RICERCA ITALIANA     

Multiscale modelling and development of process reactors for polymeric nanoparticle precipitation

Università degli Studi di Palermo

Abstract

Nanoparticle production is currently receiving a great research interest. This is due to the wide potential application field, which includes the production of adhesives, pigments, catalysts as well as of new effective drugs.
The present research programme is aimed at contributing to the field of nanoparticle production technology through the three main workpackages listed below.

1) Nanoparticle-production process and apparatus development. Five different precipitation reactors (standard stirred tank, static mixer, impinging stream , vortex and Couette cell reactors) will be experimentally investigated and critically compared by three of the research units.

2) Set up of advanced models aimed at simulating the performance of precipitation reactors. The models developed will be validated by comparison with the experimental data produced under workpackage 1. Once validated, the models developed will be employed for process and apparatus improvement and scale up.

3) One of the research units will set up a procedure for obtaining polymeric nanoparticles for targeted drug delivery. The procedure will be passed to the other units that in this way will be able to test the performance and viability of the precipitation reactors investigated and the modelling capabilities developed, on a real nanoparticle production process. <<<

Principal Investigator

Alberto Brucato Università degli Studi di PALERMO

Research Objectives

The general aim of the project is that of developing advanced models and reliable apparatuses for the production of nanoparticles with targeted characteristics.
Modelling nanoparticle production processes is a very complex task, as it involves the set up on a number of sub-models aimed at describing the many elementary processes that contribute to the final result, at the many time and length scales involved.

A further complication arises from the number of processes devised so far for nanoparticle production, with the related need to set up specific sub-models. As a consequence, the choice was made here to orientate the efforts towards a specific class of nanoparticle production processes, namely that of polymeric nanoparticles for pharmaceutical applications. This choice was suggested by both the scientific interest in elucidating a still quite obscure range of phenomena as well as in view of the great application potential of such processes.

In comparison with standard preparations, nano-dispersions of pharmaceutically active organic compounds show impressive increase in dissolution rate, improvement in biological response, and the possibility of highly selective physiological action, which are achievable only with particle sizes in the middle or lower nanometer range (50 - 500 nm). The nanoparticles employed for therapeutic applications are often made of a polymeric structure that contains the active principle. Two main classes are usually distinguished: nanospheres or pseudo-latex particles, formed by a homogeneous polymeric structure in which the active compound is bound by absorption or adsorption and nanocapsules, where the polymer forms a hollow spherical shell, which encloses the active principle (often dissolved in an oily phase) .
The advantages of the therapeutic use of these vectors include the ability of providing substances that are poorly soluble in water, the modification of the biological affinity of the active principle, the possibility of controlled release of the active compound, the improvement of the therapeutic efficiency by increasing the amount of drug that reaches the specific action site and decreasing the concentration in non-specific sites. Finally, by including particular molecular groups in the polymeric structure, these nanoparticles may be protected from degradation or inactivation in the body .
Clearly, the set up of a suitable formulation for nanoparticluate drugs require an expertise not available to the engineers involved in the project. It is for this reason that a pharmaceutical chemistry unit (Turin-University) with specific expertise has been involved in the project.

The nanoparticle production technique considered in this project is the "solvent displacement method" that consists in mixing an aqueous solution with a hydrophilic organic solvent in which polymer and active principle have been previously dissolved. The difference in solubility of the organics between the initially pure solvent and the final solvent-water solution induces precipitation .
The method was devised originally by the group of Fessi at Claude Bernard University (Lyon) and it has been known for more than fifteen years. During this period the process has been adapted to different organic actives and different polymers, but its set-up and tuning is still essentially empirical, based on a trial and error procedure coupled with qualitative reasoning to select operating conditions and reactor type. The ability to predict quantitatively the outcome of the process is still lacking, as well as the understanding required to design and control the process in order to fit the properties of the product for specific applications. This situation is partly due to the fact that the phenomena that affect more significantly the final process result occur at very different length and time scales: molecular scale (i.e., molecular interactions among solute molecules as well as with solvent molecules), nanoscale (i.e., clusters and nuclei formation, agglomeration phenomena), microscale (i.e., micromixing issues), production apparatus length scale (i.e., macromixing issues). Therefore the knowledge needed to master organic nanoparticle precipitation requires a multidimensional (hence multidisciplinary) approach. Such an approach should be set up if rational and reliable production methods were to be developed, as well as for reliably scaling up the results obtained at the laboratory scale to the industrial scale .

Issues related to the process are crucial too, in fact they determine the success of co-precipitation, with the production of particles formed by both polymer and active principle (rather than the formation of particles made by polymer only and by active principle only), as well as the size of the obtained product. The process is characterised by the interplay of different simultaneous phenomena, including precipitation of dissolved macromolecules, precipitation and growth of the organic active principle. Moreover, the high reaction rates make the process mixing sensitive, that is, different mixing histories result in different properties of the final product .

Various reactor configurations will be tested in the project, in order to compare their performances and optimise the process. An additional aim is to develop a methodology that, through accurate modelling of mixing and of the mechanisms of precipitation, growth and aggregation of the nanoparticles, allows one to identify the operating conditions that lead to production of pseudo-latex particles with desired characteristics. This methodology can also be used for the optimisation of their properties, resulting in a reliable method for simulation-based process design valid for polymer-based nanoparticles. The use of several reactor types will help us in this task, allowing the investigation of particular aspects of the process. We plan to obtain a general approach for process and reactor design to tailor the operation to specific targets for the product (e.g., size distribution, ratio between active and polymer, type of coating, by selecting properly the operating conditions and the reactor configurations.

It is finally worth remarking that although the project activities will be oriented to a well defined scope, most of the sub-models and knowledge developed (flow fields, macro and micro mixing, population balance solutions etc) is susceptible of application to many other fields, either involving or not precipitation reactions, and should therefore be regarded as worthwhile objectives on their own. <<<

Timescale

24 months

National and international background

As already specified, this project is aimed at devising processes, apparatuses and reliable simulation models for the set up of nanoparticle “process design” procedures and at testing the methodologies developed on a real benchmarck as the production of a new nanoparticle drug.

In order to increase the efficiency of therapeutic agents, the active principle should reach its target organ (or tissue) quickly and in sufficient amount, and it should remain there for a long time, avoiding distribution to other areas, thus reducing toxic effects (side effects). Unfortunately, drug distribution throughout the body is often slow, and the fraction of the drug that reaches the target is inadequate; doses must be increased and thus side effects become significant. This problem is typical of anti-tumour drugs. A recent innovation to increase drug selectivity towards cancer cells while reducing toxicity on normal tissues is the association of the drug with a nano-particulate carrier, usually made of polymer, able to deliver the drug and to shield it from degradation by the host .
These nanoparticles are classified either by the nature of polymer or by the preparation method (Couvreur et al, 1995) .
As said in the previous section, the objective of this research project is threefold: we intend to produce a nanoparticulate product well suited for optimal drug delivery in anti-tumour application, to set-up an efficient production process and to use the obtained expertise to develop a method for process design that can be subsequently applied to generic polymer-based nanoparticles .

The investigation will be oriented towards polymeric nanoparticles made of cyanoacrylates (e.g. poly alkyl cyanoacrylate, PACA). This class of materials has been used for a long time in the preparation of polymeric carriers with features of controlled drug release because of high reactivity of the monomers, biodegradability and, most importantly, very low toxicity (Couvreur et al, 1979). The efficacy of nanoparticles in intravenous administration may be limited if these are recognised and removed by macrophages of the reticuloendothelial system organs, such as liver, spleen, bone marrow. To increase the blood survival time, the nanoparticle surface must be shielded with highly hydrophilic, flexible and non-ionic polymers, such as polyethylene glycols (PEG). The hydrophilic region around the nanoparticle created by PEG prevents the adsorption of proteins from plasma, reducing the number of particles that can be recognised by macrophages.
Another advanced approach to reach the target with increased efficiency is active targeting (Brannon-Peppas &amp; Blanchette, 2004), which consists in equipping the nanoparticle surface with "agents" that are recognised by specific receptors expressed only by tumours, or over-expressed by them. One of these targeting agents, studied in depth in recent years, is folic acid: high levels of expression of folic acid have been observed in tumour tissues (ovarian, kidney, uterus, testis, brain, colon and lung cancers), as shown by Garin-Chesa et al (1993). Furthermore, the receptor in normal tissues appears unapproachable by intravenous-administered systems, thus minimising side effects. Several different research groups have used the folate receptor as tumour target so far, employing different anticancer or diagnostic agents (Stella et al., 2000; Sudimack &amp; Lee, 2000; Wang &amp; Low, 1998) .

The nanoparticles will be prepared by using a method known as "nanoprecipitation" or "solvent-displacement method". The polymer and the organic active principle are dissolved in a hydrophilic solvent (acetone, THF, low chain alcohols) and subsequently the solvent is mixed with water. The smaller solubility of the polymer and of the organic drug in water induces birth and growth of the nanoparticles. Nanocapsules are prepared in a similar way: a suitable oil in which the active compound is highly soluble is added to the solution phase. Mixing with water in this case leads to a nanoemulsion and core-shell particles are then formed with the active compound-oil solution as core .
The industrial advantages of this process rest upon the use of water miscible, toxicologically acceptable solvents, differently from other methods that adopt lipophilic or amphiphilic solvents (emulsification-evaporation, emulsification-diffusion) that could have toxic action if not fully removed during the formulation. The studied process appears superior to the production by supercritical fluid too, both because of the simpler equipment and of the superior capacity of producing particles in the lower and middle nanometer range (Horn &amp; Rieger, 2001) .

The precipitation from a molecular solution by solvent displacement was developed originally by the group of Fessi at Claude Bernard University at Lyon (Fessi et al, 1989; Thioune et al, 1997) and it has been known for more than fifteen years, but the complexity of the phenomena involved still makes the development of the relevant methods a highly empirical task. The considered process is characterised by the interplay of several simultaneous phenomena: the polymeric support is primarily precipitated, through physical condensation of dissolved macromolecules, followed by precipitation and growth of the organic active principle. As outlined in the reviews published on this subject, most of the information reported in the scientific literature concerns inorganic nano-particles, while little is known about the mechanisms of particle formation of organic systems. Actually the physico-chemical elementary processes of particle formation and the bonding of dissolved active compounds onto or into the particulate phase are still essentially unexplained (Horn &amp; Rieger, 2001; Texter, 2001). Moreover, also the mixing of the reactants is an important part of the process that needs to be controlled. Intense mixing, however, can only be achieved by turbulent flow, which is notorious for its erratic and unpredictable nature. In addition, the process involves phenomena occurring at very different scales, ranging from the nano-scale of the particles to the macro-scale of the reactor, which require different theoretical approaches .

As far as the flow field and the concentration distribution at macroscale level are concerned, in recent years the solution of Reynolds averaged equations by means of Computational Fluid Dynamics (CFD) started providing a sound way for modelling turbulent mixing issues, although results are often affected by lacks of accuracy and of universality of turbulence models (Ciofalo et al, 1996) .
Steps forwards may be obtained by the adoption of Large Eddy Simulation (LES) turbulence models (Alcamo et al, 2005), but at the expense of a very large computational demand. An even higher computational load is required by Direct Numerical Simulation (DNS) of turbulence. Nevertheless, for the smallest reactor configurations used in the project and for conditions of moderately turbulent flow, this approach can be affordable. DNS does not use any closure assumption for turbulence and therefore it should be regarded as having the same reliability of a true experiment, but giving results with detail and level of information highly superior (Moin &amp; Mahesh, 1998). The classical DNS approach rely on pseudospectral methods and is applicable to simple geometric configurations only. More recent approaches make use of finite difference algorithms (Verzicco &amp; Orlandi, 1996) and have been applied to complex geometries (Cerbelli et al 2001; Marchioli et al, 2003; Sbrizzai et al, 2004). In addition coupling of DNS with Lagrangian methods allows accurate calculation of particle dispersion, by direct integration of the equation of motion (Maxey &amp; Riley, 1983), provided that Brownian effects are included for submicron particles (Sbrizzai et al, 2005) .

The need for the characterisation of micromixing derives from the fact that the reactions of precipitation and condensation are very fast and occur before the reactants are mixed uniformly at the molecular scale (Baldyga &amp; Bourne, 1999); the reaction-diffusion mechanism takes place at the so-called Batchelor length scale, which can be even smaller than one micrometer for the systems we are interested in. For liquid systems this length scale is approximately 30 times smaller than the Kolmogorov length scale resolved by DNS. This fact excludes any possibility of direct approach to micromixing, which, as a consequence, has to be described by simplified modelling. Subgrid-scale models have been developed for implementation in CFD codes for single-phase reaction schemes (Baldyga, 2004) as well as for precipitation reactions (Marchisio et al, 2001a) .
Accurate micro-mixing models describe the fluctuating and chaotic behaviour of turbulence by means of the probability density function (pdf). The "full pdf" method solves directly the transport equation of the pdf adopting stochastic techniques. They are very accurate, but often excessively demanding in terms of computational resources and time. The computational load can be reduced while retaining good accuracy by prescribing a proper functional form for the pdf ("presumed pdf"). These methods include beta-pdf (Baldyga &amp; Orchiuch, 2001) and finite mode-pdf (Fox, 1998). The latter method has been extensively validated by comparison with full pdf results and with experimental data concerning several reacting systems under different operating conditions (Marchisio et al, 2001b; Marchisio &amp; Barresi, 2003) .

Another element of primary importance is the description of the evolution of the solid precipitate. This can be achieved by solving the population balance equation (PBE). This equation is a continuity statement for the property distribution function of the particles. It is usually written in terms of one or more properties of the particulate system (the internal coordinates), such as particle size or volume. For our case, where the population is formed by polymer-based nanoparticles of an active compound, the population has to be characterised in terms of at least two internal coordinates, which characterise the amount of polymer and the amount of active principle .
The solution of the PBE is normally obtained by discretisation of the population density function into a number of classes that are simply treated as additional passive scalars. The number of classes to be used in a typical problem with a single internal coordinate is around 50-100 (Vanni, 2000), which is intractable inside a CFD code. The situation is even worse with two or more internal coordinates. The population balance equation can however be solved through the method of moments, which is based on the solution of the transport equations of the moments of the particle property distribution. The original version of the method, formulated in the .60s (Hulburt &amp; Katz, 1964), suffered from the so-called closure problem, which has been solved only recently by using a quadrature approximation (Marchisio et al, 2003; McGraw, 1997). The use of this method reduces the number of solved equations from the 50-100 required by the method of classes to 8-10, with a comparable accuracy on the integral properties of the population, thus allowing its implementation in CFD codes. Very recently the method has been extended to problems with two internal coordinates (Marchisio &amp; Fox, 2005) .

Concerning the elementary particle evolution processes, while the theoretical framework for Brownian aggregation is relatively well established, one has to rely on often quite uncertain empirical kinetic equations for the rates of nucleation, growth and condensation, as there is presently no way of deducing these rates from independent (e.g. thermodynamic) data and molecular features. This situation depends on the complexity of the phenomena taking place at nanoscale level and the difficulty of setting up suitable observation techniques due to the shortness of the length and time scales involved. Probably significant advancements in this field will be made when molecular dynamic simulations tools will become powerful enough to properly deal with the above phenomena, but meanwhile the set up of euristic kinetic models based on experimental evidence has little alternative for design and development purposes .
As shown by Johnson (2003), the success of the synthesis of polymer based nanoparticles requires adequate matching of the characteristic times of the elementary steps involved, which include mixing. Although bench scale laboratory preparations of polymer based nanoparticles are often performed in beakers agitated by magnetic stirrers, industrialisation and scaling of the process requires much better control of mixing. This can be obtained only with reactor configurations that give reproducible and easily scalable control of mixing times. Unfortunately no systematic comparison of reactor configurations has been done so far for polymer based nanoparticles and therefore this will be one of the tasks of our project. The starting point is in classical nucleation theory, which indicates the need of high supersaturation (and hence of very intense mixing) for producing particles in the submicron range, by favouring nucleation over growth (Dirksen &amp; Ring, 1991) .

The classical stirred tank was the first configuration investigated in the literature, since it was used already by the group of Fessi. It is the common choice for mixing problems, and probably the most studied configuration in the literature. Unfortunately, it is characterised by non-homogenous distribution of turbulence (and therefore of micromixing times), with intense turbulence in the stirrer region and much lower values elsewhere. The consequence is a non-uniform population of particles, with broad distribution of size and of properties. Due to this region Fessi himself considered a T-mixer for better control of precipitation (Briancon et al, 1999).
An evolution of this mixer is the confined impinging jet reactor, proposed originally by Mahajan &amp; Kirwan (1996). This configuration allows extremely rapid mixing, with mixing times even smaller than the characteristic times for nucleation, growth and aggregation. Since the properties of the product depend on the slower processes, they become nearly independent from mixing, giving rise to a more uniform precipitate, as shown by Johnson (2003).
The vortex reactor may be regarded as a promising alternative to the former, as it overcomes the need for equal flow rates of the two reacting liquid streams.
A viable alternative to the impinging jet and vortex reactors could be the static mixer. This device, too, allows to reach very short mixing time in comparison to conventional reactors (Streiff et al, 1999). Prediction of the operation of turbulent mixers still partly relies on experimental work. As pointed out by Das et al (2005), the flow pattern in a static mixer is very complex to be described quantitatively and the efforts to use CFD codes are restricted to few applications so far (Fradette et al, 1998; Rauline et al, 1998) .
Reactors based on high shear rate (for mixing promotion) have also been considered in the literature for carrying out complex fast reactions and for precipitation reactions (Rousseaux et al, 1999 and 2001). The configuration used in this project adopts a Couette geometry, which is different from the rotating disk arrangement by Rousseaux, but should give higher flow field uniformity (and thus a more homogeneous product), while retaining the capability of high shear rate.

All the above mentioned reactor types will be the subject of specific investigations by some research unit, with the aim of comparing their performance and robustness on scale-up, for the nanoprecipitations of interest to the project. <<<