1. Introduction

Micro- and nanoparticles are recognized as versatile research tools in the fields of cell studies [1,2], drug delivery [3–5], self-assembly [6,7], barcoding [8], encoding [9,10], and anti-counterfeiting [11]. Particle-based cell culture and manipulation platforms are particularly promising as they could enable breakthrough applications in single-cell analysis and selection. For example, self-steering microparticles can act as carriers for cells [1] in continuous flow cytometers or imaging flow cytometers, improving the ability to analyze morphological features in potential cell or microtissue/organoid-based therapies in a high-throughput manner. Such microparticle carriers containing cells of interest can be recovered following rapid analysis and machine intelligence directed sorting [12], obviating current challenges in selecting cells from surfaces in standard microscopy systems. To develop cell carriers to be effective in cell culture and high-throughput sorting, it is essential to be able to control the size and shape of the microparticles, as well as their chemical composition, to improve cell adherence to specific regions of interest on each microcarrier. The size of a microcarrier and nature of its cell-adherent surface dictates whether the carrier will be useful in single-cell studies. Prior work in creating microcarriers for high-throughput cell analysis used microcarriers many times larger than the target cells (∼400 µm compared to ∼10 µm cell size) [1], which would be limited to multi-cell aggregates or cell cluster studies. In addition, the geometry of cell carriers can be utilized to protect adhered cells while tuning the carrier’s hydrodynamic behavior. Microcarriers that passively self-align in the flow could eliminate the need for active measures (e.g., magnetic fields) or complex fluidic components (multiple confining 3D co-flows) to orient the carrier, which helps ensure uniform cell velocities for consistent imaging [1]. Reducing the need for outside mechanisms of control can lead to increased throughput, as analysis and sorting channels can be more easily parallelized, ultimately enabling the biomanufacturing of cells and microtissues at scales relevant for therapeutic applications. Finally, precise control over chemically functionalized regions of the microcarrier can be used to design the location to which cells adhere. This multi-material fabrication ability allows for patterning adherent cell shelters within the features of the printed particles to protect adhered cells from the environment or fluid shear stress as well as enable drugs to be delivered to targeted cells [13,14].

The goal of this work is to push the limits of flow lithography techniques to meet the aforementioned manufacturing needs of emerging cell-based microcarrier technologies. We begin with an overview of the state-of-the-art in flow lithography and a discussion of our scanning two-photon continuous flow lithography system, and show new developments for this system that enable the desired microscale precision and multi-material fabrication capabilities. We demonstrate these features by creating shaped 3D multi-material microparticles at the critical length-scale necessary for single-cell analysis and at a fabrication throughput that enables large-scale production.

1.1 Flow lithography

Extensive applications of custom microparticles have fueled the research and technology development of approaches required to fabricate them. One promising technique for fabricating such particles is microfluidic lithography that uses liquid prepolymer resin, which flows in a microchannel in conjunction with a light source to induce photopolymerization of the microparticles. Spherical [15] or slightly deformed spherical (i.e., disk or rod) [16] microparticles were the most common geometries that could be achieved in early forms of this technique because they could be fabricated using droplet microfluidics. In this method, droplets of photosensitive materials are generated by suspension in an immiscible fluid in a microchannel. These droplets are then exposed to light with a wavelength that initiates the photopolymerization process required to solidify them [17]. Using the same approach, slightly deformed spherical microparticles could also be achieved by varying the design of the microchannels. For example, disc-shaped microparticles have been achieved by flattening the droplets in wide and short channels before curing them [16].

To further diversify the shapes of microparticles, conventional lithography techniques were coupled with microfluidics-based fabrication, and continuous flow lithography (CFL) was introduced [18]. This method used a microfluidic channel with a steady flow of photocurable resin and a UV source that is exposed to certain regions of the microchannel through an optical mask, which produces solid microparticles that are extrusions of the mask patterns. However, the printing resolution of this technique was low, especially at the edges of microparticles, as the flow caused smearing. This smearing occurred because the duration of polymerization was accompanied by significant motion of the prepolymer solution. To overcome this issue, low flow rates were inevitably used, which sacrificed system throughput. To achieve better shape accuracy and higher throughput, a follow-up method called stop-flow lithography (SFL) was introduced [19] with the same procedure, except that the flow was stopped during the exposure step. After polymerization, the flow resumed to wash the microparticles and flow in fresh resin to the fabrication site. This technique also yields extruded versions of 2D patterns on the mask but has a higher resolution and throughput, which depends on the size of the particle and the channel. A variation of SFL uses opaque magnetic microparticles in the resin to induce a gradient in the ultraviolet (UV) light at the exposed areas, thus enabling nonlinear curing along the axial direction [20]. Building off of CFL and SFL, lock-release lithography [21] was introduced in which positive features on the microchannel were used as molds to add constraints to the geometry of the particles made by projection lithography. Particles were subsequently released by pressure-induced channel deformation. This method allows for multiple medium exchanges or multi-material patterning of the particles as they can be fixed in place during the process.

Despite the high throughput of these systems, each has limitations in terms of geometry and shape of the particles that can be fabricated. In addition, the height of the microparticle depends on the channel geometry as the polymerization process is not localized axially, which causes the light to cure the whole column of exposed photopolymer. The improved vertical resolution was achieved by implementing optofluidic maskless lithography that uses a digital micro-mirror device (DMD) to project patterns on a membrane-mounted channel, with the channel height being controlled by applying pressure on the membrane [22]. This method can be used in a layer by layer fashion to achieve 3D microparticles. By changing the medium between steps, multi-material particles are also made possible. One disadvantage of this approach, however, is that it requires complex microfluidics. A similar layer by layer approach, called vertical flow lithography [23], uses a microfluidic device with four symmetric horizontal inlets and one vertical outlet, allowing for more fabrication flexibility in terms of material distribution within each layer. Thus, tapered shapes can be achieved within layers using this approach. This technique has a throughput of 4 particles per minute, which is low compared to previously discussed state-of-the-art SFL. Recently, the same approach was used to fabricate micro-tubes [24] in vertical channels using two-photon lithography, which dramatically enhanced the resolution.

Another technique that deviates from layer by layer fabrication but still uses projection lithography is called optical transient liquid molding [25], in which photopolymer resin streams are shaped by software-aided inertial flow engineering [26] to fabricate microparticles with concurrent UV exposure through a mask. In this technique, a fully-3D microparticle shape is formed by the union of two extruded 2D shapes: one shape from the optical mask and the other from the sculpted flow. However, this method is limited in terms of particle size (≥ 100 µm) and complexity, as one of the extruded 2D shapes is restricted to what is possible through inertial flow sculpting [26].

All of the approaches discussed above are based on projection lithography and one-photon polymerization, which cannot achieve submicron resolution. Two-photon polymerization can address this issue in that it localizes the polymerization to small portions of the laser’s focal point that are above the threshold for nonlinear two-photon absorption, resulting in printing resolution of less than 100 nm. Usage of two-photon polymerization in microfluidic lithography was first demonstrated by Lasa et al. [27]. This group fabricated bi-material micro springs in a stream of two co-flowing photopolymers by moving a piezoelectric stage in a circular motion as the polymers flowed past a fixed laser spot. Despite the high resolution of the microparticles created by this method and its ability to create multi-material particles, the method is limited in terms of i) throughput, which is estimated to be 0.2 particles per minute and is governed primarily by the speed of the piezoelectric stage [23], ii) particle size, which is governed by the speed of the mechanical laser shutter, and iii) particle geometry complexity, which is governed by the ability to compensate for the flow velocity. These issues, however, were addressed in our previously published work, where we introduced the scanning two-photon continuous flow lithography (STP-CFL) technique [28].

1.2 Scanning two-photon continuous flow lithography

In STP-CFL (enabled by the system shown in Fig. 1), galvanometer mirrors are used to rapidly scan a femtosecond pulsed laser beam in a continuous flow of resin such that the laser follows the microparticle and compensates for its continual movement while it is being printed. To perform this flow compensation, first, the flow velocity is determined by taking successive images of a flowing microparticle at a known time difference. It is worth noting that due to the small size of fabricated microparticles, a uniform velocity profile was assumed across each microparticle. Next, the laser scan path is updated by shifting each scan point downstream, considering its position in the path, the scan rate, and the flow velocity. This technique also allows for the layer-by-layer fabrication of arbitrary 3D shaped microparticles by synchronized movement of the galvanometers for XY plane scanning and piezoelectric stage for Z-axis scanning of the laser focal point. We utilized an acousto-optic modulator to rapidly attenuate or turn off the laser beam entirely. We demonstrated the fabrication of single material particles of approximately 10 µm in size, at the rate of 31 two-dimensional particles per second. This throughput is comparable to the rates reported for projection lithography techniques such as SFL; however, STP-CFL results in an order of magnitude higher resolution. The 3D particles were fabricated at the rate of 15 microparticles per second, which is currently unrivaled by other two-photon lithography-based techniques.

 

Fig. 1. STP-CFL system schematic. The system uses a high-speed camera (HS Cam), lenses (L), a beam splitter (BS), a tube lens (TL), a dichroic filter (DF), a dichroic mirror (DM), a 4-F telescope (4-F), galvanometer mirrors (Galvo), a beam expander (BE), a beam block (BB), an acousto-optic modulator (AOM), a femtosecond pulsed laser (fs), a mirror (M), a microscope objective (MO), a microfluidic device (MF), and a light emitting diode (LED).

Download Full Size | PDF

In this work, we advance the capabilities of the STP-CFL technique to realize multi-functional microparticles by printing at the interface of two or more co-flowing streams of prepolymer resins that contain different components. Furthermore, we introduce a new printing approach to achieve out-of-plane features of these microparticles without axial scanning of the laser spot. We demonstrate this technique by fabricating bi-functional microparticles with the geometry of self-aligning single-cell carriers. The asymmetric dumbbell shape of this design (Fig. 2) allows for their self-alignment in Stokes flows [29,30]. Moreover, the 3D architecture these microparticles encompass is designed to have a sheltered area to protect cells from the shear stress experienced in downstream analysis workflows, such as microfluidic-based flow cytometry platforms. To ensure the cells only adhere to the shelter area, a bi-functional structure is needed in which cell-adhesive material is exclusively used in the shelter region. STP-CFL is well-suited to fabricate such cell carriers due to its ability to fabricate custom 3D microparticles at the scale of a single cell. In addition, by including co-flows of different polymer precursors in the fabrication channel, multi-functional microparticles are made possible.

 

Fig. 2. Schematic of microfluidic channel used for bi-functional STP-CFL (left) and design of cell carrier with distinct shelter area (right). Lengths are not to scale.

Download Full Size | PDF

2. Materials

The liquid prepolymer solution consists of 99% w/w poly (ethylene glycol) diacrylate (PEGDA Mn∼575; Sigma-Aldrich) as the monomer, and 1% w/w Irgacure 369 (BASF) as the photoinitiator. To provide a proper visual demonstration of the bi-functional microparticle fabrication capability of STP-CFL, blue fluorescent microspheres of size 0.5 µm (Fluoro-Max, ThermoFisher) were included in the precursor solution of the middle flow to produce the shelter region.

3. Methods

3.1 Fabrication

To create the three co-flows required for STP-CFL of the cell carrier microparticles (Fig. 2), three-inlet microchannels (width=1200 µm, depth=130 µm) were created using soft lithography methods. SU-8 2100 negative photoresist was used to fabricate the master mold for the microchannels on silicon wafers (UniversityWafer, Inc.), using a standard photolithography process. Polydimethylsiloxane (PDMS) elastomer and the curing agent (Sylgard 184, Dow Corning) were mixed at a 10:1 ratio by weight, poured onto the mold, degassed, and then cured at 65°C overnight. After cutting and peeling the device, 1.5 mm holes were punched at the inlets and outlet. The PDMS and No. 1 thickness cover glass (Electron Microscopy Sciences) were air plasma treated (Plasma Cleaner, Harrick Plasma) and bonded to create an enclosed microchannel. Tygon tubing (OD = 0.06”) was used to connect the microchannel inlets to 1 mL plastic syringes (BD) that were controlled using a syringe pump (Chemyx Fusion 100) to stabilize the location and width of the co-flows, a Y-connector (IDEX Health and Science) was used to split a single flow from the syringe pump driven at 7 µL/min to shape the two outer flows, and the middle flow was driven at 0.05 µL/min to shape a narrow stream of approximately 10 µm wide (Fig. 3).

 

Fig. 3. Optical image of the high-speed camera depicting STP-CFL of cell carrier microparticles in three laminar co-flows (see Visualization 1).

Download Full Size | PDF

To precisely monitor the fabrication process at high flow rates in the channel, our previously used STP-CFL system [28] was upgraded with a Phantom VEO-440 highspeed camera to produce the work of this paper. We also updated the flow velocity measurement method that runs before each round of microparticle printing: the prepolymer medium is exposed to the static laser spot for 50 ms, which results in a solidified line of photopolymer in the channel (Fig. 4). By measuring the length of this line, the flow velocity is determined (ranging between 500–1300 µm/s based on the flow rate), and the code automatically updates the scanning data to compensate for the movement of microparticles in the channel while they are being printed.

 

Fig. 4. Line printed by exposing the flowing medium to a static laser beam to measure flow velocity.

Download Full Size | PDF

Since the desired cell carriers are about five times larger than the microparticles in our original STP-CFL work, each particle’s scan data contains considerably more scanning points. As a result, each particle travels a longer distance during polymerization, and at the same time, may be exposed to larger velocity gradients in the channel due to the presence of Poiseuille flow conditions. It was observed that the velocity gradient could exert moments on the microparticles in a way that the layers would misalign during the fabrication (Fig. 5). To resolve this issue, we utilized wider (1200 µm versus 200 µm) and deeper (130 µm versus 50 µm) channels and printed at the middle planes of each dimension to achieve a minimal velocity gradient across the particle.

 

Fig. 5. Brightfield image of a poorly fabricated microparticle resulting from the velocity gradient induced movement of the microparticle in the channel during fabrication.

Download Full Size | PDF

To print the out-of-plane features of the shelter region with high resolution, we present an alternative approach to the axial scanning of the stage. This new method relies on the fact that the out-of-plane thickness of a printed voxel in two-photon polymerization is correlated with the exposure time of the laser beam at any given location [31]. Therefore, in the absence of a z-scanning stage, out-of-plane features can be created by tuning the exposure time (i.e., point density) in different regions. To utilize this method, we first slice the cell carrier geometry into three layers stacked in the z-direction (Fig. 2). Then we use STP-CFL to scan the points in each layer without moving the laser spot out of the plane. Consequently, the thickness of different regions grows in proportion to the laser exposure time at each point. Eliminating the Z-stage scanning is beneficial as it could introduce flow perturbations and limits overall throughput due to the lower scan rate of piezoelectric stages when compared to scanning mirror galvanometers. This new technique also allows for continuous variation of thickness in applications where rough surfaces created by stage stepping are not favorable [31]. It is worth noting that microparticles that can be fabricated with the presented approach are axially symmetric and cannot be hollow. Particles deviating from these rules can be fabricated with our original STP-CFL approach published before [28].

3.2 Microparticle recovery

Immediately after microparticle fabrication, the contents of the collection tube (microparticles along with precursor solution) were passed through a 10 µm cell strainer (CellTrics), followed by a 0.1% (w/v) Pluronic F-127 in phosphate-buffered saline solution (PBSP) in order to remove excess microspheres for imaging purposes. Next, the cell strainer was flipped, and microparticles were recovered by flushing with more PBSP. PBSP was also used to pre-coat all tubes, pipette tips, and the strainer to reduce microparticle loss through adsorption to the plastic surfaces.

3.3 Confocal and fluorescence microscopy

Unreacted acrylate groups of cell carrier microparticles were modified with biotin groups after fabrication by reacting microparticles in PBSP with 0.44–0.55 mg/ml of Acrylate-PEG-Biotin (MW = 5000, Nanocs) with a final concentration of 0.072–0.091% lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) photoinitiator. The Acrylate-PEG-Biotin reaction vial was exposed to UV light (15.5 mW/cm², Omnicure S2000) for one minute while being stirred. Microparticles were then washed three times, incubated with Streptavidin conjugated with Alexa Fluor 568 (Invitrogen) for 15 minutes to allow the orthogonal binding of biotin to Alexa Fluor 568 Streptavidin, and then washed again. This post-fabrication modification allows the entire microparticle, in addition to the blue microsphere embedded shelter region, to be imaged using confocal microscopy.

Confocal images of cell carrier microparticles in coverglass wells were taken on a Leica SP8 confocal microscope using a 40x oil immersion objective. The 405 nm laser was used to image the blue fluorescent microbeads embedded in the shelter region of microparticles. The 552 nm laser was used to image the entire Alexa Fluor 568 Streptavidin-coated microparticle body. Z stacks with a step size of 0.346 or 0.5 µm were taken, and the 3D reconstruction of each microparticle was clipped along the major axis to reveal the cross-section. Heights of the shelter region and adjacent region were extracted manually on ImageJ.

Microparticles were also imaged using fluorescence microscopy to allow for higher throughput image acquisition for basic features such as microparticle length and bead localization. Free microparticles were imaged in PBSP on cover glass using an inverted microscope (Nikon, Eclipse Ti-S fluorescence microscope) with a 40x objective lens in brightfield and DAPI channel (Fig. 6). Alexa Fluor 568 Streptavidin modified microparticles were also imaged in TRITC channel, and the binary signal from this channel was used to measure microparticle major axis length using a custom MATLAB code.

 

Fig. 6. Fluorescent microscopy of cell carriers after post-processing. Images indicate the fluorescent bead pattern is aligned with the cell-shelter region. Measurements indicate ∼10% of increase in dimensions due to swelling of the hydrogel microparticles (detailed size distribution graph in Supplement 1, Fig. S1).

Download Full Size | PDF

4. Results and discussion

A 10x slowed down movie of the fabrication process is provided in Visualization 1, where cell carrier microparticles of size 50 µm (measured using optical microscopy) are printed with a throughput of 10 microparticles per second using a femtosecond pulsed laser with 20 mW of power at 690 nm. The beam is scanning the points of the scan path at 15 kHz using the galvanometer mirror system. We have demonstrated the maximum throughput of the system by printing the cell carrier microparticles in flows with velocities as high as 1300 µm/s, but we note that the maximum flow rate and throughput of the STP-CFL technique are now limited by the scan rate of the galvanometer mirrors. Consequently, the maximum rate of microparticle fabrication is directly proportional to the size and number of points in the scan path of each microparticle. By utilizing faster galvanometer mirrors, it would be possible to increase fabrication throughput for a given microparticle geometry. In addition, the characteristic microparticle dimension can conceivably be as large as the field of view, which is currently 200 µm by 180 µm in our system. However, the goal of this work was to perform multi-material particle fabrication at a length-scale similar in size to adherent cells, which we successfully achieved.

By exploiting the exposure time at different regions of the cell carrier microparticles during fabrication, we were able to produce microparticles with height variation along with a custom 2D outline, opening the possibility of rapid manufacturing of complex 3D shaped microparticles. In this case, we were able to create a cell shelter region with a length-scale matching many adherent cell types (∼10–15 um), with a sheltered depth of ∼1.5 µm based on cross-sectional data obtained via confocal microscopy (Fig. 7). Compared to the height of the region adjacent to the cell shelter, the shelter region was measured to be 18–24% shorter depending on the power density used by the laser during fabrication. This height difference can be further tuned by changing the point density in the sliced 3D microparticle geometry at each region based on the desired design. This kind of height tunability is a powerful and unique feature of the new STP-CFL approach presented in this work, significantly improving the fidelity in replicating subtle microparticle geometries. We envision that this could be utilized for designing modular microcarrier “parts” that protect different cell types and morphologies, or even purposely expose the cells to as-designed fluid shear stress.

 

Fig. 7. Confocal microscopy of cell carriers. Top view of the cell carriers (left) and the 3D render of the microparticle (right)

Download Full Size | PDF

In addition to more complex geometry at sub-100 µm length-scales, we demonstrate the ability to create multi-functional microparticles through the introduction of precursor coflows using multiple inlets that lead to a single straight fabrication channel. We illustrated this using fluorescent microbeads in the middle stream to define the shelter region (see Figs. 2, 3, and 6). Small molecules can also be added in the future to allow for a cell-adhesive composition that is contained within the shelter region, provided that (i) mass diffusion of the added molecules does not significantly alter the designed distribution before the particles have polymerized, and (ii) the cross-linked matrix of the hydrogel particle effectively retains the dopant (e.g., through steric interaction, covalent bonds, or other means). We note that the flowing streams should have similar material properties (i.e., density and viscosity) to achieve reliable steady-state behavior both in flow and after the particles are fabricated, minimizing the effects of fluid-fluid and fluid-structure interaction on the two-photon polymerization. Current multi-material formulations for bio-compatible microcarriers predominantly use similar PEG-based hydrogels, so we do not anticipate this being an issue for our intended application. However, the range of usable material differences (e.g., two immiscible polymers) should be explored in future work.

5. Conclusion

We have successfully demonstrated the fabrication of high-resolution multi-functional 3D cell carrier microparticles by bringing new capabilities to STP-CFL. Out-of-plane features can now be created with good fidelity without sacrificing the speed of fabrication. The future steps of this work include studying the self-alignment behavior of the fabricated microparticles, implementing cell adhesive material in the shelter region, loading cells on the carriers, and performing flow cytometry for downstream cell studies. More generally, we envision the STP-CFL system as an approach to rapidly prototype and analyze complex multi-material particles not only for bioengineering applications but also for use in studying novel particle-particle interactions in additive manufacturing and self-assembly.