How can you stabilize enzymes in pellets made from microcrystalline cellulose?

How can you stabilize enzymes in pellets made from microcrystalline cellulose?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I want to make pellets consisting of mainly the following:

  • microcrystalline cellulose
  • saccharose
  • rice starch
  • ascorbic acid
  • glycerin

and an enzyme as active component.

Is it furthermore necessary to add protease inhibitors or other stabilizers to remain the activity of the enzymes for a periode of at least one year OR is the dry environment conservation enough?

The enzyme is called histaminase and is extracted from kidney cortex through a series of centrifugation, filtration and dialysation. To be more precise, it is a liquid protein extract containing small amounts of this enzyme. No buffer etc. is added to the protein extract so far. The whole purification process is done at 4°C. I don't know how long the enzyme is stable without further preparation. Probably not that long because some proteases should remain in the natural homogenate from porcine kidney cortex. The other components of the pellets are of pharmaceutical quality. The aim of this experiment is to feed the pellets to dogs with digestive complaints. For this purpose the pellets will get a enteric coating to survive acid environment of the stomach.

As you can probably guess from my comment, there are a lot of factors to take into account when thinking about protein stability. Here's what I would recommend you do.

First, make sure your extract is in a neutral buffer, like PBS, HEPES, or something similar. This is a fairly standard procedure when working with proteins. Next, since this is a kidney extract, there are likely to be gobs of proteases around, so I would strongly recommend adding broad-spectrum protease inhibitors, and keeping the extract as cold as possible to inhibit any enzyme activity. I would also add BSA at 10 mg/ml, both to stabilize the proteins and keep them from precipitating while in solution, and to lessen the chances that any proteases present will attack your enzyme of interest.

Once you have your solution stabilized, I would aliquot it and lyophilize it. Proteases are most active in solution, so once dried you'll reduce their activity even more. You can now add the inert ingredients (microcrystalline cellulose, etc.), mix thoroughly, press your tablets, and add the enteric coating. Finally, once everything is prepared, I would store the final product at as cold a temperature as is convenient. Again, this will inhibit any enzyme activity and prevent degradation.

I don't know if you have a functional assay to determine activity of the enzyme of interest, but if you don't I'd highly recommend developing one if at all possible. This way, you can test the activity of the fresh preparation of extract and compare it to various stages of processing into tablets, especially the finished goods, to ensure that you haven't lost much activity. I would also test random tablets throughout the year to make sure they haven't lost activity in storage, otherwise you won't be able to compare your experimental results.

Biodegradation of nanocrystalline cellulose by two environmentally-relevant consortia.

Biodegradation of nanocellulose versus non-nanocellulose was compared.

Consortia derived from anaerobic digester versus wetland were employed.

Sulfuric acid hydrolyzed nanocellulose degraded faster than microcrystalline cellulose.

Distinct microbial community shifts occurred during degradation of two celluloses.

Nanocellulose was more biodegradable, but likely via different pathways.

Characteristic microcrystalline cellulose extracted by combined acid and enzyme hydrolysis of sweet sorghum

Microcrystalline cellulose (MCC) has been widely used in medicine, food and cosmetic industries. In this study, a combination method by using hydrochloric acid hydrolysis and fibrolytic enzyme purification was studied to extract MCC from sweet sorghum. The response surface methodology was employed to optimize the hydrochloric acid hydrolysis condition and therefore to maximize the MCC yield and cellulose content (purity). The optimal conditions for the hydrochloric acid hydrolysis were determined to be acid concentration of 7.0%, liquid–solid ratio of 17.3:1, time of 90 min, and temperature of 40 °C. Under such conditions, the yield and cellulose content of acid-extracted MCC were 81.8% and 93.2%, respectively. For enzyme refining of acid-extracted MCC, the optimum conditions were enzyme dosage of 4000 U/g substrate and time of 2 h, with which the yield, cellulose content and DP of the refined MCC were 80.03%, 99.80% and 287, respectively which were comparable to that of the commercial MCC (Lowa ® PH101). Scanning electron microscopy, X-ray diffraction, Fourier transmission infrared spectroscopy, thermogravimetric analysis (TGA), and 13 C NMR were used to characterize the refined MCC. The refined MCC demonstrated rod-shaped morphologies, and had a series of characteristic absorption peaks and chemical groups pertain to cellulose as similar to the Lowa ® PH101. The X-ray diffraction pattern and 13 C NMR spectrum reflected that the refined MCC had typical cellulose I structure. TGA indicated that the refined MCC had good thermal stability. This study showed sweet sorghum is a potential low-cost raw material for MCC production, and the combined acid-enzyme extraction method is promising to extract high purity MCC from cellulosic substrate.

This is a preview of subscription content, access via your institution.

A review of bioreactor technology used for enzymatic hydrolysis of cellulosic materials

Cellulases are costly, a principal challenge of enzymatic hydrolysis of cellulosic materials for bioethanol production. For an efficient cellulase production, fungi are preferred over bacteria due to their permeation capability and versatile substrate consumption. Some limitations in the enzymatic hydrolysis step prevent the process to be economically feasible. Different strategies have been investigated to overcome these limitations, including genetic engineering, enzyme recycling, high solid loadings, pretreatment technologies, supplementation of cellulases with additives and application of nanomaterials for improving the thermal and pH stability of cellulases. Several studies have been performed in various bioreactors with the target to procure higher yields of glucose in the enzymatic hydrolysis step. The key factors for designing a bioreactor include efficient mixing, sufficient mass transfer, low shear stress, low foaming problems and low consumption of water and energy. In this scenario, different bioreactor configurations, including stirred tank bioreactor, horizontal rotating tubular bioreactor, airlift bioreactor, membrane bioreactor, reciprocating plate bioreactor, solid-state fermentation bioreactors have been reviewed for cellulase production with the aim to investigate main factors for designing a bioreactor.

Graphical abstract

This is a preview of subscription content, access via your institution.


The sacrificial biomass templating technique was used to enhance the sorption performance of CaO-based pellets that were prepared via an extrusion-spheronization method. Five types of biomass materials were used as the templates: microcrystalline cellulose, corn starch, rice husk, sesbania powder, and lycopodium powder. It is found that the addition of biomass templates is effective to improve the cyclic CO2 sorption capacity of the CaO-based pellets. However, two opposite enhancement tendencies of CO2 uptake were observed with the increment of biomass addition. For microcrystalline cellulose, corn starch, and rice husk, more addition amounts would result in better improvement of CO2 sorption performance of the CaO-based pellets. It is attributed to the generated porous microstructure and large amounts of small grains. However, for sesbania powder and lycopodium powder, a decreasing enhancement tendency of the CO2 sorption performance was found with the increasing addition amount. It is probably due to the accelerated sintering of the sorbent because of the presence of excessive amounts of alkali metal elements. Moreover, all biomass-templated CaO-based pellets possess a high anti-attrition capacity.

Immobilization and enzymatic properties of glutamate decarboxylase from Enterococcus faecium by affinity adsorption on regenerated chitin

Glutamate decarboxylase (GAD, EC is an important enzyme in gamma-aminobutyric acid biosynthesis and DL-glutamic acid resolution. In this study, the Enterococcus faecium-derived GAD was successfully immobilized by regenerated chitin (RC) via specific adsorption of cellulose-binding domain (CBD). The optimal binding buffer was 20 mmol/L phosphate buffer saline (pH 8.0), and the RC binding capacity was 1.77 ± 0.11 mg cbd-gad /g rc under this condition. The ratio of wet RC and crude enzyme solution used for immobilization was recommended to 3:50 (g/mL). To evaluate the effect of RC immobilization on GAD, properties of the immobilize GAD (RC-CBD-GAD) were investigated. Results indicated RC-CBD-GAD was relatively stable at pH 4.4–5.6 and temperature − 20–40 °C, and the optimal reaction pH value and temperature were pH 4.8 and 50 °C, respectively. When it was reacted with 5 mmol/L of follow chemical reagents respectively, the activity of RC-CBD-GAD was hardly affected by EDTA, KCl, and NaCl, and significantly inactivated by AgNO3, MnSO4, MgSO4, CuSO4, ZnSO4, FeCl2, FeCl3, AlCl3, CaCl2, and Pb(CH3COO)2. The apparent Km and Vmax were 28.35 mmol/L and 147.06 μmol/(gRC-CBD-GAD·min), respectively. The optimum time for a batch of catalytic reaction without exogenous pH control was 2 h. Under this reaction time, RC-CBD-GAD had a good reusability with a half-life of 23 cycles, indicating that it was very attractive for GABA industry. As a novel, efficient, and green CBD binding carrier, RC provides an alternative way to protein immobilization.

This is a preview of subscription content, access via your institution.

Materials and methods


Gecarcoidea natalis (Pocock 1888) and Discoplax hirtipes(Dana 1851) were collected from rainforest in the Australian Territory of Christmas Island, Indian Ocean and airfreighted to Sydney where they were maintained at 25°C on a 12 h:12 h light:dark cycle. Tapwater was provided for drinking and the crabs were fed with fallen leaves of Ficus macrophylla Desf. ex Pers. subspecies macrophylla.


Digestive juice used in measurements was taken from the foreguts of experimental animals as follows. The crabs were held ventral side up on a polystyrene board, and a fine polythene tube was inserted into the cardiac stomach via the mouth and oesophagus. A small plastic wedge was used to prevent the mandibles from cutting the tube. Up to 2 ml of dark brown digestive juice could be collected by gentle suction with a 2-ml syringe attached to the tubing. The procedure did not harm the crabs. Fluid was centrifuged at 10 000 g for 5 (min, to remove food debris, and the supernatant was used for analyses. Fluid could be stored at 4°C for several days without loss of enzyme activity.

Measurement of enzyme activities

Cellulase activities

Total cellulase activity and activities of β-1,4-glucosidase(cellobiase EC and EG (EC were measured in digestive juice taken from the cardiac stomach of Gecarcoidea natalis and Discoplax hirtipes using modified versions of the methods of Schulz et al. (1986) and Hogan et al.(1988). Reactions and incubations were carried out at 40°C in 1.5-ml Eppendorf centrifuge tubes in an Eppendorf thermomixer. Measurement at 40°C allowed direct comparison with data for cellulase activities of other invertebrates. Absorption values of the samples were measured using an LKB Ultraspec II spectrophotometer. Activities of the enzymes are presented per ml of digestive juice. Expression per mg of protein is not meaningful in this situation where crude juice is used since the bulk of the protein does not represent the enzyme of interest,is highly variable and may even be dietary in origin. It is likely that the volume of fluid in the foregut remains relatively constant. Hanes plots derived from the data on cellulase activity at different substrate concentrations were used to determine if enzyme activity followed Michaelis-Menten kinetics. Where this was established, the kinetic parameters(Km, Vmax) were then calculated from the plots.

β-1,4-glucosidase. Activity was measured as the rate of production of glucose from cellobiose (Cat. No. C-7252 Sigma Chemical Corp.,St Louis, MO, USA). Digestive juice (25 μl) was mixed with 25 μl of 0.1 mol l -1 acetate buffer (pH 5.5) and 50 μl of either 2.92, 14.61,29.21 or 58.4 mmol l -1 cellobiose in the same buffer, and the mixture was incubated at 40°C for 30 min. The reaction was stopped by the addition of 25 μl of 0.3 mol l -1 tri-chloro acetic acid, and excess acid was neutralized with 5 μl of 2.5 mol l -1 K2CO3. Precipitated protein was pelleted by centrifugation at 10 000 g for 10 min. A blank (75 μl buffer plus 25 μl digestive juice) and a standard (50 μl of 7 mol l -1 glucose in buffer + 25 μl digestive juice + 25 μl buffer)were prepared for each sample analysed. This enabled correction for the background absorption due to the digestive juice at the wavelength measured.

Glucose concentration was measured in 50 μl (G. natalis) or 25μl (D. hirtipes) samples of the incubation mixture using a commercial glucose assay kit (Sigma Cat No. 510-A). The 50 μl or 25 μl samples were diluted to a total of 100 μl with water in a 1.5-ml micro test tube. 1 ml of the colour reagent supplied with the kit was then added, and the mixture was vortexed and incubated at 37°C for 30 min. After incubation,the absorbance of the samples was read at 445 nm.

Glucono- d -lactone competitively inhibits β-1,4-glucosidase(Scrivener and Slaytor, 1994 Shewale and Sadana, 1981 Santos and Terra, 1985). The inhibitory constant of glucono- d -lactone on β-1,4-glucosidase was also determined by measuring β-1,4-glucosidase activity in the presence of 20 mmol l -1 glucono- d -lactone and 0–58.43 mmol l -1 cellobiose.

Endo-β-1,4-glucanase. Activity was measured as the rate of production of reducing sugars from the substrate, carboxymethyl cellulose (Sigma Cat. No. C-5678). Digestive juice (20 μl) was mixed with 80 μl buffer and 100 μl of 0.1, 0.5, 1 and 2% (w/v) carboxymethyl cellulose in the same buffer. Blanks contained 20 μl of digestive juice and 180 μl of buffer, while standards contained 20 μl of digestive juice plus 100 μl glucose (13 mmol l -1 ) in buffer and 80 μl buffer. The buffer was 0.1 mol l -1 acetate buffer, pH 5.5, containing 30 mmol l -1 of the β-1,4-glucosidase inhibitor glucono- d -lactone. Samples, standards and blanks were incubated at 40°C for 10 min and the reaction stopped by the addition of 25 μl of 0.3 mol l -1 HCl. Excess acid was then neutralized by the addition of 5 μl of 2.5 mol l -1 K2CO3. The reducing sugar produced during the incubation was measured as glucose equivalents by the tetrazolium blue method of Jue and Lipke(1985) using 5 mmol l -1 glucose as a standard. Absorption of samples, standards and blanks was read at 660 nm.

Total cellulase activity. Total cellulase activity was measured as the rate of production of glucose from microcrystalline cellulose (Sigmacell 20). Digestive juice (50 μl) was mixed with 100 μl of either 0.1, 0.5, 1 or 2% (w/v) Sigmacell 20 (Sigma Cat. No. S3504) made up in buffer. Suspension of the cellulose was ensured by vortexing the stock cellulose immediately before pipetting. Blanks contained digestive juice and buffer while standards contained digestive juice, buffer and 7 mmol l -1 glucose. The buffer used was 0.1 mol l -1 acetate, pH 5.5. The mixture was incubated and agitated for 60 min at 40°C in an Eppendorf thermomixer before the reaction was stopped by the addition of 25 μl of 0.3 mol l -1 tri-chloro acetic acid. The excess acid was neutralized with 5μl of 2.5 mol l -1 K2CO3 before assay of glucose. The incubation mixture was centrifuged (10 000 g for 10 min) and the glucose concentration determined in a 25 μl or 50 μl aliquot of the supernatant as described for β-1,4-glucosidase.

The Km values for EG and CBH are given as mg substrate ml -1 reaction mixture since the substrates (carboxymethyl cellulose and cellulose) consist of carbohydrate polymers of varying length and do not have a uniform molecular mass.

Hemicellulase activities

Activities of the hemicellulase enzymes laminarinase[endo-β-1,3-glucanase (EC], licheninase [endo-β-1,3 1,4 glucanase (EC], xylanase [endo-β-1,4-xylanase (EC and 1,4-β- d -xylan xylanhydrolase (EC] were measured in the digestive juice from D. hirtipes and G. natalis.

Laminarinase. Laminarinase activity was measured as the production of reducing sugars from the hydrolysis of laminarin (from Laminaria digitata Sigma Cat. No. L-9634). Digestive juice (20 μl) was mixed with 50 μl of 1% (w/v) laminarin and 130 μl of 0.1 mol l -1 Na acetate buffer, pH 5.5. Blanks and standards were run at the same time. Blanks consisted of 20 μl of digestive juice and 180 μl of assay buffer, while standards consisted of 20 μl of digestive juice, 100 μl of 13 mmol l -1 glucose and 80 μl of assay buffer. Samples, blanks and standards were incubated with agitation at 40°C for 10 min. The reaction was stopped by the addition of 50 μl of 0.3 mol l -1 HCl and neutralized with 10 μl of 2.5 mol l -1 K2CO3. Reducing sugars were measured in a 10 μl aliquot as described above.

Licheninase. The activity of licheninase was measured by the production of reducing sugars from the hydrolysis of lichenin (from Cetraria islandica Sigma Cat. No. L-6133). Digestive juice (20μl) was mixed with 100 μl of 0.1% (w/v) lichenin and 80 μl of 0.1 mol l -1 Na acetate buffer, pH 5.5. To correct for the background absorbance of the digestive juice, blanks and standards were run at the same time. Digestive juice (20 μl) and 180 μl of assay buffer constituted the blank while 20 μl of digestive juice plus 100 μl of 13 mmol l -1 glucose and 80 μl of assay buffer constituted the standard. Samples, blanks and standards were incubated with agitation at 40°C for 10 min. The reaction was stopped with 50 μl of 0.3 mol l -1 HCl and the mixture was then neutralized with 10 μl of 2.5 mol l -1 K2CO3. Reducing sugars were measured in a 10 μl sample as described above.

Xylanase. Xylanase activity was measured as the production of reducing sugars from the hydrolysis of xylan (from birchwood, BetulaSigma Cat. No. X-0502). Digestive juice (20 μl) was incubated with 100μl of 1% (w/v) xylan and 80 μl of 0.1 mol l -1 Na acetate buffer, pH 5.5. Blanks (20 μl of digestive juice and 180 μl of buffer)and standards (20 μl of digestive juice, 100 μl of 13 mmol l -1 glucose and 80 μl of assay buffer) were run at the same time. Samples, blanks and standards were incubated with agitation at 40°C for 60 min. After this period, the reaction was stopped by precipitating the protein with 50 μl of 0.3 mol l -1 HCl and was neutralized by the addition of 10 μl of 2.5 mol l -1 K2CO3. Reducing sugars were measured in a 10 μl sample of this reaction mixture as described above.

The activities of β-1,4-glucosidase, EG and total cellulase were measured over a pH range of 4–9. Acetate buffer was used to maintain pH values of 4 and 5.5, and Tris buffer was used for the higher pH values of 7–9. The pH of gut fluid was measured anaerobically at 25°C using freshly drawn samples of fluid from the foregut and a Radiometer G299a capillary pH electrode and Radiometer PHM 73 meter (Radiometer, Copenhagen,Denmark).


The concentrations of protein in samples of digestive juice were measured using a BioRad protein assay kit and bovine γ-globulin standard (BioRad,Hercules, CA, USA).


Statistical comparisons (ANCOVA and one- and two-way ANOVA with Tukey's HSD post hoc tests) were made using the statistical computing package Systat 7 (Systat Software Inc., Richmond, CA, USA) to calculate the statistical probabilities.


Biofuels represent one of many important renewable energy alternatives to fossil fuels with the potential to decrease anthropogenic effects on climate change. Cellulosic biofuels derive energy from chemical bonds stored by plants in the form of cellulose, a polymer of glucose, and a primary structural component of plant cell walls (Somerville et al., 2004 ). Cellulose-rich biomass can be produced with fewer inputs than first-generation biofuel crops, such as starch-rich corn however, cellulose is difficult to break down (Dinan, 2014 ). Once cellulose is depolymerized into glucose, the sugar can be microbially or chemically transformed into fuels and chemicals such as ethanol, butanol, or other gasoline, jet fuel, and diesel alternatives.

While cellulose is abundant, accessing the sugar within is challenging. To biochemically degrade biomass, several enzymes work in concert, the most abundant of which is cellulase enzyme Cel7A (Payne et al., 2015 ). Cel7A cellobiohydrolase enzymes depolymerize cellulose into its fundamental repeating unit of two glucose molecules—cellobiose. These enzymes suffer from inhibition by this product, which lingers in the enzymes’ active sites and thus delays their catalytic cycles (Silveira and Skaf, 2015 ). Cellobiose accumulates over the course of a reaction unless removed by an enzyme such as β-glucosidase, which cleaves cellobiose yielding two glucose molecules (Payne et al., 2015 ). Cel7A experiences mixed inhibition by cellobiose the molecule can both competitively compete with a cellulose chain for binding in the substrate-binding sites as well as noncompetitively inhibit the enzyme by retarding processive motion as a result of persisting in the product-binding site (Jalak et al., 2012 Payne et al., 2015 ). Measurements on crystalline cellulose show that Cel7A loses half of its activity in the presence of cellobiose concentrations on the order of 2.6 mM (Teugjas and Väljamäe, 2013 ) to 19 mM (Murphy et al., 2013 ). Product inhibition is particularly nefarious in the enzymatic conversion of lignocellulosic biomass to glucose under the high substrate loadings required for commercial manufacture of biofuels, and represents a barrier to achieving the high product yields necessary for an efficient process (Bu et al., 2012 Payne et al., 2015 ). Unfortunately, addressing this issue with the product inhibition-relieving enzyme β-glucosidase alone is not a comprehensive solution. Beta-glucosidase activity is limited by its own product inhibition from glucose, as well as by gluconic acid (generated by lytic polysaccharide monooxygenase (LPMO) activity) (Payne et al., 2015 ).

Enzymes are one of the most expensive components of a biochemical cellulosic biofuels process (Klein-Marcuschamer et al., 2012 ). Therefore, improving the efficiency of Cel7A by ameliorating product inhibition may result in a lower enzyme requirement for the process and consequently a cheaper renewable fuel that is more cost-competitive with fossil fuels. To this end, several research groups have investigated ways to make Cel7A enzymes less prone to cellobiose inhibition. The prevailing strategy to mitigate product inhibition has been to perturb the binding of cellobiose in the enzyme active site via site-directed mutagenesis of the residues most responsible for this interaction (Hanson et al., 2014 Payne et al., 2015 Silveira and Skaf, 2015 ). Mutations in Trichoderma reesei (Hypocrea jecorina) Cel7A (TrCel7A) residues R251 and R394 reportedly resulted in reduced product inhibition (Hanson et al., 2014 ). A quintuple TrCel7A mutant (E223S/A224H/L225V/T226A/D262G) designed in 2001 to alter the pH optimum was similarly found to both relieve cellobiose inhibition and diminish overall cellulase activity (Becker et al., 2001 ). More recently, computational point mutation studies in the same enzyme found that mutating residues R251, D259, D262, W376, or Y381 to alanine significantly weakened the calculated binding of cellobiose in the enzyme (Bu et al., 2011 Payne et al., 2015 ). However, for many of these residues, no experimental evidence verifying this has ever been demonstrated.

Recent molecular dynamics (MD) simulations performed by Silveira and Skaf computationally investigated the effects of various Cel7A mutations on cellobiose binding, as well as any induced structural perturbations to the enzyme (Silveira and Skaf, 2015 ). These simulations built upon previous calculations (Bu et al., 2011 ) and together point to a handful of mutations predicted to disrupt cellobiose binding affinity (Silveira and Skaf, 2015 ). The aim of our study, was to produce a Cel7A variant with reduced cellobiose inhibition. We experimentally generated mutants identified by MD simulations (Silveira and Skaf, 2015 ) and evaluated their ability to hydrolyze microcrystalline cellulose in the presence of cellobiose.


Normal plant development depends critically on the interactions between different components of the plant cell wall. This dynamic structure defines the plant morphological architecture and is responsible for cell shape, cell adhesion and organ cohesion [1]. Plant cell walls are initiated by the synthesis, secretion, modification and crosslinking of individual wall components– cellulose, hemicellulose, pectin and hydroxyproline-rich glycoproteins- and are synthesized by the coordinated action of a myriad of glycosyltransferases. Understanding the underlying mechanisms involved in the assembly of a complex polysaccharide network and elucidating their biological roles is not a trivial task [2], and remains to date a key goal for scientists interested in the manipulation of plant cell wall structure to better understand its physiological functions and allow for its commercial exploitation.

One model system that is gaining increasing recognition and significance for the study of cell wall polysaccharide interactions is the Arabidopsis seed coat epidermis (SCE), also referred to as Mucilage Secretary Cells (MSC) [1]. The SCE is an excellent model system for understanding the genetic basis of cell wall biosynthesis, secretion, assembly and modification [3, 4] because large amounts of cell wall polysaccharides can be extracted with ease and analyzed in a short timeframe. Between 5- and 8-days post anthesis (DPA), large amounts of pectins are secreted to the apoplastic space at the junction of the outer tangential and radial primary walls, forming a donut-shaped pocket of mucilage around a cytoplasmic column [4]. The epidermal cells then synthesize a volcano-shaped secondary wall (9 to 11 DPA) called the columella, which protrudes through the center of the mucilage pocket and connects to the primary wall. When dry, mature seeds imbibe water, rapid mucilage expansion ruptures the tangential SCE to release the polysaccharide-rich mucilage that is organized in two distinct layers: an outer, water soluble non-adherent layer and an inner, adherent layer that remains tightly attached to the seed coat surface. Arabidopsis mucilage is composed primarily of unbranched Rhamnogalacturonan-1 (RG-I), with small quantities of Homogalacturonan (HG), cellulose, and arabinoxylan found in the inner layer [3, 5]. Several attempts have been made to better understand the functional roles of the glycosyltransferases involved in cell wall biosynthesis, secretion, and delivery of the mucilage polymers through the analysis of mucilage mutants. In recent years, genetic mutants that lack functional enzymes required for mucilage biosynthesis and extrusion have been identified and characterized, but many others await functional investigation.

Work to date has identified several genes/proteins involved in mucilage biosynthesis, including a fasciclin-like arabinogalactan-protein (AGP) named SALT OVERLY SENSITIVE 5 (SOS5), GALT2 and GALT5, two galactosyltransferases responsible for initiating glycosylation of AGPs, and a receptor-like kinase called FEI2 individual as well as higher order mutants corresponding to these genes/proteins are characterized by mucilage pectin repartitioning and the marked absence of cellulosic rays, while the diffuse cellulose staining remains intact [6]. As SOS5, which is also known as FLA4, is the only well characterized AGP reported to be involved in mucilage biosynthesis, the contribution of the SOS5 glycan moieties and potentially other AGPs to mucilage formation is far from complete and presents an enigma worth unraveling.

AGPs are a family of hydroxyproline-rich glycoproteins that are extensively glycosylated with type II AGs that are covalently attached to hydroxyproline residues in the AGP protein backbone [7, 8]. An individual type II AG glycan consists of a β-1,3-galactan backbone with β -1,6-galactosyl branches that are decorated with arabinosyl residues and often with other minor sugar residues, such as glucuronic acid (GlcA), rhamnose (Rha), and Fuc [8, 9]. Although their exact roles in mucilage formation are still unclear, the interaction of AGPs with wall polysaccharides, their involvement in intracellular signaling cascades, and their influence on a wide variety of biological processes are known [8, 10]. Notably, the complexity of the cell wall polymer network with respect to AGPs is perhaps best illustrated by the finding that AGPs form covalent linkages to both RG- I and arabinoxylan [10].

Three glucuronosyltransferases (GLCATs), GLCAT14A, GLCAT14B and GLCAT14C were functionally characterized and found to transfer GlcA residues to AGPs [11], while two additional GLCATs (GLCAT14D and GLCAT14E) were also reported to be involved in the glucuronidation of AGPs [12]. Here, we present evidence that two GLCATs (GLCAT14A and GLCAT14C) belonging to family GT14 in the Carbohydrate-Active Enzymes (CAZy) classification system ( [13]) are critically important in mucilage matrix formation in Arabidopsis.

Some modification of cellulose nanocrystals for functional Pickering emulsions

Cellulose nanocrystals (CNCs) are negatively charged colloidal particles well known to form highly stable surfactant-free Pickering emulsions. These particles can vary in surface charge density depending on their preparation by acid hydrolysis or applying post-treatments. CNCs with three different surface charge densities were prepared corresponding to 0.08, 0.16 and 0.64 e nm −2 , respectively. Post-treatment might also increase the surface charge density. The well-known TEMPO-mediated oxidation substitutes C6-hydroxyl groups by C6-carboxyl groups on the surface. We report that these different modified CNCs lead to stable oil-in-water emulsions. TEMPO-oxidized CNC might be the basis of further modifications. It is shown that they can, for example, lead to hydrophobic CNCs with a simple method using quaternary ammonium salts that allow producing inverse water-in-oil emulsions. Different from CNC modification before emulsification, modification can be carried out on the droplets after emulsification. This way allows preparing functional capsules according to the layer-by-layer process. As a result, it is demonstrated here the large range of use of these biobased rod-like nanoparticles, extending therefore their potential use to highly sophisticated formulations.

This article is part of the themed issue ‘Soft interfacial materials: from fundamentals to formulation’.

1. Introduction

Emulsion stabilization is conventionally achieved by the addition of amphiphilic surfactants to reduce interfacial tension. However, it is now well established that surfactant-free dispersions can be stabilized by dispersed solid particles to form the so-called Pickering emulsions [1–4] for which colloidal-size particles may be irreversibly anchored at the oil–water interface. Pickering emulsions typically require an interfacial solid material that exhibits an affinity for the two phases of the emulsion [3,5] and present the double advantage of being extremely stable and requiring a very small quantity of particles. Research efforts are currently focused on the development of environmentally friendly renewable materials [6–8]. Cellulose constitutes the most abundant renewable polymer resource available today. Therefore, solid cellulosic particles with their low carbon footprint and low density are of particular interest for various applications such as composites, cosmetics, pharmaceutics or medical implants.

As a chemical raw material, cellulose has been used in the form of fibres or derivatives for nearly 150 years for a wide spectrum of products and materials in daily life. In the 1950s, it was demonstrated that cellulose fibres are composed of microfibrils that can be defibrillated and produce long semi-crystalline wires constituted of crystalline rod-like residues alternating with less ordered amorphous domains. The solid crystalline part may be recovered from preferential hydrolysis of the amorphous regions of cellulose fibres. This hydrolysis leads to highly crystalline solid rod-like particles known as cellulose nanocrystals (CNCs) [9,10]. In the materials science community, CNCs have reached a high level of attention that does not appear to be diminishing. These biopolymeric assemblies warrant such attention not only because of their inherent biodegradability, renewability and sustainability, in addition to their abundance, but also because of their impressive physical and chemical properties.

Nanocrystals of various dimensions can be obtained depending on the source. The most common sources include, among others, cellulose fibres from cotton, ramie, hemp, flax, hardwood and cotton linter pulp, microcrystalline cellulose, bacterial cellulose and tunicates [11–13]. The CNCs generally used are obtained from wood and cotton through sulfuric acid hydrolysis. This hydrolysis results in charged nanoparticles with a length of approximately 160–200 nm and a cross section of approximately 7–25 nm, making them natural anisotropic rods that can be used as platforms for various modifications. This platform can appear under various more or less sulfated forms depending on the hydrolysis condition. It may even be totally desulfated which leads to more aggregated systems as revealed by visual opacification of the suspension [14,15].

CNCs have recently been used as emulsion stabilizers by taking advantage of their self-assembling ability at the oil–water interface, similar to particle-stabilized Pickering systems [16,17]. They thus lead to the formation of ultrastable and monodispersed oil-in-water emulsions. Depending on the biological source, CNCs can form crystals with various aspect ratios, up to several micrometres long, but with a section that is always around 10–30 nm. They all present the ability to stabilize oil-in-water emulsions [13]. However, depending on the morphology and the quantity of nanocrystals involved, they can be used to prepare either individual droplets or three-dimensional networks with interconnected droplets, as well as emulsions of varying coverage, thereby modulating the porosity of the interface and visco-elasticity of the emulsion. CNC is a crystalline colloid it is then a perfectly ordered solid particle with well-defined faces. The stabilization at the interface is attributed to the less hydrophilic crystalline plane of the crystal. This plane is of minor importance in terms of surface area as it is located at the corner of the cellulosic crystals. Defined as (2 0 0) crystalline plane for cotton Iβ allomorph, it is considered flat, exposing at the surface a repetition of the CH groups of the glucosyl moieties without hydroxyl functions [17,18]. Neutron scattering experiments showed clearly that rigid nanoparticles can be densely adsorbed at the oil–water interface without deforming it at the nanoscale confirming the postulate that interactions involve only the crystal surface and oil, the particle residing in water only [19]. Above these dimension variations, CNCs are versatile solid platforms for significant chemical reaction by covalent and non-covalent surface modification as reported in several reviews [20,21].

This work focuses on the modulation of the surface chemistry of CNCs and their chemical transformations for emulsion processing. It shows that their amphiphilic character is maintained after various modifications. Several ways are illustrated for CNC stabilized emulsions providing versatile surfactant-free functional emulsions in order to control the droplet dispersion in formulations.

2. Experimental procedure

(a) Sample preparation

Sulfated cotton CNCs were prepared from Whatman filters (grade 20 Chr), based on a previous process using sulfuric acid hydrolysis at two concentrations (58 or 64%) and different conditions. The slightly sulfated CNC-S was hydrolysed with 58% H2SO4 and kept at 65°C under stirring for 20 min. The medium sulfated CNC-M and the highly sulfated CNC-H samples were hydrolysed with 58% and 64% H2SO4, respectively, and kept under stirring at 70°C for 20–30 min. After hydrolysis, the prepared suspensions were systematically washed by centrifugation, dialysed to neutrality against Milli-Q water, and deionized using mixed bed resin (TMD-8). The final dispersions were sonicated for 10 min (ultrasonic Qsonica Q700 Misonix, Inc., NY, USA), filtered and stored at 4°C.

TEMPO-mediated oxidation of CNCs was performed using 4-acetamido-2,2,6,6-tetramethyl-1-piperinidyloxy radical (TEMPO) as a catalyst, according to the method described by Saito et al. [22]. The CNCs were suspended in water containing TEMPO (0.075 mmol g −1 of cellulose) and sodium bromide (1.25 mmol g −1 of cellulose). The TEMPO oxidation was started by adding the desired amount of sodium hypochlorite solution (3.6 mmol g −1 of cellulose) and was continued at room temperature while stirring. The pH was kept at 10 by adding sodium hydroxide. The TEMPO-oxidized CNCs were subsequently thoroughly washed with water by centrifugation, dialysis and filtration.

Hydrophobic surface modification of CNCs was done with quaternary ammonium salts. The pH of the CNC suspension (0.1 wt%) was adjusted to 10 using NaOH aqueous solution in order to have the carboxyl groups on the surface of the CNCs fully dissociated. The CNC suspension was then added dropwise into desired amount of stearyltrimethylammonium chloride solution (0.1 wt%). The suspension was kept at 60°C for 3 h, and stirred at room temperature overnight. The suspension was then dialysed against Milli-Q water to remove NaCl formed during the adsorption and unbound quaternary ammonium salts. The final suspension was freeze dried and redispersed in toluene using an ultrasonic treatment for 1 min. The suspensions were then centrifuged for 10 min at 20 000g to remove possible aggregated excess of quaternary ammonium salts. The resulting pellet was easily redispersed in the desired amount of organic solvent with an ultrasonic treatment.

(b) Characterization of the cellulose nanocrystals

Conductometric titrations were performed to determine the surface charge density according to Kalashnikova et al. [17] with a few modifications. Titration was performed on a total of 10 ml of a cellulose suspension at 10 g l −1 in water for CNC sulfated and neutrals, and in 5 mM HCl for CNC-TEMPO, with freshly prepared NaOH. Sulfate groups being only positioned at the surface, results in millimole per gram were calculated from dimensions obtained by microscopies and given as an average amount of elementary charge per square nanometre (e nm −2 ).

Dimensions were determined by image analysis. For transmission electron microscopy (TEM), the suspensions were sonicated just before deposition on a substrate of a freshly glow-discharged carbon-coated copper grid and negatively stained with phosphotungstic acid at 1%w/v adjusted to pH 6. The grids were observed with a Jeol JEM 1230 TEM at 80 kV. For atomic force microscopy (AFM), a drop of CNC suspension freshly sonicated and filtered at 0.1 g l −1 was deposited on freshly cleaved mica covered with a positive polyelectrolyte (poly(allylamine hydrochloride), PAH). The sample dried under ambient conditions was analysed using tapping mode AFM (Innova AFM, Bruker, Santa Barbara, CA, USA).

Transmission wide-angle X-ray scattering (WAXS) diffractograms of samples lyophilized and dried in a desiccator were recorded with a Bruker D8 Discover diffractometer (Madison, WI, USA) using Cu Kα1 radiation (λCu Kα1=1.5405 Å) produced by a sealed tube at 40 kV and 40 mA. The average dimension of the crystal perpendicular to the diffracting planes with hkl Miller indices, Dhkl, was evaluated using Scherrer’s expression from the width at half-height of intensity.

(c) Emulsion preparation

Oil-in-water (o/w) emulsions were prepared using hexadecane as oil phase and a 20/80 oil/aqueous phase ratio. Practically, emulsions were sonicated with an ultrasonic device with a dipping titanium probe close to the surface (pulsed 2 W ml −1 applied power). The emulsions were all visualized after dilution by light microscopy (BX51 Olympus). Average droplet diameter was measured by image analysis using I mage J software and compared with the drop size distribution determined by laser light diffraction using a Malvern 2000 granulometer equipped with a HeNe laser (Malvern Instruments, UK). The diameter was expressed as surface mean diameter D(3,2) (the Sauter diameter). Similar results were obtained via granulometer and I mage J analysis. Inverse water-in-oil (w/o) emulsions were prepared at a 20/80 water/hexadecane ratio mixed using an ultrasonic device.

Scanning electron microscopy (SEM) images were prepared as previously described [13] from 20/80 styrene/water emulsions obtained by sonication and degassed with nitrogen before polymerization. The resulting beads were washed by repeated centrifugation to reduce the amount of artefactual small droplets appearing during curing. Dried beads were metalized with platinum and visualized with a JEOL 6400F instrument (IMN-Nantes).

3. Results and discussion

(a) Cellulose nanocrystals modulated in surface charge density

When mixing an aqueous suspension of colloidal CNCs and an oil phase, an ultrastable Pickering emulsion is produced. CNCs are known to serve as platform for several modifications but it is not possible to predict whether or not they might still allow preparing stable emulsions once modified. Basically, CNCs are obtained by acid hydrolysis using sulfuric acid. According to the hydrolysis parameters (mainly concentration in acid, reaction temperature and time), various degrees of substitution might be reached where the hydroxyl groups exposed at the surface are replaced by sulfate ester groups. The surface-modified samples are generally characterized by conductometric titration for the resulting surface charge density. Samples with different sulfate charge densities were prepared and characterized. As a result, three samples were prepared, a slightly sulfated one (CNC-S), one with medium substitution degree (CNC-M) and a third one highly substituted (CNC-H). They were characterized for their surface charge, size and crystallinity (table 1). Some aggregations were observed for CNC-S as illustrated in figure 1, but no clear size variation was noticed for the different samples. The three CNC samples had length of 156±53 nm and width of 16±8 nm giving a similar aspect ratio around 10 (mainly from 7 to 14) whatever the acidic treatment. Since the sulfate charges are known to be responsible for crystal destructuration, these post-hydrolysis treatments were followed by WAXS to evaluate an eventual variation at a crystal level. WAXS diagrams showed a high crystallinity of 85±3% for the three samples. As a result, the various treatments maintained the same crystalline organization.

Table 1. Characteristics of CNCs with three surface charge densities. Crystalline plane dimensions and crystallinity for the elementary unit as determined by WAXS, according to the Miller index maximum and the dimensions of the nanoparticles as determined by AFM.

Figure 1. TEM images of (a) slightly, (b) medium and (c) highly charged CNCs.

These three CNCs were tested for their ability to stabilize emulsions. As a result, in pure water, repulsion between the charged CNC prevented a sufficient surface density of CNC to stabilize the oil–water emulsions instead coalescence occurs and the oil and water phases separate. However adding salt, the different samples were all able to efficiently stabilize o/w emulsions with same drop diameters for more than a year (figure 2). The diameter decreased with mp at low CNC concentration in accordance with the limited coalescence process. The drop diameter is then controlled by the amount of CNC available to stabilize the interface. This domain is followed by stabilization of the diameter values at higher mp. When the same values are plotted as 1/D versus mp, a linear behaviour appears at low mp and a deviation arises at a critical concentration of 7 mg ml −1 of hexadecane [19]. This change indicates a coverage variation that is possibly due to the limited flexibility of the particles hindering the decrease in radius of curvature and thereby in drop diameter. It results in a denser coverage of the drops.

Figure 2. Mean D[3,2] Sauter diameter with the concentration of particles given in milligrams of CNC per millilitre of hexadecane, the dispersed phase, for the three charged CNCs at a 30/70 hexadecane/water ratio.

Styrene-in-water emulsions with 0.05 M NaCl were prepared with CNCs bearing various surface charge densities and polymerized. SEM images of the resulting beads are shown in figure 3. Some artefactual small droplets arising during polymerization are visible. However, the average size and size distribution of the larger beads is similar to the hexadecane-in-water emulsions allowing comparison. These images confirmed the identical organization at the surface of the individual droplets, showing the CNC curved along the droplets creating a dense armoured layer in all cases.

Figure 3. Scanning electron microscope images of polymerized PS/W Pickering emulsion droplets stabilized by slightly sulfated CNC-S (a,d), medium sulfated CNC-M (b,e) and highly sulfated CNC-H (c,f).

Based on the previous WAXS results, four characteristic planes of cellulose I, namely (1-10), (110), (102/012) and (200) [11,17,23] were identified. We proposed that the process of stabilization involves this crystalline regular organization of CNCs forming a faceted surface. The (2 0 0) edge planes do not bear charges since they are only composed of CH groups, maintaining the more hydrophobic surface available for interface stabilization [18]. It indicates similar exposure of the crystalline planes, regardless of the charge density. The present results illustrate the low susceptibility of surface charge modulation of CNCs on their ability to adsorb at the interface. It differs from the tendency to aggregation generally observed at low surface charge density such as that illustrated in figure 1, important aggregation seems not to occur on a two-dimensional organization just varying the amount of charges. It appears then possible to modify the surface chemistry of CNCs and subsequently of droplets, without changing any other interfacial parameter.

(b) Surface modification of cellulose nanocrystal for functional droplets

These CNCs might then further be used as a platform for subsequent modifications. As previously described, the sulfuric acid used for the preparation results in sulfate moieties at the surface of the CNCs. Hydroxyl groups have low reactivity, and sulfate groups might be used for chemical modification but they are unfortunately rather labile, being, in particular, readily removed under mild alkaline conditions. Oxidation of the CNCs, whether sulfated or not, with TEMPO results in high anionic surface charges with high densities (1.7 glucose units nm −2 ) arising from dissociated C6-carboxyl groups formed on the surface [24,25]. The use of this technique has been the subject of a number of studies since the first reports of De Nooy et al. in the 1990s, who showed that only the hydroxymethyl groups of polysaccharides were oxidized, whereas the secondary hydroxyls remained unaffected (figure 4).

Figure 4. Schematic of a two-step modification of CNCs.

When submitting the TEMPO-mediated oxidized CNC with an oil phase to emulsification, it appeared clearly that such carboxylated CNC kept stabilizing efficiently oil–water interfaces. Figure 5 shows an example of emulsions stabilized with both unmodified and TEMPO-oxidized CNCs at 5 g l −1 in the aqueous phase in the presence of NaCl to reduce repulsions between the CNCs, with a 20/80 (oil/water) ratio. The average diameter increased from 5 μm for sulfated CNCs to 15 μm using TEMPO-oxidized CNCs. Such modified CNCs are not aggregated due to the highly charged surface, the reason is then attributed to unadsorbed CNCs maintained in suspension in the continuous phase. However, a stable emulsion was prepared. TEMPO oxidation improves chemical modification of cellulose through the presence of carboxylate groups. This opportunity to produce highly functionalized droplets reveals that a much larger range of functional emulsions might be reached as one could be inspired by examples of modification described in several issues dealing with more or less sophisticated modifications [21,26,27].

Figure 5. Compared Sauter diameter distribution, optical microscopy images and photographs of 80/20 hexadecane/water (v/v) emulsions stabilized by unmodified and TEMPO-oxidized CNCs.

A major relevant point is that CNC is a hydrophilic particle. Equivalently to the Bancroft rule for surfactant-stabilized emulsions, the continuous phase for Pickering emulsions is the one in which the particles are preferentially dispersed. This means that CNCs are able to stabilize direct o/w emulsions while predominantly hydrophobic particles should be used to stabilize reverse w/o emulsions allowing compatibility with apolar organic media. Several studies have developed such modifications [28–30] mainly solvent based or using harmful systems, which limit the extension of these finding to industrial scale-up. Efforts to reach an environmentally friendly procedure are needed. A surface modification performed in aqueous solution is more acceptable. For instance, modification using quaternary ammonium salts bearing long alkyl, phenyl, glycidyl and diallyl groups via adsorption was developed based on TEMPO-oxidized CNCs [31]. In this study, hydrophobically modified CNCs were prepared by exchanging the counterions of the Na carboxylate groups of TEMPO-oxidized CNCs with quaternary alkylammonium groups. The successful ionic exchange was proved (i) by the presence of a new peak at 1480 cm −1 in Fourier transform infrared spectra (not shown) corresponding to the trimethyl groups of the quaternary ammonium and (ii) by the dispersability of the modified CNCs in toluene and hexadecane. As illustrated in figure 4, stearyltrimethylammonium chloride was used to graft alkyl chains on CNCs bearing carboxylic acid groups on the surface. The w/o emulsions were prepared using Milli-Q water and 2–4 g l −1 modified CNC dispersed in hexadecane. It resulted in stable inverse emulsions (figure 6) showing that simple hydrophobic modification enables the formation of inverse w/o emulsions.

Figure 6. Water-in-hexadecane emulsion stabilized hydrophobically by CNCs modified by quaternary alkylammonium groups. (Online version in colour.)

As a result, CNC proves to be a versatile nanorod with a large range of various surface modifications. The various modifications obtained here indifferently from native CNC with various surface charge densities, after TEMPO oxidation leading to a highly carboxylated CNC, or hydrophobically modified using acceptable processes, can be used to produce w/o or o/w emulsions and various kinds of functional droplets.

(c) Multilayer drop preparation

It appears then of interest to check if such emulsion would be preserved if modification is carried out directly on the droplet. Coming back to the pristine unmodified sulfated CNCs, negative charges are exposed to the surrounding area. Because these emulsions are highly stable, they can be dropped in a different solution and recovered by centrifugation. Another way of modifying the chemistry consists then of modulating the surface using electrostatic interactions as used in layer-by-layer systems. The change from a negative to a positive surface was carried out by mixing the emulsion with a positively charged polymer, in order to add an extra layer using PAH. The strong association between these two oppositely charged polymers results in the entropy gain resulting from the release of counterions and water molecules. As a result, it is possible to rinse the emulsion by simple repeated centrifugations and keep an isolated positively charged bead. Figure 7 shows SEM images of droplets with a negative monolayer of sulfated CNCs, before (figure 7a) and after coating with PAH (figure 7b). These images revealed a very different surface with jammed CNCs, confirming that PAH was deposited on the surface.

Figure 7. SEM images of polymerized PS/W emulsions stabilized by sulfated CNC (a,d), after addition of a hydrosoluble positive polyelectrolyte (PAH) (b,e), and after a second deposition of CNC following a layer-by-layer process (c,f).

CNCs appear then of great interest as a rigid armour to stabilize droplets that might be functionalized at the surface on demand with non-surface-active agents. It is possible to change the surface chemistry in a second step with multi-layered systems. To check the modification of the surface chemistry and validate the ability to make multi-layered shells, an additional negatively charged layer of alginate previously stained with fluorescent die was used. As shown on the optical microscope with fluorescence detection (figure 8), when the sample is illuminated with a fluorescent light, stained alginate is revealed on the surface of the droplets.

Figure 8. Optical fluorescence microscopy after deposition of a PAH/stained alginate bilayer evidencing a multi-layered shell.

The surface of the capsules was reinforced by adding another layer of CNC. Figure 7c shows that CNCs are again appearing clearly on SEM images and the absence of visible droplets by optical fluorescence microscopy showed that the negatively charged alginate did not associate anymore.

These results reveal that such Pickering emulsions constitute good substrates for capsule preparation. It means that taking into account the high stability of CNCs at the oil–water interface, not only modified CNCs can be used to produce functional emulsions, but also post-modifications can be carried out for surface adapted to the required formulation revealing the high versatility of these biobased Pickering emulsions. CNCs appear then as highly relevant particles to process a large range of highly stable surfactant-free functional emulsions.

4. Conclusion

CNCs are highly versatile nanorods able to stabilize a large range of long-term Pickering emulsions. They appear as platforms for preparation of biobased functional colloidal particles. It is shown that besides the surface charge density variation, a lot of different modifications such as carboxylation might be carried out maintaining the ability of CNCs to stabilize oil–water interfaces. Hydrophobically modified CNCs allow stabilizing w/o emulsions that might also improve the processability and performances of nanocellulose-based materials in apolar media. It is finally illustrated that the high stability of such CNC stabilized emulsion also allows modification of preformed emulsions with an extra layer leading to capsules according to a layer-by-layer process. The different routes described in the present article (figure 9) aim at paving the way for innovative complex formulations and materials in very different targeted applications.

Figure 9. Schematic of the various emulsions based on pristine or modified CNC. (Online version in colour.)

Competing interests

We declare we have no competing interests.


D.S. gratefully acknowledges the financial support of INRA and INRA Transfert (Verniscell project). This work has been also funded by the local council programme MATIERES and is a contribution to the Labex Serenade (no. ANR-11-LABX-0064) funded by the ‘Investissements d’Avenir’ French Government programme of the French National Research Agency (ANR) through the A*MIDEX project (no. ANR-11-IDEX-0001-02).