Lithographically patterned micro-nozzles for controlling fluid flow profiles for drug delivery and in vitro imaging applications

Abstract

Precisely controlling delivery of drugs and other reagents is important for intravital microscopy studies. In this work, photolithographic integration of micro-nozzles onto a microfluidic platform was performed to tune the fluid flow profile and depth of penetration into biological tissue mimics. Performance characteristics were measured by correlating the flow rate through the device to the applied pressure and/or delivery of dyes into solution and agarose gel-based phantom tissue. From these results, the implementation of micro-nozzles was demonstrated to significantly improve the lateral dispersion of delivered fluid and increase the depth of penetration into phantom tissue.

Introduction

Direct observation and manipulation of cells and tissues is a valuable component of biological and medical research. However, the limitations of current approaches have become increasingly evident as the influence of the cellular microenvironment continues to grow.^[1–4] Conventional 2D and 3D in vitro assays have been shown to poorly reflect the topography, diversity, and heterogeneity present in vivo.^[5–7] Conversely, despite the intrinsic advantages of in vivo techniques, experimental approaches yielding cellular resolution (histology, fluorescence activated cell sorting, etc.) often use end-stage assays that require correlation across a population.^[8]

Intravital imaging is a non-destructive, optical sectioning technique that is capable of serial imaging live animals with single cell resolution, using a large field of view and depth of focus.^[9] This technique provides key information about tissue structure and the potential to understand sub-cellular and molecular events triggered by drugs and other diffusible, molecular effectors. Using this approach, researchers have systemically introduced therapeutics to observe the chain of events that occur upon the arrival of drugs or effectors at a target site.^{[10, 11]}

The configuration of intravital imaging, particularly the installation of an externally interfaceable device proximal to the target cells or tissue, also presents an opportunity for the direct study of drug delivery in the interstitial compartment. This is advantageous for the assessment of local therapeutic thresholds while minimizing systemic off-target effects and reducing experimental drug volumes.^[12] Additionally, interstitial barriers and interactions with novel drug formulations could be probed independently of current systemic drug delivery technology. This is particularly relevant in the emerging field of nanomedicine, where transport properties vary significantly from conventional therapeutics.^{[13, 14]} Crucially, to achieve this degree of experimental control in vivo, there is a need for high-resolution fluid delivery systems that are compatible with intravital imaging.

To demonstrate the potential for high-resolution drug delivery studies in vivo, we integrated pressure-driven microfluidic technology with a custom-fabricated intravital imaging window. Here, microfluidics present several advantages to fluid delivery including reduced reagent usage, high spatial resolution, and well-controlled laminar flow.^{[15, 16]} Convective transport is used for its greater control over distribution profiles than those obtained from passive transport methods.^{[17, 18]} A similar approach, convection enhanced delivery (CED), has been used by other groups to improve drug delivery in the brain, but has yet to see widespread use in other locations.^{[18, 19]} A critical feature of CED is the direct insertion of a catheter into the target tissue to conduct fluid into the tissue, rather than across the surface, where an intravital window is located. This serves to improve experimental control by preventing the topological flow of delivered fluids away from the target region. Thus, to minimize the influence of topography at the tissue interface with the microfluidic intravital window, a microfluidic conduit into the tissue is needed.^[12]

In this study, we used a microfluidic device suited for simultaneous controlled drug delivery and observation to assess the delivery characteristics of novel micro-nozzle structures. Here we report on a novel fabrication and assembly approach for integrating micro-nozzles with glass microfluidic structures using a combination of dry-film photoresist and greyscale photolithography to improve the drug/reagent delivery profile. To achieve fluid conduction into tissue, we explored the use of tapered microstructures (micro-nozzles) to offset the opening of the device away from the device-tissue interface. Micro-nozzles were fabricated using dry-film photoresist technology, which allows for material preprocessing prior to integration. Validation of delivery was performed in agarose tissue phantom hydrogels to allow for simple monitoring of convective flow and diffusion in a substrate with tunable mechanical properties.^{[17, 20]} In particular, we show that micro-nozzles improve the delivery of fluids into hydrogels by reducing lateral dispersion and increasing the depth of penetration into hydrogels by fluid focusing.

Materials and methods

Materials

Laser-cut glass coverslips (borosilicate #1.5) were custom fabricated by Potomac Photronics. Alconox® detergent, acetone, and isopropyl alcohol were ordered from Sigma-Aldrich. The surface crosslinking agent, 2-(3,4-epoxycyclohexyl) ethyltrimethoxysilane was ordered from Gelest, Inc. Dry-film photoresist sheets of SUEX® and SU-8 Developer were purchased from DJ Microlaminates and Kayaku Advanced Materials, respectively. Polyethylene terephthalate glycol (PETG) shim stock was purchased from Cole Parmer (9513K24) and Membrane Switch Spacer 7959MP (dry adhesive) from 3 M. Metal tubing (30 gauge) was obtained from Ziggy’s Tubes and Wires (30R304–36), and cyanoacrylate from The Original Super Glue Corporation.

Microfluidic circuits were built from 1 to 16″ O.D. PEEK tubing with varying inner diameter. Circuits were driven by either syringe pump (Pump 33DDS, Harvard Apparatus) or pressure controller with inline FS3 Flow Sensors (OB1 MKII, Elveflow). Various fittings and adapters were purchased from IDEX Health & Science.

Agarose tablets were purchased from Bioline and trypan blue solution was obtained from Sigma-Aldrich. Fluorescein isothiocyanate-dextran (FITC) and tetramethylrhodamine (TRITC) with molecular weights of 20 kDa and 155 kDa, respectively, were obtained from Sigma-Aldrich.

Device fabrication

Microfluidic intravital imaging windows were fabricated with planar fluid outlets or outlets with integrated micro-nozzles Fig. 1(a–c).

Figure 1.

(a) Exploded and (b) cross-sectional views of a (c) microfluidic device used to compare outlet properties, SEM images of (d) planar and (e) micro-nozzle outlets.

Briefly, microfluidic devices with a 200 μm circular outlet were fabricated by patterning 20-μm-thick dry film SU-8 photoresist (DJ Microlaminates, SUEX) onto laser-cut glass coverslips (Potomac Photonics) and connected to an inlet hub composed of 30-gauge metal tubing (Ziggy’s Tubes and Wires, 30R304–36), dry adhesive (3 M, 7959MP), cyanoacrylate (The Original Super Glue Corporation), and PETG shim stock (Cole Parmer, 9513K24). Glass coverslips had a 150 μm via and a laser cut microchannel that was 50 μm deep, 5 mm long, and varied from 150 μm wide at the via to 200 μm wide at the center of the coverslip. The dry adhesive and the PETG stock used for the inlet hub were cut into their final shapes using a CO₂ laser cutter (Dremel, LC40). Micro-nozzles, fabricated using a post-exposure lamination step with 200-μm-thick SUEX to achieve sloped sidewalls, were integrated prior to attachment of the inlet hub via multilayer alignment and bonding in a Sky 325R6 laminator, as shown in Fig. 1(d) and (e). Sloped sidewalls were obtained by proximal exposure of the dry-film photoresist through a diffractive zone plate pattern with a 500 μm air gap using 365 nm light. Alignment on the microfluidic device was performed prior to the post-exposure bake step to allow for crosslinking to the surface. A Hitachi S-4800 scanning electron microscope (SEM) operated at 2 kV was used to assess micro-nozzle geometry. The thickness of the SUEX photoresist controls the final micro-nozzle height and 200 μm was selected to be compatible with the depth of imaging available in intravital imaging setups¹.^[21] The radial component of the micro-nozzles is constrained primarily by the desired lumen diameter and mask resolution (5 μm). In this study, a 140 μm diameter lumen was used to demonstrate a system that is compatible with the delivery of large biomolecules and cells. A comprehensive process flow is outlined in Fig. S1 [Online Resource 1].

Experimental setup

Once devices were complete, fluid reservoirs, consisting of either sealed polystyrene tubes or preloaded syringes, were connected using various fittings and tubing segments. A programmable pressure controller (Elveflow OB1 MK2) with inline flow sensors (Elveflow MFS3) was used to deliver fluid from prefilled 50 mL Falcon tubes and 1.5 mL centrifuge tubes (Eppendorf) for calculating the fluidic resistance of devices. Alternatively, syringes were fit into a dual-channel programmable syringe pump (Harvard Apparatus, 33DDS). Microfluidic circuits are depicted in Fig. S2 [Online Resource 1].

Lateral fluid dispersion testing

Fluid conduction was assessed in a solid tissue phantom composed of an agarose hydrogel (1% w/v). Agarose is commonly used to mimic the porous media properties of tissues due to its tunable permeability.^{[17, 20]} Agarose tablets were dissolved in deionized water by heating in a microwave at 30 s intervals. The solution was then poured into 90 mm square polystyrene petri dishes (Simport, 229-D210–16) and allowed to gel at room temperature. For extended imaging sessions, a thin coating of ultrapure water was applied to the hydrogel surface to prevent dehydration-associated surface roughening. Excess fluid was removed before device placement and testing. Microfluidic devices were primed with trypan blue solution and placed onto the agarose hydrogel prior to imaging. Delivery of 2 μL volume of trypan blue was performed at 100 nL/min. The dye formed an approximately circular area at the outlet of the device with no interfacial flow. Light transmission data were collected using brightfield microscopy (Nikon SMZ800 and Andor Neo 5.5 sCMOS camera). A timelapse was recorded at 60 s interval for 71 min.

3D fluid delivery testing

Next, 3D timelapse data in free solution was collected using confocal microscopy (Leica, TCS SP5) during the injection of 20 kDa FITC-dextran dissolved in ultrapure water (0.6 mg/mL). Flow rates were established by programming the syringe pump for 100 nL/min or 1 μL/min with a target volume of 100 nL. For imaging of fluid delivery into agarose gel-based phantom tissues, 155 kDa TRITC-dextran was infused into 1% agarose hydrogels that were prepared with 20 kDa FITC-dextran. Imaging was performed in 6-well and 12-well tissue culture plates (Corning, Costar 3506). XZ-stacks were collected to resolve the depth of dye penetration with minimal diffusion between images, and z-stacks were collected to observe a larger field of view.

Image analysis

Brightfield image data were collected using NIS-Elements software (Nikon) and on the confocal as LIF files using Leica Application Suite Advanced Fluorescence software (Leica Microsystems). The Nikon microscope was used to record the change in grayscale intensity due to trypan blue delivery and the Leica confocal microscope was used to measure the fluorescence of FITC-dextran in the green channel and TRITC-dextran in the red channel. Background normalization for brightfield images was performed in NIS-Elements and the intensity profile across the device window diameter was exported.

Post-processing intensity analysis was done with Excel (Microsoft) and FIJI.^[22–24] For lateral dispersion testing, the decrease in recorded intensity was interpreted as the resulting absorbance of the dye in the gel. Relative absorbance was calculated as a function of lateral distance from the outlet by averaging the planar outlet data across all samples and subtracting the measured intensity from the average maximum intensity. Absorbance was then normalized with respect to the outlet, which was observed to have the highest absorbance due to the presence of undiluted dye in the microchannel.

For 3D fluid delivery data, LIF files were imported into FIJI and axial line profiles were plotted along the outlet center, into the media for XZ images. Z-stacks were visualized in orthogonal planes through the outlet to obtain axial fluorescence data. To compare fluid transport from planar outlets vs. micro-nozzles, the intensity data were analyzed at the z-height closest to the outlet opening.

Results and discussion

Device characterization

To understand the effect of micro-nozzles on fluid delivery, two different types of microfluidic devices were fabricated. Control devices consisted of laser-cut microchannels in glass substrates, sealed by SU-8 films that were photolithographically patterned to yield a planar, circular fluid outlet. Micro-nozzle structures were integrated above planar fluid outlets using an additional greyscale photolithography process. Figure 1(d) shows planar outlets with a 200 μm circular outlet centered on the microfluidic device. Fabricated micro-nozzles, shown in Fig. 1(e), were positioned directly over planar outlets and were 200 μm tall, with an inner diameter of 130 μm at the tip.

The fluidic resistance of the micro-nozzles was determined by plotting flow rate as a function of the applied pressure as shown in Fig. S3 [Online Resource 1]. Data were collected using a microfluidic circuit without a device connected, with control devices, and with micro-nozzle integrated devices. Linear correlations were obtained for each condition, validating that the devices were operating in the laminar flow regime. Next, a linear regression was performed using the least squares method to obtain fluidic resistance for each condition. Micro-nozzles were found to contribute an additional 0.5 mBar min/μL of hydrodynamic resistance to the microfluidic device corresponding to a 57% increase in total device flow resistance.

Comparison of fluid delivery

Different fluid delivery profiles into agarose phantom tissue were observed between planar and micro-nozzle devices, Fig. 2(a, b). For micro-nozzle devices, dye was directly delivered into agarose hydrogel with low lateral dispersion. For the planar outlet devices, however, a rapid lateral dispersion of dye was observed at the interface between the agarose gel and the microfluidic device. We hypothesize that this is due to a localized buildup of pressure in the agarose near the outlet that eventually overcomes the resistance to flow along the interface. Similar behavior was never observed for devices with a micro-nozzle, Fig. 2(a, b). We suggest that micro-nozzle structures abrogate this effect by pressing against the agarose to block interfacial flow or by penetrating the hydrogel to better distribute the hydraulic pressure away from the interface. The lateral distribution of trypan blue dye concentration, shown in Fig. 2(c, d), was correlated to the measured change in optical density. Here, the lateral distance required for a 50% decrease in concentration was 68% smaller for devices with micro-nozzles. This indicates that micro-nozzles improved localized dye delivery by limiting the radial spread. For biological applications, this should significantly expand control of reagent concentrations in solid tissues.

Figure 2.

Timelapse images of trypan blue dye injected with (a) planar outlet control devices and (b) integrated micro-nozzles at 100 nL/min for 20 min. The relative absorbance (c) before and (d) after dye dispersion occurs is plotted as a function of lateral distance from the outlet center.

During fluid delivery experiments, we observed that the presence of a fluid layer on top of the agarose hydrogel prevented controlled delivery into the phantom tissue (data not shown). Thus, the micro-nozzle must make direct contact or penetrate into the hydrogel to yield the desired fluid flow distribution. For future studies, longer micro-nozzle structures could be implemented to avoid this issue (Fig. S4, Online Resource 1).

While micro-nozzle structures physically place the fluid output into direct contact with the agarose hydrogel, the height of the nozzle also affects the flow velocity (vs. the planar outlet) by introducing additional potential energy inside the micro-nozzle lumen. In free solution, this effect was observed to improve the depth of dye injection [Fig. 3(a), (b)]. The distance for FITC-dextran fluorescence to decrease by 50% in free solution with a micro-nozzle was found to be 58% further at 100 nL/min and 70% further at 1 μL/min than planar outlets [Fig. 3(c)].

Figure 3.

Cross-sectional views of the injection of 100 nL of 20 kDa FITC-dextran into free solution using (a) planar outlet control devices and (b) integrated micro-nozzles. The relative fluorescent intensity in (c) free solution and (d) 1% agarose hydrogels is plotted as a function of axial distance from the outlet center. Cross-sectional confocal scans of 155 kDa TRITC-dextran (red) injection into 20 kDa FITC-dextran (green)-labeled agarose hydrogels using (e) control devices and (f) micro-nozzle-integrated devices. Device surfaces (white dashed line) are emphasized for clarity.

Confocal imaging of TRITC dye delivery (red) into agarose hydrogels prepared with FITC-dextran (green) allowed visualization of fluid transport in the context of a phantom tissue. Timelapse data support the two methods of fluid delivery observed with brightfield microscopy. The distance for relative fluorescence of TRITC-dextran to fall by 50% in 1% agarose hydrogels with a flow rate of 1 μL/min was 134% larger with a micro-nozzle [Fig. 3(d)]. For control devices, TRITC dye was found to localize at the hydrogel surface, forming a rectangular cross section in XZ image plane [Fig. 3(e)]. This dispersion is consistent with superficial flow disrupting the conduction of dye into the tissue. Devices with micro-nozzles were capable of generating concentration profiles of fluorophore with less lateral dispersion [Fig. 3(f)]. While a quantitative analysis of dye concentration was not performed, the reduction in observable lateral dye dispersion at the micro-nozzle outlet supports our hypothesis that convective transport into the tissue mimic is improved. For follow-up studies, characterization of fluorophore luminescence intensity in agarose using serial dilutions could allow for quantification of local dye concentrations. Measurement of fluid pressure at constant flow rates in tissue mimics studies could also be used to characterize the poroelastic behavior of target systems.

Conclusion

In this study, we have shown that the integration of micro-nozzles with microfluidic intravital imaging windows improves the conduction of fluid into phantom tissue by reducing the dispersion of delivered fluids and increasing the depth of penetration. Lateral dye dispersion was reduced by 68% in 1% agarose gel phantom tissues and the depth of penetration for FITC-dextran in free solution was improved by > 50% for flow rates between 0.1 and 1 μL/min. Used in conjunction with intravital imaging, these advances could significantly improve the management of drug transport in the interstitial compartment to enable biochemical studies in vivo with cellular resolution.

Data availability

Data will be made available on a reasonable request.