**4.1 Microfluidic systems for the production of cationic liposomes**

Among different microfluidic geometries, hydrodynamic focusing (HF) is a promising technology and liposomes can be produced in sizes ranging from 50 to 500 nm (Jahn et al., 2004, 2007; Wagner et al., 2002). The HF consists of a device with four-microchannel intersection geometry (Figure 4A). The organic solvent, miscible in water (isopropanol, ethanol), containing dispersed lipids are injected in the middle stream and hydrodynamically compressed by two aqueous (or buffered) streams. The precise upstream flow rates are achieved using syringe pumps that control the position of the focused stream in the downstream channel and, consequently, the process quality parameters (Baldas & Caen, 2010). The use of stereo microscopy is recommended to monitor the process as presented in the schematic representation of the experimental apparatus in Figure 4B. The laminar flow rate allows the formation of a well defined region between the two miscible fluids. The interfacial forces between two miscible solvents (for example, water and isopropanol or water and ethanol) control the convective-diffusive process and liposome self-assembly (Jahn et al., 2010). From the phenomenological point of view, the continuous flow mode allows the continuous diffusion of water and alcohol, reducing lipid solubility, which causes thermodynamic instability and generates liposomes (Jahn et al., 2008). The continuous flow mode also increases productivity.

The HF geometry for liposome production is based on the conventional method of ethanol injection (bulk) adapted for microfluidic systems. Conventional ethanol injection consists of the controlled addition of an ethanol/lipid stream in a tank reactor containing buffered water under controlled agitation. The advantage is the use of ethanol; compared to other organic solvents (e.g., chloroform and methanol), it is less harmful and, depending on the lipid concentration, there may be no need for post-treatment size reduction (Kremer & Esker, 1977). Some of the disadvantages are low lipid concentration in ethanol (higher concentrations require post-treatment for size reduction) and a difficulty achieving reproducibility (Wagner et al., 2002).

The representative parameters for liposome size and polydispersity control in HF are channel geometry (deep and width), volumetric flow buffer/alcohol rate-ratio (FRR), and total volumetric flow rate (buffer+alcohol flow rates). Solvent diffusivity in water is another important parameter (Jahn et al., 2004, 2007, 2010). Basically, the narrower the microchannel width, the smaller the liposome size. For total volumetric flow rate, which promotes flow velocity at the channel, higher FRRs (which corresponds to an increased proportion of buffer

the general key for homogenation and reducing liposome size. Mechanical stirring, extrusion through orifices (French press), extrusion through membranes, high-pressure impactor, and

Bottom-up processes are low energy processes inspired by biological systems and used for the development of functional nanomaterials, such as supramolecular structures, selfaggregated monolayers, Langmuir-Blodgett films, aggregated peptide nanotubes, and deposited polyelectrolites or proteins in multilayers (Mijatovic et al., 2005; Shimomura &

Microfluidic systems have been the main representatives of bottom-up processing for liposome production in continuous and scaled up processes. Different microfluidic systems can be applied to the production of liposomes and giant liposomes (Ota et al., 2009; Shum et al., 2008; Wagner et al., 2002). DNA complexation on the external surface of liposomes also

Among different microfluidic geometries, hydrodynamic focusing (HF) is a promising technology and liposomes can be produced in sizes ranging from 50 to 500 nm (Jahn et al., 2004, 2007; Wagner et al., 2002). The HF consists of a device with four-microchannel intersection geometry (Figure 4A). The organic solvent, miscible in water (isopropanol, ethanol), containing dispersed lipids are injected in the middle stream and hydrodynamically compressed by two aqueous (or buffered) streams. The precise upstream flow rates are achieved using syringe pumps that control the position of the focused stream in the downstream channel and, consequently, the process quality parameters (Baldas & Caen, 2010). The use of stereo microscopy is recommended to monitor the process as presented in the schematic representation of the experimental apparatus in Figure 4B. The laminar flow rate allows the formation of a well defined region between the two miscible fluids. The interfacial forces between two miscible solvents (for example, water and isopropanol or water and ethanol) control the convective-diffusive process and liposome self-assembly (Jahn et al., 2010). From the phenomenological point of view, the continuous flow mode allows the continuous diffusion of water and alcohol, reducing lipid solubility, which causes thermodynamic instability and generates liposomes (Jahn et al., 2008). The

The HF geometry for liposome production is based on the conventional method of ethanol injection (bulk) adapted for microfluidic systems. Conventional ethanol injection consists of the controlled addition of an ethanol/lipid stream in a tank reactor containing buffered water under controlled agitation. The advantage is the use of ethanol; compared to other organic solvents (e.g., chloroform and methanol), it is less harmful and, depending on the lipid concentration, there may be no need for post-treatment size reduction (Kremer & Esker, 1977). Some of the disadvantages are low lipid concentration in ethanol (higher concentrations require post-treatment for size reduction) and a difficulty achieving

The representative parameters for liposome size and polydispersity control in HF are channel geometry (deep and width), volumetric flow buffer/alcohol rate-ratio (FRR), and total volumetric flow rate (buffer+alcohol flow rates). Solvent diffusivity in water is another important parameter (Jahn et al., 2004, 2007, 2010). Basically, the narrower the microchannel width, the smaller the liposome size. For total volumetric flow rate, which promotes flow velocity at the channel, higher FRRs (which corresponds to an increased proportion of buffer

microchannel microfluidizer are equipment used in high-energy processes.

constitute a promising technology for the production of gene vaccines.

**4.1 Microfluidic systems for the production of cationic liposomes** 

continuous flow mode also increases productivity.

reproducibility (Wagner et al., 2002).

Sawadaishi, 2001).

to lipid/solvent flow) correspond to smaller liposome diameter (Janh et al., 2010). However, caution is required in this flow rate analysis as the final alcohol content in the liposome colloidal dispersion can change with FRR and as a consequence of particle size. Alcohol can be used to disrupt liposomes and, at low concentrations, liposome size can be increased due to alcohol incorporation into the bilayer. Precise analysis can be performed if the alcohol is removed after liposome processing. By decreasing the total volumetric flow rate, the residence time can be increased and a lower average vesicle diameter and narrower size distribution obtained. This behavior indicates that if increasing the total volumetric flow rate, the microchannel length must be longer to complete the alcohol diffusion, otherwise large particles will be obtained due to bulk mixing downstream of the channel (Janh et al., 2010).

Fig. 4. (A) Schematic representation of hydrodynamic focusing (HF) in a microfluidic device with four-channel intersection geometry. The organic solvent (miscible in water) containing the dispersed lipids is injected in the middle stream and hydrodynamically compressed by two aqueous (or buffered) streams. The flow is in the x direction and alcohol diffusion is in the y direction. (B) Schematic representation of the experimental apparatus for liposome production using HF: (1) microfluidic device; (2) water syringe pumps; (3) syringe pump for lipids/ethanol stream; (4) collector flask; (5) stereo microscopy; (6) computer data acquisition (adapted from Zhang et al., 2008).

The lipid concentration in alcohol is also another important parameter for controlling size. The development of vaccines for *in vivo* applications requires the highest drug-loading capacity. In terms of gene vaccines, the DNA-loading capacity correlates with the cationic lipid content, defined by the molar charge rate (R+/-) between the cationic charges (from the cationic lipid) and negative charges (from the DNA). The R+/- and total lipid content are project parameters for scaling up (or scaling out) processes. As an example, the R+/- of the tuberculosis gene vaccine is 10 and the total lipid concentration 64 mM (Rosada et al., 2008). These parameters require cationic liposome production in a high lipid concentration under unusual microfluidic conditions. Aiming to explore the effect of lipid concentration on microfluidic HF processes, we investigated the influence of high lipid content (100 mM) and lipid composition in ethanol (EPC or EPC/DOTAP/DOPE) as a function of average size. We understand that the studied molar concentration is greater than the lipid solubility in ethanol (approximately 4 mM for EPC) (New, 1990) and, in this case, the lipid dispersion offers an additional barrier to ethanol diffusion into water. Based on this assumption, we simulated the required microchannel length for complete ethanol diffusion from a central stream (after HF) as a function of the lipid concentration in the ethanol stream considering

Technological Aspects of Scalable Processes for the

Lipid composition Total flow velocity

*FRR: volumetric flow buffer/alcohol rate-ratio. Total lipid concentration of 100 mM in ethanol stream.* 

faster mixing (Knight et al., 1998).

flow rates.

Production of Functional Liposomes for Gene Therapy 279

According to the simulation, ethanol dispersion requires the lowest length for ethanol diffusion for a lipid concentration of 100 mM (among the simulated concentrations). Experimental evaluations with this lipid concentration were performed at different total flow velocities (total flow rate) and lipid compositions (EPC and EPC/DOTAP/DOPE – cationic liposomes) as presented in Table 1. The average diameter is strongly influenced by the total flow velocity and lipid composition. Decreasing the flow velocity decreases the average size, and this behavior is not dependent on the FRR. This parameter is associated with the residence time inside the microchannel, suggesting that higher residence time is required to control liposome size. Another interesting parameter is the lipid composition. The viscosity of the ethanol dispersion was 1.07 and 1.21 cP for EPC/DOTAP/DOPE and EPC, respectively. Lower viscosity probably offers lower mass transfer resistence, reflecting smaller diameters (and standard deviations) and polydispersity index values. This difference reflected in HF is confirmed by stereomicroscopic observation along the microchannel. A high polydispersity index (Table 1) can also indicate the possibility for

further process optimization in terms of flow velocity, FRR, and lipid concentration.

EPC 5.6 6 342.8 ± 134.5 0.427 ± 0.140

EPC/DOTAP/DOPE 8.1 10.6 162.1 ± 88.13 0.606 ± 0.157

Table 1. Average liposome size (Z-average) obtained at different lipid compositions and

The electrostatic interactions between DNA and cationic liposomes produce particles with different sizes and morphology (Mannisto et al., 2002; Oberle et al., 2000) that depend on R+/-, buffer ionic strength, order of component addition, reaction conditions, and the type of lipids (Mount et al., 2003; Zelphati et al., 1998). In this context, HF can also be used to control the diffusion process for DNA compaction, producing well organized aggregates. The flow velocity is the major parameter controlling the aggregation process (Dootz et al., 2006). Otten et al. (2005) investigated the HF microfluidic device to produce cationic liposome-DNA complexes. The DNA solution is introduced in the central stream and the cationic liposome stream introduced at a lateral position. The average liposome size was 200 nm (composed of 1:1 DOTAP and DOPC with a lipid concentration of 25 mg.mL-1), and the DNA was calf thymus (5 mg.mL-1). The flow velocity was 100 mm.s-1, varying with vLiposome = 13vDNA and vLipossomas = 130vDNA, where v is the flow velocity. The authors concluded that the complex is formed in two steps. The first step relates to the formation of a multilamellar complex, followed by the second step in which DNA is organized inside the lamellae. The central stream can be focused according the FRR, and reducing the diffusional length allows

**4.2 Microfluidic systems for electrostatic complexation of DNA** 

(cm/min) FRR Z-average (nm) Polidispersity

4 7 271.6 ± 37.5 0.408 ± 0.145

6.1 7.7 92.92 ± 10.14 0.441 ± 0.028

index

the mass continuity equation (according to Figure 4A), without a chemical reaction (Equation 18). This simulation was performed based on a microfluidic glass device produced by the wet photolithographic process. The microchannels were etched with HF solution; upstream channels measured 140 ± 1 µm in width and downstream channels 200 ± 1 µm after the T junction. The microchannel was 50 ± 2 µm deep and the diffusion length 5 cm.

$$\frac{\partial \mathbf{C}\_E}{\partial t} + \left(\upsilon\_x \frac{\partial \mathbf{C}\_E}{\partial x} + \upsilon\_y \frac{\partial \mathbf{C}\_E}{\partial y} + \upsilon\_z \frac{\partial \mathbf{C}\_E}{\partial z}\right) = D\_{\rm EW} \left(\frac{\partial^2 \mathbf{C}\_E}{\partial x^2} + \frac{\partial^2 \mathbf{C}\_E}{\partial y^2} + \frac{\partial^2 \mathbf{C}\_E}{\partial z^2}\right) \tag{18}$$

Where DEW is the ethanol diffusion coefficient in water and CE is the ethanol concentration. The Cartesian coordinates are used because the microchannel area presents rectangular geometry. Considering laminar flow in x direction and mass transfer along the y direction (Figure 4A), Equation 18 can be adapted for short times according to the Higbie penetration model (Higbie, 1935):

$$N\_E = \sqrt{\frac{4.D\_{EA}}{\pi \text{t.}}} . (\text{C}\_{E\_0} - \text{C}\_{E\text{o}}) \tag{19}$$

Where *t* is the quotient between the difference in flow rates (water and ethanol) and distance x. The influence of lipid concentration was considered an additional barrier for ethanol diffusion due to the presence of lipid aggregates dispersed in the ethanol stream. In this case, the ethanol diffusion depends on the concentration gradient through ethanol/water streams, as well as its hydrophobic characteristics, with similar behavior as ethanol diffusion from aqueous solution to a phospholipid bilayer (Galindo-Rodriguez et al., 2004). The hydrophobic characteristics of the lipid can be expressed as the partition coefficient octanol/water (Po/w= 0,478). In this context, DEW was corrected by the partition coefficient according to Equation 20.

$$D\_{EW} = 0.478 \, ^\circ D\_{EW} \tag{20}$$

Where DEW´ is the effective diffusion coefficient.

Figure 5 presents the simulation of the distance (x) for total ethanol diffusion as a function of ethanol flow rate (at a fixed FRR of 6.26 and 20). Increasing the lipid concentration in the ethanol stream, the distance x to complete ethanol diffusion will increase.

Fig. 5. Simulation of the distance for total ethanol diffusion into the water stream. The lipid (EPC) concentration was considered as an additional barrier to ethanol diffusion. (A) FRR = 6.26. (B) FRR = 20.

the mass continuity equation (according to Figure 4A), without a chemical reaction (Equation 18). This simulation was performed based on a microfluidic glass device produced by the wet photolithographic process. The microchannels were etched with HF solution; upstream channels measured 140 ± 1 µm in width and downstream channels 200 ± 1 µm after the T junction. The microchannel was 50 ± 2 µm deep and the diffusion

> *E E E E EEE <sup>x</sup> <sup>y</sup> <sup>z</sup> EW C C C C CCC vvv D t xyz xyz* ∂ ∂ ∂ ∂ ∂∂∂ + + + = ++ ∂ ∂∂∂ ∂∂∂

Where DEW is the ethanol diffusion coefficient in water and CE is the ethanol concentration. The Cartesian coordinates are used because the microchannel area presents rectangular geometry. Considering laminar flow in x direction and mass transfer along the y direction (Figure 4A), Equation 18 can be adapted for short times according to the Higbie penetration

Where *t* is the quotient between the difference in flow rates (water and ethanol) and distance x. The influence of lipid concentration was considered an additional barrier for ethanol diffusion due to the presence of lipid aggregates dispersed in the ethanol stream. In this case, the ethanol diffusion depends on the concentration gradient through ethanol/water streams, as well as its hydrophobic characteristics, with similar behavior as ethanol diffusion from aqueous solution to a phospholipid bilayer (Galindo-Rodriguez et al., 2004). The hydrophobic characteristics of the lipid can be expressed as the partition coefficient octanol/water (Po/w= 0,478). In this context, DEW was corrected by the partition coefficient

Figure 5 presents the simulation of the distance (x) for total ethanol diffusion as a function of ethanol flow rate (at a fixed FRR of 6.26 and 20). Increasing the lipid concentration in the

Fig. 5. Simulation of the distance for total ethanol diffusion into the water stream. The lipid (EPC) concentration was considered as an additional barrier to ethanol diffusion. (A) FRR =

ethanol stream, the distance x to complete ethanol diffusion will increase.

0 4. .( ) . *EA E E E <sup>D</sup> N CC* π

222 222

*<sup>t</sup>* = − <sup>∞</sup> (19)

, 0,478 \* *D D EW* <sup>=</sup> *EW* (20)

(18)

length 5 cm.

model (Higbie, 1935):

according to Equation 20.

6.26. (B) FRR = 20.

Where DEW´ is the effective diffusion coefficient.

A B

According to the simulation, ethanol dispersion requires the lowest length for ethanol diffusion for a lipid concentration of 100 mM (among the simulated concentrations). Experimental evaluations with this lipid concentration were performed at different total flow velocities (total flow rate) and lipid compositions (EPC and EPC/DOTAP/DOPE – cationic liposomes) as presented in Table 1. The average diameter is strongly influenced by the total flow velocity and lipid composition. Decreasing the flow velocity decreases the average size, and this behavior is not dependent on the FRR. This parameter is associated with the residence time inside the microchannel, suggesting that higher residence time is required to control liposome size. Another interesting parameter is the lipid composition. The viscosity of the ethanol dispersion was 1.07 and 1.21 cP for EPC/DOTAP/DOPE and EPC, respectively. Lower viscosity probably offers lower mass transfer resistence, reflecting smaller diameters (and standard deviations) and polydispersity index values. This difference reflected in HF is confirmed by stereomicroscopic observation along the microchannel. A high polydispersity index (Table 1) can also indicate the possibility for further process optimization in terms of flow velocity, FRR, and lipid concentration.


*FRR: volumetric flow buffer/alcohol rate-ratio.* 

*Total lipid concentration of 100 mM in ethanol stream.* 

Table 1. Average liposome size (Z-average) obtained at different lipid compositions and flow rates.
