Flow-assembled chitosan membranes in microfluidics: recent advances and applications

Abstract

The integration of membranes in microfluidic devices has been extensively exploited for various chemical engineering and bioengineering applications over the past few decades. To augment the applicability of membrane-integrated microfluidic platforms for biomedical and tissue engineering studies, a biologically friendly fabrication process with naturally occurring materials is highly desired. The in situ preparation of membranes involving interfacial reactions between parallel laminar flows in microfluidic networks, known as the flow-assembly technique, is one of the most biocompatible approaches. Membranes of many types with flexible geometries have been successfully assembled inside complex microchannels using this facile and versatile flow-assembly approach. Chitosan is a naturally abundant polysaccharide known for its pronounced biocompatibility, biodegradability, good mechanical stability, ease of modification and processing, and film-forming ability under near-physiological conditions. Chitosan membranes assembled by flows in microfluidics are freestanding, robust, semipermeable, and well-aligned in microstructure, and show high affinity to bioactive reagents and biological components (e.g. biomolecules, nanoparticles, or cells) that provide facile biological functionalization of microdevices. Here, we discuss the recent developments and optimizations in the flow-assembly of chitosan membranes and chitosan-based membranes in microfluidics. Furthermore, we recapitulate the applications of the chitosan membrane-integrated microfluidic platforms dedicated to biology, biochemistry, and drug release fields, and envision the future developments of this important platform with versatile functions.

Graphical Abstract

Flow-assembled chitosan membranes in microfluidics are freestanding, semipermeable, and robust, finding broad applications including substrate immobilization, gradient generation, cell culture, synthetic ecosystem, and drug delivery.

1. Introduction

Microfluidics has been intensively applied in a variety of analytical, bioengineering, and chemical engineering studies thanks to its low reagent consumption and fabrication cost, quick reaction time, and high sensitivity and controllability.^1,2Integrated microfluidic platforms for life science applications demand enhanced biocompatibility and biological functionality. Towards the biofunctionalization of microdevices to minimize their intrinsic difference from biological components, using an environmentally friendly synthesis process with biocompatible materials is highly desired.^3,4 Membrane technology offers a precise separation process with various driving forces (e.g., concentration gradient, electrical force, pressure difference, and thermal variation) in a cost-saving, operation-efficient, and function-versatile manner as compared to traditional separation techniques.^2,5 The integration of membrane functionality into microfluidics has converged their inherent advantages for broader applications.⁶

There are several options to integrate membranes into microfluidic networks with respective pros and cons for each. Direct incorporation of the commercial membrane into microdevices is problematic due to insufficient sealing and unwanted leakage that can lead to chemical compatibility issues. The preparation of the membrane as a part of the microdevice fabrication process is complicated and exorbitant.^7,8 The in situ preparation of the membrane in microfluidic networks, stemmed from interfacial reactions between converging laminar flows and referred to as the flow-assembly technique, has emerged as a promising alternative.^9–11 The accurate manipulation of multiphase flows in microfluidic networks enables a highly programmable formation of a wide range of polymeric membranes such as alginate,¹² chitosan,^3,13 nylon,^14,15 palladium-complex,¹⁶ and polyacrylamide¹⁷membranes. Among them, freestanding chitosan membranes (CMs) or chitosan-based membranes (CBMs) assembled by flows inside microfluidic devices are the prominent candidates to integrate biology with inorganic devices thanks to their excellent characteristics, which is the subject of the present review.

Chitosan is a derivative of the second abundant biopolymer chitin that comprises linear N-acetyl glucosamine and β−1,4-linked d-glucosamine units. Owing to its pronounced biocompatibility, biodegradability, low cost, ease of modification and processing, non-toxicity, and good absorption properties, chitosan has been broadly applied in a diverse range of biomedical, bio-micro-electromechanical systems (bioMEMS), tissue engineering, and drug delivery applications.^18–21 Besides the mentioned biological and physiochemical significance, chitosan has been eminent for its pH-dependent solubility. Chitosan is water-soluble under acidic conditions and becomes insoluble with gel-forming properties when the pH of the surrounding environment is higher than its pK_a (~6.3), making its gelation close to that under physiological conditions. Therefore, chitosan is an ideal candidate for biological and biomedical applications.^22,23 Most importantly, its abundant amine groups enable the undemanding immobilization of bioactive reagents (e.g. biomolecules, other polymers, or nanoparticles) and biological components (e.g. cells or tissues) to the chitosan backbone,²⁴ thus augmenting the biocompatibility and bioactivity of the chitosan-integrated microchips.

Previous review articles have explored the uses of chitosan in bioMEMS applications²⁴ and membrane technology,⁵ CMs in separation applications,²⁵ and CMs as absorptive membranes.²⁶ Common strategies to integrate CMs with microdevices including solution casting, spin casting, electrodeposition, and nanoim-printing have also been recapitulated.^24,27 To the best of our knowledge, no review paper has yet been dedicated to reporting the flow-assembly of CMs and CBMs and their integrated microfluidic platforms for practical applications. As depicted in Scheme 1, the precise control of two fluidic flows, specifically an acidic chitosan solution and a basic buffer solution, inside microchannels enables the versatile, rapid, and reliable formation of the CM as desired in microchips. The flow-assembled CM is freestanding, semi-permeable, robust, and well-aligned in microstructure,^3,13,28 and it is easy to decorate them with bioactive reagents.^7,29–34 In this review, we commence with a general explanation of the formation mechanism of the flow-assembled CM in Section 2. Next, Section 3 reports the characteristics of the flow-assembled CM, followed by a summary of recent advances and modifications in the flow-assembly of CMs and CBMs in microfluidics in Section 4. Lastly, Section 5 discusses the implementations of the CM-integrated microfluidic platforms in diverse applications ranging from biochemistry to biology to drug release screening. We envision that flow-assembled CMs will emerge as an important platform and provide tremendous benefits for multidisciplinary applications, and this timely review article can aid in directing the future developments of this platform.

Scheme 1.

Flow-assembled chitosan membranes (CMs) in microfluidics. (A) Molecular transition of chitosan around its pK_a at 6.3. (B) Flow-assembly of the CM between parallel flows. (C) Key features of the assembled CM. (D) CM in microfluidics for versatile applications. Adapted with permission from ref. 3, 30 and 35–37.

2. Flow-assembly of chitosan membranes in microfluidics

Chitosan is a polysaccharide containing many amine groups that are protonated in a low pH environment, making chitosan water-soluble. When the surrounding pH value becomes higher than chitosan’s pK_a around 6.3, the amine groups on chitosan are deprotonated, inducing a sol–gel transition to form a hydrogel or a membrane-like structure as depicted in Fig. 1A(a).³⁸ Based on this unique pH-responsive property of chitosan, it is possible to assemble CMs in microdevices with a localized pH gradient at the flow interface with the flow-assembly technique.²³ In general, there are two tactics with variations and further optimizations to flow-assemble CMs in microfluidic channels, as discussed below.

Fig. 1.

Mechanisms of the flow assembly of the CM in microfluidics. (A) Direct gelation with basic buffer: (a) schematic representation of CM formation with a localized pH gradient established at the flow interface of a basic buffer solution and an acidic chitosan solution; (b) microfluidic device design and the pH gradient generated at the flow interface in microchannels, visualized with a pH indicator solution; (c) the microscopic image of a fabricated CM. (B) Gelation across a polyelectrolyte complex membrane (PECM) layer: (a) chemical structures and electrostatic interactions between alginate and chitosan chains; (b) the formation of the PECM at the interface of an alginate solution drop and a chitosan solution drop; (c) experimental setup to balance the pressure and expel a naturally trapped air bubble in a hydrophobic PDMS aperture between two microchannels; (d) a PDMS device containing several microchannels bonded to a glass slide; (e) dissipation of the air bubble and formation of the permeable PECM allowing hydroxyl ions to diffuse through the PECM and, ultimately, forming the CM. (A) is adapted with permission from The Royal Society of Chemistry;³ (B) is adapted with permission from Elsevier.¹³

The first in situ fabrication of CMs in microfluidics was demonstrated at the converging flow interface between an acidic chitosan solution and a basic buffer solution, where a stable pH gradient was established at the flow interface to directly assemble a CM as illustrated in Fig. 1A. The established pH gradient was visualized by adding a pH indicator in the middle microchannel as demonstrated in Fig. 1A(b). A well-distinguished pH gradient transitioning from pH of 4 (pink) to pH of 10 (blue) was established where the two fluidic streams converged, triggering the sol–gel transition of chitosan, and ultimately forming a CM at the interface. The formation of the freestanding CM commenced from the upstream nucleation point and propagated to the anchoring point situated downstream.^3,23 Fig. 1A(c) shows a 60-μm-thick CM formed along the interface of the two converging flows of acidic chitosan and basic buffer solutions. The fabricated CM was usually long (up to 4 mm) and thick in this scenario. The key to the successful CM formation was to establish stable pH gradients at the flow interface with an appropriate device design and pumping strategy. One challenge of this direct assembly approach was the deposition of chitosan residues in the downstream channels, which could disrupt the pressure balance between the flow streams and dislocate the membrane anchoring on the device. One solution to this challenge will be further discussed in Section 4.

Fig. 1B shows the second approach to flow-assemble CMs involving the use of alginate in the basic solution. This approach utilized air bubbles to initiate the biofabrication of a CM on a thin polyelectrolyte complex membrane (PECM), which was generated as the alginate and the chitosan solutions came into contact.¹³ The PECM was spontaneously formed with electrostatic interactions between the negatively charged carboxyl groups of the alginate chains and the positively charged amino groups of the chitosan chains as depicted in Fig. 1B(a). The PECM acted as a barrier to prevent the diffusion of either the alginate or the chitosan chains, evidenced by a clear interface established between a drop of chitosan solution and a drop of alginate solution (Fig. 1B(b)). The ionic reaction between the alginate and chitosan solutions to form a PECM in a microfluidic device is depicted in Fig. 1B(c–e). The chitosan and alginate solutions were slowly introduced into two separate microchannels, where one set of outputs was blocked, and the other set of outputs was connected by air-filled tubing (Fig. 1B(c and d)). Initially, as the chitosan and alginate solutions came near the aperture, an air bubble was trapped within the aperture due to the hydrophobicity of the polydimethylsiloxane (PDMS) device (Fig. 1B(e)–(i)). Then, the pressure between the flow fronts of the alginate and chitosan solutions was balanced through the air-filled tubing. Continuous pumping of the solutions increased the pressure inside the microchannels, which dissipated the trapped air bubble through the gas-permeable PDMS layer, and the alginate and chitosan solutions finally came into contact and spontaneously formed a PECM at the solution interface (Fig. 1B(e)–(ii)). Once the PECM was formed, the CM was built on the PECM with the localized pH gradient generated by the continuous diffusion of hydroxyl ions from the alginate side via the PECM (Fig. 1B(e)–(iii)).¹³ This approach enables the formation of a relatively shorter and thinner CM than the one assembled directly at the flow interface. The microchip pattern with a long flow interface is no longer required to maintain the stable localized pH gradient, which allows much room for the customization and optimization of the microchip design for broader applications.^{13,29,35,39–42} It is important to note that the addition of the PECM introduces carboxyl groups in the alginate chains to one side of the CM besides the already existing amine groups inside and on the surface of the CM, thus expanding the capability to immobilize versatile biomolecules into the CM using either amine or carboxyl chemistry. The contribution of the PECM in protecting the alginate hydrogel with embedded cells will also be discussed in Section 5.

3. Characterization of flow-assembled chitosan membranes

3.1. Physicochemical properties of flow-assembled chitosan membranes

To better utilize the flow-assembled CM-integrated microfluidic platforms, it is critical to understand the characteristics of the fabricated CM. This section reviews the key physicochemical properties of the flow-assembled CM and how some of these properties might change relative to the fabrication conditions.

First and foremost, the impact of the flow rates on the growth of the fabricated membranes was studied. Fig. 2A(a and b) shows that as the flow rate of the buffer solution was fixed while varying the flow rate of the chitosan solution and vice versa, respectively, no significant effect on membrane thickness was observed.³ Meanwhile, Fig. 2A(c) reveals that if there was a significant difference in the flow rates of the basic and chitosan solutions, an increase in the total flow rate during fabrication resulted in a significant decrease in the membrane thickness.³¹ The increase in flow rates likely narrows the pH gradient and the time it appears at the interface. Furthermore, the faster the flow rates, the higher the shear stresses emerge at the membrane surface, and the shorter the time for the chitosan chains to anchor, ultimately leading to thinner and denser membranes.^43,44 Therefore, it is possible to control the thickness of the flow-assembled CM by varying the total flow rates while maintaining the flow rate ratio between the basic and the chitosan solutions.

Fig. 2.

Key physicochemical properties of the flow-assembled CM in microfluidics. (A) Growth curves of the CM formed with direct gelation between adjacent flows: time-dependent growth of membrane thickness at (a) varied chitosan flow rates with a fixed 200 μL min⁻¹ buffer solution; (b) varied buffer flow rates with a fixed 30 μL min⁻¹ chitosan solution; and (c) membrane growth tested under five different total flow rates (Q_t): 2.7 (orange), 5.4 (blue), 8.1 (gray), 10.8 (yellow), and 13.5 mL h⁻¹ (green); Q_t = flow rate of the basic solution (Q_b) + flow rate of the chitosan solution (Q_p) and the ratio of Q_b/Q_p = 20. (B) Permeability of the CM formed with direct gelation: (a and b) free diffusion of FITC (molecular size <1 nm); (c) partial transport of TRITC-labeled antibodies (molecular size of 7–10 nm); (d and e) complete stop of FITC-labeled polystyrene nanospheres (particle size of 20 nm). (C) The birefringence of the CM formed across the PECM: (a and c) the membrane thickness, (b and d) the birefringence signals, and (e) the birefringence across the normalized membrane thickness of the CM formed with gelation across the PECM at various pH values of the alginate solution such as 10.5, 10.7, 11.0, 11.5, and 12. (D) Adhesion strength characterization of the CM formed with gelation across the PECM using the ideal gas law principle: (a) experimental setup; (b) pressure measurement of an approximately 30-μm-thick CM before it burst at 0.67 atm pressure; (c) critical pressure in linear relationship with membrane thickness. (A)(a and b) and (B) are adapted with permission from The Royal Society of Chemistry³; (C) is adapted with permission from IOP Publishing;²⁸ and (A(c)) and (D) are adapted with permission from Elsevier.^13,31

Second, Luo et al. investigated the average pore size of the fabricated CM through permeability tests with fluorescein isothiocyanate (FITC), tetramethylrhodamine (TRITC)-labelled antibodies, and FITC-labelled polystyrene nanospheres, among which the particle size ranged from less than 1 nm to 20 nm in diameter. Fig. 2B(b, c and e) shows that the fluorescein (less than 1 nm in size) freely passed through, TRITC-labelled antibodies (size of 7–10 nm) partially diffused through, and FITC-labelled polystyrene nanospheres (20 nm in diameter) were completely stopped by the membrane. The results suggest that the pore size of the flow-assembled CM is within the nanometer range around the size of proteins or antibodies.³ In a later study, Luo and colleagues yielded similar results in the permeability of the flow-assembled CM fabricated using their newly developed air-initiated biofabrication process.¹³ In summary, the flow-assembled CM is permeable to small molecules with a molecular weight cut-off of a few nanometers while it physically separates flow streams, which can be used as a reaction site for biomolecular immobilization and enzyme catalysis.^3,13,23

Third, the microstructure and polymer chain alignment of the fabricated membranes and the contributing fabrication parameters were investigated. Li et al. used quantitative polarized light microscopy (qPLM) to examine the birefringence signals and determine the effects of pH and the flow rate on the flow-assembled CM’s microstructural organization and polymer chain alignment.²⁸ Birefringence is an inherent optical property of anisotropic materials that can reveal their crystal microstructures and polymer chain alignment.⁴⁵ qPLM is a powerful means to study the birefringence of many anisotropic materials by generating optical retardance location maps relative to the microscale-level organization and alignment of birefringent macromolecules.⁴⁶ Optical retardance generated by qPLM digital image processing is a parameter proportionally correlated with birefringence signals of materials where a higher optical retardance represents a higher crystalline and alignment order of materials.⁴⁷ Details on how to obtain the optical retardance map of a CM can be found in Section 4.3 and previous reports.^28,29,39 Herein, the authors figured out that the flow-assembled CM was highly aligned along the flow direction of the chitosan solution inside the PDMS microfluidic network. Furthermore, the optical retardance signals of the flow-assembled CM increased significantly relative to the alginate solution’s pH, but were less dependent on the flow rates.²⁸ Notably, the optical retardance signal reduced dramatically from the PECM side to the alginate side of the membrane despite the increased molecular density in the membrane.²⁸ Fig. 2C(a) shows a set of typical CMs after 10 minutes of their assembly by flows with varied pH values of the alginate solution while the pH of the chitosan solution was fixed. Notably, the thickness of the CM increased significantly with the increased pH of the alginate solution. Fig. 2C(b) displays the corresponding birefringence signals of CMs in (a), which were apparently observed for all CMs assembled under different conditions with a steady decrease in optical retardance across the membrane growth direction. Furthermore, Fig. 2C(d) shows that as the pH of the alginate solution increased, the higher flux of hydroxyl ions and the sharper pH gradient ultimately led to higher optical retardance in the CM. On the other hand, the birefringence signal was less sensitive to the variation of the flow rates of the polymer solutions.²⁸

Fourth, the adhesion strength of the freestanding CM to the PDMS device was determined by employing the simple ideal gas law as previously reported. The general principle of the characterization method is shown in Fig. 2D(a). By connecting compressible air in a leak-tight syringe with incompressible liquid in the tubing to the microchannel, the hydrostatic pressure acting on the membrane was reflected by the decreased air volume.^13,39,48 The relationship between the critical pressure to detach the CM from the PDMS and the CM thickness is shown in Fig. 2D(b and c). The authors reported that a linear correspondence was observed between the critical pressure and the membrane thickness: the thicker the membrane, the higher the critical pressure. Notably, the typical 54-μm-thick freestanding CM could withstand a hydrostatic pressure of up to 1.1 atmospheric pressure, suggesting that the anchoring of the flow-assembled CM on the PDMS device is robust.¹³ It is worth pointing out that so far only the adhesion strength of the CM to PDMS has been characterized, while the intrinsic mechanical properties of the flow-assembled CM are yet to be further investigated. Presumably, CMs can be fabricated in pure PDMS microchannels by replacing the bottom glass slide with a planar PDMS, so that the fabricated CM can be harvested for further characterization.

Lastly, since chitosan is a well-known pH-responsive biopolymer, the flow-assembled CM also possesses this respective property. The deprotonation of chitosan amine groups to transit soluble chitosan chains into insoluble CM at pH higher than 6.3 is reversible. This indicates that the insoluble CM can be readily dissolved if the pH of the surrounding environment falls below 6.3. Therefore, once the fabrication is completed, the CM must be maintained in an aqueous environment with a pH value higher than its pK_a. It has been reported in previous studies that CM was quickly dissolved by half within 20 seconds when an acidic solution (pH = 2) was introduced into the microchannels.³⁹ It should be noted that in the presence of the PECM, formed through the electrostatic interactions between the positively charged amine groups on the chitosan chains and the negatively charged carboxyl groups on the alginate chains, the CM exhibits complex degradation behavior. The bonds present in the PECM are highly stable at physiological pH yet become labile under mild acidic conditions. Specifically, at pH around 5.5, the insoluble CM will be protonated and becomes free of positively charged amino groups, leading to swelling, followed by gradual degradation of the CM.⁴⁹ Such responses of the CM to the pH of the surrounding environment can be utilized for controlled drug release, which will be described in Section 5.5. In another scenario, if the pH falls below 3.5, the pK_a of the alginate, the CM experiences faster degradation while the alginate becomes insoluble, resulting in a partially insoluble structure as previously reported.^50,51 On the other hand, the CM can also be crosslinked with glutaraldehyde or terephthalaldehyde to improve the strength and acidic resistance for more diverse applications.^30,39

3.2. Characterization approaches for flow-assembled chitosan membranes

Due to their tiny size, it is challenging to determine the characteristics of CMs in microfluidic devices.⁵⁵ In this section, we will list the most commonly used approaches to characterize the physicochemical properties of the flow-assembled CMs and CBMs.

One most convenient spectroscopic approach to visualize CMs and CBMs in a microfluidic chip is fluorescence microscopy. Flow-assembled CMs can be readily recognized under transmitted light microscopy, while chitosan molecules can also be conjugated with fluorescein to distinguish them from other polymers.¹³ Furthermore, using fluorescence microscopy, fluorescent dyes can be conjugated to biomolecules to confirm the presence or visualize their distribution within the fabricated membranes.^30,37 Fluorescence microscopy also aids in characterizing the permeability of CMs,^3,13 CBMs,³⁰ modified CMs,^7,29,39,55 and the generated chemical gradients.^29,35,52 Most importantly, the fluorescence microscopy technique is compatible with most microfluidic devices and can provide real-time observation and assessment of membrane fabrication and functionalities.

Furthermore, it is also challenging to determine the microstructural organization of the fabricated membranes in a microchip using conventional characterization methods due to the tiny size of the structure.^24,55 Therefore, advanced spectroscopic techniques such as qPLM and scanning electron microscopy (SEM) have also been used for membrane characterization. To obtain the optical retardance correlated with the birefringence of the CM, images of the membrane under sequent analyzer angles of 1–2° per step are taken with the birefringence signal of interest as the CM goes from the brightest to the lowest brightness area. The optical retardance map of the membrane is then obtained by fitting the birefringence signal versus the analyzer angle to a second-order polynomial.^28,29,39 SEM can also be employed to image the subnano- or microscale morphology of the CM and CBM. However, a specialized sample must be prepared for SEM observation since the conventional microfluidic devices are strongly bound and do not allow the electron beam to penetrate through them. Furthermore, the specialized sample must possess good electrical conductivity to obtain high-resolution images. To facilitate the extraction of the fabricated CM and CBM for SEM observation, specialized microfluidic devices can be fabricated with PDMS microfluidic channels as the top layer, and tape^7,56 or PDMS²⁹ as the bottom layer (PDMS-tape or bilayer PDMS devices, respectively). Additionally, to enhance the conductivity of the sample, the colloidal silver liquid can be added around the sample as a ground connector,²⁹ or a PDMS-silicone glass device can be used.⁵⁷

4. Developments and optimizations in the flow-assembly of chitosan membranes

With growing interest in the flow-assembled CMs and CBMs in microfluidics, more and more developments and optimizations have emerged in the literature. Table 1 summarizes the main advances that have been reported in the literature and their main purposes. When the flow-assembly of CMs was initially reported in 2010, an expertise pressure balancing technique through a specific syringe pump strategy was required to establish a stable pH gradient for membrane assembly. The original process is not user-friendly and generally yields a success rate of around 60% even for experienced researchers,⁴¹ and it is challenging to ensure a proper membrane attachment at the downstream.⁵³ Several advances have been developed to enhance the reliability of the flow-assembly process, among which are the employment of an add-on vacuum layer,⁴¹ the use of microchips containing circular pillars,⁵² and the addition of an extra outlet.⁵³

Table 1.

Summary of recent advances in the flow-assembly of CMs and CBMs in microfluidics

Advances	Purposes	Ref.
Add-on vacuum layer	To dissipate air bubbles in the aperture(s) through the gas-permeable properties of PDMS to initiate the flow-assembly of the CM	41
Microchips with small circular pillars	To position the initial chitosan meniscus that can be advanced down with the introduction of alginate solution, thereby preventing the trapping of air bubbles in the aperture(s) and directly enabling the flow-assembly of CM	52
Extra outlet	To serve as an anchoring point to guide CM formation	53
Crosslinking the CM with glutaraldehyde (GA)	To prevent the disruption of CM and PDMS pillars’ interactions caused by anti-adhesion agents (i.e. Pluronic F-127), resist acidic dissolution, and improve the adhesion strength of the CM to the PDMS device	39
Tuning the CM′s porosity with co-assembled nanoparticles as a sacrificial template	To actively manipulate the porosity of the flow-assembled CM inside microchannels, contributing to their semi-permeability and selectivity regarding application needs	29
In situ fabrication of a poly(N-isopropylacrylamide) (PNIPAM) nanogel-containing CM	To utilize chitosan as an embedded substrate to construct the PNIPAM nanogel-containing membrane with self-regulated permeability abilities, thanks to the reversible swelling or shrinking volume transitions relative to the changes in temperature and ethanol concentration of PNIPAM nanogels.	30 and 54
In situ fabrication of carbon nanoparticle–chitosan (CN–CS) composite membrane	To incorporate the absorption abilities of carbon nanoparticles (CNs) into chitosan (CS), creating an on-chip CN–CS composite membrane, which can perform the dual functions of absorption and dialysis	7
Fabrication of the hybrid collagen–chitosan membrane	To construct a potential extracellular matrix-like biomembrane, which possesses the good biocompatibility of collagen and great mechanical strength and processability of chitosan for on-chip cell cultures	31

Although the flow-assembly of the CM offers a rapid, facile, and reliable strategy to integrate biopolymer membranes with microfluidics, the solubility in acidic solutions and low-molecular weight cut-off (a few nanometers) of the flow-assembled CM might limit the applications of CM-integrated microfluidic platforms. To overcome these problems, several studies have attempted to tune the properties of the fabricated CM by crosslinking with glutaraldehyde³⁹ or tuning its pore size with co-assembled nanoparticles as sacrificial materials.²⁹ Meanwhile, the main inherent advantage of chitosan is its high affinity to bioactive reagents and biological components for the biological functionalization of microdevices. Studies have exploited this ability of chitosan as a substrate to successfully immobilize bioactive materials such as poly(N-isopropylacrylamide) nanogels,³⁰ carbon nanoparticles,⁷ or collagen³¹ into microfluidics. This enhances the functionality of the synthesized CBMs and the applicability of the CM-integrated microfluidic platforms for many biological and cellular studies. These developments are discussed further in this section.

4.1. Technical advances in the flow-assembly process

As aforementioned, one challenge of the direct assembly of the CM at the flow interface shown in Fig. 1(A) was the deposition of chitosan residues in the downstream channels that could disrupt the pressure balance between the flow streams and dislocate the membrane anchoring point. One solution to this challenge was to include an extra acidic input at the downstream connecting to the downstream microchannels as featured in Fig. 11A(b). The extra acidic flow continuously cleansed out any deposited chitosan residue and automatically balanced the pressure between the flowing streams. The extra acidic input greatly improved the success rate of the CM as the acidic flow rate could be much higher than those of the polymer and basic buffer flow streams, and the flow rates could be adjusted as needed to clean any downstream residues. One drawback of the extra input is that the polymer and buffer flow channels compartmented by the CM are no longer physically separated at the downstream, which limits the applications when complete compartmentalization is desired.

Fig. 11.

CM-integrated platforms as synthetic ecosystems for cell–cell signaling studies. (A) CM-facilitated assembly of multiple cell populations: (a) alginate molecular structure chelated with Ca²⁺; (b) assembling cells in alginate hydrogels with Ca²⁺ diffusing through the CM; (c) various configurations to assemble multiple cell populations in alginate hydrogels: (i) two CMs enclosing one E. coli population (green), (ii) one middle CM sandwiched by two E. coli populations (red and green), (iii) three layers of E. coli populations (blue, green, and red) sequentially assembled on one side of the CM, and then (iv) on the other side of the middle CM. Scale bars: 200 μm. (B) Stratified biofilm mimics for observing and controlling bacterial signaling: (a) schematic representation of flow dynamics that impact multicellularity signaling between E. coli’s transmitting and reporting cells separated by the middle CM. Bacteria were cultured with flows at 0.05, 0.1, 0.2, 0.2 (control), 0.3 or 0.4 μL min⁻¹ flow rates in channel A and no flows in channel B. (b) Representative fluorescence images of the transmitting and reporting cells for the case of 0.05 μL min⁻¹ flow rate; (c) the growth of reporting cells over time; (d) fluorescence intensity of reporting cells over time with deferred (0.1 and 0.2 μL min⁻¹), extinguished (0.3 and 0.4 mL min ⁻¹) or no (control at 0.2 μL min ⁻¹) DsRed protein production with increasing flow rates. (C) Modulation of distal cell–cell signaling: (a) schematic representation of quorum sensing (QS) between the transmitting and reporting cells; (b) schematic signaling flux from the transmitter to the reporter cells either enhanced or reduced by modulator cells; (c) schematic illustration of distally connected cell–gel composites in two microchannels; (d) typical cell optical density (OD) (left axis) and fluorescence intensity of the reporter cells (right axis) over time showing various signaling modulation effects; (e) estimated AI-2 concentration within the reporter cell–gel composites (left axis, solid) and estimated AI-2 concentration per cell OD (right axis, dotted). (D) Inter-kingdom synthetic ecosystems: (a) assembly of six separate yeast cell populations (labeled 1 to 6) along with the CM (referred to as fluitrodes in the study) by alternatively introducing yeast–alginate mixture solution, crosslinking with Ca²⁺, rinsing with PBS, and followed by enclosing with a protective PECM layer; (b) multilayered bacteria and yeast separated with the PECM on one single fluitrode; (c and d) extraction of bacteria and/or yeast by vacuuming for downstream analyses using Pluronic treatment to compromise the CM. Scale bars: 50 μm. (A and B) are adapted with permission from ScienceDirect.³⁶ (C) is adapted with permission from The Royal Society of Chemistry.⁷⁴ (D) is adapted with permission from Wiley.⁴⁰

In addition, it is noted that the downstream flow in the microchip represented in Fig. 1A(c) generally prevents sufficient membrane attachment at the downstream point, which reduces the success rate of CM formation. Thus, to better secure the formation of CM at the downstream point, Tibbe et al. designed a new microchip with an extra outlet, serving as an anchoring point to guide the membrane formation.⁵³ Furthermore, Jia et al. reported that by using microchips with low aspect ratio microchannels, the reliability of the flow-assembled CMs can be enhanced. With the use of low aspect ratio microchannels and by adjusting the flow rate of both flow streams relative to their viscosity, the pressure at the flow interface can simply be balanced, thus creating a stable localized pH gradient for the CM assembly.³⁷

To improve the reliability of the flow-assembly process shown in Fig. 1(B), a technical innovation using an add-on PDMS vacuum layer as needed was developed to dissipate air bubbles from small apertures. This technical advancement utilized the gas-permeable properties of PDMS, the typical material used for the microdevice fabrication process, to dissipate the air bubbles trapped in the aperture as schematically depicted in Fig. 3A(a and b). Once the air bubbles were vacuumed out by withdrawing a connected syringe or a squeezed nasal aspirator (Fig. 3A(a)), the chitosan and the alginate solutions came into contact to spontaneously form a PECM, followed by the formation of a CM by restarting the flows.^13,41 The idea to vacuum the air bubbles trapped inside a small aperture through the PDMS layer was adapted from the de-bubble process reported previously,⁵⁸ except that the add-on vacuum layer was not plasma bonded to the bottom PDMS microchips. This not only provides a rapid (usually from 9 to 20 minutes) and versatile strategy (applicable to different PDMS microchips) to actively remove air bubbles inside microchannels, but also allows the add-on vacuum layer to be reused as many times as possible. Using this technical advancement, the success rate of the flow-assembly of CM has been significantly increased to almost 100% for not only experienced users but also recruits. With the easy add-on vacuuming process, arrays of CMs in a three-channel network, as shown in Fig. 3A(c), were reliably fabricated by the introduction of solutions,⁴¹ which were used for the generation of static gradients; this is further discussed in Section 5. The CM assembled by this air-bubble-steering method was well-controlled by the flow and pH of the polymer solutions, and the membrane growth curve was similar to that with the pressure-balancing approach.^35,41 Most importantly, the properties and functionalities of the fabricated CM using this newly developed approach remain unchanged; the CM is freestanding and strongly adhered to the PDMS device, selectively permeable to small molecules and ions, and chemically communicating between the CM-separating compartments.^13,35

Fig. 3.

Optimizing the flow assembly of CM technique in microfluidics. (A) Gelation across PDMS by dissipating air bubbles trapped in the apertures using an add-on vacuum layer: (a) 3D schematic representation of a microfluidic chip with an add-on vacuum layer on top; (b) A–A^′ cross-section showing the dissipation of air bubbles through PDMS upon vacuuming; (c) in situ biofabrication of arrays of CMs facilitated with the add-on vacuum chamber: (i) air bubbles trapped in the apertures being vacuumed, and (ii) air bubbles were dissipated, allowing the interaction between chitosan and alginate macromolecules to form PECMs. (B) Direct gelation in the PDMS device with small circular pillars: three stages of the in situ formation of the CM without trapping air bubbles. Scale bars: 100 μm. (A and B) are adapted with permission from The Royal Society of Chemistry.^41,52

Meanwhile, Gu et al. reported a modification in the microchannel design, through which it was possible to prevent the trapping of air bubbles in the apertures, thus directly enabling the formation of the CM. In particular, the authors adapted the previously developed gradient generator design with small circular pillars as schematically illustrated in Fig. 3B. By carefully positioning the convex meniscus that emerged at the apertures when introducing the chitosan solution followed by the introduction of the alginate solution, the authors enabled the direct interaction of the two solutions to form the PECMs without the intervention of air bubbles.⁵² Next, hydroxyl ions continuously diffused from the alginate side via the PECMs to the chitosan side, creating the localized pH gradients for the formation of the arrays of CMs on the PECMs as previously described. However, the effects of the reported procedure on the growth rate of the membranes and their versatility to other microchip designs remain unclear and could be an interesting topic for future studies.

4.2. Modification of the properties of the flow-assembled chitosan membranes

The flow-assembled CM is freestanding, robust, well-aligned, and semipermeable to small molecules and ions. These unique characteristics have made the flow-assembled CM a promising platform for a variety of applications which are mentioned in Section 5. Nevertheless, several challenges still exist and have remained unresolved until recently. First, the fabricated CM can be easily detached when the PDMS microchip channels are treated with an anti-adhesion agent, such as Pluronic F-127, for the prevention of biomolecular and cellular adsorption on the PDMS device.³⁹ Second, the flow-assembled CM cannot be used in an aqueous environment where the pH is above 6.3, and 1× phosphate-buffered saline (PBS) is usually needed as a maintenance buffer.^3,13 Last but not least, the low-molecular weight cut-off of a few nanometers, the size of the antibodies, of the flow-assembled CM poses a problem if mass transport of macromolecules is needed.²⁹

To tackle the first two problems, Hu et al. used glutaraldehyde (GA) to crosslink the fabricated CM for enhanced resistance to anti-adhesion agents and acidic environment, and the properties of the GA-treated CM (GTCM) were investigated. First, after treating the CM with 10% GA to convert the CM into a GTCM, no obvious morphological changes were observed in the PBS solution (Fig. 4A(a)). The CM quickly dissolved in the acidic environment within 20 seconds (Fig. 4A(b)–(i, ii)), while for the GTCM, no shrinking or swelling was observed over an hour under the same circumstances (Fig. 4A(b)–(iii, iv)). These findings indicate that the GA crosslinking of the CM significantly increased the acidic resistance of the GTCM, which can expand the applicability of the CM-integrated microfluidic platforms. Second, the effects of GA treatment on the molecular organization of the flow-assembled CMs were examined through the measurement of optical retardance. Fig. 4A(c) shows the net optical retardance of both the CM and GTCM, revealing a significant decrease of about 40% after the GA crosslinking. This significant decrease in the optical retardance confirmed that GA treatment had an impact on the microstructural arrangement of the flow-assembled CMs.³⁹ Third, the adhesion robustness of the GTCM, determined by the pressure measurement approach reported by Luo et al.,¹³ was notably strengthened. Fig. 4A(d) shows the average critical pressures of the CM and GTCM before and after treatment with Pluronic F-127. The results suggest that GA crosslinking not only enhances the adhesion strength of the pure CM but also counteracts the robustness-compromising effects of Pluronic F-127 treatment.³⁹ Importantly, no significant change was found in the permeability of the membranes before and after GA crosslinking. Despite the above-mentioned findings, it is worth noting that the GA treatment would consume chitosan’s amine groups and could limit the ability to modify the CM with biomolecules and other substances.²⁴ Additionally, the use of GA as a crosslinker could raise an unwanted biocompatibility issue due to the residual cross-linker, and therefore, it must be considered carefully before usage.

Fig. 4.

Tuning CM properties. (A) Crosslinking the CM with glutaraldehyde (GA): (a) no obvious morphological changes between the CM and GA-treated CM (GTCM); (b) acidic resistance of the GTCM as compared to that of the CM: (i and ii) CM was dissolved by the acidic solution within a few seconds, while (iii and iv) the GTCM remained unchanged under the same condition for one hour; (c) optical retardance of the CM and GTCM; (d) the adhesion robustness measured as normalized critical pressure per membrane thickness of the CM and GTCM and those treated with Pluronic (CM_Pluronic and GTCM_Pluronic). (B) Tuning CM porosity with nanoparticles (NPs) as templates: (a) schematic representation of the tuning process to form porous the CM (pCM) by co-assembling NPs in the CM, crosslinking the CM with GA and dissolving nanoparticles with dimethyl sulfoxide (DMSO); (b) permeability of the membrane characterized with FITC-labeled dextran (F-dextran) molecules of various sizes (4, 10, and 70 kDa), and the percentages of different F-dextran passing through the tested membranes. (A and B) are adapted with permission from The Royal Society of Chemistry.^29,39

To improve the applicability of the integrated membranes for mass transport of macromolecules, it is highly recommended that the porosity of the membranes should be manipulated according to application needs. Co-assembled polystyrene nanoparticles, as a sacrificial template, were investigated to manipulate the porosity of the flow-assembled CM for broader applications as schematically depicted in Fig. 4B(a). Briefly, the CM with polystyrene nanoparticles (CM-nps) was flow-assembled in microchannels and treated with GA.⁵⁹ Then, dimethyl sulfoxide (DMSO) was used to remove the incorporated nanoparticles, resulting in the porous CM (pCM). Next, permeability tests with FITC-dextran (F-dextran) molecules of different sizes revealed the enlargement in the pore size of the pCM in comparison with that of the CM. Fig. 4B(b)–(i, iii) depicts the fluorescence images of the CM and pCM from the polystyrene nanoparticles of different sizes (25 and 200 nm; pCM25 and pCM200, respectively) observed in the permeability tests, while Fig. 4B(b)–(iv) quantitatively shows the corresponding F-dextran passing through the CM, pCM25, and pCM200 in terms of percentage. These results suggest that the procedure has successfully tuned the porosity of the CM, as the pCM showed improved permeability to macromolecules, confirming the capability to actively tune the porosity of the CM as application demands.

Besides, the crystalline structures and polymer chains alignment determined through optical retardance signals showed that GA treatment significantly influenced the crystallization of the flow-assembled CM, which is in agreement with the study of Hu et al.,³⁹ while DMSO treatment induced little impact on the microstructure of GTCM.²⁹ Similar tendencies were exhibited by the pCM that underwent the same treatment. The fact that no variation was observed in the optical retardance of the GTCM when treated with DMSO suggested that higher F-dextran transported across the pCM was not because of its altered microstructural organization. Instead, the variations in the mass transport of F-dextran of different sizes were probably induced by the difference in the interconnected pores instead of the crystallization of the membranes.²⁹ Further investigations to confirm these hypotheses should be considered, and future studies to prove the practical usability of the pCM in sorting a mixture of biomolecules are of interest. Furthermore, the concentration of the incorporated nanoparticles is another key factor that determines the porosity of the fabricated pCM, and further optimizations should be considered to attain the desired porosity and pore distribution for the fabricated membranes in the future.

4.3. Immobilization of bioactive reagents

One of the most important roles of chitosan in bioMEMS is to immobilize biomolecules, biological components, and other substances into microdevices. The abundant amine groups enable covalent attachment between chitosan and a variety of biomolecules and biological components, thus adding functionalities of microdevices for broader applications.²⁴ In this section, we summarize some of the modifications of the flow-assembled CM with biomolecules and other polymers (e.g., collagen) in microchips. Another exemplar of CM immobilized with mesoporous silica nanoparticles (MSNs) for personalized medical applications is discussed in Section 5.

In the first example, a carbon nanoparticle–chitosan (CN–CS) composite membrane was successfully fabricated in a microfluidic chip using the flow-assembly technique for simultaneous adsorption and dialysis applications.⁷ Fig. 5(a) depicts the fabrication process where the chitosan solution and buffer solution containing carbon nanoparticles (CNs) were introduced into the microchip through the A and B inlets, respectively. The polymerization reaction occurred at the flow interface initiating the formation of the CN–CS membrane under various flow conditions. During membrane formation, the membrane was thinner in the upstream and then thicker in the downstream, and the growth of the membrane was diffusion-limited, which resulted in uneven distribution of CNs inside the formed CN–CS membrane along with the flow interface. To improve this, CNs should be mixed with the chitosan solution instead of the buffer solution, which would result in a more uniform CN–CS membrane. Next, Fig. 5(b) shows that the growth of the CN–CS membranes could be classified into two stages: (I) convection-driven growth (where the growth rate of the membrane was fast, and the growth of the membrane was strongly affected by the convection transport of reactants along the flow direction) and (II) diffusion-driven growth (where the growth rate of the membrane was slow, and the growth of the membrane was influenced by the diffusion transport of reactants across the formed membrane). Additionally, the authors observed that the growth rate and thickness of the CN–CS membrane were, in general, larger than those without CNs, as shown in Fig. 5(c).

Fig. 5.

Fabrication of carbon nanoparticle–chitosan (CN–CS) composite membranes. (a) Schematic representation of the fabrication setup; (b) the averaged CN–CS membrane thickness over time. The black line indicates the average value of four repeated experiments; (c) comparison of the growth curves with and without CN; (d) creatinine adsorption of the CN–CS composite membranes and the control (without CN). Adapted with permission from Elsevier.^7,55

Furthermore, it is reported that as the flow rate increased, the permeability of the CN–CS membrane initially increased and then reduced. This was due to the fact that the longer the reaction time (which occurred with the small flow rate), the more compact the membrane. However, as the flow rate increased, the reaction rate reduced, leading to a less compact membrane. Additionally, a porosity-correlated mass transfer model was used to theoretically simulate the urea transport across the fabricated CN–CS membrane and the approximate porosity of the membrane was determined by fitting the theoretical data to the experimental results. The pore size of the CN–CS membrane was roughly determined to be smaller than 3 nm. Lastly, with the addition of CNs, the formed CN–CS membrane exerted strong creatinine adsorption, while the creatinine adsorption of the blank CM was not significant as shown in Fig. 5(d). Despite the promising results, it is important to note that the blood compatibility of the fabricated CN–CS membrane must be considered carefully and improved for future use of this membrane system as a micro-hemodialyzer.⁷

The second example explored the immobilization of poly(N-isopropylacrylamide) (PNIPAM) nanogels to form a PNIPAM nanogel-containing CM in a microchip as generally depicted in Fig. 6. Chitosan was used as a substrate material to embed PNIPAM nanogels into the membrane, enabling control over the permeability of the fabricated membrane³⁰ through the reversible swelling/shrinking volume transitions correlated with the variations in temperature^60,61 and ethanol concentration (C_E)^62,63 of PNIPAM nanogels. To construct such a desired membrane, a water-phase solution comprising chitosan and PNIPAM nanogels and an oil-phase solution composed of terephthalaldehyde were introduced into two converging microchannels. The nanogel-containing CM was in situ formed in microdevices through crosslinking reactions between chitosan and terephthalaldehyde instead of the localized pH gradient at the interface of the flow and trapped the PNIPAM nanogels inside the formed membrane. Fig. 6A(a) shows the fabricated CM and nanogel-containing CM assembled by the interfacial crosslinking between chitosan and terephthalaldehyde. The embedded nanogels were tagged with a red fluorescent dye to visualize their presence inside the constructed membrane. Fig. 6A(b and c) shows the blank CM and the nanogel-containing CM, in which the presence of PNIPAM nanogels in the crosslinked CM was confirmed with red fluorescence, which demonstrates the successful flow-assembly of nanogel-containing CM in microdevices.³⁰

Fig. 6.

Fabrication of the PNIPAM nanogel-containing CM. (A) Morphological characterization of the fabricated nanogel-containing CM: (a–c) microscopic images of the (a1 and b) blank CM and (a2 and c) nanogel-containing CM (scale bars: 250 μm). (B) Schematic illustrations representing the reversible swelling/shrinking transitions of PNIPAM nanogels in the CM in response to changes in its volume-phase transition temperature (VPTT) and ethanol concentration (C_E), therefore controlling the permeability of the nanogel-containing CM. Adapted with permission from The Royal Society of Chemistry.³⁰

The swelling and shrinking responses of the PNIPAM nanogels correlated with the variations in temperature and C_E were observed before being embedded into the constructed membrane. The results revealed that as the temperature increased from 25 to 40 °C (volume phase transition temperature – VPTT), the mean diameter of the nanogels reduced significantly and the critical ethanol concentration (C_C) value was determined to be around 8%, the point above which the nanogels experienced a significant decrease in size.³⁰ The nanogel content in the fabricated membrane also played an important role in controlling its self-regulated permeability. As the nanogel concentration increased, the more dramatic temperature-responsive permeability control was attained, and the optimal nanogel concentration was determined to be 40 wt%.³⁰ Next, the self-regulated permeability of the fabricated nanogels containing CM relative to the changes in temperature (Fig. 6B(d)) and C_E (Fig. 6B(e) in the microfluidic chip was examined using FITC. Briefly, at temperatures lower than the VPTT (T < VPTT), the nanogels in the membrane swelled and reduced the permeability of the membrane to FITC (Fig. 6B, a → b), while at temperatures higher than the VPTT (T > VPTT), the nanogels significantly shrank and increased the permeability of the membrane to FITC (Fig. 6B, b → a). Similarly, at 25 °C, as the C_E becomes lower than the C_C (C_E < C_C), the nanogels swelled and reduced the permeability of the membrane (Fig. 6B, b → c), while the C_E was higher than the C_C (C_E > C_C), the nanogels significantly shrank and increased the permeability of the membrane (Fig. 6B, c → b). Most importantly, the authors confirmed that the volume transitions of the nanogels that enabled self-regulation over the nanogel-containing membranes were reversible and repeatable. This enables a smart membrane platform for the development of micro-detectors, separators, sensors, or controlled release models. On the downside, this approach uses a crosslinking mechanism to enable such a membrane system;³⁰ therefore, great care has to be taken to neutralize the residual crosslinker for biological applications.

Besides being the substrate to immobilize biomolecules into microdevices, chitosan can be modified with other polymers to enhance the physicochemical and biological properties for broader applications. Collagen is a commonly used material for a wide range of biomedical and tissue engineering applications thanks to its excellent biocompatibility, biodegradability, and non-immunogenicity.^64–66 However, the molecular alterations of the collagen structure occurring during the extraction process generally result in poor mechanical strength.^67,68 Therefore, chitosan with excellent mechanical strength can serve as a support material for collagen–chitosan composites with enhanced biocompatibility inherited from collagen.⁶⁹ Rosella et al. reported a new microfluidic platform containing a collagen–chitosan hybrid membrane and explored its role as an extracellular matrix (ECM) support for biological applications.³¹ Fig. 7A(a) shows the illustrative design of the microfluidic system for membrane fabrication, in which four parallel experiments can be conducted simultaneously. Additionally, the effects of the flow rate on the growth rate of the hybrid membrane were determined. As shown in Fig. 7A(b and c), the solution flow rate significantly affected the growth of the hybrid membranes, in which the thickness of the fabricated membranes significantly decreased as the flow rate was increased and the thickness of the upstream membrane tends to be smaller than that of the downstream one. This was due to the fact that the faster the flow rates, the shorter the fabrication time, and the narrower the pH gradient, and ultimately, the thinner the membranes. Moreover, the faster flow rates induced higher shear forces at the surface of the formed membrane, leading to thinner and denser membranes as mentioned above.^7,31 Furthermore, polymer chains might also undergo repulsive interactions with the membrane surfaces under shear stresses, thus reducing the thickness of the membrane.^43,44

Fig. 7.

Fabrication of collagen-based matrices. (A) Fabrication of collagen–chitosan hybrid membranes: (a) design of a parallel membrane synthesis system with 4 X-channels; (b) formation of hybrid collagen–chitosan membranes for three total flow rates (Q_total): 1.03 (orange), 2.1 (yellow), and 5 mL h⁻¹ (green); (c) width measurements of the CM at the upstream (solid line) and downstream (dashed line) positions in the X-channel for the fastest (red) and the slowest (blue) flow rates used (b). (B) Fabrication of collagen microgels along with a CM in microfluidics: (a) experimental setup to introduce a collagen–alginate mixture and divalent calcium ion (Ca²⁺) solutions into two parallel microchannels separated by a CM. Ca²⁺ ions diffuse across the CM crosslinked with the alginate and form an aligned collagen–alginate composite microgel adjacent to the CM; (b) images of the aligned collagen–alginate microgels (green dashed) formed with different calcium concentrations (0.025, 0.1, 0.5, and 1.0 M); (c) the effects of calcium concentrations on the collagen-based microgel thickness over time; (d) its compression from peak thickness to steady-state thickness at 60 seconds after the initial growth. (A) is adapted with permission from Elsevier.³¹ (B) is adapted with permission from IOP Publishing.⁷⁰

On the other hand, the biofabrication of a stable and highly aligned collagen matrix mimicking that of the native tissue ECM is a long-standing design goal for biomedical and tissue engineering research.^71–73 To achieve this goal, Correa et al. demonstrated the fabrication of aligned collagen–alginate microgels through ionic crosslinking with divalent calcium ions. In the fabrication process, the CM was used as a barrier membrane to prevent the mixing of the collagen–alginate co-polymer solution and calcium chloride solution pumped into two parallel microchannels, as schematically illustrated in Fig. 7B(a). Then, the calcium ions diffused through the cross-channel CM to interact with alginate in the collagen–alginate solution and crosslinked with the copolymer to form a hydrogel-like structure. Fig. 7B(b) shows the fabricated collagen–alginate gels with different calcium concentrations formed within seconds as isolated islands in the single-aperture PDMS devices. Fig. 7B(c) shows the gel thickness characterization results, revealing that the gel thickness increased significantly with the increased calcium concentration. It is also reported that the higher the calcium concentration, the lower the gel compression (Fig. 7B(c)). Furthermore, the aligned and stable collagen network was confirmed with the birefringence signal and circumferential texture orientation of the fabricated collagen–alginate microgels. Minor variations were observed in the dimensions of the fabricated gels after alginate was removed from the microgel structure. The collagen concentration of 8 mg mL⁻¹ was determined to be the optimal concentration to form the most stable and asymmetric collagen microstructure matrix. The study reveals that the experimental parameters (co-polymer, calcium concentration, and solution flow rate) can affect the alignment and stability of collagen gel formation in microfluidic devices. Such an alignment approach by ion diffusion through the CM enables gel formation with various microstructures, and can be valuable for lab-on-a-chip and tissue-on-a-chip applications.⁷⁰ Notably, the biofabrication of localized collagen gel for cell seeding along the CM provides, for the first time, the spatiotemporal controllability with chemicals in flow that normally is only controllable with temperature, which can be further explored in important cellular studies and tissue engineering modelling.

5. Applications of flow-assembled chitosan membranes in microfluidic platforms

The flow-assembly of CMs and CBMs is a relatively new platform technology with vast potential to explore. In the previous sections, we have presented the flow-assembled CM fabrication approach, recent advances in the flow-assembly of CM and CBM platforms, and the characteristics of the fabricated CM. This section is dedicated to reviewing the practical uses of CM-integrated microfluidic platforms for a variety of applications from biochemistry to biology to drug screening including static gradient generators,³⁵ platforms for shear-free cell culturing,^31,53 constructing synthetic ecosystems for cell–cell communication studies,^36,40,74 and drug-screening.

5.1. Static gradient generators

Chemical gradients play an essential role in directing cellular activities during chemotaxis, differentiation, inflammation, and many other biological processes.^75–78 Microfluidic-based gradient generators that enable highly controllable and quantifiable chemical gradients in a time- and cost-saving manner can be promising alternatives over conventional gradient generation models.⁷⁹ Compared to flow-based steady gradient generators, diffusion-based static gradient generators are more favorable for cellular studies thanks to their minimal convection and shear stress induced by laminar flows.^80–82 Chemical gradients can be readily established through diffusion-based transport along central microchannels connecting two side microchannels containing solutions of high concentration (source) and low concentration (sink). Various strategies have been developed to mitigate the convective flow while enabling the diffusion of small molecules to establish chemical gradients, thus providing a convection-free culturing microenvironment. Luo et al. innovated a static gradient generator comprising fourteen parallel CMs flow-assembled in a three-channel microfluidic device.³⁵ The middle microchannel acts as a static gradient chamber separated by the two side microchannels (the source and sink microchannels, respectively) with two parallel CM arrays that were 400 μm apart. The flow-assembly of parallel CMs commenced with trapped air bubbles in the apertures, followed by membrane formation as previously reported.¹³ The fabricated CM was uniform, strong, and semi-permeable, which enables the diffusion of small molecules to establish the static gradients in the middle channel.

Fig. 8A(a)–(i) illustrates the experimental setup to establish static gradients in the middle microchannel of the three-channel microfluidic devices described above. Briefly, FITC and PBS solutions were pumped into the source (left) and sink (right) side microchannels, respectively, while the central microchannel was filled with PBS and maintained under stop flow conditions. Fig. 8A(a)–(ii–iv) displays the evolution of static chemical gradients at 20, 60, and 600 seconds. The fluorescence intensities across the middle channel at specific time points were measured and are plotted in Fig. 8A(b). It is clear that the chemical gradients evolved and an approximately linear static gradient was established after five minutes, and maintained well after ten minutes.³⁵ Furthermore, this gradient generator was employed to establish the static gradient of α-factor to monitor the morphological changes of yeast over time that is discussed in Section 5.2. In the follow-up study, the porosity of the fabricated CM was manipulated to enlarge the size of the pores, resulting in the pCM-based gradient generator that enabled the generation of macromolecule (F-dextran) gradients similar to the FITC gradients generated using the CM-based gradient generator.²⁹ This suggests that the proposed microfluidic-based gradient generator can be a facile and versatile platform technology for studies where static chemical gradients are desired.

Fig. 8.

CM-integrated microfluidic platforms for biochemistry processes and analyses. (A) Generation of a static gradient in a three-channel microfluidic device composed of parallel and semipermeable CMs: (a)–(i) experimental setups to generate a static gradient in the middle microchannel and (a)-(ii–iv) the established static gradients over time; (b) the evolution of fluorescent gradients of the plot profiles indicated in (a) over time. (B) Measurement and mitigation of convection in the CM-incorporated microfluidic gradient generator: (a) schematic top-view (xy plane) of the three-channel microfluidic device and side-view (xz plane) of the solute-driven flows in the middle channel; (b) fluorescence image of the steady gradients across the center channel with a superimposed plot profile of the fluorescence intensity. Scale bar: 100 μm. Scale bar: 100 μm. (B and A) are adapted with permission from The Royal Society of Chemistry.^35,52

It should be noted that even when there is no external flow, solute gradients can still generate steady convection by themselves due to the presence of buoyancy-driven and diffusioosmotic flows. Gu et al. adapted the above-mentioned gradient generator containing in situ biofabricated parallel CM arrays to investigate the fluidic flows induced by steady solute gradients relative to various system parameters including gradient magnitude, viscosity of fluid, microchannel dimensions, and solute type.⁵² Fig. 8B(a) shows a similar microfluidic design with W = 400 μm, L = 2 mm, and H = 35–120 μm to that used in the study by Gu et al. On applying different concentration solutions, c₁ and c₃, into the left- and right-side microchannels, respectively, the stable concentration gradient was established after ten minutes. The fluorescence intensity reduced linearly across the middle channel and the measured gradient was about 50% smaller than (c₁ – c₃)/W due to the resistance induced by the CM (Fig. 8B(b)). Through the modeling and experimental results, the authors suggested that some ways to minimize the buoyancy-driven flows were to increase the viscosity of the solution, reduce the height of the channel to increase the viscous resistance to fluid motion, and decrease the difference in gravitational pressure. Furthermore, the buoyancy flows were reported to be temperature-dependent, where such flows would be 30% larger at 37 °C as compared to those that emerged at normal room temperature (22 °C). Meanwhile, the diffusioosmotic flows were independent of the channel’s height and magnitude of the concentration difference but usually occurred in short microchannels derived from small concentration differences. To avoid such flows, it is better to employ concentration gradients where the magnitude of minimum concentration and maximum concentration was comparable. These system parameters should be designed carefully to mitigate the effects of buoyancy and diffusioosmotic flows for better quantification of cell chemotaxis and phoretic motions within the colloidal solution.⁵² It is worth noting that the solute gradient-induced convection is particularly true when the solute concentration and the gradient difference are in the high range of tens to hundreds of millimolars, in which steady flows of several microns per second were predicted even within the small channels of the microfluidic systems. Fortunately, cellular activities such as bacterial chemotaxis in reality are most apparent in the range of micromolar to a few millimolars.^83–86 Within this biological range, the buoyancy-driven and diffusioosmotic flows are minimal, and the gradients generated with parallel CM arrays in the three-channel networks can be fairly assumed to be static.

To conclude, microfluidic devices offer many advantages in generating concentration gradients for biochemical and cellular studies.^87,88 For instance, microfluidic-based gradient generators can produce reproducible, predictable, and quantifiable gradients at low sample cost yet with fast response time.^88–90 The above-described static gradient generator comprised parallel CMs in a three-channel PDMS device that possesses many properties of a good gradient generator model. The fabrication process of this platform is rapid, facile, and robust, and the static gradients are quickly generated (within a few minutes) and stable over time in the static chamber.³⁵ However, when a high range of solute concentration and gradient difference is needed, several system parameters should be taken into consideration to mitigate the buoyancy and diffusioosmotic flows as discussed above.

5.2. CM-based platforms for cell and tissue culture

Over the past two decades, integrated microfluidic platforms have been widely exploited for many biomedical and tissue engineering applications owing to their low reagent consumption, rapid fabrication, high sensitivity and controllability, and economic effectiveness.⁹¹ These microfluidic platforms in combination with tissue engineering and cell biology applications have enabled the development of organ-on-a-chip systems. Several exemplar organ-on-a-chip systems include gut-on-a-chip,⁹² liver-on-a-chip,⁹³ and lung-on-a-chip.⁹⁴

Using the flow-assembly platform technology, Rosella et al. developed microfluidic platforms consisting of biomembranes that acted as scaffolds closely resembling the native ECM for organ-on-a-chip applications.³¹ In their study, three types of biomembranes were investigated including collagen, chitosan, and hybrid collagen–chitosan membranes. Moreover, CM and chitosan were used as a scaffold to support cell proliferation and as a substrate to enhance the mechanical strength of hybrid collagen–chitosan membranes, respectively. The system parameters (e.g. biopolymer type and solution flow rate) involved in the membrane fabrication process that affected the width, uniformity, and swelling ratio of the membrane were well-characterized. The results showed that the properties of the fabricated membranes were flow-dependent, revealing an opportunity to customize and optimize the properties of the membrane accordingly. Furthermore, the cell biocompatibility of the fabricated biomembranes was tested with 3T3 fibroblast cells. The cells were injected and cultured in the corresponding biomembrane-integrated microfluidic platform for seven days and stained with live/dead assays. Fig. 9A(a) shows the representative live/dead stained images of 3T3 fibroblasts at day 7 on collagen, chitosan, and hybrid collagen–chitosan membranes, respectively. As shown in Fig. 9A(b), collagen membranes supported a significantly greater level of cell viability as compared to CMs. Notably, there were no significant differences in cell viability among the collagen and hybrid collagen–chitosan membranes. This suggested that the hybrid collagen–chitosan membranes possibly possessed not only great mechanical properties and processability of chitosan but also excellent biocompatibility of collagen, showing the potential for biological studies of such membrane systems.³¹

Fig. 9.

CM- or CBM-integrated two-channel platforms for cellular studies. (A) Collagen–chitosan hybrid membranes for on-chip cell culture: (a) live/dead (green/red) stained images of 3T3 fibroblast cells one week after being seeded on (i) collagen, (ii) chitosan, and (iii) collagen–chitosan hybrid membranes; (b) cell viability for each membrane type (* indicates p < 0.05 using one-way ANOVA analysis, n = 4). (B) Temporary CBM for organ-on-a-chip applications: (a and d) astrocytes are homogeneously distributed within the extracellular matrix-like gel (Matrigel) in one channel marked with an asterisk (*); (b and e) 18 h after seeding the astrocytes, the temporary CM is completely removed using acetic acid (pH = 5.0). The morphology of the astrocytes is not influenced by the removal of the CM; (c and f) subsequently, the channel is coated with a fibronectin solution, followed by the seeding of brain–endothelial cells in the empty channel. The endothelial cells reach almost 100% confluence after 24 h. (A) is adapted with permission from Elsevier.³¹ (B) is adapted with permission from Wiley.⁵³

An advantage of using CMs in the organ-on-a-chip field is their ability to be removed with a mild acidic solution to enable a direct cell–cell interface. Previously, it was challenging to recreate a native interface between parenchymal and vascular endothelium tissues in blood–brain barrier microfluidic models. While commercial polymer membranes might not mimic the stiffness, porosity, and interconnected porous structure of the native basement membrane, the utilization of the CM as the temporary membrane could create direct contact between two tissue types, thus enabling the coculture of multiple cell types in the absence of a synthetic membrane.⁵³ Fig. 9B shows the overall experimental process. Initially, the CM was flow-assembled at the interface between the acidic chitosan and the basic buffer solutions as described above. Next, human astrocytes in Matrigel were seeded and cultured for one day at the bottom microchannel (Fig. 9B(a and d)), followed by the removal of the CM using a mild acidic solution (Fig. 9B(b and e)). Subsequently, the other microchannel was coated with a fibronectin solution prior to the seeding of brain endothelial cells, resulting in a human astrocytes-brain endothelial cell coculture that enabled a direct cell–cell interaction of the two cell types inside the microfluidic chip (Fig. 9B(c and f)). Thus, the proposed membrane fabrication and removal process can be employed to produce membrane-free cocultures in microdevices for broader organ-on-a-chip applications. Perfusion with flow for long-term culturing is yet to be integrated. For future studies, the possible cross-invasion of both cell types over long-term culturing without the barrier membrane is worth investigating.

5.3. CM-based platforms for cellular signalling studies

Besides the mammalian cell culture applications, CM-integrated microfluidic platforms have been demonstrated to be useful for a variety of bacterial chemotropism and bacterial cell–cell signalling studies.^95,96 For instance, the ability to generate static gradients within the middle microchannel of a three-channel microfluidic platform containing parallel CMs is utilized to study the evolution of the morphological changes of yeast under mating pheromone α-factor gradients over time,³⁵ and to monitor the chemotropism between adjacent populations of multiple types of yeasts.⁴²

For the first application, it was reported that the target strain of the yeast would react with the mating pheromone α-factor and change to an elongated shmoo-shaped yeast with the tips intrinsically grown towards the source of the α-factor. On applying the growth media (YPD) containing the in vitro α-factor and normal YPD solutions to the source and sink microchannels, respectively, as depicted in Fig. 10A(a), no obvious morphological change of yeast was noted for the first 100 minutes. Then, the yeast started to grow into elongated shapes and stopped dividing after three hours. Fig. 10A(b, d and e) shows the zoomed-in images of the middle microchannel where the morphology of the yeast shows a gradient-dependent transition in the aspect ratio (length ratio of long axis to short axis). The gradient-dependent transition was well-distinguished within the middle microchannel where more yeasts were elongated from the source (right) towards the sink (left) side microchannels. Fig. 10A(f) shows the quantification of the average aspect ratio of the yeasts over time in five separate blocks, as indicated in Fig. 10A(c). A gradient of an increasing aspect ratio of the yeasts was observed across the middle microchannel from the source to the sink side microchannels over time. Thus, the flow-assembly of the CM provides a rapid and facile platform technology for static gradient generation studies and is promising for many biological studies where convection-free static gradients are vital.³⁵

Fig. 10.

CM-integrated three-channel platforms for cellular studies. (A) Monitoring the morphological changes of yeasts in response to static gradients of in vitro alpha factor (α-factor): (a) yeast cells in the middle microchannel were exposed to the static gradients of α-factor with continuous flows of growth media (YPD) and α-factor in YPD in the sink (left) and the source (right) side microchannels, respectively; (b) zoomed-in image of the red rectangle in (a); (c) morphology of the yeasts in the five separate blocks to quantify the average aspect ratio; (d and e) zoomed-in shmoo shapes of the yeasts in the left and right red rectangles in (b), respectively; (f) Average aspect ratio of the yeasts in the five separate blocks over time. (B) Monitoring the in vivo chemotropism between adjacent populations of yeasts: (a) schematic representation of the A- and α-yeasts assembled side-by-side in the alginate hydrogel assembled in the middle microchannel of a three-channel microfluidic device; (b) relative uniform distribution of A- and α-yeasts one hour after assembly; (c) distribution and morphologies of A- and α-yeasts eight hours after assembly; (d) proliferation ratio of A-yeasts in the four indicated blocks in (b and c); (e) the projection ratio of A-yeasts in the same four blocks; (c–e) together shows the chemotropism of A-yeasts closer to α-yeasts as compared to continuous proliferation of A-yeasts away from α-yeasts. (A) is adapted with permission from The Royal Society of Chemistry.³⁵ (B) is adapted with permission from Biomicrofluidics.⁴²

Numerous microfluidic platforms have been developed to monitor the chemotropic responses of cells relative to the in vitro generated chemical gradients of synthetic mating peptides, and to provide new insights into a better understanding of chemotropism in multiple cell types.^97–101 Nevertheless, the chemotropic response of cells to in vivo chemical gradients is dynamic and cannot be closely modeled with synthetic mating pheromones. Additionally, cells in nature tend to be heterogeneous, which may affect the chemotropic responses or the established chemical gradients. Vo et al. proposed a CM-based microfluidic platform that contains populations of multiple yeast strains, the mating A- and α-yeasts, to monitor their mutually chemotropic responses with spatiotemporal programmability and sensitivity. Fig. 10B(a) shows the schematics of A- and α-yeasts positioned side-by-side in alginate hydrogels in the previously developed three-channel microfluidic platform containing parallel CM arrays to allow nutrient diffusion from the side microchannels. The strain A of the yeasts was positioned at the left-hand side separated from the α-yeasts at the right-hand side, where the yeast distribution was relatively even after one hour (Fig. 10B(b)). After eight hours, A-yeasts in blocks 1 and 2 that were far from α-yeasts continued to grow while those in proximity with α-yeasts formed shmoo structures and underwent cell cycle arrest, indicating the mating response to in vivo pheromone gradients (Fig. 10B(c)). Furthermore, Fig. 10B(c) also shows that the proliferation of α-yeasts was independent of the closeness to A-yeasts.

Fig. 10B(d) quantitatively shows a gradual decrease in the proliferation ratios of the A-yeasts from blocks 1 to 4 over the first five hours, indicating that the pheromone secreted by the α-yeasts had initiated the mating responses and induced cell proliferation of A-yeasts. In contrast, the projection ratios increased steadily from block 1 to 4, illustrating the chemotropic responses of the A-yeasts towards the pheromone emitted from the α-yeasts and induced morphological changes (Fig. 10B(e)). These suggested that the chemotropic responses of the A-yeasts relative to the mating pheromone secreted from the α-yeasts were spatially dependent, while α-yeasts were not as sensitive to A-factor as A-yeasts to α-factor. Furthermore, the authors reported that the chemotropic responses of the A-yeasts were dependent on cell density, with individual A-yeasts developing into shmoo shapes tending to direct more precisely toward the α-yeasts. Meanwhile, a cluster of A-yeasts received less pheromones from the α-yeasts and tended to proliferate more except that those around the edge of the cluster could sense pheromone better, and therefore shifted from proliferation to cell cycle arrest and then to shmoo formation. Lastly, the direction of shmoo projection of the A-yeasts was concentration-dependent in microscale-spatial resolution. In sum, the work demonstrated that the CM-based microfluidic platform could be used to assemble and investigate population-scale, spatial-sensitive cell–cell signalling behaviors similar to in vivo and chemotropism between multiple cell types.⁴²

5.4. Biofabricated synthetic ecosystems

Furthermore, communications between multiple cell types in a complex heterogeneous microenvironment not only direct their biological phenotypes but also reassemble the synthetic environments of native ecosystems spanning different length scales. To enable such communications, Luo et al. presented an approach to biofabricate multiple populations of cells in microfluidic devices with spatiotemporal programmability.³⁶ As schematically illustrated in Fig. 11A(b), the authors utilized the permeability of the CM that allowed Ca²⁺ ions to diffuse from one microchannel through the membrane and interact with alginate containing cell solution on the other side. The Ca²⁺ ions would ionically crosslink with the carboxylate groups on the guluronic acid residue of alginate to biofabricate a three-dimensional cell–gel composite (Fig. 11A(a and b)). The thickness of the formed cell–gel composite could be controlled with the diffusion time of Ca²⁺ ions via the CM or the concentration of Ca²⁺, and then the uncrosslinked alginate solution would simply be rinsed with deionized water. The proposed biofabrication of multiple cell types in microdevices provides a facile, rapid, and versatile platform system for in vivo cell–cell communications in the modelled heterogeneous microenvironmental studies. The semipermeable CM not only acts as a barrier between the two cell populations but also acts as a supporting backbone for the alginate hydrogels. Furthermore, this platform system allows independent fluidic access to each cell layer and manipulations over generated biological signalling by varying the system parameters such as the fluidic flows inside the microchannels.³⁶ It is also possible to assemble multiple layers of alginate hydrogels containing multiple cell types in various configurations as shown in Fig. 11A(c). Thanks to the spatiotemporal programmability in assembling cell populations in the microenvironmental length scale of biological relevance, and such platform systems were used to innovate synthetic ecosystems for biological applications.

First, a platform system consisting of two adjacent cell populations of transmitting and reporting cells separated by a CM in between was developed, which allowed for direct observation and manipulation of autoinducer-2 (AI-2) quorum sensing (QS) signalling.³⁶ Fig. 11B(a) schematically depicts the assembly of the transmitting and reporting cells containing alginate hydrogels in microchannels B and A, respectively. Briefly, the transmitting cells simultaneously secreted AI-2 signalling molecules and constitutively expressed GFP for direct observation. The reporting cells then sensed the secreted AI-2 and instantaneously produced DsRed, evidence of QS response¹⁰² and an example for many AI-2-induced behaviors.¹⁰³ It was reported that during the first five hours of culturing, the cell density of both strains (transmitting and reporting cells) increased rapidly and started to leak out of the alginate hydrogels at around the 15th hour. As seen from Fig. 11B(b–d), after the first five hours, the AI-2 concentration secreted by the transmitting cells reached the adequate amount that the reporting cells sensed and expressed the QS responses by emitting DsRed and reached a plateau after 11 hours. In the representative case of 0.05 μL min⁻¹ flow rate in the reporting channel shown in Fig. 11B(b), a gradient of QS response from the upstream flow to downstream was observed at the time points of 5th and 8th hour. This was due to the shear stress applied on the upstream channel that carried away the AI-2 and delayed the QS behaviors of the upstream reporting cells. By varying the flow rate of nutrients supplied in microchannel A and stopping the flow in microchannel B, the authors observed a significant trend that as the flow rate increased, the DsRed reduced dramatically. Particularly, the DsRed fluorescence intensity increased steadily at the flow rates of 0.05, 0.1, and 0.2 μL min⁻¹, while that remained relatively low at the flow rates of 0.3 and 0.4 μL min⁻¹ (Fig. 11B(d)). This suggested that the AI-2 level in the case of high flow rates (0.3 and 0.4 μL min ⁻¹) never reached the adequate amount, since the high flow rate continuously diluted or created high shear forces throughout the microchannel and disposed of the AI-2 secreted by the transmitting cells. The results demonstrated that QS responses in the stratified biofilms could be simulated and manipulated by adjusting the flow conditions using the proposed platform technology that allowed for cell–cell interaction studies and small-molecule drug discovery.³⁶

Another notable application of CM-integrated microfluidic systems is the modulation of E. coli cell–cell signalling in close and distal proximity.⁷⁴ The developed microfluidic system consisted of two individual microdevices that were connected by flexible tubing to transmit AI-2 signalling molecules, which were produced by the transmitting cells residing in the upstream device, to the modulating cells, which either amplified (enhancer) or attenuated (reducer) the signals, before transmitting to the reporting cells. Both the modulating and the reporting cells were in the downstream device (Fig. 11C(a–c)). Such a microfluidic system allowed for the modulation of longitudinal transport of small molecules produced by E. coli that closely resembled distant signalling pathways observed in the human intestinal tract.^74,95 The results reported that the molecular signals could be transmitted from the transmitting to the reporting cells located at a long commuting distance. Fig. 11C(d) shows the growth rate of the reporting cells in optical density (left axis) and fluorescence intensity (right axis). Overall, the reporting cells grew consistently over time, and they took approximately two hours for signal transport and four hours for QS responses. There were significant differences in fluorescence intensity among different configurations, in which the fluorescence intensity was the highest when the reporting cells were assembled with an enhancer (red), followed by that with a clear alginate hydrogel (control, yellow). Meanwhile, the fluorescence intensity of the reporting cells assembled alongside with the reducer reduced tremendously (blue) and was at the lowest with two reducers (black). Furthermore, the concentration of AI-2 available for the reporting cells in various configurations (assembled with enhancer, control, and reducer) was also approximated using numerical simulation that showed a strong agreement with the real-time fluorescence intensity from the reporting cells (Fig. 11C(e)). Hence, the reported microfluidic system reassembles a synthetic ecosystem mimicking the human gastrointestinal tract, and can be applied for various cell populations such as epithelial cells and manipulating effector molecules (glucose, hormones, and ions) with spatiotemporal control.⁷⁴

Despite the promising applications of such biofabricated platforms in monitoring QS responses in close and distal proximity, several problems remain to be solved. First, the mechanical strength of alginate hydrogels is relatively weak that can easily be delaminated from the PDMS device under strong laminar flows.³⁶ For this, the length and thickness of the assembled alginate hydrogels have to reach a certain size to withstand shear forces and secure their location within microfluidic channels. Second, the shear stress induced by laminar flows can remove the signaling molecules, resulting in the transient gradient with QS responses as observed in Fig. 11B(b). Furthermore, the alginate hydrogels can be degraded quickly without the continuous supply of Ca²⁺, thus leading to an unwanted mixing of multiple cell populations. To avoid this scenario, the growth medium is usually supplied with a low concentration of Ca²⁺. This, however, produces a new problem as Ca²⁺ is a sensory ion for gene expression of biofilm-associated growth and can alter bacterial adhesion and biofilm formation,^104–106 which may influence the experimental results.

Pham et al. developed a new microfluidic platform consisting of 12 individually addressable CMs, referred to as fluitrodes, to enable the programmable assembly of multiple cell populations in alginate hydrogels in the length scale of biological relevance (Fig. 11D(a)). The concept of individually addressable fluitrodes allowed the separate delivery of nutrients and signaling molecules, enabled the assembly of more cell populations (up to 12 different cell species) in one single device, and facilitated the release of embedded cells for further bioassays and molecular analyses.⁴⁰ Specifically, six alginate hydrogels containing yeasts of the same strain were sequentially assembled on the six individually addressable fluitrodes by the biofabrication process described above and in the previous publication.⁴⁰ To protect the assembled alginate hydrogels and prevent the leakage of cells, the authors enclosed the six microbial constructs with the PECM, as illustrated in Fig. 11D(a), by allowing chitosan to interact with the alginate hydrogels. The enclosure of the PECM also significantly improved the mechanical robustness of the assembled alginate hydrogels under a strong laminar flow rate of around 1250 μL min⁻¹. Next, it was demonstrated that multiple species of cell populations could be assembled side-by-side on one single fluitrode, enabling the monitoring of different species’ responses to the same types of nutrients or signaling molecules in juxtaposition. The side-by-side alginate hydrogels were also separated with the PECM in between to prevent the unwanted mixing of different cell types (Fig. 11D(b)). Lastly, this fluitrode platform allowed for the sequential release of embedded cells for downstream analyses. For this purpose, the CM was compromised and detached from the apertures using Pluronic F-127, enabling the sequential extraction of embedded cells (from both separate yeast populations and multilayers of yeast and bacterial cells) with vacuum pressure as shown in Fig. 11D(c and d). No significant effect, however, was observed on the viability of bacteria embedded in the CM-supported alginate hydrogel and maintained under physiological conditions despite chitosan being a prominent antibacterial and antifungal agent. This agrees with previous studies that reported no inhibitory effects of chitosan/alginate composites on bacterial growth.^107,108 The main reason for this can be due to the mode of inhibitory action of positively charged chitosan. Similar to the antibacterial mechanism of metallic nanoparticles,^21,109,110 amino groups carrying positively charged chitosan can easily attract and penetrate the negatively charged cell membrane of bacteria, then disrupt the respiratory process and cause bacterial death.^111,112 Therefore, the culture environment should be maintained at physiological pH to prevent the protonation of amino groups that may affect bacterial viability and intervene experimental results.

In summary, the developed fluitrode platform provides real-time observation of in vitro synthetic ecosystems for cell–cell communication in their complex heterogeneous microenvironment, and can broaden the applications of CM-integrated microfluidic platforms in high-throughput drug screening.⁴⁰ Furthermore, this study addressed some of the challenges experienced in previous works with the addition of the PECM to protect and secure the assembled alginate containing cells in hydrogels in the absence of extra Ca²⁺ ions, and to prevent the mixing of different cell types.

5.5. Drug delivery

Drug delivery systems and drug screening are among the most important applications of microfluidic platforms. The superior advantages of low sample consumption, fast reaction time, cost-saving, high throughput, and reproducibility can aid in the development of new drugs and strategies for efficient drug delivery.^11,113,114 Jia et al. have explored the CBM-integrated microfluidic platforms to program complex release of nanocarriers.³⁷

Mesoporous silica nanoparticles (MSNs) are a nanoparticle-based drug delivery system that has gained significant research interest thanks to their high surface areas, large pore volume, and tunable porosity that can protect therapeutic agents for controlled and targeted drug delivery. The controlled release of MSNs is based on a variety of physical, chemical, and biological stimuli.^115,116 However, the controlled release based on the pH-responsiveness of MSNs usually requires complicated chemical grafting methods that can result in pore blockage or toxicity issues.¹¹⁷ Jia and colleagues proposed, for the first time, a pH-responsive CM containing MSNs for drug delivery in microfluidics.³⁷

Fig. 12(a) illustrates the schematic design of the microfluidic device with multiple upstream microchannels for rapid changing of different chitosan containing MSNs, base, and PBS solutions, allowing for the synthesis of multi-layered CM containing MSNs. For instance, a bilayer CM containing positive MSNs with FITC-tag (MSN_F⁺ green) and those with rhodamine B-tag (MSN_F⁺ red) was assembled at the flow interface as shown in Fig. 12(b). Different types of MSNs embedded in the layer-by-layer CM remained unmixed over the tested time frame (4 hours). Next, the release of the embedded MSNs using a mild acidic solution (pH = 5.0) over time is demonstrated in Fig. 12(c). A gradual decrease in the membrane thickness and subsequent release of the embedded MSNs were observed within the first ten minutes and completely dissolved after 15 minutes with the erosion rate of around 480 μm h⁻¹. To enable delayed dosing and sustained drug release, the authors developed two-MSN-capped CMs separated by a pure CM as depicted in Fig. 12(d)–(i). The release profile of such a membrane system in an acidic solution (pH = 5.9) could be prolonged for up to two hours as shown in Fig. 12(d)–(ii). Lastly, a complex 7-layer CM capped with MSNs was also assembled by flows with the release profile in an acidic solution (pH = 5.9) that lasted for up to four hours (Fig. 12(e)). Hence, by accurately manipulating the precursor solution flow rates for highly programmable membrane formation, MSN-embedded CMs with complex layered architectures for customizable drug release were successfully presented. The presented flow-assembled CBM-based platform for complex release profiles of embedded therapeutic agents is attractive and can be applied for a wide range of biomedical applications and personalized therapies.³¹ Importantly, the release profile of CM-embedded therapeutic agents can be easily manipulated by treating the membrane with crosslinking agents.

Fig. 12.

CM-integrated microfluidic platform for complex release profiles of nanocarriers: (a) schematic representation of the X-channel showing the inlets of basic (red), chitosan (light and dark blue), and PBS solutions for step-wise control over the flow-assembly of the membrane (yellow) at the aperture; (b) fluorescence image of a bilayer CM containing positively charged mesoporous silica nanoparticles (MSN⁺) with FITC-tag (MSN_F⁺) and with rhodamine B-tag (MSN_R⁺), respectively; (c) cross-sectional profiles during membrane dissolution at (i) 0 min, (ii) 4 min and (iii) 10 min after pH of 5.0 solution was introduced to the bottom channel at 1.0 mL h⁻¹; (d)–(i) fluorescence image of 3-layer CM with the first and third layers containing MSN_F⁺; (d)-(ii) normalized instantaneous (solid) and cumulative (dashed) MSN_F⁺ release profiles during dissolution; (e)-(i) Fluorescence image of the 7-layer CM with MSN_F⁺ (green) in layers 1, 3, 5, and 7, with MSN_R⁺ (red) in layers 2, 4, and 6; (e)-(ii) instantaneous MSN release profiles during dissolution. A solution (pH = 5.9) was introduced to the bottom membrane side at 1.0 mL h⁻¹ for dissolution in (d)-(ii) and (e)-(ii). Adapted with permission from The Royal Society of Chemistry.³⁷

6. Conclusions and future perspectives

Chitosan has been demonstrated as a valuable material for broad biological functionalization in bioMEMS. Using laminar flows to deposit freestanding CMs inside microfluidic networks is a rapid, facile, and versatile approach to integrate biology (or biological materials) with inorganic microdevices. Furthermore, the ability of in situ fabricating freestanding biopolymer membranes inside microchannels not only overcomes the unwanted leakage, insufficient sealing, complex and expensive fabrication process but also enhances the biological friendliness of integrated bioMEMS. In this section, we summarize the key ideas that this review aims to convey and disclose some possible directions for future studies.

First, by creating the localized pH gradient, the freestanding, robust, well-aligned, and semipermeable CM can be readily formed at the interface of the converging flows of the chitosan solution and a countering basic buffer or alginate solution in a spatiotemporally controlled manner. While the use of a basic buffer solution requires a stable flow interface achieved through precise pressure-balancing, the other tactic enables the facile formation of the CM on PECM generated upon the spontaneous contact between chitosan and alginate macromolecules. The addition of the PECM to the flow-assembly of the CM significantly enhances the capability to immobilize a wide variety of biomolecules or biological components thanks to the presence of both amine and carboxyl functional groups. Furthermore, the PECM aids in stabilizing and separating the subsequently biofabricated synthetic ecosystems of multispecies entrapped in alginate hydrogels, providing closely resembled microenvironments for cell–cell communication/signaling studies.^40,42 Besides the basic buffer and alginate solutions, crosslinking agents such as glutaraldehyde,^29,39 terephthalaldehyde,^118,119 and tripoly-phosphate^120,121 can be explored in the flow assembly of CMs in microfluidics.

To enhance the reliability of the fabrication process, our group developed two improvement strategies: one is to include an extra downstream acidic input in the device design to rinse any undesired residue deposition,³⁶ and the other is to use an add-on vacuum layer that can dissipate air bubbles trapped in apertures through the gas-permeable PDMS layer.⁴¹ Meanwhile, other groups designed microchips with circular pillars and precise pumping to skip the formation of air bubbles,⁴⁴ or included an extra outlet to act as an anchoring point for the membranes.⁵³ On the other hand, the solubility in slightly acidic solutions and low-molecular weight cut-off of the flow-assembled CM (only a few nanometers) might limit the applications of their integrated microfluidic platforms. To overcome these problems, several works have been conducted to modify the properties of the CM for broader applications.^29,39 Additionally, many bioactive materials including PNIPAM nanogels,³⁰ carbon nanoparticles,⁷ or collagen³¹ have been successfully immobilized in microfluidics using chitosan as an embedded substrate to enhance the functionality and applicability of the fabricated integrated microfluidic platforms.

Since the first invention in 2010, the flow-assembly of CM has gained more and more attention from scientific communities as a promising platform technology. The flow-assembled CM-integrated microfluidic platforms have been widely exploited in a variety of important biochemical and biological applications, which include static gradient generators,³⁵ platforms for shear-free cell culturing,^31,53 and synthetic ecosystems for cell–cell communication studies.^36,40,74 Additionally, multi-layered CMs have recently been developed to investigate the complex release profiles of mesoporous silica nanoparticles for personalized medical applications.³⁷ Last but not least, the horizontal layout of the CM- and CBM-integrated microfluidic platforms not only allows direct visualization of cellular interactions and high-resolution live imaging but also enables a simpler quantification process as compared to the existing sandwiched platforms.^35,40,122

The current review reports the recent progress in the flow-assembly of CMs in microfluidics and the implementation of this promising platform technology that provide insights into and open many opportunities for future research and applications. First, future studies can focus on developing feasible characterization approaches since the physicochemical properties of CMs and CBMs remain difficult to be investigated due to the tiny size of these membranes and intrinsic enclosure nature of microdevices. Second, more works can be carried out to immobilize biomolecules and biological entities to chitosan and/or alginate backbones, if a PECM is present, for broader applications. For example, the biofabrication of three-dimensional hydrogel microenvironments with embedded cells using the semipermeable CM as an architecture provides a unique assembly strategy with spatiotemporal programmability and opens the door for future cell–cell signaling studies of multiple cell populations or species in synthetic microbiomes. Finally, it is highly desired that the biofabrication of freestanding CMs by flows can be scaled up similar to the interfacial electrofabrication of the CM using distal electrodes.⁵¹ The capability to fabricate freestanding CMs in three-dimensional geometry will increase the surface area of CMs for enhanced loading of biomolecules or biological components for more diverse applications.²⁴

Biographies

Khanh L. Ly

Khanh L. Ly received her BSE in Biomedical Engineering from the International University, Vietnam National University in Ho Chi Minh City, Vietnam, in 2018. She received her MS in Biomedical Engineering from the Catholic University of America in 2020. She is currently pursuing her PhD in Biomedical Engineering while working at the Integrated BioMicroFluidics Lab at the Catholic University of America. Her research is focused on microfabrication that uses biological or biomimetic materials and processes for construction of biopolymer membranes and tissue-mimicking models in microfluidics.

Piao Hu

Piao Hu is a PhD candidate in the Mechanical Engineering Department of the Catholic University of America. She has a BS degree in Materials Science and Engineering from Jilin University (2016) and a MS degree in Mechanical Engineering from the Catholic University of America (2018). Her research involves micro-electro-mechanical systems with biopolymer materials, and electrophysiological characterization of membrane proteins.

Le Hoang Phu Pham

Le Hoang Phu Pham is a PhD candidate from the Catholic University of America. He is working as a research assistant in the Integrated BioMicroFluidics Lab directed by Dr Xiaolong Luo. His expertise is in micro-biofabrication, and microfluidic device design and control. His interest is in developing microfluidic platforms based on the utilization of freestanding biopolymer chitosan membranes for biological studies of intercommunication between microbes.

Xiaolong Luo

Dr Xiaolong Luo is an associate professor in the Department of Mechanical Engineering at the Catholic University of America (CUA). His main research interests are biofabrication in microfluidics for biomedical research and practical applications. He was a recipient of the NSF CAREER award in 2016, and the Provost Young Faculty Scholar’s Award at CUA in 2017. Before joining CUA in 2013, he received his BE degree in Mechatronics from Zhejiang University, MS degree in Mechanical Engineering from Temple University, and PhD degree in Bioengineering from the University of Maryland, and he was a research associate at the University of Maryland.