Fractionation of Nanoparticle Matter in Red Wines Using Asymmetrical Flow Field-Flow Fractionation.

The particle matter of wine is mainly composed of wine colloids and macromolecules. The present work develops a methodology using asymmetrical flow field-flow fractionation coupled with multi-angle light scattering, differential refractive index detector, and ultraviolet detector (AsFlFFF-MALS-dRI-UV) for the fractionation and determination of the molar mass, the hydrodynamic radius, and the apparent densities of the aggregates and macromolecules present in wine samples. The results from a set of six Argentinian high-altitude wines showed two main populations: the first population composed of wine colloids with higher UV-specific absorptivity and the second population composed of polysaccharides, such as arabinogalactans. The conformation results showed that population 1 consists of small and dense particles, while population 2 showed high molar masses and lower densities. The results demonstrated the use of AsFlFFF as a new, effective method for the fractionation and characterization of wine colloids and wine macromolecules in red wines with further potential applications.

Introduction

Wine particle matter consists of wine colloids (e.g., tannins) and wine macromolecules (polysaccharides and proteins). They are naturally present in grapes, released by microorganisms and/or generated as a result of certain treatments.[1−3]

Wine colloids are made up of a complex and heterogeneous group of compounds, such as condensed tannins[4] and other phenols,[5,1] with mannoproteins,[6] pectins, or gums. Tannins are water-soluble polyphenolic biomolecules that can interact with proteins and various other organic compounds that can form a precipitate.[7] Wine colloids are important in winemaking not only because of their antioxidant activity but also because they form complexes with proteins that may modify the astringency.[8] Wine colloids are involved in the stability of the wine.[9] These colloidal particles may grow over time, possibly through Ostwald ripening, and when the particles become sufficiently large (>1 μm), the process may lead to sedimentation.

The total phenolic fraction in red wines has been suggested to be within a range of 35–40% of the wine particle matter.[10] The most common interactions are hydrogen-bonding and hydrophobic interactions, as reported in the case of the complexation between proanthocyanidins and proteins.[11] The moiety of anthocyanins is mainly made up of the glucosides of delphinidin, malvidin, cyanidin, petunidin, and peonidin combined with their acylated forms.[12−14]

The influence of solvents, such as the ethanol content, on the tannin size distribution has been evaluated.[4] The alcohol content enhances tannin solubility as well as its capacity to form metastable colloidal aggregates. The higher the alcohol content, the higher the solubility of the tannins, thereby leading to the stabilization of smaller colloidal aggregates with a less polydisperse distribution.[5] This may be caused by the alcohol reducing the impact of hydrophobic interactions and reducing the tendency of aggregation. However, studies have reported that an increase in ionic strength, up to 100 mM of the medium, reduces the electrostatic repulsion, enhancing the tannin self-aggregation rate.[5]

The major components of the wine macromolecular material are polysaccharides, such as arabinogalactans[15−17] (about 40% of the total soluble polysaccharides) and mannoproteins.[6,17,2] Wine macromolecules are involved in sensory properties.[18,19] Many efforts are used to control these macromolecules by means of purification and filtration techniques.[20,21]

There is still a lack of suitable methodology for the characterization of the size and conformation of the particle matter that is present in wines. To date, only a few studies have investigated tannins in extracts and the wine particle matter in wines as a result of the difficulties in separating and isolating the macromolecules and colloidal aggregates.[4,22,23]

A previous study suggests that the asymmetrical flow field-flow fractionation (AsFlFFF) technique may be useful for the development of a systematic method for separation and characterization of wine colloids and macromolecules.[4] AsFlFFF uses the interactions of the particles with a flow field close to a membrane as a way of separating them. This makes it possible to run the separation in a liquid with a set ethanol concentration, pH, and electrolyte and avoid the aggregates from changing during the separation.

Separation in the AsFlFFF operates based on the diffusion coefficient of the analyte, which provides its hydrodynamic radius.[24] From the addition of different detectors, such as multi-angle light scattering (MALS), differential refractive index (dRI), and ultraviolet (UV), detailed information can be obtained over the entire molar mass distribution of the particles and polymers. Furthermore, conformational information, such as apparent density (ρ̑ app) and hydrodynamic radius (rH), can be obtained throughout the size distribution by a combination of different detector and elution properties.[25]

This work is intended to evaluate if AsFlFFF can be used as a method for fractionation of red wine colloids and macromolecules. The aim is to develop the AsFlFFF methodology to analyze the particle sizes and compositional and microstructural properties using the information from combinations of detector responses and elution data. In addition, the interpretations are supported by chemical analysis of the wine fractions from AsFlFFF and dialysis experiments.

Materials and Methods

Samples

The wines used in the present study were purchased from Cafayate, Salta, Argentina, in June 2018. The selected wines are known as high-altitude wines from vineyards located around 2000 m above sea level (m.a.s.l.) in bodegas surrounding Cafayate, Calchaquí Valle, Salta, Argentina. Table summarizes their characteristics, including grape variety, origin, vintage, and altitude of the red wines. All of these selected wines are rather bold, medium tannic, dry, and medium acidic quality wines with comparable character.

Table 1

List of Grape Variety, Origin, Vintage, and Altitude of Wines in This Study

number	code	sample	grape variety	region	origin	vintage	altitudea
1	CAB	Domingo Molina	Cabernet Sauvignon	Salta	Argentina	2013	2200
2	MAL-1	Quebrada de las Flechas	Malbec	Salta	Argentina	2015	1900
3	MAL-2	El Tapao del Cese	Malbec	Salta	Argentina	2016	1920
4	MALCAB-1	San Pedro de Yacochua	Malbec, 80%; Cabernet Sauvignon, 20%	Salta	Argentina	2013	2035
5	MALCAB-2	Tapadito	Malbec, 70%; Cabernet Sauvignon, 30%	Salta	Argentina	2012	1920
6	TANN	Coquena	Tannat	Salta	Argentina	2015	2042

The altitude is given in meters above sea level (m.a.s.l.).

Chemicals

The water used was Milli-Q water (Millipore Corporation, Molsheim, France) with a conductivity of 18 MΩ cm–1. High-performance liquid chromatography (HPLC)-grade methanol and formic acid were used as the solvents for the HPLC analysis and were obtained from Sigma-Aldrich. Ethanol (99.5%) was obtained from VWR (Fontenay-sous-Bois, France). The phenolic standards, cyanidin chloride, malvidin chloride, delphinidin chloride, p-coumaric acid, ellagic acid, resveratrol, gallic acid, caffeic acid, polydatin, epicatechin gallate, epicatechin, vanillin, and vanillic acid, were purchased from Extrasynthese (Genay, France). Bovine serum albumin (BSA) was purchased from Sigma-Aldrich (St. Louis, MO, U.S.A.). Tartaric acid, NaN3, and NaNO3, p.a. grade, were from Merck (Darmstadt, Germany). Folin–Ciocalteu reagent was purchased from Sigma-Aldrich (St. Louis, MO, U.S.A.), and sodium carbonate was purchased from Merck (Darmstadt, Germany). HCl was purchased from VWR International (Rue Carnot, France). The sugars used as standards were l(+)-rhamnose monohydrate, d(+)-galactose, and d(+)-glucose from Merck, l(+)-arabinose from VWR Chemicals, d(+)-xylose from PanReac AppliChem (Darmstadt, Germany), and d(+)-mannose from Sigma-Aldrich.

AsFlFFF System and Operation

The AsFlFFF instrument was an Eclipse 3+ separation system (Wyatt Technology, Dernbach, Germany). The system was connected to a Dawn Heleos II MALS detector and an Optilab T-rEX dRI detector (both from Wyatt Technology), both operated at a wavelength of 658 nm. The UV detector (Jasco Corporation, Tokyo, Japan) was operated at a wavelength of 280 nm. An Agilent 1100 series isocratic pump with an in-line vacuum degasser and an Agilent 1100 autosampler (Agilent Technologies, Waldbronn, Germany) delivered the carrier liquid and handled sample injection onto the AsFlFFF channel, respectively. Different ultrafiltration membranes, such as regenerated cellulose (RC) and poly(ether sulfone) (PES), with 5 and 10 kDa cutoff were evaluated. The RC ultrafiltration membrane with cutoff at a molar mass of 10 kDa (Microdyn-Nadir GmbH, Wiesbaden, Germany) was used to minimize the adhesion of wine material evaluated as discoloration. The 10 kDa cutoff corresponds to a radius of 2.70 nm and an elution time of 1.50 min after the void (from a protein calibration). The channel used was a long channel (Wyatt Technology) having a trapezoidal geometry (tip–tip length of 26.5 cm and inlet and outlet widths of 2.6 and 0.6 cm, respectively). The spacer formed a trapezoidal channel and had a thickness of 350 μm. The actual channel thickness was determined to 309 μm, obtained by calibration with BSA according to the procedure described in the literature.[25]

AsFlFFF Theory

The separation in AsFlFFF takes place in a channel along which a carrier liquid flow is pumped, resulting in a laminar flow with a parabolic velocity profile. The sample is injected into the channel via a sample inlet and transported in an axial flow along the channel by the carrier liquid. The cross-flow is a force flow perpendicular to the axial flow direction, causing a concentration polarization at the membrane surface. The polarization effect depends upon the diffusion coefficient of the molecules or particles. Rapidly diffusing objects are, on average, further away from the membrane surface, while slowly diffusing objects are, on average, closer to the membrane surface. The flow profile of the axial flow provides the separation, with the smaller objects eluted earlier and the larger objects eluted later.[26] It is common to use cross-flows decaying with time to obtain a more even selectivity and shorten analysis times for samples with wide size distributions.

AsFlFFF and Data Analysis

For the fractionation of red wine samples, various combinations of the parameters were examined. The separation conditions used different cross-flow profiles, i.e., constant, linear, and exponential decay, including changes in the decay rate. Detector flow, injected volume, and accumulation wall membrane type were also varied.

A detector flow rate of 0.5 mL/min was used and kept constant during analyses. The calculated void time[25] (t0) was 0.7 min, and the sample injection volume was 100 μL, with an injection flow of 0.2 mL/min. The initial focus and injection step were set to 5 min with an additional 5 min of focusing to allow for enough time to inject all of the sample into the channel. The initial cross-flow rate during elution was set to 2.5 mL/min with an exponential decay of 4 min (half-life), reaching a constant cross-flow of 0.22 mL/min at 31 min for 10 min. At the end of each analysis, a blank was run to flush the channel. The blank was also used for the dRI baseline subtraction as a result of the drift caused by the pressure variation of the cross-flow decay. The liquid carrier was an aqueous solution with pH 3.6 and 13% (v/v) ethanol with 20 mM tartaric acid (30 mM ionic strength) as a model of wine-like conditions. To further identify the chemical properties of the objects of a characteristic wine sample based on UV absorption at 280 nm and the MALS signal, fractions were collected from 14 repeated injections for one of the samples (CAB) to allow for more detailed chemical analysis.

Light scattering data were processed using ASTRA software, version 6.1.7.17 (Wyatt Technology). For mass calculation, the Berry fitting[27,28] was used to assign the data and dn/dc values for polyphenols, proteins, and polysaccharides in 13% ethanol were calculated and described in section .

Hydrodynamic Radii Determination and Apparent Densities

The hydrodynamic radius (rH) was estimated directly from the elution time and AsFlFFF retention theory[25] bywhere f(t) is a function that depends upon the axial flow rate along the channel, cross-flow, channel geometry, viscosity, temperature, and diffusion coefficient of eluting analytes. i refers to each data collecting time in the fractogram. The function f(t) was evaluated using a MATLAB app as described in the literature.[25]

The apparent densities refer to the average concentration of particle matter (in a particle) or polymer segments (of a polymer) within a spherical shape of the same hydrodynamic radius.[29] The densities obtained from molar mass and the hydrodynamic radius assume a homogeneous distribution of mass and a spherical shape. The apparent density, ρ̂, for component i of the sample is calculated aswhere m is the molar mass and V(r) is the volume of a sphere with hydrodynamic radius rH. Because the density refers to the concentration of material in the corresponding sphere rather than an actual density, the density obtained should be considered an apparent property. The mass-weighted average apparent density of a population was obtained usingwhere m is the mass flow in each time slice and j refers to the peak number.

Determination of dn/dc

The specific refractive index increment, dn/dc, is a crucial parameter for the molar mass determination and the quantifications of dRI data obtained in the AsFlFFF analyses. Thereby, it influences all quantitative parameters, such as molar mass, density, UV-specific absorptivity, and determinations of concentrations. The dn/dc value is obtained assuming that the partial molar properties of the components are constant within the range of concentrations used.

The chemical components of the colloidal and macromolecular fractions of the wine solids have different dn/dc values. The principal solid components of wine, polysaccharides, proteins, and polyphenols, have dn/dc values of 0.15, 0.185, and 0.248 or 0.266 mL/g according to the literature.[4,30] Precisely, the later values were reported for water-soluble tannins of 0.248 and 0.266 mL/g for water-insoluble tannin.[4]

In the wine model carrier, 13% ethanol, dn/dc can be obtained assuming equal specific refractive index at the actual ethanol concentration using eq where (dn/dc)LC refers to the specific refractive index increment of the material in the carrier liquid (LC), 13% ethanol and pH 3.6, or in aqueous solution (aq), n refers to the refractive index, and ρ is the partial density of the component in the aqueous solution.

The equation is a simple extrapolation assuming constant and equal partial densities of the solute in aqueous media and 13% (w/w) ethanol. The assumption is supported by previous results.[31] It can also be noted that the correction factor in eq is small and has a limited impact on the results. A constant density of 1.5 g/mL has been used for all three principal wine components. dn/dc values of 0.14, 0.18, and 0.26 mL/g were used for polysaccharides, proteins, and polyphenols in the wine model carrier, respectively. The material in the AsFlFFF analyses consists of, in principal, two populations: one dense, UV-absorbing population of smaller objects (r of about 3 nm), wine colloids, and one less dense, UV-neutral population of larger objects (10–13 nm), wine macromolecule population (see the Results and Discussion). The challenge is that two peaks with different chemical compositions were found with a limited knowledge of their composition. However, the separation of the colloidal and macromolecular matter through dialysis allows for determination of the refractive index increment of the total solids of the retentate (dn/dc)TS. In the analyses, it is assumed that the contributions of the components are additive and equal to the mass average of the componentswhere WC refers to wine colloids, WM refers to wine macromolecules, and TS refers to total solids.

The concentration of each species is obtained from the area, A, of the integrated dRI signal.Applying eq in eq gives a second-order eq . The equation contains two unknowns, (dn/dc)WC and (dn/dc)WM. To be able to solve the equation, we assume that the wine macromolecular fraction mainly consists of polysaccharides and, thus, (dn/dc)WM being 0.14 mL/g (in the carrier liquid). With this assumption, we obtain the solution for (dn/dc)WC as eq .The refractive index, n, of the total solids from the dialyses and freeze drying was measured in the carrier liquid of the AsFlFFF using a digital refractometer (HI 96801, Hanna Instrument, Woonsocket, RI, U.S.A.). The properties of eq are illustrated in the Supporting Information.

Error Estimation of dn/dc

The key experimental parameter is the specific refractive index increment of the total colloidal and macromolecular material, (dn/dc)TS, of the wine obtained after freeze drying of the retentate after dialyses of the wine samples. The experimental error from the measurements has been estimated to be within the range of ±0.005 mL/g (about 3% relative error). The comparable high error may be there as a result of a slight turbidity originating from a somewhat incomplete solubility after freeze drying. This error in the measured (dn/dc)TS propagates through eq to the estimated (dn/dc)WC.The propagated errors as a function of (dn/dc)TS and the ratio AWM/AWC for all samples are shown in Table . The general properties of the error function are shown in section 3 of the Supporting Information. The errors range from 15% obtained for the medium range of (dn/dc)TS and low area ratio (e.g., sample CAB) to 40% for the samples with a low (dn/dc)TS and a small area of the wine colloid population (MAL-1 and MALCAB-2). These errors have consequences for all quantitative estimations of the wine colloids. Most of the results are directly proportional to (dn/dc)WC; thus, the error of the concentration of the wine colloids (cWC), specific absorptivity (εWC), molar mass (MWWC), and apparent density (ρ̂WC) are equal to the estimated error for (dn/dc)WC. The conclusion is that only differences in magnitude can be interpreted.

Table 2

Error Estimation as a Result of Error in Measured (dn/dc)TS

	(dn/dc)TSa (mL/mg)	AWM/AWCb	E(dn/dc)WC (%)	EcTSc (%)
CAB	0.204	0.20	15	11.6
MAL-1	0.158	2.20	37	9.4
MAL-2	0.200	1.34	18	5.4
MALCAB-1	0.181	1.46	23	7.0
MALCAB-2	0.163	6.97	43	3.2
TANN	0.203	1.19	17	5.5

From Table .

From Table .

Refers to the error in the concentration of total solids obtained in the AsFlFFF analysis.

Table 4

Retained Solids, dn/dc of Total Solids, and Total Protein Content

sample	retained total solidsa (mg/mL)	dn/dcb of total solids (mL/g)	protein contentc (mg/mL)
CAB	3.19	0.204	0.107
MAL-1	2.29	0.158	0.194
MAL-2	2.71	0.200	0.083
MALCAB-1	3.77	0.181	0.111
MALCAB-2	1.82	0.163	0.085
TANN	4.09	0.203	0.099

Total concentration of the solids determined as mass after dialysis in relation to the initial mass volume.

dn/dc of the total solids present in wine determined by a digital refractometer.

Total protein content expressed in milligrams per milliliter.

Table 3

Average Values Obtained from the AsFlFF MALS–dRI Signals for the Wine Colloids (WC) and Wine Macromolecules (WM)

code	elution time (min)		rHa (nm)		MWb (×103, g/mol) range		MWc (×103, g/mol) average		dRI aread (×106, Ai)
population	WC	WM	WC	WM	WC	WM	WC	WM	WC	WM
CAB	14.0–18.5	19.5–26.0	2.9	11.2	13–53	81–306	47.1 (0.7)	157 (0.4)	35.7	7.24
MAL-1	14.5–17.0	18.5–28.0	2.7	13.1	18–28	60–1480	22.2 (1.9)	277 (0.4)	6.78	14.9
MAL-2	14.0–17.5	18.0–27.0	2.8	11.6	9–32	52–578	25.1 (1.2)	200 (0.4)	11.1	14.9
MALCAB-1	14.0–17.5	18.5–26.0	2.8	11.5	19–49	75–844	44.0 (1.1)	252 (0.4)	17.6	25.7
MALCAB-2	15.0–18.0	19–28.0	3.5	13	15–34	96–2040	24.5 (2.2)	303 (0.4)	1.85	12.9
TANN	14.8–17.5	20–27.0	2.9	10.4	13–38	62–879	26.3 (1.1)	205 (0.4)	21.4	25.4

Hydrodynamic radii, rH, is estimated at the peak mode from the MALS fractogram.

MW range is the molar mass range on which the MW is based.

MW is the weight-average molar mass from the MALS distribution, and the fitting error is in parentheses.

Area of the population peak based on the dRI signal and elution time.

In Table , the error estimation as a result of error in the measured (dn/dc)TS is estimated to be the range of the values obtained at a measurement considering the uncertainty range of the (dn/dc)TS measurements. The sensitivity increases with a decreasing (dn/dc)TS and increasing area ratio; AWM/AWC, (dn/dc)TS values, and dRI areas are taken from Tables and 4.

Absolute Concentration Determination

The absolute concentration of the material (mg/mL) in the population is calculated using the following equation:where the dn/dc value is the characteristic value for each population j, Fout is the detector flow equal to the exit flow of the channel, and Vinj is the injection volume.

The extinction coefficient (ε0) of the material in the peaks can be obtained by comparing the ratio of the UV intensity of the peak with the intensity of the actual material in the peak (obtained from the dRI signal) and using BSA as a reference. The extinction coefficient is given bywhere IUV, is the UV intensity of peak j, IUV,BSA is the UV intensity of BSA, εBSA is 0.66 mL mg–1 cm–1, and IdRI, is the dRI intensity of peak j.

Retained Total Solids, Dialysis, and Freeze Drying

The samples were concentrated and purified from soluble components through a dialysis experiment that was performed for 6 days a using standard regenerated cellulose (RC) tubing with 3.5 kDa molecular weight cutoff (Spectra/Por from Roth, Karlsruhe, Germany). The dialyses were against the carrier liquid during 4 days, followed by dialyses against deionized water for 2 days. The liquid was changed twice daily. Then, the samples were frozen at −20 °C in a 1–2 cm thick layer and freeze-dried (Epsilon 2-6D LSC plus, Osterode, Germany) for 6 days. Approximately 25–30 mL of each wine sample resulted in about 50–100 mg of lyophilized material.

The solid content after freeze drying was determined gravimetrically after correcting for the dilution of the wine during the dialysis. A parallel experiment was run using a 14 kDa membrane. The magnitude of the results was similar but on average 10% lower.

Protein Content Determination

The nitrogen content was analyzed by nitrogen combustion with the Dumas method.[32] Protein was calculated using the factor N × 6.25. A protein analyzer (Flash EA, 1112 Series, Thermo Electron Corp., Waltham, MA, U.S.A.) was used for the protein quantification. Approximately 50 mg of each sample was weighed into a tin capsule and introduced to a combustion reaction in automated equipment. Nitrogen oxides formed are determined by gas chromatography. Aspartic acid was used as a reference.

Measurement of the Total Phenolic Content

The total phenolic content (TPH), was determined according to the literature[33] The method is based on the oxidation of the phenolic compounds to phenolates using the Folin–Ciocalteu reagent in a saturated solution of sodium carbonate, resulting in a blue molybdenum–tungsten complex. The Folin–Cicocalteu reagent was diluted 10 times (2.5 mL), and sodium carbonate (1 mL) and 50 μL of the wine sample (diluted 25 times) were mixed for 5 s and heated for 30 min at 45 °C. The absorbance at 765 nm was read after cooling at room temperature. The absorbance of each wine sample was compared to a standard calibration curve made from gallic acid. The phenolic content was measured in the wine samples before and after dialysis.

Recovery Methods

The recovery values of the AsFlFFF analyses were evaluated in three different ways. (i) The analyzable fraction: The dRI area analyzed (wine colloid and wine macromolecule area) is compared to the total dRI area that also includes non-separated areas between peaks and the area of material eluted after the cross-flow was stopped during the AsFlFFF analysis. (ii) The recovery of the UV absorbing material of the AsFlFFF analysis: The total analyzed material obtained from the integrated UV signal of the sample is compared to the total integrated UV signal of the non-separated sample (without cross-flow) passing through the AsFlFFF system. (iii) The total mass recovery: The total recovered mass in the AsFlFFF experiments is compared to an alternative method of determining the total wine colloids and wine polymers. The colloidal and macromolecular fraction from wine was isolated using dialysis over a 3.5 kDa membrane, followed by freeze drying (referred to as “total solids”).

Identification of the Phenolic Compounds through Reverse-Phase High-Performance Liquid Chromatography (RP-HPLC)

The polyphenols were analyzed after acid hydrolysis that was performed for 1 h at 90 °C, by refluxing 1.5 M HCl, to hydrolyze the glycosidic bonds of the condensed or complex structures.

The phenolic compounds were identified using HPLC (Agilent Technologies 1260 Infinity II, Palo Alto, CA, U.S.A.), equipped with a quaternary pump with an auto injector and degasser (G1311C), a column oven set at 25 °C, and a diode array detector (DAD, G1315D). The elution of phenolic and anthocyanin compounds was monitored at wavelengths of 280 and 530 nm, respectively. The column was a 3.0 mm × 100 mm × 2.7 μm Halo C18 reversed-phase column (Hichrom, Wilmington, DE, U.S.A.). The flow rate was 0.6 mL/min, and the injection volume was 20 μL. The mobile phase consisted of two eluents: (A) 0.1% formic acid/water and (B) methanol. The gradient was achieved as follows: 25% B at 0 min, 90% B after 10 min, 25% B after 16 min, and 95% B constant for 4 min to reach 20 min. The method is a modification of a method described previously in the literature.[34]

The identification of polyphenols and anthocyanins was performed through comparisons to standards using the retention time (tr) and the ultraviolet–visible (UV–vis) spectra. For this purpose, standard solutions for the corresponding phenolic compounds were prepared, and their characteristic retention times and their UV–vis spectra were recorded.

Identification of Sugar Compounds in the Polysaccharide Fraction Using High-Performance Anion-Exchange Chromatography (HPAEC)

The polysaccharides from the collected fractions in both peaks were hydrolyzed using the standardized method for acid hydrolysis,[35] with a slight modification. In short, 2 mL of the collected fraction was hydrolyzed by adding 75 μL of 72% sulfuric acid. The samples were then incubated for 1 h at 120 °C.

The concentration of monomeric sugars was measured using HPAEC coupled with pulsed amperometric detection in an ICS-3000 chromatography system (Dionex Corp., Sunnyvale, CA, U.S.A.). Deionized water was used as the eluent at a flow rate of 1 mL/min. The standards were treated with acid to obtain the sugar recovery factor for each sugar. Anhydrous corrections of 0.88 for pentoses and 0.90 for hexoses were used.

Results and Discussion

Fractionation of Wine Samples Using AsFlFFF

The results of the fractionation in the red wine samples are shown in Figure . In this figure, different types of particle and macromolecular matter can be detected by the complementary detectors. The figure shows the dRI signal (blue) that is proportional to the concentration. It also shows the MALS signal (red) that is proportional to the concentration and exponentially proportional to the radius. The UV detector, at 280 nm, providing information about the presence of the UV-active groups, such as polyphenols and/or proteins, is shown in green.

Figure 1

Fractograms from AsFlFFF showing the molar mass (MW) determined by MALS–dRI detection (black circles) and hydrodynamic radius (rH) (pink circles): (a) CAB, (b) MAL-1, (c) MAL-2, (d) MALCAB-1, (e) MALCAB-2, and (f) TANN. MALS signal at 90° scattering angle (red trace), UV at 280 nm (green trace), and dRI response (blue trace).

The fractograms show the presence of two main populations within all samples, although the two populations vary from one sample to another. The fractograms are reproducible (after repeated runs and at different occasions). The peak retention time varies less than 2% for the populations (from 14 repeated runs of sample CAB). The samples have been analyzed at different occasions (9 months between them, samples stored in closed bottles, after sampling of 1–2 mL using a 0.8 mm syringe and the bottle stored at +4 °C) with equal results.

Molar mass and rrms radius are obtained from the MALS signal by fitting the angular dependence of the intensity. The error of the fitting is provided as an experimental error in Table on the molar mass data. The rrms radius of the first population was not possible to determine as a result of isotropic light scattering, and thus, these data are not included in further analyses.

In Figure a, the CAB sample, the first population shows higher signals in the MALS, dRI, and UV detectors, with its maximum at about 17 min, while population 2 shows lower signals for the MALS–dRI–UV detectors. Conversely, Panels b–f of Figure showed higher signals for the MALS–dRI, confirming a higher content of the larger fraction in these particular wine samples, although the UV signal remains low.

In Figure , the first population is UV-absorbing in all samples and the second population is UV-neutral. Thus, it can be concluded that the first and second populations are chemically different. The first population can be assumed to be containing wine colloids with partly aromatic structures, and the second population can be assumed to be containing mainly wine macromolecules without aromatic structures.

Table shows that the hydrodynamic radius (rH) results of the wine colloids and wine macromolecules range from 2.7 to 3.5 nm and from 10.4 to 13.1 nm, respectively. The weight-average molar mass (MW) ranges from 22 to 47 × 103 g/mol for the wine colloids, WC, and from 157 to 303 × 103 g/mol for the wine macromolecules, WM. The MW values show a broad range because the molar mass ratio between two populations differs from approximately 3 times (CAB) to 13 times (MAL-1 and MALCAB-2).

The samples were analyzed gravimetrically after dialysis to evaluate the total amount of colloidal and macromolecular material present in the wine. Table shows the results for the total retained solids after dialysis. The results appear in a range of 1.82 mg/mL for MALCAB-2 to 4.09 mg/mL for TANN. The retained solids refer to the total mass of wine colloids and the wine macromolecules obtained in the retained fraction. The solid material obtained after freeze drying is fluffy and bluish/reddish in its character.

The specific refractive index increment, dn/dc, was determined for the solids obtained in the retentate after dialysis. The dn/dc values found for these total solids ranged from 0.158 for MAL-1 to 0.204 for CAB. The protein content in the solids is shown to be rather low, and the values range from 0.085 mg/mL for MALCAB-2 to 0.194 mg/mL for MAL-1.

From (dn/dc)TS, (dn/dc)WC is obtained using eq . Table shows the (dn/dc)WC values for the wine colloids in population 1 using eq . The values range between 0.187 mL/g for MAL-1, to 0.246 ml/g for MAL-2. (The MALCAB-2 value is considered as an outlier due to a high error bar in its calculation in addition to the low concentration of the peak). The 0.187 mL/g value found is close to a typical dn/dc for proteins. Indeed, the protein analysis shows the highest value found, 0.194 mg/mL, for MAL-1, agreeing with the actual concentration of the wine colloids of that population. For most of the wine samples, the dn/dc values are higher than 0.18 mL/g, the value for protein in the carrier liquid, indicating the presence of polyphenols in the colloidal fraction rather than just proteins.

Table 5

Values of Wine Colloids of dn/dc, Concentration, ρ̑ app Obtained from the AsFlFFF–MALS–dRI Analysis, and the Calculated Specific Absorptivity for the Wine Colloids (WC) and Wine Macromolecules (WM)

	dn/dca (mL/g)	absolute concentrationb (mg/mL)		total concentrationc (mg/mL)	apparent density,d ρ̑ app (kg/m3)		specific absorptivity,e ε (mL mg–1 cm–1)
code	WC	WCf	WMg	WC + WM	WC	WM	WC	WM
CAB	0.212	0.88	0.26	1.14	885	34.0	7.15	0.69
MAL-1	0.187	0.18	0.53	0.71	469	89.6	1.90	0.41
MAL-2	0.246	0.23	0.53	0.76	539	62.6	5.66	0.50
MALCAB-1	0.219	0.40	0.92	1.32	1246	103	7.41	0.40
MALCAB-2	0.251	0.04	0.46	0.50	397	64.3	0.78	0.05
TANN	0.245	0.44	0.91	1.34	551	77.8	6.75	0.47

dn/dc refers to the wine colloid refractive index calculated from eq .

Absolute concentration of the material in the populations according to eq .

Total concentration is the sum of peak 1 and peak 2 in milligrams per milliliter.

ρ̑ app is the apparent density. Calculations are based on rH and MW.

ε is the specific absorptivity based on eq .

dn/dc used for wine colloids according to eq .

dn/dc = 0.14 for polysaccharides in 13% ethanol according to the literature.[30]

Different concentrations of wine colloids and wine macromolecules are estimated for the samples from dRI area normalizing the area to dn/dc; an average result of approximately 1 mg/mL was found as the total concentration. The CAB sample is the only sample where the presence of the wine colloids shows a considerably higher concentration than the wine macromolecular fraction. MAL, MALCAB, and TANN samples show the wine macromolecules as the major matter of these wines.

To further describe the nature of the matter of the populations, the apparent densities have been estimated by taking the mass and comparing it to the volume estimated from the size assuming spherical geometry and using the hydrodynamic radius. Thus, the apparent density becomes a measure of the concentration of the molecular matter in the volume that the object occupies in the liquid.

Table shows that the apparent densities for the WC fraction vary between 400 and 1200 kg/m3 and for WM fraction vary between 30 and 100 kg/m3. Hence, WC is shown to be on average approximately 10 times denser than WM. Thus, these results show that the wine colloids are smaller and denser compared to the higher molar mass wine macromolecules.

The densities of the wine colloids are between 400 and 1200 kg/m3, with most of them higher than what is observed for typical proteins,[29] 400 kg/m3. The densities of the wine macromolecules are lower than what has been observed previously for gum arabic,[36] about 100–500 kg/m3, but similar to what has been observed for arabinogalactan,[37] 70–150 kg/m3. Thus, the apparent density of the wine colloids being smaller than the tannin material density (about 1500 kg/m3) may act as an indication that the wine colloids are not fully compact particles and possibly somewhat fractal, while the wine macromolecules have an apparent density within the range of different polysaccharides.

The specific absorptivity of the fractions is shown in Table . The results show that the wine colloids are characterized by high specific absorptivity (ε) in most of the samples, with values that range between approximately 1 and 7. However, the specific absorptivity (ε) is low in both MAL-1 and MALCAB-2 samples, 1.90 and 0.78, respectively. In these cases, the UV absorbance, the apparent density, and the peak concentration of P1 are also low (see Tables and 7).

Table 7

Total Phenols, Non-dialyzable Fraction, Soluble Fraction, and the Absorbance from the Dialysis Experiment

code	TPHa wine (mg/mL)	TPH retantateb (mg/mL)	dialysate fractionc (%)	retentate fractiond (%)	total abse wine	total abs dialysatef	dialysate fractiong (%)	retantate fractionh (%)
CAB	2.44	0.83	66	34	39.6	27.74	70	30
MAL-1	1.92	0.25	87.2	12.8	28.6	24.16	84.5	15.5
MAL-2	2.72	0.64	76.6	23.4	36.8	24.93	67.7	32.3
MALCAB-1	2.28	0.91	60.2	39.8	42.8	24.76	57.8	42.2
MALCAB-2	1.74	0.12	93	7	34.4	23.97	69.7	30.3
TANN	2.82	1.19	57.8	42.2	48.1	30.87	64.2	35.8

Expressed as milligrams of gallic acid equivalents (GAE) per milliliter of original wine.

Refers to the portion of the total phenols that was not dialyzable and remains present inside the membranes.

Dialysate fraction refers to the percentage of total phenols that was released during the dialysis experiment.

Retentate fraction refers to the percentage of phenols that were retained after dialysis.

Total absorbance of the wine samples at 280 nm.

Total absorbance of the dialysate accumulated after 4 days of dialysis.

Fraction of the total absorbance present in the dialysate in percentage.

Retained fraction of total absorbance in percentage.

Furthermore, if we take together the results of the total protein content and the specific absorptivity listed in Tables and 5, respectively, it is possible to note that the wine colloid population in MAL-1 and MALCAB-2 is rather dominated by proteins.

Table shows various aspects of the total recovery of the analysis. The analyzable fraction of the AsFlFFF experiments varies between 85 and 94%. The main part of the non-included material is larger than the separable range of the analyses under the present flow conditions (diameters between 2.7 and 71 nm). The samples were filtered through a 0.45 μm filter before the analysis, and thus, the non-analyzable fraction under the present flow conditions should be limited to between 71 and 450 nm.

Table 6

Recovery Values Using Different Methods

code	total area analyzeda fraction (%)	UV recoveryb (%)	total mass recoveryc (%)
CAB	93.7	15.8	38.2
MAL-1	87.7	1.9	35.5
MAL-2	84.7	4.2	33.0
MALCAB-1	87.4	9.6	40.0
MALCAB-2	91.5	0.2	29.9
TANN	88.8	7.2	37.0

Refers to the ratio of the sum of the dRI area fraction of the peaks with the total dRI area.

UV recovery calculated the UV area peak at 280 nm after fractionation with the area of peaks without cross-flow.

Total recovery at the AF4 relative to the total mass of the retained solids.

The results of the UV recovery show values lower than 20% (section 1 of the Supporting Information). However, there is a low-molecular UV-absorbing material in addition to the UV-active material connected to the wine colloid and macromolecular fractions, as confirmed with the dialysis experiment.

Total mass recovery ranged from 30 to 40% as a result of the comparison of the AsFlFFF–dRI to a quantitative dialysis experiment. There are no indications that substantial material sticks to the membrane because no discoloration and no changes in the elution properties could be observed after 10 repeated elutions using the same membrane. However, a recovery of 30–50% can be considered representative of the sample.

Total Phenols and Absorbance from the Soluble and Retentate Fractions from the Dialysis Experiment

Table shows the results of the total phenols present in the wine as well as the fraction of phenols that remained as a retentate after the dialysis process. The retentate can be interpreted as the phenolic fraction present in the aggregates of the wine colloid fraction. The results of the total phenols range between 1.74 mg/mL for MALCAB-2 and 2.82 mg/mL for TANN. These results are consistent with other results reported in South American red wines.[34] The values of total phenols of the retentate range between 0.246 mg/mL for MAL-1 and 1.19 mg/mL for TANN. The TANN sample showed the highest amount of total phenols and total phenols retained.

Using the relationship between the amount of total phenols and the fraction of the total phenols retained, the soluble amount and retained amount can be estimated. The soluble fraction shows values that ranged from 58% for TANN to 93% for MALCAB-2. This shows that an average of approximately 75% of the polyphenols is dialyzed and the polyphenols of the MALCAB-2 sample are the most soluble. This can be supported by the fact that, in the AsFlFFF fractograms for the latter, the UV–dRI area at peak 1 was practically negligible.

The results of the non-dialyzable or retained fraction are shown to be less than 30% on average for all samples.

In addition, the absorbance was also compared during the dialysis experiment. Table shows the results of the total absorbance of the wine prior to dialysis compared to the total absorbance released in 4 days. The results show a similar trend with an average of 70% soluble or released absorbance as opposed to 30% retained. In this case, the MAL-1 sample was shown to be the most soluble, 84.5%, and the MALCAB-1 sample was shown to be the most retained, 42%.

Qualitative Identification of the Wine Colloid Population Using HPLC–DAD and Sugar Monosaccharides Using HPAEC

Phenolic and Anthocyanin Identification

The wine colloid fraction of the CAB sample was collected during repeated injections using a fraction collector. The tannic polymer was digested to allow for analysis of the primary polyphenols building the structures.

The resulting chromatogram showed the presence of anthocyanins and phenolic compounds that were detected at a wavelength of 530 and 280 nm, respectively. However, several peaks were not possible to identify, despite attempts to compare the UV spectra and retention times to different external standards, as described in section .

The results in Figure confirmed the presence of three anthocyanins and two phenolic acids in the collected fractions. By comparison of the retention time and the UV–VIS spectra to the standards, the following identifications were possible: delphinidin at 6.9 min, cyanidin at 7.7 min, and malvidin at 8.4 min as anthocyanins. Phenolic compounds, such as p-coumaric acid at 7.4 min and ellagic acid at 8.7 min at 280 nm, were also found (additionally, the UV spectra of all of the unidentified peaks are provided in the section 2 of the Supporting Information).

Figure 2

HPLC chromatograms of population 1 of the CAB wine sample for the anthocyanins delphidin, cyanidin, and malvidin glucosides, with detection at 530 nm. At 280 nm, the phenolic compounds detected are p-coumaric acid and ellagic acid. The UV–Vis spectra of all detected compounds are also shown.

Generally, the polymerized compounds of Cabernet Sauvignon red wines are mainly composed of tannins. The anthocyanins, such as malvidin, delphinidin, and cyanidin, are linked to sugars through glycosidic bonds and are acylated with cinnamic acids, such as p-coumaryl structures.[13]

Specifically, the identification of malvidin-n-O-glucoside in Figure is due to the fact that it is one of the main anthocyanins present in red wines, where its form and its concentration depend upon the grape variety.[38] The low concentration found may be attributed to the fact that malvidin is more sensitive to the thermal degradation than other anthocyanins, such as cyanidin.[39] On the other hand, cyanidin-n-O-glycoside and delphinidin-n-O-glycoside are originally found in grape skins and seeds, respectively. In our fractions, they come from the hydrolysis of tannin while the free anthocyanins present in the wine dragged through the membrane. On the other hand, in acidic and heating media, proanthocyanidins and prodelphinidins can generate condensation products, such as cyanidin and delphinidin, respectively.[40]

With respect to the origin of the phenolic acid compounds found, p-coumaric acid is a hydroxycinnamic acid found in the pulp and mainly in the skin of the grapes, present as a tartaric acid ester.[40,41] However, phenolic acids, such as p-coumaric acid, are derived from the esterification of glucose in position 6 in Vitis vinifera L. grapes.[30,42] Phenolic acids, such as hydroxycinnamic and p-coumaric acids, are commonly part of the esterification of the polysaccharides in the cell walls.[43] With regard to the precedence of ellagic acid, this compound is a hydrolyzed product of the tannins, consisting of a carbohydrate core attached to ellagic acid that may either come from wooden barrels or from the addition of enological tannins.[40,44] Ellagic acid has been previously reported as an unstable compound that, together with some anthocyanins, such as delphinidin 3,5-diglucoside, affects color and contributes to the formation of insoluble sediments.[45]

The results confirmed the presence of polyphenols in the wine colloids after acid hydrolysis. These results show the potential portion of phenolic compounds that can be part of the aggregates in this fraction. However, for a more precise phenolic identification, it is suggested to control the degree of depolymerization of the aggregates (hydrolysis) as well as to implement other techniques that allow for a more detailed characterization of all of the phenolic compounds that may be present in the wine colloid fraction.

Monosaccharide Identification

The monosaccharide composition of the two populations of the CAB sample is shown in Table . The monosaccharides were identified by comparing the elution times to external standards (chromatograms in section 1 of the Supporting Information). In both populations, six peaks were identified in the following order: arabinose, rhamnose, galactose, glucose, xylose, and mannose; the results are expressed in terms of the ratio of the total sugar content detected in percentage form. The total peak content is expressed in percentage by comparison to the absolute concentration of the peak fraction, and the total value found is expressed in milligrams per milliliter.

Table 8

Total and Relative Concentration of the Monosaccharides after Hydrolysis

CAB population	arabinose (%)	rhamnose (%)	galactose (%)	glucose (%)	xylose (%)	mannose (%)	glucuronic acid (%)	total peaka content (%)	totalb (mg/mL)
WC	11	18	21	36	<1	14	NDc	5	0.05
WM	31	0.8	42	2	ND	20	ND	100	0.26

Percentage of the polysaccharide in the colloidal and macromolecular fraction by comparison to the absolute concentration of the wine colloids and wine macromolecules, respectively, in Table .

Total value found in the fraction by HPAEC in milligrams per milliliter.

ND = not detectable (<1%) relative percentage of the total sugar content.

The higher relative values of the glucose content, 36% in the wine colloids, are related to the hydrolysis of the polysaccharides as well as to a lesser extent the loss of the sugar units in the phenolic glycosides, particularly those in the flavonol 3-glucoside molecules. Glycosidic forms of sugars in flavonoids are at position 3 in the molecule. This linkage is characteristic of the grapes containing mainly monoglycosidic rather than diglycosidic forms. Previous research found that eight monoglycosides and three diglycosides have been characterized.[38] The most common sugar linked to the anthocyanidin is glucose, followed by rhamnose, xylose, galactose, and arabinose, which agrees well in the present study.[46]

The wine macromolecular fraction has a high content of galactose, followed by arabinose, suggesting a high content of arabinogalactan. Because the proportion of galactose and arabinose is 50 and 40%, respectively, this suggests that arabinogalactan[2] may be 80% of the macromolecular population of the CAB sample. Previous studies have shown evidence that the carbohydrate moiety of the polymeric material of the wine mainly consists of arabinogalactans and mannoproteins.[47,48]

Arabinogalactan proteins represent approximately 40% of the total soluble polysaccharides (300 mg/L).[49,16] Mannoproteins and arabinogalactans constitute the main polysaccharides of the wine polymeric material that have more selectivity to bind with the p-coumaroyl anthocyanin structures.[47] In addition, the presence of mannose may be linked to the wine mannoproteins found in the cell walls of the yeast, which are made up of 20% protein and 80% d-mannose.[10]

The results in Table do not show any presence of uronic acids, such as glucuronic and/or galacturonic acids, possibly as a result of the loss of these monosaccharides during the high pH hydrolysis step.

The percentage of polysaccharides present in wine colloids is 5%, which agrees with the high specific absorptivity values found and the statement that wine colloids mainly consist of polyphenols in this particular sample (CAB), while the percentage of polysaccharides found in the wine macromolecular fraction is 100% for the same wine sample.

Relevance of the Results toward the Properties of Wine

The study shows that a range of interesting properties of potential relevance to wine can be characterized, including molar mass, hydrodynamic radius, and UV-specific absorptivity. Furthermore, by measurement of the refractive index of the retained material in a dialysis experiment, it has been possible to obtain a relevant dn/dc to quantify the wine colloidal and wine macromolecular fractions. This also allows for quantitative estimation of the specific absorptivity and apparent particle density, although the estimation of the refractive index increment (dn/dc) involves the assumption that the wine macromolecular fraction in principle is similar to polysaccharides and subtraction that magnifies the error when (dn/dc) of the total dry matter is converted into a refractive index of the wine colloid matter (Table ). However, the relative error is large for small fractions of colloids with low specific absorptivity, low density, and low concentration, and the errors do, thereby, not change the interpretations.

To show that the AsFlFFF analyses provide a representative description of the properties of the colloidal and macromolecular material of the wine, two examples are shown: absorptivity of the wine colloid fraction in the AsFlFFF as a function of the retained total phenols in the dialysis experiment (Figure a) and total colloidal and macromolecular solids in the AsFlFFF experiment as a function of the total solids in the dialysis experiment (Figure b).

Figure 3

(a) Absorptivity of the wine colloid fraction from the AsFlFFF as a function of the retained total phenols from the dialysis experiment (Tables and 7) and (b) total colloidal and macromolecular solids in the AsFlFFF experiment as a function of the total solids in the dialysis experiment (Tables and 5).

The six wine samples (Table ) show the same precedence of origin and are rather similar in terms of character and design. However, the analysis shows that the character of the wine colloids differs between the samples (Figure ). In the figure, the UV-specific absorptivity is shown as a function of the concentration of the wine colloids in the samples. The trend is that the colloidal fraction of wines with a significant concentration of wine colloids also has higher UV-specific absorptivity of wine colloids, which is interpreted as a higher fraction of polymerized tannins. Other properties follow the same interpretation. With the apparent density increasing, (dn/dc)WC is increasing to a value expected for polyphenols in 13% ethanol solution (0.26 mL/g), and in the dialysis experiment, it was possible to observe that a larger fraction of the polyphenols was retained. Thus, the wines can be characterized in three different groups: high specific absorptivity, high apparent density, and high phenolic content with a higher colloidal fraction (CAB, MALCAB-1, and TANN), low specific absorptivity, low total phenols, low apparent density, and high soluble polyphenol fraction (MAL-1 and MALCAB-2), and intermediate values in its properties (MAL-2).

Figure 4

Properties of wine colloids described as specific absorptivity (UV at 280 nm) as a function of the concentration of wine colloids. The properties are in terms of apparent density (ρ̑ app), refractive index increment (dn/dc), and fraction of colloidal polyphenols (ϕ, retained fraction of polyphenols; Table ).

If the total phenol concentration is compared, it is possible to notice that the wines are designed with about a similar free polyphenol concentration, while the wines with a large fraction of polyphenols present in the wine colloid fraction have a higher concentration of total phenols (Figure ). The tannic character of these wines is rather comparable, and thus, it can possibly be expected that the winemakers have chosen processes and combinations of raw materials compensating for or using the formation of wine colloids to achieve a balanced tannic character.

Figure 5

Total phenol and soluble phenols (dialysate) as a function of the concentration of wine colloids.

The wine macromolecular fraction is less variable compared to the wine colloid fraction (Figure ). It can be noticed that the specific absorptivity follows the molar mass. High molar mass is correlating with low specific absorptivity, suggesting that the largest macromolecules are more or less pure arabinogalactans.

Figure 6

Molar mass of the wine macromolecular fraction as a function of the absorptivity.

The concentrations of the wine macromolecules are low, 0.09–0.03%; thus, despite the low apparent density, the impact on the wine viscosity will be very small (between 1 and 3%, assuming validity of the Stokes–Einstein equation).

The findings of this study suggest a methodology using AsFlFFF in conjunction with multiple detectors and the developed analysis as a tool for separation and characterization of fundamental properties of wine colloids and wine macromolecules of red wines. However, the results also shows that dn/dc of the wine colloids varies between the wines and, thereby, needs to be determined in a separate experiment.

In this set of red wine samples, a significant variation in the properties of the wine colloid fraction as well as in the wine macromolecular fraction can be observed. Most likely, this is consequence of different grape varieties and different processes used by the winemakers.