Cellulose nanofibers as excipient for the delivery of poorly soluble drugs

Cellulose nanofibers as excipient for the delivery of poorly soluble drugs

Accepted Manuscript Title: Cellulose Nanofibers as Excipient for the Delivery of Poorly Soluble Drugs Authors: Korbinian L¨obmann, Anna J. Svagan PII:...

1MB Sizes 0 Downloads 65 Views

Accepted Manuscript Title: Cellulose Nanofibers as Excipient for the Delivery of Poorly Soluble Drugs Authors: Korbinian L¨obmann, Anna J. Svagan PII: DOI: Reference:

S0378-5173(17)30930-4 http://dx.doi.org/10.1016/j.ijpharm.2017.09.064 IJP 17045

To appear in:

International Journal of Pharmaceutics

Received date: Revised date: Accepted date:

16-8-2017 20-9-2017 22-9-2017

Please cite this article as: L¨obmann, Korbinian, Svagan, Anna J., Cellulose Nanofibers as Excipient for the Delivery of Poorly Soluble Drugs.International Journal of Pharmaceutics http://dx.doi.org/10.1016/j.ijpharm.2017.09.064 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Cellulose Nanofibers as Excipient for the Delivery of Poorly Soluble Drugs Korbinian Löbmanna* and Anna J. Svaganb* a

The University of Copenhagen, Dept. of Pharmacy, Universitetsparken 2, Copenhagen, Denmark.

b

Royal Institute of Technology (KTH), Dept. of Fibre and Polymer technology, Teknikringen 56-58,

Stockholm, Sweden. *Corresponding [email protected], Tel: +46 8 790 60 00; [email protected], Tel.: +45 35 32 05 41

Graphical Abstract

ABSTRACT Poor aqueous solubility of drugs is becoming an increasingly pronounced challenge in the formulation and development of drug delivery systems. To overcome the limitations associated with these problematic drugs, formulation scientists are required to use enabling strategies which often demands the use of new excipients. Cellulose nanofibers (CNFs) is such an excipient and it has only recently been described in the pharmaceutical field. In this review, the use of CNF in drug formulation with a focus on poorly soluble drugs is featured. In particular, the aim is to describe and discuss the many unique properties of CNFs, which make CNFs attractive as excipients in pharmaceutical sciences. Furthermore, the use of CNF as stabilizers for crystalline drug nanoparticles, as a matrix former to obtain a long-lasting sustained drug release over several weeks and as a film former with immediate release properties for poorly soluble drug are reported.

1

Finally, the preparation of pharmaceutical CNF foams together with poorly soluble drugs is highlighted; foams, which offer a sustained drug delivery system with positive buoyancy. Keywords: Cellulose nanofibers; drug delivery; poorly-soluble drugs; immediate release; sustained release

2

1.

Introduction

The development of oral drug delivery systems to improve the efficacy of new drugs is plagued by their often very low aqueous solubility. Around 40% of marketed drugs and up to 70% of drug candidates in the pipeline of the pharmaceutical industry show poor aqueous solubility, consequently resulting in unsatisfactory treatment for the patient when given orally (Babu and Nangia, 2011; Eddershaw et al., 2000; Hauss, 2007). The problems associated with poor aqueous solubility can potentially be overcome by using enabling drug formulation approaches. The most frequently used strategies in the pharmaceutical industry are (i) particle size reduction (nano-sizing), (ii) the use of the amorphous form, and (iii) lipid based drug delivery systems. All of these approaches have led to products on the market. For example, the antinausea drug aprepitant as a nano-particulate formulation in Emend® (Hargreaves et al., 2011), the HIV drugs ritonavir and lopinavir as an amorphous formulation in the medicine Kaletra® (Vasconcelos et al., 2016), or the HIV drug saquinavir as a lipid based drug delivery system in Fortovase® (Porter et al., 2008). However, all of these approaches also come along with drawbacks. Particle size reduction often does not lead to the desired dissolution increase because of limitations in size reduction (Merisko-Liversidge et al., 2003). Furthermore, because of the high surface energy, nanoparticles tend to agglomerate or aggregate and stabilizing excipients are needed to avoid the formation of bigger agglomerates or aggregates (Peltonen and Strachan, 2015). The main disadvantage of amorphous formulations is that they are physically unstable, and the solubility advantage is lost upon recrystallization (Hancock and Zograf, 1997). For this purpose, the amorphous form usually also requires the addition of stabilizing excipients such as polymers (Vasconcelos et al., 2016), mesoporous silica (Laitinen et al., 2013) or in form of smaller interacting molecules such as in co-amorphous formulations (Dengale et al., 2016). Lipid based drug delivery systems, on the other hand, frequently have chemical stability problems, scalability challenges and often only a low drug loading capacity (Muellertz et al., 2010; Porter et al., 2008). The poor solubility of some drugs is often further complicated if the drugs have a narrow absorption window such as a site-specific absorption only in the stomach or the upper intestine. The transit time in these parts of the gastro intestinal tract are often variable and usually comparatively short, hence, adding to the low drug

3

absorption and bioavailability despite the attempts of using enabling technologies. Drugs with a limited absorption window include for example levodopa (Hoffman et al., 2004), riboflavin (Hoffman et al., 2004; Klausner et al., 2002) and the poorly soluble drug furosemide (Chungi et al., 1979). In such special cases, a gastro-retentive drug delivery system, which has a prolonged residence time in the stomach, may enable an improvement in bioavailability when releasing the drug in a sustained matter to the absorption site. The challenges associated with the formulation for a given (poorly soluble) drug usually depend on the specific requirements of the drug. In order to develop drug delivery systems with desirable properties, especially for problematic drugs, the use and development of new excipients (and formulation approaches) is often necessary. One promising excipient in this regard is cellulose nanofibers (CNFs). Interest in CNF as excipient in drug formulations has increased in the past few years, because of their unique rheological, barrier and physico-chemical properties which allow CNFs to stabilize oil/water and air/water interfaces, as well as their large surface area-to-volume that offer possibilities for positive molecular interactions with poorly-soluble drugs or stabilization of nanoparticles. Additionally, recent studies have demonstrated the possibility to, in a facile way, structure CNF based materials (particles, capsules, Pickering stabilized lipophilic droplets, films and foams) with tailored drug release properties. In this article the unique properties of CNF, and its use as excipient in drug formulations for poorly water soluble drugs, are reviewed. 2.

Cellulose nanofibers (CNFs) – what is it?

There is a saying “a dear child has many names”, and this certainly holds true for cellulose nanofibers or cellulose nanofibrils, CNFs. CNFs have changed name several times since they were first reported in the early 1980’s (Turbak et al., 1983), and even today the nomenclature used in literature frequently leads to misunderstandings and ambiguities. Other names include microfibrillated cellulose (MFC), nanofibrillated cellulose (NFC), nanocellulose, and cellulose microfibrils, to mention a few. Indeed, a standardized nomenclature is urgently needed and, fortunately, such a nomenclature is presently under preparation by TAPPI (TAPPI, WI 3021). According to the definition by TAPPI, CNF is a type of cellulose nanofiber that contains both crystalline regions and amorphous regions, with dimensions of 5-30 nm in width and aspect

4

ratio (=length/width) usually greater than 50 (TAPPI, WI 3021). Cellulose nanofibers should not be confused with cellulose nanocrystals (Habibi et al., 2010) or microcrystalline cellulose (Thoorens et al., 2014). Cellulose nanocrystals, CNC, are cellulose nanoparticles that consists of predominately pure crystalline cellulose, with dimensions of 3-10 nm in width and aspect ratio greater than 5 but usually less than 50 (TAPPI, WI 3021). In other words, CNC are only the shorter crystalline segments that are found in CNF. Hence, CNC can be made from CNF by removing the amorphous parts that connect these crystalline segments in the CNF structure. However, CNC is typically produced by acid hydrolysis of pulp fibers, filter paper or other cellulosic materials (Habibi et al., 2010). Microcrystalline cellulose, MCC, is an old and traditional excipient in pharmaceutical applications that was first commercialized under the brand name Avicel® in the early 1960’s (Thoorens et al., 2014). MCC is, amongst others, used in tablets as a binder enabling tablet production via direct compression. MCCs are obtained via hydrolysis of cellulose and the product is commonly prepared by spray drying the neutralized aqueous slurry of hydrolysed cellulose. The result is a dry powder of particles that are agglomerates of hydrolysed cellulose. The size of the MCC particles depends on processing conditions but is quite large, 50200 µm, that is, several orders of magnitude larger compared to CNF. Both MCC and CNC differ in properties compared to CNF, due to differences in aspect ratio (=length/width), size and structure. Differences are found in mechanical, barrier and rheological properties and these properties are described for CNF in the following sections. 2.1.

Preparation of CNF

Cellulose nanofibers are produced from cellulosic materials derived from different botanical origin, e.g. wood, hemp, flax, cotton, algae or animals such as tunicate animals. In addition there is a certain type of bacteria, the most efficient and studied producer being Acetobacter xylinum (Gluconacetobacter xylinum), that are able to secrete nanoscale cellulose ribbons as an extracellular metabolite (Chawla et al., 2009). This so-called bacterial cellulose (BC), is devoid of hemicellulose, lignin and pectin, whereas CNF derived from other sources typically contains both hemicellulose and lignin. Bacterial cellulose is considered as a promising material for implants and scaffolds in tissue engineering, due to its biocompatibility, mechanical

5

strength and chemical and morphological controllability (Petersen and Gatenholm, 2011). Bacterial cellulose is also a food-grade product and sold as the popular dessert “Nata-de-coco” in Asia. Nata-de-coco typically consists of cut cubes of BC emerged in some type of fruit juice or syrup. In Fig. 1a cubes of nata-de-coco are presented. The isolation of cellulose nanofibers from the different cellulosic origins is performed using mechanical treatment, often in combination with some chemical or enzymatic pre-treatment prior to the disintegration step. The most common chemical pre-treatments are perhaps those that render the pulp fibers (pulp fibers are used when the botanical origin is wood) charged, i.e. anionic and cationic. This modification results in electrostatic repulsion between the fibers, which is also beneficial in the subsequent mechanical treatment step as it further promotes the disintegration process into nanofibers. The resulting nanofibers possess charged functional groups on the surfaces and are typically finer (thinner CNF width and the resulting CNF suspension is transparent) compared to CNF obtained from unmodified or enzymatically treated pulp fibres (thicker CNF that are not fully separated, the obtained CNF suspension is milky/opaque). The mechanical processing, using a high-pressure homogenizer or grinding instrument, is an energy consuming step that induces high shearing forces that cleaves the macroscopic cellulose structures along the longitudinal axis of the cellulose microfibrillar structure. Caution should be taken during the mechanical processing so that the obtained cellulose nanofibers are not too short, because the end-product will then more resemble cellulose nanocrystals, see the previous definition of cellulose nanocrystals (CNC). The length of the nanofiber will depend on the degree to which the material has been exposed to mechanical processing, for example the number of passages through the high-pressure device. If the processing is performed correctly, the final product is a low solid content suspension that has the appearance of a highly viscous gel. In Fig. 1b such a gel-like material is shown and in Fig 1c an atomic force microscopy image of individual nanofibers is shown. In Fig. 1d a film from anionic CNF is also presented demonstrating the optical transparency. The cellulose nanofibers presented in Fig. 1b and 1c were derived from softwood pulp. Today, there are many different types of CNF reported in literature. As mentioned previously, CNF can be made by enzymatic hydrolysis and mechanical shearing (Pääkkö et al., 2007). Anionic CNFs are usually

6

made by 2,2,6,6,-tetramethylpiperidine-1-oxyl radical (TEMPO)-mediated oxidation or carboxymethylation of pulp fibres and then disintegration into TEMPO-CNF (Saito et al., 2007) and carboxymethylated CNF (Wågberg et al., 2008), respectively. Cationic CNFs are frequently made by reacting the cellulosic material with glycidyltrimethylammonium chloride prior to disintegration (Pei et al., 2013). Recently, new types of CNFs have been produced that also form cross-links between the obtained nanofibers, such as phosphorylated CNF that can form cross-links between phosphate groups (Noguchi et al., 2017), and dialdehyde CNF that will form hemiacetal cross-links (Larsson et al., 2008). The water content of the TEMPO-CNF in water suspension in Fig. 1b is quite high, 99.1. wt%, which is an indirect consequence of processing difficulties and the high viscosity of CNF suspensions with higher solid contents. The presence of water is also advantageous for the further use of CNF, because upon drying CNF will undergo irreversible aggregation, which will create problems when trying to re-disperse the dry CNF in different solvents again. The phenomenon is called ‘hornification’ and is due to hydrogen bonds that are formed between the CNF nanofibres during drying, and that cannot be broken during rewetting. Recent studies have, however, shown that the hornification can be prevented by mixing the CNF with small molecules like NaCl. In this case, the salt will block hydrogen interactions between the nanofibers. Such nanofiber/salt complexes are then easily redispersed in water (Missoum et al., 2012), but then the end product will also contain NaCl. On the other hand, reducing the water content may be attractive from an application point of view and also from the perspective of transportation. 2.2. Structure of Cellulose nanofibers A simplified image of the arrangement of fibrils, microfibrils and cellulose in the cell wall of a microscopic plant cell is given in Fig.2a. The repeating cellobiose units in cellulose, consist of two anhydroglycose rings connected via a β-1,4, glycosidic bond as shown in Fig. 2a. These cellulose molecules are packed together in parallel into semicrystalline microfibrils that are held together via inter- and intramolecular hydrogen bonds and van der Waals forces. The microfibrils are further arranged into macroscopic plant cell walls, as shown in Fig. 2a. Cellulose nanofibers can be individual microfibrils or a bundle of a couple of microfibril entities. The size of a microfibril depends highly on the cellulosic source and for wood the width is typically ca. 4 nm

7

and the length is in the micrometer range. The width corresponds to 36 cellulose chains packed together (Endler and Persson, 2011). Other experimental data have proposed a different model for microfibrils from wood, consisting of about 24 chains, and where the microfibrils structure is twisted, see Fig. 2b and 2c (Fernandes et al., 2011). The microfibrils will contain both crystalline and less ordered regions, and the less ordered (amorphous) regions are increasing towards the surface of the microfibrils (Fernandes et al., 2011). The less ordered regions will also be found in segments along the length of the microfibrils (Battista and Smith, 1962). The prevalent crystalline solid state form in nature is the cellulose-I crystal form, which consists of two allomorphs, the cellulose Iα and Iβ form, in different ratios depending on the botanical source (Schenzel et al., 2009). The native cellulose-I crystal form is not the thermodynamically most stable polymorph, thus if one successfully dissolves nano- or macrofibers and allows the dissolved cellulose to recrystallize (regenerated cellulose), cellulose-II crystals will form. In other words, it is not possible to obtain thin nanofibers like CNF from dissolved cellulose using a bottom-up approach such as for example electrospinning. Thus, once the cellulosic material is dissolved, also the many unique properties due to the specific arrangement of cellulose chains (cellulose I crystalline form) within the CNF, will be lost, e.g. the mechanical properties, amphiphilic surfaces, nanofiber structure and size etc. Exposure to high alkali conditions (mercerization reaction) will also result in a partial or complete transformation from the cellulose I to the cellulose II form. In the cellulose-II the arrangement of cellulose chains within microfibrils will be antiparallel, which will be an energy minimum (Schenzel et al., 2009). Fortunately, it is very difficult to dissolve microfibrils in water or aqueous solvents or even to melt them – a microfibril will degrade (around 220 ᵒC) prior to melting (Eichhorn et al., 2010). This is due to the strong inter and intra-molecular hydrogen bonds as well as the amphiphilicity and hydrophobic molecular interactions (Medronho et al., 2012). The large surface area of microfibrils or cellulose nanofibers is rich in hydroxyl groups, and these can be utilized for further functionalization of the cellulose nanofibers (see above). 2.3. The hydrophobic and hydrophilic surfaces of microfibrils – Pickering stabilization and beyond

8

Unmodified cellulose nanofibers exhibit amphiphilic properties, which are a consequence of the molecular arrangement of the cellulose molecules in the microfibril structure (Kalashnikova et al., 2012). The parallel arrangement of the cellulose chains (Cellulose I form) in the microfibrils gives raise to both hydrophilic and hydrophobic surfaces, which makes microfibrils naturally amphiphilic in character, see Fig 2d (Kalashnikova et al., 2012; Mazeau and Wyszomirski, 2012). The amphiphilic properties of CNF (and CNC) was suggested to enable them to stabilize the interface of oil droplets (dodecane or hexadecane) in water, see recent studies by Gestranius et al and Kalashnikova et al. (Gestranius et al., 2017; Kalashnikova et al., 2013; Kalashnikova et al., 2011). The stability of the prepared emulsions was very high compared to common surfactant based emulsions and the resulting emulsions could be centrifuged without the oil-droplets coalescing. The oil droplets were stabilized via a Pickering mechanism. Pickering emulsions are oil-in-water or water-in-oil emulsions where the interface is stabilized by particles instead of surfactants. The Pickering stabilization phenomenon can also be extended to include the gas/liquid interface, giving raise to very stable gas bubbles. The described phenomenon are named after S.U. Pickering, who was the first to describe it in 1907 (Pickering, 1907). The main advantage with Pickering stabilization is that the solid particles provide the interface of droplets and bubbles with high resistance against coalescence or fusion and (debatable) coarsening or Ostwald ripening. Additionally, since there is no need for surfactants, they are attractive in pharmaceutical applications were surfactants potentially show adverse effects such as irritation (Chevalier and Bolzinger, 2013). Particle stabilized emulsion might be interesting in for example formulations for controlled drug delivery to the skin or for improving the stability of lipid based drug delivery systems. For solid particles to adhere to and be accumulated at the surface of emulsion droplets (or air-bubbles), the partial wetting of particles by both oil and water (or air and water) is required, see Fig. 2d. Also, the oil/water interface will bend in the direction of the poorly wetting liquid, facilitating droplet-formation (Hunter et al., 2008). The preferentially wetting phase will form the continuous phase. Solid particles that are totally wet by water (or oil) will not adsorb because they will be dispersed in the aqueous phase (or oil phase) (Chevalier and Bolzinger, 2013). On the other hand, particles that are strongly adsorbed at an oil/water (or air/water)

9

interface will require a large energy to remove them from the interface. The detachment energy, ∆𝑮𝒓𝒆𝒎𝒐𝒗𝒆 , needed to remove adsorbed spherical particles from an interface into the bulk phase can be described according to (Hunter et al., 2008): ∆𝑮𝒓𝒆𝒎𝒐𝒗𝒆 = 𝝅𝑹𝟐 𝜸𝑶𝑾 (𝟏 ± 𝒄𝒐𝒔𝜽)𝟐

(1)

where R is the radius of the particles, 𝜸𝑶𝑾 , is the initial oil (O)-water (W) interfacial surface tension and 𝜽 is the particle contact angle. The detachment energy is largest at angles (𝜽) around 90°, that is, when the particles are equally (or almost equally) wetted by both phases (Hunter et al., 2008). For stable emulsions (O/W or W/O) to form, the contact angles should be 30° < 𝜽 < 150° (Hunter et al., 2008). There are many particles that fulfil the partial wetting conditions, including some CNF (and CNC) (Gestranius et al., 2017; Kalashnikova et al., 2013; Kalashnikova et al., 2011). Additionally, surface modification can be made to the solid particles so that the condition of partial wetting by water and oil is met. Such groups can be covalently grafted to the surface. Also surfactants, or more preferably (from a pharmaceutical perspective) small surfactant-like drug molecules such as certain poorly-soluble drug molecules, can adsorb to the CNF surface and modify the surface energy of CNF sufficiently to allow the Pickering stabilization of oil/water interface in emulsions. Surface-modification with such molecules will turn the particle surface more hydrophobic. In the case of CNF, the aspect ratio is also expected to further improve the stability of the interface compared to spherical particles (Cervin et al., 2015). Wet-stable foams (with air-water interfaces) will be attained in a more narrow particle contact range (𝜽), compared to emulsions. The contact angles should be below 90°, in particular in the region 60° - 70°, as at these contact angles the particles will be better at preventing two adjacent gas-bubbles from coalescence, by retarding the thinning and the drainage of the liquid in the interfilm between the bubbles, see Hunter et al for further details (Hunter et al., 2008). Indeed, cellulose nanofibers can stabilize the air-water (A/W) interface via Pickering stabilization, but for that to occur (e.g.) an additional surfactant is needed to render the CNF surface more hydrophobic (Cervin et al., 2013; Svagan et al., 2016a). It was recently shown that poorlysoluble drug molecules can be used for this purpose as well (Bannow et al., 2017; Lobmann et al., 2017; Svagan et al.). In this way the surface energy of CNF will change and facilitate the accumulation of CNF at

10

the air-water interface. To obtain a wet-stable foam, Gibbs stability criterion for air-bubbles in a liquid needs to be fulfilled. The Gibbs stability criterion (Stocco et al., 2009; Stocco et al., 2011) states that the complex viscoelastic modulus, E, of the air-water interface should be equal to or exceed half the surface tension, γ, of the suspension containing the nanoparticles; E ≥ γ/2

(2)

Indeed, CNF modified with the right amount of a poorly-soluble drug will fulfil the stability criterion (at both high and low experimental oscillation frequencies, E might depend on the oscillation frequency), whereas this was not possible with a pure CNF suspension or drug solution/suspension alone (Bannow et al., 2017). These CNF-based wet-stable foams could then also successfully be dried down and solid cellular materials with closed cells were obtained. 2.4. Rheological properties CNF based suspensions exhibit complex rheological properties and studies on several different types of CNF suspension can be found in the literature. The prevalent nanofiber network within a CNF suspension can be broken by shearing and will flow over time, so-called shear-thinning behaviour (Herrick et al., 1983; Lasseuguette et al., 2008; Pääkkö et al., 2007). At the same time, the CNF suspension displays thixotropic time-dependent behaviour, that is, the viscosity will reach an equilibrium value when holding the shear rate constant for some (finite) time (Iotti et al., 2011; Lasseuguette et al., 2008). The “characteristic” shape for the flow curve of CNF suspension is shown in Fig. 3 (Tatsumi et al., 2002). The shear stress is constant at low shear rates, for all concentration examined (0.05 - 0.5 wt%). Above a specific threshold, the shear stress increases with shear rate, which might be due to the orientation of the CNF along the flow lines, causing a decrease in the viscosity (Iotti et al., 2011). The values of the shear stress in the plateau region (at low shear rates), the so-called yield stress (σy) of the suspension, increases with increasing CNF content, see Fig. 3. In general, the finer the CNF, the higher the viscosity. The concentration at which a gel will form, is inversely proportional to aspect ratio (= length/width) of the nanofibers (Tatsumi et al., 2002). The elastic modulus (G’, storage modulus) of a CNF gel at rest can be described as G’ ∝ kϕa, where ϕ is the volume

11

fraction of nanofibers, k is inversely proportional to the aspect ratio of the nanofibers and a depends on the volume fraction (Hill, 2008; Saarikoski et al., 2012; Tatsumi et al., 2002). The rheological properties of CNF suspension will depend both of the nanofiber network characteristics and the organization of the nanofiber in the network (Saarikoski et al., 2012), that is, the presence of flocs of nanofibers. Flocculation of nanofibers will affect the rheological properties both during resting and processing (Saarikoski et al., 2012). Additionally, the rheological properties depend on the homogeneity of the CNF size distribution (Hubbe, 2007), which is a consequence on origin of the cellulosic material and the processing conditions (pre-treatment steps, and degree of fibrillation) during production. 2.5. Mechanical properties of CNF The cellulose crystals have an extraordinary stiffness, the Young’s modulus of the crystal is in the range 100 – 160 GPa (Nishino et al., 1995; Sturcova et al., 2005). Thus, by breaking down the macroscopic fibres (e.g. pulp fibres) into cellulose nanofibers, improved mechanical properties can be achieved because the amount of amorphous material will be reduced. In other words, the produced nanofibers will have a much higher proportion of crystallinity compared to the starting material. By reducing the amount of amorphous material, the moisture absorption and swelling will also be reduced in a dry material that is made from CNF (Eichhorn et al., 2010). The mechanical and reinforcing properties in CNF-based materials will depend on the aspect ratio (=length/width) of the final CNF. Thus, there will be an optimal size beyond which the mechanical properties will decrease. A high aspect ratio is desirable as it enables a better stress transfer between matrix material and the reinforcing nanofibers. The reinforcing effect can be predicted using the Halpin-Tsai equation (Affdl and Kardos, 1976) and an excellent example is given by Eichhorn et al. (Eichhorn et al., 2010), where the elastic modulus of the composite material is calculated as a function of the aspect ratio of the filler with different fibre moduli, see Fig 4. The volume fraction of the filler was fixed to 50 vol%. If the aspect ratio is smaller than 10, then no major benefits will be achieved for nanofibres (elastic moduli of 100140 GPa) compared to macrofibers (modulus of 40-60 GPa). At this low aspect ratio, the nanofibers should be considered as cellulose nanocrystals, CNC (see previous definitions). However, when the aspect ratio is

12

larger than 50 a strong reinforcement effect will appear. At aspect ratios larger than 100, the modulus will reach an upper–limit case value. According to the calculations, the modulus for a composite reinforced by a nanofiber with sufficiently high aspect ratio should be more than 3-fold when compared to a macrofibre reinforced composite. Unfortunately, the Halpin-Tsai equation does not take into account fibre-fibre interactions. However, in reality the cellulose nanofibers will form a percolating network which is held together via strong hydrogenbonding in the matrix material. This network will improve the mechanical reinforcement (Svagan et al., 2007; Svagan et al., 2008) and the importance of the network formation has been studied in several papers (Nair and Dufresne, 2003a, b; Nair et al., 2003). The mechanical strength of individual cellulose nanofibers is also extraordinary and it was recently investigated in a study by Saito et al (Saito et al., 2013). The mean strength for cellulose nanofibers derived from wood was in the order of 1.6 - 3 GPa, the exact value was influenced by the method used to determine the width of the CNF. The mean strength can be compared to that of multiwalled carbon nanotubes and ultrastrong para-aramid fibers (Kevlar). To obtain adequate improvements in mechanical properties for different matrix materials, it is vital to obtain a homogenous dispersion of the nanofibers in the matrix material. This is achieved by mixing the CNF using appropriate techniques as well as using adequate dispersion solvents. Unmodified CNF tends to aggregate in solvents other than water, and therefore it is advantageous to work with water as the processing medium. Dimethylformamide, dimethyl sulfoxide or N-methyl pyrrolidine have also shown to be good dispersion media (van den Berg et al., 2007; Viet et al., 2007). Adaptation to non-aqueous processing conditions could be achieved by adsorbing surfactants to the surface of cellulose nanofibers or by chemical modification of the CNF. 2.6. Barrier properties Dry films made from neat CNF exhibit particularly high oxygen barrier properties at low relative humidity and this could potentially be used to improve the oxidative stability of oxygen sensitive molecules during storage (Svagan et al., 2016b). Aulin et al. (Aulin et al., 2010) measured an oxygen barrier coefficient of

13

0.0006 mL µm m-2 day-1 kPa-1 at 0% RH. Similar values were reported by Fukuzumi et al (Fukuzumi et al., 2011). These values are 2 to 3 orders of magnitude lower than for films made from ethylene vinyl alcohol (EVOH) at 0%RH (Lange and Wyser, 2003), which is one of the best petro-chemically derived oxygen barrier material. The excellent barrier properties of neat CNF films depend on a number of different factors. First, the crystalline regions of the CNF will be impermeable to small molecules like oxygen and CNF has a high degree of crystallinity of ca. 50 - 80% (Svagan et al., 2016b). Secondly, the strong hydrogen bonding between nanofibers, makes it difficult for permeating molecules to penetrate the material. High barrier properties are typically obtained for materials with strong hydrogen bonds between polymer chains (this also applies to EVOH). These hydrogen bonds can however be disrupted by small polar molecules, such as water, that act as plasticizers. Indeed the barrier properties of CNF based films quickly decrease above 70 % RH (Aulin et al., 2010). The packing of the CNF in the film will also influence the barrier properties, and this will depend on the type of nanofiber and the preparation steps used in the production of the CNF material. Because CNF are nanofibers, there will inevitably be small pores present in all types of neat CNF films, however these pores do not necessarily need to be interconnected throughout the thickness of the material (Fukuzumi et al., 2011) which also explains the high oxygen barrier properties obtained for neat CNF films in the dry state (0 % RH). As mentioned previously, due to the plasticizing effect of water, the barrier properties might quickly be lost for small permeating molecules, such as drugs, when water is present in significant amounts (Svagan et al., 2016a). However, chemical modification to the nanofibers or chemical crosslinking of CNF might potentially improve the barrier properties further in the wet state. Protection from moisture could also be achieved by for example coating the CNF surface with a hydrophobic film. 2.7. Toxicity aspects Until now, there have been several studies investigating the biocompatibility and toxicity of cellulose nanofibers based materials. However, to give a general answer this question is rather difficult, as it depends on a number of factors such as the surface chemistry (due to surface modification), topography and physical form, e.g. is the CNF in the form of a solid material (films and foams), suspension or particles. In principle if

14

unmodified CNF is dried (into a neat film, foam or microparticle) and hornification (irreversible hydrogen bonding) takes place, then it is reasonable to consider the toxicity of such a CNF material comparable to that of a material made from MCC. In other words, the toxicity aspect should be more of an issue when the CNF are present as individual nanofibers in suspension (at the same time note that BC, which is a network of cellulose nanofibers in water, is considered biocompatible and food-grade). Recent studies by Colic et al (Colic et al., 2015) proposed that unmodified CNF are suitable for use in implantable biomaterials. They showed that CNF (lengths of a couple of micrometers and diameters 10 to 70 nm) are cytocompatible, with non-inflammatory and non-immunogenic properties. These were tested according to the present ISO criteria (ISO10993-5, 2009). Concentrations up to 1 mg mL-1 CNF were tested. Similarly, the findings in the review by Jorfi and Foster concluded that the studies performed up to date give little evidence of serious toxic effects of CNF on the cellular and genetic level (Jorfi and Foster, 2015). 3.

The “stabilization” effect of CNF – interactions with drug molecules and nanoparticles

The hydrophilic and hydrophobic surfaces and the large surface area of CNF, described previously, can potentially be used to adsorb hydrophobic drug molecules, nanoparticles or other hydrophobic moieties to CNF and by this means prevent them from aggregating or crystallizing. However, more research is needed in order to obtain in-depth understanding of such interactions with CNF, as these will depend on the physicochemical properties of the CNF used (Jackson et al., 2011). Indeed, there are many different types of CNFs with completely different properties. Previous studies have confirmed the presence of positive molecular interactions between certain CNF and poorly-soluble drugs. Kolakovic et al investigated drug interactions with a CNF type that was slightly anionic due to a high amount of hemicellulose content (23 wt% xylan) (Kolakovic et al., 2013). Positive molecular interactions were observed in binding studies or isothermal titration calorimetry experiments for the poorly-soluble drug beclomethasone dipropionate and the peptide nafarelin acetate but not for the poorly-soluble drug indomethacin. The molecular interaction with CNF was a combination of electrostatic and hydrophobic interactions (Kolakovic et al., 2013). However, in a recent study by Lobmann et al (Lobmann et al., 2017) positive molecular interactions were also observed between the drug indomethacin and a cationic CNF (with a much lower hemicellulose

15

content). This is contrary to the results by Kolakovic et al. (Kolakovic et al., 2013) and demonstrates the importance of selecting the right type of CNF. Molecular dynamics (MD) simulations were performed that showed that the indomethacin molecules had a clear affinity to CNF that was irreversible on the timescale of the simulations (50 ns). In the MD simulations, the drug oriented planar to the hydrophobic cellulose surface, similar to what has been proposed for carbohydrate binding modules of cellulases. At the hydrophilic surfaces, the drug, to some extent, intercalated between the cellulose chains suggesting that it is more advantageous to bind at a site where the drug molecule can lay more flat. A direct binding between the ionic indomethacin and the positively charged side groups of the CNF did not take place, however the cationic side groups exerted an attractive force on the anionic groups of indomethacin (Lobmann et al., 2017). Positive molecular interactions between indomethacin and a neutral CNF were also proposed by Gao et al (Gao et al., 2014). Additionally, the study by Bannow et al (Bannow et al., 2017) further confirmed positive interactions between indomethacin and cationic CNF. A rapid increase of the amorphous indomethacin fraction was also observed in the final CNF-based materials at processing pHs slightly above the pKa (pKa = 4.42) of indomethacin. The same rapid trend was not observed with increasing processing pH in the absence of CNF, also see the supporting information in ref (Bannow et al., 2017), suggesting that the apparent solubility of the drug was most likely enhanced and the amorphous state of the drug was stabilized in the presence of CNF (Bannow et al., 2017). Storage stability studies were however not performed. In the case of nanoparticles, CNF could potentially be used to prevent nanoparticles from aggregating in aqueous environments. Indeed, CNF has been applied to stabilize crystalline nanoparticles of poorly soluble drugs and preserve the nanoparticulate morphology, both in aqueous suspension during storage but also upon freeze drying of these suspensions (Valo et al., 2013; Valo et al., 2011). For this purpose, hydrophobin fusion proteins were initially adsorbed to the surface of the drug nanoparticles. The hydrophobin fusion protein was genetically modified with cellulose binding domains, which were able to specifically bind to the hydrophobic sites of CNF. By this means, the modified drug nanoparticles were able to bind to non-modified CNF from bleached birch pulp and were stabilized against agglomeration (in aqueous suspension during storage but also upon freeze drying) (see Fig. 5). In the case of crystalline itraconazol nanoparticles, no change in the

16

morphology of the nanoparticles was observed for more than 10 months and the resulting highly porous freeze dried nanoparticle formulation showed an increased in vitro and in vivo performance compared to a plain microcrystalline itraconazol suspension (Valo et al., 2011). In a follow up study, the authors studied the release of the poorly water soluble drug beclomethasone hydrochloride from nanoparticle loaded highly porous freeze dried CNF aerogels prepared from a set of different CNF types (Valo et al., 2013). Depending on the origin as well as the chemical and mechanical modification of the CNF, the surface properties of the CNFs was altered. These included for example, the degree of hemicellulose residues on the CNF fibres or chemical modification such as TEMPO oxidation in order to introduce carboxylate groups on the CNF surface. Hence, the beclomethasone hydrochloride nanoparticles also showed differences in their affinity to interact and bind to the CNF fibres. In the case of red pepper CNF aerogels, the drug nanoparticles only showed limited affinity to the CNF fibres and an immediate release profile similar to plain nanoparticles was obtained. On the other hand, a sustained drug release was obtained from aerogels freeze dried from BC, quince seed CNF and TEMPO- CNF since the drug nanoparticles were immobilized into the matrix of the aerogel as a result of interactions between the nanoparticles and the CNF fibres. 4.

CNF-based particles and capsules

CNF microparticles can be obtained by simply spray drying a CNF suspension. Plain CNF microparticles can be for example used as an alternative filler to MCC in the preparation of tablets (Kolakovic et al., 2011). By adding a dissolved drug to the CNF suspension prior to the spray dying process, drug loaded CNF microparticles can also be obtained. By this means, drug loaded CNF matrix particles with the drugs indomethacin, metoprolol tartrate and verapamil hydrochloride with a sustained release of up to 60 days could be prepared (Kolakovic et al., 2012a). The drug release was mainly controlled by diffusion through the CNF matrix and followed the model for diffusion from spherical matrix objects as developed by Baker and Lonsdale (Tanquary et al., 1974). However, even though all three drugs were comparatively small molecules and one may assume a similar diffusion through the CNF network, differences in the release kinetics were

17

found for the three drugs. The authors suggested that these differences were a result of the solubility of the drug in the dissolution medium as well as the affinity and interactions of the drug with the CNF. The preparation of drug loaded CNF matrix particles seems a promising way to obtain a longer-lasting (several days) sustained release, however, the preparation of these microparticles under the present processing conditions showed several drawbacks. Due to the high viscosity of a CNF suspension even at low CNF concentrations, spray drying was only possible at a very low total solid content of 0.5% (w/v). In addition, CNF suspensions have one of the highest water-holding capacities due to the high affinity of water to the CNF, meaning that the short drying times in the spraying tower required high drying temperatures (outlet temperature above 120C) to allow water removal from the CNF fibers and to obtain a dry product. This made the process unfeasible for thermo-labile drugs and drugs with low melting points. Furthermore, due to differences in the drying kinetics, fractions of the drug and CNF are being dried separately, resulting in very low drug loadings between 4.5 to 15.1 wt% and low entrapment efficacies between 22.5 and 29.9%. Thus, other techniques than spray-drying should be investigated in order to incorporate the drug into a CNF matrix in a more efficient way. CNF based capsules were recently successfully prepared for the first time (Kaufman et al., 2017; Paulraj et al., 2017; Svagan et al., 2016b; Svagan et al., 2014). The capsules diameters were typically in the micrometre range. The smallest size that could be obtained was ca. 600 nm in diameter (Svagan et al., 2014). Kaufman et al prepared water-core capsules from an anionic CNF, a cationic polyelelectrolyte and polyacrylic acid (Kaufman et al., 2017). Their results demonstrated that CNF incorporation provides a facile route for producing strong yet flexible microcapsule shells. In the studies by Svagan and Paulraj et al, the capsules were inspired by the composition or structuring of plant cell walls (Paulraj et al., 2017; Svagan et al., 2014). In nature, cellulose nanofibers, in form of microfibrils, are used to impart mechanical reinforcement to the cell wall of plant cells (=natural cellulose nanofiber capsule). For example, the primary cell wall of parenchyma cells in fruits and vegetables is highly hydrated and contains a soft matrix of polysaccharides; pectins and hemicelluloses. Despite this, the cell wall needs to withstands high mechanical stresses (Cosgrove, 2005). The load bearing components are the cellulose microfibrils, that are further interlocked

18

with hemicellulose molecules. The cellulose accounts for 15-30% of the dry mass of the primary cell wall. Inspired by this, microcapsules based on cellulose and pectin were prepared by the layer-by-layer technique (Paulraj et al., 2017). The resulting capsules had a switchable permeability, which responded to the changes in salinity, and this enabled facile loading/unloading of fluorescent-labelled dextrans. In the study by Svagan et al (Svagan et al., 2014), a mixture of (short) CNF and CNC and monomer (the monomer was a diisocyante that formed urethane and urea bonds upon reaction with the CNF/CNC and water, respectively) was used to co-assemble capsules at the O/W interface of oil droplets in water (continuous phase) in a facile way. In this case, the CNF/CNC stabilized the O/W interface via a Pickering stabilization, thus no extra surfactants were needed. The resulting capsules had a nanofibrous exterior, see Fig. 6, and the mechanical properties (measured at low relative humidity) were significantly enhance due to the CNF/CNC (Svagan et al., 2014). The oil-core of such capsule could potentially be used for delivery of poorly-soluble drugs and/or protection of oxygen-sensitive molecules during storage (Svagan et al., 2016b). 5.

CNF-based drug loaded films

A CNF film (or foam, see below) is perhaps the simplest and most straightforward material that can be made into a drug delivery system. Such a drug loaded CNF film can be obtained by filtration of a CNF/drug suspension and subsequent drying (Gao et al., 2014; Kolakovic et al., 2012b). Alternatively, the CNF/drug suspension can be cast on a surface, followed by a drying step (Lobmann et al., 2017; Svagan et al., 2016a). Kolakovic et al. prepared drug loaded CNF films, with unmodified CNF obtained from bleached birch pulp, using the filtration method (Kolakovic et al., 2012b). First, the CNF suspension was mixed with the poorly water soluble drugs beclomethasone diproprionate, indomethacin or itraconazol. During filtration, the CNF fibres collapsed on the filter membrane and formed a layered CNF network with the undissolved drug particles (particle size 10-50 µm) incorporated into the CNF matrix. In the subsequent drying step, the collapsed CNF fibres underwent hornification and formed a drug loaded CNF film. Using this approach, much higher drug loadings (20-40 wt%) and entrapment efficacies (>90%) could be achieved compared to the spray dried CNF microparticles. Similar to the CNF microparticles (see above), the obtained CNF films showed a long-lasting sustained release of up to 3 months (see Fig 7). All drugs showed a long-lasting

19

release, however, differences in the release kinetics of the three drugs were again attributed to different drug solubilities in the dissolution medium as well as specific interactions between the drug and the CNF fibres. In the case of indomethacin, the release followed a Higuchi kinetic and was mainly limited by the drug diffusion through the CNF matrix. On the other hand, the release of beclomethasone diproprionate and itraconazol followed more a zero-order release kinetic since these drugs strongly bind to and interact with the CNF once they dissolve (Kolakovic et al., 2013) and the release is mainly determined by the desorption of the drug from the CNF fibres. Due to the long-lasting drug release, such systems are less feasible as oral drug delivery systems, however, have been suggested to be useful as drug delivery systems for implants, transdermal patches or ocular applications. A similar long-lasting release of up to 30 days was also observed by Gao et al. when preparing indomethacin loaded CNF films, using CNF obtained from wood powder of popular trees (Gao et al., 2014). In contrast to the long-lasting sustained release from CNF films described above, Svagan et al. (Svagan et al., 2016a) and Löbmann et al. (Lobmann et al., 2017) reported immediate drug release from drug loaded CNF films prepared by simply casting a drug/CNF suspension followed by a drying step. The CNF used was obtained from bleached sulphite pulp and chemically modified with glycidyltrimethylammonium chloride to obtain a cationic surface charge on the CNF fibres. As model drugs, riboflavin and indomethacin have been incorporated into such CNF films. The release from the riboflavin films was slightly lower but comparable to the release from a marketed riboflavin tablet (see Fig 8). The small difference was explained by the fast disintegration of the tablet, whereas the morphology of the film remained unchanged and the drug needed to diffuse through the loose CNF network in the film. The authors furthermore performed a diffusion study and found that the diffusivity of riboflavin through a plain cationic CNF/surfactant film was approx. one order magnitude slower compared to small molecules in water at 20°C (Svagan et al., 2016a). In the case of the films containing the poorly water soluble drug indomethacin (Lobmann et al., 2017), the drug was pipetted from an ethanol solution into a CNF suspension, which resulted in one fraction of the drug adsorbing the CNF fibres and one fraction of the drug precipitating as nanocrystals. The resulting indomethacin CNF suspension had a strong tendency to create a foam upon mixing and a degassing step

20

needed to be included in order to obtain a bubble-free film upon drying. As a result of the drug adsorption to the CNF fibres, the film contained the drug in an amorphous form (adsorbed to the CNF fibres) as well as nanocrystals (drug precipitate, approx. 100nm width and 100-5000 nm length) embedded in the lamellar structured CNF film. From these films, indomethacin was then completely released within 10-20 min. The immediate drug release from these films was somewhat surprising compared to the long-lasting sustained release over several days from the indomethacin-CNF films prepared by Kolakovic et al. (Kolakovic et al., 2012b) and Gao et al. (Gao et al., 2014). The authors suggested that the difference in release kinetics are due to (i) the amorphous nature of a large fraction of indomethacin, (ii) the remaining fraction of indomethacin being present as nanocrystals in the CNF film, (iii) the different types of CNF used, which might have resulted in differences in porosity and tortuosity, (iv) a one order of magnitude thinner CNF film in the study by Löbmann et al. (Lobmann et al., 2017). Furthermore, the drug dissolution rate from a film loaded with 21 wt% and 51 wt% indomethacin was faster compared to the pure crystalline drug. The 21wt% film showed in addition a faster dissolution compared to the pure amorphous drug, whereas the drug dissolution from the 51 wt% film was similar to the dissolution of the pure amorphous drug (Fig. 9). The authors suggested that CNF apparently enhanced the dissolution of the drug, however, the exact mechanism behind the dissolution enhancement remained unknown. The reason for the slower release from the 51wt% film may be a result of the higher fraction of crystalline drug compared to the 21wt% film, which had a higher fraction of amorphous content. 6.

CNF-based drug loaded foams

Dry CNF-based foams (CNF-based cellular solid materials) can be prepared using a number of different protocols and these typically involve mainly three steps; (i) the preparation of a gel (ii) creation of the 3-D structure, e.g. via freezing or solvent-exchange or foaming in the presence of surfactants (iii) removal of the solvent. The removal of the solvent can be done with e.g. supercritical drying, freeze-drying, oven-drying or ambient conditions. Depending on the chosen processing route, the nano- to macro-structure of the final material will vary which will in turn influence the final properties of the cellular solid, e.g porosity, mechanical properties, barrier properties and drug release kinetics. In the first studies on dry CNF-based

21

foams or aerogels, the materials were prepared via ice-templating/freeze-drying (Paakko et al., 2008; Svagan et al., 2008), as this processing technique takes advantage of the large quantities of water present in the CNF gel. An aerogel is an ultralight and open pore structure material derived from gels in which the liquid has been replaced with gas (Ulker and Erkey, 2014). During the freezing step, the ice crystals nucleate and grow and separate from the CNF and other solid components of the suspension, which will be confined to the space between the ice crystals (Svagan et al., 2010). The size and growth of the ice crystals will depend on the freezing rate and rapid freezing, e.g. by rapid immersion of the dispersion in liquid nitrogen, can also preserve the structure of the original dispersion. The ice-crystals are then sublimated in the freeze-drying step and a cellular solid material is created. CNF aerogels obtained via ice-templating/freeze-drying have been prepared by Valo et al. (Valo et al., 2013; Valo et al., 2011) in order to stabilize drug nanoparticles as outlined above. Supercritical drying has also been used to form, e.g. CNF-based aerogels (Kobayashi et al., 2014). The supercritical fluid, e.g. CO2, is used to exchange the solvent in the CNF-gel. The carbon dioxide is then removed in a depressurizing step (using a slow depressurization rate) to produce the CNF aerogel. As water is not miscible with carbon dioxide, the water in CNF based gel need to be exchanged to an intermediate solvent, e.g. ethanol (Kobayashi et al., 2014). Both supercritical drying and the ice-templating/freeze-drying technique give cellular solid materials with an interconnected and open cell structure. However, to obtain a closed-cell cellular solid material, a different processing route is required. As was described previously, CNF allow a Pickering stabilization of air bubbles and given the right circumstances allow the preparation of wet stable CNF foams. Upon drying of such a wet-stable foam, a dry closed-cell cellular solid material is obtained that is also buoyant (Cervin et al., 2016), see Fig 10a and b. One advantage with a porous drug delivery system with closed cells, from the perspective of the release kinetics of drugs, is that if the closed cells (that are filled with air) of the foam remain intact throughout the release, then there will be a decrease in the diffusivity of the small molecules through such a material. The decrease in diffusivity is due to a more tortuous diffusion path, see Figure 10c, and consequently, the release kinetics will be slowed down. The decrease in diffusivity through a closed-cell

22

foam compared to a solid material can be understood by considering the closed cells or “air-bubbles” as impermeable objects. If the air-bubbles are approximated as spheres, then the diffusivity in such a foam material, Dcomp, in relation to the diffusivity in solid material devoid of bubbles, D0, can then be expressed as: 𝑫𝒄𝒐𝒎𝒑 = 𝑫𝟎 ∙

𝟏−𝛟 𝟏+

𝛟 𝟐

(3)

where 𝛟 is the volume fraction of spheres. Equation (3) was derived for impermeable spheres that were periodically arrayed (Jonsson et al., 1986; Maxwell, 1881; Westrin and Axelsson, 1991). Indeed, if 𝛟 > 0, then Dcomp < D0, i.e. the diffusivity will decrease. For dry CNF based cellular solids the volume fraction of closed air-filled cells will be around 𝛟 ≈ 0.9 (Svagan et al., 2016a). On the other hand, if the closed cells are punctured, faster release kinetics would be expected due to the larger surface area in contact with the dissolution medium. Such release kinetics is also expected for cellular solid materials with interconnected cells (open cells), such as those obtained by ice-templating/freeze-drying or supercritical drying. A comparative study on the drug release kinetics from CNF-based cellular solids with fixed outer dimensions, density, porosity but with different cellular structures (open or closed-cells) is however up to date missing in the literature. The first study on dry drug-loaded CNF foams with closed-cell was recently published by Svagan et al. (Svagan et al., 2016a). The authors obtained the CNF foams by mixing a cationic CNF suspension with the edible surfactant lauric acid sodium salt as a foaming agent. In a second step, the water-soluble model drug riboflavin was added and incorporated into the wet-stable foam structure. Upon drying, cellular solid materials with drug loadings of up to 50 wt% were achieved. The foam structure was flexible, could be prepared in different shapes and thicknesses and could be cut in differently sized pieces (see Fig 10a). The drug was released from the CNF foams in a sustained matter and the release was depending on the dimensions of the foam, where thin foams (thickness 0.6 – 0.7 mm, specific surface area = area/volume= 33 cm-1) did release the drug content within 24 hours whereas thicker foams (thickness 8 mm, specific surface area of 6.25 cm-1) only released approx. 50% of the drug content in the same time (see Fig 8). The entrapped air bubbles remained intact during the dissolution and were responsible for the positive buoyancy for at least 24 hours (see Fig 10b). The drug had to diffuse along the cell walls of the foam, which represents a long and

23

tortuous diffusion path (see Fig 10c). The authors found that the diffusion coefficient of drug through the foam was more than one order of magnitude lower compared to the diffusivity in a film (see above). The positive buoyancy of the CNF foams may offer the possibility to obtain gastric retention as a CNF floating device. Since riboflavin is only absorbed from upper intestine, the authors suggested that such a CNF foam could potentially be used as a floating device for gastric retention in order to obtain a prolonged release of riboflavin to the narrow absorption side and hence obtain a higher bioavailability. Whereas in the study above, a surfactant was used to obtained a cellular solid material with CNF, Löbmann et al. were able to prepare foams with cationic CNF and the hydrophobic drug indomethacin, as a result of the positive molecular interactions between the drug and the CNF fibres (Lobmann et al., 2017). The drug indomethacin was added to the CNF suspension from an ethanol solution, upon mixing adsorbed to the CNF fibres and changed the surface energy of the CNF, and subsequently allowed the accumulation of the CNF fibres at the air/water interface. By this means, a wet stable foam was obtained via a Pickering stabilization and upon drying a solid cellular material was obtained. On the other hand, neither a neat CNF suspension nor indomethacin suspension alone was able to stabilize a foam, but only the combination of both. The positive interactions between the drug and the CNF fibres in the CNF suspension and the accumulation of the CNF/indomethacin complex at the air/water interface could further be shown using molecular dynamics simulations. As seen in Fig 11, all indomethacin molecules were adsorbed to the CNF-model either directly or in form of indomethacin clusters. Once adsorbed to the CNF-model, the indomethacin appeared irreversibly bound to the CNF (for the 50ns duration of the simulation). Subsequently, the formed CNF/indomethacin adduct migrated (also irreversibly for the 50ns duration of the simulation) towards the interface with an empty space, representing the air/water interface, in such a way that the hydrophobic surface of CNF partially covered with indomethacin was facing the empty space. Therefore, the authors concluded that the use of a hydrophobic drug that is able to modify the surface energy of the CNF sufficiently can be used as a foaming agent. Under the applied conditions, stable CNF/indomethacin foams could be prepared for a drug loading of 21 wt%. However, at a higher drug load of 51 wt%, the foam was no longer stable in the wet state and collapsed as a result of the higher fraction of ethanol and free indomethacin

24

in solution, which were competing with the surface modified CNF for the air/water interface (see further explanation below). Similar to the riboflavin foams above, a sustained drug release over a period of 24h was obtained for the 21 wt% foam. In a follow up study, the concept of preparing a sustained release CNF foam with buoyancy using a poorly soluble drug as foaming agent was expanded to the drug furosemide (Svagan et al.). The authors suggested that furosemide is a more realistic model drug for such a system, since a floating CNF foam could potentially be used as a gastric retentive system because furosemide has a very site specific absorption in the stomach and upper intestine. Hence, prolonged drug release to these absorption sites may increase the bioavailability of the poorly soluble drug. The CNF foam was stabilized by positive interactions between the drug and CNF, however, a stable dry foam was obtained for drug loadings at 21 and 51 wt%. In both cases, the drug was present in amorphous form (adsorbed to the CNF fibres and responsible for the surface modification of CNF and foaming properties) as well as in nanocrystalline form embedded in the CNF matrix. Compared to a marketed tablet formulation, both foams showed a diffusion controlled sustained release, with the foam with 21 wt% drug loading having a slightly faster release compared to the foam with 51 wt% drug loading. The differences in the release kinetics were assigned to the different dimensions of the foam pieces as well as the drug being mainly present in the amorphous form in the foam with 21 wt%, whereas the foam with 51 wt% contained a larger fraction of the drug in crystalline form. From the studies above, it is evident that a successful preparation of a drug loaded CNF foam seems to be dependent on the properties of drug itself but also the drug concentration. In the case of indomethacin, a CNF/drug suspension at high drug loading of 51 wt% had unfavourable properties in order to obtain a stable foam (due to large amounts of ethanol). On the other hand, for CNF/furosemide suspensions stable foams could be obtained even at drug loadings of 51 wt%. Hence, there seem to be conditions under which for a given drug one can successfully prepare wet-stable foams from CNF/drug suspensions that upon drying result in a cellular solid material. For this purpose, Bannow et al. performed a systematic study on the influence of processing parameters and drug content on the foam-ability and material structure of coprocessed cationic CNF and indomethacin suspensions followed by drying (Bannow et al., 2017). As

25

discussed above, the surface energy of the CNF needs to changed, i.e. hydrophobically modified with indomethacin, to enable a wet-stable foam. Hence, a certain amount of indomethacin appears necessary to achieve the required surface modification of the CNF fibres. However, if too much indomethacin was in solution, the fast adsorbing indomethacin is competing with the surface modified CNF for the air/water interface, and hence, is destabilizing the foam, leading to a foam coarsening and collapse. Since indomethacin is a weak acid (pKa = 4.42), the pH will play a crucial role for the ionization and the solubility of the drug. It was found that at pH values below the pKa of indomethacin, the amount of dissolved indomethacin was too low to sufficiently adsorb to nanofibres and modify the surface energy. Above the pKa of the drug, the solubility of indomethacin drastically increased and the preparation of wet-stable CNF foams was possible for drug loadings from 1 to 46 wt%. However, for the suspension with high drug loading (21 and 46 wt%) the pH range for successful foaming was very narrow (pH 3.9 – 6). At these high drug loadings, the amount of dissolved indomethacin exceeded a critical value (pH > 6) and the foaming was destabilized by a too large fraction of free indomethacin in solution, as described above. Hence, there is a very sensitive balance between the amount of indomethacin adsorbed to the CNF and free indomethacin in solution in order to prepare a successful indomethacin loaded CNF foam. Fig 12 summarizes the fine interplay between pH and indomethacin loading for the preparation of indomethacin/CNF cellular solid materials. 7.

Conclusions and Outlook

In the past, cellulose nanofibers and bacterial cellulose have shown to be promising candidates for a wide range of biomedical applications, from simple wound dressings to tissue engineering scaffolds. There are already nanocellulose products on the market for use in wound healing. However, in comparison very few studies (ca. 20 relevant articles, search words “drug” and “cellulose nanofiber” or “microfibrillated cellulose”, web of science, 2017 august) have investigated their potential as excipients for drug delivery of poorly-soluble drugs. The purpose with the present review was to present the many intrinsic properties of cellulose nanofibers that make CNFs interesting as excipient candidates. These properties include unique colloidal properties, high surface area-to-volume and surface chemistry, excellent mechanical properties, rheological properties, (oxygen) barrier

26

properties in the dry state, lack of toxicity and biodegradability. In particular, recent studies have demonstrated the presence of positive molecular interactions between CNF and poorly-soluble drugs as well as CNF and drug nanoparticles and in some cases enhanced apparent solubility of poorlysoluble drugs. Both fast and prolonged (i.e. tailored release kinetics) release of poorly-soluble drugs could be achieved in a facile way by selecting appropriate cellulose nanofibers and structuring the CNF-based material (particles, capsules, Pickering stabilized lipophilic droplets, films, aerogel, wetstable foams and cellular solid materials with closed cells). It was recently shown that cellular solid materials with closed cells could be self-assembled using a poorly-soluble drug and CNF only. These foams demonstrated prolonged drug release (hours or days) and could potentially be used as gastroretentive drug delivery system (for site-specific drug release) due to buoyant properties. On the other hand, a fast release could also be achieved by formulating aerogels containing nanoparticles or films from poorly-soluble drugs and the right cellulose nanofibers as matrix material. Prolonged release (for months) was also possible from films by selecting another CNF type. The unique molecular arrangement of the CNF also enables the stabilization of oil-droplets in water (Pickering stabilization) – such lipid based droplet could be used for delivery of poorly-soluble drugs. Indeed, CNF is truly a versatile excipient with respect to modulating drug release kinetics. However, to fully prove its application in the delivery of poorly-soluble drugs, further research is needed, aimed at better understanding its interaction with poorly-soluble drugs, and also the performance of such formulation in in vitro and in vivo disease state models. Acknowledgements A.J.S. would like to acknowledge SSF (ICA14-0045) and the Wallenberg Wood Science Centre for financial support. References Affdl, J.C.H., Kardos, J.L., 1976. The Halpin-Tsai equations: A review. Polymer Engineering & Science 16, 344-352. Aulin, C., Gallstedt, M., Lindstrom, T., 2010. Oxygen and oil barrier properties of microfibrillated cellulose films and coatings. Cellulose 17, 559-574. Babu, N.J., Nangia, A., 2011. Solubility Advantage of Amorphous Drugs and Pharmaceutical Cocrystals. Cryst Growth Des 11, 26622679.

27

Bannow, J., Benjamins, J.W., Wohlert, J., Löbmann, K., Svagan, A.J., 2017. Solid nanofoams based on cellulose nanofibers and indomethacin—the effect of processing parameters and drug content on material structure. International Journal of Pharmaceutics 526, 291-299. Battista, O.A., Smith, P.A., 1962. MICROCRYSTALLINE CELLULOSE. Industrial & Engineering Chemistry 54, 20-29. Cervin, N.T., Andersson, L., Ng, J.B.S., Olin, P., Bergstrom, L., Wagberg, L., 2013. Lightweight and Strong Cellulose Materials Made from Aqueous Foams Stabilized by Nanofibrillated Cellulose. Biomacromolecules 14, 503-511. Cervin, N.T., Johanson, E., Larsson, P.A., Wagberg, L., 2016. Strong, Water-Durable, and Wet-Resilient Cellulose Nanofibril-Stabilized Foams from Oven Drying. Acs Applied Materials & Interfaces 8, 11682-11689. Cervin, N.T., Johansson, E., Benjamins, J.W., Wagberg, L., 2015. Mechanisms Behind the Stabilizing Action of Cellulose Nanofibrils in Wet-Stable Cellulose Foams. Biomacromolecules 16, 822-831. Chawla, P.R., Bajaj, I.B., Survase, S.A., Singhal, R.S., 2009. Microbial Cellulose: Fermentative Production and Applications. Food Technology and Biotechnology 47, 107-124. Chevalier, Y., Bolzinger, M.A., 2013. Emulsions stabilized with solid nanoparticles: Pickering emulsions. Colloid Surface A 439, 23-34. Chungi, V.S., Dittert, L.W., Smith, R.B., 1979. Gastro-Intestinal Sites of Furosemide Absorption in Rats. International Journal of Pharmaceutics 4, 27-38. Colic, M., Mihajlovic, D., Mathew, A., Naseri, N., Kokol, V., 2015. Cytocompatibility and immunomodulatory properties of wood based nanofibrillated cellulose. Cellulose 22, 763-778. Cosgrove, D.J., 2005. Growth of the plant cell wall. Nat Rev Mol Cell Bio 6, 850-861. Dengale, S.J., Grohganz, H., Rades, T., Lobmann, K., 2016. Recent advances in co-amorphous drug formulations. Adv Drug Deliver Rev 100, 116-125. Eddershaw, P.J., Beresford, A.P., Bayliss, M.K., 2000. ADME/PK as part of a rational approach to drug discovery. Drug Discov Today 5, 409-414. Eichhorn, S.J., Dufresne, A., Aranguren, M., Marcovich, N.E., Capadona, J.R., Rowan, S.J., Weder, C., Thielemans, W., Roman, M., Renneckar, S., Gindl, W., Veigel, S., Keckes, J., Yano, H., Abe, K., Nogi, M., Nakagaito, A.N., Mangalam, A., Simonsen, J., Benight, A.S., 2010. Review: current international research into cellulose nanofibres and nanocomposites. Journal of Materials Science 45, 1-33. Endler, A., Persson, S., 2011. Cellulose Synthases and Synthesis in Arabidopsis. Molecular Plant (Oxford University Press / USA) 4, 199-211. Fernandes, A.N., Thomas, L.H., Altaner, C.M., Callow, P., Forsyth, V.T., Apperley, D.C., Kennedy, C.J., Jarvis, M.C., 2011. Nanostructure of cellulose microfibrils in spruce wood. Proceedings of the National Academy of Sciences 108, E1195-E1203. Fukuzumi, H., Saito, T., Iwamoto, S., Kumamoto, Y., Ohdaira, T., Suzuki, R., Isogai, A., 2011. Pore Size Determination of TEMPOOxidized Cellulose Nanofibril Films by Positron Annihilation Lifetime Spectroscopy. Biomacromolecules 12, 4057-4062. Gao, J.L., Li, Q., Chen, W.S., Liu, Y.X., Yu, H.P., 2014. Self-Assembly of Nanocellulose and Indomethacin into Hierarchically Ordered Structures with High Encapsulation Efficiency for Sustained Release Applications. Chempluschem 79, 725-731. Gestranius, M., Stenius, P., Kontturi, E., Sjoblom, J., Tammelin, T., 2017. Phase behaviour and droplet size of oil-in-water Pickering emulsions stabilised with plant-derived nanocellulosic materials. Colloid Surface A 519, 60-70. Habibi, Y., Lucia, L.A., Rojas, O.J., 2010. Cellulose Nanocrystals: Chemistry, Self-Assembly, and Applications. Chem Rev 110, 34793500. Hancock, B.C., Zograf, G., 1997. Characteristics and significance of the amorphous state in pharmaceutical systems. J Pharm Sci-Us 86, 1-12. Hargreaves, R., Ferreira, J.C.A., Hughes, D., Brands, J., Hale, J., Mattson, B., Mills, S., 2011. Development of aprepitant, the first neurokinin-1 receptor antagonist for the prevention of chemotherapy-induced nausea and vomiting. Ann Ny Acad Sci 1222, 40-48. Hauss, D.J., 2007. Oral lipid-based formulations. Adv Drug Deliver Rev 59, 667-676. Herrick, F.W., Casebier, R.L., Hamilton, J.K., Sandberg, K.R., 1983. Microfibrillated cellulose: morphology and accessibility, J. Appl. Polym. Sci.: Appl. Polym. Symp.;(United States). ITT Rayonier Inc., Shelton, WA. Hill, R.J., 2008. Elastic Modulus of Microfibrillar Cellulose Gels. Biomacromolecules 9, 2963-2966. Hoffman, A., Stepensky, D., Lavy, E., Eyal, S., Klausner, E., Friedman, M., 2004. Pharmacokinetic and pharmacodynamic aspects of gastroretentive dosage forms. International Journal of Pharmaceutics 277, 141-153. Hubbe, M.A., 2007. Flocculation and Redispersion of Cellulosic Fiber Suspensions: A Review of Effects of Hydrodynamic Shear and Polyelectrolytes. Bioresources 2, 296-331. Hunter, T.N., Pugh, R.J., Franks, G.V., Jameson, G.J., 2008. The role of particles in stabilising foams and emulsions. Advances in Colloid and Interface Science 137, 57-81. Iotti, M., Gregersen, O.W., Moe, S., Lenes, M., 2011. Rheological Studies of Microfibrillar Cellulose Water Dispersions. J Polym Environ 19, 137-145. ISO10993-5, 2009. Biological evaluation of medical devices. Part 5. Test for in vitro cytotoxicity. Jackson, J.K., Letchford, K., Wasserman, B.Z., Ye, L., Hamad, W.Y., Burt, H.M., 2011. The use of nanocrystalline cellulose for the binding and controlled release of drugs. Int J Nanomed 6, 321-330. Jonsson, B., Wennerstrom, H., Nilsson, P.G., Linse, P., 1986. Self-Diffusion of Small Molecules in Colloidal Systems. Colloid Polym Sci 264, 77-88. Jorfi, M., Foster, E.J., 2015. Recent advances in nanocellulose for biomedical applications. J Appl Polym Sci 132, n/a-n/a.

28

Kalashnikova, I., Bizot, H., Bertoncini, P., Cathala, B., Capron, I., 2013. Cellulosic nanorods of various aspect ratios for oil in water Pickering emulsions. Soft Matter 9, 952-959. Kalashnikova, I., Bizot, H., Cathala, B., Capron, I., 2011. New Pickering Emulsions Stabilized by Bacterial Cellulose Nanocrystals. Langmuir 27, 7471-7479. Kalashnikova, I., Bizot, H., Cathala, B., Capron, I., 2012. Modulation of Cellulose Nanocrystals Amphiphilic Properties to Stabilize Oil/Water Interface. Biomacromolecules 13, 267-275. Kaufman, G., Mukhopadhyay, S., Rokhlenko, Y., Nejati, S., Boltyanskiy, R., Choo, Y., Loewenberg, M., Osuji, C.O., 2017. Highly stiff yet elastic microcapsules incorporating cellulose nanofibrils. Soft Matter 13, 2733-2737. Klausner, E.A., Lavy, E., Stepensky, D., Friedman, M., Hoffman, A., 2002. Novel gastroretentive dosage forms: Evaluation of gastroretentivity and its effect on riboflavin absorption in dogs. Pharmaceut Res 19, 1516-1523. Kobayashi, Y., Saito, T., Isogai, A., 2014. Aerogels with 3D Ordered Nanofiber Skeletons of Liquid-Crystalline Nanocellulose Derivatives as Tough and Transparent Insulators. Angew Chem Int Edit 53, 10394-10397. Kolakovic, R., Laaksonen, T., Peltonen, L., Laukkanen, A., Hirvonen, J., 2012a. Spray-dried nanofibrillar cellulose microparticles for sustained drug release. International Journal of Pharmaceutics 430, 47-55. Kolakovic, R., Peltonen, L., Laaksonen, T., Putkisto, K., Laukkanen, A., Hirvonen, J., 2011. Spray-Dried Cellulose Nanofibers as Novel Tablet Excipient. Aaps Pharmscitech 12, 1366-1373. Kolakovic, R., Peltonen, L., Laukkanen, A., Hellman, M., Laaksonen, P., Linder, M.B., Hirvonen, J., Laaksonen, T., 2013. Evaluation of drug interactions with nanofibrillar cellulose. Eur J Pharm Biopharm 85, 1238-1244. Kolakovic, R., Peltonen, L., Laukkanen, A., Hirvonen, J., Laaksonen, T., 2012b. Nanofibrillar cellulose films for controlled drug delivery. Eur J Pharm Biopharm 82, 308-315. Laitinen, R., Lobmann, K., Strachan, C.J., Grohganz, H., Rades, T., 2013. Emerging trends in the stabilization of amorphous drugs. International Journal of Pharmaceutics 453, 65-79. Lange, J., Wyser, Y., 2003. Recent innovations in barrier technologies for plastic packaging - a review. Packaging Technology and Science 16, 149-158. Larsson, P.A., Gimaker, M., Wagberg, L., 2008. The influence of periodate oxidation on the moisture sorptivity and dimensional stability of paper. Cellulose 15, 837-847. Lasseuguette, E., Roux, D., Nishiyama, Y., 2008. Rheological properties of microfibrillar suspension of TEMPO-oxidized pulp. Cellulose 15, 425-433. Lobmann, K., Wohlert, J., Mullertz, A., Wagberg, L., Svagan, A.J., 2017. Cellulose Nanopaper and Nanofoam for Patient-Tailored Drug Delivery. Adv Mater Interfaces 4. Maxwell, J.C., 1881. A treatise on electricity and magnetism. Clarendon press. Mazeau, K., Wyszomirski, M., 2012. Modelling of Congo red adsorption on the hydrophobic surface of cellulose using molecular dynamics. Cellulose 19, 1495-1506. Medronho, B., Romano, A., Miguel, M.G., Stigsson, L., Lindman, B., 2012. Rationalizing cellulose (in)solubility: reviewing basic physicochemical aspects and role of hydrophobic interactions. Cellulose 19, 581-587. Merisko-Liversidge, E., Liversidge, G.G., Cooper, E.R., 2003. Nanosizing: a formulation approach for poorly-water-soluble compounds. Eur J Pharm Sci 18, 113-120. Missoum, K., Bras, J., Belgacem, M.N., 2012. Water Redispersible Dried Nanofibrillated Cellulose by Adding Sodium Chloride. Biomacromolecules 13, 4118-4125. Muellertz, A., Ogbonna, A., Ren, S., Rades, T., 2010. New perspectives on lipid and surfactant based drug delivery systems for oral delivery of poorly soluble drugs. J Pharm Pharmacol 62, 1622-1636. Nair, K.G., Dufresne, A., 2003a. Crab shell chitin whisker reinforced natural rubber nanocomposites. 1. Processing and swelling behavior. Biomacromolecules 4, 657-665. Nair, K.G., Dufresne, A., 2003b. Crab shell chitin whisker reinforced natural rubber nanocomposites. 2. Mechanical behavior. Biomacromolecules 4, 666-674. Nair, K.G., Dufresne, A., Gandini, A., Belgacem, M.N., 2003. Crab shell chitin whiskers reinforced natural rubber nanocomposites. 3. Effect of chemical modification of chitin whiskers. Biomacromolecules 4, 1835-1842. Nishino, T., Takano, K., Nakamae, K., 1995. ELASTIC-MODULUS OF THE CRYSTALLINE REGIONS OF CELLULOSE POLYMORPHS. Journal of Polymer Science Part B-Polymer Physics 33, 1647-1651. Noguchi, Y., Homma, I., Matsubara, Y., 2017. Complete nanofibrillation of cellulose prepared by phosphorylation. Cellulose 24, 1295-1305. Paakko, M., Vapaavuori, J., Silvennoinen, R., Kosonen, H., Ankerfors, M., Lindstrom, T., Berglund, L.A., Ikkala, O., 2008. Long and entangled native cellulose I nanofibers allow flexible aerogels and hierarchically porous templates for functionalities. Soft Matter 4, 2492-2499. Paulraj, T., Riazanova, A.V., Yao, K., Andersson, R.L., Mullertz, A., Svagan, A.J., 2017. Bioinspired Layer-by-Layer Microcapsules Based on Cellulose Nanofibers with Switchable Permeability. Biomacromolecules 18, 1401-1410. Pei, A., Butchosa, N., Berglund, L.A., Zhou, Q., 2013. Surface quaternized cellulose nanofibrils with high water absorbency and adsorption capacity for anionic dyes. Soft Matter 9, 2047-2055. Peltonen, L., Strachan, C., 2015. Understanding Critical Quality Attributes for Nanocrystals from Preparation to Delivery. Molecules 20, 22286-22300.

29

Petersen, N., Gatenholm, P., 2011. Bacterial cellulose-based materials and medical devices: current state and perspectives. Applied Microbiology and Biotechnology 91, 1277-1286. Pickering, S.U., 1907. CXCVI.-Emulsions. Journal of the Chemical Society, Transactions 91, 2001-2021. Porter, C.J.H., Pouton, C.W., Cuine, J.F., Charman, W.N., 2008. Enhancing intestinal drug solubilisation using lipid-based delivery systems. Adv Drug Deliver Rev 60, 673-691. Pääkkö, M., Ankerfors, M., Kosonen, H., Nykänen, A., Ahola, S., Österberg, M., Ruokolainen, J., Laine, J., Larsson, P.T., Ikkala, O., Lindström, T., 2007. Enzymatic Hydrolysis Combined with Mechanical Shearing and High-Pressure Homogenization for Nanoscale Cellulose Fibrils and Strong Gels. Biomacromolecules 8, 1934-1941. Saarikoski, E., Saarinen, T., Salmela, J., Seppala, J., 2012. Flocculated flow of microfibrillated cellulose water suspensions: an imaging approach for characterisation of rheological behaviour. Cellulose 19, 647-659. Saito, T., Kimura, S., Nishiyama, Y., Isogai, A., 2007. Cellulose nanofibers prepared by TEMPO-mediated oxidation of native cellulose. Biomacromolecules 8, 2485-2491. Saito, T., Kuramae, R., Wohlert, J., Berglund, L.A., Isogai, A., 2013. An Ultrastrong Nanofibrillar Biomaterial: The Strength of Single Cellulose Nanofibrils Revealed via Sonication-Induced Fragmentation. Biomacromolecules 14, 248-253. Schenzel, K., Almlöf, H., Germgård, U., 2009. Quantitative analysis of the transformation process of cellulose I → cellulose II using NIR FT Raman spectroscopy and chemometric methods. Cellulose 16, 407-415. Stocco, A., Drenckhan, W., Rio, E., Langevin, D., Binks, B.P., 2009. Particle-stabilised foams: an interfacial study. Soft Matter 5, 22152222. Stocco, A., Rio, E., Binks, B.P., Langevin, D., 2011. Aqueous foams stabilized solely by particles. Soft Matter 7, 1260-1267. Sturcova, A., Davies, G.R., Eichhorn, S.J., 2005. Elastic modulus and stress-transfer properties of tunicate cellulose whiskers. Biomacromolecules 6, 1055-1061. Svagan, A.J., Benjamins, J.W., Al-Ansari, Z., Bar Shalom, D., Mullertz, A., Wagberg, L., Lobmann, K., 2016a. Solid cellulose nanofiber based foams - Towards facile design of sustained drug delivery systems. J Control Release 244, 74-82. Svagan, A.J., Jensen, P., Dvinskikh, S.V., Furo, I., Berglund, L.A., 2010. Towards tailored hierarchical structures in cellulose nanocomposite biofoams prepared by freezing/freeze-drying. J Mater Chem 20, 6646-6654. Svagan, A.J., Koch, C.B., Hedenqvist, M.S., Nilsson, F., Glasser, G., Baluschev, S., Andersen, M.L., 2016b. Liquid-core nanocelluloseshell capsules with tunable oxygen permeability. Carbohyd Polym 136, 292-299. Svagan, A.J., Musyanovych, A., Kappl, M., Bernhardt, M., Glasser, G., Wohnhaas, C., Berglund, L.A., Risbo, J., Landfester, K., 2014. Cellulose Nanofiber/Nanocrystal Reinforced Capsules: A Fast and Facile Approach Toward Assembly of Liquid-Core Capsules with High Mechanical Stability. Biomacromolecules 15, 1852-1859. Svagan, A.J., Müllertz, A., Löbmann, K., Floating solid cellulose nanofibre nanofoams for sustained release of the poorly soluble model drug furosemide. J Pharm Pharmacol, n/a-n/a. Svagan, A.J., Samir, M.A.S.A., Berglund, L.A., 2007. Biomimetic polysaccharide nanocomposites of high cellulose content and high toughness. Biomacromolecules 8, 2556-2563. Svagan, A.J., Samir, M.A.S.A., Berglund, L.A., 2008. Biomimetic foams of high mechanical performance based on nanostructured cell walls reinforced by native cellulose nanofibrils. Adv Mater 20, 1263-+. Tanquary, A.C., Lacey, R.E., Southern Research Institute (Birmingham Ala.), 1974. Controlled release of biologically active agents. Plenum Press, New York,. TAPPI, WI 3021. Standard Terms and Their Definition for Cellulose Nanomaterials. Tatsumi, D., Ishioka, S., Matsumoto, T., 2002. Effect of fiber concentration and axial ratio on the rheological properties of cellulose fiber suspensions. J Soc Rheol Jpn 30, 27-32. Thoorens, G., Krier, F., Leclercq, B., Carlin, B., Evrard, B., 2014. Microcrystalline cellulose, a direct compression binder design environment-A review. Int J Pharmaceut 473, 64-72. Turbak, A.F., Snyder, F.W., Sandberg, K.R., 1983. Microfibrillated cellulose, a new cellulose product: Properties, uses and commercial potential., 37 ed. Wiley, New York. Ulker, Z., Erkey, C., 2014. An emerging platform for drug delivery: Aerogel based systems. J Control Release 177, 51-63. Valo, H., Arola, S., Laaksonen, P., Torkkeli, M., Peltonen, L., Linder, M.B., Serimaa, R., Kuga, S., Hirvonen, J., Laaksonen, T., 2013. Drug release from nanoparticles embedded in four different nanofibrillar cellulose aerogels. Eur J Pharm Sci 50, 69-77. Valo, H., Kovalainen, M., Laaksonen, P., Hakkinen, M., Auriola, S., Peltonen, L., Linder, M., Jarvinen, K., Hirvonen, J., Laaksonen, T., 2011. Immobilization of protein-coated drug nanoparticles in nanofibrillar cellulose matrices-Enhanced stability and release. J Control Release 156, 390-397. van den Berg, O., Capadona, J.R., Weder, C., 2007. Preparation of Homogeneous Dispersions of Tunicate Cellulose Whiskers in Organic Solvents. Biomacromolecules 8, 1353-1357. Vasconcelos, T., Marques, S., das Neves, J., Sarmento, B., 2016. Amorphous solid dispersions: Rational selection of a manufacturing process. Adv Drug Deliver Rev 100, 85-101. Westrin, B.A., Axelsson, A., 1991. Diffusion in Gels Containing Immobilized Cells - a Critical-Review. Biotechnol Bioeng 38, 439-446. Viet, D., Beck-Candanedo, S., Gray, D., 2007. Dispersion of cellulose nanocrystals in polar organic solvents. Cellulose 14, 109-113. Wågberg, L., Decher, G., Norgren, M., Lindström, T., Ankerfors, M., Axnäs, K., 2008. The Build-Up of Polyelectrolyte Multilayers of Microfibrillated Cellulose and Cationic Polyelectrolytes. Langmuir 24, 784-795.

FIGURE CAPTIONS:

30

Fig. 1 (A) Commercially available cubes of nata-de-coco (marketed as coconut gel) made from bacterial cellulose. Nata-de-coco is a dessert that originates from the Philippines. In (B), a suspension of anionic cellulose nanofibers in water, the solid content is 0.9 wt%. The anionic CNF was prepared via a 2,2,6,6,tetramethylpiperidine-1-oxyl radical (TEMPO)-mediated oxidation of pulp fibres from soft wood. (C) Atomic force microscopy image of TEMPO-CNF, adapted with permission from ref (Svagan et al., 2016b). (D) A transparent film made from TEMPO-CNF, reprinted with permission from (Fukuzumi et al.,2009). Copyright 2009 American Chemical Society. Fig. 2 (A) Simplified image, showing the arrangement of fibrils, microfibrils and cellulose in the cell walls of plants, from ref (Moore, 1998). Copyright 1998 The McGraw-Hill Companies, Inc. (B) Model of the crosssection of a microfibril (from wood), showing the orientation and packing of cellulose molecule chains. (C) Proposed models of microfibrils from spruce. (B) and (C) are adapted from Ref (Fernandes et al., 2011). Copyright 2011 National Academy of Sciences. (D) Proposed model of the adsorption of a cellulose nanocrystal/microfibril at the oil/water interface, reprinted with permission from (Kalashnikova et al., 2012). Copyright 2012 American Chemical Society. Fig. 3 Logarithmic plot of the shear stress, σ, as a function of shear rate, 𝛾̇, for a CNF suspension, adapted from ref (Tatsumi et al., 2002). Copyright 2002 The Society of Rheology, Japan. The CNF is derived from purified wood pulp. Different CNF concentrations, 0.05 - 0.5 wt%, were tested. Fig. 4 The predicted composite modulus (Ec), for a composite material, as a function of the aspect ratio (L/d) of the fibres (reinforcement) with different fibre moduli (40-140 GPa). The values are predicted from the Halpin-Tsai equation which assumes a unidirectional composite sample, with no fibre-fibre interactions. The matrix is polypropylene and the volume percent of filler is 50 vol%. The figure is adapted with permission from ref (Eichhorn et al., 2010). Fig. 5 Transmission electron microscopy image of crystalline itraconazole nanoparticles adsorbed to CNF in suspension. Each nanoparticle is associated with CNF. In order to facilitate the adsorption of the nanoparticles to the CNF, the nanoparticles have been coated with genetically modified hydrophobin fusion proteins with cellulose binding domains. The figure is adapted with permission from Ref (Valo et al., 2011). Fig. 6 Oil-core capsules with capsule walls consisting of (short) cellulose nanofibers and cellulose nanocrystals and a poly urea/urethane matrix, from ref (Svagan et al., 2014). The CNF/CNC both stabilized the oil/water interface (Pickering stabilization) during the co-assembly of the capsules, and improved the mechanical properties (Young’s modulus) of the capsule wall in the resulting capsules. Fig. 7 Long-lasting sustained release from beclomethasone diproprionate loaded CNF films with different drug loadings: 20 wt% (BECLO20), 30 wt% (BECLO30) and 40 wt% (BECLO40). The figure is adapted with permission from Ref (Kolakovic et al., 2012). Fig. 8 Drug release from a commercial riboflavin tablet (10 mg riboflavin) as well as riboflavin loaded CNF films (containing 14 wt% drug) and foams (14 wt% or 51 wt%) in Fasted-state simulated gastric fluid (FaSSGF, pH 1.6, pepsin content: 450 U mL-1) at 37 C. The amount (in mg) of riboflavin in all CNF-based samples was equal. From the commercial tablet and the CNF film, an immediate release of the drug can be seen. From the CNF foams, a sustained release was observed. The sample dimensions of the foams influenced the release kinetics, as shown in the release of riboflavin from the foam, 14 wt% (specific surface area of 33 cm-1) and thick foam, 14 wt% (specific surface area of 6.25 cm-1). As a result of the diffusion controlled drug release, the thick foam with a smaller specific surface area showed a slower drug release. On the other hand, when the foams had approx. the same thickness but different drug loading, i.e.

31

foam, 14 wt% and foam, 50 wt%, the drug release was found to be similar. The figure is adapted with permission from Ref (Svagan et al., 2016a). Fig. 9 Intrinsic dissolution profiles of indomethacin loaded CNF-films with 21 wt% and 51wt% drug load compared to those of crystalline indomethacin (-form) and amorphous indomethacin. The arrows indicate the timepoint when all drug was released from the films. The figure is adapted and modified with permission from Ref (Lobmann et al., 2017). Fig. 10 (A) Riboflavin loaded CNF foams (14 wt%) prepared in different shapes. (B) Positive buoyancy of a representative CNF foam. (C) Simplified illustration of the long and tortuous diffusion paths of the drug riboflavin along the wetted cell walls of the CNF foam. The figure is adapted with permission from Ref (Svagan et al, 2016a). Fig. 11 Snapshot from a molecular dynamic simulation showing the final configuration of an aqueous phase/empty space interface simulation. The indomethacin molecules are coloured red (ionic) and yellow (non-ionic) representing the charge distribution of the drug at the foaming pH 4.9 from the experimental setup. The CNF is coloured in green and its cationic modification is coloured in blue. Ethanol and water are coloured in grey and illustrated as light blue dots, respectively. The figure is adapted and modified with permission from Ref. (Lobmann et al., 2017) Fig. 12 Chart showing the pH conditions at different indomethacin contents that can be used for a successful preparation of indomethacin/CNF foams (green dots). The red dots indicate conditions at which no stable foam structure could be obtained. The figure is adapted with permission from Ref (Bannow et al., 2017).

32

Figr-1

33

Figr-2

34

Figr-3

35

Figr-4

36

Figr-5

37

Figr-6

38

Figr-7

39

Figr-8

40

Figr-9

41

Figr-10

42

Figr-11

43

Figr-12

44