Nanoscience and nanotechnologies for biobased materials, packaging and food applications: New opportunities and concerns

In the past decades, research into nanoscience and nanotechnologies has grown explosively and stimulated a large panel of scientific and technological fields. The boosting effect comes from either the reinvestigation of scientific fields by considering the nanoscale as a relevant level for improving our knowledge or from the extraordinary development of new tools that have democratized access to the nanoscale, such as AFM which is now a routine tool in most labs. This stimulating research has also reached the field of food science and biobased products. The INRA’s Science for Food & Bioproducts Engineering division (‘CEPIA’) has been engaged in this huge challenge, and selected results are presented here. ACCEPTED MANUSCRIPT


A C C E P T E D M A N U S C R I P T
4 renewable materials that are raw components for both food and non-food applications. Plants cell walls thus offer amazing case studies of the importance of nanoscale in biological functions. For instance, the mechanical properties of wood are intimately related to cellulose microfibril orientation at nanometric scale (Salmen 2004). Sap transport and cavitation resistances are linked to nanoscale vessel architecture and the size of the torus present in the pits (Cochard 2006;Lens, Tixier et al. 2013). Nanostructured surfaces of biopolymers thus provide simple, dense and anisotropic systems that can be used as tunable models of the interface and surface to mimic naturally-existing materials and investigate both fundamental hypotheses of plant cell wall biology such enzymatic degradation processes (Josefsson, Henriksson et al. 2007;Ahola, Turon et al. 2008;Cerclier, Guyomard-Lack et al. 2013) or polymer interactions (Holmberg, Berg et al. 1997;Stiernstedt, Nordgren et al. 2006;Winter, Cerclier et al. 2010;Villares, Moreau et al. 2015). Nanostructured surfaces also serve to investigate complex industrial processes such as pulp production or paper coatings through a simplified approach. Model surfaces or thin films can then be studied with various powerful analytical tools offered by surface science that cannot be applied to real substrates such as plant cell wall or manufactured products. Nanostructured biopolymer surfaces and thin films are also functional materials per se, and the fine control of their architectures can afford amazing functional properties. For instance, thanks to their nanometric dimensions lower than the wavelength of light, nanostructured surfaces and thin films can display optical properties such antireflectivity or structural colors (Cathell and Schauer 2007;Cerclier, Guyomard-Lack et al. 2011;Azzam, Moreau et al. 2014;Azzam, Moreau et al. 2015). Finally, as biopolymers are biological molecules, they can be expected to possess characteristic features such as specific recognition, sensitivity to stimuli or susceptibility to enzymes. Enzymes can thus be used as modifying agents for patterning, or in reverse, biopolymer thin films can be used as simplified systems

Preparation of biobased nanostructured surfaces or thin films
The preparation of nanostructured surfaces of biopolymers generally uses the three core methods employed for polymer surface or thin film fabrication. The first is the so-called " Langmuir-Blodgett" method-named from the inventors-which is a long-standing well-established method for the preparation of thin films/surfaces (Langmuir 1920;Blodgett 1934;Blodgett 1935). The first step consists in dissolving the polymer in a suitable solvent then spreading the solution on the surface of a nonsolvent liquid (usually water). In the second step, after evaporating the dissolving solvent and compressing the molecules on the surface, the film is transferred from the liquid/air interface onto a solid surface by moving a solid substrate upward or downward through the surface while keeping the surface pressure constant by moving the barrier in an appropriate manner. This approach has been applied for instance in cellulose nanocrystal (CNC) thin film fabrication (Habibi, Foulon et al. 2007) where negatively-charged, rod-like CNCs formed stable layers at the air-water interface in the presence of a cationic amphiphilic molecule. The layers formed can be efficiently transferred onto silicon wafers to afford well-structured surfaces with low roughness.
The second widespread method for preparing polymer surfaces or thin films is the spin-coating process.
In this method, a polymer solution is deposited on a solid substrate and the solvent is removed with the aid of high-speed spinning to afford uniform polymer films. Film properties can easily be tailored by adjusting experimental parameters such as spin velocity, acceleration or solution viscosity. This method has been applied for instance to prepare cellulose model surfaces from both dissolved cellulose or dispersed nanocellulose (Kontturi, Thune et al. 2005;Cranston and Gray 2006;Kontturi, Johansson et al. 2007; Winter, Cerclier et al. 2010).
The latest popular method is layer-by-layer (LbL) self-assembly originally developed for the preparation of polymer-based nanostructured surfaces (Decher 1997). It consists in the alternate deposition of polymers or particles presenting attractive interactions. Electrostatically-driven polymer assemblies,

A C C E P T E D M A N U S C R I P T
6 including synthetic and/or biopolymers, are the most frequent assemblies found in the LbL literature, but there are also reports of a buildup process through non-electrostatic interactions (Picart, Lavalle et al. 2001;Boddohi, Killingsworth et al. 2008;Olszewska, Kontturi et al. 2013). Using this buildup strategy, the elaboration of multilayered structures based on nanocellulose such as CNC or cellulose nanofibrils (CNF) has also been investigated, and changes of processing parameters can be tuned to obtain a large variety of structural and physical properties (Valentin, Cerclier et al. 2010;Moreau, Beury et al. 2012;Martin and Jean 2014;Decher and Schlenoff 2012;Borges and Mano 2014;Jean 2014;Richardson, Björnmalm et al. 2015;Moreau, Villares et al. 2016).
In the context of plant cell wall biopolymers, the fabrication of bioinspired assemblies that can mimic interactions and phenomena occurring in plant cell walls is a powerful approach for gaining insight into cell wall organization. Electrostatic-based LbL has been successfully applied for the elaboration of extensin/pectin thin films in order to investigate the force and quality of interactions between the two polymers (Valentin, Cerclier et al. 2010). Recently, Dammak et al. (Dammak, Moreau et al. 2013) also fabricated multilayered thin films based on electrostatic interactions from the alternate deposition of CNC and cationic xylan (CX) solutions that were synthesized from oat spelt xylans after reaction with epoxypropyltrimethylammonium chloride (Schwikal, Heinze et al. 2011) in order to have minimum cationic charge for adsorption to the CNC layer.
Biomimetic cellulose nanocrystals/xyloglucan assemblies have also been prepared by the LbL method (Jean, Heux et al. 2008;Winter, Cerclier et al. 2010). Multilayers are stabilized by the interaction between cellulose and xyloglucan, which is assumed to be mediated by hydrogen bonding, hydrophobic interactions and van der Waals dispersion forces, and deposition is achieved for up to eight bilayers.
Construction of the films can be only achieved when xyloglucan (XG) concentration is low enough to allow two cellulose nanocrystal surfaces to crosslink. At higher concentrations, the connection by one chain of two surfaces cannot be achieved. The upper phase is removed during the washout, demonstrating the dual adhesive and antiadhesive capacities of XG that have been investigated via in vivo and in vitro studies (Stiernstedt, Nordgren et al. 2006;Winter, Cerclier et al. 2010;Cosgrove 2014).  A C C E P T E D M A N U S C R I P T 9 patterning from pure and regenerated (trimethylsilyl cellulose, TMSC) cellulose. By applying a patterned mask (microchannels = 40-100 µm width spaced at 50 µm) onto the TMSC surface and using the common regeneration procedure (HCl vapors), regeneration of TMSC to cellulose is achieved in the exposed areas only. Furthermore, filling stamp capillaries with cellulase solution on these patterned films allows selective degradation of pure cellulose thin films, whereas pure TMSC cannot be digested.

ACCEPTED MANUSCRIPT
These techniques make it possible to create biopolymer film patterns with specific surface topography and chemistry (e.g. hydrophilic/hydrophobic domains, crystalline/amorphous regions), and to selectively functionalize the surfaces through available reactive groups of biopolymers.

Thin films as biomimetic model and biomass degradation model surfaces
The literature reports a number of approaches to investigate cell wall polymer organization and interactions using physical, spectroscopic and biochemical techniques. These approaches on living A C C E P T E D M A N U S C R I P T 10 in contact, and XG would act as a one-molecule-thick adhesive between them. Therefore, mechanical properties and wall extension would be controlled at limited regions of close contact between cellulose microfibrils, mediated by the XG chains (Park and Cosgrove 2012).
In an attempt to gain a deeper understanding of the adsorption of XG onto the cellulose surface, we developed a kinetic model that gives more insight into the arrangement of polymer chains on the cellulose surface (Villares, Moreau et al. 2015). The model takes into account two contributions to the adsorption process: (i) at low XG concentration or short adsorption time, every XG molecule reaching the uncovered CNC surface is directly adsorbed and has time to rearrange before new molecules arrive, likely as a monolayer or pancake conformation; (ii) when a certain coverage is reached, the preformed layer prevents further molecules adsorbing on the surface by steric hindrance and the unavailability of binding sites. The overall adsorption process can therefore be described as the contribution of XG adsorption onto the uncovered CNC surface and the contribution of adsorption onto pre-covered CNC surfaces after XG rearrangement (Wågberg, Ödberg et al. 1988;Couture and Vandeven 1991;De Witt and Van de Ven 1992) (Fig. 2).
We prepared model cellulose surfaces using spin-coating to deposit one layer of CNC onto gold-coated quartz crystals, and then monitored XG adsorption in situ by quartz crystal microbalance with dissipation (QCM-D  Schematic description of XG adsorption onto CNC surfaces at very diluted (a) and more concentrated (b) XG regimes.

ACCEPTED MANUSCRIPT
In fact, the two CNC/XG complexes fabricated from low and high XG concentrations, respectively, show different enzymatic accessibility. We previously demonstrated that CNC/XG assemblies prepared from concentrated solutions experienced a higher mass loss upon incubation with a XG-specific endoglucanase than CNC/XG assemblies fabricated at low XG concentration (Dammak, Quemener et al. 2015). The different arrangement of XG chains at the cellulose surface was shown to be crucial for certain applications, such as the stabilization of colloidal suspensions. Colloidal stability hinges on avoiding aggregation and flocculation. When bacterial cellulose is coated with XG at low concentration, XG chains can cross-link cellulose nanocrystals and therefore promote association, which results in aggregation and consequently low colloidal stability. In contrast, for the CNC/XG assemblies prepared from XG at high concentration, the loops and tails formed by the XG chains at the cellulose surface prevent aggregation by steric stabilization (Winter, Cerclier et al. 2010).
We have also studied the impact of molar mass on adsorption behavior (Villares, Bizot et al. 2016) and demonstrated that a decrease in XG molar mass from 10.28·10 5 g.mol -1 to 1.04·10 5 g.mol -1 has impacts on XG conformation on the CNC surface. High-molar-mass XG (10.28·10 5 g.mol -1 ) formed loops and tails on CNCs whereas low-molar-mass XG (1.

A C C E P T E D M A N U S C R I P T
12 was confirmed by the greater accessibility of high-molar-mass XG loops compared to low-molar-mass XG trains upon hydrolysis of CNC-XG-CNC assemblies by an XG-specific endoglucanase.

Enzymatic detection
Plant biomass transformation processes to produce monosaccharides and oligosaccharides need a far better enzymatic treatment to degrade lignocellulosic materials. However, the cellulose microfibrils in plant cell walls are encrusted in lignin and hemicelluloses in a complex architecture that, together with the crystallinity of cellulose, makes cellulosic biomass recalcitrant to hydrolysis to fermentable sugars.
Enzyme cocktails need to be optimized and new enzymes discovered, so a critically important challenge is to develop easy-to-use, sensitive and rapid screening tests applicable to a wide range of hydrolytic enzymes (Arantes, Gourlay et al. 2014;Lee, Hamid et al. 2014;Alvarez, Reyes-Sosa et al. 2016).
Traditional methods to determine specific enzyme activity are based on detecting the reducing sugars formed during biopolymer hydrolysis in solution (Somogyi 1937;Nelson 1944;Somogyi 1945).
Assemblies of biopolymers onto a solid surface at nanoscale can prove valuable model systems for investigating enzyme hydrolysis. By controlling the deposition conditions as well as the polymer solution characteristics, it becomes possible to prepare thin films of appropriate thickness, composition and porosity, and biopolymer thin films with the appropriate layer architecture are able to produce strong interference colours. Indeed, when film/layer architecture is tuned to appropriate values of thickness and refractive index, there is an interference that occurs between light reflected from the air-film interface and from the film-substrate interface ( Fig. 3) (Hiller, Medel et al. 2002). These structurally colored films were achieved from spin-coated or dip-coated polyelectrolyte/CNC multilayer films (Cranston and Gray 2006;Podsiadlo, Sui et al. 2007;Jean, Dubreuil et al. 2008;Moreau, Beury et al. 2012). Alginate films fabricated by spin-coating were then exposed to aerosolized CaCl2 solution to enable layer crosslinking (Cathell and Schauer 2007). All these works showed that homogenous film surfaces of various nanometer-scale thicknesses can be produced with a wide variety Biopolymer films displaying this remarkable optical property served to design highly sensitive detectors for biomass-hydrolyzing enzymes (Cerclier, Guyomard-Lack et al. 2011;Dammak, Moreau et al. 2013).
Indeed, when films are in contact with specific enzymes able to hydrolyze biopolymers, there is a change of film color due to the thickness decrease of the degraded layer.

A C C E P T E D M A N U S C R I P T
14 xylanase at different concentrations (Dammak, Moreau et al. 2013). This new assay for enzymatic activity detection based on color change of nanostructured biopolymer films is a simple yet highly sensitive method compared to commonly used methods. It offers new opportunities to detect a broad range of enzymes specific to biomass polymers. and spots corresponding to Cellulyve enzyme degradation of the film at an enzyme concentration of (1) 5000 µg/mL, (2) 50 µg/mL -1 , (3) 25 µg/mL -1 , (4) 20 µg/mL -1 , (5) 15 µg/mL -1 and (6) 10 µg/mL -1 .

Conclusion
Progress in creating biobased surfaces and thin films has produced useful tools for obtaining functional and/or chemical information from plant cell wall biopolymers. The ability of such biological molecules to naturally self-organize in nature makes it possible to fabricate many novel nanoscale materials inspired from specific natural properties and chemical complementarity.

Nanocomposite materials for food packaging applications: improving functional properties and safety concern
In the extensive field of nanoscience R&D, food packaging is without a doubt one of the most promising and advanced R&D areas, posting a compound annual growth rate of 12.7% for commercial applications (Bumbudsanpharoke and Ko 2015). The development of engineered nanomaterials (ENM) for food packaging applications aims either to enhance the mechanical and barrier properties of conventional or biobased packaging materials and/or provide innovative active and intelligent functionalities. As defined in the food contact material framework regulation 1935/2004, active and intelligent packaging deliberately incorporates active or intelligent components intended to release or absorb substances into, onto or from the packaged food or its surrounding environment, or to provide the intended information, with their conditions of use. A large portfolio of nanoparticles spanning organic/mineral nanospheres, nanotubes and nanosheets is being used to drive the development of such innovative packaging materials. Indeed, consumer acceptance and regulatory attention are easier to manage with such "outof-food" applications than "inside" food applications (Siegrist, Stampfli et al. 2008;Duncan 2011).
Active ENM packaging is based on introducing substances inside the material matrix, or attached to the surface or trapped in between different layers, to maintain the quality and safety and extend the shelf life of the food. Oxygen absorber is one such substance. These innovative packagings have attracted much attention in the food industry as a way to replace conventional food preservation systems (Cho, Lee et al. 2011) in which levels of additives and physical-chemical treatments can lead to adverse effects that are avoidable with the use of efficient active material (Kanmani and Rhim 2014). Nanotechnology provides more efficient activity of the active substance at lower concentration, but also facilitates their

A C C E P T E D M A N U S C R I P T
16 incorporation into the packaging materials without detrimental impact in other packaging properties such as transparency, mechanical performance and barrier performance.
By providing accurate and easy-to-read information on food safety and quality, a new generation of intelligent ENM packaging materials has a key role to play in limiting/avoiding the safety and waste issues resulting from the margin of safety and misunderstanding of conventional food date labels. The unique optical, chemical, biochemical and electrical properties of nanoscale particles offer breakthrough foundations to develop cost-efficient 'smart' labels for food packaging applications.
Nanotechnology-based non-invasive sensors have radically changed the speed and accuracy of smallmolecule detection. Nanotechnologies can be easily embedded into the packaging material itself or into a label or a coating, to add an intelligent function to food packaging (Bowles and Lu 2014). These socalled "intelligent packagings" are able to provide users with appropriate in-time information on the quality and safety of the food, either indirectly by monitoring storage conditions such as timetemperature history, or directly by monitoring the presence of food degradation markers. These indirect or direct indicators of food quality and safety are able to integrate multiple gas and volatile sensors useable on flexible substrate such as packaging materials. Recent special attention to wireless sensors could potentially lay the foundation for the truly convergent wireless sensor ad hoc networks of the future (Gontard, Peyron et al. 2017 is based on a suitable risk assessment approach taking into account the "specific properties of the engineered nanomaterials in addition to those common to the equivalent non-nanoforms". The EU had already edited statutory texts in this direction along with technical guidance mentioning nanomaterials. In particular, the EU Plastics Regulation ( 18 technologies that engineer substances in particle sizes that exhibit chemical and physical properties significantly diverging from those at a larger scale-nanoparticles included-should be risk-assessed on a case-by-case basis until more is known about the novel technology. On this basis and taking into account the lack of knowledge about their potential toxicity (oral exposure to engineered nanomaterials hasg received less attention than the dermal or inhalation pathways), the concept of functional barrier used to prevent migration of contaminants, which are not evaluated by health authorities, cannot be applied in the specific case of PNC packaging including nano-dimensional substances. This statement notably differentiates nanoform substances from non-nanoform susbstances which are authorized for use behind a functional barrier, provided that they fulfil certain criteria and their migration remains below a given detection limit.
It is important to point out that, in the case of food packaging applications, the risk assessment related to ENM goes beyond the nanometric paradigm to encompass two distinct issues: 1) the migration of the nanoparticles themselves or their constitutive elements 2) the impact of nanoparticle addition on the migration of other contaminants (i.e. monomers, additives or processing aids) which can potentially demonstrate toxicity.
The comprehensive evaluation of any PNC's suitability for contact with food products must necessarily include these two investigations.

Migration of nanoparticles and exposure assessment
Considering the first point, and in absence of any harmonized procedure, the difficulty of assessing the release of ENMs from packaging into a complex food matrix is the key challenge for controlling human exposure. Until more data are available and the assessment becomes standardized, the case-by-case approach will fatally involve a disparity of criteria and a high degree of uncertainty. However, new analytical methods and equipment are yet now to measure/detect minute fractions of migrating inorganic nanoparticles (Duncan 2011) .Lots of investigations have been conducted on nano-silver which is widely incorporated into a range of commercial active packagings due to its recognized antimicrobial properties (Duncan 2011). The net result of the work effort converges on the common finding that silver

Impact of ENM incorporation on the inertia of food-contact packaging
As mentioned above, the safety assessment of PNC in terms of human exposure cannot be restricted to the a study of ENM migration alone. The influence of ENM on the inertia of the nanocomposite packaging material remains a question given very little attention, despite being a key issue dictating the suitability of packaging to be in contact with food. Regulation (EC) 10/2011 (European Commission 2011), sets the requirement of inertia for the plastic material as (1) an overall migration limit (to 60 mg kg -1 of food), i.e. the total mass released by the packaging material during a migration test, and (2) a specific migration limit determined on the basis of toxicological data on every substance authorized for use in the manufacture of plastic materials (monomers, process aids and additives such as plasticizers, antioxidants, light stabilizers, etc.).
In the case of homogeneous material (considered as a continuous medium), migration phenomena fit well-established physical laws and can be described by diffusion models, most often based on Fick's second law. In this respect, the EU regulation authorizes the recognized diffusion models based on experimental data to determine overestimated migration levels, thereby avoiding expensive and timeconsuming experiments. This strategy is not easy to transpose to the specific case of NPC which can be considered a heterogeneous medium including impermeable particles and which present transport properties involving more complex physical laws. While the diffusion coefficient of a substance migrating in polymer blends is mainly influenced by it molecular characteristics (molar volume, shape and polarity) and the mobility of the polymeric chains (itself related to temperature, crystallinity of material, chain The stability of polylactide-based PNC had been extensively researches (Plackett, Andersen et al. 2003;Paul, Delcourt et al. 2005;Roy, Hakkarainen et al. 2012), but many studies were conducted over a long period of time at ambient temperature under controlled relative humidity, whereas the migration test conditions recommended by the EU regulation (generally 10 days of contact with the food simulant at 40°C) dramatically increase the degradation rate of polymeric materials, and ENM incorporation could jeopardize the suitability of the PNC for food contact applications.
The impact of ENM on the potentially health-hazardous migration of chemical substances such as plastic additives is the second safety issue to factor into the assessment of packaging material compliance. The migration process can be considered as a combination of the diffusion of chemical additives through the polymers and the desorption of the diffused molecules from the polymer surface to the food or food simulants. In principle, the addition of an impermeable filler in a polymer matrix is expected to affect the solubility and diffusivity of a penetrant molecule (i) by introducing a path tortuosity when penetrants are forced to migrate around impermeable fillers and (ii) by modifying the interaction properties of the PNC due to the sorption/adsorption capacity of the nanoclays and their possible affinity for the diffusing molecules. Efforts have recently been led to enhance the knowledge of migration mechanisms by investigating the impact of ENM on the diffusion coefficient and partition coefficient of a panel of 8 lowmolecular-weight molecules representative of plastic additives in food-packaging systems including PNC (linear low-density polyethylene including 5% of MMT) and different food-simulating liquids (Nasiri, Peyron et al. 2016). Results showed that apparent diffusivity (Dap) was significantly reduced by the incorporation of MMT in the LLDPE-based packaging in the case of contact with aqueous simulants but not when when PNC is set in (Bharadwaj 2001) compared to the values predicted from tortuosity-based models assuming a regular arrangement of parallel nanoplatelets (Nielsen 1967;Cussler, Hughes et al. 1988;Moggridge, Lape et al. 2003;Lape, Nuxoll et al. 2004)  based on "worst-case" scenario using an overestimation of the diffusion coefficient remain applicable to the risk assessment of such nanocomposite packaging.
In contrast, caution should be exercised in the case of low-barrier PNC e(specially polyolefin-based materials) that reach equilibrium within a few days and for which foodstuff contamination is correlated to the partition coefficient. Regarding the contrasted influence of MMT inclusion on the partitioning of substances, the case-by-case approach recommended by the current EU regulation for exposure evaluation on PNCs remain suitable to ensure consumer protection against chemical contamination.

Conclusion
Given the diversity of nanocomposite materials used for packaging applications, the assessment of consumer exposure is a key part of safety concerns that should be led via a multi-criteria analysis.
Beyond the simple question of nanoparticle migration, the impact of incorporating ENM in a food packaging material has to be investigated in terms of their influence on the inertia of the polymeric

A C C E P T E D M A N U S C R I P T
26 nanosubstances to underpin universal risk management, research now needs to progress on identifying the physical law(s) controlling PNC transport properties and then developing the next generation of food nanopackaging solutions using a "safe-by-design" approach with direct benefits for consumer safety.

Controlling the presence of the food additive titanium dioxide and its fate after ingestion
A range of inorganic additives including titanium dioxide (TiO2) is available for food applications, nutraceuticals and supplements. TiO2 is used as a white pigment in processed foods, as described in the recent EFSA report (EFSA 2016). TiO2 has been in the spotlight for years, as it is totally exogenous to our organism, has been classified as possibly carcinogenic by inhalation to humans by the International Agency for Research on Cancer (IARC) since 2006, and it may be formed very small particles that have raised concerns due to their potential ability to cross biological barriers. Here we present the approach used in our lab to analyze the characteristics of food-grade titanium dioxide extracted from food products, together with pointers on the fate of the particles after ingestion.

Identifying titanium dioxide in the coating of sweets
Searching for very small particles in a food matrix can be a really hard task. Investigating titanium dioxide makes it easier, as it is usually at the surface of confectionery and gives a strong signal by Raman spectroscopy when in anatase form. By chance, this form is widely used in food additives (Peters, van Bemmel et al. 2014;Yang, Doudrick et al. 2014;Dudefoi, Terrisse et al. 2017). We thus used this method to discriminate food products containing TiO2. The procedure was applied to a kind of sweets formed of a stick of gum surrounded by a layer of liquorice which is itself covered by a white coating (see a crosssection of one piece in Fig. 8). When the piece was illuminated with a monochromatic laser beam (Fig.   8), the light, which was inelastically scattered, yields the Raman scattering and its wavelength depended on molecular vibrations and phonons at the surface layer. The spectrum of a piece of the sweet is given  Fig. 8, which shows a lot of peaks between 2800-3000 and 50-1500 cm -1 corresponding to the vibration modes of organic molecules (ν(C−Η) and δ(CH2)). At low wavenumbers, it also shows the typical fingerprints of TiO2 anatase, centred at 144, 399, 513 and 639 cm -1 (Ohsaka, Izumi et al. 1978),

ACCEPTED MANUSCRIPT
confirming the presence of TiO2 in the coating of the sweet. Once the presence of TiO2 was detected, the confectionery coating was removed by the protocol shown in Fig. 9, based on the insolubility of TiO2 in water. In a first step, the coating was stripped from the sweet by placing it in deionized water for 30 seconds and hand-shaking. Once the white layer of the coating was separated from the gum, a waterwashing step was performed to separate TiO2 from physically adsorbed organic molecules. The washed coating was then characterized by FT-Raman spectroscopy, transmission electronic microscopy (TEM), inductively coupled plasma atomic emission spectrometry (ICP-AES) and X-ray diffraction (XRD) to learn more about the amount, shape, structure and purity of the nanoparticles. The washed coating was analyzed by FT-Raman spectroscopy (Fig. 8). All fingerprints of organic molecules disappeared after washing the extracts. Only the typical fingerprints of TiO2 anatase were visible, confirming that TiO2 was easily separated from the other ingredients.

Fig. 9.
Steps in the characterization of TiO2 particles extracted from the coating of sweets (protocol adapted from Chen et al. (2013)).
As rutile is less easily identified by Raman than anatase, the nature of the crystalline phase of TiO2 in the coating of the sweet was also confirmed by XRD measurements (Fig. 10). The main crystalline structure detected was the anatase phase (2θ = 25.30°) with some traces of rutile (reflection (110) at 2θ = 27.41°).
This composition is similar to that found in chewing-gums (Chen, Cheng et al. 2013). The diffractogram exhibits one additional signal at 2θ = 28.96° which can be attributed to talc, a natural magnesium silicate used as anti-caking agent.

Characterization of nanoparticles by transmission electron microscopy
The size and percentage of nanoparticles contained in the washed coating were determined by TEM in the same conditions as in a previous study performed in our lab (Dudefoi, Terrisse et al. 2017). Analysis of the TEM images (Fig. 11) showed the presence of 34% TiO2 nanoparticles in the sweet coating, which is consistent with values found in native E171 powders or food products (Weir, Westerhoff et al. 2012;Chen, Cheng et al. 2013;Dudefoi, Terrisse et al. 2017). This amount is below the cut-off threshold (50%) above which the labelling of nanomaterial is required in accordance with the current definition of nanomaterial. In addition to TiO2, we also identified larger objects (>500 nm) containing Mg, Si and Al (determined by energy-dispersive X-ray spectroscopy EDX), which could correspond to the talc identified by XRD. The presence of aluminium in talc is usual and dependent on the kind of deposits. 0.03% of the total weight of the sweet, in accordance with the values found by Lomer et al. (2000) in various food products (0.045-0.782%) and Wang et al. (2013) in chewing gums. The presence of Al, Si and Mg were confirmed by ICP-AES measurements, at 9.15, 72.78 and 35.14 mg/g of washed coating, respectively. The presence of P seems typical of food-grade forms of TiO2 and was identified as a phosphate species (Dudefoi, Terrisse et al. 2017 Considering both the percentage of nanoparticles and the amount of TiO2 within the coating of this sweet, we estimated the amount of nano-TiO2 per gram of gum and per gum ( Table 2). The ingestion of one sweet represents an ingestion of 0.05 mg of TiO2 NPs, i.e. 3.8x10 13 nanoparticles. This value is close to that found in other food products (4.10 13 in Lomer et al.(2000)). According to a study on the consumption of TiO2 in Germany, the ingestion of TiO2 from chewing-gums represents a small amount compared to any other intake (Bachler, von Goetz et al. 2015). For example, savoury sauces, dressings, soft drinks and cheese represent more than 75% of the intake of titanium dioxide in adults, whereas consumer perceptions are more focused on confectionery. Daily median intake is thus between 0.5 and 1.0 mg/kg b.w. for all age groups (from toddlers to seniors), except for A C C E P T E D M A N U S C R I P T 32 children between 3 and 9 years who have the highest titanium intake of all age classes (approximately 2.0 mg/kg b.w.) (Bachler, von Goetz et al. 2015). As noted by the authors, the actual amount of ingested TiO2 may be overestimated due to a bias towards products with high titanium levels. Moreover, exposure is significantly different between countries due to a significant difference in consumption habits, for example between the US and the UK (Weir, Westerhoff et al. 2012). Furthermore, ICP devices are expected to gain sensitivity and food composition is expected to change, and we have experienced these trends ourselves: i) TiO2 was not previously found in Wrigley-brand chewing gums (Lomer, Thompson et al. 2000) yet recently found in work performed at our lab and elsewhere (Athinarayanan, Alshatwi et al. 2015;Dudefoi, Terrisse et al. 2017)); ii) TiO2 is no longer found in Smarties, based on work by us on the listed ingredients and a previous review (Rydström Lundin 2012).

Physical-chemical properties of titanium dioxide E171
The physical-chemical characteristics of particles, including morphology (spherical, cylindrical), size (smaller or larger than <100), charge (negative, neutral or positive), structure (crystallinity), agglomeration (aggregates, agglomerates, primary particles) and surface composition, are assumed or demonstrated to play a role in nanoparticle uptake through the gut (EFSA Scientific Committee 2011).
For this reason, we set out to characterize the washed coating of sweets and chewing-gums. However, the amount of TiO2 per sweet was very small (0.88 mg/sweet) and its extraction from the coating could not be done without the co-extraction of inorganic substances or with a chemical treatment that would have modified the surface chemistry of the extracted particles. To get round this problem and obtain the physicochemical properties of TiO2 particles, we turned to the raw powder of food-grade TiO2 additive E171. We got several samples sold in Europe and analyzed them by a large panel of tools including TEM, XRD, laser scattering particle size distribution analysis, X-ray photoelectron spectrometry (XPS), zeta potential measurements, Raman and Infrared spectroscopies, ICP-AES, diffuse reflectance infrared Fourier-transform spectroscopy (DRIFTS) and surface area analysis. As almost nothing is known about these food-grade particles, we compared their properties with the well-known P25 particles which serve as reference (Orts-Gil, Natte et al. 2013 A C C E P T E D M A N U S C R I P T 33 nanoparticles (ranging from 17 to 36% in number) below the threshold (50%) defining the labelling of nanomaterial (Dudefoi, Terrisse et al. 2017), in accordance with the results found on the sweets. These powders are also characterized by a low specific surface area, a pure crystalline-phase anatase, a low isoelectric point (around 4.1) related to the phosphate found at its surface, and agglomeration in water.
Comparison of the physical-chemical characteristics of both E171 and P25 samples of TiO2 is shown in Fig. 12. The P25sample, which is a non-food grade material but serves as a reference material, appears very different from the food-grade sample. E171 is not therefore the best material to use when studying human exposure by ingestion, or should be at least investigated in parallel to the reference material P25. This chart also emphasises that the concern about nano-forms is not only related to the physical form (NPs) but also to the the chemistry of the particles.

Fig. 12.
Radar chart showing specific surface area (SSA), isoelectric point (IEP), percentage of rutile (the complementary part of anatase to 100%), threshold in reflectance (G), surface composition of elements relative to titanium (O, P and K), the granulometric diameter at pH 2 and 9 (D2 and D9 respectively) of titanium dioxide E171 (dark grey stars and area, data averaged for several batches described individually in Dudefoi et al. (2017)) and P25 (light grey squares and area).

Transformations in the gastro-intestinal tract
When the process of chewing TiO2-coated chewing-gums was over, only a small amount of TiO2 remained in the gum (maximum 7.2 ± 0.6%), which means more than 93% of the intake of TiO2 into the body happened after only 10 minutes of chewing, i.e. 1.5 mg for one piece of chewing-gum (Chen, Cheng et al. 2013). Our previous results on TiO2 extraction from sweets showed that TiO2 was easily separated from the other ingredients and could thus be considered as naked when entering into contact with digestive fluids. While the digestive system is able to transform most food into absorbable structures, inorganic particles like TiO2cannot be metabolized (Zhao, Hallin et al. 2014). Nevertheless, they do undergo the different physical-chemical conditions along the alimentary tract. In the oral cavity, particles already had a high agglomeration tendency (Teubl, Schimpel et al. 2015), but 10% were still available in the nanosized range and penetrated the upper and lower buccal epithelium, independently of hydrophilicity (Tay, Fang et al. 2014;Teubl, Leitinger et al. 2014;Teubl, Schimpel et al. 2015). Once introduced directly in the gastric fluid, TiO2 NPs agglomerate anyway (Jones, Morton et al. 2015). In all the digestive fluids, TiO2 particles could still exist as primary entities, but the percentage of nano-TiO2 is lower (Chen, Cheng et al. 2013), especially in the gastric fluid. This was attributed to the presence of proteins and electrolytes (Chen, Cheng et al. 2013;Jones, Morton et al. 2015). Further agglomeration was observed when TiO2 was dispersed in polymeric or elemental foods (Jones, Morton et al. 2015).

A C C E P T E D M A N U S C R I P T
35

Crossing biological barriers
The amount of TiO2 that gets absorbed by the body is actually extremely low (Cho, Kang et al. 2013;Jovanović 2015). This was first proved by detecting up to 43.2±4.6 µg/L of TiO2 in the blood of volunteers after oral ingestion of 22.9 mg of TiO2 incorporated in gelatin capsules (Böckmann, Lahl et al. 2000). If the ingested amount had totally crossed the intestinal barrier, the amount for an adult with 6 L of blood would amount to 3.8 mg/L, i.e. roughly 100 times more than the amount found in reality, which confirms the low absorption rate. This translocation was confirmed by Pele et al (2015) but disconfirmed by Jones et al. (2015). The discrepancy can be explained by the time of tests, the formulation, and the type of particles. First, the collection of human blood samples in the study of Jones et al. (2015) was performed 2, 4, 24 and 48h after dosing while peak level of TiO2 was noted between 4 and 12h for all volunteers in the study of Böckmann et al. (2000). Second, levels in blood depend on type of formulation (TiO2 encapsulated in gelatin capsules or not; Böckmann, Lahl et al. 2000). Third, the particles used in these two studies were different from a physical-chemical point of view, which may modify particle reactivity, agglomeration and the pathway/kinetics used to cross the physiological barriers.
Many studies converge on the fact that the main mechanism of TiO2translocation involves the M-cellrich layer of Peyer patches, which are epithelial cells specialized for the transcytosis of macromolecules and particles (Powell, Faria et al. 2010;Bergin and Witzmann 2013;Brun, Barreau et al. 2014;Bellmann, Carlander et al. 2015;Higashisaka, Yoshioka et al. 2015;McCracken, Dutta et al. 2016). However, the mechanism of translocation is still under debate, since contradictory results were obtained with in vitro cells (Koeneman, Zhang et al. 2010

Biodistribution and bioaccumulation
Once it has entered the bloodstream, TiO2 can diffuse by blood circulation throughout the organism.
where it is finally mainly found in liver, the main organ where exogenous chemicals are metabolized and eventually excreted, and kidney, the organ involved in the excretion of metabolic wastes (Chang, Zhang et al. 2013). The most recent studies tend to confirm this fact (Cho, Kang et al. 2013; Park, Lee et al. , with sometimes poor clearance (Umbreit, Francke-Carroll et al. 2012;Hummel, Kindermann et al. 2014;Disdier, Devoy et al. 2015) and sometimes a decrease of TiO2 levels down to control levels (Fabian, Landsiedel et al. 2008). Studies on children suspected of having inflammatory bowel disease showed that the amount of pigment in Peyer patches in the terminal ileum becomes denser with increasing age (Hummel, Kindermann et al. 2014), indicating poor clearance.

Conclusion
Titanium dioxide is a food additive widely used in processed foods to improve the color of end-products.
After negative toxicity tests, titanium dioxide found its way into the composition of processed foods a very long time ago, is not labeled as a nanomaterial due to its low number of nanoparticles, and is mostly (95%) excreted from the body. In contrast to nanofoods, including nanoliposomes or solid lipid nanoparticles which are used as delivery systems for vitamins or micronutrients, TiO2 as food additive has no nutritional value but has a role in the perception of food. Following an in-depth re-evaluation of this compound based on the data available in early 2016, the European Food Safety Agency had no scientific reason to reject this inorganic compound as a food additive. Scientific questions that have yet to be answered include the impact of such particles from long-term exposure (as in work developed by Bettini et al. (2017)) and the fate of the non-excreted fraction (location, clearance, effect) as well as its nanoparticle content. Given the demonstrated differences between E171 and P25 materials, we strongly urge researchers studying the biological effects of ingested TiO2 to consider the food-grade forms in addition to P25 forms. Learning whether particles develop a corona of adsorbed proteins in the digestive fluids is one of the goals currently being tackled in our lab to better understand the absorption of this compound through the intestinal barrier.

Future for nanotechnology
Nanotechnology offers promising new possibilities in the agriculture and food industry where it has concrete applications, but there is also a lack of awareness and uncertainty about the balance between potential benefits on one side and potential risks and possible long-term side effects on the other.
Consumer perception of health and safety concerns is central to the uptake or rejection of this new technology. The number of questions, not still well taken into account and financed, on the identification of nanomaterials in complex matrices, on their short-and long -term behavior in humans and the environment has increased both citizen and scientist concerns but at the same time fueled a lack of transparency in commercial communications and in institutional documents.
For the future, research should aim to ensure that nano-devices will help develop cheap and highly integrated diagnostic systems of higher sensitivity, specificity and reliability for chemical, biological and medical analysis to play a major role in biodegradable packaging, protective coatings, food safety monitoring, and also ensure that nanotoxicology will provide a thorough nanoscale hazard and risk analysis before nanotechnology is exploited by industry and used by consumers. Nano should not be commercialized and marketed in food and agriculture until nano-specific safety guidelines and laws have been issued, nano-specific standard specifications have been agreed, and nano-specific regulatory issues for food and environment technology have been established with public input to decision-making.
Better consumer awareness on nanoscience and nanotechnologies is equally vital.