these flavonols are in agreement with the results reported by Gawlik-Dziki et al. (2014), as well as the mass fragments and UV-vis maximum absorptions for the same plant material. Regarding the amounts found for the quercetin derivatives, they ranged from 0.265 ± 0.005 mg/g dry extract for the O-hexoside form to 0.3822 ± 0.0003 mg/g dry extract for the O- pentoside one, yielding a total amount of flavonoids of 1.293 ± 0.004 mg/g dry extract. These compounds were the least abundant class of molecules found.

The following group of compounds is based on a tetralone structure. Tetralone derivatives are widely distributed in Juglans spp. (Liu et al., 2010; Wang et al., 2017). Particularly, the tetralone derivatives found in green husks of J. regia usually occur as hydroxyl, hexosyl and

89 Characterization of the plant material CHAPTER 2 hydroxybenzoyl derivatives, with several isomeric alternatives (Zhou et al., 2015). In this work, peak 3 ([M – H]– at m/z 339), peak 5 ([M – H]– at m/z 177) and peak 7 ([M – H]– at m/z 193) were identified as dihydroxytetralone hexoside, dihydroxytetralone and trihydroxytetralone, respectively. These identifications were possible by comparison with literature data, describing the different fragmentation patterns of individual compounds found in J. mandshurica samples (Jin-Hai et al., 2018; Wang et al., 2017). Peaks 8 and 9 ([M – H]– at m/z 491), and peak 10 ([M – H]– at m/z 507) showed similar MS2 fragments at m/z 271, 211 and 169. Following the observations of Jin-Hai et al. (2018) and Wang et al. (2017), the compounds were identified as tetralone galloyl hexoside derivatives, specifically, dihydroxytetralone galloyl hexoside isomers (peaks 8 and 9) and trihydroxytetralone galloyl hexoside (peak 10). Furthermore, this group of molecules yielded in total 14.2 ± 0.2 mg/g dry extract, being the main group of compounds present in walnut green husks. Dihydroxytetralone was the most abundant compound (7.1 ± 0.4 mg/g dry extract), about half of the total amount of tetralone derivatives. Interestingly, dihydroxytetralone galloyl- hexoside isomers were the second most abundant compounds (2.8 ± 0.1 mg/g dry extract and 1.486 ± 0.003 mg/g dry extract, for isomers 1 and 2, respectively), and dihydroxytetralone hexoside was the least abundant tetralone derivative (0.42 ± 0.02 mg/g dry extract).

Peaks 4, 13 and 16 were classified as naphthalene derivatives. This type of molecules has already been found in J. mandshurica fruits (Jin-Hai et al., 2018), husks (Jin-Hai et al., 2018; Zhou et al., 2015) and branches (Wang et al., 2017). As with tetralone derivatives, natural naphthalene also occurs with hydroxyl, hexosyl and hydroxybenzoyl substitutions (Sun et al., 2012). Peak 4 ([M – H]– at m/z 409) was identified due to its MS2 fragments, which comply to the previous fragmentation pattern descriptions of Jin-Hai et al. (2018) and Wang et al. (2017) for a trihydroxynaphthalene hexoside; thus, the loss of -72u was unknown and therefore, peak 4 was assigned to a trihydroxynaphthalene hexoside derivative. On the other hand, peak 13 ([M – H]– at m/z 489) was identified as trihydroxynaphthalene galloyl- hexoside, while peak 16 ([M – H]– at m/z 487) as jugnaphthalenoside A. These assumptions take into account the identifications according to the previous fragmentation patterns reported by Jin-Hai et al. (2018). Regarding the quantification, trihydroxynaphthalene

90 Characterization of the plant material CHAPTER 2 hexoside derivative and trihydroxynaphthalene galloyl-hexoside were present in similar amounts (4.2 ± 0.2 mg/g dry extract and 4.7 ± 0.3 mg/g dry extract, respectively), considerably higher than the concentration obtained for jugnaphthalenoside A (0.76 ± 0.02 mg/g dry extract). Naphthalene derivatives were the second group of compounds present in higher concentration.

The results obtained show that walnut green husks are an important source of molecules with biological activity, with potential application in several areas including food, cosmetic and pharmaceutical industries (Balasundram et al., 2006; Butler et al., 2014; Newman and Cragg, 2016).

2.3.3.2. Bioactivity of the hydroethanolic extracts The antioxidant activity of walnut green husks was evaluated by two in vitro assays using cells and tissues. The extract potential for the inhibition of the lipid peroxidation (evaluated by the thiobarbituric acid reactive substances – TBARS assay) was assessed by using porcine brain tissues, while the maintenance of the erythrocyte integrity by the extract action was performed by the oxidative haemolysis inhibition (OxHLIA) assay. The results are presented in Table 2.9, and the anti-haemolytic action of the extracts is presented in Figure 2.7.

The mean concentration leading to half of the malondialdehyde formed during the peroxidation of the unsaturated fatty acids present in the brain tissues, in the TBARS assay, was IC50 =101 ± 4 µg/mL. In the OxHLIA) assay, the inhibition of the erythrocyte’s membrane damage promoted by the free radical AAPH was achieved at IC50 = 80 ± 4 µg/mL. The IC50 values obtained by the OxHLIA assay were lower than the ones for the TBARS, contrarily to the results found for the positive control, with IC50 = 19 ± 0.6 µg/mL and 5.8 ± 0.6 µg/mL, respectively.

91 Characterization of the plant material CHAPTER 2

Table 2.9. Antioxidant activity, NO formation inhibition and cytotoxicity of J. regia green husks (mean ± SD, n = 9).

Hydroethanolic extract Positive control

Antioxidant activity (EC50 values, µg/mL)

TBARS inhibition 101 ± 4 5.8 ± 0.6

OxHLIA 80 ± 4 19.6 ± 0.6

Anti-inflammatory activity (EC50 values, µg/mL)

Nitric oxide (NO) production 56 ± 3 16 ± 1

Cytotoxicity (GI50 values, µg/mL)

Tumour cell lines

MCF-7 (breast carcinoma) 26 ± 1 1.21 ± 0.02

NCI-H460 (non-small lung cancer) 41 ± 2 0.9 ± 0.1

HeLa (cervical carcinoma) 34 ± 2 1.03 ± 0.09

HepG2 (hepatocellular carcinoma) 24 ± 2 1.1 ± 0.09

Non-tumour cell line

PLP2 (porcine liver primary cells) 87 ± 4 2.3 ± 0.2 Trolox was used as positive control for the antioxidant properties, dexamethasone for the anti-inflammatory activity, while ellipticine for the cytotoxicity assays.

100

75

PBS 3.75 µg/mL 7.5 µg/mL 50 P (%) P 15 µg/mL 30 µg/mL 60 µg/mL 120 µg/mL 25

0 100 150 200 250 300 350 400 Time (min) Figure 2.7. Kinetic behavior of J. regia green husks extract during the erythrocyte heamolysis (OxHLIA assay).

92 Characterization of the plant material CHAPTER 2

To our knowledge, the effect of walnut green husks against the lipid oxidation through the TBARS assay was only reported by Bagheri et al. (2012). The hydroalcoholic walnut green husk extract was able to inhibit the LDL oxidation with a dose-dependency from 0.2 to 20 µg/mL. Regarding the anti-haemolytic potential, Carvalho et al. (2010) also evaluated the dose-dependency between different concentrations of the walnut green husks methanolic extract and the time of the AAPH free radical exposure. The IC50 mean value was 127 ± 14 µg/mL after 3 hours of reaction (Δt ≈ 60 minutes), which is in agreement with the results obtained in this work.

The anti-inflammatory potential was assessed by using macrophages cells (Table 2.9). After recognition of a pathogen, macrophages release a series of toxic molecules as an immune defense, being one of them nitric oxide (NO). In this context, NO is a recognized intercellular marker in the immune system, being involved in immunologically mediated diseases and inflammation (Tripathi et al., 2007). A concentration of 56 ± 3 µg/mL inhibited half of the NO produced by the macrophage’s cultures (EC50). This result was 3.5-fold higher than the pure standard used for the positive control, dexamethasone, with EC50 = 16 ± 1 µg/mL. To the best of our knowledge, this is the first time this assay was applied to the extracts of walnut husks. The inhibition of the NO production is frequently carried out by simple chemical reactions using sodium nitroprusside, but in those cases only antioxidant activity is evaluated (Akbari et al., 2012; Fernández-Agulló et al., 2013; Ghasemi et al., 2011; Sharma et al., 2013). It is usually applied as an antiradical indicator because the source of NO is not provided by cells (e.g. macrophages). In this context, some studies about the NO scavenging activity of walnut green husks can be found. Akbari et al. (2012) showed that the methanolic extracts of six green husks genotypes were able to capture, on average, about 69.18 ± 1.57% of the NO free radicals. The results for the NO inhibition in sample extracts from India (obtained with different organic solvents) were all > 500 µg/mL (Sharma et al., 2013). Ghasemi et al. (2011) obtained IC50 values ranging from 141 to 2890 µg/mL for methanolic extracts depending on the J. regia cultivar. Regarding the NO scavenging activity of other samples from Portuguese origin, the aqueous extract of a Mellanaise walnut husk variety resulted in EC50 values of 960 ± 130 µg/mL (Fernández-Agulló et al., 2013). Though the results cannot be directly compared, as they were obtained in different

93 Characterization of the plant material CHAPTER 2 assays, the concentration values from the chemical assay were all higher than the ones obtained in this study with the cellular model.

Beyond the intrinsic characteristics of the plant material (geographic location, time of collection), the differences found in literature are also due to the selected extraction conditions (extraction technique, solvent, time), which are selective factors during the extraction of compounds from natural matrices (Sampaio et al., 2016; Zhang et al., 2018b).

Regarding the cytotoxicity of walnut green husks, the results for the tumor and non-tumor cell cultures are presented in Table 2.9. The extract showed similar activity against the four tumor cell lines (non-small lung cancer, breast, cervical and hepatocellular carcinomas), with lower GI50 values (higher cytotoxic potential) against the HepG2 (hepatocellular carcinoma) and MCF-7 (breast carcinoma) cells, reaching 24 ± 2 µg/mL and 26 ± 1 µg/mL, respectively. Slightly higher concentrations providing half of the cell proliferation were obtained against HeLa (cervical carcinoma) and NCI-H460 (non-small lung cancer) cell lines, with GI50 values of 34 ± 2 µg/mL and 41 ± 2 µg/mL, respectively. Concerning the results of the hepatotoxicity, the extract was toxic for the primary liver cells; however, at higher concentrations compared to the tumor cell lines, with GI50 = 87 ± 4 µg/mL.

To the best of our knowledge, it is the first time that the cytotoxic potential of J. regia green husks extracts is studied against the proliferation of the selected cell cultures.

Nevertheless, the cytotoxic effect of pure isolated compounds from Juglans mandshurica Maxim extracts was evaluated for HepG2 cells (Yang et al., 2019; Zhou et al., 2015).

According to Zhou et al. (2015), lower IC50 values were found for naphthoquinone derivatives than the ones obtained for tetralone, and the non-glycosylated forms were more active against the HepG2 cells than the glycosylated ones: dihydroxynaphthoquinone isomers (7.33 ± 0.52 µM to 15.37 ± 1.63 µM) < dihydroxy tetralone (regiolone: 56.87 ± 4.27 µM) < trihydroxy naphthoquinone hexoside isomers (78.61 ± 2.38 µM to 83.32 ± 4.54 µM), while dihydroxy- and trihydroxytetralone hexoside isomers were not active against the cell culture. Similar results were found for regiolone in Juglans cathayensis Dode samples against the same cell line (42.56 µM) by Li et al. (2008b). The authors also presented results about the regiolone toxicity against HeLa (30.48 µM) and HL-60 (40.21 µM) cell cultures.

94 Characterization of the plant material CHAPTER 2

Furthermore, two tetralone dimers (juglanone A and B) were isolated from J. regia pericarps (husks) by Li et al. (2013a). The cytotoxicity of the pure compounds was assessed against a series of human cell lines (A549, MCF-7, BEL-7402, HeLa, COLO205, BGC-823, and SK-OV-3). Juglanone A showed a higher inhibition rate for the MCF-7 cells (94.16 ± 0.37 %), while juglanone B revealed higher activity against HeLa (92.62 ± 2.00 %). Also, the five cyclic diarylheptanoids isolated by Yang et al. (2019) from the ethanolic (95 %) extract resulted in IC50 values in the range of 27.72 ± 3.71 µM (Juglanin H) to 383.54 ± 29.57 µM (Juglanin J).

Regarding the anti-proliferative potential of the green husks from J. regia, Carvalho et al.

(2010) reported IC50 values of 285 ± 51 µg/mL and 496 ± 71 µg/mL for A-498 and 769-P (human kidney carcinomas) cell lines, respectively. For the Caco-2 (colorectal carcinoma) cells, the IC50 was higher than 500 µg/mL. The increase on the TUNEL-positive apoptotic cell count of different extracts (methanol, n-hexane and chloroform) from this J. regia by- product was assessed for PC-3 (human prostate cancer) cells by Alshatwi et al. (2012), and the n-hexane extract was the most active. Moreover, the recent findings of Soto- Maldonado et al. (2019) showed the antiproliferative and cytotoxic potentials of ethanolic extracts against the HL-60 cells, in which the plant extract presented higher activity than pure juglone. Overall, walnut green husks have shown to be a rich natural source of bioactive compounds with anti-proliferative potential, either in terms of the final extract or of their pure components.

The antimicrobial properties of the hydroethanolic extract of walnut green husks were also studied. Five Gram-negative (Escherichia coli, Klebsiella pneumoniae, Morganella morganii, Proteus mirabilis, and Pseudomonas aeruginosa) and three Gram-positive (Enterococcus fecalis, Listeria monocytogenes, and methicillin-resistant Staphylococcus aureus) bacteria obtained from clinical isolates were used (Alves et al., 2014). The results of the antibacterial activity of the extract and positive controls are presented in Table 2.10. Both type of bacteria were susceptible to the ethanolic extract, being the lowest MIC (higher anti- bacterial potential) obtained for the methicillin-resistant Staphylococcus aureus (MRSA) strain (MIC = 5 mg/mL). The double of inhibitory concentration (10 mg/mL) was found to be effective against one Gram-positive (L. monocytogenes) and one Gram-negative (E. coli)

95 Characterization of the plant material CHAPTER 2 bacteria. However, under the experimental concentration range, the extract was not able to inhibit P. aeruginosa (MIC > 20 mg/mL), being the remaining species susceptible to a concentration of 20 mg/mL (K. pneumoniae, M. morganii, P. mirabilis, and E. fecalis). In all cases, the minimum bactericidal concentration was higher than 20 mg/mL.

The antibacterial potential of Portuguese walnut green husks was previously evaluated by Oliveira et al. (2008). The authors evaluated the decoction extracts of several J. regia cultivars against different Gram-negative and Gram-positive bacteria with positive inhibitions against S. aureus (MIC = 0.1 mg/mL), Bacillus cereus (MIC = 0.1 to 1 mg/mL), Bacillus subtilis (MIC = 0.1 to 10 mg/mL) and P. aeruginosa (MIC = 100 mg/mL). Later, Keskin et al. (2012) evaluated the activity of aqueous extracts by the disk diffusion method, obtaining higher inhibition zones against Pseudomonas fluorescens (15 mm) and lower for B. subtilis and P. aeruginosa (8 mm). The water extracts, studied by Fernández-Agulló et al. (2013), have shown higher potential against B. cereus (MIC = 20 mg/mL), but lower for the Gram-negative P. aeruginosa and E. coli (MIC = 100 mg/mL). On the other hand, Sharma et al. (2013) evaluated the antibacterial activity of walnut green husks extracts obtained by different solvents (ethanol, ethyl acetate and water). Generally, the ethanolic extract showed higher diameter of inhibition against the assessed bacteria (E. coli, K. pneumoniae and S. aureus), while the ethyl acetate extract was most effective against B. subtilis. Finally, the results obtained by Zhang et al. (2014) for the chloroform fraction of Chinese samples showed the lowest MIC values against E. coli, P. aeruginosa, B. subtilis and S. aureus (6.25, 1.56, 3.13 and 3.13 mg/mL, respectively). In general, the MIC values obtained in the present work were higher in comparison with those reported in literature. These differences can be caused by several factors, such as the extraction methodology and solvents applied, and other factors related to the plant material, as previously discussed. Moreover, the selected microorganisms can also be a determinant factor, because the strains used in this work are multi-resistant clinical isolates, which may demand higher extract concentrations to inhibit their growth.

96 Characterization of the plant material CHAPTER 2

Table 2.10. Antibacterial properties of J. regia green husks against Gram-positive and Gram- negative bacteria (mean, n = 6).

Ethanolic Ampicillin Imipenem Vancomycin extract

MIC MBC MIC MBC MIC MBC MIC MBC Gram-negative bacteria Escherichia coli 10 >20 <0.15 <0.15 <0.0078 <0.0078 n.t. n.t. Klebsiella pneumoniae 20 >20 10 20 <0.0078 <0.0078 n.t. n.t. Morganella morganii 20 >20 20 >20 <0.0078 <0.0078 n.t. n.t. Proteus mirabilis 20 >20 <015 <0.15 <0.0078 <0.0078 n.t. n.t. Pseudomonas aeruginosa >20 >20 >20 >20 0.5 1 n.t. n.t.

Gram-positive bacteria Enterococcus faecalis 20 >20 <0.15 <0.15 n.t. n.t. <0.0078 <0.0078 Listeria monocytogenes 10 >20 <0.15 <0.15 <0.0078 <0.0078 n.t. n.t. MRSA 5 >20 <0.15 <0.15 n.t. n.t. 0.25 0.5 MRSA: Methicillin-resistant Staphylococcus aureus; n.t.: not tested.

2.3.4. Conclusions Walnut green husks presented a diverse phytochemical profile including different classes of phenolic compounds and naphthalene derivatives, including tetralone derivatives. The extract was richer in tetralone derivatives (14.2 ± 0.2 mg/g) compared to the other bioactive phytochemicals such as phenolic acids and flavonoids (3.72 ± 0.03 mg/g). The hydroethanolic extract presented relevant antioxidant, anti-inflammatory, cytotoxic and antibacterial potentials, being all these assays carried out using in vitro cell models. In this regard, the antiproliferative potential of walnut green husks hydroethanolic extract was assessed for the first time for the selected cell cultures (MCF-7, NCI-H460, HeLa, HepG2 and PLP2). Moreover, it was also the first time that the anti-inflammatory properties of the extract were studied using a specific cell model (RAW264.7). Thus, this study contributes to increase the knowledge about J. regia walnut green husks since, as far as we know, it was the first time that this set of combined results are presented, showing in general consistent results with the few studies available in literature.

97

3. Extractions with alternative solvents

Extractions with alternative solvents CHAPTER 3

3.1. Enhanced extraction of phenolic compounds using choline chloride based deep eutectic solvents from Juglans regia L. Vieira, V., Prieto, M. A., Barros, L., Coutinho, J.A.P., Ferreira, I.C.F.R., Ferreira, O., 2018. Ind. Crops. Prod. 115, 261-271.4

Abstract

The extraction of phenolic compounds from walnut leaves (Juglans regia L.) was optimized using heat-assisted extraction and deep eutectic solvents based on choline chloride and carboxylic acids. A preliminary solvent screening was performed using a selected group of carboxylic acids as hydrogen bond donors, showing that the highest extraction yield of phenolic compounds was obtained using choline chloride mixtures with butyric or phenylpropionic acid at a mole ratio 1:2, with 20 % of water (w/w). The extraction conditions (time, temperature and water proportion) were then optimized by an experimental design, assisted by response surface methodology. To evaluate the response, the three most abundant compounds identified by HPLC (neochlorogenic acid, quercetin 3- O-glucoside and quercetin O-pentoside) were quantified. Additionally, the solid/liquid ratio effect at the optimal conditions, in dose-response format, was studied in view of its upscale, not showing any significant decrease until 140 g/L. The results here presented provide valuable information towards the design of a process in a pre-industrial form for the extraction of phenolic compounds from J. regia leaves using deep eutectic solvents.

Keywords: Phenolic compounds; Deep eutectic solvents; Heat-assisted extraction; Response Surface Methodology; Juglans regia L..

4 My contribution to this work was the measurement of all experimental data, the interpretation of the RSM results and writing of the manuscript.

101 Extractions with alternative solvents CHAPTER 3

3.1.1. Introduction Deep eutectic solvents (DES) were first introduced in 2003 by Abbott and co-workers (Abbott et al., 2003) reporting mixtures of urea and quaternary ammonium salts, in particular choline chloride (CC). Since then many other DES have been reported and their use proposed for a wide range of applications (Francisco et al., 2012; Smith et al., 2014). A DES is a mixture between two (or more) starting materials (hydrogen bond acceptor - HBA and hydrogen bond donor - HBD) where the eutectic temperature of the mixture is considerably lower than that of either of the constituents (Abbott et al., 2004). Some authors prefer to call them low transition temperature mixtures (Francisco et al., 2013; Jancheva et al., 2017).

The replacement of volatile and toxic organic solvents by greener and more performant solvents is one the most important challenges of our days for the chemical industry in general, and biorefineries in particular (Pena-Pereira and Namieśnik, 2014). Choi et al. (2011) introduced the DES as an alternative media to extract secondary metabolites, such as phenolic compounds, from natural matrices instead of conventional organic solvents. Nowadays, this idea is being expanded to create designer solvents, using several combinations of HBA and HBD, with tunable properties to selectively dissolve and extract natural and bioactive compounds from plants, oils or biomass, valorizing natural products or wastes as a source of valuable compounds (Dai et al., 2016; Nam et al., 2015; Paradiso et al., 2016). A very recent review on the application of deep eutectic solvents for the extraction of phenolic compounds can be found in literature, describing the possibility of using DES both as solvent and formulation media with potential cosmetics, pharmaceutical, or food applications (Ruesgas-Ramón et al., 2017).

This new generation of solvents entails several characteristics, which support their insertion in different industries. First, they are simple to prepare from cheap starting materials (Dai et al., 2016). Then, the HBA and HBD can be selected to be less toxic than organic volatile solvents and also biodegradable (Mbous et al., 2017a). Moreover, in general, these solvents have low volatility and flammability (Dai et al., 2016). The addition of water to DES, for extraction purposes, is a well-established procedure to reduce the viscosity of the solvents and improve the mass transfer of bioactive molecules from the

102 Extractions with alternative solvents CHAPTER 3 natural matrices (García et al., 2016; Gu et al., 2014). Regarding temperature, its increase leads to a lower viscosity and, consequently, to higher extraction yields (Dai et al., 2016; Jancheva et al., 2017).

In the present study, the performance of choline chloride (CC) based DES combined with carboxylic acids (CA) is investigated for the extraction of phytochemical compounds from the leaves of walnut trees. Walnut leaves were chosen as a plant model as they stand out as a significant source of bioactive compounds such as phenolic acids and flavonols (Santos et al., 2013; Vieira et al., 2017). These leaves are well known in traditional medicine due to their health benefits arising from their antioxidant, antitumor (Santos et al., 2013), antiproliferative (Carvalho et al., 2010), anti-inflammatory and antinociceptive (Erdemoglu et al., 2003) activities. A DES screening of mixtures of CC and CA was carried by heat- assisted extraction (HAE). Fifteen acids were selected to assess the effect of the alkyl chain length, the number of carboxylic acid groups, and the additional presence of hydroxyl and/or phenyl groups over the extraction performance. The best two DES selected will be used to optimize the extraction conditions of time (t), temperature (T) and water content (S) by response surface methodology (RSM). The work here presented can be better comprehended as a continuation of a previous work from other authors (Vieira et al., 2017), in which they optimized the extraction of bioactive compounds from walnut leaves using hydro-alcoholic mixtures.

3.1.2. Material and methods

3.1.2.1. Standards and reagents HPLC-grade acetonitrile and anhydrous citric acid were from Fisher Scientific (Lisbon, Portugal). The phenolic compounds standards 5-O-caffeoylquinc acid (Extrasynthèse, Genay, France) and quercetin 3-O-glucoside, as well as glutaric acid (99 wt%) and glycolic acid (99 wt%) were purchased from Sigma-Aldrich. Butyric acid (99 wt%), lactic acid (88-92 wt%) and valeric acid (99 wt%) were purchased from Riedel-de Haen. Acetic acid (99.5 wt%) was obtained from Labsolve JMGS, propionic acid (99 wt%) from Merck, DL-malic acid (99.5 wt%) from Panreac, malonic acid (98 wt%) from Fluka, phenyl-acetic acid (99 wt%) from

103 Extractions with alternative solvents CHAPTER 3

Alfa Aesar. Choline chloride (98 wt%), 3-phenylpropionic acid (99 wt%), 4-phenylbutyric acid (99 wt%) and 5-phenylvaleric acid (99 wt%) from ACROS Organics. All the other chemicals were of analytical grade and purchased from common sources.

3.1.2.2. Plant material Juglans regia L. (walnut) dried leaves were purchased from Soria Natural, S.A., Spain. According to the distributor, the leaves were collected in Soria (Spain) in June 2014 and naturally dried in a room with controlled humidity. The samples were reduced to a fine dried powder (60 to 20 mesh) and stored in a desiccator protected from light for subsequent assays.

3.1.2.3. Screening analysis of DES

Preparation of DES DES were prepared using the heating-stirring method proposed by Abbot and co-workers (Abbott et al., 2004) with slight modifications. Firstly, the water amount present in the starting materials was measured using a Metrohm 831 Karl Fisher coulometer (data not shown). Then, each component was accurately weighed (±10-4 g) to a round-bottom flask. The mixtures were stirred in a water bath (from 50 to 80 °C) at 600 rpm until a homogenous solvent was obtained. The resulting DES are listed in Table 3.1.

Conditions for the screening extraction by heat-assisted extraction For the screening extractions, DES were prepared with 20 % (w/w) of water except for phenylbutyric and phenylvaleric acids that were prepared with 5 % (w/w) of water due to their low solubility in water. For comparison purposes, the DES CC:phenylpropionic acid was also prepared with 5 % of water.

The solid-liquid heat assisted extractions (HAE) were performed using a Carousel 12 Plus Reaction Station™ (Radleys Tech). This equipment allows stirring and temperature control within ±0.5 °C, with protection from light. It is also coupled to a condensation system, avoiding the loss of solvent. The powdered samples (0.15 g) were extracted with 5 mL of each solvent, during 60 min at 50 °C and 600 rpm. A control extraction using the

104 Extractions with alternative solvents CHAPTER 3 conventional solvent ethanol:water (1:1, v/v) was also performed. After extraction, the mixtures were centrifuged at 6000 rpm during 10 min at room temperature and the supernatant was filtered through a Whatman n°4 for further analysis. For the extracts obtained with choline chloride:aromatic acids, a previous dilution was made with methanol due to thermal and physical stability.

3.1.2.4. Chromatographic analysis of the main phenolic compounds Extract solutions were two-fold diluted with water and filtered through 0.2 µm disposable LC filter disks (30 mm, regenerated cellulose). The DES composed by CC and aromatic acids were diluted with methanol to avoid precipitation at room temperature. The samples were analysed as previously applied (Vieira et al., 2017) using a Shimadzu 20A series UFLC (Shimadzu Corporation, Kyoto, Japan) with a quaternary pump and a photodiode array detector (PDA) coupled to an LC solution software data-processing station using a Waters

Spherisorb S3 ODS-2 C18, (3 μm, 4.6 mm × 150 mm) column operating at 35 °C for separation. Double online detection was carried with a diode array detector (DAD) operating at 280 and 370 nm as preferred wavelengths and the target phenolic compounds were identified according to their UV spectra and retention time (Barros et al., 2013). The results were expressed in mg per g of dry weight (mg/g dw).

3.1.2.5. HAE optimization by RSM of the selected DES, experimental design, model analysis and statistical evaluation An illustrative diagram of the different steps carried out to obtain an optimal phenolic extract from J. regia L. is presented in Figure B1 (Appendix B). The solid-liquid extractions were performed with the HAE using the same equipment above described for the DES screening.

Experimental design The study of the impact of all independent variables was carried using one-factor-at-a-time, to identify the most influent, and to determine the initial range of the processing variables.

Through the analysis of this experimental results (data not shown), X1 (t, min), X2 (T, in °C) and X3 (S, in %) were chosen as variables for the RSM design. Therefore, the combined

105 Extractions with alternative solvents CHAPTER 3 effect of these variables on the extraction of the three main phenolic compounds present in J. regia (maximizing responses individually or globally) was studied using circumscribed central composite design (CCCD) with 5 levels of each factor. The experimental design is based on 20 independent combinations, 6 of which are replicas at the central point of the experiment Box & Hunter (1957). The mathematical expressions used to calculate the design distribution, code and decode the tested variables can be found all detailed in the supplemental section (Table B1, Appendix B). Once the optimal conditions (X1, X2 and X3) were optimized, the study was advanced furthermore with the study of the S/L condition

(X4, in g/L).

Mathematical model The response surface models were fitted by means of least-squares calculations using the following second-order polynomial equation:

nnnn −1 Ybb=+++ Xb X Xb X 2 0 iiijijiii [2.1] iiji====1121 ji

In this equation, Y represents the dependent variable (response variable) to be modelled, the independent variables are Xi and Xj, b0 is the constant coefficient, bi the coefficient of linear effect, bij the coefficient of interaction effect, bii the coefficients of quadratic effect and n is the number of variables. The responses used for evaluating the extraction process were: 1) the main HPLC compounds identified of as the phenolic acid 3-O-caffeoylquinic acid (P1) and two flavonols, quercetin 3-O-glucoside (P2), and quercetin O-pentoside (P3); and 2) the total HPLC compounds determined (P1+P2+P3).

Procedure to optimize the variables to a maximum response A simplex method was used to maximize the model process by solving non-linear problems to optimize the response criteria selected (P1, P2, P3 and total HPLC) (Heleno et al., 2016; Pinela et al., 2016). To avoid unnatural conditions of the variables, certain limitations were imposed (i.e., times lower than 0).

106 Extractions with alternative solvents CHAPTER 3

Fitting procedures and statistical analysis The statistical analysis and fitting was obtained using a Microsoft Excel spreadsheet in three phases as explained in detail previously (Vieira et al., 2017). Briefly: 1) Coefficients measurement was achieved using the nonlinear least-square (quasi-Newton) in Microsoft Excel (Kemmer and Keller, 2010); 2) Coefficients significance was obtained via ‘SolverAid’ (Prieto et al., 2014); and 3) Model reliability was confirmed by applying Fisher F-test (α = 0.05) (Pinela et al., 2017) and the ‘SolverStat’ macro was used to make assessment of parameter and model prediction uncertainties (Prieto et al., 2015).

3.1.3. Results and discussion

3.1.3.1. DES screening to investigate the effect of CA molecular structure on the extraction of phytochemical compounds Phenolic compounds are complex molecules, and their extraction from a solid matrix requires compatible solvents. The responses used to evaluate the efficiency of the extraction were the specific quantification of the phenolic compounds performed by HPLC- DAD (P1, P2 and P3) as described before (Vieira et al., 2017). The phenolic profile of J. regia leaves extracts is in good agreement with the previously reported by Santos et al. (2013) (Figure A2, Appendix A). These compounds were identified by comparison of their UV spectra and retention times with those obtained in previous findings, as well as with the commercial standards.

In this work, DES formed by CC and CA were evaluated to investigate the effect of the acid’s molecular structure on the extraction of compounds from J. regia L. To study the effect of the chemical structure of the hydrogen bond donor upon the extraction efficiency of the DES, different acid groups were chosen: linear monocarboxylic acids (acetic acid, propionic acid, butyric acid and valeric acid), aromatic acids (phenylacetic, 3-phenylpropionic, 4- phenylbutyric acid and 5-phenylvaleric acids), dicarboxylic acids (malonic and glutaric acids), hydroxy monocarboxylic acids (glycolic acid and lactic acid), one hydroxy dicarboxylic acid (malic acid) and one hydroxy tricarboxylic acid (citric acid). The CA:CC mole ratio was chosen according to the eutectic composition found by Abbott et al. (2004): 2:1

107 Extractions with alternative solvents CHAPTER 3 for monocarboxylic acids (phenylacetic and 3-phenylpropionic acids); 1:1 for dicarboxylic acids (malonic acid), and 1:2 for tricarboxylic acids (citric acid). To decrease the systems viscosity, a given amount of water (20 %, w/w) was added to the DES. For the systems CC:phenylbutyric acid and CC:phenylvaleric acid, only 5 % w/w water was added because of miscibility issues. For comparison purposes, CC:phenylpropionic acid was studied at 5 % and 20 % water content. In this screening, the following extraction conditions were adopted: 1 hr extraction time, 50 °C and 600 rpm stirring speed. In addition, water, pure ethanol and ethanol:water (1:1, v/v) were also used as extraction solvents at the same conditions, for comparison purposes.

108 Extractions with alternative solvents CHAPTER 3

Table 3.1. Composition, proportions, properties and responses of the set of DES used in the initial screening for the selection of the HBD. Solvent features Responses

Composition Proportions HPLC-DAD C1:C2 C3 P1 P2 P3 TOTAL C1 C2 C3 (mol:mol) (%, w/w) (mg/g dw) (mg/g dw) (mg/g dw) (mg/g dw) Reference solvents -- -- H2O -- 100 5.4 1.3 2.1 8.8 -- EtOH ------0.4 4.83 5.0 10.3 -- EtOH H2O -- 50 4.9 11.3 8.6 24.9 Deep eutectic solvents (DES) CC Acetic acid H2O (1:2) 20 4.9 13.1 10.3 28.2 CC Propionic acid H2O (1:2) 20 7.8 13.8 11.7 33.3 CC Butyric acid H2O (1:2) 20 7.3 14.4 12.0 33.7 CC Valeric acid H2O (1:2) 20 5.4 11.0 9.5 25.9 CC Glycolic acid H2O (1:2) 20 6.2 10.1 6.3 22.6 CC Lactic acid H2O (1:2) 20 5.7 9.9 6.8 22.3 CC Phenylacetic acid H2O (1:2) 20 4.9 14.7 11.6 31.1 CC 3-Phenylpropionic acid H2O (1:2) 20 5.4 15.4 13.4 34.3 CC Malic acid H2O (1:1) 20 5.0 8.6 5.6 19.2 CC Malonic acid H2O (1:1) 20 5.6 10.5 5.3 21.2 CC Glutaric H2O (1:1) 20 4.4 9.7 7.0 21.2 CC Citric acid H2O (2:1) 20 4.4 6.5 4.1 14.9 CC 3-Phenylpropionic acid H2O (1:2) 5 2.5 6.3 5.5 14.3 CC 4-Phenylbutyric acid H2O (1:2) 5 3.8 8.3 6.7 18.8 CC 5-Phenylvaleric acid H2O (1:2) 5 2.1 6.5 5.2 13.8 Components composition: C1 (HBA component), C2 (HBD component) and C3 (water component). Solvent properties: number of carbons (C), hydroxyl group (-HO), carboxyl group (-COOH) and phenyl group (Ph). Responses: HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (TOTAL, P1+P2+P3); n.d.: not determined.

109 Extractions with alternative solvents CHAPTER 3

40

35

30

) dw

25

20

Extraction yield (mg/gyield Extraction 15

10

5

0

Neochlorogenic acid Quercetin 3- O -glucoside Quercetin O -pentoside

Figure 3.1. Extraction yield (mg/g dw) of 3-O-caffeoylquinic acid (P1), quercetin 3-O-glucose (P2) and quercetin O-pentoside (P3) obtained in the initial screening of the choline chloride based DES described in Table 3.1, with the extraction conditions: 1 hr extraction time, 50 °C, 600 rpm and 20 % (w/w) of water (solvents with * only contain 5 % (w/w) of water).

Figure 3.1 shows the results obtained for the extraction yield using the selected set of DES and reference solvents. Detailed data are reported in Table 3.1. The following conclusions were derived:

1) The total amount of phenolic compounds detected (P1+P2+P3) for the reference solvents, at the screening extraction conditions, were 8.8 ± 0.1 mg/g dw for water, 10.3 ± 0.01 mg/g dw for ethanol and 24.9 ± 0.1 mg/g dw for ethanol:water (1:1, v/v). The phenolic profile showed significant differences depending on the solvent applied. Water extracted mainly the polar compound P1 (3-O-caffeoylquinic acid) and only traces of the quercetin derivatives (P2 and P3). Meanwhile, when pure ethanol was used, only a small fraction of P1 is detected (~0.4 mg/g) favoring the extraction of less polar compounds (P2 and P3). These results were in accordance

110 Extractions with alternative solvents CHAPTER 3

to the results previously reported by Vieira et al. (2017), in which the mixture ethanol/water was the most adequate combination. 2) For DES with 20 % water content, monocarboxylic acids (aliphatic or aromatic) led to higher extraction yields (phenylpropionic > butyric > propionic > phenylacetic > acetic > valeric > glycolic > lactic acid), followed by dicarboxylic acids (malonic, glutaric and malic acid) and the tricarboxylic acid (citric acid). For the solvents prepared with only 5 % of water, in general, lower extraction yields are obtained. However, an increase is observed going from phenylpropionic acid to phenylbutyric acid, followed by a decrease when using phenylvaleric acid. 3) Regarding the structural differences, some tendencies can be observed by looking at the solvents containing 20 % of water. The addition of a hydroxyl group (-OH) to the acid structure, leads to lower extraction yields (e.g. pairs glycolic/acetic acids or lactic/propionic acids). Regarding the alkyl chain length of the linear monocarboxylic acids, the increase until four carbon atoms seems to be favorable. The presence of a phenyl group also seems to promote a slight increase of the extraction yield (see phenylacetic/acetic acids or phenylpropanoic/propanoic acids). Overall, for the acids based DES, the higher the number of -OH and -COOH groups present in the solvent, the lower the extraction yield, being the target molecules preferentially extracted by acids with three/four carbons in its linear alkyl chain. 4) Analyzing the individual extraction response of the main phenolic components determined by HPLC (P1, P2 and P3), additional conclusions can be drawn. The component P1 was preferentially extracted by CC:propionic acid (7.84 ± 0.04 mg/g dw) and CC:butyric acid (7.30 ± 0.03 mg/g dw). Nevertheless, P2 and P3 are preferentially extracted by DES based on CC and aromatic acids reaching 15.44 ± 0.01 mg/g dw, and 13.37±0.13 mg/g dw, respectively, using CC:phenylpropionic acid. Good results are also obtained for CC:butyric acid and CC:phenylacetic acid. 5) All effects combined, the solvents that presented the higher extraction yields were the mixtures of CC combined with phenylpropionic (with 20 % water) (34.3 ± 0.2

111 Extractions with alternative solvents CHAPTER 3

mg/g dw) or butyric acid (33.7 ± 0.4 mg/g dw), closely followed by CC and propionic acid (33.3 ± 0.12 mg/g dw). The mixtures of CC with phenylpropionic acid (PPA) or butyric acid (BA) were then selected for further optimization.

3.1.3.2. Response criteria for the RSM analysis and statistical verification The extraction of target compounds from matrices requires a specific consideration due to the intrinsic nature and stability of these target compounds. Therefore, to maximize the response extraction of these compounds, it is essential to identify the response effects caused by the main variables involved using the minimum time, energy and solvent consumption and designing the most cost-effective and profitable extraction system (Dai and Mumper, 2010). The RSM design allows optimizing all the variables simultaneously considering interactive reactions and to predict the most efficient conditions. This is achieved by using second order polynomial models with interactions, that are able to describe and maximize the response criteria selected, based on the experimental range tested (Bezerra et al., 2008; Ferreira et al., 2007; Kalil and Maugeri, 2000).

The RSM experiment was designed based on the preliminary tests described above, using five levels of extraction t, T and S as independent variables to efficiently optimize the HAE process. The coded values and their natural values are presented in Table B1 (Appendix B). Once the optimal conditions (t, T and S) were optimized, the study was further advanced towards the study of the effects of the S/L.

Experimental conditions of BA and PPA, and chromatographic response results (HPLC-PDA) are presented in Table 3.2 for the CCCD RSM design. The quantification results obtained from the chromatographic methodology were used as response criteria to optimize the solvents conditions by RSM. The obtained parametric fitting values, confidence intervals and statistical information are presented in Table 3.3.

112 Extractions with alternative solvents CHAPTER 3

Table 3.2. Results of the response surface experimental plan for the BA and PPA optimization process of independent variables of time (t), temperature (T), and water proportion (S). Response criteria comprise the HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (TOTAL, P1+P2+P3). Experimental design CC:BA CC:PPA

Coded values CC:BA CC:PPA HPLC content HPLC content

X1: t X2: T X3: S X1: t X2: T X3: S P1 P2 P3 TOTAL P1 P2 P3 TOTAL X1 X2 X3 min °C % °C w % mg/g dw mg/g dw mg/g dw mg/g dw mg/g dw mg/g dw mg/g dw mg/g dw 1 -1 -1 -1 60.4 42.2 20.3 60.4 42.2 4.1 5.0 12.1 10.5 27.6 1.2 3.6 3.1 7.9 2 1 -1 -1 149.6 42.2 20.3 149.6 42.2 4.1 6.0 14.0 12.0 32.0 2.3 7.2 6.2 15.7 3 -1 1 -1 60.4 77.9 20.3 60.4 77.9 4.1 6.3 14.4 12.5 33.2 2.3 7.1 5.8 15.3 4 1 1 -1 149.6 77.9 20.3 149.6 77.9 4.1 5.9 13.5 11.2 30.6 3.2 10.2 7.1 20.4 5 -1 -1 1 60.4 42.2 79.7 60.4 42.2 16 6.5 11.3 9.0 26.8 4.7 14.2 12.1 31.0 6 1 -1 1 149.6 42.2 79.7 149.6 42.2 16 6.6 11.3 9.0 26.9 4.8 14.8 12.9 32.5 7 -1 1 1 60.4 77.9 79.7 60.4 77.9 16 7.2 12.1 9.6 28.9 4.5 14.2 11.1 29.8 8 1 1 1 149.6 77.9 79.7 149.6 77.9 16 6.5 11.0 8.4 25.9 4.6 13.8 9.3 27.6 9 -1.68 0 0 30 60 50 30 60 10 6.4 14.1 12.2 32.8 3.0 9.5 8.2 20.8 10 1.68 0 0 180 60 50 180 60 10 6.8 15.0 12.6 34.4 4.7 14.9 12.3 31.9 11 0 -1.68 0 105 30 50 105 30 10 6.6 15.2 13.1 35.0 2.7 8.4 7.4 18.5 12 0 1.68 0 105 90 50 105 90 10 6.6 14.2 10.3 31.1 3.9 12.4 7.8 24.0 13 0 0 -1.68 105 60 0 105 60 0 2.6 2.6 6.7 2.6 1.1 4.0 3.5 8.6 14 0 0 1.68 105 60 100 105 60 20 7.2 5.6 5.5 18.2 5.0 15.8 13.1 34.0 15 0 0 0 105 60 50 105 60 10 6.6 14.6 12.5 33.6 3.8 12.2 10.3 26.3 16 0 0 0 105 60 50 105 60 10 6.4 14.3 12.3 32.9 4.0 13.1 11.1 28.3 17 0 0 0 105 60 50 105 60 10 7.3 16.0 13.6 36.9 4.0 12.7 10.8 27.5 18 0 0 0 105 60 50 105 60 10 6.4 14.1 12.3 32.8 3.5 11.6 9.8 25.0 19 0 0 0 105 60 50 105 60 10 6.8 15.3 13.0 35.1 3.6 11.7 9.7 25.0 20 0 0 0 105 60 50 105 60 10 6.5 14.8 12.5 33.8 3.7 11.8 9.9 25.4

113 Extractions with alternative solvents CHAPTER 3

Table 3.3. Estimated coefficient values obtained from the Box polynomial model, parametric intervals and numerical statistical criteria for each parametric response criteria of the extractions systems tested (BA and PPA). Response criteria comprise the HPLC quantification of 3-O- caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (P1+P2+P3).

HPLC content

Parameters P1 P2 P3 Total

CC:BA

Intercept b0 6.71 ±0.18 15.17 ±0.18 12.69 ±0.44 34.52 ±0.63 b1 ns ns ns ns Linear effect b2 0.16 ±0.03 ns -0.32 ±0.09 ns b3 0.77 ±0.17 ns -0.75 ±0.09 1.47 ±0.20 b11 ns ns 0.09 ±0.07 0.26 ±0.20 Quadratic effect b22 ns ns -0.40 ±0.09 ns

b33 -0.62 ±0.16 -3.32 ±0.18 -2.17 ±0.09 -7.23 ±0.20 b12 -0.13 ±0.02 -0.30 ±0.18 ns -0.91 ±0.20 Interactive effect b13 ns -0.24 ±0.18 ns ns b23 ns ns ns ns

Statistics (R²) 0.8340 0.8445 0.8576 0.9229

CC:PPA

Intercept b0 3.80 ±0.15 12.03 ±0.34 10.22 ±0.25 26.18 ±0.60 b1 0.37 ±0.14 1.23 ±0.26 ns 2.34 ±0.47 Linear effect b2 0.30 ±0.14 1.00 ±0.26 -0.32 ±0.09 1.46 ±0.47 b3 1.27 ±0.14 3.42 ±0.26 -0.75 ±0.09 7.07 ±0.47 b11 ns ns 0.09 ±0.07 ns Quadratic effect b22 -0.36 ±0.13 -0.39 ±0.26 -0.40 ±0.09 -1.43 ±0.45

b33 ns -0.95 ±0.26 -2.17 ±0.09 -2.43 ±0.45 b12 ns ns ns -0.97 ±0.61 Interactive effect b13 -0.19 ±0.18 -0.62 ±0.35 ns -1.50 ±0.61 b23 -0.19 ±0.18 -0.61 ±0.35 ns -1.29 ±0.61

Statistics (R²) 0.9407 0.9609 0.9671 0.9491

Theoretical response surface models predicted The RSM mathematical models were obtained by fitting each individual response (experimental values in Table 3.2) to the second-order polynomial model of Eq. [2.1] using nonlinear least-squares estimations. The resulting models for each assessed extraction technique are the following, for the CC:BA system:

BA 2 [3.1] YP1 =6.71 + 0.16 T + 0.77 S − 0.62 S − 0.13 tT

114 Extractions with alternative solvents CHAPTER 3

BA 2 [3.2] YStTtSP2 =−−−15.173.320.200.30

BA 222 [3.3] YTStTSP3 =−−+−−12.690.320.750.090.402.17

BA 22 [3.4] YStStTPPP123++ =++−−34.521.470.267.230.91

And for the CC:PPA system:

PPA 2 [3.5] YtTSTtSTSP1 =+++−−−3.800.370.301.270.360.190.19

PPA 22 [3.6] YP2 =12.03 + 1.23 t + 1.00 T + 3.42 S − 0.39 T − 0.95 S − 0.62 tT − 0.61 TS

PPA 22 [3.7] YtTStTP3 =−−+−−10.220.320.750.090.402.17

PPA 22 [3.8] YtTSTStTtSTSPPP123++ =+++−−−−−26.182.341.467.071.432.430.971.501.29

These equations translate the response patterns for individual measurement of phenolic compounds and Table 3.3 shows the complexity of possible sceneries. Not all the parameters of Eq. [2.1] were used for building the model since some coefficients were non- significant (ns). The significant parametric coefficients were tested with intervals of confidence at the 95 % level (α = 0.05) and the correlation coefficients were always higher than 0.87.

The models presented in Eq. [3.1] to [3.8] are workable showing a good agreement between the experimental and predicted values, which proves that they can be used for the prediction and optimization stages. Additionally, all independent variables used show significant variations, which means that they are needed for complete understanding of the behavior of responses. The significant parametric values found in those models cannot be linked with physical or chemical significance, but they are valuable to predict the results of untested operation conditions (Ranic et al., 2014). The sign of the parametric value determines part of the response; for positive effects, the response is higher at the high level and when a factor has a negative effect, the response is lower at high level. Additionally, the higher of the parametric value, the more significant the weight of the governing variable is.

115 Extractions with alternative solvents CHAPTER 3

Certain features of the parametric values of the variables display similar effects in both DES tested: 1) the relevance of the significant parametric values can be ranked as a function of the variables involved in the following decreasing form as S>>T>t; 2) Linear, quadratic and interactive parametric effects of the equations developed play a relevant role on the responses achieved; and 3) All the interactive parametric values show a negative effect, which reveals that in all of them the increase of the interacting variables reduces the expected response (antagonistic action).

However, other characteristics are more specific for each of the solvents used: 1) For CC:PPA the linear effect of the variables T and S show strong values, meanwhile the effect of t is negligible in almost all response criteria, but for CC:BA the linear effect of t increases, reaching nearly the relevance of the variable T. 2) The quadratic or non-linear effect shows a strong effect in the S variable for CC:BA, meanwhile in the CC:PPA the strongest effect is registered in the T variable; and 3) Regarding the interactive effects, for the CC:PPA the interactions of the variable t with the other variables (tT and tS) showed a minor relevance, meanwhile for the CC:BA the tT interaction showed a strong significance in the description of the behavior of almost all responses.

Overall effects of the extraction variables on the target responses Although parametric results can depict the patterns of the responses, 3D and 2D graphical representations support their understanding. Figure 3.2 shows the graphical results in terms of the response surfaces of the total compounds content in mg/g dw determined by HPLC (P1+P2+P3) for both DES. Each column is divided into two subsections (A and B). Part A shows the joint graphical 3D analysis as a function of each of the variables involved. Each net surface represents the theoretical three-dimensional response surface predicted with the second order polynomial of Eqs. [3.4] and [3.8], respectively. The statistical design and results are described in Table 3.2. Estimated parametric values are shown in Table 3.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 3.4). Part B illustrates the goodness of fit, using two basic graphical statistic criteria. The first one, the ability to simulate the changes of the response between the predicted and observed data; and the second one,

116 Extractions with alternative solvents CHAPTER 3 the residual distribution as a function of each of the variables. Note all the differences in the axes scales.

In similar terms, Figure 3.3 illustrates the results (mg/g dw) for the identified HPLC peaks of P1 (3-O-caffeoylquinic acid content), P2 (quercetin 3-O-glucoside content) and P3 (quercetin O-pentoside content). Figure 3.3 illustrates a 3D response surface plots predicted with their respective second order polynomial equation described by Eq. [2.1]. Each of the net surfaces, representing the theoretical three-dimensional response surface, were built with Eqs. [3.1], [3.2] and [3.3] for the CC:BA system and Eqs.[3.5], [3.6] and [3.7] for CC:PPA. The subsection related with the statistical results of the individual responses is eluded in those figures, but estimated parametric values, confidence of intervals and statistical coefficients of the fitting procedure performed are shown in Table 3.3.

Table 3.4. Operating optimal conditions of the variables involved that maximize the response values for RSM using a CCCD for each of the extracting techniques assessed (BA and PPA), for the three individual response value formats of each compound and for all the compounds. Response criteria comprise the following: 1) HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2), quercetin O-pentoside content (P3) and total HPLC content (TOTAL, P1+P2+P3).

Optimal variable conditions Optimum Criteria (mg/g dw) X1: t (min) X2: T (°C) X3: S (%)

Individual optimal variable conditions for 3-O-caffeoylquinic acid (P1):

CC:BA *180.0 *30.0 68.6 7.0±1.4 mg/g dw CC:PPA *180.0 59.5 *20.0 6.0±3.9 mg/g dw

Individual optimal variable conditions for the quercetin 3-O-glucoside (P2):

CC:BA *180.0 *30.0 49.3 16.0±4.2 mg/g dw CC:PPA *180.0 67.5 16.6 15.6±5.1 mg/g dw

Individual optimal variable conditions for the quercetin O-pentoside (P3):

CC:BA *180.0 52.7 44.9 13.1±3.1 mg/g dw CC:PPA *30.0 53.8 *20.0 14.8±4.9 mg/g dw

Global optimal variable conditions for the total HPLC compounds (P1+P2+P3):

CC:BA *180.0 *30.0 53.0 37.9±4.0 mg/g dw CC:PPA *30.0 55.7 *20.0 31.7±4.2 mg/g dw

117 Extractions with alternative solvents CHAPTER 3

Observing all the plots in Figure 3.2 it is possible to verify that the amount of extracted material increases to an optimum value and, in most cases, decreases as a function of each assessed independent variable. Similar behavior can be found in Figure 3.3, and in consequence in almost all combinations the optimum value can be found in single points, which allows computing the conditions that lead to an absolute maximum.

118 Extractions with alternative solvents CHAPTER 3

BUTYRIC ACID PHENYLPROPIONIC ACID

1

Figure 3.2. Response surfaces of the total compounds content determined by HPLC (P1+P2+P3) for the BA and PPA systems optimization. Part A: Shows the joint graphical 3D analysis as a function of each of the variables involved. Each of the net surfaces represents the theoretical three- dimensional response surface predicted with the second order polynomial of Eqs. [3.4] and [3.8], respectively. The statistical design and results are described in Table 3.2. Estimated parametric values are shown in Table 3.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 3.4). Part B: To illustrate the goodness of fit, two basic graphical statistic criteria are used. The first one, the ability to simulate the changes of the response between the predicted and observed data; and the second one, the residual distribution as a function of each of the variables.

119 Extractions with alternative solvents CHAPTER 3

Figure 3.3. Joint graphical 3D analysis as a function of each the variables involved for BA and PPA systems for the HPLC quantification of 3-O-caffeoylquinic acid content (P1), quercetin 3-O-glucoside content (P2) and quercetin O-pentoside content (P3).

120 Extractions with alternative solvents CHAPTER 3

Numerical individual and global optimal conditions that maximize the extraction, statistical analysis and experimental verification of predictive models The performed optimization of the HAE response by RSM provides a strong solution that minimizes the errors with a short number of experimental trials as has been demonstrated elsewhere (Roselló-Soto et al., 2015; Wong et al., 2015). The multivariable fitting decreases the number of parameters needed to analyze the response leading to better estimations, reducing their interval of confidence and allowing predicting the response behavior. In addition, by applying a simple procedure (considering restrictions to the experimental ranges) the individual optimal conditions are found, as well as the maximal response values (Table 3.4) in relative (marked with * when the optimal value may be outside of the experimental range studied) or absolute optimal conditions were found for all the responses.

Finally, Figure 3.4 simplifies the interpretation of the influence of the independent variables in a bi-dimensional representation of the optimal conditions of the RSM variables. The line represents the variable response pattern when the others are located at the optimal values, the operating conditions that maximize the extraction conditions are represented with dots (); these singular conditions are represented in Table 3.4 to clarify not only the optimal independent variables conditions but also the optimal values predicted. The conditions that lead to the optimal values were re-checked to ensure the accuracy of the presented results. Generally, CC:BA gave significantly higher values than CC:PPA probably due to the lack of solubility of these compounds in the water that can be incorporated into CC:PPA.

121 Extractions with alternative solvents CHAPTER 3

Figure 3.4. Summary of the effects of all variables assessed for BA and PPA systems, showing the individual 2D responses of all studied responses as a function of all the variables assessed. The variables in each of the 2D graphs were positioned at the optimal values of the others (Table 3.4). The dots () presented alongside each line highlight the location of the optimum value. Lines and dots are generated by the theoretical second order polynomial models of Eqs. [3.1]-[3.4] for CC:BA and [3.5]-[3.8] for CC:PPA.

122 Extractions with alternative solvents CHAPTER 3

3.1.3.3. Dose-response analysis of the solid-to-liquid effect at the optimum conditions The effect of the variable S/L was studied in a dose-response format for all the HAE responses at the optimal global conditions predicted for the BA and PPA solvents by the polynomial Eqs. [3.1] to [3.8] (Table 3.4). Preliminary results performed, showed that the experimental limit value at lab-scale was close to 140 g/L. Therefore, in both techniques an experimental dose-response test was designed to evaluate the S/L behavior between 5 g/L to 140 g/L using ten different concentrations (data not shown). In all cases, the S/L effect can be described by a simple linear relationship with an intercept and in all cases the linear relation shows decreasing patterns as a function of the increase of S/L. However, in all cases, the decreasing patterns explained by the parametric coefficient of the slope were non-significant at confidence interval level of 95 % (α = 0.05) and the decreasing effect was rejected for further analysis. Although the conclusions of the analysis are excluded, in global terms it can be postulated that the increase of the S/L has a slight effect on the phenolic compound extraction and no saturation effects seem to appear at least up to 140 g/L.

3.1.3.4. Comparison with other responses reported using non-DES To the best of our knowledge there are no previous works that have used DES as alternative solvents for the extraction of these compounds. The solid-liquid extraction of the compounds here studied is usually carried out using organic solvents such as methanol, ethanol, acetone and ethyl acetate (Kerton, 2009). In particular, the most common solvents used by researchers are aqueous methanol and ethanol (Bravo and Mateos, 2008). In the scientific literature, there are several techniques that also describe the extraction of these compounds with non-DES solvents, in which the relevant ones are summarized hereafter:

- Compound P1 is found in small amounts in several natural matrices. Using the same plant source (J. regia leaves) authors have found similar amounts as those here described (7.0 and 6.0 mg/g dw for CC:BA and CC:PPA, respectively). As example, Santos et al. (2013) using hydro-alcoholic mixtures found 6.41 mg/g dw, meanwhile

123 Extractions with alternative solvents CHAPTER 3

Vieira et al. (2017) achieved a maximum of 10.7 mg/g dw. In other natural sources such as in Vaccinium floribundum Kunth berries were found 0.015 mg/g dw (Prencipe et al., 2014), 0.76 and 0.89 mg/g dw in Lonicera japonica Thunb. flowers (Li et al., 2014) and 17.2 mg/g dw on Solanum betaceum skin extract (Orqueda et al., 2017).

- For compound P2 the quantities recovered using J. regia leaves by other authors (Vieira et al., 2017) (13.2 and 15.2 mg/g dw, using hydro-alcoholic mixtures) are slightly lower than those found here using DES (16.0 and 15.6 mg/g dw for CC:BA and CC:PPA, respectively). In other material sources, higher values are reported such as 33.40 mg/g dw in Securigera securidaca flowers (Ibrahim et al., 2015) and 36.7 mg/g dw in Dasiphora fruticosa (L.) Rydb. (Tomczyk et al., 2012).

- For compound P3 the quantities recovered using J. regia leaves by other authors (Vieira et al., 2017) (12.7 and 15.1 mg/g dw, using water/ethanol mixtures) are similar to those found here using DES (13.1 and 14.8 mg/g dw for CC:BA and CC:PPA, respectively). Lower amounts were recovered from other material sources such as 0.0133 mg/g dw on Rosa canina hips (Cunja et al., 2015) and 0.101 mg/g fw in V. floribundum berries (Prencipe et al., 2014).

Compared to more traditional water/ethanol reference solvent, similar or higher extraction yields were obtained using the selected DES.

This work provides valuable insights on the use of DES formed between choline chloride and carboxylic acids, not only as extracting solvents, but also as potential formulation media, avoiding an intermediate recovery step of the bioactive compounds from the solvent. In this regard, the design of the final composition of the solvent should take into consideration the compatibility issues with the desired final application.

3.1.4. Conclusions The combined effects of three independent variables were studied to maximize the intrinsic responses on the J. regia chemical composition. A CCCD of three factors with five- levels of 16 combinations and 6 replicates at the center of the experimental domain was

124 Extractions with alternative solvents CHAPTER 3 successfully applied for the solvent optimization by RSM. Polynomial responses were designed and experimentally verified on 3-O-caffeoylquinic acid (P1), quercetin 3-O- glucoside (P2), quercetin O-pentoside (P3) and total HPLC (sum of the most abundant compounds: P1, P2 and P3) extraction. Comparatively, for all the assessed responses, CC:BA results in higher extraction yields than CC:PPA.

The water content was clearly the most relevant extraction variable, followed by temperature and, lastly, extraction time. Considering the study of the solid/liquid ratio effect, an important variable from an industrial point of view, no significant decreasing patterns were observed for any of the studied responses in the 5 to 120 g/L range. Because of the increased demand for industrial production of bioactive compounds, the effect of each variable on the optimum value is a practical information. Usually, laboratory scale up units are slightly less efficient than those obtained at industrial level. However, the profiles of the physical-chemical variables involved remain and therefore, can be used as a valuable illustration to transfer the results found at a lab-scale to an industrial environment. Other variables can be included to the models developed to reduce the industrial costs (i.e. energy, materials and time consumption) highly influencing the profitable viability of the process.

Therefore, this work studied the effect of the acid’s molecular structure (as hydrogen bond donor in a CC based DES) on the extraction of phytochemical compounds from J. regia L., followed by the optimization of the extraction conditions by RSM using HAE for the best DES.

125 Extractions with alternative solvents CHAPTER 3

3.2. Insights on the extraction performance of alkanediols and glycerol: using Juglans regia L. leaves as a source of bioactive compounds Vieira, V., Prieto, M. A., Barros, L., Calhelha, R.C., Coutinho, J.A.P., Ferreira, I.C.F.R., Ferreira, O., 2018. Under preparation.5

Abstract

Glycerol and alkanediols are being studied as alternative solvents to extract phytochemicals from plant material, often as hydrogen bond donors in natural deep eutectic solvents (NADES). Many of those alcohols are liquid at room temperature but their direct use as solvent for extraction purposes is scarce. In this work, glycerol and a series of alkanediols (1,2-ethanediol, 1,2-propanediol, 1,3-propanediol, 1,3-butanediol, 1,2-pentanediol, 1,5- pentanediol and 1,2-hexanediol) were used for the extraction of phenolic compounds from Juglans regia L. leaves, a rich source of this class of bioactive compounds. The extraction yield of the major compounds was quantified and the bioactivity of both extracts and pure solvents was evaluated by measuring the antioxidant, anti-inflammatory and cytotoxic activities. The solvent showing the best combined results (1,3-propanediol, PPD) was selected for further studies with eutectic mixtures. An experimental design, assisted by response surface methodology, was applied to obtain the highest extraction yield, selecting as independent variables the composition of the ternary mixtures of PPD, choline chloride (CC) or betaine (Bt), and water (W). The extracts obtained under optimized conditions, with CC:PPD:W (0.06:0.42:0.52, mole fraction) or Bt:PPD:W (0.08:0.25:0.67) presented similar extraction yields of phenolic compounds to the binary PPD:W (0.46:0.54). Interestingly, the bioactive profile was quite different. Only the extracts of the ternary mixtures presented anti-inflammatory activity, having also higher cytotoxicity, suggesting additional studies to evaluate the phytochemical composition of the extracts beyond phenolic compounds.

5 My contribution to this work was the measurement of the experimental data, the interpretation of the RSM results and writing of the manuscript.

126 Extractions with alternative solvents CHAPTER 3

Finally, this work contributes for the development of new applications of bioactive extracts in alternative solvents, with potential application as ingredients in the pharmaceutical or cosmetic areas.

Keywords: Alkanediols, Glycerol, Phenolic compounds; Juglans regia L.; Bioactivity; Eutectic mixtures.

3.2.1. Introduction Plants are important sources of bioactive phytochemicals, with a wide range of applications in the food, pharmaceutical and cosmetic sectors as natural ingredients, e.g. replacing synthetic additives or improving the quality of a final product (Rostagno and Prado, 2013). In particular, the leaves of Juglans regia L. (walnut tree) have been reported as a rich source of phenolic compounds, especially hydroxycinnamic acids and flavonols (Gawlik-Dziki et al., 2014; Santos et al., 2013), with potential application in the cosmetics or pharmaceutical areas (Ribeiro et al., 2015). In fact, they stand out for their bioactive properties (antioxidant, anti-inflammatory, antidiabetic, anti-proliferative, and antibacterial, among others) (Jahanban-Esfahlan et al., 2019; Panth et al., 2016).

Plant extracts are frequently obtained by extraction processes using conventional volatile organic solvents. However, alternative solvents that are more efficient, sustainable and safe are being proposed in the literature and those include natural deep eutectic solvents (NADES) (Vanda et al., 2018) (Dai et al., 2013a). Depending on the components of the eutectic solvent, it is possible to obtain liquid mixtures at room temperature using inexpensive starting materials, avoiding additional synthesis and purification steps (Benvenutti et al., 2019). Furthermore, similarly to ionic liquids, they can be designed to have the desired physical properties to extract specific target molecules. A considerable number of NADES are formed by components that are liquid at room temperature such as glycerol, ethylene glycol, propanediols (1,2 and 1,3 isomers), butanediols (1,2, 1,3 and 1,4 isomers) and triethylene glycol (Benvenutti et al., 2019; Cunha and Fernandes, 2018; Huang et al., 2019; Jablonský et al., 2018; Zainal-Abidin et al., 2017). Their use as pure solvents remains scarce, and their performance is rarely compared with the corresponding NADES with a few exceptions (Lavaud et al., 2016; Ozturk et al., 2018). Concerning the use of

127 Extractions with alternative solvents CHAPTER 3 alternative alcohols as solvents to extract phytochemicals from plants, only ethylene glycol, propanediols and glycerol were reported so far. The solvents most commonly applied, often mixed with water, were 1,2-propanediol (Baranauskaitė et al., 2016; Chulasiri et al., 2011; Gaweł-Bȩben et al., 2015; Jimtaisong and Krisdaphong, 2013; Lahucky et al., 2010; Lavaud et al., 2016; Manconi et al., 2017; Moraes et al., 2016; Stojiljković et al., 2016; Tubtimdee and Shotipruk, 2011) and glycerol (Apostolakis et al., 2014; Baranauskaitė et al., 2016; Huang et al., 2018; Karakashov et al., 2015; Lavaud et al., 2016; Philippi et al., 2016; Shehata et al., 2015), and less frequently ethylene glycol (Kim et al., 2007).

In this study, aqueous solutions of glycerol and a set of alkanediols (1,2-ethanediol, 1,2- propanediol, 1,3-propanediol, 1,3-butanediol, 1,2-pentanediol, 1,5-pentanediol and 1,2- hexanediol) were studied to extract phenolic compounds from walnut leaves. These compounds were chosen not only to evaluate the effect of increasing the alkyl chain of the diols on the extraction yield, but also considering their potential application in different areas such as pharmaceutical, food or cosmetics. For example, ethanol, 1,2-ethanediol, 1,2- propanediol, 1,3-butanediol and glycerol are listed as allowed ingredients for cosmetic formulations by the European Commission (European Commission, 2006). A recent patent published by Lavaud et al. (2016) about alternative solvents for food and cosmetic applications also includes 1,2-ethanediol, 1,2-propanediol, 1,3-propanediol, 1,2- pentanediol and glycerol, highlighting these ingredients as “bio-sourced natural compounds”, and the possibility to obtain them from renewable sources. Finally, 1,5- pentanediol was also studied as a preserving and humectant ingredient for dermatological care products (Foti et al., 2010; Gallo et al., 2003; Kerre, 2008) and 1,2-hexanediol was already proposed as a component for the preparation of topical formulations aiming to retard the percutaneous absorption of active pharmaceutical ingredients (Li et al., 2011). After the extraction process, the antioxidant, anti-inflammatory and cytotoxic activities of both extracts obtained and pure solvents were evaluated. The best solvent selected in this preliminary stage was used to form eutectic mixtures with betaine (Bt) or choline chloride (CC). To do so, an experimental design, assisted by response surface methodology was applied to obtain the highest phenolic extraction yield, selecting as independent variables

128 Extractions with alternative solvents CHAPTER 3 the composition of the eutectic mixture. Again, the bioactivity of the extracts obtained under optimized conditions was also evaluated.

Overall, this work aims to find alternative solvents with a desirable profile to obtain liquid extracts that could be directly used in different applications, highlighting the importance of walnut leaves as a source of bioactive compounds.

3.2.2. Material and methods

3.2.2.1. Plant material and extract preparation HPLC-grade acetonitrile, 1,2-ethanediol (99.94 %) and glycerol (99.88 %) were obtained from Fisher Scientific. The standards 5-O-caffeoylquinic acid, p-coumaric acid and quercetin 3-O-glucoside, and the solvent 1,3-butanediol (99.5 %) were purchased from Sigma-Aldrich. Choline chloride (98 %), 1,5-pentanediol (98 %), 2,2'-Azobis(2-methylpropionamidine) dihydrochloride (AAPH, 98 %) were obtained from ACROS Organics. 1,3-propanediol (≥ 99.8 %) was supplied from DuPont & Lyle BioProducts. 1,2-pentanediol (> 98 %) was purchased from TCI, while 1,2-hexanediol (97 %) and betaine (anhydrous, 98 %) were from Alfa AESAR. 1,2-propanediol (99.5 %) was purchased from Panreac. Finally, fluorescein sodium salt (98.5 – 100.5 %) was obtained from Fluka Analytical. All the other chemicals were of analytical grade and purchased from common sources.

3.2.2.2. Plant material The Juglans regia L. (walnut) leaves were purchased in a dried state from Soria Natural, S.A., Spain. According to the distributor, the leaves were collected in Soria (Spain) during June 2014, and naturally dried in a room with controlled humidity. Before the extractions, the samples were milled (60 to 20 mesh) and stored in a desiccator protected from light for subsequent assays.

3.2.2.3. Chromatographic analysis of the main phenolic compounds l The extract solutions were diluted with water and filtered through 0.2 µm disposable LC filter disks (30 mm, regenerated cellulose). The samples were analysed using a Shimadzu 20A series UFLC (Shimadzu Corporation, Kyoto, Japan) with a quaternary pump and a

129 Extractions with alternative solvents CHAPTER 3 photodiode array detector (PDA) coupled to an LC solution software data-processing station. A Waters Spherisorb S3 ODS-2 C18, (3 μm, 4.6 mm × 150 mm) column was used, operating at 35 °C. The chromatographic method was previously described by the authors (Vieira et al., 2017). A diode array detector (DAD) was used at 280 and 370 nm wavelengths. The target phenolic compounds were identified according to their UV spectra and retention time (Barros et al., 2013). For the quantitative analysis, a baseline to valley integration with baseline projection mode was used to calculate peak areas and external standards were used for quantification. The results were expressed in mg per g of dry plant (mg/g dry plant).

3.2.2.4. Extraction methodology

Preparation of solvents Firstly, the water amount in the components used to prepare the different solutions was measured using a Metrohm 831 Karl Fisher coulometer (data not shown). For the preparation of the solvents, each component was accurately weighed (±10-4 g). In the preliminary screening, aqueous solutions of different alcohols were prepared containing 20 % (in weight) of water. The deep eutectic solvents were prepared by the heating method, in a water bath at 50 °C (±0.1 °C), stirring the starting materials until the last crystal disappeared.

Extraction technique The heat assisted extractions (HAE) were performed using a heating equipment with magnetic stirring (Carousel 12 Plus Reaction Station™, Radleys Tech). The powdered samples (0.3 g) were extracted with 10 mL of each solvent, during 120 min at 50 °C and 600 rpm. After extraction, the mixtures were filtered through a Whatman n°4 paper for further analysis.

3.2.2.5. Cytotoxicity assays J. regia leaves extracts (30 mg/mL) and solvents were studied regarding their inhibitory growth activity of the HeLa (cervical carcinoma) cell line and the non-tumour PLP2 cells

130 Extractions with alternative solvents CHAPTER 3

(porcine liver primary cells), by using the sulforhodamine B assay. The extracts were diluted to obtain a stock solution of 10 mg/mL and the concentration of the working solutions ranged from 500 to 7.8 µg/mL, while the concentration of the solvents ranged from 4 % to 0.0156 %. Ellipticine was used as positive control. The experimental protocol related to the cell cultures was previously described by Barros et al. (2013) and Abreu et al. (2011). The concentration needed to reach 50 % of the growth inhibition effect (GI50) was used for quantification purposes.

3.2.2.6. Anti-inflammatory activity The walnut leaves extracts (30 mg/mL) and the pure solvents were tested for the lipopolysaccharide induced nitric oxide (NO) production assay according to Corrêa et al. (2015), using a mouse macrophage-like cell line (RAW264.7). The extracts were diluted to obtain a stock solution of 10 mg/mL and the concentration of the working solutions ranged from 500 to 7.8 µg/mL. Again, the concentration of the solvents ranged from 4 % to 0.0156 %. Then, the anti-inflammatory activity was assessed by measuring the nitrite concentration in the cell culture medium, using the Griess Reagent System kit. Dexamethasone was used as a positive control. The concentration providing 50 % of the inhibition of NO production (EC50) was determined.

3.2.2.7. Oxygen radical absorbance assay (ORAC) The antioxidant activity of extracts and solvents was measured by the ORAC assay. Firstly, the stock solutions (extracts and solvents) were successively diluted with distilled water (from 6 mg/mL to 0.09 mg/mL dry plant; 16 % to 0.25 % of solvent). Then, 50 µL of each were placed in a 96 well microplate with 200 µL of APPH reactive solution (PBS buffer solution, pH 6.25, 1M; fluorescein 4 nM; distilled water at 37 °C). Gallic acid was used as a control.

Fluorescence was measured by a FLx800 Spectroflourimeter. The emission was set at 528 nm and the excitation at 485 nm. The experiment was carried during 120 min with records at each 4 min, maintaining the chamber at 37 °C.

131 Extractions with alternative solvents CHAPTER 3

The results were calculated by determining the area under the resulting curve for the fluorescein decay and expressed as nM Fluorescein per millilitre of extract or solvent (nM F/mL).

Regarding the quantification methodology, the area under the curve (AUC) (Eq. 3.9), computed by any numerical integration method such as the trapezoidal rule, proved to be a highly robust criterion, able to summarize in a single and direct value the global feature of any kinetic profile.

in=−1 Rt11 Rtnn AUCRt=++  ii [3.9] 22i=2 where i is the number of data measured over time t, Ri are the responses along an arbitrary time series, and Δt is the interval of each measurement.

The AUC of a concentration-response of an antioxidant was normalized against the AUC obtained with the control, leading to the formulation of the relative area units or protected substrate ( P ) in percentage (Eq. 3.10), as defined similarly by other authors (Dávalos, 2004; Huang et al., 2002; Naguib, 2000) for antioxidant responses.

AUCAUCCA− 100 PAS( ) = 0  [3.10] AUCSC 0 where AUCC and AUCA are the area units corresponding to the kinetic profiles found in the absence (control, C) and presence of an antioxidant concentration A, respectively, and S0 is the initial substrate in the reaction (4 nM F/ mL).

The relationship in Eq. (3.10) establishes that AUCC (control) is also the maximum response achievable; consequently the values obtained were also standardized in percentage. In addition, by normalizing the response, the results obtained are less dependent on the experimental conditions, which, in practice, is one of the common problems when analysing the efficacy of response factors.

The asymptotic variation of as function of an antioxidant compound suggests that some radical-generating property of the system can be saturated (Gieseg and Esterbauer, 1994).

132 Extractions with alternative solvents CHAPTER 3

This type of concentration-response patterns should behave in a similarly accumulative way with a number of different antioxidant compounds found in the extract material. Therefore, in general, this patterns can be adjusted by a group of mathematical expressions (mechanistic or not) that translates the pattern of the response into parameters that allow to deduce the meaning and/or quantify the effect of the dependent variable in a simple and global mode. Previous researchers discussed the applicability of different mathematical expressions (Prieto et al., 2014); therefore, following their views, the Weibull cumulative distribution function was selected (Weibull and Sweden, 1951). Thus, the variation of P as function of increasing concentrations of an antioxidant (A) can be described satisfactorily using the Weibull model rearranged for our own purposes as follows in Eq. (3.11).

a 1−a 2Vm PtAPA( ,1expln2) =−−m ( )  [3.11] Pa m

The parameter Pm is the averaged maximum protected substrate, asymptotic value of the response (nM of F), which is specific of each A agent. The parameter Vm corresponds to the average amount of protected molecules per gram of extracted material (nM of protected substrate/g extract). The parameter a is a shape parameter related to the slope that can produce potential profiles (α < 1), first order kinetic ones (α = 1) and a variety of sigmoidal profiles (α > 1). The parameter Vm was the quantification solution used to compare and analyse the results obtained.

3.2.2.8. Optimization of the solvent composition (eutectic mixtures)

Experimental design After the preliminary screening of the binary mixtures containing an alcohol and water, 1,3- propanediol (PPD) was selected for further studies, to form eutectic mixtures with either betaine (Bt) or choline chloride (CC). The effect of water (W) addition was also considered. Therefore, a triangular study was designed for modelling ternary mixtures (Neto et al., 2007). Thus, x (Bt or CC, mole fraction), y (PPD, mole fraction) and z (W, mole fraction) were chosen as variables. An extended Simplex Centroid Design (SCD) was used. Therefore, the

133 Extractions with alternative solvents CHAPTER 3 combined effect of these variables on the recovery of the four main phenolic compounds from J. regia leaves was evaluated in order to maximize their individual and global yields. Nevertheless, the high melting temperatures of CC (324 °C) and Bt (290 °C) (Fernandez et al., 2017; Patil and Tatke, 2015) and the miscibility limits of the ternary mixtures impose some restraints to the composition variables. Preliminary solubility tests carried out at room temperature (data not shown), allowed us to estimate a liquid phase region in which the experimental extraction conditions could be designed as schematically shown in Figure B2. Therefore, x the mole fraction of betaine varied between 0 and 0.08, and the mole fraction of choline chloride varied between 0 and 0.3.

The adopted model is a central composite-like for a limited design. To do that, the design entails the vertex of the quadrilateral, central point and intermedia points covering the quadrilateral surface. Overall, eleven independent combinations were set, the conditions (x, y and z) were optimized, and the responses found were compared with the experimental ones.

Mathematic model for the triangular mixture design The response models were fitted by means of least-squares calculation using the following quadratic model:

nnnn −1 2 Ybb=+++ Xb0 X Xbiiijijiii X iiji====1121 [2.1] ji

In which, Y represents the dependent variable (response variable) to be modelled, the independent variables are Xi and Xj, b0 is the constant coefficient, bi the coefficient of linear effect, bij the coefficient of interaction effect, bii the coefficients of quadratic effect and n is the number of variables.

A simplex method was used to maximize the model process by solving non-linear problems to optimize the 3-O-caffeoylquinic acid, trans 3-O-coumaroylquinic acid, quercetin 3-O- glucoside, quercetin O-pentoside and total HPLC content (Heleno et al., 2016; Pinela et al., 2016b). To avoid unnatural conditions of the variables certain limitations were imposed (i.e., mole fractions lower than 0).

134 Extractions with alternative solvents CHAPTER 3

3.2.2.9. Statistical analysis In the bioactivity assays, duplicates of each extract were assayed and three repetitions of each methodology were performed, being the results expressed as mean values and standard deviations (SD). One-way ANOVA followed by Tukey’s HSD test (p = 0.05) were used to analyse the results. Furthermore, significant differences between two samples were established by applying a Student´s t-test, with p = 0.05 (SPSS v. 23.0 program).

In the specific case of the ORAC assay, the results are expressed as mean and confidence intervals (CI) as they better traduce the analysis of kinetic procedures.

3.2.3. Results and discussion

3.2.3.1. Preliminary screening using alcohols

Extraction yields First, aqueous solutions of glycerol and several diols (1,2-ethanediol, 1,2-propanediol, 1,3- propanediol, 1,3-butanediol, 1,2-pentanediol, 1,5-pentanediol and 1,2-hexanediol) containing 20 % of water (w/w) were used to extract phenolic compounds from J. regia L. leaves. Water and ethanol + water (80:20, w/w) solvents were also applied for comparison purposes.

The responses used to evaluate the efficiency of the extraction for each solvent were quantified by HPLC-DAD of the four main phenolic compounds (3-O-caffeoylquinic acid, trans 3-p-coumaroylquinic acid, quercetin-3-O-glucoside and quercetin-O-pentoside). The typical chromatograms obtained by applying conventional solvents (water and ethanol + water) are depicted in Figure 3.5.

The extraction yield of the main phenolic compounds can be consulted in Table 3.5 (screening part). As can be seen, hydroalcoholic solvents (except 1,2-hexanediol) exhibited higher extraction yields than water. This group of diols performed similarly to ethanol (27.8 ± 0.1 mg/g dry plant), ranging the phenolic content from 23.9 ± 0.3 mg/g dry plant (1,3- butanediol) to 30.5 ± 0.2 mg/g dry plant (1,2-ethanediol). The 1,2-ethanediol extract was the only one exceeding the extraction yields obtained by the reference solvent (ethanol).

135 Extractions with alternative solvents CHAPTER 3

Propanediol extracts (1,2 and 1,3 isomers) have shown equivalent extraction yields. In contrast, with the more polar glycerol (1,2,3-propanetriol) only a total content of 18.3 ± 0.4 mg/g dry plant was obtained. Regarding the general phenolic profile of pentanediol extracts, 1,2-pentanediol extracted the highest amount of quercetin derivatives (24.5 mg/g dry plant), considering all solvents. In contrast, the 1,5-pentanediol liquid extract was richer in phenolic acids (5.69 mg/g dry plant) than the 1,2 isomer (1.60 mg/g dry plant). Finally, 1,2-hexanediol was the poorest solvent (5.73 ± 0.09 mg/g dry plant), with lower extraction yields than water (14.1 ± 0.3 mg/g dry plant).

Table 3.5. Quantification of the main phenolic compounds present in J. regia leaves (mean ± SD): 3-O- caffeoylquinic acid (P1), trans 3-p-coumaroylquinic acid (P2), quercetin 3-O-glucoside (P3), quercetin O- pentoside (P4) and total HPLC content (PT, P1 +P2 + P3 + P4).

P1 P2 P3 P4 PT Solvent (mg/g dry plant) (mg/g dry plant) (mg/g dry plant) (mg/g dry plant) (mg/g dry plant)

Screening

Water 5.16 ± 0.06b 1.07 ± 0.03d 4.3 ± 0.2h 3.59 ± 0.07h 14.1 ± 0.3g Ethanol 4.52 ± 0.01d 1.23 ± 0.02b 11.7 ± 0.1c 10.32 ± 0.06c 27.8 ± 0.1b 1,2-Ethanediol 5.79 ± 0.08a 1.36 ± 0.02a 12.6 ± 0.1b 10.73 ± 0.08b 30.5 ± 0.2a 1,2-Propanediol 4.96 ± 0.06c 1.14 ± 0.01c 11.3 ± 0.2d 9.8 ± 0.2d 27.2 ± 0.4c 1,3-Propanediol 5.30 ± 0.04b 1.22 ± 0.04b 11.1 ± 0.1d 9.6 ± 0.07de 27.3 ± 0.3c 1,3-Butanediol 4.46 ± 0.01d 1.03 ± 0.01d 9.85 ± 0.2f 8.6 ± 0.2f 23.9 ± 0.3e 1,2-Pentanediol 1.30 ± 0.02g 0.300 ± 0.002g 13.0 ± 0.3a 11.45 ± 0.2a 26.0 ± 0.5d 1,5-Pentanediol 4.47 ± 0.07d 1.22 ± 0.01b 10.75 ± 0.2e 9.5 ± 0.1e 25.9 ± 0.4d 1,2-Hexanediol 3.00 ± 0.05f 0.8 ± 0.03f 1.93 ± 0.04i nd 5.73 ± 0.09h Glycerol 4.3 ± 0.1e 0.94 ± 0.04e 6.98 ± 0.2g 6.1 ±0.1g 18.3 ± 0.4f

Optimization

PPD:W 4.66 ± 0.08 1.12 ± 0.03a 10.7 ± 0.2b 9.5 ± 0.2a 26.0 ± 0.4b Bt:PPD:W 4.62 ± 0.01 1.11 ± 0.02a 11.11 ± 0.06a 9.52 ± 0.05a 26.4 ± 0.1a CC:PPD:W 4.6 ± 0.1 1.02 ± 0.01b 10.5 ± 0.1c 8.95 ± 0.03b 25.1 ± 0.2c

Calibration curves: 3-O-caffeoylquinic acid: 5-O-caffeoylquinic acid (y=118879x – 181046; r2=0.9992); trans 3-p-coumaroylquinic acid: p- coumaric acid (y=120011x + 1×106; r2=0.9973); quercetin 3-O-glucoside and quercetin O-pentoside: quercetin 3-O-glucoside (y=98385x + 143369; r2=0.9978). Optimized solvents (mole fractions): PPD:W, 0.4582:0.5418; Bt:PPD:W, 0.08:0.2473:0.6727; CC:PPD:W, 0.0583:0.4182:0.5235. Different letters represent significant differences (p < 0.05).

136 Extractions with alternative solvents CHAPTER 3

Figure 3.5. Chromatographic records at 280 nm (part A) used for the quantification of phenolic acids (P1: 3-O-caffeoylquinic acid; P2: trans 3-p-coumaroylquinic acid), and, at 370 nm (part B) for the quantification of quercetin derivatives (P3: quercetin 3-O-glucoside; P4:quercetin O-pentoside) present in J. regia L. leaves samples. Representation of conventional solvents extracts (water and ethanol + water) solvents and graphical comparison with the extract obtained applying 1,3- propanediol aqueous solution during the screening process.

Cytotoxicity of extracts and solvents To assist the development of different applications using these extracts, cytotoxicity assays were applied not only to the liquid extracts, but also to the pure solvents. The cytotoxicity was evaluated using a human tumour cell line (HeLa) and a non-tumour porcine live primary cell culture (PLP2). The HeLa cell line represent a suitable in vitro model to study cytotoxicity that was already applied to new solvents (Xia et al., 2018) such as NADES (Hayyan et al., 2016; Radošević et al., 2016) and ionic liquids prepared from biomaterials (Gouveia et al., 2014). On the other hand, PLP2 cells can be used to evaluate one specific type of cytotoxicity – the hepatotoxicity (Abreu et al., 2011). Therefore, the potential toxicity occurring during/after metabolization is here represented by the porcine liver primary cells (Rached et al., 2018). The results of the cytotoxicity assays are summarized in Table 3.6. (screening part). It is important to mention that none of the extracts was as

137 Extractions with alternative solvents CHAPTER 3

active as the positive control (GI50 ellipticine = 1.0 ± 0.1 and 2.3 ± 0.2 µg/mL for HeLa and PLP2, respectively).

138 Extractions with alternative solvents CHAPTER 3

Table 3.6. Bioactive properties of J. regia leaves. Cytotoxicity in human tumour cell lines HeLa and non-tumour liver primary cells PLP2 (GI50). Anti-inflammatory activity by the inhibition of the NO production (EC50). Antioxidant activity by the ORAC assay (EC50). Quantification for extracts and solvents (mean ± SD), except for ORAC results as they are presented as mean ± CI.

HeLa PLP2 NO production ORAC Solvent Extract Solvent Extract Solvent Extract Solvent Extract Solvent (µg/mL) (%) (µg/mL) (%) (µg/mL) (%) (nM F/mL) (nM F/mL)

Screening

Water > 500 > 4 >500 > 4 > 500 > 4 3.3 ± 0.4 - Ethanol 245 ± 14b > 4 > 500 > 4 > 500 > 4 45 ± 6 11 ± 2 1,2-Ethanediol 97 ± 10e > 4 142 ± 5b > 4 > 500 > 4 54 ± 7 10 ± 1 1,2-Propanediol 292 ± 24a > 4 > 500 > 4 > 500 > 4 60 ± 8 11 ± 2 1,3-Propanediol 216 ± 10c > 4 > 500 > 4 > 500 > 4 66 ± 9 13 ± 2 1,3-Butanediol 257 ± 12b > 4 > 500 > 4 > 500 > 4 58 ± 8 16 ± 2 1,2-Pentanediol 151 ± 12d 0.63 ± 0.04a 232 ± 13a 0.89 ± 0.04a > 500 > 4 31 ± 4 7 ± 1 1,5-Pentanediol 212 ± 4c 0.49 ± 0.02b 141 ± 3b 0.59 ± 0.02b > 500 > 4 60 ± 8 16 ± 2 1,2-Hexanediol 37 ± 2f 0.36 ± 0.02c 48 ± 5b 0.285 ± 0.005c > 500 > 4 31 ± 4 8 ± 1 Glycerol 88 ± 4e > 4 143 ± 5b > 4 > 500 > 4 22 ± 3 7 ± 1

Optimization

PPD:W 285 ± 13a > 4 > 500 > 4 > 500 > 4 61 ± 6 12 ± 1 Bt:PPD:W 103 ± 3b > 4 158 ± 6* > 4 183 ± 9* > 4 76 ± 8 16 ± 4 CC:PPD:W 80 ± 1c > 4 131 ± 1* > 4 134 ± 1* > 4 71 ± 5 13 ± 3

Positive controls: Ellipticine (GI50 values): HeLa: 1.03 ± 0.09 µg/mL; PLP2: 2.3 ± 0.2 µg/mL; Dexamethasone (EC50 values): 16 ± 1 µg/mL; Gallic acid (nM F/mL): 48 ± 6 nM F/mL. Results expressed in mean values ± standard deviation (SD). Optimized solvents (mole fractions): PPD:W, 0.4582:0.5418; Bt:PPD:W, 0.08:0.2473:0.6727; CC:PPD:W, 0.0583:0.4182:0.5235. Different letters represent significant differences (p < 0.05). *Means statistical differences obtained by a t-student test.

139 Extractions with alternative solvents CHAPTER 3

As can be seen, the extracts obtained using standard solvents (water and ethanol) have no toxicological impact in PLP2 cells, allowing the normal cell proliferation under concentrations up to 500 µg/mL. The same was observed for the aqueous extract applied to the HeLa cell culture; however, for the ethanolic extract a GI50 of 245 ± 14 µg/mL was obtained. This result is in good agreement with the study made by Santos et al. (2013) using methanolic extracts (GI50 = 294.87 ± 9.36 µg/mL dry extract) of walnut leaves against the same cell line. Regarding the extracts showing cytotoxic potential only against the tumour cell line, but not for the non-tumour one (1,2- and 1,3-propanediol, and 1,3-butanediol),

1,3-propanediol demonstrated to be the most effective solvent (GI50 = 216 ± 10 µg/mL and > 500 µg/mL, respectively for the HeLa and PLP2 cell cultures). Interestingly, the extracts obtained using 1,2-ethanediol and glycerol showed toxicity against the selected control cell culture (PLP2), with GI50 values of 142 ± 5 µg/mL and 143 ± 5 µg/mL, respectively. Nevertheless, the pure solvents (without extracts) were absent of cytotoxicity up to concentrations of 4 %. Besides the cytotoxicity of 1,2-ethanediol and glycerol extracts against the PLP2 cells, the GI50 values for the HeLa cells were lower (GI50 = 97 ± 10 µg/mL and 88 ± 4 µg/mL, respectively); thus, the extracts were more active against the cervical human tumour cell line. Finally, the most cytotoxic extract was obtained with 1,2- hexanediol, being the concentrations cells significantly lower than the ones obtained with the remaining solvents (GI50, HeLa = 37 ± 2 µg/mL). The same was observed for the pure solvent, constituting a strong cytotoxic agent even when highly diluted (GI50, PLP2 = 0.285 ±

0.005 % and GI50, HeLa = 0.36 ± 0.02 %). This was also observed for both pentanediol isomers which presented activity for concentrations lower than 1 %. The 1,5-pentanediol extract showed significant higher GI50 values against the HeLa cell line (GI50 = 212 ± 4 µg/mL) than glycerol and 1,2-ethanediol, but the effective GI50 for the PLP2 cells was lower (141 ± 3

µg/mL), which is not desirable, as ideally GI50, PLP2 > GI50, HeLa.

To complete the above discussion, it is important to clarify the contribution of the solvent to the final toxicity of the extract. For that reason, the toxicity of the solvent was evaluated in a concentration range that includes the concentration of the solvent in the liquid extracts studied in the bioactivity studies. When both extracts and solvents presented cytotoxicity (1,2-pentanediol, 1,5-pentanediol and 1,2-hexanediol), it was found that the

140 Extractions with alternative solvents CHAPTER 3 concentrations providing cytotoxicity for the pure solvents are similar to the ones present in the extracts (HeLa = 0.41 %, 0.57 % and 0.10 %; PLP2 = 0.63 %, 0.38 % and 0.13 %, respectively). Thus, the toxicity found is probably due to the presence of the solvent. On the other hand, for the remaining extracts, the bioactivity could be attributed (mainly) to the presence of bioactive substances since the solvents do not show toxicity against the studied cell lines up to 4 % concentration, which is higher than the amount of solvent present in the analysed liquid extracts.

The cytotoxicity of walnut leaves extracts were previously studied by Santos et al. (2013) for several cell lines, including HeLa and PLP2 cell cultures. The authors reported the absence of toxicity of extracts obtained by decoction (GI50 > 500 µg/mL), a specific type of aqueous extract, which is in good agreement with our results.

Antioxidant and anti-inflammatory activity of extracts and solvents The antioxidant activity of the tested extracts of walnut leaves was investigated through the oxygen radical absorbance capacity (ORAC) assay. As some of the selected solvents may exhibit antioxidant properties, they were also studied. The principle of this method consists on the inhibition of the peroxyl radical induced oxidation (induced by thermal decomposition) of a probe, generally 2,2-azobis (2-amidino-propane) dihydrochloride (AAPH). The antioxidant will shell the fluorescein (F) degradation, retarding the fluorescence decay generated by AAPH (Mellado-Ortega et al., 2017). In this sense, higher ORAC values (expressed as nM F/mL) indicate a higher antioxidant action of an extract or solvent. The results are listed in Table 3.6 (screening part), and Figure 3.6 shows the antioxidant responses of the ORAC assay based in the kinetic dose-response effect of the liquid extracts (solvents & extract) and pure solvents. The ORAC results of the liquid extracts ranged from 3.3 ± 0.4 nM F/mL (water extract) to 66 ± 9 nM F/mL obtained with the extract in 1,3-propanediol. As expected, the extracts using ethanol (a common solvent for the extraction of bioactive/ phenolic compounds) presented an antioxidant capacity 14- fold higher (45 ± 6 nM FmL) than water, similar to the values obtained using the selected control (48 ± 6 nM F/mL for gallic acid), though lower than the one obtained with 1,3- propanediol. Glycerol, 1,2-pentanediol and 1,2-hexanediol resulted in intermediate values.

141 Extractions with alternative solvents CHAPTER 3

Regarding the antioxidant potential of solvents, they also present some antioxidant activity, approximately five times lower than the corresponding liquid extracts, showing a slower kinetic behaviour, traduced by the lower amounts of nM of fluorescein protected per one millilitre of solvent compared to the ones achieved by the extracts & solvent. For instance, as can be seen in Part B of Figure 3.6, the limiting case of water (section B10) allows a completely decay on the fluorescein signal while B6 (1,2-pentanediol), B8 (1,2- hexanediol) and B9 (glycerol) presented similar kinetics, however, with lower antioxidant potential (Part C) than the other solvents.

To the best of our knowledge, the evaluation of the antioxidant activity of extracts by the ORAC assay was performed for the first time for these type of liquid extracts. Finally, as presented in Table 3.6 (screening part), none of the extracts or solvents inhibited the NO production by macrophages, i.e., none of the samples had anti-inflammatory potential.

142 Extractions with alternative solvents CHAPTER 3

Figure 3.6. Parts A and B show the application of the antioxidant responses of the ORAC assay based in the kinetic dose-response effect of the solvents & extract (mg F dw/mL of extraction) and solvents (%), respectively. The dots () are the experimental data series and the lines the fittings. Part C shows the numerical values of the rate parameter of the response as assessment criteria. The numerical notation 1 to 10 represents the solvents (ethanol, 1,2-ethanediol, 1,2-propanediol, 1,3- propanediol, 1,3-butanediol, 1,2-pentanediol, 1,5-pentanediol, 1,2-hexanediol, glycerol and water), respectively.

143 Extractions with alternative solvents CHAPTER 3

3.2.3.2. Eutectic mixtures based on 1,3-propanediol

Optimization of the extraction yield of phenolic compounds The best combined results in terms of extraction yield, antioxidant potential, and non-toxic effects for non-tumour cells, but anti-proliferative activity against the tumour cells were obtained with 1,3-propanediol (PPD). Thus, this solvent was selected to form eutectic mixtures with two of the most reported hydrogen bond acceptors in literature: choline chloride (CC) and betaine (Bt). Previous studies showed that the solvent composition is one of the main factors influencing the extraction of phenolic compounds from the leaves of J. regia (Vieira et al., 2018, 2017). In this context, an experimental mixture SCD design was developed to optimize the recovery of the phenolic fraction from walnut leaves. The selected independent variables were the mole fractions of the three compounds (CC or Bt, PPD and W) used to form eutectic mixtures within an existing liquid phase region at room temperature.

The parametric fitting values, confidence intervals and statistical data are presented in Table B2. All coefficients showed significant parametric intervals at the 95% confidence level (α = 0.05) and the correlation coefficients were always higher than 0.9068. The parametric values to describe the optimum conditions of solvent are listed in Table B2. The models for each individual response were constructed fitting the quadratic model of Eq. (2.1) to the experimental values in Table B2 using nonlinear least-squares estimations. As can be seen, not all the parameters of Eq. (2.1) were used for building the model since some coefficients were non-significant (data not shown). Therefore, linear and interactive effects played an important role to describe the behaviour of the extraction yield vs solvent composition.

The optimization of the target response assumed different profiles depending on the nature of the solvent. The optimum conditions for the total HPLC is the combined effect of two distinct tendencies (one for the phenolic acids, and another for the quercetin derivatives). Thus, as quercetin derivatives are present in higher amounts (comparatively to the phenolic acids) in J. regia L. leaves, the favourable extraction yields for phenolics (total HPLC) assumed conditions quite similar to the ones found for quercetin derivatives.

144 Extractions with alternative solvents CHAPTER 3

These findings are in good agreement with the previous reports using the same plant material, during optimization processes using aqueous ethanol mixtures and aqueous solutions of deep eutectic solvents composed of choline chloride and carboxylic acids (Vieira et al., 2018, 2017).

Finally, the conditions of solvent composition found to be optimum to extract phenolic compounds (total HPLC) from J. regia leaves were the following mole fractions (x : y : z):

A = 0 : 0.4582 : 0.5418;

B = 0.08 : 0.2473 : 0.6727; in which x is composed by betaine (0 < x <0.08);

C = 0.0583 : 0.4182 : 0.5235; in which x is composed by choline chloride (0 < x <0.3).

The phenolic profile of the extracts obtained under optimized conditions is very similar (Figure B3), with a slight quantitative variation. Regarding the final amounts of phenolic compounds, the extraction yields obtained by applying the PPD:W and Bt:PPD:W solvents were similar (26.0 ± 0.4 and 26.4 ± 0.1 mg/g dry plant, respectively). A higher content of quercetin 3-O-glucoside was obtained in the Bt-based solvent extract (11.11 ± 0.06 mg/g dry plant), but lower in the CC-based extract (10.5 ± 0.1 mg/g dry plant). In contrast, the concentrations of 3-O-caffeoylquinic acid were equivalent, independently of the solvent used for extraction.

Under the optimized conditions, the experimental results obtained by HPLC-DAD ranged from 25.1 ± 0.2 to 26.4 ± 0.1 mg/g dry plant (CC and Bt-based solvent extracts, respectively). Regarding the total content in phenolic acids, it ranged from 5.6 mg/g dry plant to 5.78 mg/g dry plant using CC:PPD:W and PPD:W solvents, respectively. The total content in flavonols ranged between 19.45 mg/g dry plant and 20.63 mg/g dry plant, for CC:PPD:W and Bt:PPD:W, respectively.

Simplifying, besides the variety of the solvent constitution, namely the PPD+W ratio and the presence (or not) of a third element (Bt or CC) to form an eutectic mixture, the gains (besides the significant differences) in terms of phenolic yields ranged from (absolute) values with the same magnitude.

145 Extractions with alternative solvents CHAPTER 3

In summary, under the extraction conditions studied, the presence of a third element (Bt or CC) does not appreciably increase the extraction yield of phenolic compounds. In the literature, the number of studies comparing this type of binary (W + alkanediol) and ternary mixtures (W + alkanediol + CC or Bt) is scarce. Ozturk et al. (2018) reported that a higher total phenolic content was obtained from Citrus x sinensis fruit peel extracts, using an aqueous solution of ethylene glycol when compared to aqueous solutions of ethylene glycol with choline chloride. Under the studied conditions, aqueous ethylene glycol also doubled the extraction yield obtained with aqueous ethanol. On the other hand, the eutectic mixtures using glycerol:CC:W were able to extract 6-fold higher amounts of total phenolic compounds compared to the aqueous glycerol solution. A recent patent, showed the increased yields achieved by eutectic systems (Bt : 1,2-propanediol or PPD or 1,2- pentanediol : W) compared to the binary mixtures of water and diol (Lavaud et al., 2016). Higher extraction yields of rosmarinic acid, oleuropein, amentoflavone and total phenolic contents were obtained from Rosmarinus officinalis leaves, Olea europaea leaves, Selaginella pulvinata aerial parts and Tillandsia usneoides aerial parts, respectively.

It is also important to mention that both studies evaluated specific ratios of the starting materials, while in the present study an experimental design was carried out to cover a larger range of solvent composition. Therefore, it would be recommended to compare the results obtained using eutectic mixtures with the pure starting materials. If they are solid compounds at room temperature, the hydrotropy effect could be checked by using their aqueous solutions as extraction solvents.

In summary, the extraction yields of phenolic compounds obtained in this work are very similar in the three extracts obtained under optimized conditions. Nevertheless, the phytochemical complexity of the extracts beyond phenolic compounds (e.g. tocopherols, organic acids, mono and oligosaccharides, or even diarylheptanoids) may have an impact on the bioactivity of the final formulations, as studied in the next section.

Cytotoxicity of the optimum mixtures (extracts and solvents) The cytotoxic effect of PPD-based solvents and extracts obtained under optimized conditions was assessed for the HeLa and PLP2 cells, as in the preliminary screening. The

146 Extractions with alternative solvents CHAPTER 3 results are presented in Table 3.6 (optimization part). The most active extract was obtained with the CC-based solvent, having the lowest GI50 values for both cell cultures (GI50, HeLa =

80 ± 1 µg/mL and GI50, PLP2 = 131 ± 1 µg/mL), followed closely by the Bt-based extract. The binary mixture of PPD and W (0.78:0.22, w/w) does not present cytotoxic effects over the

PLP2 cells (GI50 > 500 µg/mL), but it does against the Hela ones, however, with lower impact

(GI50 = 285 ± 13 µg/mL). These results are consistent with the values obtained in the screening step for the PPD:W (0.80:0.20, w/w).

As can be seen in the chromatograms depicted in Figure B3, the phenolic profile is the same for all the extracts, suggesting that the cytotoxicity found could be caused by the presence of unknown (non-phenolic) molecules that the eutectic mixtures are capable to recover from walnut leaves. It is important to add that the solvents do not induce cytotoxic effects in the concentration range as they are present in the extracts (GI50 > 4 %) and that, for all the extracts, GI50,PLP2 > GI50,HeLa.

The use of eutectic solvents has been motivating some authors to explore their potential cytotoxicity. Even so, studies using cell lines are scarce. Most studies are focused on the solvent toxicity (Hayyan et al., 2016; Macário et al., 2019; Paiva et al., 2014; Radošević et al., 2018) and less on the eutectic extracts (Lavaud et al., 2016; Radošević et al., 2016).

Paiva et al. (2014) have shown that CC:carboxylic acids eutectic solvents had higher impact on the cell viability (higher cytotoxicity) than the ones composed by CC and sugars. The authors used L929 fibroblast-like cells in their study (Paiva et al., 2014). Later, Hayyan et al.

(2016) found significant lower IC50 values for the CC+malonic acid (1:1, mol/mol) eutectic mixture (IC50 HeLa S3 = 20 ± 8.4 mM) comparatively to the ones obtained with CC:glycerol: water (1:2:1, mole fractions) (IC50 HeLa S3 = 427 ± 11 mM), while the IC50 values for the solvents using sugars ranged from 166 ± 11 mM (CC:sucrose:water, 4:1:4, mole fractions) to 182 ± 7.6 mM (CC:glucose:water, 5:2:5, mole fractions). Recently, Radošević et al. (2018) reported the cytotoxicity of several eutectic solvents in cell cultures, including the HeLa cells and a non-tumor embryotic kidney cell culture (HEK293T). With the exception of the mixture CC:oxalic acid (1:1 mol/mol, and 10 to 30 % wt of water) with EC50, HeLa = 330.90 ± 29.75 µg/mL, none of the assessed solvents showed cytotoxic potential (up to 2000 µg/mL).

147 Extractions with alternative solvents CHAPTER 3

More recently, Macário et al. (2019) studied the cytotoxic effect of CC-based solvents (butanoic acid, hexanoic acid, ethylene glycol, 1-propanol and urea) for human skin cells. The authors verified an increase on the HaCaT cells viability, and some of them promoted an analogous action over the MNT-1 cells as well (CC:ethylene glycol and CC:1-propanol, both at 1:1, mol/mol), highlighting the potential use of CC-based solvents in cosmetic and pharmaceutical formulations (Macário et al., 2019), even though CC is not listed by the European Commission for topical use (European Commission, 2006).

On the other hand, considering the bioactivity of extracts, Radošević et al. (2016) studied the cytotoxic effect of the resulting extracts of grape skins (Vitis vinifera L.) obtained with different families of eutectic solvents. Interestingly, the results follow the tendency already found for the pure solvents by Hayyan et al. (2016). Concerning the results obtained for the

HeLa cell cultures, the authors also found higher EC50 values (lower cytotoxicity) for the extract obtained with CC:glycerol (1:2, mol/mol) (EC50 = 106 ± 3 µg/mL) than with CC+malic acid (1:1, mol/mol) (EC50 = 18 ± 3 µg/mL) for extraction. Thus, none of the studies compared the bioactivity of the pure solvent and the resulting extract, as suggested by the present work.

Anti-inflammatory and antioxidant activity of extracts and solvents The results of the anti-inflammatory and antioxidant assays are presented in Table 3.6

(optimization part). The eutectic extracts were able to inhibit the NO production (EC50 = 183 ± 9 µg/mL and 134 ± 1 µg/mL, for the Bt and CC based solvents, respectively), whereas the

PPD:W extract had no anti-inflammatory activity (EC50 > 500 µg/mL). These results cannot be granted to the solvent since none of them showed anti-inflammatory potential under the studied concentration range (up to 4 %). This effect could result from the synergistic effect of extract and solvent, or more likely to the presence of other phytochemicals as previously suggested.

Regarding the antioxidant potential, slightly better results were obtained for the eutectic extracts (EC50 = 76 ± 8 nM F/mL and 71 ± 5 nM F/mL for the Bt and CC based solvents extracts, respectively) compared to the binary solvent extract PPD:W (EC50 = 61 ± 6 nM F/mL).

148 Extractions with alternative solvents CHAPTER 3

Lavaud et al. (2016) also presented an example in which similar phenolic profiles, obtained by applying eutectic and non-eutectic solvents, produced different bioactivities. That is the case of the eutectic extract (Bt:glycerol at 40:60, mol/mol and 25 % wt of water) of Crithmum maritimum that exhibited an inhibition of melanin synthesis on human epidermal melanocytes whereas the extract obtained with the mixed solvent glycerol:W (50 %, wt) had no activity.

3.2.4. Conclusions Aqueous solutions of several alkanediols were able to extract similar amounts of phenolic compounds as ethanol, a conventional volatile solvent for these target compounds. Among alkanediols, 1,3-propanediol showed the best combined results considering the extraction yield and bioactivity. In fact, the extract of 1,3-propanediol showed higher antioxidant activity than the ethanolic one. Both extracts had similar cytotoxicity, being active against the human cervical carcinoma cell line, but not for the non-tumour porcine liver primary cell culture. None of the aqueous mixtures used during the screening step showed anti- inflammatory activity. Furthermore, the solvents until 4 carbon atoms were not toxic for both cell cultures, which extends their potential range of applications.

Regarding the extracts obtained with eutectic solvents composed by 1,3-propanediol, they showed similar phenolic profile. However, differences were found in their bioactivity, as higher antioxidant and cytotoxic potentials were obtained. Moreover, only eutectic extracts showed anti-inflammatory effects. Most likely, the presence of other bioactive molecules beyond phenolics or the combined effect of solvent and extract induced those effects. Therefore, further studies should be carried out to obtain the phytochemical profile of other families of compounds and, also, to optimize the composition of the solvent to achieve the desired bioactivity.

Considering the lower vapour pressure of alkanediols and glycerol, their use could be advantageous in some applications. In those cases, the preconcentration of extracts due to the volume fluctuations while using conventional volatile organic solvents could be avoided.

149 Extractions with alternative solvents CHAPTER 3

Overall, this work supports the use of liquid extracts of walnut leaves as a source of important phytochemicals with bioactive properties with potential application as pharmaceutical or cosmetic ingredients.

150

4. Solubility of target molecules

Solubility of target molecules CHAPTER 4

4.1. Solvent and temperature effects on the solubility of syringic, vanillic or veratric acids: experimental, modeling and solid phase studies Vilas-Boas, S.M., Vieira, V., Brandão, P., Alves, R.S., Coutinho, J.A.P., Pinho, S.P., Ferreira, O., 2019. J. Mol. Liq. 289, 111089. 6

Abstract

The solubility of syringic acid, vanillic acid and veratric acid was measured in pure water and eleven organic solvents (methanol, ethanol, 1-propanol, 2-propanol, 2-butanone, ethyl acetate, acetonitrile, dimethylformamide, 1,2-propanediol, 1,3-propanediol and 1,3- butanediol), at 298.2 K and 313.2 K. Besides the solubility data, the melting temperatures and enthalpies of the solutes were determined by differential scanning calorimetry, while powder and single X-ray diffraction was used to resolve the solute solid structure, before and after the solubility studies.

For modeling purposes, the NRTL-SAC model, also combined with the Reference Solvent Approach (RSA), and the Abraham solvation model were applied to describe the solid-liquid equilibria of the binary systems. A set of solvents was used to estimate the model parameters and afterwards, solubility predictions were carried out for binary systems not included in the correlation step. Better results were obtained using the Abraham solvation model with average relative deviations (ARD) of 15 % for the correlation set and 26 % for the predictions, which are more satisfactory than the results with the NRTL-SAC (33 % for the correlation and 59 % for the predictions) or NRTL-SAC combined with RSA (30 % for the correlation and 59 % for the predictions).

6 My contribution to this work was the experimental measurement of the solid-liquid equilibria data, the collection of literature data, the interpretation of the obtained results, and writing of the manuscript, all in collaboration with VBSM.

153 Solubility of target molecules CHAPTER 4

Keywords: Phenolic acids, melting properties, NRTL-SAC, Abraham solvation model, X-ray diffraction.

4.1.1. Introduction Among numerous phenolic compounds present in vegetables, fruits and aromatic herbs, phenolic acids are an important class as they show relevant pharmacological, biological and organoleptic activities, usually presenting high-value added (Heleno et al., 2015; Nollet and Gutierrez-Uribe, 2018). In this context, reliable solubility data of phenolic acids and derivatives in water and organic solvents are essential to the design of separation processes, such as extraction or crystallization, in the food, cosmetic, chemical and pharmaceutical industries (Nollet and Gutierrez-Uribe, 2018; Stalikas, 2007).

Following our previous work on the solubility of hydroxybenzoic acids (gallic, α-resorcylic, gentisic and protocatechuic acids) (Vilas-Boas et al., 2018), in this work our focus is on benzoic acids containing the methoxy group, namely syringic, vanillic and veratric acids. Interesting structural differences can be seen in Figure 4.1, compared to vanillic acid (4- hydroxy-3-methoxybenzoic acid), in veratric acid (3,4-dimethoxybenzoic acid) a methoxy group replaces the hydroxyl group, while syringic acid (4-hydroxy-3,5-dimethoxybenzoic acid) has an additional methoxy group.

(a) (b) (c)

Figure 4.1. Chemical structures of: (a) syringic acid; (b) vanillic acid and (c) veratric acid.

Exploring their properties and applications, syringic acid presents strong antibacterial properties (Kong et al., 2008), skin carcinogenesis protective action (Ha et al., 2018), and potential effect on the prevention of diabetic cataract (Wei et al., 2012). In addition, along with vanillic acid, syringic acid also presents potential hepatoprotective effects, suppressing hepatic fibrosis in injured livers (Itoh et al., 2010). Vanillic acid also presents

154 Solubility of target molecules CHAPTER 4 biological properties: specific inhibitor of the snake venom enzyme 5’-nucleotidase (Dhananjaya et al., 2009), potential use to regulate chronic intestinal inflammation (Kim et al., 2010), and prevention of obesity by activating thermogenesis in brown and white adipose tissues (Han et al., 2018). Veratric acid can be used to treat skin disorders due to its ability of protecting UVB-induced skin injuries (Shin et al., 2013). Additionally, studies show that veratric acid has significant antihypertensive, anti-inflammatory and antioxidant properties (Choi et al., 2012; Saravanakumar et al., 2015; Saravanakumar and Raja, 2011). In this work, the solubility of the selected acids was measured in water and organic solvents (methanol, ethanol, 1-propanol, 2-propanol, 2-butanone, ethyl acetate, acetonitrile, dimethylformamide – DMF, 1,2-propanediol, 1,3-propanediol and 1,3-butanediol) at 298.2 K and 313.2 K. For many of the studied binary systems, no solubility data have been reported yet, but whenever possible, the new solubility data were critically compared to the data available in literature. Besides a few studies concerning veratric (Bowen et al., 2013; Li et al., 2013b), vanillic (Noubigh et al., 2008a, 2007; Noubigh and Abderrabba, 2016; Zhang et al., 2016b), and syringic acids (Noubigh et al., 2008b; Noubigh and Akermi, 2017; Queimada et al., 2009) solubility, other works focused on aromatic acids with similar structures, such as o-anisic acid (2-methoxybenzoic acid) (Apelblat and Manzurola, 1998), p-anisic acid (4-methoxybenzoic acid) (Apelblat and Manzurola, 1998), 2,4- dimethoxybenzoic acid (Tian et al., 2015), 3,5-dimethoxybenzoic acid (Wang and Zhang, 2009) and 3,4,5-trimethoxybenzoic acid (Erin Hart et al., 2018). In order to obtain a more complete picture of the solid-liquid equilibrium (SLE), the melting properties of the solutes were measured by Differential Scanning Calorimetry (DSC) as well as solid phase studies were carried out by X-Ray Diffraction (XRD). The empirical selection of solvents only is quite restrictive, as it requires substantial experimental work and high cost (Sheikholeslamzadeh and Rohani, 2012). An effective alternative is the use of predictive and semi-predictive thermodynamic models to estimate SLE diagrams. In this context, the NRTL-SAC model proposed by Chen and Song (2004) has shown reliable results for predictions of solubility of phenolic compounds in water and organics solvents (Feng et al., 2014; Mota et al., 2012, 2009; Sheikholeslamzadeh and Rohani, 2012; Vilas-Boas et al., 2018). Alternately, due to the high uncertainty of the

155 Solubility of target molecules CHAPTER 4 melting enthalpies of the studied acids, the reference solvent approach (RSA) was combined with the NRTL-SAC model (Abildskov and O’Connell, 2003; Mota et al., 2012) under the same conditions as mentioned before. A third approach followed in this work was the application of the Abraham solvation model (Abraham, 1993; Abraham et al., 2010, 2004), which has already been successfully employed to predict SLE for several substituted benzoic acids (Acree et al., 2017c, 2017b, 2017a; Charlton et al., 2005; Daniels et al., 2003; Hoover et al., 2005, 2004b, 2004a; Stoval et al., 2005; Ye et al., 2011) in different organic solvents.

4.1.2. Experimental

4.1.2.1. Chemicals In the solubility measurements of aqueous systems, ultrapure water (resistivity of 18.2 MΩ.cm, free particles ≥ 0.22 μm and total organic carbon < 5 μg.dm-3) was used. The solids were kept in a desiccator to avoid water contamination. All the compounds were used as received from suppliers and are listed in Table 4.1.

Table 4.1. Mass purity (%) and source of the organic compounds used in this work.

Component Mass Purity (%) Source

Syringic Acid 99.1a Alfa Aesar

Vanillic Acid 99.2a Merck KGaA

Veratric Acid 99.2a Acros Organics

Methanol ≥ 99.9 Honeywell

Ethanol ≥ 99.9 Carlo Erba

Isopropanol ≥ 99.8 Honeywell

1-Propanol ≥ 99.5 Carlo Erba

2-Butanone ≥ 99.5 Sigma Aldrich

Ethyl Acetate ≥ 99.7 Carlo Erba

Acetonitrile ≥ 99.9 Sigma Aldrich

DMF ≥ 99.9 Carlo Erba

1,2-Propanediol ≥ 99.5 Sigma Aldrich

1,3-Propanediol ≥ 99.8 DuPont

1,3-Butanediol ≥ 99.45 Sigma Aldrich

a The purity was obtained in the certificate of analysis issued by the manufacturer.

156 Solubility of target molecules CHAPTER 4

4.1.2.2. Solubility experiments The solubility experiments were carried out by the isothermal shake-flask method, which is described thoroughly elsewhere (Ferreira and Pinho, 2012; Vilas-Boas et al., 2018). Saturated solutions containing a small amount of solid in excess were prepared in a flask with around 50 ml of solvent. The flasks, covered with aluminum foil to avoid light degradation, were placed on plate stirrers inside a thermostatic bath (maximum temperature deviation of ± 0.1 K). From preliminary experiments, it was found that the minimum stirring and settling times to reach the equilibrium were 30 hours and 12 hours, respectively. After reaching equilibrium, three samples of around 4 cm3 were extracted from the mother solution by using plastic syringes coupled to a polypropylene filter (0.45 μm pore size). The gravimetric method was employed to quantify the solubilities in water, methanol, ethanol, 1-propanol, 2-propanol, 2-butanone, ethyl acetate, acetonitrile and DMF. In these cases, the solubilities were determined by weighting (Denver Instruments, precision of ± 0.1 mg) the collected samples and the reminiscent solid after solvent evaporation. For systems with 1,2-propanediol, 1,3-propanediol and 1,3-butanediol, solvents with higher boiling temperatures, the solubilities were determined by refractometry (Abbemat 500, Anton Paar) with a reproducibility within ± 5x10-5. Each collected sample was first diluted in the same solvent, and the refractive index measured at least three times to calculate the solubility. The calibration curves (R2 ≥ 0.998) used to correlate the concentration with the refractive index were obtained using six standard solutions with known compositions. Since the solubility of veratric acid in water at 298.2 K is lower than 1x10-3 g/g of solvent, the gravimetric method is not suitable to determine the solubility. Alternatively, the solubilities were determined by UV-Vis spectroscopy at 258 nm (T70, PG Instruments). Each sample, after being diluted in a mixture of ethanol-water (65:35 mass ratio), was placed in cuvettes (5 mm optical path) and the absorbance read at least three independent times. As before, a calibration curve (R2 ≥ 0.999) was used to calculate the solubility of veratric acid in water.

157 Solubility of target molecules CHAPTER 4

4.1.2.3. Melting properties Melting temperatures and enthalpies were determined by DSC (204 F1 Phoenix, NETZSCH) using a nitrogen flowing system. Samples from 2 up to 10 mg (± 0.1 mg) were hermetically sealed into aluminum crucibles and placed, along with a reference cell, to be heated or cooled at a rate of 1 K/min or 2 K/min, respectively. The experiments were performed from 293.15 K to 493.15 K for syringic and vanillic acids, and from 293.15 K to 473.15 K for veratric acid, where at least three runs were considered for the final results. To ensure the reliability of the results, an external calibration using 11 eleven compounds (ultra-pure water, 4-nitrotoluene, naphthalene, benzoic acid, diphenylacetic acid, indium, anthracene, tin, caffeine, bismuth and zinc) was performed for the two transition properties. In all cases, the onset value was considered as the melting temperature.

The phase change was also investigated by a visual method, which makes easier to identify any eventual degradation. Each sample was placed in an automatic glass capillary device (M-565 Büchi, 50-60 Hz, 150 W, temperature resolution: 0.1 K) and heated at the same conditions as employed for the DSC measurements. The melting temperatures were registered when the last crystal disappeared and the observed visual changes were recorded. For each compound, three independent measurements were also carried out.

4.1.2.4. Solid-phase studies

Samples The pure acids from the manufacturer as well as the solids crystallized after evaporation of a set of selected solvents were analyzed by powder or single X-Ray diffraction.

Powder and single X-ray diffraction Powder XRD data were collected on a X’Pert MPD Philips diffractometer, using Cu-Ka radiation (λ = 1.5406 Å), with a curved graphite monochromator, a set incident area of 10 mm2, and a flat plate sample holder, in a Bragg–Brentano para-focusing optics configuration. Intensity data were collected by the step counting method (step 0.02o and time 5 s) in the range 5o < 2θ < 50o.

158 Solubility of target molecules CHAPTER 4

The cell parameters of suitable crystals of the solutes provided from suppliers as well the samples obtained after crystallization from water, methanol, acetonitrile, dimethylformamide and 2-butanone solvents were determined on a Bruker SMART Apex II diffractometer equipped with a CCD area detector, with monochromated Mo-Kα radiation (λ = 0.71073 Å) and operating at 150(2) K. The selected crystals were placed at 40 mm from the CCD and the spots were measured using different counting times (varying from 10 to 80 s).

4.1.3. Thermodynamic modeling

4.1.3.1. The NRTL-SAC model Due to its semi-predictive nature and successful application in previous studies (Chen and Crafts, 2006; Ferreira and Pinho, 2012; Mota et al., 2012, 2009; Sheikholeslamzadeh and Rohani, 2012; Vilas-Boas et al., 2018), the NRTL-SAC model was chosen to describe the solubility of the solutes studied in this work in pure solvents. A detailed description of the model, with its fundaments and equations, is presented elsewhere (Chen and Crafts, 2006; Chen and Song, 2004). Briefly, each molecule is described by four conceptual segments (hydrophobic X, hydrophilic Z, polar attractive Y+, and polar repulsive Y-), representing different molecular surface interactions. For 63 organic solvents, these parameters are already available in the literature (Chen and Crafts, 2006; Chen and Song, 2004). Therefore, only the molecular descriptors of the solute are required to calculate its solubility in a pure solvent. Assuming pure solid phase and that the triple point can be replaced by the melting point, and neglecting the heat capacities change, the solubility of a solid solute in a liquid solvent can be determined by the following equation (Prausnitz et al., 1999):

훥푚퐻 푇푚 ln 푥푆 = (1 − ) − ln 훾푆 [4.1] 푅푇푚 푇 where 푥푆 is the mole fraction solubility of the solute S, R is the ideal gas constant, T is the absolute temperature, Tm is the melting temperature of the solute, 훥푚퐻 its melting

159 Solubility of target molecules CHAPTER 4

enthalpy, and 훾푆 is the activity coefficient of the solute S, calculated in this case using the NRTL-SAC model (Chen and Song, 2004). In equation [4.1], the accuracy of the melting properties plays a crucial role for the reliability of the estimates. However, in many cases these properties are unavailable or present high uncertainty, which lead us to combine the NRTL-SAC model to the Reference Solvent Approach (RSA) proposed by Abildskov and O’Connell (2005, 2003). This methodology can be briefly described by the following equation:

ln 푥푆푖 = ln 푥푆푗 + ln 훾푆푗(푇, {푥푆}푗) − ln 훾푆푖(푇, {푥푆}푖) [4.2]

where 푥푆푖 is the mole fraction solubility of the solute S in a pure solvent i, 푥푆푗 is the solubility of the same solute in a pure reference solvent j, 훾푆푖(푇, {푥푆}푖) is the activity coefficient of the solute in solvent i, while 훾푆푗(푇, {푥푆}푗) is the activity coefficient of the solute in the reference solvent j and T is the temperature of the system.

In this methodology, the experimental solubility of a solute in a chosen reference solvent is used along with the NRTL-SAC model to determine the other variables in equation [4.2]. For a given set of data, the optimal reference solvent is found by:

푚푖푛푗 | ∑ 훿 푙푛푥푆,푖푗| = 푚푖푛푗 | ∑ (푙푛푥푆푖 + 푙푛훾푆푖) − 푁(푙푛푥푆푗 + 푙푛훾푆푗)| [4.3] 푖=푑푎푡푎 푖=푑푎푡푎 where ∑푖=푑푎푡푎 훿 푙푛푥푆,푖푗 is the error associated to the mole fraction solubilities of solute Ss in all the solvents assuming a reference solvent j and N is the number of the experimental data points in a given set.

4.1.3.2. The Abraham solvation model The Abraham solvation model is based on two linear free energy relationships (LFERs), as described in detail by Abraham and co-workers (Abraham, 1993; Abraham et al., 2010, 2004). One LFER quantifies the solute partition between two condensed phases (Eq. 4.4) and, the other, the partition between a gas phase and an organic solvent (Eq. 4.5):

푙표푔(푃푆) = 푐 + 푒퐸 + 푠푆 + 푎퐴 + 푏퐵 + 푣푉 [4.4]

160 Solubility of target molecules CHAPTER 4

푙표푔(퐾푆) = 푐 + 푒퐸 + 푠푆 + 푎퐴 + 푏퐵 + 푙퐿 [4.5]

In these equations, the uppercase descriptors (E, S, A, B, V and L) represent the Abraham solute descriptors, where E is the solute excess molar refractivity, S refers to the solute dipolarity/polarizability, A and B account for the overall solute hydrogen bond acidity and basicity, V is the solute’s McGowan characteristic molecular volume and L is the logarithm of the gas-to-hexadecane partition coefficient at 298 K. The lower case regression coefficients and constants represent condensed phase properties, already established for a large number of solvents.

For the solubility calculations, only the first LFER quantifying the solute transfer between the organic solvent and water was considered in this work. In that case, the partition between water and a solvent (푃푆) is defined as the ratio between the molar solubilities in the organic solvent (푆푠) and in water (푆푤) as follows:

푆푠 푃푆 = [4.6] 푆푤

As discussed by Abraham et al. (2010), the application of Equation [4.6] is subject to some constraints, namely: (a) the solid phase in equilibrium with both water and organic solvent is the same; (b) the secondary medium activity coefficient of the solute in the two phases is near unity ; the same (undissociated, if ionizable) chemical species should be present in each phase.

Excepting two solvents (1,3-propanediol and 1,3-butanediol), the Abraham descriptors of the solvents studied in this work are available in recent literature (Abraham et al., 2015, 2010; Stovall et al., 2016). Therefore, using a given set of solubility data in solvents with high chemical diversity, the solute’s coefficients can be estimated by multiple linear regression analysis. Regarding the solutes, for veratric acid the descriptors were already estimated by Bowen et al. (2013). For all the solutes, the descriptor V can be calculated independently by the molecule structure, as described elsewhere (Abraham and McGowan, 1987). In addition, parameter E can also be calculated from the solute’s refractive index, which can be experimentally obtained for liquid solutes or estimated using ACD free software for solid solutes, as described in detail by Abraham et al. (2004).

161 Solubility of target molecules CHAPTER 4

4.1.4. Results and discussion

4.1.4.1. Experimental solubilities The measured solubilities of syringic acid, veratric acid and vanillic acid in water, methanol, ethanol, 1-propanol, 2-propanol, ethyl acetate, 2-butanone, acetonitrile, DMF, 1,2- propanediol, 1,3-propanediol and 1,3-butanediol at 298.2 K and 313.2 K are presented in Table 4.2.

Table 4.2. Experimental solubility (g of solute/100 g of solvent) in water and organic solvents at 298.2 K and 313.2 K.a,b Syringic Acid Vanillic Acid Veratric Acid Solvent 298.2 K 313.2 K 298.2 K 313.2 K 298.2 K 313.2 K

Water 0.142 ± 0.001 0.231 ± 0.004 0.128 ± 0.002 0.269 ± 0.005 0.050 ± 0.001 0.098± 0.002

Methanol 11.480 ± 0.036 16.237 ± 0.028 18.264 ± 0.035 23.607 ± 0.012 4.424 ± 0.026 7.767 ± 0.037

Ethanol 5.562 ± 0.002 8.006 ± 0.019 11.947 ± 0.015 16.329 ± 0.063 3.051 ± 0.020 5.396 ± 0.006

2-Propanol 2.294 ± 0.010 3.739 ± 0.002 7.009 ± 0.016 9.884 ± 0.124 1.992 ± 0.006 3.682 ± 0.005

1-Propanol 2.593 ± 0.008 4.083 ± 0.008 6.435 ± 0.009 8.716 ± 0.017 1.992 ± 0.012 3.875 ± 0.009

2-Butanone 2.658 ± 0.012 3.665 ± 0.006 5.518 ± 0.026 7.249 ± 0.009 3.019 ± 0.024 4.734 ± 0.005

Ethyl Acetate 1.006 ± 0.008 1.446 ± 0.005 2.193 ± 0.007 3.110 ± 0.006 1.498 ± 0.018 2.411 ± 0.002

Acetonitrile 0.951 ± 0.009 1.665 ± 0.003 1.383 ± 0.018 2.165 ± 0.004 1.483 ± 0.081 2.656 ± 0.011

DMF 66.963 ± 0.009 81.527 ± 0.087 88.945 ± 0.220 92.413 ± 0.057 52.024 ± 0.096 63.17 ± 0.017

1,2-Propanediol 3.894 ± 0.047 5.808 ± 0.054 7.514 ± 0.129 10.205 ± 0.151 1.872 ± 0.036 3.088 ± 0.079

1,3-Propanediol 4.353 ± 0.028 6.193 ± 0.045 7.377 ± 0.096 9.513 ± 0.038 1.434 ± 0.054 2.201 ± 0.046

1,3-Butanediol 3.757 ± 0.026 5.784 ± 0.033 6.643 ± 0.149 9.967 ± 0.058 1.607 ± 0.038 2.921 ± 0.018 a:Temperature and pressure standard uncertainties are u(T) = 0.10 K and ur(p) = 0.05, respectively; b: Standard deviations are placed after plus-minus sign.

Each data point presented in Table 4.2 is an average of at least 3 independent measurements. The low variation coefficients, always inferior to 3.8 %, indicate that the standard deviations are much lower than the solubility values. The solubility of the acids is, in general, considerably higher in organic solvents than in water. In nine of the studied binary systems (water, methanol, ethanol, 1-propanol, 2-propanol, DMF, 1,2-propanediol, 1,3-propanediol and 1,3-butanediol), the solubility at 298.2 K ranks: vanillic acid > syringic

162 Solubility of target molecules CHAPTER 4 acid > veratric acid. With the exception of DMF, all those solvents present at least one hydroxyl group, capable of forming stronger hydrogen bonds with the carboxylic group, and also to the hydroxyl groups, which are more accessible in vanillic acid, while veratric acid only presents two hydrogen bond acceptor methoxy substituents.

For ethyl acetate and 2-butanone, vanillic acid is still the most soluble one, but less polar veratric acid presents higher solubilities than syringic acid. In acetonitrile, the most soluble compound is veratric acid, followed by syringic acid, and finally vanillic acid. In fact, the importance of hydrogen bonds is now weaker (the solvents contain only hydrogen bond acceptor groups) than in the previous set of solvents, contributing to the changes observed in the solubility order.

In the case of systems containing alcohols, the solubility decreases with an increase of the alky chain of the alcohol. Surprisingly, much higher solubilities were found for systems with DMF, behavior that was also observed for dihydroxybenzoic acids (Vilas-Boas et al., 2018). In all cases, the solubility increases with temperature.

4.1.4.2. Data analysis Whenever possible, the solubilities of the selected acids were compared to literature data. In Figure 4.2, a comparison of the solubilities of syringic, vanillic and veratric acids in water and ethanol is presented. A complete overview of this coherence analysis is presented in Table C1 and in Figure C1 to Figure C4 of Appendix C. Figure 4.2 shows that in general the solute solubility in water and ethanol agrees with the data found in literature, presenting deviations lower than 1.0 g/100 g of solvent, with the exception of the system syringic acid + ethanol, where the solubility measured by Noubigh and Akermi (2017) is much higher, while in water is 3 to 4 times higher than the solubility measured in this work and by Queimada et al. (2009). Due to the higher disagreement observed in the systems containing ethyl acetate and ethanol, solubility experiments were here performed twice, reaching always very similar solubility values between the two independent measurements.

163 Solubility of target molecules CHAPTER 4

a) b)

)

c) d)

)

e) f)

Figure 4.2. Comparison of the experimental solubility data obtained in this work with literature (Bowen et al., 2013; Li et al., 2013b; Noubigh et al., 2008b, 2008a; Noubigh and Abderrabba, 2016; Noubigh and Akermi, 2017; Queimada et al., 2009; Zhang et al., 2016b) data: (a) syringic acid + water; (b) syringic acid + ethanol; (c) vanillic acid + water; (d) vanillic acid + ethanol; (e) veratric acid + water; (f) veratric acid + ethanol.

For vanillic acid, the solubilities in water (0.128 g/100 g of water at 298.2 K and 0.269 g/100 g of water at 313.2 K) measured in this work are quite similar to the literature average values (0.151 g/100 g of water and 0.271 g/100 g of water at 298.2 K and 313.2 K, respectively). However, the differences concerning the solubility in some organic solvents are relevant, especially at 313.2 K. Solubility experiments were repeated for systems containing ethyl acetate, methanol and 2-propanol and no significant changes were observed.

Particularly, the solubility of vanillic acid in methanol published by Noubigh and Abderrabba (2016) is much higher than the values found in this work. In that study, the

164 Solubility of target molecules CHAPTER 4 solubility of vanillic acid in methanol at 298.2 K is 7.1 times higher than the solubility of the same compound in ethanol, which is quite unexpected. For instance, in previous studies (Noubigh et al., 2015; Noubigh and Akermi, 2017) the same group reported, for syringic and protocatechuic acids, solubility values 1.8 and 1.6 times higher in methanol than in ethanol, respectively.

For veratric acid, the solubility in water at 298.2 K (0.051 g/100 g of solvent) measured in this work is comparable to the value reported by Bowen and coworkers (Bowen et al., 2013) (0.058 g/100 g of solvent). In organic solvents, the solubilities obtained in this work are also in good agreement with the data reported by Li et al. (2013b) and Bowen et al. (2013). In the first case, the authors employed the robust laser monitoring method, that dynamically detects the temperature at which the last solid particles disappears, eliminating the issues related to the chemical analysis, stirring and settling times.

4.1.4.3. Melting properties The melting temperatures and enthalpies found in literature (Booth et al., 2012; Li et al., 2013b; Manic et al., 2012; Queimada et al., 2009; Reggane et al., 2018; Thipparaboina et al., 2016; Zhang et al., 2016b) along with the values measured in this work by DSC, and by the visual method, are summarized in Table 4.3. Exemplificative thermograms for the DSC measurements are shown in Figure C5 and Figure C6 (Appendix C).

The melting temperatures obtained by DSC in this work agree very well with the values from the literature. The melting temperatures measured by the visual method are slightly higher than the values measured by DSC, probably due to the employed methodologies. While in the visual method the melting temperature was considered as the point in which the last crystal disappears, the onset temperatures were registered in the DSC measurements.

The melting enthalpies from this work, however, present less consistency with the values reported by other authors, being larger for all the studied solutes. This fact could be related to possible degradation, as discussed below.

165 Solubility of target molecules CHAPTER 4

Table 4.3. The melting temperatures and enthalpies of the studied acids from the literature and measured in this work. a -1 a Compound Tm/K ΔmH/kJ·mol Methodology Reference 480.3 ± 0.6 33.7 ± 1.8 DSC (Queimada et al., 2009) 482.5 28.1 DSC (Booth et al., 2012) Syringic Acid 481.6 ------DSC (Thipparaboina et al., 2016) 482.3 ± 0.1 ------VM This Work 480.9 ± 0.3a 40.3 ± 0.6 DSC This Work

484.9 ± 0.2 32.8 ± 0.1 DSC (Zhang et al., 2016b)

484.7 25.6 DSC (Booth et al., 2012) Vanillic Acid 480.7 ± 0.2 29.1 ± 0.6 DSC (Manic et al., 2012) 481.15 ------DSC (Reggane et al., 2018) 484.9 ± 0.12 ------VM This Work 483.3 ± 0.3a 34.63 ± 0.8 DSC This Work

453.1 ± 0.2 29.6 ± 0.1 DSC (Li et al., 2013b) Veratric Acid 455.75 ± 0.17 VM This Work 453.5 ± 1.3a 34.64 ± 0.3 DSC This work a: The experimental onset temperatures were considered as melting temperatures in this work; VM: visual method.

DSC experiments detected that syringic acid presents a solid-solid transition at (462.6 ± 1.1) K, before reaching the fusion transition. This behavior is reproducible for runs with independent samples, but it was not observed in experiments with successive runs. The visual method gives some insights on possible degradation of syringic acid upon melting, due to a slight modification in the solid coloration from white to a light yellow.

In the case of vanillic acid, the melting enthalpy reported by Manic et al. (2012) is closer to the value obtained by us. Nevertheless, possible decomposition of the sample was identified by the visual method starting at 481.6 K. An illustrative image of this phenomena is shown in Figure C7 of Appendix C.

From the three solutes, veratric acid was the only one which did not present any modification on the fusion peak shapes over successive runs or any visual degradation. For this compound, a transition was identified at (422.9 ± 0.4) K, reproducible when the heating rate was set as 1 K/min. Even so, the melting enthalpy measured is 5 kJ/mol higher than that measured by Li et al. (2013b).

166 Solubility of target molecules CHAPTER 4

4.1.4.4. Solid phase studies Solid samples of syringic, vanillic and veratric acids as received from suppliers and those recrystallized from the saturated solutions of water, methanol, 2-butanone, acetonitrile and dimethylformamide were analyzed by powder and single crystal X-ray diffraction.

From a single crystal, it was found that syringic acid received from the supplier as well syringic acid recrystallized from 2-butanone, acetonitrile and dimethylformamide crystalized in monoclinic C system with cell parameters of a = 26.64 Å, b = 4.14 Å, c = 15.84 Å, β = 96.30°, which are comparable to those published in CCDC database (CCDC number: 1450484) (Thipparaboina et al., 2016). Figure C8 (Appendix C) shows the comparison of the experimental powder X-ray diffraction of the syringic acid and the samples obtained from the solutions of 2-butanone, acetonitrile and dimethylformamide which all are similar to the powder pattern calculated from the single-crystal X-ray diffraction data of syringic acid published by Thipparaboina et al. (2016) with CCDC=1450484. The samples of syringic acid recrystallized from water and methanol show X-ray powder patterns similar to syringic acid from supplier; however, the crystallinity is too low.

Suitable crystals of vanillic acid recrystallized from water were analyzed by single crystal X- ray diffraction exhibiting the following cell parameters: a = 3.90 Å, b = 17.69 Å, c = 11.38 Å, and β = 95.21°, which are comparable to those published in CCDC database (CCDC number: 277423) (Kozlevčar et al., 2006). The comparison of the powder X-ray diffraction patterns of all the vanillic acid samples shown in Figure C9 (Appendix C) reveals that they are similar to the powder pattern calculated from the structure solved from single-crystal X-ray diffraction by Kozlevčar et al. (2006) (CCDC number of 277423).

Likewise, all samples of veratric acid studied in this work crystallized in the triclinic system P showing cell parameters of a = 4.89 Å, b = 8.52 Å, c = 11.36 Å, α = 101.53°, β = 101.78° and γ = 105.83°, comparable to those reported in CCDC number of 207337 and published in 2002 by Pinkus et al. (2002). The experimental powder X-ray diffraction of the veratric acid samples obtained from the solutions of water, methanol, acetonitrile, dimethylformamide and 2-butanone (Figure C10 of Appendix C) are similar to the powder pattern calculated from the single-crystal X-ray diffraction data of veratric acid with CCDC

167 Solubility of target molecules CHAPTER 4 number of 207337 (Pinkus et al., 2002). However, for the experimental powder patterns the intense peaks at (15.61 and 26.85)/2θ, are split.

4.1.4.5. Thermodynamic Modelling

NRTL-SAC and NRTL-SAC + RSA The model calculations were performed using the MATLAB software version R2013a in order to optimize the NRTL-SAC segment descriptors for syringic acid, vanillic acid, and veratric acid. The optimization algorithm was the MATLAB routine Isqnonlin, which is based on the nonlinear least-squares curve fitting of the objective function. In this work, the objective function (F) was:

푒xp cal푐 2 푥푖 − 푥푖 퐹 = ∑ ( 푒xp ) [4.7] 푥푖 푖 where 푥푖 is the mole fraction solubility in the solvent i and the upper scripts “exp” and “calc” mean the experimental and calculated values, respectively.

To begin, the experimental solubilities in seven solvents (water, methanol, ethanol, 2- propanol, ethyl acetate, acetonitrile and 2-butanone) were used to correlate the four conceptual segment parameters (X, Y+, Y-, Z) for each solute. After, those were used to predict the solubility in 2-propanol and DMF. The average relative deviation (ARD) was calculated for each binary system as:

exp calc 1 |푥푖 − 푥푖 | 퐴푅퐷(%) = ∑ exp ∗ 100 [4.8] 푁푃 푥푖 푖 where NP is the number of data points.

As mentioned before, the high uncertainty observed in the fusion properties, particularly regarding the enthalpies, can lead to difficulties or even hamper the parameter estimation. For that reason, the RSA proposed by Abildskov and O’Connell (2005, 2003) was combined with the NRTL-SAC model. The optimized molecular descriptors of the solutes obtained for both correlation approaches (NRTL-SAC or NRTL-SAC + RSA) as well as the reference solvent

168 Solubility of target molecules CHAPTER 4

(when applicable), the outlier solvent (presenting the highest ARD) and the global ARD are presented in Table 4.4.

Table 4.4: NRTL-SAC optimized parameters, reference solvent, outlier solvent and ARD (%) for each acid using water and six organic solvents in the fitting. Compound X Y- Y+ Z Reference Solvent Outlier ARD (%)

NRTL-SAC

Syringic acid 0.287 0.840 0.642 0.000 ----- Methanol 41

Vanillic acid 0.424 0.885 0.304 0 ----- Methanol 38

Veratric acid 0.495 0.000 0.202 0.531 ----- 2-Butanone 20

NRTL-SAC + RSA

Syringic acid 0.496 0.000 0.400 1.625 Methanol Ethyl acetate 35

Vanillic acid 0.587 0.059 0.000 1.405 Acetonitrile Ethyl acetate 33

Veratric acid 0.491 0.000 0.239 0.561 Ethanol 2-Butanone 22

The results above show that NRTL-SAC is an adequate model to correlate the solubility of the studied compounds, presenting ARDs varying between 20 and 41 %, which is very satisfactory for a semi-predictive model with a reduced number of parameters to be estimated for each solute. The combination of the NRTL-SAC model with RSA also showed reliable correlation, resulting in lower ARD for syringic and vanillic acids. In previous works, the authors have applied the NRTL-SAC model to predict the solubility of drug molecules in water and organic solvents (Mota et al., 2012, 2009; Vilas-Boas et al., 2018), obtaining ARD similar to those found in this work. Figure 4.3 relates the calculated and experimental solubility data for the correlations performed by NRTL-SAC and NRTL-SAC + RSA.

The NRTL-SAC segment descriptors obtained by the conventional approach and with the use of RSA are in closer agreement for veratric acid, the only compound that showed good thermal stability upon melting.

169 Solubility of target molecules CHAPTER 4

Figure 4.3. Comparison between experimental and calculated solubility for correlation: a) NRTL- SAC; b) NRTL-SAC + RSA.

As can be seen in Figure 4.3, the NRTL-SAC and NRTL-SAC + RSA describe well the solubility in water (maximum ARD of 22 %). The results obtained using NRTL-SAC only present higher ARDs for the systems containing methanol and 2-butanone. For a few cases, the model is limited in giving the exact order of magnitude of the solubilities, leading to lower calculated values. On the other hand, the correlations performed using NRTL-SAC + RSA represent better the solubilities in alcohols, but delivering higher ARDs in the cases of ethyl acetate or 2-butanone.

After, solubility predictions in 1-propanol and DMF were carried out. The obtained ARDs were 22 % and 15 % for 1-propanol, and 94 % and 73 % for DMF, applying NRTL-SAC or NRTL-SAC + RSA, respectively. Although the difference was not so pronounced in the correlation step, the ARDs found are considerably lower when the NRTL-SAC + RSA estimated parameters are used.

The prediction analysis was also extended to the binary systems previously studied by different authors (Bowen et al., 2013; Noubigh and Abderrabba, 2016; Noubigh and Akermi, 2017; Zhang et al., 2016b). In Figure 4.4, a summary of predictions found using NRTL-SAC and NRTL-SAC + RSA is presented.

170 Solubility of target molecules CHAPTER 4

1 1 a b

0,1

0,1 calc

0,01 calc

x x

0,01 0,001

0,0001 0,001 0,0001 0,01 1 0,001 0,01 0,1 1 exp xexp x

Figure 4.4. Comparison between experimental and calculated solubility data for prediction: a) NRTL-SAC; b) NRTL-SAC + RSA.

Regarding the NRTL-SAC predictions, the highest ARDs are observed for ethylene glycol, isobutanol and dimethylformamide, whereas the model provides good estimates for the solubilities in propyl acetate and for alcohols. In the case of NRTL-SAC + RSA, the three highest predictions are the systems with tetrahydrofuran, 1,4-dioxane and 2-butanol. All the systems with other alcohols and propyl acetate are also well described by this methodology. For both, NRTL-SAC and NRTL-SAC + RSA approaches, the ARD for all the predictions is 59 %, and the models generally underestimate the calculated solubilities.

4.1.4.6. Abraham solvation model Since no modification in the crystalline structures was observed by X-ray diffraction and the measured solubilities were generally not too high, the Abraham solvation model was applied to describe the SLE of the studied acids. To simultaneously solve the set of LFER (equation 4.4) for each solute, the multiple linear regression model was applied. Initially, the same set of solvents (water, methanol, ethanol, 2-propanol, 2-butanone, ethyl acetate and acetonitrile) employed in NRTL-SAC was selected to correlate the solutes parameters.

171 Solubility of target molecules CHAPTER 4

푒푥푝 All experimental solubilities were converted from mole fraction (푥푆 ) to molar solubilities 푒푥푝 (푆푆 ) applying:

푒푥푝 푒푥푝 푥푆 푆푠 ≈ 푒푥푝 푒푥푝 [4.4] [푥푆 푉푆표푙푢푡푒 + (1 − 푥푆 )푉푆표푙푣푒푛푡]

The molar volumes of the solutes were calculated from the solid densities obtained by crystallographic studies (Kozlevčar et al., 2006; Pinkus et al., 2002; Thipparaboina et al., 2016). The values for syringic, vanillic and veratric acids are 134.0 cm3 mol-1, 115.5 cm3 mol- 1 and 126.9 cm3 mol-1, respectively. For the solvents, the molar volumes were calculated based on the density values found in literature (Table C2 of Appendix C). The obtained ARDs for the correlation step were 21 % for syringic acid, 15 % for vanillic acid and 5 % for veratric acid. Predictions were also performed for the solubility data measured in this work (1- propanol, DMF, 1,2-propanediol) and for the data available in literature (Bowen et al., 2013; Noubigh and Abderrabba, 2016; Noubigh and Akermi, 2017; Zhang et al., 2016b), showing a global ARD of 43 %. For these first set of simulations, the parameters obtained are in Table C3 (Appendix C).

The average deviations are in the expected range for semi-predictive models, but for the solubility in DMF the estimates are really very poor. Therefore, it was decided to test the impact of including the solubility data in DMF during the correlation step. After re- estimating the solute descriptors, the ARD found for each solute was 20 % for syringic acid, 13 % for vanillic acid and 11 % for veratric acid, while for predictions the global ARD, also considering the literature data, was 26 %. Table 4.5 presents the parameters obtained in this second correlation round. The correlation deviations for veratric acid including DMF is slightly higher, but its reduction in the predictions is much more significant (from 43 % to 26 %). Since DMF is structurally different from the other solvents used, its presence in the correlation step provides more robust solute parameters, improving the accuracy of the predictions for a larger number of solvents.

The importance of the chemical diversity of the solvents in the reliability of the fitted parameters has been shown previously (Abraham et al., 2010; Acree et al., 2017b; Bowen et al., 2013). Following this idea, a similar approach was performed for the NRTL-SAC and

172 Solubility of target molecules CHAPTER 4

NRTL-SAC + RSA simulations, incorporating DMF in the parameter fitting. However, no considerable changes were observed neither in the optimized solute descriptors or global ARDs. In Table 4.5, the estimated Abraham solute descriptors (including DMF), global correlation ARD and the outliers are presented.

Table 4.5. Estimated solute parameters in the Abraham solvation model, outlier solvent and ARD (%) for each acid using water and seven organic solvents in the correlation database. Compound E S A B V Outlier ARD (%) Syringic Acid 1.123 1.757 0.808 0.878 1.390 Methanol 20 Vanillic Acid 1.144 1.452 0.846 0.647 1.190 Methanol 13 Veratric Acid 0.881 1.646 0.630 0.733 1.331 2-Butanone 11

To the best of our knowledge, among the acids here studied the Abraham solvation model was only used to describe the SLE of veratric acid (Bowen et al., 2013; Sedov et al., 2015). In this case, the authors employed a set of 54 partition data including condensed phase and gas-liquid data, a considerable larger number than we used in this work. The solute parameters reported by Bowen et al. (2013) are very close to the values presented in Table 4.5. In Figure 4.5 and Figure 4.6, a summary of the correlation and prediction results is presented, relating the calculated solubility to the experimental solubility data.

1 Methanol Ethanol 2-Propanol 2-Butanone 0,1 Ethyl Acetate Acetonitrile DMF

calc 0,01 x

0,001

0,0001 0,0001 0,001 0,01 0,1 1 exp x Figure 4.5. Comparison between experimental and calculated solubility in the organic solvents used in the estimation of the Abraham solute parameters.

173 Solubility of target molecules CHAPTER 4

0,1

0,01

calc x

0,001

0,0001 0,0001 0,001 0,01 0,1 xexp 1-Propanol 1,2-Propanediol 1-Butanol 2-Butanol Isobutanol 1-Pentanol 1-Hexanol 1-Heptanol 1-Octanol 1-Decanol 2-Pentanol 2-Methyl-2-propanol 3-Methyl-1-butanol 2-Methyl-1-butanol Ethylene Glycol Acetone 1,4 Dioxane Diethyl ether Dibutyl ether Methyl Acetate

Figure 4.6. Comparison between experimental and calculated solubility data in the predicted set using the Abraham solvation model.

The outliers in the predictions were 2-butanol and dibutyl ether (ARD of 79 % and 53 % respectively). For all the other predicted solvents, the ARD found were inferior to 40 %. Particularly, for 1,2-propanediol, the Abraham solvation model provides reasonable predictions (ARD of 39 %) even though no diol was used in the fit. In general, the model describes well the SLE of binary systems containing alcohols (ARD of 20 %) and esters (ARD of 19 %).

Despite the Abraham solute parameters describe quite well the SLE of a large number of binary systems, it is relevant to analyze the descriptor’s values to check if they reflect the chemical properties of the solute. Hoover and coworkers (Hoover et al., 2004b) previously correlated the Abraham solute’s descriptors for 2-methoxybenzoic acid (S = 1.410, A = 0.450 and B = 0.620) and 4- methoxybenzoic acid (S = 1.250, A = 0.620 and B = 0.520) by using a similar approach employed in this work (the authors used solubility data in alcohols and in ethers to fit the solute parameters).

For syringic, vanillic and veratric acids, the parameters S, A and B are higher than the parameters reported for 2-methoxybenzoic and 4-methoxybenzoic acids, even though the value order changes for each parameter. In the case of the acidity descriptor (A), the presence of hydroxyl groups increases the H-bond acidity of the solute, whereas

174 Solubility of target molecules CHAPTER 4 intramolecular hydrogen bonds tend to reduce it. Vanillic acid (with one hydroxyl and one methoxy substituents attached to the aromatic ting) presents the highest value for A, followed by syringic acid (one hydroxyl and two methoxy substituents) and veratric acid (two methoxy groups). The close proximity of the hydroxyl and methoxy groups in syringic and vanillic acids likely leads to the formation of intramolecular H-bonds, being stronger for syringic acid.

The absence of a hydroxyl-substituent in veratric acid reduces its acidity compared to the other two studied phenolic acids. The correlated value for this compound (A = 0.630) is quite close to the value found for 4-methoxybenzoic acid but considerably higher than the value for 2-methoxybenzoic-acid. The methoxy-substituents in the aromatic ring should increase the electron density of the aromatic ring through resonance, increasing the H- bond acidity of the carboxyl group. The presence of a methoxy group in the position 2 of the ring, however, can also lead to intramolecular hydrogen-bond formation, reducing the acidity of this acid compared to veratric acid and 4-methoxybenzoic acid.

Regarding the solute’s basicity, the values of B progressively increase in the following order: syringic acid > veratric acid > vanillic acid. In the case of syringic acid, the presence of two methoxy-substituent and one hydroxyl-substituent groups leads to a higher number of available lone electron pairs on the oxygen atoms than it is observed for the other two acids. In the case of vanillic acid, the presence of intramolecular H-bonds should decrease the basicity of the substituents (methoxy and hydroxyl groups). As expected, the basicity parameters reported for 2-methoxybenzoic acid and 4-methoxybenzoic acid were lower than the values correlated to syringic, vanillic and veratric acids, which is probably related to the lower electron density present in the monosubstituted methoxybenzoic acids.

Even if the parameters give good consistency with the chemical structure of the solutes, it is relevant to mention that they are an average representation of the different conformations of the solute in each solvent, which are also different among all the solvents used for correlation (Hoover et al., 2004b).

175 Solubility of target molecules CHAPTER 4

4.1.5. Conclusions In this work the solubility of syringic acid, vanillic acid and veratric acid was measured in water and in eleven organic solvents at 298.2 and 313.2 K using the isothermal shake-flask method. For all the studied binary systems, an increment on the solubility was observed when increasing the temperature. On all the possible occasions, the measured solubility was compared to the data available in literature. In the case of veratric acid, the solubilities measured in this work are in good agreement with the literature data. For vanillic and syringic acids, some inconsistencies were found.

Melting properties measurements of the three acids were also carried out by DSC and by a visual capillary method. For all acids, the melting temperatures determined by DSC strongly agree with the literature. The visual method gave some indications about degradation of syringic and vanillic acids upon the melting, which probably explains the higher uncertainties observed for the melting enthalpies. Additionally, solid-solid transitions, with small phase change enthalpies were identified for syringic acid, at (462.6 ± 1.1) K, and veratric acid, at (422.9 ± 0.4) K.

From the solid phase studies, it was possible to conclude that crystals of syringic, vanillic and veratric acids obtained from the manufacturer and after the evaporation from selected solvents (water, methanol, 2-butanone, acetonitrile and DMF) are comparable to structures previously reported in the CSD-system (Kozlevčar et al., 2006; Pinkus et al., 2002; Thipparaboina et al., 2016).

NRTL-SAC was successfully applied to describe the SLE of the selected solutes in aqueous and in organic binary systems, showing global ARD of 33 % for correlation and 59 % for prediction. For this set of systems, the combination of NRTL-SAC with the reference solvent approach did not introduce any significant improvement. The Abraham solvation model was also studied in the correlation and prediction of the solubility in organic solvents at 298.2 K. Using solubility data for the same set of solvents used in the estimation of NRTL- SAC solute parameters, plus DMF, the model descriptors S, A and B were fit by multilinear regressions. The global ARD in the correlation and prediction steps were 15 % and 26 %, respectively.

176 Solubility of target molecules CHAPTER 4

Generally, the NRTL-SAC and Abraham models provided satisfactory results, allowing to estimate the solubility order of magnitude in several solvents by using a limited set of experimental data. However, Abraham solvation model provides more accurate solubility values than NRTL-SAC. On the other hand, the NRTL-SAC ability to calculate the variation of the solubility with temperature is a very important feature.

177

5. Conclusions and future work

Conclusions and future work CHAPTER 5

This thesis aimed to explore sustainable strategies for the extraction of added-value compounds from natural sources, especially phenolic compounds, using alternative solvents to the conventional volatile organic compounds.

Firstly, the phytochemical profile and bioactive properties of Juglans regia L. (walnut) leaves and green husks were studied using aqueous ethanolic mixtures as solvents, as pure water is a poor solvent for many phenolic compounds. Phenolic acids, flavonoids, tetralone and naphthalene derivatives were extracted, identified and quantified as described in Chapter 2.

In the case of the walnut leaves, the maceration extraction was compared with the microwave assisted extraction, using conventional hydroethanolic mixtures as solvent and the extraction conditions of both systems were optimized using an experimental design assisted by a response surface methodology. The global optimum conditions for the extraction time, temperature and ethanol content (v/v, %) were 115.6 min, 61.3 C and 50.4 % for ME; and 3.0 min, 107.5 C and 67.9 % for MAE. The MAE technique proved to be a much faster alternative for the extraction of the target compounds, at higher temperature, with similar or higher extraction yields when compared to the conventional maceration process. This is a very important result as, for many applications, a dry extract is needed as a final product.

When considering the walnut leaves, the impact of seasonal variation in their phytochemical profile was evaluated applying the previous optimized conditions of extraction for the maceration technique. The extracts of green leaves contained a higher amount of total phenolic compounds and phenolic acids, while yellow samples were richer in flavonoids. Overall, green leaves showed higher antioxidant and antibacterial activities than the yellow samples, though the major difference found was in the results of the anti- inflammatory activity as only the extracts of green leaves were bioactive in this dimension.

While studying the aqueous ethanolic extract of walnut green husks, remarkable bioactive results were obtained, especially for the antioxidant, anti-inflammatory and cytotoxic potentials, with efficient inhibitions up to concentrations of 100 µg/mL. Furthermore, the study of the phytochemical profile showed a higher abundancy on naphthalene derivatives,

181 Conclusions and future work CHAPTER 5 including tetralone derivatives, and, contrarily to the leaves, the extract was poor in phenolic compounds like phenolic acids and flavonoids. Also, comparing with the leaves extracts from the same geographic origin and time collection, the green husks samples presented higher anti-inflammatory and cytotoxic activities, reflecting the differences found in the phytochemical profile, mainly the presence of naphthalene derivatives (including tetralones).

The search for more efficient and sustainable extraction processes was the goal of the work presented in Chapter 3, using walnut leaves as plant model. First, eutectic mixtures of choline chloride with a series of carboxylic acids were screened. Higher extraction yields were obtained with mixtures of choline chloride and butyric acid or phenylpropanoic acid when compared to conventional solvents such as water, or ethanol. Again, the extraction conditions were optimized by an experimental design assisted by response surface methodology. Under the optimized conditions (180 min of extraction time, 53 wt% of water, 30 °C), higher extraction yields of phenolic compounds were obtained with choline chloride and butyric acid (1:2).

In a second stage, glycerol and a series of alkanediols were studied taking into consideration the potential application of the extract of walnut leaves in the pharmaceutical or cosmetics areas. Again, some of the proposed solvents had a similar performance to ethanol, in terms of the extraction yield of phenolic compounds. Also, C2 up to C4 diols showed satisfactory extraction capacity, and the resulting extracts were active against the assessed tumour cell line, but not against the non-tumour cell culture, being the solvents non-toxic under the studied conditions. In this regard, this study represents a step towards the development of new applications using this type of solvents and extracts. Finally, the evaluation of aqueous mixtures of 1,3-propanediol with choline chloride or betaine as solvents resulted in extracts with significantly higher bioactivity, though similar amounts of phenolic compounds were obtained. In fact, the resulting extracts presented anti-inflammatory potential as well as a higher cytotoxic potential. These results suggest more detailed studies of the phytochemicals present in the extracts, including additional sequential separation and purification steps of the crude extract (e.g. liquid-liquid extractions, different chromatographic elution grades), as other bioactive

182 Conclusions and future work CHAPTER 5 molecules were already reported in the literature beyond phenolics (e.g. diarylheptanoids, ascorbic acid derivatives, alkaloids, terpenes, steroids, etc.). The potential synergistic effect between betaine or choline chloride and the compounds extracted by the solvent should also be studied.

As future work and considering the results discussed above, it would be interesting to evaluate the stability of the extracts obtained with these alternative solvents. In addition, biocompatibility studies of the pure DES remain scarce. Therefore, further analysis to evaluate their cyto- and eco-toxicity should be carried out. Also, the study of the bioactivity of the extracts (obtained with DES and other alternative solvents) should be designed according to the final application of the liquid extract (e.g. the use of skin cell cultures for the cosmetic applications, gastric and metabolic organs cells for food applications, etc.). In the particular case of walnut leaves, the optimization of dermocosmetic formulations using the most promising extracts (DES or pure alkanediols) should be further studied.

Finally, the solubility studies in this thesis were focused in three aromatic acids, representative of an important class of plant metabolites, the phenolic acids, as that information is essential for the design of their extraction and purification processes. The solid-liquid equilibria was modelled using the NRTL-SAC and Abraham models, and satisfactory results were obtained, using only a limited set of experimental data in the correlation step. The Abraham solvation model provides more accurate solubility values than NRTL-SAC, though it does not describe the variation of the solubility with temperature. The extension of these experimental and modelling studies to other phenolic model compounds should be considered in the future, due to their importance in the biorefineries context.

183

6. List of publications

List of publications CHAPTER 6

List of Publications from PhD

Papers in international scientific periodicals with referees

1. Vieira, V., Pereira, P., Pires, T. C. S. P., Calhelha, R. C., Alves, M. J., Ferreira, O., Barros, L., Ferreira, I. C. F. R., 2019. Phenolic profile, antioxidant and antibacterial properties of Juglans regia L. (walnut) leaves from the Northeast of Portugal. Ind. Crops Prod. 134, 347-355.

2. Vilas-Boas, S. M., Vieira, V., Brandão, P., Alves, R. S., Coutinho, J. A. P., Pinho, S. P., Ferreira, O., 2019. Solvent and temperature effects on the solubility of syringic, vanillic or veratric acids: experimental, modeling and solid phase studies. J. Mol. Liq. 289, 111089.

3. Vieira, V., Prieto, M. A., Barros, L., Coutinho, J. A. P., Ferreira, I. C. F. R., Ferreira, O., 2018. Enhanced extraction of phenolic compounds using choline chloride based deep eutectic solvents from Juglans regia L.. Ind. Crops Prod. 115, 261-271.

4. Vieira, V., Prieto, M. A., Barros, L., Coutinho, J. A. P., Ferreira, O., Ferreira, I. C. F. R., 2017. Optimization and comparison of maceration and microwave extraction systems for the production of phenolic compounds from Juglans regia L. for the valorization of walnut leaves. Ind. Crops Prod. 107, 341-352.

187 List of publications CHAPTER 6

Submitted papers

1. Vieira, V., Pereira, C., Abreu, R. M. V., Calhelha, R. C., Alves, M. J., Coutinho, J. A. P.,

Ferreira, O., Barros, L., Ferreira, I. C. F. R., 2019. Phytochemical profile of Juglans regia L. husks: valorization of walnut residues through their bioactive potential Food Chem., submited.

Manuscripts under preparation

1. Vieira, V., Prieto, M. A., Barros, L., Barros, L., Calhelha, R. C., Coutinho, J. A. P., Ferreira, O., Ferreira, I. C. F. R. Insights on the extraction performance of polyols and their versality: using Juglans regia L. Leaves as a source of bioactive compounds.

188

7. Bibliography

Bibliography CHAPTER 7

Abbott, A.P., Boothby, D., Capper, G., Davies, D.L., Rasheed, R.K., 2004. Deep eutectic solvents formed between choline chloride and carboxylic acids: Versatile alternatives to ionic liquids. J. Am. Chem. Soc. 126, 9142–9147.

Abbott, A.P., Capper, G., Davies, D.L., Rasheed, R.K., Tambyrajah, V., 2003. Novel solvent properties of choline chloride/urea mixtures. Chem. Commun. 99, 70–71.

Abdoli, M., Hashem, F.D., Goushegir, A., Shirazi, M.T., Nakhjavani, M., Shojaii, A., Rezvani, S., Mahlooji, K., 2017. Anti-hyperglycemic effect of aqueous extract of Juglans regia L. leaf (walnut leaf) on type 2 diabetic patients: A randomized controlled trial. Adv. Integr. Med. 4, 98–102.

Abildskov, J., O’Connell, J.P., 2005. Thermodynamic method for obtaining the solubilities of complex medium-sized chemicals in pure and mixed solvents. Fluid Phase Equilib. 228–229, 395–400.

Abildskov, J., O’Connell, J.P., 2003. Predicting the solubilities of complex chemicals I. Solutes in different solvents. Ind. Eng. Chem. Res. 42, 5622–5634.

Abraham, M.H., 1993. Scales of solute hydrogen-bonding: their construction and application to physicochemical and biochemical processes. Chem. Soc. Rev. 22, 73–83.

Abraham, M.H., Acree, W.E., Earp, C.E., Vladimirova, A., Whaley, W.L., 2015. Studies on the hydrogen bond acidity, and other descriptors and properties for hydroxyflavones and hydroxyisoflavones. J. Mol. Liq. 208, 363–372.

Abraham, M.H., Ibrahim, A., Zissimos, A.M., 2004. Determination of sets of solute descriptors from chromatographic measurements. J. Chromatogr. A 1037, 29–47.

Abraham, M.H., McGowan, J.C., 1987. The use of characteristic volumes to measure cavity terms in reversed phase liquid chromatography. Chromatographia 23, 243–246.

Abraham, M.H., Smith, R.E., Luchtefeld, R., Boorem, A.J., Luo, R., Acree, W.E., 2010. Prediction of solubility of drugs and other compounds in organic solvents. J. Pharm. Sci. 99, 1500–1515.

Abreu, R.M. V, Ferreira, I.C.F.R., Calhelha, R.C., Lima, R.T., Vasconcelos, M.H., Adega, F., Chaves, R., Queiroz, M.J.R.P., 2011. Anti-hepatocellular carcinoma activity using human

191 Bibliography CHAPTER 7

HepG2 cells and hepatotoxicity of 6-substituted methyl 3-aminothieno[3,2-b]pyridine-2- carboxylate derivatives: In vitro evaluation, cell cycle analysis and QSAR studies. Eur. J. Med. Chem. 46, 5800–5806.

Acree, W.E., Barrera, M., Abraham, M.H., 2017a. Comment on “measurement and correlation of the solubility of p-coumaric acid in nine pure and water + ethanol mixed solvents at temperatures from 293.15 to 333.15 K.” J. Chem. Eng. Data 62, 578–583.

Acree, W.E., Bowen, K.R., Horton, M.Y., Abraham, M.H., 2017b. Computation of Abraham model solute descriptors for 3-methyl-4-nitrobenzoic acid from measured solubility data. Phys. Chem. Liq. 55, 482–491.

Acree, W.E., Ramirez, A.M., Cheeran, S., Martinez, F., 2017c. Determination of Abraham model solute descriptors and preferential solvation from measured solubilities for 4- nitropyrazole dissolved in binary aqueous-organic solvent mixtures. Phys. Chem. Liq. 55, 605–616.

Akbari, V., Jamei, R., Heidari, R., Esfahlan, A.J., 2012. Antiradical activity of different parts of Walnut (Juglans regia L.) fruit as a function of genotype. Food Chem. 135, 2404–2410.

Alberti, Á., Riethmüller, E., Béni, S., 2019. Characterization of diarylheptanoids: An emerging class of bioactive natural products. J. Pharm. Biomed. Anal. 147, 13–34.

Almasi, M., 2015. Densities and viscosities of binary mixtures containing ethyl formate and 2-alkanols: Friction theory and free volume theory. J. Chem. Eng. Data 60, 714–720.

Almeida, I.F., Fernandes, E., Lima, J.L.F.C., Costa, P.C., Bahia, M.F., 2008. Walnut (Juglans regia) leaf extracts are strong scavengers of pro-oxidant reactive species. Food Chem. 106, 1014–1020.

Alshatwi, A.A., Hasan, T.N., Shafi, G., Syed, N.A., Al-Assaf, A.H., Alamri, M.S., Al-Khalifa, A.S., 2012. Validation of the antiproliferative effects of organic extracts from the green husk of Juglans regia L. on PC-3 human prostate cancer cells by assessment of apoptosis-related genes. Evidence-Based Complement. Altern. Med. 2012, 103026.

Alves, M., Ferreira, I., Lourenço, I., Costa, E., Martins, A., Pintado, M., 2014. Wild mushroom extracts as inhibitors of bacterial biofilm formation. Pathogens 3, 667–679.

192 Bibliography CHAPTER 7

Amaral, J.S., Seabra, R.M., Andrade, P.B., Valentão, P., Pereira, J.A., Ferreres, F., 2004. Phenolic profile in the quality control of walnut (Juglans regia L.) leaves. Food Chem. 88, 373–379.

Apelblat, A., Manzurola, E., 1998. Solubilities of L-aspartic, DL-aspartic, DL-glutamic, p- hydroxybenzoic, o-anistic, p-anisic, and itaconic acids in water from T = 278 K to T = 345 K. J. Chem. Thermodyn. 29, 1527–1533.

Apostolakis, A., Grigorakis, S., Makris, D.P., 2014. Optimisation and comparative kinetics study of polyphenol extraction from olive leaves (Olea europaea) using heated water/glycerol mixtures. Sep. Purif. Technol. 128, 89–95.

Asgary, S., Parkhideh, S., Solhpour, A., Madani, H., Mahzouni, P., Rahimi, P., 2008. Effect of ethanolic extract of Juglans regia L. on blood sugar in diabetes-induced rats. J. Med. Food 11, 533–538.

Athanasiadis, V., Grigorakis, S., Lalas, S., Makris, D.P., 2017. Highly efficient extraction of antioxidant polyphenols from Olea europaea leaves using an eco-friendly glycerol/glycine deep eutectic solvent. Waste and Biomass Valorization. DOI:10.1007/s12649-017-9997-7.

Bagheri, S., Ahmadvand, H., Khosrowbeygi, A., 2012. Various antioxidative activities of Iranian walnut (Juglans regia L.) green husks hydroalcoholic extract and its effects on LDL oxidation. J. Biol. Act. Prod. from Nat. 6, 330–340.

Bajić, D.M., Ivaniš, G.R., Visak, Z.P., Živković, E.M., Šerbanović, S.P., Kijevčanin, M.L., 2013. Densities, viscosities, and refractive indices of the binary systems (PEG200 + 1,2- propanediol, +1,3-propanediol) and (PEG400 + 1,2-propanediol, +1,3-propanediol) at (288.15 to 333.15) K and atmospheric pressure: Measurements and modeling. J. Chem. Thermodyn. 57, 510–529.

Bajić, D.M., Živković, E.M., Šerbanović, S.S., Kijevčanin, M.L., 2014. Volumetric and viscometric study of binary systems of ethyl butyrate with alcohols. J. Chem. Eng. Data 59, 3677–3690.

Bakirtzi, C., Triantafyllidou, K., Makris, D.P., 2016. Novel lactic acid-based natural deep eutectic solvents: Efficiency in the ultrasound-assisted extraction of antioxidant

193 Bibliography CHAPTER 7 polyphenols from common native Greek medicinal plants. J. Appl. Res. Med. Aromat. Plants 3, 120–127.

Balasundram, N., Sundram, K., Samman, S., 2006. Phenolic compounds in plants and agri- industrial by-products: Antioxidant activity, occurrence, and potential uses. Food Chem. 99, 191–203.

Baranauskaitė, J., Jakštas, V., Ivanauskas, L., Kopustinskienė, D.M., Drakšienė, G., Masteikova, R., Bernatonienė, J., 2016. Optimization of carvacrol, rosmarinic, oleanolic and ursolic acid extraction from oregano herbs (Origanum onites L., Origanum vulgare spp. hirtum and Origanum vulgare L.). Nat. Prod. Res. 30, 672–674.

Barreira, J.C.M., Rodrigues, S., Carvalho, A.M., Ferreira, I.C.F.R., 2013. Development of hydrosoluble gels with Crataegus monogyna extracts for topical application: Evaluation of antioxidant activity of the final formulations. Ind. Crops Prod. 42, 175–180.

Barros, L., Pereira, E., Calhelha, R.C., Dueñas, M., Carvalho, A.M., Santos-Buelga, C., Ferreira, I.C.F.R., 2013. Bioactivity and chemical characterization in hydrophilic and lipophilic compounds of Chenopodium ambrosioides L. J. Funct. Foods 5, 1732–1740.

Beiki, T., Najafpour, G.D., Hosseini, M., 2018. Evaluation of antimicrobial and dyeing properties of walnut (Juglans regia L.) green husk extract for cosmetics. Color. Technol. 134, 71–81.

Benvenutti, L., Zielinski, A.A.F., Ferreira, S.R.S., 2019. Which is the best food emerging solvent: IL, DES or NADES ? Trends Food Sci. Technol. 90, 133–146.

Bernal-García, J.M., Guzmán-López, A., Cabrales-Torres, A., Estrada-Baltazar, A., Iglesias- Silva, G.A., 2008. Densities and viscosities of (N,N-dimethylformamide + water) at atmospheric pressure from (283.15 to 353.15) K. J. Chem. Eng. Data 53, 1024–1027.

Bessada, S.M.F., Barreira, J.C.M., Barros, L., Ferreira, I.C.F.R., Oliveira, M.B.P.P., 2016. Phenolic profile and antioxidant activity of Coleostephus myconis (L.) Rchb.f.: An underexploited and highly disseminated species. Ind. Crops Prod. 89, 45–51.

194 Bibliography CHAPTER 7

Bezerra, M.A., Santelli, R.E., Oliveira, E.P., Villar, L.S., Escaleira, E.A., Escaleira, L.A., 2008. Response surface methodology (RSM) as a tool for optimization in analytical chemistry. Talanta 76, 965–977.

Bi, D., Zhao, Y., Jiang, R., Wang, Y., Tian, Y., Chen, X., Bai, S., She, G., 2016. Phytochemistry, bioactivity and potential impact on health of Juglans: the original plant of walnut. Nat. Prod. Commun. 11, 869–880.

Bonaccorsi, P., Caristi, C., Gargiulli, C., Leuzzi, U., 2008. Flavonol glucosides in Allium species: A comparative study by means of HPLC-DAD-ESI-MS-MS. Food Chem. 107, 1668– 1673.

Booth, A.M., Bannan, T., McGillen, M.R., Barley, M.H., Topping, D.O., McFiggans, G., Percival, C.J., 2012. The role of ortho, meta, para isomerism in measured solid state and derived sub-cooled liquid vapour pressures of substituted benzoic acids. RSC Adv. 2, 4430– 4443.

Bowen, K., Stephens, T., Lu, H., Satish, K., Shan, D., Acree Jr., W.E., Abraham, M.H., 2013. Experimental and predicted solubilities of 3,4-dimethoxybenzoic acid in select organic solvents of varying polarity and hydrogen-bonding character. Eur. Chem. Bull 9, 577–583.

Box, G., Hunter, J., 1957. Multi-factor experimental designs for exploring response surfaces. Ann. Math. Stat. 28, 195–241.

Bravo, L., Mateos, R., 2008. Analysis of flavonoids in functional foods and nutraceuticals, in: Hurst, W.J. (Ed.), Methods of Analysis for Functional Foods and Nutraceuticals. CRC Press, Boca Raton.

Bubalo, M.C., Ćurko, N., Tomašević, M., Ganić, K.K., Redovniković, I.R., 2016. Green extraction of grape skin phenolics by using deep eutectic solvents. Food Chem. 200, 159– 166.

Butler, M.S., Robertson, A.A.B., Cooper, M.A., 2014. Natural product and natural product derived drugs in clinical trials. Nat. Prod. Rep. 31, 1612–1661.

Canosa, J., Rodríguez, A., Tojo, J., 1999. Binary mixture properties of diethyl ether with alcohols and alkanes from 288.15 K to 298.15 K. Fluid Phase Equilib. 156, 57–71.

195 Bibliography CHAPTER 7

Cao, J., Chen, L., Li, M., Cao, F., Zhao, L., Su, E., 2018. Efficient extraction of proanthocyanidin from Ginkgo biloba leaves employing rationally designed deep eutectic solvent-water mixture and evaluation of the antioxidant activity. J. Pharm. Biomed. Anal. 158, 317–326.

Carocho, M., Ferreira, I.C.F.R., 2013. A review on antioxidants, prooxidants and related controversy: Natural and synthetic compounds, screening and analysis methodologies and future perspectives. Food Chem. Toxicol. 51, 15–25.

Carocho, M., Morales, P., Ferreira, I.C.F.R., 2015. Natural food additives: Quo vadis? Trends Food Sci. Technol. 45, 284–295.

Carro, L., Torrado, M., Raviña, E., Masaguer, C.F., Lage, S., Brea, J., Loza, M.I., 2014. Synthesis and biological evaluation of a series of aminoalkyl-tetralones and tetralols as dual dopamine/serotonin ligands. Eur. J. Med. Chem. 71, 237–249.

Carvalho, A.M., Morales, R., 2010. Persistence of wild food and wild medicinal plant knowledge in a Northeasten region of Portugal, 2nd ed., in: Prado-de-Santayana, M., Pieroni, A., Puri, R. (Eds.), Ethnobotany in the New Europe: People, Health and Wild Plant Resources. Berghahn Books, New York, pp. 147–171.

Carvalho, M., Ferreira, P.J., Mendes, V.S., Silva, R., Pereira, J.A., Jerónimo, C., Silva, B.M., 2010. Human cancer cell antiproliferative and antioxidant activities of Juglans regia L. Food Chem. Toxicol. 48, 441–447.

Castro, V.I.B., Craveiro, R., Silva, J.M., Reis, R.L., Paiva, A., Ana, A.R., 2018. Natural deep eutectic systems as alternative nontoxic cryoprotective agents. Cryobiology 83, 15–26.

Charlton, A.K., Daniels, C.R., Wold, R.M., Pustejovsky, E., Acree, W.E., Abraham, M.H., 2005. Solubility of crystalline nonelectrolyte solutes in organic solvents: mathematical correlation of 3-nitrobenzoic acid solubilities with the Abraham general solvation model. J. Mol. Liq. 116, 19–28.

Chatrabnous, N., Yazdani, N., Tavallali, V., Vahdati, K., 2018. Preserving quality of fresh walnuts using plant extracts. LWT - Food Sci. Technol. 91, 1–7.

196 Bibliography CHAPTER 7

Chemat, F., Vian, M.A., Cravotto, G., 2012. Green extraction of natural products: Concept and principles. Int. J. Mol. Sci. 13, 8615–8627.

Chen, A.Y., Chen, Y.C., 2013. A review of the dietary flavonoid, kaempferol on human health and cancer chemoprevention. Food Chem. 138, 2099–2107.

Chen, C.-C., Crafts, P.A., 2006. Correlation and prediction of drug molecule solubility with the NRTL-SAC model. Comput. Aided Chem. Eng. 21, 859–864.

Chen, C.-C., Song, Y., 2004. Solubility modeling with NRTL segment activity coefficient model. Ind. Eng. Chem. Res 43, 8354–8362.

Choi, W.S., Shin, P.G., Lee, J.H., Kim, G. Do, 2012. The regulatory effect of veratric acid on NO production in LPS-stimulated RAW264.7 macrophage cells. Cell. Immunol. 280, 164– 170.

Choi, Y.H., van Spronsen, J., Dai, Y., Verberne, M., Hollmann, F., Arends, I.W.C.E., Witkamp, G.-J., Verpoorte, R., 2011. Are natural deep eutectic solvents the missing link in understanding cellular metabolism and physiology? Plant Physiol. 156, 1701–1705.

Choi, Y.H., Verpoorte, R., 2019. Green solvents for the extraction of bioactive compounds from natural products using ionic liquids and deep eutectic solvents. Curr. Opin. Food Sci. 26, 87–93.

Chulasiri, M., Wanaswas, P., Sriaum, D., Nakamat, S., Wongkrajang, Y., Kongsaktrakoon, B., Phornchirasilp, S., Songchitsomboon, S., Leelarungrayub, D., 2011. Utilizing hydroglycolic extract from myrobalan fruits to counteract reactive oxygen species. Int. J. Cosmet. Sci. 33, 371–376.

Ciniglia, C., Sansone, C., Panzella, L., Napolitano, A., D’Ischia, M., 2012. Effects of walnut husk washing waters and their phenolic constituents on horticultural species. Environ. Sci. Pollut. Res. 19, 3299–3306.

Ciocirlan, O., Croitoru, O., Iulian, O., 2014. Density and refractive index of binary mixtures of two 1-alkyl-3- methylimidazolium ionic liquids with 1,4-dioxane and ethylene glycol. J. Chem. Eng. Data 59, 1165–1174.

197 Bibliography CHAPTER 7

Ciocirlan, O., Teodorescu, M., Dragoescu, D., Iulian, O., Barhala, A., 2010. Densities and excess molar volumes of the binary mixtures of cyclohexanone with chloroalkanes at temperatures between (288.15 and 318.15) K. J. Chem. Eng. Data 55, 968–973.

Clifford, M.N., Johnston, K.L., Knight, S., Kuhnert, N., 2003. Hierarchical scheme for LC-MSn identification of chlorogenic acids. J. Agric. Food Chem. 51, 2900–2911.

Clifford, M.N., Knight, S., Kuhnert, N., 2005. Discriminating between the six isomers of dicaffeoylquinic acid by LC-MSn. J. Agric. Food Chem. 53, 3821–3832.

Comuzzi, C., Polese, P., Melchior, A., Portanova, R., Tolazzi, M., 2003. SOLVERSTAT: a new utility for multipurpose analysis. An application to the investigation of dioxygenated Co (II) complex formation in dimethylsulfoxide solution. Talanta 59, 67–80.

Conklin, K.A., 2000. Dietary antioxidants during cancer chemotherapy: impact on chemotherapeutic effectiveness and development of side effects. Nutr. Cancer 37, 1–18.

Corrêa, R.C.G., De Souza, A.H.P., Calhelha, R.C., Barros, L., Glamoclija, J., Sokovic, M., Peralta, R.M., Bracht, A., Ferreira, I.C.F.R., 2015. Bioactive formulations prepared from fruiting bodies and submerged culture mycelia of the Brazilian edible mushroom Pleurotus ostreatoroseus Singer. Food Funct. 6, 2155–2164.

Cosmulescu, S., Botu, M., Achim, G., Baciu, A., Gruia, M., Trandafir, I., 2014a. Polyphenol content in walnut (Juglans regia L.) mature leaves. Acta Hortic. 1050, 205–212.

Cosmulescu, S., Trandafir, I., Achim, G., Baciu, A., 2011. Juglone content in leaf and green husk of five walnut (Juglans regia L.) cultivars. Not. Bot. Horti Agrobot. Cluj-Napoca 39, 237–240.

Cosmulescu, S., Trandafir, I., Achim, G., Botu, M., Baciu, A., Gruia, M., 2010. Phenolics of green husk in mature walnut fruits. Not. Bot. Horti Agrobot. Cluj-Napoca 38, 53–56.

Cosmulescu, S., Trandafir, I., Nour, V., 2014b. Seasonal variation of the main individual phenolics and juglone in walnut (Juglans regia) leaves. Pharm. Biol. 52, 575–580.

Cruz-Vega, D.E., Verde-Star, M.J., Salinas-González, N., Rosales-Hernández, B., Estrada- García, I., Mendez-Aragón, P., Carranza-Rosales, P., González-Garza, M.T., Castro-Garza, J.,

198 Bibliography CHAPTER 7

2008. Antimycobacterial activity of Juglans regia, Juglans mollis, Carya illinoensis and Bocconia frutescens. Phyther. Res. 22, 557–559.

Cunha, S.C., Fernandes, J.O., 2018. Extraction techniques with deep eutectic solvents. Trends Anal. Chem. 105, 225–239.

Cunja, V., Mikulic-Petkovsek, M., Zupan, A., Stampar, F., Schmitzer, V., 2015. Frost decreases content of sugars, ascorbic acid and some quercetin glycosides but stimulates selected carotenes in Rosa canina hips. J. Plant Physiol. 178, 55–63.

Dai, J., Mumper, R.J., 2010. Plant phenolics: Extraction, analysis and their antioxidant and anticancer properties. Molecules 15, 7313–7352.

Dai, Y., Rozema, E., Verpoorte, R., Choi, Y.H., 2016. Application of natural deep eutectic solvents to the extraction of anthocyanins from Catharanthus roseus with high extractability and stability replacing conventional organic solvents. J. Chromatogr. A 1434, 50–56.

Dai, Y., van Spronsen, J., Witkamp, G., Verpoorte, R., Choi, Y.H., 2013a. Natural deep eutectic solvents as new potential media for green technology. Anal. Chim. Acta 766, 61– 68.

Dai, Y., Verpoorte, R., Choi, Y.H., 2014. Natural deep eutectic solvents providing enhanced stability of natural colorants from safflower (Carthamus tinctorius). Food Chem. 159, 116– 121.

Dai, Y., Witkamp, G.-J., Verpoorte, R., Choi, Y.H., 2013b. Natural deep eutectic solvents as a new extraction media for phenolic metabolites in Carthamus tinctorius L. Anal. Chem. 85, 6272–6278.

Dakkach, M., Aguilar, F., Alaoui, F.E.M., Montero, E.A., 2015. Liquid density of oxygenated additives to biofuels: 2-Butanol at pressures up to 140 MPa and temperatures from (293.15 to 393.27) K. J. Chem. Thermodyn. 89, 278–285.

Daniels, C.R., Charlton, A.K., Wold, R.M., Acree, Jr., W.E., Abraham, M.H., 2003. Thermochemical behavior of dissolved carboxylic acid solutes: solubilities of 3-

199 Bibliography CHAPTER 7 methylbenzoic acid and 4-chlorobenzoic acid in organic solvents. Can. J. Chem. 81, 1492– 1501.

Dávalos, A., 2004. Extending applicability of the oxygen radical absorbance capacity (ORAC- fluorescein) assay. J. Agric. Food Chemestry 52, 48–54.

Dávila, M.J., Alcalde, R., Atilhan, M., Aparicio, S., 2012. PρT measurements and derived properties of liquid 1,2-alkanediols. J. Chem. Thermodyn. 47, 241–259.

De Morais, P., Gonçalves, F., Coutinho, J.A.P., Ventura, S.P.M., 2015. Ecotoxicity of cholinium-based deep eutectic solvents. ACS Sustain. Chem. Eng. 3, 3398–3404.

Delaviz, H., Mohammadi, J., Ghalamfarsa, G., Mohammadi, B., Farhadi, N., 2017. A review study on phytochemistry and pharmacology applications of Juglans Regia plant. Pharmacogn. Rev. 11, 145–152.

Destandau, E., Michel, T., Elfafir, C., 2013. Microwave-assisted extraction, in: Rostango, M.A., Prado, J.M. (Eds.), Natural product extraction principles and applications. Royal Society of Chemistry, Cambridge, pp. 113–150.

Devi, K.P., Malar, D.S., Navabi, S.F., Sureda, A., Xiao, J., Navabi, S.M., Daglia, M., 2015a. Kaempferol and inflammation: From chemistry to medicine. Pharmacol. Res. 99, 1–10.

Devi, K.P., Rajavel, T., Habtemariam, S., Navabi, S.F., Navabi, S.M., 2015b. Molecular mechanisms underlying anticancer effects of myricetin. Life Sci. 142, 19–25.

Dhananjaya, B.L., Nataraju, A., Raghavendra Gowda, C.D., Sharath, B.K., D’souza, C.J.M., 2009. Vanillic acid as a novel specific inhibitor of snake venom 5′-nucleotidase: a pharmacological tool in evaluating the role of the enzyme in snake envenomation. Biochem. 74, 1315–1319.

Dolatabadi, S., Nesari, H., Mahdavi-ourtakand, M., 2018. Microbial pathogenesis evaluating the anti-biofilm and antibacterial effects of Juglans regia L. extracts against clinical isolates of Pseudomonas aeruginosa. Microb. Pathog. 118, 285–289.

Drosou, C., Kyriakopoulou, K., Bimpilas, A., Tsimogiannis, D., Krokida, M., 2015. A comparative study on different extraction techniques to recover red grape pomace polyphenols from vinification byproducts. Ind. Crops Prod. 75, 141–149.

200 Bibliography CHAPTER 7

Du, H., Li, C., Wen, Y., Tu, Y., Zhong, Y., Yuan, Z., Li, Y., Liang, B., 2014. Secondary metabolites from pericarp of Juglans regia. Biochem. Syst. Ecol. 54, 88–91.

Duan, L., Dou, L.-L., Guo, L., Li, P., Liu, E.-H., 2016. Comprehensive Evaluation of Deep Eutectic Solvents in Extraction of Bioactive Natural Products. ACS Sustain. Chem. Eng. 4, 2405–2411.

Egorov, G.I., Makarov, D.M., 2014. Volumetric properties of binary liquid-phase mixture of (water+glycerol) at temperatures of (278.15 to 323.15)K and pressures of (0.1 to 100)MPa. J. Chem. Thermodyn. 79, 135–158.

Egorov, G.I., Makarov, D.M., 2011. Densities and volume properties of (water + tert- butanol) over the temperature range of (274.15 to 348.15) K at pressure of 0.1 MPa. J. Chem. Thermodyn. 43, 430–441.

Eidi, A., Moghadam, J.Z., Mortazavi, P., Olamafar, S., 2013. Hepatoprotective effects of

Juglans regia extract against CCl4-induced oxidative damage in rats. Pharm. Biol. 51, 558– 565.

Enders, S., Kahl, H., Winkelmann, J., 2007. Surface tension of the ternary system water + acetone + toluene. J. Chem. Eng. Data 52, 1072–1079.

Erdemoglu, N., Küpeli, E., Yeşilada, E., 2003. Anti-inflammatory and antinociceptive activity assessment of plants used as remedy in Turkish folk medicine. J. Ethnopharmacol. 89, 123– 129.

Erin Hart, Klein, A., Zha, O., Wadawadigi, A., Qian, E., Dunn, S., Herron, J., Kankolongo, K., Shy’an Ryan, Acree, W.E., Abraham, M.H., 2018. Determination of Abraham model solute descriptors for monomeric 3,4,5-trimethoxybenzoic acid from experimental solubility data in organic solvents measured at 298.2 K. Phys. Chem. Liq. 56, 381–390.

Espin, S., Gonzalez-Manzano, S., Taco, V., Poveda, C., Ayuda-Durán, B., Gonzalez-Paramas, A.M., Santos-Buelga, C., 2016. Phenolic composition and antioxidant capacity of yellow and purple-red Ecuadorian cultivars of tree tomato (Solanum betaceum Cav.). Food Chem. 194, 1073–1080.

201 Bibliography CHAPTER 7

European Commission, 2006. Commission Decision of 9 February 2006 amending Decision 96/335/EC establishing an inventory and a common nomenclature of ingredients employed in cosmetic products (2006/257/EC). Off. J. Eur. Union 1–528.

Evans, B.C., Nelson, C.E., Yu, S.S., Beavers, K.R., Kim, A.J., Li, H., Nelson, H.M., Giorgio, T.D., Duvall, C.L., 2013. Ex vivo red blood cell hemolysis assay for the evaluation of pH-responsive endosomolytic agents for the cytosolic delivery of biomacromolecular drugs. J. Vis. Exp. 73, 1–5.

Faranda, S., Foca, G., Marchetti, A., Pályi, G., Tassi, L., Zucchi, C., 2004. Density measurements of the binary mixtures of 2-butanone and 2-butanol at temperatures from −10 to 80 °C. J. Mol. Liq. 111, 117–123.

Feng, Y., Tang, W., Huang, Y., Xiong, Y., Chen, L., Liu, Y., Li, Y., 2014. (Solid + liquid) phase equilibria of tetraphenyl piperazine-1,4-diyldiphosphonate in pure solvents. J. Chem. Thermodyn. 78, 143–151.

Fernández-Agulló, A., Pereira, E., Freire, M.S., Valentão, P., Andrade, P.B., González- Álvarez, J., Pereira, J.A., 2013. Influence of solvent on the antioxidant and antimicrobial properties of walnut (Juglans regia L.) green husk extracts. Ind. Crops Prod. 42, 126–132.

Fernández-Ponce, M.T., Casas, L., Mantell, C., De La Ossa, E.M., 2015. Use of high pressure techniques to produce Mangifera indica L. leaf extracts enriched in potent antioxidant phenolic compounds. Innov. Food Sci. Emerg. Technol. 29, 94–106.

Fernández-Ponce, M.T., Casas, L., Mantell, C., Rodríguez, M., de la Ossa, E.M., 2012. Extraction of antioxidant compounds from different varieties of Mangifera indica leaves using green technologies. J. Supercrit. Fluids 72, 168–175.

Fernández, L., Ortega, J., Pérez, E., Toledo, F., Canosa, J., 2013. Multiproperty correlation of experimental data of the binaries propyl ethanoate + alkanes (pentane to decane). New experimental information for vapor-liquid equilibrium and mixing properties. J. Chem. Eng. Data 58, 686–706.

202 Bibliography CHAPTER 7

Fernandez, L., Silva, L.P., Martins, M.A.R., Ferreira, O., Ortega, J., Pinho, S.P., Coutinho, J.A.P., 2017. Indirect assessment of the fusion properties of choline chloride from solid- liquid equilibria data. Fluid Phase Equilib. 448, 9–14.

Fernández, M. de los Á., Boiteux, J., Espino, M., Gomez, F.J.V., Silva, M.F., 2018a. Natural deep eutectic solvents-mediated extractions: The way forward for sustainable analytical developments. Anal. Chim. Acta 1038, 1–10.

Fernández, M. de los Á., Espino, M., Gomez, F.J.V., Silva, M.F., 2018b. Novel approaches mediated by tailor-made green solvents for the extraction of phenolic compounds from agro-food industrial by-products. Food Chem. 239, 671–678.

Ferreira, O., Pinho, S.P., 2012. Solubility of flavonoids in pure solvents. Ind. Eng. Chem. Res. 51, 6586–6590.

Ferreira, S.L.C., Bruns, R.E., Ferreira, H.S., Matos, G.D., David, J.M., Brandão, G.C., da Silva, E.G.P., Portugal, L.A., dos Reis, P.S., Souza, A.S., dos Santos, W.N.L., 2007. Box-Behnken design: an alternative for the optimization of analytical methods. Anal. Chim. Acta 597, 179–186.

Ferreira, V., Fernandes, F., Pinto-Carnide, O., Valentão, P., Falco, V., Martín, J.P., Ortiz, J.M., Arroyo-García, R., Andrade, P.B., Castro, I., 2016. Identification of Vitis vinifera L. grape berry skin color mutants and polyphenolic profile. Food Chem. 194, 117–127.

Florindo, C., Lima, F., Ribeiro, B.D., Marrucho, I.M., 2019. Deep eutectic solvents: overcoming 21st century challenges. Curr. Opin. Green Sustain. Chem. 18, 31–36.

Forino, M., Stiuso, P., Lama, S., Ciminiello, P., Tenore, G.C., Novellino, E., Taglialatela- Scafati, O., 2016. Bioassay-guided identification of the antihyperglycaemic constituents of walnut (Juglans regia) leaves. J. Funct. Foods 26, 731–738.

Foti, C., Bonamonte, D., Cassano, N., Conserva, A., Vena, G.A., 2010. Allergic contact dermatitis to propyl gallate and pentylene glycol in an emollient cream. Australas. J. Dermatol. 51, 147–148.

203 Bibliography CHAPTER 7

Francisco, M., van den Bruinhorst, A., Kroon, M.C., 2013. Low-transition-temperature mixtures (LTTMs): A new generation of designer solvents. Angew. Chem. Int. Ed. 52, 3074– 3085.

Francisco, M., van den Bruinhorst, A., Kroon, M.C., 2012. New natural and renewable low transition temperature mixtures (LTTMs): screening as solvents for lignocellulosic biomass processing. Green Chem. 14, 2153–2157.

Gallo, R., Viflizzo, G., Vecchio, F., Parodi, A., 2003. Allergic contact dermatitis from pentilene glycol in an emolient cream, with possible co-sensitization to resveratrol. Contact Dermatitis 48, 176–177.

Gamsjäger, H., Lorimer, J.W., Scharlin, P., Shaw, D.G., 2008. Glossary of terms related to solubility (IUPAC Recommendations 2008). Pure Appl. Chem. 80, 233–276.

García-Garabal, S., Segade, L., Domínguez-Pérez, M., Cabanas, M., Yañez-Vilar, S., Cabeza, O., 2015. Excess molar enthalpies of the binary systems: (Dibutyl ether + isomers of pentanol) at T = (298.15 and 308.15) K. J. Chem. Thermodyn. 87, 52–57.

García, A., Rodríguez-Juan, E., Rodríguez-Gutiérrez, G., Rios, J.J., Fernández-Bolaños, J., 2016. Extraction of phenolic compounds from virgin olive oil by deep eutectic solvents (DESs). Food Chem. 197, 554–561.

Gaweł-Bȩben, K., Bujak, T., Nizioł-Łukaszewska, Z., Antosiewicz, B., Jakubczyk, A., Karaś, M., Rybczyńska, K., 2015. Stevia rebaudiana Bert. leaf extracts as a multifunctional source of natural antioxidants. Molecules 20, 5468–5486.

Gawlik-Dziki, U., Durak, A., Pecio, L., Kowalska, I., 2014. Nutraceutical potential of tinctures from fruits, green husks, and leaves of Juglans regia L. Sci. World J. DOI:10.1155/2014/501392.

Ghasemi, K., Ghasemi, Y., Ehteshamnia, A., Nabavi, S.M., Nabavi, S.F., Ebrahimzadeh, M.A., Pourmorad, F., 2011. Influence of environmental factors on antioxidant activity, phenol and flavonoids contents of walnut (Juglans regia L .) green husks. J. Med. Plants Res. 5, 1128– 1133.

204 Bibliography CHAPTER 7

Gieseg, S.P., Esterbauer, H., 1994. Low density lipoprotein is saturable by pro-oxidant copper. FEBS Lett. 343, 188–194.

Gîrzu, M., Carnat, A., Privat, A., Fialip, J., Carnat, A., Lamaison, J., Carnat, A., Privat, A., Fialip, J., Carnat, A., Lamaison, J., 1998. Sedative effect of walnut leaf extract and juglone, an isolated constituent. Pharm. Biol. 36, 280–286.

Gouveia, W., Jorge, T.F., Martins, S., Meireles, M., Carolino, M., Cruz, C., Almeida, T. V., Araújo, M.E.M., 2014. Toxicity of ionic liquids prepared from biomaterials. Chemosphere 104, 51–56.

Gu, T., Zhang, M., Tan, T., Chen, J., Li, Z., Zhang, Q., Qiu, H., 2014. Deep eutectic solvents as novel extraction media for phenolic compounds from model oil. Chem. Commun. 50, 11749–11752.

Guarrera, P.M., 2005. Traditional phytotherapy in Central Italy (Marche, Abruzzo, and Latium). Fitoterapia 76, 1–25.

Gupta, D., 2015. Methods for determination of antioxidant capacity: a review. Int. J. Pharm. Sci. Res. 6, 546–566.

Ha, S.J., Lee, J., Park, J., Kim, Y.H., Lee, N.H., Kim, Y.E., Song, K.M., Chang, P.S., Jeong, C.H., Jung, S.K., 2018. Syringic acid prevents skin carcinogenesis via regulation of NoX and EGFR signaling. Biochem. Pharmacol. 154, 435–445.

Han, X., Guo, J., You, Y., Yin, M., Liang, J., Ren, C., Zhan, J., Huang, W., 2018. Vanillic acid activates thermogenesis in brown and white adipose tissue. Food Funct. 9, 4366–4375.

Hayyan, M., Hashim, M.A., Hayyan, A., Al-Saadi, M.A., AlNashef, I.M., Mirghani, M.E.S., Saheed, O.K., 2013. Are deep eutectic solvents benign or toxic? Chemosphere 90, 2193– 2195.

Hayyan, M., Mbous, Y.P., Looi, C.Y., Wong, W.F., Hayyan, A., Salleh, Z., Mohd-Ali, O., 2016. Natural deep eutectic solvents: cytotoxic profile. Springerplus 5, 913.

Heleno, S.A., Diz, P., Prieto, M.A., Barros, L., Rodrigues, A., Barreiro, M.F., Ferreira, I.C.F.R., 2016. Optimization of ultrasound-assisted extraction to obtain mycosterols from Agaricus

205 Bibliography CHAPTER 7 bisporus L. by response surface methodology and comparison with conventional Soxhlet extraction. Food Chem. 197, 1054–1063.

Heleno, S.A., Martins, A., Queiroz, M.J.R.P., Ferreira, I.C.F.R., 2015. Bioactivity of phenolic acids: Metabolites versus parent compounds: A review. Food Chem. 173, 501–513.

Hoover, K.R., Coaxum, R., Pustejovsky, E., Stovall, D.M., Acree, W.E., Abraham, M.H., 2004a. Thermochemical behavior of dissolved carboxylic acid solutes: part 4 – mathematical correlation of 4-nitrobenzoic acid solubilities with the Abraham solvation parameter model. Phys. Chem. Liq. 42, 339–347.

Hoover, K.R., Pop, K., Acree Jnr, W.E., Abraham, M.H., 2005. Solubility of crystalline nonelectrolyte solutes in organic solvents: mathematical correlation of 3-chlorobenzoic acid solubilities with the Abraham solvation parameter model. S. Afr. J. Chem. 58, 25–29.

Hoover, K.R., Stovall, D.M., Pustejovsky, E., Coaxum, R., Pop, K., William E. Acree, J., Abraham, M.H., 2004b. Solubility of crystalline nonelectrolyte solutes in organic solvents: Mathematical correlation of 2-methoxybenzoic acid and 4-methoxybenzoic acid solubilities with the Abraham solvation parameter model. Can. J. Chem. 82, 1353–1360.

Hosseini, S., Jamshidi, L., Mehrzadi, S., Mohammad, K., Najmizadeh, A.R., Alimoradi, H., Huseini, H.F., 2014. Effects of Juglans regia L. leaf extract on hyperglycemia and lipid profiles in type two diabetic patients: A randomized double-blind, placebo-controlled clinical trial. J. Ethnopharmacol. 152, 451–456.

Hosseinzadeh, H., Zarei, H., Taghiabadi, E., 2011. Antinociceptive, anti-inflammatory and acute toxicity effects of Juglans regia L. leaves in mice. Iran. Red Crescent Med. J. 13, 27– 33.

Huang, D., Ou, B., Hampsch-Woodill, M., Flanagan, J., Prior, R., 2002. High-throughput assay of oxygen radical absorbance capacity (ORAC) using a multichannel liquid handling system coupled with a microplate fluorescence reader in 96-well format. J. Agric. Food Chem. 50, 4437–4444.

206 Bibliography CHAPTER 7

Huang, H., Wang, Z., Aalim, H., Limwachiranon, J., Li, L., Duan, Z., Ren, G., Luo, Z., 2018. Green recovery of phenolic compounds from rice byproduct (rice bran) using glycerol based on viscosity, conductivity and density. Int. J. Food Sci. Technol. 54, 1363–1371.

Huang, J., Guo, X., Xu, T., Fan, L., Zhou, X., Wu, S., 2019. Ionic deep eutectic solvents for the extraction and separation of natural products. J. Chromatogr. A 1598, 1–19.

Huang, Y., Feng, F., Jiang, J., Qiao, Y., Wu, T., Voglmeir, J., Chen, Z.-G., 2017. Green and efficient extraction of rutin from tartary buckwheat hull by using natural deep eutectic solvents. Food Chem. 221, 1400–1405.

Huo, Y., Xia, S., Ma, P., 2007. Densities of ionic liquids, l-butyl-3-methylimidazolium hexafluorophosphate and l-butyl-3-methylimidazolium tetrafluoroborate, with benzene, acetonitrile, and 1-propanol at T = (293.15 to 343.15) K. J. Chem. Eng. Data 52, 2077–2082.

Ibrahim, R.M., El-Halawany, A.M., Saleh, D.O., Naggar, E.M.B. El, El-Shabrawy, A.E.-R.O., El- Hawary, S.S., 2015. HPLC-DAD-MS/MS profiling of phenolics from Securigera securidaca flowers and its anti-hyperglycemic and anti-hyperlipidemic activities. Rev. Bras. Farmacogn. 25, 134–141.

Itoh, A., Isoda, K., Kondoh, M., Kawase, M., Watari, A., Kobayashi, M., Tamesada, M., Yage,

K., 2010. Hepatoprotective effect of syringic acid and vanillic acid on CCl4-induced liver injury. Biol. Pharm. Bull. 33, 983–987.

Ivanov, E. V., 2014. To the issue of temperature-dependent behavior of standard molar volumes of components in the binary system (water + tetrahydrofuran) at ambient pressure. J. Chem. Thermodyn. 72, 37–43.

Izadiyan, Z., Shameli, K., Hara, H., Mohd Taib, S.H., 2018. Cytotoxicity assay of biosynthesis gold nanoparticles mediated by walnut (Juglans regia) green husk extract. J. Mol. Struct. 1151, 97–105.

Jablonský, M., Andrea, Š., Malvis, A., Jozef, Š., 2018. Extraction of value-added components from food industry based and agro-forest biowastes by deep eutectic solvents. J. Biotechnol. 282, 46–66.

207 Bibliography CHAPTER 7

Jahanban-Esfahlan, A., Ostadrahimi, A., Tabibiazar, M., Amarowicz, R., 2019. A comparative review on the extraction, antioxidant content and antioxidant potential of different parts of walnut (Juglans regia L.) fruit and tree. Molecules 24, 2133.

Jancheva, M., Grigorakis, S., Loupassaki, S., Makris, D.P., 2017. Optimised extraction of antioxidant polyphenols from Satureja thymbra using newly designed glycerol-based natural low-transition temperature mixtures (LTTMs). J. Appl. Res. Med. Aromat. Plants 6, 31–40.

Javidanpour, S., Reza, S., Tabtabaei, F., Siahpoosh, A., Morovati, H., Shahriari, A., 2012. Comparison of the effects of fresh leaf and peel extracts of walnut (Juglans regia L.) on blood glucose and β-cells of streptozotocin-induced diabetic rats. Vet. Res. Forum 3, 251– 255.

Jeng, T.L., Lai, C.C., Liao, T.C., Lin, S.Y., Sung, J.M., 2015. Effects of drying on caffeoylquinic acid derivative content and antioxidant capacity of sweet potato leaves. J. Food Drug Anal. 23, 701–708.

Jeong, K.M., Jin, Y., Yoo, D.E., Han, S.Y., Kim, E.M., Lee, J., 2018. One-step sample preparation for convenient examination of volatile monoterpenes and phenolic compounds in peppermint leaves using deep eutectic solvents. Food Chem. 251, 69–76.

Jimtaisong, A., Krisdaphong, P., 2013. Antioxidant activity of Pandanus amaryllifolius leaf and root extract and its application in topical emulsion. Trop. J. Pharm. Res. 12, 425–431.

Jin-Hai, H., Xiao-Wei, D., Guo-Dong, S., Wen-Ting, D., Wei-Ming, W., 2018. Identification and characterization of major constituents in Juglans mandshurica using ultra performance liquid chromatography coupled with time-of-flight mass spectrometry (UPLC-ESI-Q- TOF/MS). Chin. J. Nat. Med. 16, 525–545.

Kalil, S., Maugeri, F., 2000. Response surface analysis and simulation as a tool for bioprocess design and optimization. Process Biochem. 35, 539–550.

Karabegović, I.T., Stojičević, S.S., Veličković, D.T., Nikolić, N.Č., Lazić, M.L., 2013. Optimization of microwave-assisted extraction and characterization of phenolic

208 Bibliography CHAPTER 7 compounds in cherry laurel (Prunus laurocerasus) leaves. Sep. Purif. Technol. 120, 429– 436.

Karakashov, B., Grigorakis, S., Loupassaki, S., Makris, D.P., 2015. Optimisation of polyphenol extraction from Hypericum perforatum (St. John’s Wort) using aqueous glycerol and response surface methodology. J. Appl. Res. Med. Aromat. Plants 2, 1–8.

Karmakar, R., Mal, D., 2011. N-Methylanilinium trifluoroacetate-promoted Prins reaction of α-methylene-α-tetralone dimers: generation of new molecular scaffolds. Tetrahedron Lett. 52, 6098–6102.

Kashyap, D., Sharma, A., Singh, H., Sak, K., Punia, S., Mukherjee, T.K., 2017. Kaempferol – A dietary anticancer molecule with multiple mechanisms of action: Recent trends and advancements. J. Funct. Foods 30, 203–219.

Kasmi, M., Rama, P., Hodaj, B., Kukali, E., Rabeta, A., 2013. Budding of walnut (Juglans regia L.). Albanian J. Agric. Sci. 12, 465–469.

Kaume, L., Gilbert, W.C., Brownmiller, C., Howard, L.R., Devareddy, L., 2012. Cyanidin 3-O- B-d-glucoside-rich blackberries modulate hepatic gene expression, and anti-obesity effects in ovariectomized rats. J. Funct. Foods 4, 480–488.

Kemmer, G., Keller, S., 2010. Nonlinear least-squares data fitting in Excel spreadsheets. Nat. Protoc. 5, 267–281.

Kerre, S., 2008. Allergic contact dermatitis to pentylene glycol in a cosmetic cream. Contact Dermatitis 58, 122–123.

Kerton, F.M., 2009. Alternative Solvents for Green Chemistry, 1st ed. Royal Society of Chemistry, Cambridge.

Keskin, D., Ceyhan, N., Ugur, A., 2012. Chemical composition and in vitro antimicrobial activity of walnut (Juglans regia) green husks and leaves from West Anatolia. J. Pure Appl. Microbiol. 6, 583–588.

Kim, J.-M., Chang, S.-M., Kim, I.-H., Kim, Y.-E., Hwang, J.-H., Kim, K.-S., Kim, W.-S., 2007. Design of optimal solvent for extraction of bio-active ingredients from mulberry leaves. Biochem. Eng. J. 37, 271–278.

209 Bibliography CHAPTER 7

Kim, S.J., Kim, M.C., Um, J.Y., Hong, S.H., 2010. The beneficial effect of vanillic acid on ulcerative colitis. Molecules 15, 7208–7217.

Kong, W., Zhao, Y., Shan, L., Xiao, X., Guo, W., 2008. Thermochemical studies on the quantity-antibacterial effect relationship of four organic acids from Radix Isatidis on Escherichia coli growth. Biol. Pharm. Bull. 31, 1301–1305.

Kottaras, P., Koulianos, M., Makris, D., 2017. Low-Transition Temperature Mixtures (LTTMs) Made of Bioorganic Molecules: Enhanced Extraction of Antioxidant Phenolics from Industrial Cereal Solid Wastes. Recycling 2, 3. DOI:10.3390/recycling2010003.

Kozlevčar, B., Odlazek, D., Golobič, A., Pevec, A., Strauch, P., Šegedin, P., 2006. Complexes with lignin model compound vanillic acid. Two different carboxylate ligands in the same dinuclear tetracarboxylate complex [Cu2(C8H7O4)2(O2CCH3)2(CH3OH)2]. Polyhedron 25, 1161–1166.

Kuete, V., Kamga, J., Sandjo, L.P., Ngameni, B., Poumale, H.M.P., Ambassa, P., Ngadjui, B.T., 2011. Antimicrobial activities of the methanol extract and compounds from Artocarpus communis (Moraceae). BMC Complement. Altern. Med. 11, 1–6.

Kwak, J.-H., Seo, J.M., Kim, N.-H., Arasu, M.V., Kim, S.K., Yoon, M.K., Kim, S.-J., 2016. Variation of quercetin glycosides derivatives in three onion (Allium cepa L.) varieties. Saudi J. Biol. Sci. 24, 1387–1391.

Lahucky, R., Nuernberg, K., Kovac, L., Bucko, O., Nuernberg, G., 2010. Assessment of the antioxidant potential of selected plant extracts – In vitro and in vivo experiments on pork. Meat Sci. 85, 779–784.

Lavaud, A., Laguerre, M., Birtic, S., Fabiano Tixier, A.S., Roller, M., Chemat, F., Bily, A.C., 2016. International Publication Number WO 2016/162703. Eutectic extraction solvents, extraction methods by eutectigenesis using said solvents, and extracts derived from said extraction methods.

Levie, R. de, 2008. Advanced Excel for scientific data analysis, 2nd ed. Oxford University Press, Inc., New York.

210 Bibliography CHAPTER 7

Li, C.-Y., Du, H.-J., Su, X.-H., Zhong, Y.-J., Yuan, Z.-P., Li, Y.-F., Liang, B., 2013a. Juglanones A and B: Two novel tetralone dimers from walnut pericarp (Juglans regia). Helv. Chim. Acta 96, 1031–1035.

Li, C., Liu, J.-X., Zhao, L., Di, D.-L., Meng, M., Jiang, S.-X., 2008a. Capillary zone electrophoresis for separation and analysis of four diarylheptanoids and an α-tetralone derivative in the green walnut husks (Juglans regia L.). J. Pharm. Biomed. Anal. 48, 749– 753.

Li, J., Jin, S., Zu, Y.-G., Luo, M., Wang, W., Zhao, C.-J., Fu, Y.-J., 2014. Rapid preparative extraction and determination of major organic acids in honeysuckle (Lonicera japonica Thunb.) tea. J. Food Compos. Anal. 33, 139–145.

Li, N., Jia, W., Zhang, Y., Tan, F., Zhang, J., 2011. Synergistic effect of 1,4-cyclohexanediol and 1,2-hexanediol on percutaneous absorption and penetration of metronidazole. Int. J. Pharm. 415, 169–174.

Li, Q., Lu, F., Tian, Y., Feng, S., Shen, Y., Wang, B., 2013b. Solubility of veratric acid in eight monosolvents and ethanol + 1‑butanol at various temperatures. J. Chem. Eng. Data 58, 1020–1028.

Li, Y., Ding, Y., 2012. Minireview: Therapeutic potential of myricetin in diabetes mellitus. Food Sci. Hum. Wellness 1, 19–25.

Li, Y., Ruan, H., Zhou, X., Zhang, Y., Pi, H., Wu, J., 2008b. Cytotoxic diarylheptanoids from pericarps of Juglans Cathayensis Dode. Chem. Res. Chinese Univ. 24, 427–429.

Liu, J., Meng, M., Li, C., Huang, X., Di, D., 2008. Simultaneous determination of three diarylheptanoids and an α-tetralone derivative in the green walnut husks (Juglans regia L.) by high-performance liquid chromatography with photodiode array detector. J. Chromatogr. A 1190, 80–85.

Liu, L., Li, W., Koike, K., Zhang, S., Nikaido, T., 2004. New α-tetralonyl glucosides from the fruit of Juglans mandshurica. Chem. Pharm. Bull. 52, 566–569.

211 Bibliography CHAPTER 7

Liu, Q., Zhao, P., Li, X.-C., Jacob, M.R., Yang, C.-R., Zhang, Y.-J., 2010. New α-tetralone galloylglucosides from the fresh pericarps of Juglans sigillata. Helv. Chim. Acta 93, 265– 271.

Liu, X., Ahlgren, S., Korthout, H.A.A.J., Salomé-Abarca, L.F., Bayona, L.M., Verpoorte, R., Choi, Y.H., 2018a. Broad range chemical profiling of natural deep eutectic solvent extracts using a high performance thin layer chromatography–based method. J. Chromatogr. A 1532, 198–207.

Liu, Y., Friesen, J.B., McAlpine, J.B., Lankin, D.C., Chen, S.N., Pauli, G.F., 2018b. Natural deep eutectic solvents: Properties, applications, and perspectives. J. Nat. Prod. 81, 679–690.

Lockowandt, L., Pinela, J., Lobo, C., Pereira, C., Abreu, R.M. V, Calhelha, R.C., José, M., Barros, L., Bredol, M., Ferreira, I.C.F.R., 2019. Chemical features and bioactivities of cornflower (Centaurea cyanus L.) capitula: The blue flowers and the unexplored non-edible part. Ind. Crop. Prod. 128, 496–503.

Macário, I.P.E., Jesus, F., Pereira, J.L., Ventura, S.P.M., Gonçalves, A.M.M., Coutinho, J.A.P., Gonçalves, F.J.M., 2018a. Unraveling the ecotoxicity of deep eutectic solvents using the mixture toxicity theory. Chemosphere 212, 890–897.

Macário, I.P.E., Oliveira, H., Menezes, A.C., Ventura, S.P.M., Pereira, J.L., Gonçalves, A.M.M., Coutinho, J.A.P., Gonçalves, F.J.M., 2019. Cytotoxicity profiling of deep eutectic solvents to human skin cells. Sci. Rep. 9, 3932.

Macário, I.P.E., Ventura, S.P.M., Pereira, J.L., Gonçalves, A.M.M., Coutinho, J.A.P., Gonçalves, F.J.M., 2018b. The antagonist and synergist potential of cholinium-based deep eutectic solvents. Ecotoxicol. Environ. Saf. 165, 597–602.

Makar, S., Saha, T., Singh, S.K., 2019. Naphthalene, a versatile platform in medicinal chemistry: Sky-high perspective. Eur. J. Med. Chem. 161, 252–276.

Malek, N.I., Singh, A., Surati, R., Ijardar, S.P., 2014. Study on thermo physical and excess molar properties of binary systems of ionic liquids. I: [Cnmim][PF6] (n = 6, 8) and alkyl acetates. J. Chem. Thermodyn. 74, 103–118.

212 Bibliography CHAPTER 7

Manconi, M., Marongiu, F., Manca, M.L., Caddeo, C., Sarais, G., Cencetti, C., Pucci, L., Longo, V., Bacchetta, G., Fadda, A.M., 2017. Nanoincorporation of bioactive compounds from red grape pomaces: In vitro and ex vivo evaluation of antioxidant activity. Int. J. Pharm. 523, 159–166.

Manic, M.S., Villanueva, D., Fornari, T., Queimada, A.J., Macedo, E.A., Najdanovic-Visak, V., 2012. Solubility of high-value compounds in ethyl lactate: Measurements and modeling. J. Chem. Thermodyn. 48, 93–100.

Manvar, D., Fernandes, T.D.A., Domingos, J.L.O., Baljinnyam, E., Basu, A., Junior, E.F.T., Costa, P.R.R., Kaushik-Basu, N., 2015. Synthesis and biological evaluation of α-aryl-α- tetralone derivatives as hepatitis C virus inhibitors. Eur. J. Med. Chem. 93, 51–54.

Mariano, A., Canzonieri, S., Camacho, A., Mainar, A., Postigo, M., 2011. Viscometric and volumetric properties of benzene + methyl acetate, or + methyl propanoate, or + methyl butanoate binary systems at 283.15, 298.15 and 313.15K. Phys. Chem. Liq. 49, 720–728.

Martínez, M.L., Labuckas, D.O., Lamarque, A.L., Maestri, D.M., 2010. Walnut (Juglans regia L.): Genetic resources, chemistry, by-products. J. Sci. Food Agric. 90, 1959–1967.

Martins, F.C.O.L., Sentanin, M.A., de Souza, D., 2019. Analytical methods in food additives determination: Compounds with functional applications. Food Chem. 272, 732–750.

Martins, M.A.R., Pinho, S.P., Coutinho, J.A.P., 2018. Insights into the nature of eutectic and deep eutectic mixtures. J. Solution Chem. DOI:10.1007/s10953-018-0793-1.

Martins, S., Mussatto, S.I., Martínez-Avila, G., Montañez-Saenz, J., Aguilar, C.N., Teixeira, J.A., 2011. Bioactive phenolic compounds: Production and extraction by solid-state fermentation. A review. Biotechnol. Adv. 29, 365–373.

Mbous, Y.P., Hayyan, M., Hayyan, A., Wong, W.F., Hashim, M.A., Looi, C.Y., 2017a. Applications of deep eutectic solvents in biotechnology and bioengineering ̶ Promises and challenges. Biotechnol. Adv. 35, 105–134.

Mbous, Y.P., Hayyan, M., Wong, W.F., Looi, C.Y., Hashim, M.A., 2017b. Unraveling the cytotoxicity and metabolic pathways of binary natural deep eutectic solvent systems. Sci. Rep. 7, 1–14.

213 Bibliography CHAPTER 7

Mellado-Ortega, E., Zabalgogeazcoa, I., Vázquez de Aldana, B.R., Arellano, J.B., 2017. Solutions to decrease a systematic error related to AAPH addition in the fluorescence- based ORAC assay. Anal. Biochem. 519, 27–29.

Meneses, M.A., Caputo, G., Scognamiglio, M., Reverchon, E., Adami, R., 2015. Antioxidant phenolic compounds recovery from Mangifera indica L. by-products by supercritical antisolvent extraction. J. Food Eng. 163, 45–53.

Meshkini, A., Tahmasbi, M., 2017. Antiplatelet aggregation activity of walnut hull extract via suppression of reactive oxygen species generation and caspase activation. J. Acupunct. Meridian Stud. 10, 193–203.

Miliauskas, G., Venskutonis, P.R., Beek, T.A. van, 2004. Screening of radical scavenging activity of some medicinal and aromatic plant extracts. Food Chem. 85, 231–237.

Mohammadi, J., Mirzaei, A., Azizi, A., Rouzbehi, A., Delaviz, H., 2012. The effects of hydroalcoholic extract of Juglans regia leaf on histological changes of Langerhans islet in diabetic rat model. Iran. South Med. J. 4, 293–301.

Mollica, A., Zengin, G., Locatelli, M., Stefanucci, A., Macedonio, G., Bellagamba, G., Onaolapo, O., Onaolapo, A., Azeez, F., Ayileka, A., Novellino, E., 2017. An assessment of the nutraceutical potential of Juglans regia L. leaf powder in diabetic rats. Food Chem. Toxicol. 107, 554–564.

Monge, M., Montaño, D.F., Bandrés, I., Lafuente, C., Royo, F.M., 2009. Volumetric, acoustic and refractive properties at several temperatures of dibutyl ether + 1-chlorobutane system. J. Mol. Liq. 150, 73–76.

Monteiro, M.R., Kugelmeier, C.L., Pinheiro, R.S., Otávio, M.B., César, A. da S., 2018. Glycerol from biodiesel production: Technological paths for sustainability. Renew. Sustain. Energy Rev. 88, 109–122.

Moraes, M.L.L., da Silva, H.D.T., Blanes, L., Doble, P., Tavares, M.F.M., 2016. Optimization of chemometric approaches for the extraction of isorhamnetin-3-O-rutinoside from Calendula officinalis L. J. Pharm. Biomed. Anal. 125, 408–414.

214 Bibliography CHAPTER 7

Mota, F.L., Carneiro, A.P., Queimada, A.J., Pinho, S.P., Macedo, E.A., 2009. Temperature and solvent effects in the solubility of some pharmaceutical compounds: Measurements and modeling. Eur. J. Pharm. Sci. 37, 499–507.

Mota, F.L., Queimada, A.J., Andreatta, A.E., Pinho, S.P., Macedo, E.A., 2012. Calculation of drug-like molecules solubility using predictive activity coefficient models. Fluid Phase Equilib. 322–323, 48–55.

Murador, D.C., de Souza Mesquita, L.M., Vannuchi, N., Braga, A.R.C., de Rosso, V. V., 2019. Bioavailability and biological effects of bioactive compounds extracted with natural deep eutectic solvents and ionic liquids: advantages over conventional organic solvents. Curr. Opin. Food Sci. 26, 25–34.

Naguib, Y.M., 2000. A fluorometric method for measurement of oxygen radical-scavenging activity of water-soluble antioxidants. Anal. Biochem. 284, 93–98.

Nakatani, N., Kayano, S., Kikuzaki, H., Sumino, K., Katagiri, K., Mitani, T., 2000. Identification, quantitative determination, and antioxidative activities of chlorogenic acid isomers in prune (Prunus domestica L.). J. Agric. Food Chem. 48, 5512–5516.

Nam, M., Zhao, J., Lee, M., Jeong, J., Lee, J., 2015. Enhanced extraction of bioactive natural products using tailor-made deep eutectic solvents: application to flavonoid extraction from Flos sophorae. Green Chem. 17, 1718–1727.

Newman, D.J., Cragg, G.M., 2016. Natural products as sources of new drugs from 1981 to 2014. J. Nat. Prod. 79, 629–661.

Neto, B., Scarminio, I. S., Bruns, R. E., 2007. Como modelar misturas, in B. d. Neto, I. S. Scarminio, R. E. Bruns (Eds.), Como fazer experimentos: Pesquisa e desenvolvimento na ciência e na indústria. Campinas: Editora da UNICAMP, Campinas. pp. 367-398

Nollet, L.M.L., Gutierrez-Uribe, J.A., 2018. Phenolic compounds in food: characterization and analysis, 1st ed. CRC Press, Boca Raton.

Noubigh, A., Abderrabba, M., 2016. Solid–liquid phase equilibrium and thermodynamic properties of vanillic acid in different pure solvents. J. Mol. Liq. 223, 261–266.

215 Bibliography CHAPTER 7

Noubigh, A., Abderrabba, M., Provost, E., 2007. Temperature and salt addition effects on the solubility behaviour of some phenolic compounds in water. J. Chem. Thermodyn. 39, 297–303.

Noubigh, A., Akermi, A., 2017. Solubility and thermodynamic behavior of syringic acid in eight pure and water + methanol mixed solvents. J. Chem. Eng. Data 62, 3274–3283.

Noubigh, A., Aydi, A., Abderrabba, M., 2015. Experimental measurement and correlation of solubility data and thermodynamic properties of protocatechuic acid in four organic solvents. J. Chem. Eng. Data 60, 514–518.

Noubigh, A., Cherif, M., Provost, E., Abderrabba, M., 2008a. Solubility of some phenolic compounds in aqueous alkali metal nitrate solutions from (293.15 to 318.15) K. J. Chem. Thermodyn. 40, 1612–1616.

Noubigh, A., Cherif, M., Provost, E., Abderrabba, M., 2008b. Solubility of gallic acid, vanillin, syringic acid, and protocatechuic acid in aqueous sulfate solutions from (293.15 to 318.15) K. J. Chem. Eng. Data 53, 1675–1678.

Oliveira, I., Sousa, A., Ferreira, I.C.F.R., Bento, A., Estevinho, L., Pereira, J.A., 2008. Total phenols, antioxidant potential and antimicrobial activity of walnut (Juglans regia L.) green husks. Food Chem. Toxicol. 46, 2326–2331.

Orhan, I.E., Suntar, I.P., Akkol, E.K., 2011. In vitro neuroprotective effects of the leaf and fruit extracts of Juglans regia L. (walnut) through enzymes linked to Alzheimer’s disease and antioxidant activity. Int. J. Food Sci. Nutr. 62, 781–786.

Orqueda, M.E., Rivas, M., Zampini, I.C., Alberto, M.R., Torres, S., Cuello, S., Sayago, J., Thomas-Valdes, S., Jiménez-Aspee, F., Schmeda-Hirschmann, G., Isla, M.I., 2017. Chemical and functional characterization of seed, pulp and skin powder from chilto (Solanum betaceum), an Argentine native fruit. Phenolic fractions affect key enzymes involved in metabolic syndrome and oxidative stress. Food Chem. 216, 70–79.

Oswal, S.L., Oswal, P., Modi, P.S., Dave, J.P., Gardas, R.L., 2004. Acoustic, volumetric, compressibility and refractivity properties and Flory’s reduction parameters of some

216 Bibliography CHAPTER 7 homologous series of alkyl alkanoates from 298.15 to 333.15 K. Thermochim. Acta 410, 1– 14.

Ozturk, B., Parkinson, C., Gonzalez-Miquel, M., 2018. Extraction of polyphenolic antioxidants from orange peel waste using deep eutectic solvents. Sep. Purif. Technol. 206, 1–13.

Paiva, A., Craveiro, R., Aroso, I., Martins, M., Reis, R.L., Duarte, A.R.C., 2014. Natural deep eutectic solvents - Solvents for the 21st century. ACS Sustain. Chem. Eng. 2, 1063–1071.

Panth, N., Paudel, K.R., Karki, R., 2016. Phytochemical profile and biological activity of Juglans regia. J. Integr. Med. 14, 359–373.

Paradiso, V.M., Clemente, A., Summo, C., Pasqualone, A., Caponio, F., 2016. Towards green analysis of virgin olive oil phenolic compounds: Extraction by a natural deep eutectic solvent and direct spectrophotometric detection. Food Chem. 212, 43–47.

Patil, P., Tatke, P., 2015. Betaine: Chemistry, Analysis, Function and Effects, 1st ed, Food and Nutritional Components in Focus. Royal Society of Chemistry, Cambridge.

Pena-Pereira, F., Kloskowski, A., Namieśnik, J., 2015. Perspectives on the replacement of harmful organic solvents in analytical methodologies: a framework toward the implementation of a generation of eco-friendly alternatives. Green Chem. 17, 3687–3705.

Pena-Pereira, F., Namieśnik, J., 2014. Ionic liquids and deep eutectic mixtures: Sustainable solvents for extraction processes. ChemSusChem 7, 1784–1800.

Peng, F., Xu, P., Zhao, B.-Y., Zong, M.-H., Lou, W.-Y., 2018. The application of deep eutectic solvent on the extraction and in vitro antioxidant activity of rutin from Sophora japonica bud. J. Food Sci. Technol. 55, 2326–2333.

Pereira, J.A., Oliveira, I., Sousa, A., Valentão, P., Andrade, P.B., Ferreira, I.C.F.R., Ferreres, F., Bento, A., Seabra, R., Estevinho, L., 2007. Walnut (Juglans regia L.) leaves: Phenolic compounds, antibacterial activity and antioxidant potential of different cultivars. Food Chem. Toxicol. 45, 2287–2295.

217 Bibliography CHAPTER 7

Perusquía, M., Mendoza, S., Bye, R., Linares, E., Mata, R., 1995. Vasoactive effects of aqueous extracts from five Mexican medicinal plants on isolated rat aorta. J. Ethnopharmacol. 46, 63–69.

Philippi, K., Tsamandouras, N., Grigorakis, S., Makris, D.P., 2016. Ultrasound-assisted green extraction of eggplant peel (Solanum melongena) polyphenols using aqueous mixtures of glycerol and ethanol: optimisation and kinetics. Environ. Process. 3, 369–386.

Piemontese, L., Perna, F., Logrieco, A., Capriati, V., Solfrizzo, M., 2017. Deep eutectic solvents as novel and effective extraction media for quantitative determination of Ochratoxin A in wheat and derived products. Molecules 22, 121–130.

Pinela, J., Prieto, M.A., Antonio, A.L., Carvalho, A.M., Oliveira, M.B.P.P., Barros, L., Ferreira, I.C.F.R., 2017. Ellagitannin-rich bioactive extracts of Tuberaria lignosa: insights into the radiation-induced effects in the recovery of high added-value compounds. Food Funct. 8, 2485–2499.

Pinela, J., Prieto, M.A., Barreiro, M.F., Carvalho, A.M., Oliveira, M.B.P.P., Vázquez, J.A., Ferreira, I.C.F.R., 2016. Optimization of microwave-assisted extraction of hydrophilic and lipophilic antioxidants from a surplus tomato crop by response surface methodology. Food Bioprod. Process. 98, 283–298.

Pinela, J., Prieto, M.A., Pereira, E., Jabeur, I., Filomena, M., Barros, L., Ferreira, I.C.F.R., 2019. Optimization of heat- and ultrasound-assisted extraction of anthocyanins from Hibiscus sabdariffa calyces for natural food colorants. Food Chem. 275, 309–321.

Pinkus, A.G., Kautz, J.A., Ahobila-vajjula, P., 2002. Crystal structure of 3,4- dimethoxybenzoic acid. J. Chem. Crystallogr. 32, 5–8.

Pires, T.C.S.P., Dias, M.I., Barros, L., Alves, M.J., Oliveira, M.B.P.P., Santos-Buelga, C., Ferreira, I.C.F.R., 2018. Antioxidant and antimicrobial properties of dried Portuguese apple variety (Malus domestica Borkh. cv Bravo de Esmolfe). Food Chem. 240, 701–706.

Pitschmann, A., Zehl, M., Atanasov, A.G., Dirsch, V.M., Heiss, E., Glasl, S., 2014. Walnut leaf extract inhibits PTP1B and enhances glucose-uptake in vitro. J. Ethnopharmacol. 152, 599– 602.

218 Bibliography CHAPTER 7

Płotka-Wasylka, J., Rutkowska, M., Owczarek, K., Tobiszewski, M., Namieśnik, J., 2017. Extraction with environmentally friendly solvents. Trends Anal. Chem. 91, 12–25.

Prat, D., Wells, A., Hayler, J., Sneddon, H., Mcelroy, C.R., Abou-shehada, S., Dunn, P.J., 2016. CHEM21 selection guide of classical- and less classical-solvents. Green Chem. 18, 288–296.

Prausnitz, J.M., Lichtenthaler, R.N., Azevedo, E.G. de, 1999. Molecular thermodynamics of fluid-phase equilibria, 3rd ed. Prentice Hall PTR, New Jersey.

Prencipe, F.P., Bruni, R., Guerrini, A., Rossi, D., Benvenuti, S., Pellati, F., 2014. Metabolite profiling of polyphenols in Vaccinium berries and determination of their chemopreventive properties. J. Pharm. Biomed. Anal. 89, 257–267.

Prieto, M.A., Vázquez, J.A., Murado, M.A., 2015. A new and general model to describe, characterize, quantify and classify the interactive effects of temperature and pH on the activity of enzymes. Analyst 140, 3587–3602.

Prieto, M.A., Vázquez, J.A., Murado, M.A., 2014. A critical point: the problems associated with the variety of criteria to quantify the antioxidant capacity. J. Agric. Food Chem. 62, 5472–5484.

Qa’dan, F., Thewaini, A., Ali, D.A., Afifi, R., Elkhawad, A., Matalka, K.Z., 2005. The antimicrobial activities of Psidium guajava and Juglans regia leaf extracts to acne- developing organisms. Am. J. Chin. Med. 33, 197–204.

Queimada, A.J., Mota, F.L., Pinho, S.P., Macedo, E.A., 2009. Solubilities of biologically active phenolic compounds: measurements and modeling. J. Phys. Chem. B 113, 3469–3476.

Rabiei, K., Ebrahimzadeh, M.A., Saeedi, M., Bahar, A., Akha, O., Kashi, Z., 2018. Effects of a hydroalcoholic extract of Juglans regia (walnut) leaves on blood glucose and major cardiovascular risk factors in type 2 diabetic patients: a trial. BMC Complement. Altern. Med. 18, 206.

Rached, W., Zeghada, F.Z., Bennaceur, M., Barros, L., Calhelha, R.C., Heleno, S., Alves, M.J., Carvalho, A.M., Marouf, A., Ferreira, I.C.F.R., 2018. Phytochemical analysis and assessment of antioxidant, antimicrobial, anti-inflammatory and cytotoxic properties of Tetraclinis articulata (Vahl) Masters leaves. Ind. Crops Prod. 112, 460–466.

219 Bibliography CHAPTER 7

Radošević, K., Bubalo, M.C., Srček, V.G., Grgas, D., Dragičević, T.L., Redovniković, R.I., 2015. Evaluation of toxicity and biodegradability of choline chloride based deep eutectic solvents. Ecotoxicol. Environ. Saf. 112, 46–53.

Radošević, K., Čanak, I., Panić, M., Markov, K., Bubalo, M.C., Frece, J., Srček, V.G., Radovniković, I.R., 2018. Antimicrobial, cytotoxic and antioxidative evaluation of natural deep eutectic solvents. Environ. Sci. Pollut. Res. 25, 14188–14196.

Radošević, K., Ćurko, N., Gaurina Srček, V., Cvjetko Bubalo, M., Tomašević, M., Kovačević Ganić, K., Radojčić Redovniković, I., 2016. Natural deep eutectic solvents as beneficial extractants for enhancement of plant extracts bioactivity. LWT - Food Sci. Technol. 73, 45– 51.

Rajan, M., Prabhavathy, A., Ramesh, U., 2015. Natural Deep Eutectic Solvent Extraction Media for Zingiber officinale Roscoe: The Study of Chemical Compositions, Antioxidants and Antimicrobial Activities. Nat. Prod. J. 5, 3–11.

Ranic, M., Nikolic, M., Pavlovic, M., Buntic, A., Siler-Marinkovic, S., Dimitrijevic-Brankovic, S., 2014. Optimization of microwave-assisted extraction of natural antioxidants from spent espresso coffee grounds by response surface methodology. J. Clean. Prod. 80, 69–79.

Rather, M.A., Dar, B.A., Dar, M.Y., Wani, B.A., Shah, W.A., Bhat, B.A., Ganai, B.A., Bhat, K.A., Anand, R., Qurishi, M.A., 2012. Chemical composition, antioxidant and antibacterial activities of the leaf essential oil of Juglans regia L. and its constituents. Phytomedicine 19, 1185–1190.

Reggane, M., Wiest, J., Saedtler, M., Harlacher, C., Gutmann, M., Zottnick, S.H., Piechon, P., Dix, I., Müller-Buschbaum, K., Holzgrabe, U., Meinel, L., Galli, B., 2018. Bioinspired co- crystals of Imatinib providing enhanced kinetic solubility. Eur. J. Pharm. Biopharm. 128, 290–299.

Ribeiro, A.S., Estanqueiro, M., Oliveira, M.B., Lobo, J.M.S., 2015. Main benefits and applicability of plant extracts in skin care products. Cosmetics 2, 48–65.

Rice-Evans, C.A., Miller, N.J., Paganga, G., 1996. Structure-antioxidant activity relationships of flavonoids and phenolic acids. Free Radic. Biol. Med. 20, 933–956.

220 Bibliography CHAPTER 7

Roselló-Soto, E., Galanakis, C.M., Brnčić, M., Orlien, V., Trujillo, F.J., Mawson, R., Knoerzer, K., Tiwari, B.K., Barba, F.J., 2015. Clean recovery of antioxidant compounds from plant foods, by-products and algae assisted by ultrasounds processing. Modeling approaches to optimize processing conditions. Trends Food Sci. Technol. 42, 134–149.

Rostagno, M.A., Prado, J.M., 2013. Natural Product Extraction - Principles and Applications. The Royal Society of Green Chemistry, Cambridge.

Routray, W., Orsat, V., 2012. Microwave-assisted extraction of flavonoids: a review. Food Bioprocess Technol. 5, 409–424.

Ruesgas-Ramón, M., Figueroa-Espinoza, M.C., Durand, E., 2017. Application of deep eutectic solvents (DES) for phenolic compounds extraction: Overview, challenges, and opportunities. J. Agric. Food Chem. 65, 3591–3601.

Ryu, H.W., Kim, K.O., Yuk, H.J., Kwon, O.-K., Kim, J.H., Kim, D.-Y., Na, M.K., Ahn, K.-S., Oh, S.-R., 2016. The constituent, anti-inflammation, and human neutrophil elastase inhibitory activity of Gnaphalium affine. J. Funct. Foods 27, 674–684.

Rywaniak, J., Luzak, B., Podsedek, A., Dudzinska, D., Rozalski, M., Watala, C., 2015. Comparison of cytotoxic and anti-platelet activities of polyphenolic extracts from Arnica montana flowers and Juglans regia husks. Platelets 26, 168–176.

Salejda, A.M., Janiewicz, U., Korzeniowska, M., Kolniak-Ostek, J., Krasnowska, G., 2016. Effect of walnut green husk addition on some quality properties of cooked sausages. LWT - Food Sci. Technol. 65, 751–757.

Salimi, M., Ardestaniyan, M.H., Mostafapour Kandelous, H., Saeidnia, S., Gohari, A.R., Amanzadeh, A., Sanati, H., Sepahdar, Z., Ghorbani, S., Salimi, M., 2014. Anti-proliferative and apoptotic activities of constituents of chloroform extract of Juglans regia leaves. Cell Prolif. 47, 172–179.

Sampaio, B.L., Edrada-Ebel, R., Da Costa, F.B., 2016. Effect of the environment on the secondary metabolic profile of Tithonia diversifolia: A model for environmental metabolomics of plants. Sci. Rep. 6, 1–11.

221 Bibliography CHAPTER 7

Santos, A., Barros, L., Calhelha, R.C., Dueñas, M., Carvalho, A.M., Santos-Buelga, C., Ferreira, I.C.F.R., 2013. Leaves and decoction of Juglans regia L.: Different performances regarding bioactive compounds and in vitro antioxidant and antitumor effects. Ind. Crops Prod. 51, 430–436.

Santos, J.S., Deolindo, C.T.P., Esmerino, L.A., Genovese, M.I., Fujita, A., Marques, M.B., Rosso, N.D., Daguer, H., Valese, A.C., Granato, D., 2016. Effects of time and extraction temperature on phenolic composition and functional properties of red rooibos (Aspalathus linearis). Food Res. Int. 89, 476–487.

Saravanakumar, M., Raja, B., 2011. Veratric acid, a phenolic acid attenuates blood pressure and oxidative stress in L-NAME induced hypertensive rats. Eur. J. Pharmacol. 671, 87–94.

Saravanakumar, M., Raja, B., Manivannan, J., Silambarasan, T., Prahalathan, P., Kumar, S., Mishra, S.K., 2015. Oral administration of veratric acid, a constituent of vegetables and fruits, prevents cardiovascular remodelling in hypertensive rats: A functional evaluation. Br. J. Nutr. 114, 1385–1394.

Sarrou, E., Martens, S., Chatzopoulou, P., 2016. Metabolite profiling and antioxidative activity of Sage (Salvia fruticosa Mill.) under the influence of genotype and harvesting period. Ind. Crops Prod. 94, 240–250.

Schwindl, S., Kraus, B., 2017. Phytochemical study of Juglans regia L. leaves. Phytochemistry 144, 58–70.

Schwindl, S., Kraus, B., Heilmann, J., 2019. Secondary metabolites from the leaves of Juglans regia L. Biochem. Syst. Ecol. 83, 130–136.

Sedov, I.A., Stolov, M.A., Hart, E., Grover, D., Zettl, H., Koshevarova, V., Dai, C., Zhang, S., Acree, W.E., Abraham, M.H., 2015. Abraham model correlations for describing solute transfer into 2-butoxyethanol from both water and the gas phase at 298 K. J. Mol. Liq. 209, 196–202.

Shanmugam, S., Thangaraj, P., Lima, B. dos S., Chandran, R., Araújo, A.A. de S., Narain, N., Serafini, M.R., Júnior, L.J.Q., 2016. Effects of luteolin and quercetin 3-β-d-glucoside

222 Bibliography CHAPTER 7 identified from Passiflora subpeltata leaves against acetaminophen induced hepatotoxicity in rats. Biomed. Pharmacother. 83, 1278–1285.

Sharafati-Chaleshtori, R., Sharafati-Chaleshtori, F., Rafieian, M., 2011. Biological characterization of Iranian walnut (Juglans regia) leaves. Turkish J. Biol. 35, 635–639.

Sharma, P., Ravikumar, G., Kalaiselvi, M., Gomathi, D., Uma, C., 2013. In vitro antibacterial and free radical scavenging activity of green hull of Juglans regia. J. Pharm. Anal. 3, 298– 302.

Shehata, E., Grigorakis, S., Loupassaki, S., Makris, D.P., 2015. Extraction optimisation using water/glycerol for the efficient recovery of polyphenolic antioxidants from two Artemisia species. Sep. Purif. Technol. 149, 462–469.

Sheikholeslamzadeh, E., Rohani, S., 2012. Solubility prediction of pharmaceutical and chemical compounds in pure and mixed solvents using predictive models. Ind. Eng. Chem. Res. 51, 464–473.

Shi, G.-J., Li, Y., Cao, Q.-H., Wu, H.-X., Tang, X.-Y., Gao, X.-H., Yu, J.-Q., Chen, Z., Yang, Y., 2019. In vitro and in vivo evidence that quercetin protects against diabetes and its complications: A systematic review of the literature. Biomed. Pharmacother. 109, 1085– 1099.

Shi, P., Tsai, C.-L., 2002. Regression model selection – a residual likelihood approach. J. R. Stat. Soc. Ser. B (Statistical Methodology). 64, 237–252.

Shin, S.W., Jung, E., Kim, S., Lee, K.E., Youm, J.K., Park, D., 2013. Antagonist effects of veratric acid against UVB-induced cell damages. Molecules 18, 5405–5419.

Shokoohinia, Y., Rashidi, M., Hosseinzadeh, L., Jelodarian, Z., 2015. Quercetin-3-O-β-d- glucopyranoside, a dietary flavonoid, protects PC12 cells from H2O2-induced cytotoxicity through inhibition of reactive oxygen species. Food Chem. 167, 162–167.

Silveira, C.C., Braga, A.L., Kaufman, T.S., Lenardão, E.J., 2004. Synthetic approaches to 2- tetralones. Tetrahedron 60, 8295–8328.

Smith, E.L., Abbott, A.P., Ryder, K.S., 2014. Deep eutectic solvents (DESs) and their applications. Chem. Rev. 114, 11060–11082.

223 Bibliography CHAPTER 7

Sonnenberg, A., Baars, J., Hendrickx, P., 2007. IEA Bioenergy Task 42 Biorefinery.

Soto-Maldonado, C., Vergara-Castro, M., Jara-Quezada, J., Caballero-Valdés, E., Müller- Pavez, A., Zúñiga-Hansen, M.E., Altamirano, C., 2019. Polyphenolic extracts of walnut (Juglans regia) green husk containing juglone inhibit the growth of HL-60 cells and induce apoptosis. Electron. J. Biotechnol. 39, 1–7.

Stalikas, C.D., 2007. Extraction, separation, and detection methods for phenolic acids and flavonoids. J. Sep. Sci. 30, 3268–3295.

Stampar, F., Solar, A., Hudina, M., Veberic, R., Colaric, M., 2006. Traditional walnut liqueur – cocktail of phenolics. Food Chem. 95, 627–631.

Stojiljković, D., Arsić, I., Tadić, V., 2016. Extracts of wild apple fruit (Malus sylvestris (L.) Mill., Rosaceae) as a source of antioxidant substances for use in production of nutraceuticals and cosmeceuticals. Ind. Crops Prod. 80, 165–176.

Stoval, D.M., Givens, C., Keown, S., Hoover, K.R., Barnes, R., Harris, C., Lozano, J., Nguyen, M., Rodriguez, E., Jr, W.E.A.J., Abraham, M.H., 2005. Solubility of crystalline nonelectrolyte solutes in organic solvents: mathematical correlation of 4-chloro-3-nitrobenzoic acid and 2-chloro-5-nitrobenzoic acid solubilities with the Abraham solvation parameter model. Phys. Chem. Liq. 43, 351–360.

Stovall, D.M., Dai, C., Zhang, S., Acree, W.E., Abraham, M.H., 2016. Abraham model correlations for describing solute transfer into anhydrous 1,2-propylene glycol for neutral and ionic species. Phys. Chem. Liq. 54, 1–13.

Suganthy, N., Devi, K.P., Navabi, S.F., Braidy, N., Navabi, M.S., 2016. Bioactive effects of quercetin in the central nervous system: Focusing on the mechanisms of actions. Biomed. Pharmacother. 84, 892–908.

Sun, D., Yamada, Y., Sato, S., Ueda, W., 2016. Glycerol hydrogenolysis into useful C3 chemicals. Appl. Catal. B Environ. 193, 75–92.

224 Bibliography CHAPTER 7

Sun, J.-X., Zhao, X.-Y., Fu, X.-F., Yu, H.-Y., Li, X., Li, S.-M., Ruan, H.-L., 2012. Three new naphthalenyl glycosides from the root bark of Juglans cathayensis. Chem. Pharm. Bull. 60, 785–789.

Sun, T.F., Schouten, J.A., Trappeniers, N.J., Biswas, S.N., 1988. Measurements of the densities of liquid benzene, cyclohexane, methanol, and ethanol as functions of temperature at 0.1 MPa. J. Chem. Thermodyn. 20, 1089–1096.

Taha, N.A., Al-wadaan, M., 2011. Utility and importance of walnut, Juglans regia Linn: A review. African J. Microbiol. Res. 5, 5796–5805.

Takebayashi, J., Iwahashi, N., Ishimi, Y., Tai, A., 2012. Development of a simple 96-well plate method for evaluation of antioxidant activity based on the oxidative haemolysis inhibition assay (OxHLIA). Food Chem. 134, 606–610.

Talapatra, S.K., Karmacharya, B., De, S.C., Talapatra, B., 1988. (-)-Regiolone, an α-tetralone from Juglans regia: structure, stereochemistry and conformation. Phytochemistry 27, 3929–3932.

Taofiq, O., González-Paramás, A.M., Barreiro, M.F., Ferreira, I.C.F.R., 2017. Hydroxycinnamic acids and their derivatives: Cosmeceutical significance, challenges and future perspectives, a review. Molecules 22, 281.

Tapia, M.I., Sánchez-Morgado, J.R., García-Parra, J., Ramírez, R., Hernández, T., González- Gómez, D., 2013. Comparative study of the nutritional and bioactive compounds content of four walnut (Juglans regia L.) cultivars. J. Food Compos. Anal. 31, 232–237.

Tasneem, S., Liu, B., Li, B., Choudhary, M.I., Wang, W., 2019. Molecular pharmacology of inflammation: Medicinal plants as anti-inflammatory agents. Pharmacol. Res. 139, 126– 140.

Tatsis, E.C., Connor, S.E.O., 2016. New developments in engineering plant metabolic pathways. Curr. Opin. Biotechnol. 42, 126–132.

Teimori, M., Kouhsari, M., Ghafarzadegan, R., Hajiaghaee, R., 2009. Study of hypoglycemic effect of Juglans regia leaves and its mechanism. J. Med. Plants 9, 57–65.

225 Bibliography CHAPTER 7

Terpinc, P., Polak, T., Šegatin, N., Hanzlowsky, A., Ulrih, N.P., Abramovič, H., 2011. Antioxidant properties of 4-vinyl derivatives of hydroxycinnamic acids. Food Chem. 128, 62–69.

Thipparaboina, R., Mittapalli, S., Thatikonda, S., Nangia, A., Naidu, V.G.M., Shastri, N.R., 2016. Syringic acid: structural elucidation and co-crystallization. Cryst. Growth Des. 16, 4679–4687.

Tian, Y., Li, J., Wu, B., Zhang, Y., Tang, H., Li, Q., 2015. Determination and modeling of the solubility of 2,4-dimethoxybenzoic acid in six pure and isopropanol + ethyl acetate mixed organic solvents at temperatures from (288.15 to 323.15) K. J. Chem. Eng. Data 60, 1098– 1105.

Tiwari, B.K., 2015. Ultrasound: A clean, green extraction technology. TrAC - Trends Anal. Chem. 71, 100–109.

Tomczyk, M., Bazylko, A., Bonarewicz, J., 2012. Method development and validation for optimized separation of quercetin derivatives in selected Potentilla species using high- performance thin-layer chromatography photodensitometry method. J. Pharm. Biomed. Anal. 61, 265–270.

Tripathi, P., Tripathi, P., Kashyap, L., Singh, V., 2007. The role of nitric oxide in inflammatory reactions. FEMS Immunol. Med. Microbiol. 51, 443–452.

Tubtimdee, C., Shotipruk, A., 2011. Extraction of phenolics from Terminalia chebula Retz with water–ethanol and water–propylene glycol and sugaring-out concentration of extracts. Sep. Purif. Technol. 77, 339–346.

Vanda, H., Dai, Y., Wilson, E.G., Verpoorte, R., Hae, Y., 2018. Green solvents from ionic liquids and deep eutectic solvents to natural deep eutectic solvents. Comptes Rendus Chim. 21, 628–638.

Verma, R.S., Padalia, R.C., Chauhan, A., Thul, S.T., 2013. Phytochemical analysis of the leaf volatile oil of walnut tree (Juglans regia L.) from western Himalaya. Ind. Crops Prod. 42, 195–201.

226 Bibliography CHAPTER 7

Vieira, V., Pereira, C., Pires, T.C.S.P., Calhelha, R.C., Alves, M.J., Ferreira, O., Barros, L., Ferreira, I.C.F.R., 2019. Phenolic profile, antioxidant and antibacterial properties of Juglans regia L. (walnut) leaves from the Northeast of Portugal. Ind. Crop. Prod. 134, 347–355.

Vieira, V., Prieto, M.A., Barros, L., Coutinho, J.A.P., Ferreira, I.C.F.R., Ferreira, O., 2018. Enhanced extraction of phenolic compounds using choline chloride based deep eutectic solvents from Juglans regia L. Ind. Crops Prod. 115, 261–271.

Vieira, V., Prieto, M.A., Barros, L., Coutinho, J.A.P., Ferreira, O., Ferreira, I.C.F.R., 2017. Optimization and comparison of maceration and microwave extraction systems for the production of phenolic compounds from Juglans regia L. for the valorization of walnut leaves. Ind. Crops Prod. 115, 261–271.

Vilas-Boas, S., Brandão, P., Martins, M.A.R., Silva, L.P., Schreiner, T.B., Fernandes, L., Ferreira, O., Pinho, S.P., 2018. Solid phase studies and solubility of isomeric phenolic acids in water and organic solvents. J. Mol. Liq. 272, 1048–1057.

Vraneš, M., Zec, N., Tot, A., Papović, S., Dožić, S., Gadžurić, S., 2014. Density, electrical conductivity, viscosity and excess properties of 1-butyl-3-methylimidazolium bis (trifluoromethylsulfonyl) imide+ propylene carbonate binary mixtures. J. Chem. Thermodyn. 68, 98–108.

Wang, H., Zhang, W., 2009. Solubility of 3,5-dimethoxybenzoic acid, 4-cyanobenzoic acid, 4-acetoxybenzoic acid, 3,5-diaminobenzoic acid, and 2,4-dichlorobenzoic acid in Ethanol. J. Chem. Eng. Data 54, 1942–1944.

Wang, T.-M., Fu, Y., Yu, W.-J., Chen, C., Di, X., Zhang, H., Zhai, Y.-J., Chu, Z.-Y., Kang, T.-G., Chen, H.-B., 2017. Identification of polar constituents in the decoction of Juglans mandshurica and in the medicated egg prepared with the decoction by HPLC-Q-TOF MS2. Molecules 22, 1452.

Wang, X., Chen, D., Li, Y., Zhao, S., Chen, C., Ning, D., 2019. Alleviating effects of walnut green husk extract on disorders of lipid levels and gut bacteria flora in high fat diet-induced obesity rats. J. Funct. Foods 52, 576–586.

227 Bibliography CHAPTER 7

Wang, X., Zhao, M., Su, G., Cai, M., Zhou, C., Huang, J., Lin, L., 2015. The antioxidant activities and the xanthine oxidase inhibition effects of walnut (Juglans regia L .) fruit, stem and leaf. Int. J. Food Sci. Technol. 50, 233–239.

Wang, Y., Gao, H., Yan, W., 2010. Excess molar enthalpies of diethyl malonate+ (1-butanol, 2-methyl-1-propanol, 1-pentanol, n-heptane, and ethyl acetate) at T= (288.2, 298.2, 313.2, 328.2, 338.2, and 348.2 K) and p = 101.3 kPa. Fluid Phase Equilib. 291, 8–12.

Wei, X., Chen, D., Yi, Y., Qi, H., Gao, X., Fang, H., Gu, Q., Wang, L., Gu, L., 2012. Syringic acid extracted from Herba dendrobii prevents diabetic cataract pathogenesis by inhibiting aldose reductase activity. Evidence-based Complement. Altern. Med. 2012, 1–13.

Weibull, W., Sweden, S., 1951. A statistical distribution function of wide applicability. J. Appl. Mech. 18, 293–297.

Wen, Q., Chen, J.X., Tang, Y.L., Wang, J., Yang, Z., 2015. Assessing the toxicity and biodegradability of deep eutectic solvents. Chemosphere 132, 63–69.

Wichtl, M., Anton, R., 1999. Plantes thérapeutiques, 2nd ed. Tec.&Doc., Paris, pp. 291–293.

Wiczkowski, W., Szawara-Nowak, D., Topolska, J., Olejarz, K., Zieliński, H., Piskuła, M.K., 2014. Metabolites of dietary quercetin: profile, isolation, identification, and antioxidant capacity. J. Funct. Foods 11, 121–129.

Wojdylo, A., Oszmianski, J., Czemerys, R., 2007. Antioxidant activity and phenolic compounds in 32 selected herbs. Food Chem. 105, 940–949.

Wong, W.H., Lee, W.X., Ramanan, R.N., Tee, L.H., Kong, K.W., Galanakis, C.M., Sun, J., Prasad, K.N., 2015. Two level half factorial design for the extraction of phenolics, flavonoids and antioxidants recovery from palm kernel by-product. Ind. Crops Prod. 63, 238–248.

Xia, X., Wan, R., Wang, P., Huo, W., Dong, H., Du, Q., 2018. Toxicity of imidazoles ionic liquid

[C16mim]Cl to Hela cells. Ecotoxicol. Environ. Saf. 162, 408–414.

Yang, H., Ma, Y., Gao, C., Wang, B., ARuhan, Lin, C., Feng, H., Wang, L., Huang, J., Wang, J., 2019. Five novel diarylheptanoids from green walnut husks (Juglans regia L.). Fitoterapia 134, 221–225.

228 Bibliography CHAPTER 7

Ye, S., Saifullah, M., Grubbs, L.M., McMillan-Wiggins, M.C., Acosta, P., Mejorado, D., Flores, I., Acree, W.E., Abraham, M.H., 2011. Determination of the Abraham model solute descriptors for 3,5-dinitro-2-methylbenzoic acid from measured solubility data in organic solvents. Phys. Chem. Liq. 49, 821–829.

Yfanti, V., Lemonidou, A.A., 2019. Effect of hydrogen donor on glycerol hydrodeoxygenation to 1,2-propanediol. Catal. Today. DOI:10.1016/j.cattod.2019.04.080.

Zainal-Abidin, M.H., Hayyan, M., Hayyan, A., Jayakumar, N.S., 2017. New horizons in the extraction of bioactive compounds using deep eutectic solvents: A review. Anal. Chim. Acta 979, 1–23.

Zarei, H.A., Asadi, S., Iloukhani, H., 2008. Temperature dependence of the volumetric properties of binary mixtures of (1-propanol, 2-propanol and 1,2-propanediol) at ambient pressure (81.5 kPa). J. Mol. Liq. 141, 25–30.

Zhang, J., Yang, L., Ruan, X., Sheng, Z., Yu, M., Zheng, B., Zhang, J., Li, X., Wang, Q., 2018a. Synthesis and herbicidal activity of 4,8-DHT and its derivatives. Ind. Crop. Prod. 111, 755– 767.

Zhang, K., Pei, Z., Wang, D., 2016a. Organic solvent pretreatment of lignocellulosic biomass for biofuels and biochemicals: A review. Bioresour. Technol. 199, 21–33.

Zhang, Q., 2015. Effects of extraction solvents on phytochemicals and antioxidant activities of walnut (Juglans regia L.) green husk extracts. Eur. J. Food Sci. Technol. 3, 15–21.

Zhang, Q.W., Lin, L.G., Ye, W.C., 2018b. Techniques for extraction and isolation of natural products: A comprehensive review. Chin. Med. 13, 1–26.

Zhang, W., He, K., Pu, Q., 2014. Antimicrobial and antioxidant activities of extracts from walnut green husks. Chinese J. Appl. Environ. Biol. 20, 87–92.

Zhang, X., Huang, H., Zhao, X., Lv, Q., Sun, C., Li, X., Chen, K., 2015. Effects of flavonoids-rich Chinese bayberry (Myrica rubra Sieb. et Zucc.) pulp extracts on glucose consumption in human HepG2 cells. J. Funct. Foods 14, 144–153.

229 Bibliography CHAPTER 7

Zhang, Y., Guo, F., Cui, Q., Lu, M., Song, X., Tang, H., Li, Q., 2016b. Measurement and correlation of the solubility of vanillic acid in eight pure and water + ethanol mixed solvents at temperatures from (293.15 to 323.15) K. J. Chem. Eng. Data 61, 420–429.

Zhao, B.-Y., Xu, P., Yang, F.-X., Wu, H., Zong, M.-H., Lou, W.-Y., 2015. Biocompatible deep eutectic solvents based on choline chloride: Characterization and application to the extraction of rutin from Sophora japonica. ACS Sustain. Chem. Eng. 3, 2746–2755.

Zhao, M., Jiang, Z., Liu, T., Li, R., 2014. Flavonoids in Juglans regia L. leaves and evaluation of in vitro antioxidant activity via intracellular and chemical methods. Sci. World J. 2014. DOI:10.1155/2014/303878.

Zhou, Y., Yang, B., Jiang, Y., Liu, Z., Liu, Y., Wang, X., Kuang, H., 2015. Studies on cytotoxic activity against HepG-2 Cells of naphthoquinones from green walnut husks of Juglans mandshurica Maxim. Molecules 20, 15572–15588.

Zhu, Z., Koltunov, K.Y., 2016. A convenient synthesis of 8-hydroxy-1-tetralones from naphthalene-1,8-diol. Mendeleev Commun. 26, 79–80.

Zhuang, B., Dou, L.-L., Li, P., Liu, E.-H., 2017. Deep eutectic solvents as green media for extraction of flavonoid glycosides and aglycones from Platycladi Cacumen. J. Pharm. Biomed. Anal. 134, 214–219. DOI:10.1016/j.jpba.2016.11.049.

Zorenc, Z., Veberic, R., Stampar, F., Koron, D., Mikulic-Petkovsek, M., 2017. Thermal stability of primary and secondary metabolites in highbush blueberry (Vaccinium corymbosum L.) purees. LWT - Food Sci. Technol. 76, 79–86.

230

8. Appendix

Appendix A: Supporting information of Chapter 2

A description of the mathematical expressions is presented to calculate the design distribution and to decode and code the ranges of the variables tested (presented in Table A1) for the circumscribed central composite design (CCCD):

The number of repetitions n0 at the centre point is calculated by the following formulas for k factors based on the uniform precision:

k+3 + 9 k2 + 14 k − 7 ( ) 2  = ; where: kk [A2.1] nfloork0 =+−−  2222 42(k + ) ( )

where floor designates the highest integer value smaller than the argument. The number of experiments n for k factors is given as:

k nk= +2 + 2 1 [A2.2] Independent variable coded values and natural ones of the CCCD are coded and decoded by the following expressions:

vvvvcnn=−( 0 ) and vvvvnnc=+0  [A2.3]

where vn and vc are the natural (n) and the coded (c) values in the centre of the

experimental domain, v0 is the initial value and Δvn is the increment of vn per unit of vc.

233

Table A1. Coded and natural values of the optimization parameters used in the response surface analysis applied in ME and MAE. Three independent variables, extraction time (X1: t), temperature or power (X2: T); and solvent proportion (X3: S) are combined in five level experimental design of 14 independent variable combinations and 6 replicates on the centre of the experimental domain (20 data points).

NATURAL VALUES

CODED VALUES X1: t (min) X2: T (°C) X3: S (%)

MACERATION EXTRACTION

-1.68 30 30 0 -1 54.3 42.2 20.3 0 90 60 50 +1 125.7 77.9 79.8 +1.68 150 90 100

MICROWAVE ASISTED EXTRACTION

-1.68 3 60 0 -1 11.5 85 20.3 0 24 120 50 +1 36.5 155 79.8 +1.68 45 180 100

234

Figure A1. Diagram of the different steps carried out in order to obtain an optimal phenolic extract from Juglans regia L. using the following quantification responses: total phenolics and flavonols, quercetin derivatives, 3-O-caffeoylquinic acid and yield.

235

Figure A2. The phenolic profile of J. regia leaves. The main compounds present in the matrix were one phenolic acid, 3-O-caffeoylquinic acid (P1) and two flavonols, quercetin 3-O-glucoside (P2) and quercetin O-pentoside (P3). This profile is according to the previous reported by Santos and co-workers (Santos et al., 2013).

236

Figure A3. Graphical representation of the 3D analysis as a function of each the variables involved for ME and MAE systems for the response determinations of quercetin O-pentoside content (P3) and total flavonols (FLAV). Each of the net surfaces represents the theoretical three-dimensional response surface predicted with the second order polynomial of Eqs. [2.4], [2.8], [2.11] and [2.15], respectively. The statistical design and results are described in Table 2.2. Estimated parametric values are shown in Table 2.3. The binary actions between variables are presented when the excluded variable is positioned at the optimum of the experimental domain (Table 2.4).

237

238

Appendix B: Supporting information of Chapter 3

Table B1. Coded and natural values of the optimization parameters used in the response surface analysis applied in BA and PPA. Three independent variables, extraction time (X1: t), temperature or power (X2: T); and water proportion (X3: S) are combined in five level experimental design of 14 independent variable combinations and 6 replicates on the center of the experimental domain (20 data points).

NATURAL VALUES

CODED VALUES X1: t (min) X2: T (°C) X3: S (%)

BUTYRIC ACID

-1.68 30 30 0 -1 60.4 42.2 20.3 0 105 60 50 +1 149.6 77.9 79.8 +1.68 180 90 100

PHENYLPROPIONIC ACID

-1.68 30 30 0 -1 60.4 42.2 4.1 0 105 60 10 +1 149.6 77.9 16.1 +1.68 180 90 20

239

Figure B1. Diagram of the different steps carried out to obtain an optimal phenolic extract from Juglans regia L. using the following quantification responses: total phenolics and flavonols, quercetin derivatives and 3-O-caffeoylquinic.

240

Table B2. Results of the response surface experimental plan for the eutectic mixtures with betaine and choline chloride. Optimization process of the independent variables: betaine or choline chloride (x), 1,3-propanediol (y) and water (w); expressed as mole fractions. Response criteria (dependent variables) were the results obtained by HPLC-DAD analysis for the contents of 3-O-caffeoylquinic acid (P1), trans 3-p-coumaroylquinic acid (P2), quercetin 3-O-glucoside (P3), quercetin O-glucoside (P4) and total HPLC content (Total, P1 + P2 + P3 + P4). Bt:PPD:W CC PPD:W Bt:PPD:W CC:PPD:W

x y z x y z P1 P2 P3 P4 PT P1 P2 P3 P4 PT Experimental values for the optimization process 1 0 1 0 0 1 0 1.63 0.37 4.47 3.48 9.94 1.65 0.37 4.53 4.53 10.18 2 0 0 1 0 0 1 4.95 1.07 3.98 3.03 13.03 4.92 1.05 3.94 3.00 12.92 3 0 0.67 0.33 0 0.67 0.33 4.96 1.15 10.73 9.45 26.29 4.86 1.13 10.63 9.29 25.21 4 0 0.33 0.67 0 0.33 0.67 5.62 0.81 8.91 7.51 22.85 5.62 0.80 8.96 7.52 22.90 5 0.08 0.92 0 0.3 0.7 0 0.88 0.24 2.82 2.19 6.14 0.87 0.19 2.55 2.04 5.46 6 0.08 0 0.92 0.3 0 0.7 4.95 1.11 8.14 7.99 22.19 2.29 0.46 5.33 4.03 12.10 7 0.08 0.3 0.62 0.3 0.25 0.45 0.88 0.24 2.82 2.19 6.14 1.76 0.35 4.51 3.43 10.05 8 0.08 0.6 0.32 0.3 0.45 0.25 4.25 1.12 11.15 9.90 26.43 1.23 0.25 3.28 2.43 7.18 9 0.04 0.165 0.795 0.15 0.225 0.625 4.98 1.23 10.14 10.20 26.56 4.63 1.00 10.46 9.05 25.14 10 0.04 0.46 0.5 0.15 0.35 0.5 4.19 1.10 10.64 9.34 25.27 3.95 0.81 9.47 7.92 22.15 11 0.04 0.795 0.165 0.15 0.625 0.225 1.93 0.48 5.55 4.41 12.37 2.61 0.54 6.78 5.32 15.25 Parametric and statistical values b1 4.37±0.38 0.21±0.02 2.21±0.06 1.36±0.05 8.15±1.41 0.67±0.11 0.07±0.03 0.19±0.03 0.67±0.11 2.02±0.40

Parametric b2 1.66±0.01 0.43±0.01 4.29±0.23 3.45±0.03 9.84±0.03 1.81±0.36 0.44±0.09 4.95±0.10 1.81±0.25 11.26±0.23 values b3 4.83±0.21 0.95±0.07 3.44±1.97 2.60±0.22 11.82±0.21 4.89±0.10 1.01±0.02 3.86±0.62 4.89±0.98 12.60±2.52 obtained b12 8.64±0.24 -0.15±0.01 0.89±0.53 -0.03±0.00 -7.92±0.54 5.82±1.05 1.42±0.03 13.89±1.94 5.82±0.35 30.09±0.60

(Eq. 4) b13 1.47±0.24 2.66±0.63 28.60±1.97 29.95±5.89 62.67±9.93 -6.45±1.16 -0.98±0.20 5.83±0.23 -6.45±0.13 0.86±0.09 b23 9.28±0.04 1.41±0.03 28.51±1.51 25.62±0.22 64.83±0.19 9.61±0.77 0.87±0.07 26.85±2.69 9.61±0.19 62.76±10.04

Statistical values R² 0.9422 0.9172 0.9155 0.9404 0.9502 0.9758 0.9845 0.9068 0.9482 0.9373

241

(z) Water

x

RSM

(x) Betaine or Choline cloride (y) 1,3-Propanediol

Figure B2. Schematic representation of the existing liquid phase equilibria (at 25 °C) of the three components used to design mixtures (solvents) to extract phenolic compounds from Juglans regia leaves. 0 < xbetaine < 0.08; 0 < xcholine chloride < 0.3.

242

Figure B3. Chromatographic records at 280 nm (parts A) used for the quantification of phenolic acids, and, at 370 nm (parts B) for the quantification of quercetin derivatives present in J. regia L. leaves samples. Comparison between the optimized solvent extracts (1,3-propanediol + water (PPD:W), betaine + 1,3-propanediol + water (Bt:PPD:W) and choline chloride+ 1,3-propanediol + water (CC:PPD:W). Main phenolics identified and quantified: 3-O-caffeoylquinic acid (P1), trans 3- p-coumaroylquinic acid (P2), quercetin 3-O-glucoside (P3) and quercetin O-pentoside (P4).

243

244

Appendix C: Supporting information of Chapter 4

Table C1. Overview of the literature solubility data for the syringic acid, vanillic acid and veratric acid for the binary systems addressed in this work.

a Temperature range Solubility range (g/100 g of Solute Solvent Np Reference (K) solvent)b 6 293.15 – 318.15 0.41 – 1.16 (Noubigh et al., 2008b) water 5 288 – 323 0.11 – 0.59 (Queimada et al., 2009) 2 298.15 -313.15 0.142 – 0.231 this work 9 283.15 – 323.15 8.59 – 21.18 (Noubigh and Akermi, 2017) methanol 2 298.15 -313.15 11.48 – 18.26 this work syringic 9 283.15 – 323.15 5.01 – 14.01 (Noubigh and Akermi, 2017) acid ethanol 2 298.15 -313.15 5.56 – 8.01 this work 9 283.15 – 323.15 1.73 -5.72 (Noubigh and Akermi, 2017) 2-propanol 2 298.15 -313.15 2.29 – 4.08 this work ethyl 9 283.15 – 323.15 0.41 – 1.16 (Noubigh and Akermi, 2017) acetate 2 298.15 -313.15 1.01 – 1.45 this work 6 293.15 – 318.15 0.12 – 0.28 (Noubigh et al., 2008a) water 7 293.15-323.15 0.12 - 0.44 (Zhang et al., 2016b) 2 298.15 -313.15 0.13 – 0.27 this work (Noubigh and Abderrabba, 6 293.15 – 318.15 84.96-134.49 methanol 2016) 2 298.15 -313.15 18.26 – 23.61 this work (Noubigh and Abderrabba, 6 293.15 – 318.15 10.82 – 21.29 2016) ethanol vanillic acid 7 293.15-323.15 11.21 - 23.92 (Zhang et al., 2016b) 2 298.15 -313.15 11.95 – 16.33 this work 7 293.15-323.15 7.35 – 15.72 (Zhang et al., 2016b) 2-propanol 2 298.15 -313.15 7.01 – 9.88 this work 7 293.15-323.15 5.71 – 16.22 (Zhang et al., 2016b) 1-propanol 2 298.15 -313.15 6.43 – 8.72 this work (Noubigh and Abderrabba, ethyl 6 293.15 – 323.15 3.16 – 6.56 2016) acetate 2 298.15 -313.15 2.19 – 3.11 this work 1 298.15 0.06 (Bowen et al., 2013) water 2 298.15 -313.15 0.05 – 0.09 this work 1 298.15 4.38 (Bowen et al., 2013) methanol 2 298.15 -313.15 4.42 – 7.77 this work 1 298.15 2.87 (Bowen et al., 2013) ethanol 10 278.15 – 323.3 1.39 – 8.19 (Li et al., 2013b) veratric 2 298.15 -313.15 3.05 – 5.40 this work acid 1 298.15 1.89 (Bowen et al., 2013) 2-propanol 10 278.37 – 323.2 0.73 – 5.65 (Li et al., 2013b) 2 298.15 -313.15 1.99- 3.68 this work 1 298.15 1.96 (Bowen et al., 2013) 1-propanol 10 278.37 – 323.2 0.82 – 5.46 (Li et al., 2013b) 2 298.15 -313.15 1.99- 3.88 this work 1 298.15 1.69 (Bowen et al., 2013) ethyl 10 278.3 – 323.37 0.77 – 3.30 (Li et al., 2013b) acetate veratric 2 298.15 -313.15 1.50 – 2.41 this work acid 1 298.15 3.08 (Bowen et al., 2013) 2-butanone 10 278.98 – 323.49 1.73 – 6.37 (Li et al., 2013b) 2 298.15 -313.15 3.02 – 4.73 this work

245

Table C2. Densities used to convert the units of the solubilities data simulated by the Abraham Solvation Model to molar solubilities.

Solvent Density (g cm-3) Reference 2-methyl-2-propanol 0.75946 (Egorov and Makarov, 2011) 1,2-propanediol 1.03259 (Bajić et al., 2013) 1,4-dioxane 1.02792 (Ciocirlan et al., 2014) 1-butanol 0.80580 (Dávila et al., 2012) 1-decanol 0.82550 (Dávila et al., 2012) 1-heptanol 0.81850 (Dávila et al., 2012) 1-hexanol 0.81490 (Dávila et al., 2012) 1-octanol 0.82140 (Dávila et al., 2012) 1-pentanol 0.81070 (Dávila et al., 2012) 1-propanol 0.79980 (Dávila et al., 2012) 2-butanol 0.80201 (Faranda et al., 2004) 2-butanone 0.79989 (Faranda et al., 2004) 2-methyl-1-butanol 0.81498 (García-Garabal et al., 2015) 2-Pentanol 0.80530 (Almasi, 2015) 3-methyl-1-butanol 0.87362 (Bajić et al., 2014) acetone 0.78433 (Enders et al., 2007) acetonitrile 0.77693 (Huo et al., 2007) butyl acetate 0.87600 (Malek et al., 2014) cyclohexanone 0.94276 (Ciocirlan et al., 2010) dibutyl ether 0.76409 (Monge et al., 2009) diethyl ether 0.73530 (Canosa et al., 1999) diisopropyl ether 0.71830 (Dakkach et al., 2015) dimethylformamide 0.94460 (Bernal-García et al., 2008) ethanol 0.78480 (Dávila et al., 2012) ethyl acetate 0.89471 (Oswal et al., 2004) ethylene glycol 1.10996 (Ciocirlan et al., 2014) glycerol 1.25791 (Egorov and Makarov, 2014) isobutanol 0.79841 (Wang et al., 2010) isopropanol 0.78093 (Zarei et al., 2008) methanol 0.78655 (Sun et al., 1988) methyl acetate 0.92736 (Mariano et al., 2011) propyl acetate 0.88255 (Fernández et al., 2013) propylene carbonate 1.19902 (Vraneš et al., 2014) tetrahydrofuran 0.88201 (Ivanov, 2014) water 0.99704 (Egorov and Makarov, 2014)

246

1,4 Noubigh et al.al.(2008b) [1] 1,2 Queimada et al.al. [2](2009) This Work 1,0 0,8 0,6

0,4 0,2

Solubility 100of(g/g solvent) 0,0 280 290 300 310 320 330 Temperature (K) Figure C1. Comparison of the experimental solubility data of syringic acid in aqueous systems available in literature (Noubigh et al., 2008b; Queimada et al., 2009) and obtained in this work.

247

1,6 22 a Noubigh and AkermiAkemi [3](2017) b 1,4 20 This Work Noubigh and AkermiAkemi [3](2017) 1,2 This Work 18 1,0 16 0,8 14 0,6 12 0,4 0,2 10

Solubility 100of(g/g solvent) 0,0 8 Solubility 100of(g/g solvent) 280 290 300 310 320 330 280 290 300 310 320 330 Temperature (K) Temperature (K) 16 6,0 d Noubigh and AkermiAkemi [3](2017) c Noubigh and AkermiAkemi [3](2017) 5,5 14 This Work This Work 5,0 12 4,5 4,0 10 3,5 8 3,0 2,5 6 2,0

4 Solubility 100of(g/g solvent) Solubility 100of(g/g solvent) 1,5 280 290 300 310 320 330 280 290 300 310 320 330 Temperature (K) Temperature (K) Figure C2. Comparison of the experimental solubility data of syringic acid available in literature (Noubigh and Akermi, 2017) and obtained in this work for the following organic solvents: a) ethyl acetate; b) methanol; c) ethanol; d) 2-propanol.

248

0,5 a 142 Noubigh et al. (2008a)[4] Noubigh and Abderrabba [6](2016) b 122 Zhang etet al. (2016b)[5] This Work 0,4 This Work 102 82 0,3 62

0,2 42

22 Solubility 100of(g/g solvent) Solubility 100of(g/g solvent) 0,1 2 290 300 310 320 330 290 295 300 305 310 315 320 Temperature (K) Temperature (K) 26 c 18 Zhang et al. (2016b)[5] d 23 16 This Work 20 14 17 12 14 10 11 Zhang et al. (2016)[5] 8 Noubigh andand Abderrabba [6](2016) 8 This Work

Solubility 100of(g/g solvent) 6 5 Solubility 100of(g/g solvent) 290 300 310 320 330 290 300 310 320 330 Temperature (K) Temperature (K) 7 17 Noubigh andand Abderrabba (2016)[6] f Zhang etet al. (2016b)[5] e This Work 15 This Work 6 13 5 11 4 9 3

7 Solubility 100of(g/g solvent) Solubility 100of(g/g solvent) 5 2 290 300 310 320 330 290 295 300 305 310 315 320 Temperature (K) Temperature (K) Figure C3. Comparison of the experimental solubility data of vanillic acid available in literature (Noubigh et al., 2008a; Noubigh and Abderrabba, 2016; Zhang et al., 2016b) and obtained in this work for the following solvents: a) water; b) methanol; c) ethanol; d) 2-propanol; e) 1-propanol; f) ethyl acetate

249

9 a 6,5 Bowen et alal.. ([7]2013) Bowen etet alal.. ([7]2013) b 8 Li et et alal.. ( [8]2013b) 5,5 Li et et alal.. ( [8]2013b) 7 This Work This Work 4,5 6 5 3,5

4 2,5 3 1,5 2

Solubility 100of(g/g solvent) 0,5 1 Solubility 100of(g/g solvent) 275 285 295 305 315 325 275 285 295 305 315 325 Temperature (K) Temperature (K) 6,5 c 3,5 d Bowen et al.al. ([7]2013) Bowen etet alal.. ([7]2013) 5,5 Li et et alal.. ([8]2013b) 3,0 LiLiet etal al.. ( 2013[8] b) This Work This Work 4,5 2,5

3,5 2,0

2,5 1,5

1,5 1,0

0,5 Solubility 100of(g/g solvent) 0,5 Solubility 100of(g/g solvent) 275 285 295 305 315 325 275 285 295 305 315 325 Temperature (K) Temperature (K) 7,5 e 8 f Bowen etet alal.. ([7]2013) Bowen et alal.. ([7]2013) 6,5 LiLiet etal al.. ( 2013[8] b) This Work This Work 7 5,5

4,5 6

3,5 5 2,5

1,5 4 Solubility 100of(g/g solvent) Solubility 100of(g/g solvent) 275 285 295 305 315 325 290 295 300 305 310 315 320 Temperature (K) Temperature (K)

250

0,10 g Bowen et al.al. [7](2013) 0,09 This Work

0,08

0,07 0,06

0,05

0,04 Solubility 100of(g/g solvent) 290 295 300 305 310 315 320 Temperature (K) Figure C4. Comparison of the experimental solubility data of veratric acid available in literature (Bowen et al., 2013; Li et al., 2013b) and obtained in this work for the following solvents: a) ethanol; b) 2-propanol; c) 1-propanol; d) ethyl acetate; e) 2-butanone (f) methanol (g) water.

251

a) b)

c)

Figure C5. Illustrative DSC fusion thermograms of: a) syringic acid; b) vanillic acid; c) veratric acid.

252

a) b)

c)

Figure C6. Exemplificative thermograms of successive runs for: a) syringic acid; b) vanillic acid; c) veratric acid.

253

Figure C7. Screenshot of the visual method measurement for vanillic acid when a change in the solid coloration (possible degradation) is observed.

254

Figure C8. Comparison of the experimental X–ray powder diffraction patterns of recrystallized syringic acid from solutions with DMF, acetonitrile and 2-butanone with the powder diffraction pattern of syringic acid from the manufacturer and the simulated powder diffraction pattern reported by. Thipparaboina et al. (2016).

255

Figure C9. Comparison of the experimental X–ray powder diffraction patterns of recrystallized vanillic acid (from solutions with DMF, acetonitrile, 2-butanone, methanol and water) with the powder diffraction pattern of vanillic acid from the manufacturer and the simulated diffraction powder pattern reported by Kozlevčar et al. (2006).

256

Figure C10. Comparison of the experimental X–ray powder diffraction patterns of recrystallized veratric acid (from solutions with DMF, acetonitrile, 2-butanone and methanol) with the powder diffraction pattern of veratric acid from the manufacturer and the simulated powder diffraction pattern reported by Pinkus et al. (2002).

257

Table C3. Predicted solute parameters in the Abraham solvation model, outlier solvent and ARD (%)

Compound E S A B V Outlier ARD (%) Syringic acid 1.123 1.570 0.660 0.913 1.390 methanol 21 Vanillic acid 1.144 1.442 0.839 0.648 1.190 ethyl acetate 15 Veratric acid 0.881 1.136 0.225 0.829 1.331 methanol 6

258