Modification of starch_ A review on the application of “green” solvents and controlled functionalization

Starch is a polysaccharide widely present in nature and characterized by a wide range of applications. This often implies the necessity for various novel properties with respect to those of native starch, mainly achievable via chemical modification. During the last decades, products with new or enhanced properties were prepared from starch because of the adoption of “green” solvents (ionic liquids and supercritical CO2) and several new techniques (regioselective derivatization, atom transfer radical polymerization, etc.) that are characterized by controlled modification. However, reviews on these works seem very rare. In this article, the application of ionic liquids and supercritical CO2 in the modification and processing of starch is summarized. The development of regioselective derivatization and controlled grafting of starch are also reviewed in the second part of this article.


Introduction
Starch, as the most abundant natural occurrence of polysaccharides that is just second to cellulose, widely exists in the roots, seeds, leaves of various plants, and some algae. Typically, starch granules are composed of two main fractions: amylose, a kind of linear polymer of glucose unit with α-(1→4) linkage and amylopectin, which is highly branched with lots of short chains that linked through α-(1→6) linkage to the linear parts of the macromolecule (Wu, Witt, & Gilbert, 2013). The difference in structure renders these two components differ significantly with each other in both physical and chemical properties. For example, amylopectin is easier to dissolve in hot water than amylose. Another significant difference is that only amylose could complex with iodine and make their solution appear blue-black color (Fig. 1).
Because of technological developments, the application field of starch is broadened progressively and different techniques have been developed for the modification of starch to overcome its shortcomings. These include poor processability and solubility in common organic solvents, retrogradation and syneresis, low shear stress resistance and thermal decomposition. Generally, there are four categories of methods for the modification of starch: physical, chemical, enzymatic and genetic engineering respectively (Ashogbon & Akintayo, 2014;Benavent-Gil & Rosell, 2017;Halley & Avérous, 2014;Lyytikainen et al., 2018;Masina et al., 2017). While the main purpose of genetic engineering is to modify the ratio of amylopectin to amylose and their structures for specific applications, physical modifications are employed to improve the granule size and solubility of starch in water. The applications of physical and biotechnological methods in the modification of starches have been kindly reviewed by Ashogbon et al. and Halley et al. respectively (Ashogbon & Akintayo, 2014;Halley & Avérous, 2014).
Compared with physical and genetic engineering modifications, chemical methods provide more options for the functionalization of starch and therefore broaden the application field significantly, especially with the help of new emerging chemical techniques. As is known, there are three hydroxyl groups adjacent to the carbon atoms at 2, 3 and 6 positions in each anhydroglucose unit (AGU), which enable the modification of the starch could be easily achieved through the chemical reaction with various functional groups. Traditional chemical modifications, including esterification and etherification, cationization, T oxidation as well as crosslinking, are usually realized through the reaction of starch with small molecules. These derivatives are widely used in the food industry and as additives for medicine, cosmetics, coatings, oil exploiting fluids and so on. The synthesis and applications of these kinds of derivatives have been reviewed in many papers published in recent years (Ashogbon & Akintayo, 2014;Chen et al., 2015;Masina et al., 2017). In the present work, we mainly focus on the newly emerging green solvent for modification and the progress of controlled modification of starch.

"Green" solvents for starch modification
Because of the intra-/inter-molecular hydrogen bonds between the hydroxyl groups along the backbone of the molecules, starch has a poor solubility in water and most common organic solvents. The modification of starch in aqueous solutions, in most cases, is carried out in a heterogeneous way, which, together with many other side reactions, limits the degree of substitution (DS) and the further application of these derivatives (Gao, Luo, & Luo, 2012). To get a higher DS, starch is dissolved and modified in some organic solvents with the help of some catalysts. Solvents and catalysts such as dimethyl sulfoxide (DMSO) and pyridine, however, are flammable, toxic and therefore hazardous towards both human and environment. With increasing attention being paid to health, safety and environment (HSE) protection, chemists are trying to find new solvents to displace traditional organic solvents for dissolving natural macromolecules and subsequent chemical reactions. In the past decades, the focus has been mainly paid to two kinds of "green" solvents, which are ionic liquids and supercritical carbon dioxide respectively, for the dissolution and modification of natural polysaccharides such as starch, cellulose, and chitin. The applications of these two solvents for starch are summarized below.

Ionic liquids
Ionic liquid (IL) usually is composed of a combination of organic cation and smaller organic or inorganic anion. With properties such as negligible volatility, non-flammability, thermal and chemical stability, tunable polarity and so on, IL attracted the attention of scientists who work on polymer science. Many reports on the successful application of various ionic liquids for different polymerization techniques, such as free radical and ionic polymerizations as well as controlled polymerizations (Chakraborty, Jähnichen, Komber, Basfar, & Voit, 2014;Vijayakrishna, Manojkumar, & Sivaramakrishna, 2015;, have been made and reviews on this field are also available (Kubisa, 2009;Mota-Morales et al., 2018). Specifically, the application of ILs in the preparation of drug/gene delivery materials with biopolymers like starch has also been reviewed (Chen, Xie, Li, & Chen, 2018). In this present work, the applications, as well as limitations of ionic liquids as solvent and/or catalysts in the modification of starch, will be reviewed.
Various derivatizations of starches with ILs as solvents have been studied in recent years. For example, cationic corn starch was successfully synthesized by Cl) as solvent (Wang & Xie, 2010). Glycidyltrimethylammonium chloride (GTAC) was used in their reaction for the derivatization of starch and a maximum DS of 0.99 was obtained.
[BMIM]Cl was also used as a solvent for the esterification and carboxymethylation of starches. In the experiment carried out by Xie et al., corn starch was esterified with succinic anhydride and acetic anhydride with pyridine as the catalyst in [BMIM]Cl (Xie, Shao, & Liu, 2010). Succinylated starch with DS values that varied from 0.03 to 0.93 and acetylated starch with DS values ranged from 0.37 to 2.35 were successfully synthesized respectively. In another report, the carboxymethylation of corn starch in [BMIM]Cl was investigated and a maximum DS of 0.76 was obtained (Xie, Zhang, & Liu, 2011 (Bertoft & Blennow, 2016;Vanier, El Halal, Dias, & da Rosa Zavareze, 2017). Despite all these studies, however, systematic experimental comparisons of the effectiveness and efficiency of starch derivatization in ILs and other organic solvents (DMSO, N,N-dimethylacetamide (DMAc)/LiCl) are still not available. Apart from the research mentioned above, the grafting of corn starch with L-Lactide by ring-opening graft polymerization (ROP) was also studied. In the report by Xu and Wang et al.,Cl) was used as the reaction medium for the grafting of starch and the grafting efficiency of poly(L-lactide) (PLLA) was calculated to reach 30 % according to the standard curve based on FT-IR method (Xu, Wang, & Liu, 2008). Recently, Wang et al. also reported the synthesis of starch-based macroinitiator in [AMIM]Cl for the atom transfer radical polymerization (ATRP) grafting of polystyrene (PS) and poly(methyl methacrylate) (PMMA) (Wang, Yang et al., 2015). Corn starch was esterified with 2-bromoisobutyryl bromide in [AMIM]Cl at room temperature without using any additional catalysts. Compared with reactions carried out under heterogeneous conditions, the graft density and ratio were significantly improved with [AMIM]Cl as the solvent. Another IL used for the grafting of corn starch with PS was 1-ethyl-3-methylimidazolium acetate ([EMIM]Ac) (Men, Du, Shen, Wang, & Liu, 2015). In this report, PS was successfully grafted onto starch via conventional free radical polymerization. However, different from the one grafted via ATRP, the ratio between styrene and starch has a significant influence on the grafting percentage (GP), which is too high ratio (e.g. 3:1) will reduce the GP due to high monomer concentration. This is the inherent drawback of conventional free radical polymerization compared with controlled radical polymerization like ATRP.
Besides the application as solvent, several ILs, mainly 1-N-alkyl-3methylimidazolium chlorides could also act as catalysts for the esterification of high-amylose maize starch (Lehmann & Volkert, 2011). In the research of Lehmann and Volkert, different carboxylic anhydrides were used as both reactants and solvents for the esterification of maize starch, while the amount of ILs was just kept to a low value. Imidazolium-based ILs with halogenides as counter ions were proved to be good catalysts for this reaction. They also found that the degradation of maize starch could be suppressed by the addition of pyridine or 1- Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 methylimidazole as the base and the average molar mass of starch acetate could be adjusted through varying the amount of pyridine added.
Concerning the dissolution and degradation of starch, it is of great importance to get a deep knowledge of the interactions between starch molecules and ILs for the successful modification of starch in ILs. Kärkkäinen et al. compared the dissolution and degradation of six different native starches (wheat, barley, potato, rice, corn, and waxy corn) in [BMIM]Cl under conventional oil bath heating and controlled microwave heating via liquid chromatography equipped with an evaporative light scattering detector (HPLC-ELSD). (Kärkkäinen, Lappalainen, Joensuu, & Lajunen, 2011). Results suggest that although a lower temperature (80°C) was used in microwave heating compared with that of oil bath (100°C), the dissolution and degradation process were both speeded up. Amylopectin was reported to degrade faster than amylose possibly due to its branched structure. A comparison between different starches showed that amylopectin with higher molecular weight degraded slower than that with smaller molecular weight. Similarly, Lappalainen et al. studied the dissolution and depolymerization of barley starch in ten ILs ( Fig. 2) with p-TsOH as catalyst and microwave as the heating method (80°C) (Lappalainen, Karkkainen, & Lajunen, 2013). It was found that barley starch could be depolymerized into 1000-2000 Da-sized, water-soluble oligomers by dialkylimidazolium halides that analyzed. Both cation and anion of the ILs were found to have significant influences on this dissolution process, to which a smaller cation or a more nucleophilic anion was proved to be more beneficial. This was related to the steric hindrance caused by the dialkylimidazolium cation and the ability of anion to break intra-and intermolecular hydrogen bonds of starch. In their research, [NH 3 CH 2 CH 2 OH][HCOO] and [EMIM]DMP were proved to be the most suitable solvent among those ten ILs for dissolving starch with a notable decreasing in depolymerization. According to the above-mentioned reports, although we cannot rule out the influence of heating on depolymerization of starch, the different degradation speeds of starch in various ILs do suggest the composition of ILs plays a vital role in the degradation process. One possible explanation is that the residual water in starch (and ILs) resulted in the formation of acid with the anion of ILs which catalyzed the depolymerization of starch (Kärkkäinen et al., 2011). As the molecular weight of starch significantly affects its application, further researches are needed to build up systematic guidance for us to choose the right ILs for the derivation of starch. Besides the influence of composition, the presence of water was also revealed to play a special role in dissolving starches. Liu and Budtova found that the dissolution of waxy corn starch in [EMIM]Ac could be accelerated with a proper amount of water (≤ 50 wt.%) and no gelatinization was observed in this process (Liu & Budtova, 2013). Moreover, a decrease of about 20°C in the beginning and final temperature of dissolution was noticed. A possible explanation for this is that the presence of water could induce the swell of starch granules which is favorable to the diffusion of [EMIM]Ac. Mateyawa and Xie et al. also studied the influence of water on the phase transition of maize starch (waxy and regular) in [EMIM]Ac (Mateyawa et al., 2013). They concluded that a higher water content (> 25 %, mol ratio) could strengthen the interactions between [EMIM]Ac and water which reduced the available [EMIM]Ac for starch molecules, thus made gelatinization dominates the phase transition process of starch. With the decreasing of water content, more [EMIM]Ac became available and resulted in a conversion from gelatinization to dissolution for the phase transition process. However, a proper ratio (7.2 mol/mol) of water to [EMIM]Ac is required for the effective disruption of starch granules at a relative lower temperature (e.g., 40-60°C), which is in accordance with the results from other reports (Liu & Budtova, 2013;Sciarini et al., 2015). In one recent report, the lowest temperature (28°C), as far as we know, for the dissolution of maize starch (24.4 wt.% amylose, 10 wt.% in total in solution) was achieved with 6.7 mol/mol water to [  Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 starch can be dissolved in this solvent within 1 h without apparently depolymerization. This provides a "green" way for the non-water sensitive processing of starch.
In addition to its use as solvents and catalysts for the chemical modification of starch, the applications of ILs as the plasticizer for the processing of starches have also been investigated. Sankri et al. verified the feasibility of plasticizing maize starch with [BMIM]Cl and a kind of electrically conductive thermoplastic starch (TPS) was obtained (Sankri et al., 2010). Compared with samples plasticized by glycerol, TPS modified by [BMIM]Cl showed a lower water uptake and higher elongation at break in its rubbery state. The results from a control experiment in their research showed that the structure of ILs and processing conditions probably play a significant role in the properties of the final product, but this still awaits deep studies. [BMIM]Cl was also used as a plasticizer for the processing of starch, zein and their blends with glycerol plasticized samples as control groups. According to the report by Leroy et al., [BMIM]Cl is a better plasticizer for maize starch than glycerol but has no obvious effects on the plasticization of zein . However, in the blends of starch and zein, [BMIM]Cl was proved to be a good compatibilizer for the two components. This makes it possible for the production of biocomposites from raw natural polymers and has a potential application in food packaging and other areas. The performance of [EMIM]Ac as the plasticizer for the production of starch-based film was also studied in recent researches (Xie et al., 2014;Zhang et al., 2016). It's reported that compared with glycerol, [EMIM]Ac can disrupt the crystalline structure of starch more effectively and increase the mobility of amorphous region via replacing starch-starch interactions with stronger starch-[EMIM]Ac interactions. Consequently, the obtained IL plasticized TPS film displayed lower tensile strength and stiffness but higher flexibility. Better anti-aging and bio-stable performance were also noticed in the experiments when compared with films that use glycerol as the plasticizer. In another report, a transparent conductive TPS film was obtained by compression molding (8 MPa) of starch (24.4 % amylose) at 55°C (or 65°C) with [EMIM]Ac as plasticizer (Zhang, Xie, Shamshina, Rogers, McNally, Wang et al., 2017). The conductivity (> 10 −3 S/cm) of the film can be enhanced with a lower molding temperature (55°C) and a higher relative humidity (RH, 75 %) during post-processing conditioning. This was related to the weak starch-[EMIM]Ac interaction at a lower temperature, and thus strong ion-pair dissociation. High RH, on the other hand, is preferable for the transference of ions in the film.
To be successfully applied to the production of biocompatible materials, there are still some problems to be overcome for ILs. For example, ILs are challenged in terms of "green" because of their poor performance in biodegradability, biocompatibility, and sustainability (Dai, van Spronsen, Witkamp, Verpoorte, & Choi, 2013). The relatively higher price of most ILs compared with that of conventional organic solvents also precludes their applications in most areas but high valueadded products. In recent years, a new generation of environmentfriendly solvents, consisting of hydrogen bond acceptor (HBA) and hydrogen bond donor(s) (HBD), was introduced as deep eutectic solvents (DES) (Smith, Abbott, & Ryder, 2014;Tavares, Rodríguez, & Macedo, 2013). Leroy et al. reported the plasticization of corn starch/ zein blends with two choline chloride-based DES (choline chloride-urea (CC-U) and choline chloride-glycerol (CC-G)) (Leroy, Decaen et al., 2012). Their research showed that CC-U and CC-G can be used as an efficient functional plasticizer for TPS. Besides plasticization, these two DESs were also able to reduce the water sensitivity of TPS and act as a good compatibilizer for the blend of corn starch and zein. Very recently, the effect of DES (choline or betaine-based), IL ([EMIM] salts and [NH 3 CH 2 CH 2 OH][HCOO]) and glycerol on the treatment of potato starch (16.5 wt.% water content) was compared (Zdanowicz, 2020). The results showed that urea-based DES (dissolving DES) and IL can dissolve potato starch while glycerol-based DES (non-dissolving DES) just gelatinized it. When used as plasticizer dissolving DES and IL resulted in TPS with a higher degree of amorphous than that of TPS from non-dissolving DES, thus lower tensile strength and much higher elongation at break as discussed previously on ILs. Like conventional ILs, this paper revealed that the composition of DES (the anion of choline salts and other components) also affect its performance on dissolving, plasticizing of starch as well as the mechanical properties of TPS. Another report, which studied 12 different DESs with choline chloride, betaine, malic acid, proline, lactic acid as HBA respectively, suggested that the composition of DESs affected the hydrogen-bonding formation between HBA anion and starch, consequently made great difference in their capacity to dissolve starch (Cao, Nian, Li, Wu, & Liu, 2019). All these results indicated that detailed systematic studies are needed for a thorough understanding of these influences on the DES application. Recently, several CC-based two-component DESs for starch dissolution and plasticization were summarized in one review (Zdanowicz, Wilpiszewska, & Spychaj, 2018). Not only CC-based DESs, imidazole-based two-component DESs were also prepared for starch dissolution and plasticization (Zdanowicz, Spychaj, & Maka, 2016). The performances of imidazole-choline chloride (IM-CC), imidazole-glycerol (IM-G), imidazole-citric acid (IM-CA) and imidazole-malic acid (IM-MA) on the dissolving and plasticizing of potato starch and highamylose (> 68 %) starch were compared in this report. Unlike IM-CA and IM-MA, IM-CC and IM-G (ratio 7:3) turned out to be good plasticizers for both potato starch and high-amylose starch with 35 wt.%. DES in the mixture. The obtained transparent and elastic film exhibit an amorphous structure and no obvious retrogradation was observed after storage for one month. However, further evaluations like the degradation of starch in DES and mechanical tests on the obtained materials are still needed to verify the application potential of these DESs.

Supercritical CO 2
Supercritical fluids are defined as substances that are held at or above their critical point with a certain temperature and pressure, which makes the distinct interface between liquid and gas disappear. This kind of fluid possesses both the properties of gas and liquid-like low viscosity, high diffusion coefficient (Table 1) and good permeability and dissolving capacity. All the properties make the mass transfer process easier for chemical reaction or abstraction, which usually means a higher reaction rate or a faster abstraction than that in traditional solvents. Furthermore, there is a temperature and pressure zone near the critical point within which a slight change in temperature and (or) pressure will result in a significant change in density and many other properties, which enables the fluids to be tunable for different solutes. For all the reasons mentioned above, the applications of supercritical fluids have become a hot point for scientific research in recent decades. Among all the supercritical fluids that have been reported, water and supercritical CO 2 (scCO 2 ) are the most commonly used and commercially available fluids.
Compared with water (critical point: 647.14 K and 22.06 MPa), the supercritical conditions of CO 2 are relatively easy to achieve (304.12 K and 7.37 MPa) and the mild processing conditions, together with its chemical inertness makes scCO 2 an ideal solvent for many compounds that are easy to be damaged (Tommasi et al., 2017). Besides, the separation of scCO 2 by simply depressurization enables it to be more Table 1 Typical values of physical properties of gases, supercritical fluids and liquids (Kemmere & Meyer, 2006).
Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 convenient and thus more economy than using other solvents. To the best of our knowledge, scCO 2 has been successfully used as a green solvent in many areas such as the extraction of caffeine from coffee beans and capsaicin from peppers, for dry cleaning and coating (de Aguiar et al., 2013;De Marco, Riemma, & Iannone, 2018;Liu, Zhang, Li, He, & Zhang, 2014;Ratcharak & Sane, 2014). In this part of the review, the application of scCO 2 as the solvent for polymer synthesis and processing was summarized with the hope that new ideas could be enlightened for the modification of starch.
1.3.1. Physical modification of starch with scCO 2 Normally, starches exist in nature in the form of granules with various diameters, depending on their botanical sources. Polarizing microscope observations and X-ray diffraction experiments have proved the existence of crystalline structures in starch granules. The crystalline zones have a significant influence on the processing of starches because they obstruct the accessibility of amorphous zones (Wang & Copeland, 2015). Thus, to get a deep knowledge of the structural change of starch macromolecules after modification, there have been many reports on the gelatinization of starches that is related to the disruption and breakdown of crystalline structures in granules (Ai & Jane, 2015;Carlstedt, Wojtasz, Fyhr, & Kocherbitov, 2015). Most of these experiments were carried out at various temperatures with water as medium and sometimes high pressure was applied. With a diffusivity that is two orders of magnitude larger than that of liquid solvents at relatively mild conditions as indicated previously, scCO 2 is believed to be a more effective solvent for the gelatinization of starches. To the best of our knowledge, Francisco and Sivik first reported the gelatinization of cassava, potato and wheat starch (72 % wt./wt. water content) with the assistance of CO 2 under different temperatures (50-70°C) and pressures (up to 30 MPa) (Francisco & Sivik, 2002). An increase in the degree of gelatinization (DG) was found when the pressure was kept below 8 MPa, higher than which decreases in the DG were noticed. They concluded that the DG was positively related to the plasticizing effect of scCO 2 , but negatively related to pressure. Muljana and Picchioni et al. also studied the effect of scCO 2 on the gelatinization of potato starch with various water content (16.2-40 % wt./wt.) (Muljana, Picchioni, Heeres, & Janssen, 2009). In their experiments, various temperatures (50-90°C) and pressures (up to 25 MPa) were applied and a maximum DG of about 14 % (quantified by DSC) was observed at the pressure of 25 MPa and temperature of 90°C. Contrary to the conclusion made by Francisco and Sivik with higher water content starch, the DG of lower water content starch was found to increase with increasing temperature and pressure. Except for these two papers, the influence of scCO 2 on the gelatinization temperature of different starch blends was also studied. In the research of Zaidul and Noda et al., different starches such as potato and cassava starch as well as their blends with wheat starch (70 % wt./wt. water content) at a series of ratios were gelatinized by scCO 2 treatment (20 MPa, 60°C) (Zaidul, Noda, Sharif, Karim, & Smith, 2014). Decreases of 10-18°C in the gelatinization temperature for all the samples were found via differential scanning calorimetry (DSC). They attributed this phenomenon to the pH changes caused by the contact of water with CO 2 and the plasticization of starch granules caused by scCO 2 .
In addition to researches on scCO 2 induced gelatinization, many other studies on the application of scCO 2 in the processing of starches such as supercritical fluid extrusion (SCFX) and SCFX-based crosslinking have also been carried out. Chen and Rizvi et al. studied the relationships between the DG and rheology as well as the expansion properties of wheat starch-water mixtures under SCFX processing conditions (Chen & Rizvi, 2006). A reduction of 14 % in the apparent viscosity (measured between 100-200 s −1 ) was found with the injection of 0.45 g scCO 2 /100 g sample, regardless of the DG of mixtures. Their experiments also indicated a minimum DG of 80 % was required for the starch-water mixture to get a suitable gas-holding rheology for the production of extrudate with nonporous skin and desirable cellular morphology. In their later research, Rizvi and Ayoub's group reported the production of SCFX-based cross-linked starch microcellular foam (XL-SMCF) using wheat starch with epichlorohydrin (EPI) and sodium trimetaphosphate (STMP) as cross-linker respectively (Ayoub & Rizvi, 2008;Manoi & Rizvi, 2010). The results showed that using scCO 2 for the SCFX process could achieve good control over the expansion of extrudates as well as their microstructures via adjusting the injection rate of scCO 2 . XL-SMCF from this experiment showed good water-resistance, which is favorable for its future application. Concerning the water-resistance property, a dual-modification of crosslinking and acetylation of corn starch was carried out in continuous scCO 2 extrusion and a more hydrophobic material was successfully achieved (Ayoub & Rizvi, 2011). A combination of SCFX and subsequent solvent exchange process was also studied in an attempt to get a deeper understanding of the influence of processing conditions on foam microstructure and macro-properties (Patel, Venditti, Pawlak, Ayoub, & Rizvi, 2009). The solvent exchange was proved, in this experiment, to be more important than other factors like crosslinking and the existence of scCO 2 for the formation of microcellular structures that have a great contribution to the brightness of material. Moreover, the brightness was also found to be linearly related to the density of the foams. In brief, starch foams produced by all these methods based on SCFX, with properties like low density, non-porous skin, and water resistance, have potential use as biodegradable materials.
Besides SCFX, the application of scCO 2 for tissue engineering, pharmaceutical processing and the production of antimicrobial materials have been explored in recent years. For instance, a supercritical immersion precipitation technique was used by Duarte et al. to evaluate the feasibility of using chloroform or dichloromethane as solvent and scCO 2 as a non-solvent for the preparation starch-polylactic acid scaffold for tissue engineering (Duarte, Mano, & Reis, 2009). A matrix with high porosity and interconnected microstructure was obtained by supercritical assisted phase-inversion (10-20 MPa, 35-55°C, 45 min). Both pressure and blend composition were proved to play important roles in controlling the morphology of the scaffold. Pure starch aerogel was also prepared to obtain low-cost biodegradable material for applications in pharmaceuticals (Ubeyitogullari & Ciftci, 2016). Wheat starch (10 wt.%) was first gelatinized in water and solvent-exchange with ethanol followed which was removed subsequently by scCO 2 (10 MPa, 40°C). The aerogel obtained at a low CO 2 flow rate (0.5 L/min) has a maximum surface area of 59.7 m 2 /g. The reason is that low CO 2 flow can prevent the "growing" of the ethanol drop in the matrix. A similar procedure was also used in another report for the preparation of starch aerogel thermal superinsulation material (Druel, Bardl, Vorwerg, & Budtova, 2017). The obtained results suggest that starch (8 wt.%) with higher amylose content produced aerogel with lower density and higher specific surface area (254 m 2 /g). A possible explanation is amylose can form stronger networks during retrogradation, thus resist shrinkage during subsequent solvent-exchange. It should be noted that pure amylose, however, did not work due to the heterogeneous structure (resulted from the fast retrogradation progress) that cannot stand the pressure during drying (8 MPa, 37°C). In addition to the production of the scaffold and aerogel, starch can also be used for the delivery of bioactive material. For example, Varona et al. impregnated lavandin oil into n-octenyl succinate modified waxy maize starch and the influences of different operational conditions like lavandin oil to starch mass ratio, pressure as well as temperature were analyzed (Varona, Rodríguez-Rojo, Martín, Cocero, & Duarte, 2011). Higher impregnation loads were observed at a relatively low concentration of scCO 2 fluids (low pressure, e.g., 10 MPa for 2 h) and a high depressurization rate (e.g., within 1 min). It is believed that low pressure could weaken the interactions between scCO 2 and solutes in it, which is favorable to strengthen the interactions between solutes and starch. High depressurization rate, on the other hand, could facilitate the condensation of lavandin oil. The droplets formed in this process deposited in starch particles and thus increased the impregnation load. Another important factor that influences the efficiency of impregnation is solubility and diffusion of solutes in scCO 2 . For example, Comin et al. found that simple lipids like oleic acid had a higher load (optimum: 15 MPa, 40°C for 8 h) on pregelatinized corn starch than flax oil (optimum: 30 MPa, 80°C for 8 h) which is a more complex lipid mixture (Comin, Temelli, & Saldaña, 2012). They attributed this to oleic acid's higher solubility and diffusion coefficient in scCO 2 than that of flax oil. In the research carried out by Souza et al., a kind of antimicrobial film was successfully developed by impregnating cinnamaldehyde into cassava starch-based film (de Souza, Dias, Sousa, & Tadini, 2014). The highest impregnation load was achieved at a higher pressure (25 MPa, 35°C for 15 h) which is opposite to that mentioned by Varona et al. A conclusion was drawn that the solubility of cinnamaldehyde in scCO 2 played a vital role in the impregnation process. No impregnation products from the above-mentioned researches showed a porous structure. In one recent report, potato starch aerogel (185 m 2 /g) impregnated with green coffee oil was prepared via a one-pot process (Villegas, Oliveira, Bazito, & Vidinha, 2019). This procedure integrated gelatinization, retrogradation and drying (20 MPa, 40°C, 2 ml/min CO 2 ) with impregnation process. Moreover, no solvent-exchange process was needed to fabricate the aerogel, which is different from aforementioned reports. However, like conventional solvent-exchange procedure, the long retrogradation period (at least 24 h) for the buildup of internal structure to prevent shrinkage is still the efficiency barrier for the preparation of starch aerogel.

Chemical modification of starch in scCO 2
Although there are many advantages to use scCO 2 as the solvent for chemical reaction and many reports on this kind of application have been published, papers on the chemical modification of starches in scCO 2 are still very few (Table 2). Muljana et al. reported the synthesis of fatty acid-starch esters in scCO 2 with different vinyl, methyl esters as esterification reagents (Muljana et al., 2010). The highest degree of substitution (DS = 0.31, 150°C and 8 MPa for 18 h) was obtained with the combination of vinyl laurate and K 2 CO 3 . In their later work, the acetylation of potato starch with acetic anhydride (Ac 2 O) was studied in sub-and supercritical CO 2 medium (Muljana, Picchioni, Knez, Heeres, & Janssen, 2011). The DS was evaluated within a broad range of pressure (8-25 MPa) at 90°C and the highest DS of 0.29 and 0.62 (corresponding to 1 h and 24 h reaction time respectively) were obtained near the critical point of the mixture (15 MPa). Besides, the DS was found to relate with the size of starch granules, which means the mass transfer inside the granule plays a more important role than intrinsic reaction kinetics in controlling the overall conversion rates. Apart from the esterification and acetylation, the grafting of starch was also studied with scCO 2 as the solvent. Salimi et al. reported that up to 52 % poly(L-lactic acid) was successfully grafted onto pregelatinized corn starch by controlled polymerization with stannous 2-ethyl hexanoate (Sn(Oct) 2 ) as a catalyst in scCO 2 (20 MPa, 100°C for 6 h with starch/lactic acid ratio = 1:3) (Salimi, Yilmaz, Rzayev, & Piskin, 2014). High pressure, temperature and flow rate of scCO 2 were observed unfavorable to the chain growth. However, a relatively higher temperature was proved acceptable for chain growth when the flow rate of scCO 2 was maintained at a relatively lower value.

Controlled modification of starch
Chemical modification, which could expand the application of starch to areas beyond imagination (e.g., medical plasma expander and drug carrier, drilling fluid in oilfield, film, and foam in packaging), has been under investigation for a long time. As is known, the application fields of materials are determined by their intrinsic properties that are closely related to their microstructures. In the case of polymers, their microstructures are influenced by the composition and structure of both backbones and side chains or groups. According to this principle, varieties of starch derivatives with different properties can be got by: -substituting hydroxyl groups with other functional groups; -grafting side chains with different composition, length, and structure; -varying the distribution of groups or chains along the backbone.
Lots of work has been carried out following the three mentioned guidelines. Compared with traditional chemical methods, from our point of view, the controlled modification of starch is of great interest as it could provide us with functional products that have tailor-made properties for specific applications. Another advantage of controlled modification is, as will be illustrated in the following part, that some properties can even be achieved or enhanced just by the switching of reaction sites. For this reason, much research has been done on the controlled modification of starches during the last decades. In this part of the review, attention was paid to the progress on the controlled modification of starches with the hope that more efficient work could be done in the future.

Regioselective functionalization of OH-groups
For polysaccharides like starch, the existence of three hydroxyl groups with different chemical environments in the AGU provides us the possibilities to achieve derivatives that have the same side groups/ chains but different properties. For example, after preserved at room temperature for a hundred days a significant degradation can be observed in the starch acetate prepared according to the conventional procedure (acetylation of starch in an aqueous medium with acetic anhydride as the reagent). On the contrary, regioselective synthesized 2-O-acetyl starch can be stored in sterilized distilled water for over 200 days without significant change in viscosity and only a slight decrease in the DS was observed (Liebert, Kulicke, & Heinze, 2008). Significant increases in the viscosity (400 % higher), glass transition temperature and degradation temperature (34°C and 36°C higher respectively) were also observed in the 2-O-acetyl starch when compared with conventional acetylated starch that has the same DS (Dicke, 2004). A possible explanation for this is the more uniform structure resulted from regioselective derivatization enhances intra-and intermolecular interactions (hydrogen bonds or hydrophobic interactions) and thus improves the macro-properties of the products. Besides, 2-O-acetyl starch only swells in water while conventional starch acetate was soluble with similar DS. Another interesting report was that 2-O/3-Oacetyl starch had a higher biodegradation rate than 6-O-acetyl starch when exposed to α-amylase (Roesser, McCarthy, Gross, & Kaplan, 1996). Clearly, the regioselective functionalization could endow starch derivatives with novel properties with currently available reagents thus expand the application fields of starches, especially in areas like biomaterials.
As far as we know, the regioselective functionalization of polysaccharide has been studied for over 50 years. Generally, there are three approaches available now for the effective regioselective functionalization of polysaccharides (see below): -protection and deprotection approach in which bulky protecting groups such as silyl and trityl groups are employed for the preparation of 6-O or 2,6-O protected intermediates (Petzold, Koschella, Klemm, & Heublein, 2003;Wang, Yang et al., 2015); -catalyzed selective functionalization of OH groups without using protecting groups (enzymes, inorganic salts, etc.) (Chakraborty, Sahoo, Teraoka, Miller, & Gross, 2005;Wang, Qu, Wu, Men, & Liu, 2017); -regioselective replacement of OH groups with other functional groups through reactions controlled by a stereochemical mechanism (S N 2 reaction etc.) (Gao, Liu, & Edgar, 2018;Tan, Li, Gao et al., 2017).

Protection and deprotection approach
Two different regioselective approaches, in which toluenesulfonyl Table 2 The application of scCO 2 in the modification of starches and corresponding general processing conditions.

-9
Corn Stannous 2-ethyl hexanoate (Salimi et al., 2014) a: Depressurization rate was used instead of CO 2 ratio. Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 (tosyl) and triphenylmethyl (trityl) groups were employed as leaving/ protecting groups for the introduction of 3,6-anhydro rings into amylose, were reported by Whistler et al. (Whistler & Hirase, 1961). Results showed that the trityl group had better regioselectivity on the derivatization of primary hydroxyl groups than the tosyl group. However, it should be noticed that the detritylation process carried out in the acid medium may cause the breakdown of polysaccharide chains. Besides the groups mentioned above, the silyl group has also been used for the regioselective functionalization of starches. For instance, a highly regioselective approach that using thexyldimethylchlorosilane (TDSCl) as protecting reagent was developed by Petzold et al. for the protection of primary hydroxyl groups in potato starch . The silylation of potato starch was carried out in the mixture of N-methylpyrrolidone (NMP) and ammonium at -20°C. Different from their another research in which the silylation was carried out in the mixture of DMSO and pyridine , 6-O-TDS starch ether could be obtained exclusively by this approach with a DS up to 1 even when the TDSCl was excessive. The introduced protecting group can be easily removed by the treatment with tetrabutylammonium fluoride (TBAF) solution in THF (R.T. to 50°C, 14-24 h) (Petzold, Klemm, Stein, & Günther, 2002). Besides the TDS group, TBAF or tetrabutylammonium hydroxide (TBAOH) can also be used for the efficient deacylation of potato amylose tripropionate (R.T. to 50°C, 1-24 h) to obtain highly regioselective amylose-6-O-ester (Zhang, Zheng, Kuang, & Edgar, 2014). Nevertheless, the protection-deportation approach is time-consuming and especially not atom economy for largescale applications. According to the aforementioned researches, not only the structure of the protecting reagent affects the selective protection of hydroxyl groups, but the reaction medium also has a significant influence on the regioselectivity. One example is the tosylation of starch. In the researches carried out by Whistler et al. and Clode et al., pyridine was used as the reaction medium for the tosylation of amylose and results shown that this process mainly occurred at the C-6 position (Clode & Horton, 1971;Whistler & Hirase, 1961). However, after switching the reaction medium to the mixture of DMAc and LiCl the tosylation at the C-2 position was found more preferable than that occurred at C-6 and C-3 position (Dicke, Rahn, Haack, & Heinze, 2001). The influence of solvent on the regioselectivity of protecting groups was related to the supramolecular structure of the polymer formed in specific conditions, but no evidence is available now and further studies are needed to support this hypothesis.

Catalyzed regioselective functionalization
Besides the utilization of protecting groups, regioselective derivatization of polysaccharides can also be achieved directly by using enzymes, catalytic active salts and heterocyclic compounds such as pyridine to promote the reaction. As for the enzyme-catalyzed site-selective functionalization, one problem faced is that the use of polar aprotic solvents, which are used to dissolve polysaccharides, may decrease the activity of enzymes (Klibanov, 1997). To overcome this problem, an approach developed by Paradkar and Dordick was employed by Bruno et al. to incorporate Subtilisin Carlsberg into surfactant micelles for the regioselective acylation of amylose film (Bruno et al., 1995;Paradkar & Dordick, 1994). Although this research was carried out successfully, only the surface of the amylose film was accessible for the enzyme and thus the efficiency of functionalization was not satisfactory. In the research carried out by Chakraborty et al., a new approach was proposed to solve this problem (Chakraborty et al., 2005). Starch nanoparticles were incorporated into Aerosol-OT (AOT, bis(2-ethylhexyl)sodium sulfosuccinate) micelles to improve its accessibility to immobilized Candida antartica Lipase B (CAL-B). As a comparison, free CAL-B was also incorporated into AOT micelles together with starch nanoparticles in the control experiments. The esterification of AOT-coated starch nanoparticles was carried out with toluene as solvent in which vinyl stearate, ε-caprolactone, and maleic anhydride were dissolved respectively as esterification reagent (mole ratio of reagent: AGU = 3:1). Starch-6-O-esters were obtained with DS ranging from 0.8 to 0.4 after reaction at 48°C for two days. Results showed that the catalyzed esterification process, whether the CAL-B was immobilized or not, was regioselective for the primary hydroxyl groups in the AGU. However, considering the post-purification, immobilized CAL-B is preferable for this approach. An alternative to the aforementioned approach was reported by Klohr et al., by whom protease was employed to achieve the regioselective functionalization of C-2 position (Klohr, Koch, Klemm, & Dicke, 2005). In their reported approach, protease was first activated in phosphate buffer solution (pH = 7.8, c = 0.15 M) and then lyophilized before added to the starch/DMSO solution. Various esterification reagents, such as vinyl acetate, acetic anhydride or propiolactone can be used for the regioselective substitution of starch in this approach and DS up to 1 can be reached. One thing should be noted is that chemical esterification may occur besides the enzyme-catalyzed reaction when this reaction is carried out at 40°C. To substantially suppress the chemical reaction, the esterification procedure can be carried out at a lower temperature (20-25°C) or conducted in an almost anhydrous system (water content < 0.01 %).
In addition to the enzyme, several salts and heterocyclic compounds can also catalyze the regioselective functionalization of starches. The differences are that the latter one can modify starch homogenously and the regioselective site usually is C-2 instead of C-6 from that of the enzyme. For instance, nine different salts (Table 3) were tested by Dicke for the catalyzed regioselective acetylation of starches in DMSO and their performances in regioselectivity were compared with that of DMAP/pyridine (Dicke, 2004). These salts, according to their site-selectivity and the pH values of their aqueous solutions, were divided into three types. High regioselectivity at the C-2 position was shown when using neutral, weak acid or alkaline salts as catalysts, while further acetylation would occur to some extent at C-6 and C-3 position when alkaline salts were employed. No catalytic activity was found when acid salt was employed as a catalyst. The regioselectivity in DMSO was related to the formation of intramolecular hydrogen bonds between the C-2 hydroxyl groups and that in the C-3 position of the neighboring AGU (Fig. 3). The formation of these hydrogen bonds, together with other factors such as solvent and ring oxygen, was believed to make the protons of C-2 hydroxyl groups have a stronger acidic character than the other two protons. However, it should be kept in mind that there is still no evidence to prove this hypothesis. In a more recent paper, the regioselective preparation of starch-based macroinitiator for ATRP in DMSO and imidazolium-based ILs (  . According to this report, the acylation of starch with vinyl chloroacetate in DMSO with sodium carbonate as catalyst mainly occurs at Table 3 The influence of catalysts, catalyst concentration, and acetylation agents on total and partial DS of starch acetate (Dicke, 2004 Besides the type of salt, regioselectivity was also influenced by the amount of salts used as well as the type of acetylation reagent ( Table 4). The latter factor was attributed to their differences in reactivity. The regioselective substitution of C-2 hydroxyl groups was also achieved by Liebert et al., in the study of whose starches were acetylated homogeneously by acetic acid/N,N'-carbonyldiimidazole (CDI) and acetic anhydride/imidazole respectively with DMSO as solvent (mole ratio of esterification reagent: base (imidazole): AGU = 1.3:1.3 ∼ 3:1) (Liebert et al., 2008). Starch acetates with DS ranging from 0.4 to 1.23 can be obtained after a reaction at 60°C for two hours and then kept at room temperature for another 16 h. In both cases, high regioselectivity at the C-2 position can be detected and was contributed to the formation of reactive intermediate (acetic acid imidazolide) and the possible solution state of starch in DMSO. The same approach was also employed by Hampe et al. to study the influences of solvent and many other factors (such as molecular weight and the concentration of starch) on the substitution patterns of starch acetate (Hampe & Heinze, 2014). Another approach used in their experiment was acetylating different starches by acetic anhydride with molten imidazole as the solvent and the reaction was kept at 100°C for one hour. When using DMSO as the solvent, potato starch was acetylated predominately at the C-2 position and the DS increased (up to 0.71) with a higher molar ratio of acetic anhydride to AGU (reagent/ AGU ratio up to 1.0). However, an even higher reagent/AGU ratio, which could result in additional acetylation of the C-6 hydroxyl groups, was needed to get similar DS at the C-2 position for degraded potato and tapioca starches. One possible explanation proposed for this was that degradation, as well as the branching points, reduced the amount of 2,3-hydrogen bonds between adjacent AGUs thus the 2-O proton had less activity for acetylation. Another interesting found was that, besides the reagent/AGU ratio, increasing the concentration of degraded starches in DMSO could also promote the acetylation of primary hydroxyl groups. When molten imidazole was used as the solvent, in contrast to that occurred in DMSO, regioselective acetylation at the C-6 position was found at a low reagent/AGU ratio (≤ 0.5 and 0.3 for potato and degraded potato starch respectively). At higher reagent/AGU ratio, hydroxyl groups at C-2/C-3 positions could also be acetylated and resulted in a random distribution of acetyl moieties.

Regioselective replacement of OH groups
Besides regioselective esterification, etherification and so on, the hydroxyl groups could also be selectively replaced by various nucleophiles (such as oxygen, nitrogen and sulfur nucleophiles) through S N 2 reaction (as seen in Fig. 4) (Cumpstey, 2013). To achieve this purpose, a good leaving group is needed to make the target carbon atom more susceptible to the attack of nucleophiles. Commonly used leaving groups are sulfonates (RS(O) 2 O-) and halides.
In the research carried out by Whistler et al., the tosylate group was employed for the introduction of 3,6-anhydro rings into amylose (Whistler & Hirase, 1961). According to the reported procedure, amylose was first tosylated at the C-6 position (35-60°C) and then the introduction of 3,6-anhydro rings was realized through the substitution of Fig. 3. Solution state of starch (amylose) in DMSO (Dicke, 2004).  Barsi et al., 2017;Shey et al., 2006) Water TEMPO/NaBr/ NaClO C-6 0.95 (Chang & Robyt, 1996) Water TEMPO/NaClO C-6 0.68 (Bragd et al., 2000) a The most preferable reaction site. b The largest partial degree of substitution could be got before other secondary hydroxyl groups are substituted.
In addition to tosylate groups, halides were also proved good leaving groups for the regioselective derivatization of starches. In one research carried out by Cimecioglu et al., 6-halogen-6-deoxyamylose with DS up to 1 was synthesized by the reaction of amylose with Ph 3 P/ N-bromosuccinimide (NBS) and methanesulfonyl chloride respectively with DMF/LiBr (LiCl) as solvent (70°C, 2 h) (A. Cimecioglu, Ball, Kaplan, & Huang, 1994). The halides were replaced by azide (70°C, 48 h) and then Staudinger reaction was employed to reduce azide to amine. 6-amino-6-deoxyamylose with DS up to 1 can be achieved according to this approach. A similar method was also employed in some recent researches for the preparation of 1,2,3-triazolium-functionalized starch with antifungal properties (80°C, 3 h for bromination, 80°C, 24 h for azidation) Tan, Li, Gao et al., 2017;Tan et al., 2016). Although the halogenation procedure was proved to be high regioselective, degradation was reported when a similar approach was employed for the chlorination of chitin with excess Nchlorosuccinimide at a relatively higher temperature (Sakamoto, Tseng, & Furuhata, 1994). Besides, the described procedure for the preparation of 6-azido-6-deoxyamylose was also time-consuming and not economical. To overcome these problems, a more direct and mild approach was introduced by Cimecioglu et al. for the azidation of amylose (Cimecioglu, Ball, Huang, & Kaplan, 1997). Amylose was dissolved in the mixture of DMF/LiN 3 and then Ph 3 P, as well as the freshly prepared CBr 4 /DMF solution, were added successively. This homogeneous reaction system was kept at room temperature for about 18 h with the protection of nitrogen. According to this approach, 6-azido-6-deoxyamylose can be achieved with nearly all the primary hydroxyl groups substituted. Given the high price of amylose, this one-pot approach has been extended to the azidation of different starches in recent years. For example, in one recent research potato amylose and maize amylopectin were functionalized with photochromic spiropyran via the combination of this method and Cu(I)-catalyzed azide-alkyne cycloaddition (CuAAC) (Barsi et al., 2017). This modified starch could be a promising photochromic coating for substrates like paper and leather. Lithium azide, because of its high hygroscopicity and the resulting adverse effect on azidation, was replaced with sodium azide by Shey et al. (Shey et al., 2006). Not only DMF, but DMAc was also used as the solvent in their experiments. In both cases, pregelatinized starches can be derivatized with high regioselectivity at the C-6 position while no desirable result could get from those untreated starches. This was attributed to the low solubility of untreated starch granules.
Besides S N 2 substitution, another reported stereochemically controlled reaction is 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO)-mediated oxidation of starch. One example is the successful oxidization of various polysaccharides (starch included) reported by Chang et al. (Chang & Robyt, 1996). Results showed that polysaccharides, whether water-soluble or not, could be oxidized with high regioselectivity at the C-6 position by using TEMPO with sodium bromide/sodium hypochlorite as co-catalyst (DS = 0.95, pH = 10.8, 0°C, 0.75-1.8 h). The high regioselectivity of this reaction was related to the steric hindrance caused by the bulky TEMPO ion as well as the conformation of AGU. Despite the high selectivity, the practical application of this technique was hindered by the use of bromide which is considered to be hazardous to both environment and production equipment. To overcome this problem, a bromide-free approach in which TEMPO ion can be directly generated by hypochlorite was reported by Bragd et al. (Bragd, Besemer, & van Bekkum, 2000). According to this approach, high regioselectivity can be achieved with a significant suppression on the degradation of molecular chains when the reaction was carried out under suitable conditions (temperature ≤ 20°C and pH < 9). Considering the comparable reaction rate (DS = 0.68, 20°C, pH = 8.5, 0.75-2 h) to that of bromide co-catalyzed reaction (DS = 0.7, pH = 8.5, 2°C, 2-5 h), this approach should be a good alternative for the conversion of primary alcohol groups into aldehyde and carboxyl groups.
For clarity, the reported methods for regioselective modification of starch are summarized in Table 4. As suggested in the table, while most of the protection-deprotection and region-replacement methods aim at the C-6 position, catalyzed derivatization mainly generates C-2 modified starch. So far as we know, there is still no report on the C-3 selective derivatization available. The applications of green solvents like ILs are either failed (salt-catalyzed method) or not reported yet.

Controlled grafting of starch
Compared with starches modified through the substitution of hydroxyl groups with small molecules, such as hydroxyethyl starch and cationic starches, brushed starches have some unique properties which can expand the application of starch-based materials as high valueadded products. For example, amphiphilic branched starches have been synthesized as potential drug carriers with the property of self-assemble (Li et al., 2015;Wu et al., 2016;Yang et al., 2014). Besides, many other grafted-starches have also been reported for various applications such as biodegradable polymers, absorbents, and enhanced oil recovery (EOR) (Dragan, Apopei Loghin, & Cocarta, 2014;Najafi Moghaddam, Fareghi, Entezami, & Ghaffari Mehr, 2013;Song, Zhang, Ma, Wang, & Yang, 2007). As it is known, there are many other factors besides the botanical sources of starches have a significant influence on the properties of final branched products. For instance, the intrinsic properties of branch chains, as well as their grafting density and length, are directly related to the rheological and mechanical properties of modified starches. To achieve a successful application of these materials, it is of great importance to have a deep knowledge of the relationship between macroscopic properties and microstructures. Thanks to the development of techniques in controlled/living polymerization, many new approaches have been adopted for the synthesis and analysis of branched starches in recent years. In the following part of this review, the progress and open questions in the controlled modification of starch with techniques like metal-catalyzed living radical polymerization (mainly atom transfer radical polymerization (ATRP) and Cu°-mediated living radical polymerization (Cu°-mediated LRP)), reversible additionfragmentation chain transfer (RAFT), nitroxide-mediated polymerization (NMP), "click" and chemoenzymatic methods are summarized.

Metal-catalyzed living radical polymerization
Techniques used for metal-catalyzed living radical polymerization can be categorized into two main groups, which are known as ATRP and Cu°-mediated LRP (or single-electron transfer living radical polymerization (SET-LRP)), according to the differences in the reaction Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 mechanism. Metal-catalyzed living radical polymerization is one of the most widely used controlled polymerization techniques for the preparation of polymers with defined architectures, through which desirable properties as predesigned can be achieved conveniently. Many works have been done on this subject since Wang et al. reported the first paper in 1995 and the progress of this technique has been kindly reviewed (Anastasaki et al., 2016;Boyer et al., 2016;Ding, Jiang, Zhang, Cheng, & Zhu, 2015;He, Jiang, Zhang, Cheng, & Zhu, 2013;Matyjaszewski, 2012;Wang & Matyjaszewski, 1995). According to literature, most of the reported controlled grafting of starches were accomplished by ATRP (see example in Fig. 5). Surface-initiated atom transfer radical polymerization (SI-ATRP) was first employed by Liu et al. for the grafting of potato starch granules with the hope that grafted starches could be used for the production of biodegradable plastics (Liu & Su, 2005). Bromoacetyl bromide was immobilized onto the surface of starch granules in the first step then poly(n-butyl acrylate) was grafted from the surface of the macroinitiator with CuBr/1,10-phenanthroline as the catalyst in toluene. A 9.9 % of monomer conversion was achieved after polymerization for 5 h. Similar method was also used by Moghaddam et al. in the modification of starch surfaces with polystyrene (PS) and polyacrylamide (PAM) to develop biodegradable thermoplastic copolymers (Najafi Moghaddam et al., 2013). In another research reported by their group, both free radical polymerization and ATRP were used for the synthesis of starch-g-polyacrylamide and starch-g-polyhydroxyethylacrylate and the performance of copolymers in controlled release of cephalexin antibiotic were compared (Avval, Moghaddam, & Fareghi, 2013). Results showed that it would take a longer time for copolymers obtained from the ATRP method to release all the loaded drugs. It was attributed to the more orderly chain length and larger graft densities (2.51 vs. 1.89) in these polymers. Recently, SI-ATRP was also employed to prepare surface quaternized starch nanoparticles for the removal of heavy metals in waste water (Su et al., 2020). In this report, poly(dimethylaminoethyl methacrylate) was first grafted from the surface of starch nanoparticles with CuBr/2,2′-bipyridine as the catalyst in water and then quaternized with methyl groups.
To get a high graft density, Wang et al. prepared a starch-based macroinitiator in ionic liquid ([AMIM]Cl) and then grafted that with PS Fig. 5. The reaction mechanism for ATRP and an example of its application in starch modification.
Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 and PMMA respectively . Compared with products from SI-ATRP, it was reported that both the graft density and ratio were significantly improved because of the homogeneous reaction condition in preparing macroinitiator. Recently, water-soluble waxy potato starch-based macroinitiator was also homogeneously prepared in DMAc/LiCl then the controlled grafting of with acrylamide (AM) and N-isopropylacrylamide (NIPAM) was carried out via aqueous Cu°mediated LRP (Fan, Boulif, & Picchioni, 2018;Fan, Cao, van Mastrigt, Pei, & Picchioni, 2018). In contrast to conventional ATRP, copper (either generated from the disproportionation of CuBr in water or fine powder that directly added to the system) instead of CuBr was used as the catalyst in Cu°-mediated LRP. This method was proved to be more effective for controlled polymerization in the aqueous system (Gavrilov, Zerk, Bernhardt, Percec, & Monteiro, 2016;Zhang, Wilson et al., 2013).
In the reported synthesis of starch-g-polyacrylamide (St-g-PAM) via aqueous Cu°-mediated LRP, copper particle from the in-situ disproportionation of CuBr was used as catalyst and monomer conversion ranges from 25 % to 66 % with corresponding polymer chain length (degree of polymerization, DP) in the range of 200 to 600 could be obtained within two hours (Fan, Cao et al., 2018). However, the in-situ generated copper didn't work when targeting polymers with higher DP. Fine copper powder that directly added to the reaction system, on the contrary, was proved to be an effective catalyst, although reasons are still unknown, in another report for the synthesizing of starch-g-poly (acrylamide-co-N-isopropylacrylamide) (St-g-(PAM-co-PNIPAM)) with DP around 5000 (Fan, Boulif et al., 2018). As mentioned earlier, starches that modified by the ATRP method are potential candidates for drug carriers and so on. However, the biotoxicity of the commonly used copper catalyst is one of the problems that hinder their practical applications. Several techniques have been developed for decreasing the concentration of Cu(I) used for polymerization as well as the residues in final products (Aggarwal, 2010;Matyjaszewski, 2012;Wang & Matyjaszewski, 2010). Among them, two reports seem to have significant meaning for eliminating the influence of monovalent copper ions. One was reported by Wang et al., in the research of whose MMA was polymerized with FeBr 2 as the catalyst and no ligand was needed (Wang & Matyjaszewski, 2010). In contrast with traditional ATRP, their experiments were successfully carried out in several polar solvents (N-methyl-2-pyrrolidone, DMF and acetonitrile), which were believed to act as ligands for the dissolution of FeBr 2 and thus the polymerization could be controlled. One should note that, however, too strong coordination between solvent and Fe could lead to failure in controlling the polymerization. In another research reported by Aggarwal, the technique named as supported aqueous-phase catalysis (SAPC) for activator generated by electron transfer (AGET) ATRP was developed (Aggarwal, 2010). N,N,N',N',N"-pentamethyldiethylenetriamine (PMDETA) was used as ligand and its complex with CuBr 2 was loaded onto various inorganic (for example Na-Clay) and organic materials (starch and cellulose). The performance of SAPC AGET ATRP in polymerization and its ability in controlling the monovalent copper ions in final products were tested by the synthesis of linear poly(benzyl methacrylate) (PBnMA) and PMMA. Catalyst-free polymers were obtained with narrow PDI in the research, which could be taken as effective proofs for the success of the technique. What's more, branched/ star PBnMA was also successfully synthesized in his later studies, which reflected the development potential of SAPC AGET ATRP in wide areas.

RAFT and NMP
Although RAFT-mediated polymerization has advantages in preparing metal-free polymers with tailor-made structures and many research have been done on its application in the modification of cellulose and chitosan (Hufendiek, Trouillet, Meier, & Barner-Kowollik, 2014;Semsarilar & Perrier, 2010;Zhang, Wang et al., 2013), only a handful of papers were found to be related to the modification of starch by this technique (see example in Fig. 6). This may be attributed to the following two reasons: the procedure is more complex than ATRP, most of the chain transfer agents are unavailable as commercial products.
In the research carried out by Lu et al., starch-based CTA was synthesized for the controlled grafting of poly(vinyl acetate) (starch-g-PVAc) with AIBN as initiator and starch-g-poly(vinyl alcohol) (starch-g-PVA) was also prepared as derivative (Lu, Xiao, & Sun, 2012;Lu, Xiao, Xu, & Ye, 2011). It is interesting to notice that starch-g-PVAc showed the ability to self-assemble into micelles when mixed with water, while starch-g-PVA could physically crosslink into a hydrogel. All these properties were found to depend on the length of grafted chains that could be well controlled with a narrow PDI (around 1.2).
Compared with RAFT, there is much less report on the application of NMP in the modification of starch. As far as we know, the only reported research on the modification of starch via NMP was carried out by Cazotti et al. (Cazotti et al., 2020. In their work, the surface of starch nanoparticles (SNP) was modified with reactive vinyl groups for the introduction of NMP initiating groups. Afterwards, PS, PMMA, poly (methyl acrylate) (PMA), and poly(acrylic acid) (PAA) were grafted from the SNP surface respectively at different temperatures (90-120°C) in DMSO. Except for this report, most of the researches on the NMP grafting of polysaccharides focused on macromolecules such as cellulose, chitosan, and guar gum (García-Valdez, Champagne-Hartley, Saldívar-Guerra, Champagne, & Cunningham, 2015;Glasing, Bouchard, Jessop, Champagne, & Cunningham, 2017;Holmes, 2007;Karaj-Abad, Abbasian, & Jaymand, 2016;Moreira et al., 2015).

Click chemistry
The introduction of "click" chemistry to polymer science opened up a new way to synthesize various polymers with well-defined structures, especially when this was combined with controlled/living radical polymerization techniques like ATRP and so on. Reviews on different techniques that use "click" reaction to prepare polymers and various clickable initiators, monomers and polymers are available in recent years ( , Hoogenboom, 2010Huang, Qin, & Tang, 2018;Lowe, 2010;Mansfeld, Pietsch, Hoogenboom, Becer, & Schubert, 2010;Tasdelen, 2011).
When used for the modification of starch, one commonly employed "click" reaction is copper-catalyzed [3 + 2] Huisgen cycloaddition reaction (CuAAC) (Fig. 7). With this method, oligocaprolactone was grafted onto soluble starch in the research of Uliniuc et al. and amphiphilic copolymer was obtained via tuning the chain length of oligocaprolactone (Uliniuc et al., 2013). This amphiphilic starch-based copolymer was proved able to form micelles in water at low concentration, which renders it a potential candidate for applications like controlled drug delivery. In another report, covalently crosslinked Kraft pulp-starch composite was prepared via CuAAC and significantly enhanced mechanical properties were observed after the crosslink (Elchinger, Awada, Zerrouki, Montplaisir, & Zerrouki, 2014). Besides using CuAAC alone, reports on the combination of CuAAC with ATRP, RAFT, ROP and enzymatic method in the grafting of starches are available in recent years. In one research carried out by Bertoldo et al., homopolymers of methacrylate, butyl acrylate, and dimethylaminoethyl methacrylate synthesized via ATRP were "click" onto amylose separately (Bertoldo, Zampano, Terra, & Castelvetro, 2011). The reported method for functionalizing starch with alkynyl groups, in which the AGU was oxidized by sodium periodate to generate aldehyde groups and then reduced with propargylamine/sodium cyanoborohydride (see Fig. 7), could be easily and efficiently carried out in mild reaction conditions. However, the yield of this method still needs to be improved. In another report, amylose-graft-poly(n-butyl methacrylate) (Am-g-PBMA) was successfully synthesized through the combination of ATRP and CuAAC (Borsacchi et al., 2014). Subsequent Electron paramagnetic resonance (EPR) analysis revealed the presence of residual Cu (II) after the extensive purification process. The total amount of residual Cu(II) which existed in three different forms (coordinate with ligands, coordinate with triazole nitrogens and Cu(II) cluster) was calculated to be about 7 % of that introduced as catalyst. The last two forms of Cu(II) were believed to be difficult to be removed through prolonged dialysis. This may hinder the practical application of CuAAC in the production of biocompatible materials. Compared with ATRP, reports on the combination of CuAAC with RAFT and ROP is relatively fewer (Karami Ghaleseiedi, Dadkhah Tehrani, & Parsamanesh, 2018;Xiao, Lu, Xu, & Huang, 2011).
Besides CuAAC, the thiol-ene "click" reaction was also reported for synthesizing starch-based materials. One example is the synthesis of starch-based hydrogel for cell encapsulation . In this research, sulfobetaine functionalized starch acrylate could be rapidly and conveniently crosslinked with dithiol-functionalized poly(ethylene glycol) under physiological conditions. The obtained biodegradable antifouling hydrogel possibly could be a good candidate for 3D cell encapsulation and culture in vitro. In another report, corn starch was functionalized with allyl chloride and L-Cysteine respectively and hydrogel can be obtained via simply mixing these two modified starches (Li, Tan, Xu, Lu, & Wang, 2017). With good mechanical and biodegradable properties, this hydrogel could be a promising material for tissue engineering. In addition to hydrogels, light-mediated thiol-ene "click" was also employed for the preparation of macroporous peptideimmobilized antimicrobial starch sponges (Yang et al., 2019). Potato starch was first esterified with norbornene anhydride in DMSO and sponge was prepared subsequently via the lyophilization of starch hydrogel that obtained from the dithiol-functionalized poly(ethylene glycol) photocrosslinking. An antimicrobial peptide with thiol group was then immobilized on the sponge via the thio-ene photoreaction. The obtained biodegradable sponge displayed good hemostatic and antimicrobial performance.
Another "click" reaction for the modification of starch recently reported is Diels-Alder reaction. For example, this reaction was combined Fig. 6. The reaction mechanism for RAFT and an example of its application in starch modification.
Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 with organocatalytic transesterification for the synthesis of starch-farnesene bio-renewable polymer (Orzolek, Rahman, & Iovine, 2018). Trans-β-farnesene, an acyclic sesquiterpene hydrophobic building block that is produced through a sustainable large-scale fermentation process, was employed as both diene and solvent for Diels-Alder reaction to introduce ester group for the following modification of starch. 1,5,7triazabicyclo[4.4.0]dec-5-ene (TBD) was used as an organocatalyst for the homogeneous transesterification reaction in DMSO to obtain modified starch with DS from 0.3 to 2.6. The results suggest that linear amylose is more favorable for the organocatalytic transesterification than branched amylopectin due to their difference in molecular structure. This method paves a new way to modify polysaccharides. What's more, the double bond in farnesene can be kept intact after transesterification, thus potentially can be used for further modification like thiol-ene reaction. In Gabilondo's group, starch was functionalized with furfuryl isocyanate and then cross-linked with water-soluble polypropylene glycol-block-polyethylene glycol-block-polypropylene glycol-based bismaleimide and polyethylene glycol-based tetramaleimide respectively (González et al., 2018(González et al., , 2019. Graphene sheets and cellulose nanocrystals (CNC) were added separately to obtain hydrogels with properties like antimicrobial activity, conductivity or improved macroporous structure for drug delivery.

Chemo-enzymatic modification
Various enzymes have been discovered and employed for the processing of starches in the past few decades. For example, amylases are used to hydrolyze starches into small molecules like maltose and glucose, while α-glucanotransferases (AGTases) are useful for remodeling the molecular structures of starches. The properties and applications of microbial AGTases in the modification of starches have been kindly reviewed by van der Maarel et al. and as illustrated in their paper varieties of derivatives could be achieved by this treatment (van der Maarel & Leemhuis, 2013).
Although amylose, which has a helical conformation thus is expected to act as a good host molecule for the preparation of functional materials, could also be prepared by AGTases, one of the most effective ways to get the pure product is considered to be α-glucan phosphorylase (GP) catalyzed α-D-glucose 1-phosphate (G-1-P) polymerization (Kadokawa, 2014). Maltooligosaccharides, the DS of which should be large enough to be recognized by GP, are needed as a primer to initiate the polymerization. Take potato GP catalyzed polymerization as an example, maltotetraose is usually employed as the primer and the chain will propagate at the nonreducing end of the primer in a step-growth manner. Hence, this process could be taken as a living polymerization and the molecular weight of resultant amylose can be tailored by varying the molar ratio of G-1-P to primer while the PDI could be controlled within 1.2 (Kadokawa, 2014). The application of this technique in the preparation of functional polymers can be seen in the review composed by Kadokawa (Kadokawa, 2014). As implied by this review and other reports, amylose has been successfully grafted to many other polysaccharides like alginate, chitosan, and cellulose by combining GP catalyzed polymerization with certain chemical reactions (chemoenzymatic approach) (Kadokawa, Arimura, Takemoto, & Yamamoto, 2012;Kadokawa, Egashira, & Yamamoto, 2018;Omagari, Kaneko, & Kadokawa, 2010). Besides grafting, amylose block copolymers were also successfully synthesized in recent years. One example is the synthesis of poly(2-vinyl pyridine)-b-amylose (P2VP-b-amylose) which generates a copolymer with rod-coil structure (Kumar, Woortman, & Loos, 2015). In this research, maltoheptaose lactone was reacted with amino-terminated P 2 VP and then GP catalyzed polymerization was carried out. The obtained copolymer is expected to possess both the inclusion ability from amylose due to its helical conformation with a hydrophobic cavity within the helix and the intermolecular complexation ability from P 2 VP due to the strong hydrogen bond, therefore, should be a good candidate for the fabrication of selfassembly materials. Another method for the preparation of functional supramolecular materials with amylose as host molecules is the socalled "vine-twining polymerization" (Kadokawa, Tanaka, Hatanaka, & Yamamoto, 2015). According to this method, GP catalyzed polymerization is carried out in a dispersion solution of appropriate hydrophobic polymer. The hydrophobic polymer "rod" will be included in the cavity of the synthesized amylose helix. Several hydrogels were synthesized via this method and review on this now is available (Orio, Yamamoto, & Kadokawa, 2017;Tanaka et al., 2015).
In addition to the materials mentioned above, this chemoenzymatic approach has also been employed to modify the surfaces of different materials to prevent biofouling. For instance, potato phosphorylase was tested by van der Vlist et al. for the grafting of amylose brushes from the surface of silicon and Deinococcus geothermalis branching enzyme was then employed for the preparation of hyperbranched polysaccharide surface coating (van der Vlist, Schonen, & Loos, 2011). Results from Xray photoelectron spectroscopy (XPS) and spectroscopic ellipsometry revealed the successful grafting of both linear and hyperbranched brushes. A similar method was also used by Mazzocchetti et al. to graft amylose brushes from the surfaces of Si and Au, in which experiment 3aminopropyldimethyethoxysilane (APDMES) and cystamine were used respectively for the amination of the two different surfaces and then maltoheptaoses were immobilized onto the surfaces as primers to initiate polymerization (Mazzocchetti, Tsoufis, Rudolf, & Loos, 2014). Phosphorylases from rabbit muscles and potatoes were studied for the catalytic polymerization of G-1-P and it was reported that amylose brushes synthesized by rabbit muscle phosphorylases showed a better performance in preventing protein adsorption.
Despite all the advantages brought by GP catalyzed polymerization, there are still some problems faced by the industrialization of this approach. For example, the monomer used in this approach, G-1-P, is too expensive and the commonly used potato phosphorylase is not thermostable enough for practical application. Numerous research has been done to overcome these problems and these could be seen in the review composed by Ohdan (Ohdan, Fujii, Yanase, Takaha, & Kuriki, 2006). According to this article and other reports, the most suitable method available by now for the synthesis of amylose is the combining use of sucrose phosphorylase (SP) with GP in the presence of inorganic phosphate with sucrose as substrate. G-1-P, produced as the result of the SP catalyzed phosphorolysis of sucrose, was used as the substrate for GP catalyzed polymerization (Fig. 8) (Doblado-Maldonado, Gomand, Goderis, & Delcour, 2017;Qi, You, & Zhang, 2014;Waldmann, Gygax, Bednarski, Randall Shangraw, & Whitesides, 1986). Results showed that amylose can be obtained with a higher yield than directly synthesized from G-1-P. The molecular weight of amylose from this method could also be controlled by varying the molar ratio of sucrose to primer and the PDI was found to be about 1. However, no report has been found by now on the application of this method in the preparation of functional materials as that synthesized from GP catalyzed polymerization of G-1-P.

Conclusions
With the development of technology and society, increasing attention is being paid to the utilization of polysaccharides such as starch to reduce the reliance on fossil fuel that is nonrenewable. Starch, as a kind of widely existed renewable polysaccharide, has been modified for the applications in various areas that are now being extended gradually with the development of polymer science and technology. However, although megatons of starch derivatives are produced and consumed every year, the conventional way in which they are modified is still not eco-friendly enough to reduce the environmental impact (wastewater, toxic organic solvent). What's more, most of the starch products are low value-added and their performance is far from excellent for application, which means in some cases large amounts of them have to be used to achieve desirable properties. Thus, nowadays more focus is being paid to the development of new techniques that could produce "high- Y. Fan and F. Picchioni Carbohydrate Polymers 241 (2020) 116350 performance" starch derivatives in a "green" way. In the past decades, one interesting progress in the modification of starch is the introduction of "green" solvents such as ionic liquids (ILs) and supercritical CO 2 (scCO 2 ) for the modification of starch. Not only could ILs serve as reaction medium, in some cases, they can also catalyze the chemical reaction or act as a better plasticizer than traditional ones such as glycerol. Success was also obtained when using scCO 2 as the solvent for the modification of starch, in both physical (gelatinization, extrusion, foaming, and impregnation) and chemical (esterification and grafting) ways. While the introduction of "green" solvents provides us the possibilities to prepare desirable products without the significant environmental impact that caused by the use of organic solvents in traditional methods, controlled modification (regioselective functionalization of hydroxyl groups and controlled grafting) of starch makes it possible to obtain starch derivatives with tailor-made properties. Generally, regioselective functionalization could be achieved by three different methods, which are the protection-deprotection approach, catalyzed functionalization and site-selective replacement of OH-groups respectively. In these methods, regioselectivity is related to steric hindrance or supramolecular structure formed in specific solvents. As for controlled grafting, methods like ATRP, Cu 0 -mediated LRP, RAFT, click reaction and chemo-enzymatic approach are already available.
In the future, the combination of these methods could be of interest for the preparation of high value-added starch-based materials in a green and efficient way. Although there is still lots of work to be done, it is clear that the development and application of starch-based materials could benefit significantly from the progress of "green" solvent and controlled modification. The same is true for the sustainable development of human society.