Concentration and Puri fi cation of Yacon ( Smallanthus sonchifolius ) Root Fructooligosaccharides Using Membrane Technology

Fructooligosacc harides (FOS) are storage carbohydrates found in some fruits a nd vegetables that can be considered prebiotics because they reach the colon intact and are selectively fermented by probiotic bacteria like Lactobacillus spp. and Bifi dobacterium spp. (1–3). They are linear polydisperse oligomers consisting mainly of β-(2→1) fructosyl-fructose linkages, sometimes containing a starting α-d-glucose moiety, that resist hydrolysis by human small intestinal digestive enzymes, which are specifi c for α-glycosidic bonds. They have thus been classifi ed as nondigestible oligosaccharides (4) and are best characterised by their degree of polymerization, between 3 and 10 (1).


Introduction
Fructooligosacc harides (FOS) are storage carbohydrates found in some fruits a nd vegetables that can be considered prebiotics because they reach the colon intact and are selectively fermented by probiotic bacteria like Lactobacillus spp.and Bifi dobacterium spp.(1)(2)(3).They are linear polydisperse oligomers consisting mainly of β-(2→1) fructosyl-fructose linkages, sometimes containing a starting α-d-glucose moiety, that resist hydrolysis by human small intestinal digestive enzymes, which are specifi c for α-glycosidic bonds.They have thus been classifi ed as non-digestible oligosaccharides (4) and are best characterised by their degree of polymerization, between 3 and 10 (1).FOS, also known as oligofructose, are considered food ingredients and can be used in many food applications as sugar substitutes (5).Besides contributing to a well-balanced diet by increasing the fi bre content and the diversity of the fi bre sources, FOS specifi c fermentative characteristics are responsible for numerous health benefi ts.There are studies relating their consumption to a more balanced composition of intestinal microbiota (6)(7)(8)(9)(10), improved mineral absorption, especially in post-menopausal women (11) and female teenagers (12)(13)(14), and endocrine activities (15).There is also evidence supporting improvement of systemic functions, as immune functions (16,17) and lipid homeostasis (18,19), as well as the ability to reduce the risk of various diseases (4).
Oligofructose is industrially produced by partial enzymatic hydrolysis (using an endoinulinase) of inulin extracted from chicory roots or by synthesis from sucrose using fructosyltransferase (5).Studies related to new sources of FOS and production processes could be useful for allowing their application in a wider range of foods.An interesting possibility is their extraction from yacon (Smallanthus sonchifolius) roots, a perennial plant originating from the Andean region, whence it has spread to New Zealand, Japan and Brazil, which diff er from other roots by storing carbohydrates in the form of FOS, instead of starch (20).
Yacon roots consist mostly of water, which usually exceeds 80 % of fresh mass, and carbohydrates, rich in FOS, especially 1-kestose (GF2), nystose (GF3) and 1-β-d--fructofuranosylnystose (GF4).There are also 15-40 % of simple sugars, such as sucrose, fructose and glucose.Other nutrients are reported to be at low concentrations, except for potassium (20,21).It has been reported, however, that the FOS content in yacon tuberous roots signifi cantly decreases during postharvest storage, even at low temperatures (5 to 10 °C) (22,23).Considering that yacon rootstocks are highly perishable, the development of alternative techniques for extracting, concentrating and purifying yacon FOS would allow their increased consumption by addition to more frequently consumed products, such as yoghurts, cereals and drinks.It would also be possible to reach levels at which a prebiotic eff ect may occur (5-8 grams per day) (1).
A high selective separation technology is a prerequisite for purity, especially when processing complex plant materials.Membrane technology, currently encompassing mainly ultrafi ltration (UF) and nanofi ltration (NF), is a potential feasible strategy for industrial manufacture of purifi ed oligosaccharides (24,25).Membrane processes have a number of advantages compared to e.g.chromatographic purifi cation techniques, as they include low energy requirements, hence off er sustainable processing, easy modifi cation of the critical operational variables such as pressure, temperature, feed fl ow rate and agitation, and relatively easy scale-up (26).
Many authors have obtained good results using ultra-and nanofi ltration for purifi cation of oligosaccharides (27)(28)(29)(30).Kamada et al. (31) evaluated the eff ectiveness of combined UF and NF for purifying and concentrating oligosaccharides from yacon rootstock, obtaining a 25-fold concentrated retentate with 98 % of FOS purity.In a similar work, but with chicory rootstock, the NF retentate, in which mono-and disaccharide content was reduced from 9.0 % in the initial solution to 2.6 %, was obtained as a 20--fold concentrated product, indicating that the combined membrane-processing system (UF and NF) is quite promising for FOS purifi cation (32).Kamada et al. (31,32), however, did not evaluate the use of diafi ltration for FOS purifi cation.They also used freeze-dried yacon rootstock, not fresh ones, in their experiments, which caused additional costs of the process.Kuhn et al. (33) used nanofi ltration for purifying FOS present in a mixture of sugars, containing also glucose, fructose and sucrose.Performing diafi ltration, they obtained a retentate containing around 80 % of FOS.
In this study, ultra-and nanofi ltration were combined to purify fresh yacon root extract, aiming to remove suspended solids and simple sugars as glucose and fructose and concentrate the yacon fructooligosaccharides.The use of NF in combination with discontinuous diafi ltration (DF) was also evaluated.

Yacon root extract
Yacon (Sma llanthus sonchifolius) roots were cultivated in São Paulo State, Brazil, and acquired from the Supply Centre of the Rio Grande do Sul State in the city of Porto Alegre, Brazil.The roots were cleaned and selected considering the absence of visible injury and infection, and refrigerated ((8±2) °C) until use for no more than three weeks.
The yacon extract used in the combined ultrafi ltration (UF) and nanofi ltration (NF) processes was prepared in two steps, adapting the methodology described by Toneli et al. (34).First, yacon roots were sliced and kept for 20 min in a 0.5 % sodium pyrosulphite (Sigma-Aldrich Brazil, São Paulo, Brazil) solution to minimise the enzymatic browning (35).The roots were ground in a food multiprocessor (Arno, São Paulo, Brazil) and the extracted juice was kept refrigerated.Aft er this, residual saccharides were extracted from ground roots by lixiviation, with the addition of water heated at 80 °C in the ratio of 2 kg of water per 1 kg of ground roots.This mixture was kept at an average temperature of (80±2) °C for 1 h and then fi ltered through a 270-mesh sieve to remove the triturated roots.The two obtained yacon fractions, the juice and the liquid extracted from ground roots, were then fi ltered and mixed, resulting in the so-called yacon extract.In order to minimise fouling during ultrafi ltration, two diff erent fi lters were used, with 1 and 22 μm nominal pore sizes (Parker Filtration, São Paulo, Brazil), which were available in the laboratory.

UF and NF equipment
Experiments were performed in a pilot plant, WGM--KOC H PROTOSEP IV (WGM Sistemas, São Paulo, Brazil), comprising the following equipment: (i) feed tank made of glass with a volume of 1.0 L, (ii) pn eumatic pump, diaphragm type, model Versa-Matic VM50 ( Versa--Matic, Mansfi eld, TX, USA), operating with compressed air through a system comprising a FLR kit (fi lter, air regulator and lubricator), (iii) a stainless steel grade 316 housing for the fl at sheet module, allowing the installation of membranes with an eff ective area of 0.00572 m 2 , (iv) stainless steel grade 316 manometer, with a scal e from 0 to 10.5 bar, and (v) valve for pressure contro l.

Experimental design
Membranes tested for the separation of saccharides were compressed at 4.5 bar and characterised by their permeate fl uxes of water and yacon extract for at least six of the following transmembrane pressures (Δp): 0.75, 1.00, 1.25, 1.50, 1.75, 2.00, 2.75, 3.00, 3.25 and 3.50 bar.Temperature was kept constant at (25±2) °C.The volumetric fl ux of the permeate (Jp) was calculated and expressed in L/ (m 2 •h).The Δp of the procedure was determined by the permeate-fl ux (Jp) vs. Δp curve.
Aft er defi ning the operating conditions, the yacon extract was processed in two stages, combining UF and NF crossfl ow processes, according to the methodology described by Kamada et al. (31).Prefi ltered yacon extract was subjected to batch UF processing, recirculating the retentate to the feed tank in order to remove large molecules such as proteins and fi bres and suspended substances, yielding a saccharide-rich permeate.The UF experiments were performed at a transmembrane pressure of 0.75 bar, recirculation fl ow of 20 L/(m 2 •h) and 25 °C, controlled by water-cooling the feed tank.These operating conditions were determined in previous experiments.UF process was conducted until the initial feed volume was reduced by half.
In the second stage, the clarifi ed saccharide-containing permeate obtained in the UF process was processed by NF, with full recycling of retentate, in order to concentrate the oligosaccharides in the retentate and reduce glucose, fructose, sucrose and salt concentration.This operation was also done together with discontinuous diafi ltration (DF), which consisted of adding incremental volumes of distilled water to the retentate aiming to optimise the withdrawal of salts, mono-and disaccharides from the permeate while removing this added volume, and increasing the degree of FOS purifi cation.Every time the permeate volume reached 50 mL, the same volume of water was added to the retentate, totalising 200 mL of added water.The NF experiments were performed at a transmembrane pressure of 4.5 bar and recirculation fl ow of approx.300 L/h, also determined in previous experiments.

Membrane fouling, cleaning protocol eff ectiveness and membrane retention
In all UF and NF experiments, measurements of water fl ux were carried out before and aft er fi ltration to quantify the fouling formation on the membrane.A chemical cleaning procedure was performed at the end of each experiment to restore the fl ux and retention characteristics of the membrane and prevent the growth of microorganisms in the system.It consisted of rinsing with distilled water, and alkaline, acid and chlorine cleaning, taking into account the membrane pH and temperature limits.
Fouling was expressed as the percentage diff erence in water permeate fl uxes of the membranes before and after yacon extract ultrafi ltration, according to the following equation (36): /1/ where J pi is water permeate fl ux of the virgin, unfouled membrane aft er compaction and before yacon extract UF and J pf is wat er permeate fl ux aft er yacon extraction by UF and rinsing with water to remove loosely bound foulants, at the same temperature and pressure conditions.The medium fouling value was considered at the diff erent Δp.
The eff ectiveness of the ultrafi ltration cleaning protocol was measured by calculating the water fl ux recovery according to the following equation ( 37): where J pc is the water permeate fl ux aft er the application of cleaning procedure at the same pressure and temperature as J pi .The medium fl ux recovery value was considered at the diff erent Δp.
For each saccharide, the observed retention (R obs ) was calculated from the following equation, based on the permeate and bulk saccharide concentration, determined from the sample analysis: /3/ where c p is the permeate concentration and c b is the bulk concentration of a given saccharide.The retentate was accumulated at the permeate tank until the initial feed volume was reduced by half.

Analytical methods
The moisture content of fresh yacon rootstock was measured by weighing up an d drying samples at 105 °C until constant mass, according to AOAC method 984.25 (38).The characteristics of yacon extract and of UF and NF feed, retentate and permeate samples were determined by measuring soluble solid content (in °Brix), electrical conductivity, pH and concentration of FOS, glucose and fructose by high-performance liquid chromatography (HPLC).Soluble solid content was determined using a refractometer at 25 °C, according to AOAC method 932.12 (39).The pH was measured with a Digimed DM20 pH meter (Digimed Instrumentação Analítica, São Paulo, Brazil), following AOAC method 981.12 (40).Electrical conductivity was determined with a Tecnopon mCA150 conductivity meter (MS Tecnopon Equipamentos Especias Ltda., Piracicaba, Brazil).HPLC analyses were performed adapting the method described by Zuleta and Sambucett i (41) and Fenner Scher et al. (42), using a PerkinElmer series 200 chromatograph equipped with a refractive index detector (HPLC--RI; PerkinElmer Life and Analytical Sciences, Shelton, CT, USA) and Milli-Q water as the mobile phase at 0.6 mL/ min, temperature of 80 °C and a Phenomenex Rezex RHM pi pf pi Fouling 100 Flux recovery 100 monosaccharide column, 330 mm×7.8 mm, with a total run time of 14 min (Phenomenex Inc., Torrance, CA, USA).The retention times were 6.766 min for oligofructose, 9.946 min for glucose and 10.742 min for fructose.All injections were carried out at least in duplicate.Saccharide content was used for mass balance determination throughout UF and NF process.

Saccharide purity and yield
The degree of purifi cation (purity) of each saccharide (i) at UF and NF feed, permeate and retentate was calculated according to the following equation ( 25): /4/ where c s i is the concentration of saccharide i in a given stream (feed, permeate or retentate) and c s total is total saccharide concentration in the same stream.The yield of each saccharide (in g per kg of yacon) was calculated using the following equation ( 25): where m s i is the saccharide i mass (in g) in a given stream (feed, permeate or retentate) and m yacon is yacon root mass (in kg) used for preparing the yacon extract for each UF or NF experiment.

Yacon extract characteristics
The yacon extract characteristics, resulting from the mixture of yacon juice and the liquid extracted from the ground roots by hot water diff usion, are described in Table 1.The pH of yacon juice and extract was close to the values found by Fenner Scher et al. (42), who observed the pH of (6.09±0.01)when evaluating fresh yacon roots from the same origin as the ones used in this experiment.Ribeiro (43) has also shown a similar value of 5.87.
The soluble solid content of the yacon juice was close to that reported by Manrique et al. (44), 8 to 12 °Brix for yacon roots, and a litt le lower than values found by Fenner Scher et al. (42), (9.9±0.01)°Brix, and by Hermann et al. (45), which varied from 9.9 to 12.6 °Brix.Soluble solids of yacon extract of 7.5 °Brix were lower due to water dilution with the liquid extracted from yacon ground roots.
The saccharide content of yacon extract, on a dry mass (dm) basis, was: (10.55±0.02)g per 100 g of FOS, (17.30±0.03)g per 100 g of glucose and (20.60±0.01)g per 100 g of fructose.The hot water lixiviated from yacon ground root saccharide added 1.72, 2.66 and 2.91 g per 100 g of FOS, glucose and fructose, respectively, to yacon juice, which represented an increase of 21. 23 8.0-46.0g per 100 g), glucose (2.04-17.9g per 100 g) and fructose levels (9.02-43.2g per 100 g) among ecotypes and cultivation years.A tendency of a decrease of FOS degree of polimerisation (DP) is related to the cultivation of yacon in regions situated more to the north (at 56° of north latitude no inulin content was observed previously (23)).It was also found that the content of monosaccharides in yacon tubers cultivated in a plastic greenhouse was 37 % higher as compared to the ones cultivated in the fi eld (23).These diff erences emphasise the diffi culty of comparing yacon saccharides among experiments, since there are many factors infl uencing their content, such as year and method of cultivation and postharvest storage (47).Enzymes related to syn the sis and hydrolysis of oligofructose may also be involved, as already described by many authors (22,48,49).Increasing amounts of mono-and disaccharides during storage are also common in the inulin-and FOS-containing roots of chicory (Cichorium intybus L.) and Jerusalem artichoke (Helianthus tuberosus L.) (47).

Ultrafi ltration of yacon extract
The water permeate and yacon extract permeate fl uxes trough UF-10 and UF-30 membrans as functions of transmembrane pressure are shown in Fig. 1.The fl ux of water through both membranes increased linearly with transmembrane pressure, with R 2 =0.998, confi rming that membranes were properly compacted, and the yacon extract permeate fl ux was signifi cantly smaller than the water permeate fl ux under the same operating conditions.This indicates that the eff ect of concentration polarisation is signifi cant for yacon extract and this eff ect tends to increase at higher transmembrane pressures (TMP).Concentration polarisation occurs when TMP is large enough to transport the solutes with high molecular mass to the membrane surface, limiting the permeation rate by back diff usion of the solute from the membrane surface to the bulk of the feed (50).The accumulation of retained solutes at the upstream surface of the membrane leads to the reduction of eff ective pressure driving force due to the increase of fi ltration resistance and also osmotic pressure e ff ects (51).The concentration polarisation cannot be completely avoided, but its extent can be controlled by adjusting the fl uid fl ow characteristics, typically by providing low pressures and high shear rates at the membrane surface, in tangential fl ow fi ltration mode (51,52).It was noticed that, under pressures higher than 1.5 bar, yacon extract permeate fl ux became pressure-independent, indicating that limiting fl ux was reached.The limiting fl ux is the maximum fl ux that can be achieved at steady state in an operation.Yacon extract permeate fl ux curve was also used to estimate the critical fl ux, i.e. the maximum possible fl ux to minimise fouling and concentration polarisation tendency, which was determined by tracing a line from the origin and identifying the point where it became non-linear.Using this method, the chosen transmembrane operating pressure for the ultrafi ltration process was 0.75 bar, which provided a recirculation fl ow of approx.300 L/h.It was the highest TMP possible for operating the process under the critical fl ux and still providing acceptable permeate fl uxes.
When using UF-30 membrane, it took approx.200 min until initial feed volume was reduced by half, while with UF-10 membrane, this volume was reduced in 310 min.Only a slight permeate fl ux decrease was observed in both membranes during the process: when using UF-10 membrane, it decreased from 15.3 to 13.8 L/(m²•h) and when using UF-30, from 27.6 to 24.3 L/(m²•h).Kamada et al. (31) observed larger yacon extract permeate fl ux decrease, from 42 to 12 L/(m²•h), during UF with a 20-kDa NMMCO membrane.This may be explained by the higher concentration factor: while the initial volume was reduced by half in this experiment, Kamada et al. (31) reduced the fi nal retentate mass ratio to 3.3 % of the initial feed.Figs. 2 and 3 show water permeate fl uxes before (J pi ) and aft er ultrafi ltration (J pf ) with UF-10 and UF-30 membranes of yacon extract pretreated by fi ltration through a 22-μm fi lter.
These water fl uxes were used for estimating membrane fouling, a common problem of all types of membrane separation processes that arises from a deposit formation on the external surface of the membrane and/or from adsorption to and within the membrane pores, causing blocking or reduction in eff ective pore diameter (51).Fouling tendency can be reduced by working below criti-cal values of fl ux and pressure, balancing the hydrodynamic force, which drives solutes towards the pores, and the electrostatic forces opposing this motion (53).Figs. 2 and 3 also show water permeate fl ux aft er cleaning (J pc ), used for evaluating the cleaning protocol eff ectiveness.
The optimum prefi lter distributes the particles/molecules evenly between the two fi lters (prefi lter and membrane) so they both reach their maximum particle load and the process becomes more effi cient.For evaluating the two selected prefi lters, membrane fouling formation was calculated for UF-10 and UF-30 membranes, and it was observed that fouling was signifi cantly lower when 1-μm prefi lter was used, compared to the 22-μm prefi lter.When using UF-10 membrane, fouling formation was 68.67 % with 22-μm prefi lter and 26.95 % with 1-μm prefi lter; when using UF-30 membrane, values of fouling formation for 22-and 1-μm prefi lter were 52.66 and 16.16 %, respectively.This indicates that the 1-μm prefi lter was more effi cient in removing suspended particles that caused   Experiments were done in duplicate.J pi =water permeate fl ux of virgin, unfouled membrane aft er compaction; J pf =water permeate fl ux aft er yacon extract ultrafi ltration; J pc =water permeate fl ux aft er cleaning procedure Fig. 3. Water permeate fl uxes (J p ) vs. transmembrane pressure before and aft er yacon extract ultrafi ltration and aft er cleaning procedure.Membrane UF-30, 22-μm prefi lter, temperature=25 °C.Experiments were done in duplicate.J pi =water permeate fl ux of virgin, unfouled membrane aft er compaction; J pf =water permeate fl ux aft er yacon extract ultrafi ltration; J pc =water permeate fl ux aft er cleaning procedure; J pr =water permeate fl ux aft er membrane cleaning and recompaction rapid clogging of membrane pores with yacon extract, which suggests that a feed preclarifi cation is needed.A similar result was obtained by Saha et al. (36), comparing the fl ux profi les of sugarcane juice and a polysaccharide solution.Both fl uxes decreased exponentially, but the multicomponent sugarcane juice feed containing diff erent sizes of macromolecules caused severe fouling (60 %), while with polysaccharides alone, no visible fouling was observed during the experiment.
The cleaning protocol eff ectiveness was not the same for both membranes.Water fl ux recovery for UF-10 membrane aft er cleaning was partial (82.42 % with 22-μm and 81.72 % with 1-μm prefi lter), but total for UF-30 membrane (133.45and 139.81 % with 22-and 1-μm prefi lter, respectively), which shows that water permeate fl ux aft er cleaning increased compared to the fl ux of the virgin membrane aft er compaction.This increased fl ux may have been caused by UF-30 membrane decompaction, because the fl ux returned to the original values aft er a new compaction process (J pr ).Fu et al. (54) noticed that two NF membranes with diff erent properties required diff erent cleaning processes even with the same feed.Song et al. (55) examined membrane cleaning and reported that the tested cleaning agents could not achieve complete fl ux recovery because some residual foulants were strongly embedded in the concavities of membrane surface.However, Al-Amoudi et al. (53), from the results of the permeability of a fouled NF membrane before and aft er cleaning, showed that the cleaning process restored the declined fl ux close to its original value.Al-Amoudi and Lovitt (56) and Liikanen et al. (37) noticed that cleaning, especially alkaline, oft en increased the fl ux of the virgin membrane.
Table 2 shows the concentration of fructooligosaccharides, glucose and fructose in the feed, permeate and retentate aft er ultrafi ltration of yacon extract pretreated using 1-μm fi lter.The observed retention of each saccharide is also presented.Small diff erences in the initial composition of saccharides were observed between samples due to diff erent dilutions with residual water present in the system.UF was used to clarify yacon extract by removing large molecules, like proteins and other suspended solids, and allowing saccharides of low molecular mass to pass through the membrane pores and to be collected in the permeate.According to Lachman et al. (21), yacon rootstocks comprise (in %), in general: moisture 83.1, saccharides 13.8, ash 1.1, protein 1.0, fi bre 0.9 and lipids 0.1.Both membrane permeates were transparent and colourless, indicating that suspended solids of high molecular mass were probably removed.The observed retention of FOS at the UF-30 membrane (14.77%) was smaller than at the UF-10 membrane (22.35%), as expected due to its higher NMMCO.The same was observed for glucose and fructose.
Ultrafi ltration of yacon extract pretreated using 22--μm fi lter showed signifi cantly higher retentions of saccharides compared to the feed pretreated with 1-μm fi lter, especially with UF-30 membrane.With UF-10 membrane, FOS, glucose and fructose retentions were 27.00, 13.95 and 17.70 % respectively when using 22-μm fi lter, and 22.35, 5.36 and 6.10 % respectively when using 1-μm fi lter.With UF-30 membrane, saccharide retentions with 22-μm prefi lter were (in %): 47.80 of FOS, 18.90 of glucose and 25.14 of fructose; with 1-μm prefi lter these retentions were 14.77 of FOS, 0.42 of glucose and 1.65 of fructose.The increase in the observed retentions may have been caused by an incomplete removal of suspended solids by the 22-μm fi lter, leading to increased concentration polarisation layer and fouling.The increase in retention was more signifi cant for the UF-30 membrane, probably because of the rapid plugging of the larger size pores of this membrane, leading to high fouling (36).These results led to the use of 1-μm prefi ltered yacon extract in further experiments.
The pH, electrical conductivity and soluble solids were measured at ultrafi ltration feed, permeate and retentate.A slight decrease of pH was observed throughout the process with both membranes, UF-10 (6.11 in feed, 5.95 in permeate and 5.91 in retentate) and UF-30 (6.25, 6.35 and 6.17 in feed, permeate and retentate, respectively).The measurement of electrical conductivity showed a slight increase in the retentate (3.27 mS/cm at 25 °C when using UF-10 and 2.46 mS/cm at 25 °C with UF-30 membrane), indicating a higher solid content compared to feed (2.93 and 2.37 mS/cm at 25 °C with UF-10 and UF-30 membranes, respectively).Soluble solid content was 5.75 and 5.53 °Brix in UF-10 retentate and feed, respectively, and 5.08 and 5.03 °Brix in UF-30 retentate and feed, respectively.Kamada et al. (31) observed an increase in soluble solid content from 2.0 °Brix in the initial feed to 10.9 °Brix in the final retentate during yacon extract ultrafi ltration (20-kDa NMMCO membrane).A possible reason for this steep increase in soluble solid content is the higher concentration factor used by these authors, which reduced the fi nal retentate mass ratio to 3.3 % of the initial feed.Considering that high permeate fl uxes are desirable for improved process effi ciency, and low FOS retentions at this fi rst step of the process yield higher fi nal content of purifi ed FOS, for the nanofi ltration experiments only UF--30 permeate was used.

Nanofi ltration of the ultrafi ltration permeate
No irreversible fouling occurred during the experiments since there were no diff erences in the water fl ux measurements before and aft er each fi ltration run and after cleaning.The same result was found by Goulas et al.  (24) during the nanofi ltration of a commercial galactooligosaccharide mixture.
Table 3 shows fructooligosaccharide, glucose and fructose concentrations in nanofi ltration feed, permeate and retentate, obtained with and without diafi ltration, and their respective observed retentions.Some diff erences in the initial composition of saccharides were found between samples due to diff erent dilutions with residual water present in the system.Nanofi ltration was used to remove salts and mono-and disaccharides, like glucose, fructose and sucrose, concentrating yacon FOS in the retentate.It can be noticed that diafi ltration did not infl uence largely the FOS retention (it increased from 68.78 % without to 70.48 % with diafi ltration), but it decreased the retention of glucose and fructose from 40.63 to 31.61 % and from 25.64 to 18.69 %, respectively.The lower retention of monosaccharides with diafi ltration was highly desirable, allowing a greater purifi cation of FOS at the retentate.
The observed retention diff erences of simple sugars and FOS were not as high as desired, especially when using NF without diafi ltration, which may partially be explained by the small degree of polymerisation (DP) from yacon oligosaccharides.According to a study by Pedreschi et al. (57), evaluating the selective consumption of yacon saccharides by Bifi dobacterium spp.and Lactobacillus spp.probiotic strains, yacon FOS are composed of 27 % of GF2-type fructans, 54 % of GF3 and 19 % of GFn≥4 molecules; i.e. 81 % of yacon FOS are short oligomers with only two or three fructosyl units polymerised.Kuhn et al. (33) obtained similar saccharide retentions: 64, 28 and 31 % of FOS, glucose and fructose, respectively, during NF of a mixture of sugars containing FOS, glucose, fructose and sucrose with a 1000-Da NMMCO membrane, under 18 bar pressure, at room temperature and without diafi ltration.The same authors, with a 300--Da NMMCO membrane, obtained retentions of 66, 18 and 15 % of FOS, glucose and fructose respectively.Kamada et al. (31) calculated retentions of 14.0 % of mono-saccharides, 46.2 % of disaccharides, 80.9 % of trisaccharides and 91.5 to 99.9 % of saccharides of DP=4 and higher, with a 1000-Da NMMCO membrane at 50 bar during nanofi ltration of a saccharide mixture from yacon roots previously clarifi ed by UF (20-kDa NMMCO membrane).Using a 500-Da NMMCO membrane, they achieved respective retentions of 64.9, 81.6, 97.3 and 99.2 to 99.9 %.Moreno-Vilet et al. (58) studied nanofi ltration for separating inulin-type fructans from model solutions containing low molecular mass sugars (sucrose, glucose and fructose) using a pilot cross-fl ow unit.They found inulin retention values over 90 % with a 600-Da NMMCO membrane at 14 bar.Goulas et al. (24) observed retentions of 71, 45 and 11 % of raffi nose, sucrose and fructose, respectively, during nanofi ltration with a 1000-Da NMMCO membrane at 6.9 bar.
It is known that the fractionation of saccharides and oligosaccharides by nanofi ltration depends on several factors: membrane pore size distribution, fi ltration pressure and temperature, solute concentration and pH (28,29).Goulas et al. (24) reported increasing retentions of saccharides, especially monosaccharides, as pressure was increased from 6.9 to 27.6 bar, due to increased solvent fl ux and membrane compaction.The same authors also observed that an increase in temperature from 25 to 60 °C decreased retentions due to reduced viscosity and increased diff usion.The size of a monosaccharide is equal to or smaller than the cut-off sizes of the NF membranes.The calculated diameters of the monosaccharide molecules are approx.0.6-0.8nm and the reported measured pore diameters of the common commercial NF membranes are from 0.6 to 2.0 nm, including the mean pore diameter of approx.0.8-0.9nm.Thus, the comparatively small monosaccharides are the most aff ected when the total permeate fl ux changes due to the changes in pressure or temperature (29).This means that higher pressures may promote an undesirable increase in monosaccharide retention, while higher temperatures apparently decrease retentions, which may be interesting for glucose and fructose, but not for FOS.
The infl uence of pH on the nanofi ltration of saccharides was studied by Himstedt et al. (59), who evaluated the eff ect of pH on the separation of monosaccharides, disaccharides and their mixtures using membranes whose surface was modifi ed by graft ing poly(acrylic acid) (PAA) nanochains applying a UV-initiated free radical polymerisation method.They observed that pH dramatically affected membrane fl ux, rejection and selectivity, probably due to diff erent interactions between the sugar molecules and the neutral PAA chain or its negative conjugate; other possible explanation is conformational changes of the graft ed surface layer due to protonation and deprotonation of the acid groups on the polymer chains.Simulations conducted by the same authors to investigate the specifi c interactions between glucose/sucrose and the neutral/negatively charged PAA chains showed that both sugar size and PAA charge aff ect signifi cantly the sugar--polymer interactions.
Another factor infl uencing nanofi ltration is membrane material and pore size distribution.Goulas et al. (24) and Kuhn et al. (33) observed that membranes composed of polyethersulfone (less hydrophilic) appear to ex- ert bett er separation of saccharides than cellulose (more hydrophilic) membranes, as the one used in this experiment.Also, according to these authors, the small molecular size diff erence among FOS, glucose and fructose requires the use of membranes with a more uniform pore size distribution.
Van der Bruggen et al. (60) affi rm that nanofi ltration membranes for uncharged solutes are characterised by a sigmoidal rejection curve as a function of molar mass, which results in an insuffi cient separation of diff erent compounds on the basis of molecular size, as it is very diffi cult to retain one component completely and at the same time allow a second component, slightly diff erent in size or charge, to pass completely.The same authors suggest that the use of integrated continuous countercurrent recycle membrane cascades, in analogy with (conventional) separations based on thermodynamic equilibrium, may allow bett er separations between individual compounds, or fractionation of a mixture.
A slight decrease in pH could also be observed throughout both nanofi ltration processes: 7.09 in feed, 6.75 in permeate and 6.70 in retentate (NF without diafi ltration) and 6.00, 5.74 and 5.50 in feed, permeate and retentate, respectively (NF with diafi ltration).Electrical conductivity showed a slight increase in the retentate obtained with NF without diafi ltration (from 0.78 mS/cm at 25 °C in feed to 0.81 mS/cm at 25 °C in retentate) and a decrease in the retentate obtained with the diafi ltration NF (from 1.52 to 1.45 mS/cm at 25 °C in feed and retentate, respectively) because of the addition of distilled water.There was no signifi cant diff erence between soluble solid content in feed and retentate fractions in both experiments.

Mass balance
The mass balance throughout the UF and NF processes, with and without diafi ltration, is shown in Table 4. Aft er UF with diafi ltration, 63.75 % of the saccharides from the initial feed were recovered in the total permeate.Yield of saccharides in the NF fi nal retentate obtained with combined UF and NF processes associated with diafi ltration of total saccharides was calculated to be 50.89% and of FOS 51.85 %.Increased saccharide yields in the diafi ltration experiments occurred due to a higher overall initial concentration of saccharides.As previously explained, these diff erences in initial composition of saccharides between samples were caused by diff erent dilutions with residual water present in the system.Kamada et al. (31), using combined UF and NF processes, were able to recover 82.2 and 56.4 % of the initial saccharides in the UF permeate (20-kDa NMMCO membrane) and NF retentate (1000-Da NMMCO membrane), respectively.
Analysing the nanofi ltration retentate purity data, it can be noticed that the enrichment in FOS vs. simple sugars was not as high as desirable.A possible reason is that FOS, glucose and fructose molecules have similar sizes and the membrane has a pore size distribution which does not allow eff ective fractionation of these molecules.The diafi ltration process, nevertheless, allowed a slight FOS purifi cation with a purity increase from 17.44 to 19.75 %.These results may suggest that the retentions of The obtained fi nal nanofi ltration retentate, even not consisting of pure FOS, may still be applied as a functional ingredient in non-dietetic foods (because of the presence of simple sugars).Gullón et al. (61) evaluated the prebiotic potential of xylooligosaccharide (XOS) concentrates purifi ed by membrane technology from malting industry solid waste, assessing the eff ect of diff erent purity and/or molecular mass distribution.They concluded that purity of XOS concentrates did not play a signifi cant role in fermentation, whereas the sample with shorter average degree of polymerisation presented faster fermentation kinetics and led to the highest concentration of lactic acid.

Conclusions
Aft er UF, a large fraction of the saccharides in the initial feed was recovered in the total permeate.Saccharide yields in the fi nal NF retentate aft er the combined UF and NF processes were acceptable, but FOS purity was not as high as desired.Diafi ltration did not infl uence largely FOS retention in NF, but it decreased glucose and fructose retentions, which was the aim, allowing greater purifi cation of FOS at the retentate.The combined UF and NF is promising for concentrating yacon saccharides, but it did not totally purify FOS, suggesting that more diafi ltration steps are required to improve the preferential removal of mono-and disaccharides while retaining and concentrating saccharides of higher DP in the retentate.

Table 3 .
Concentration of fructooligosaccharides (FOS), glucose and fructose in the feed, permeate and retentate, and observed retention of saccharides (R obs ), in the yacon extract obtained with nanofi ltration using NF-1 membrane with and without diafi ltration

Table 4 .
Mass balance throughout the ultrafi ltration and nanofi ltration processes, wi th and without diafi ltration, in feed, permeate and retentate fractions Yield is expressed in g per kg of yacon roots glucose and fructose are FOS concentration-dependent, but these have not been experimentally evaluated.Since glucose and fructose retentions (31.6 and 18.69 %) at the diafi ltration process were smaller than FOS retention (70.48 %), as discussed previously, we suggest that further distilled water additions with increased permeate volume removal could be evaluated, as it may lead to improved glucose and fructose withdrawal and, consequently, higher FOS purity.