Lab on a Chip

This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

View Article Online

PAPER View Journal | View Issue

High-resolution low-cost LCD 3D printing for

Cite this: Lab Chip, 2024, 24, 2774
microfluidics and organ-on-a-chip devices†
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Houda Shafique,ab Vahid Karamzadeh, ab Geunyong Kim,ab Molly L. Shen,ab

Yonatan Morocz,ab Ahmad Sohrabi-Kashaniab and David Juncker *ab

The fabrication of microfluidic devices has progressed from cleanroom manufacturing to replica molding
in polymers, and more recently to direct manufacturing by subtractive (e.g., laser machining) and additive
(e.g., 3D printing) techniques, notably digital light processing (DLP) photopolymerization. However, many
methods require technical expertise and DLP 3D printers remain expensive at a cost ∼15–30 K USD with
∼8 M pixels that are 25–40 μm in size. Here, we introduce (i) the use of low-cost (∼150–600 USD) liquid
crystal display (LCD) photopolymerization 3D printing with ∼8–58 M pixels that are 18–35 μm in size for
direct microfluidic device fabrication, and (ii) a poly(ethylene glycol) diacrylate-based ink developed for
LCD 3D printing (PLInk). We optimized PLInk for high resolution, fast 3D printing and biocompatibility while
considering the illumination inhomogeneity and low power density of LCD 3D printers. We made lateral
features as small as 75 μm, 22 μm-thick embedded membranes, and circular channels with a 110 μm
radius. We 3D printed microfluidic devices previously manufactured by other methods, including an
embedded 3D micromixer, a membrane microvalve, and an autonomous capillaric circuit (CC) deployed
for interferon-γ detection with excellent performance (limit of detection: 12 pg mL−1, CV: 6.8%). We made
Received 31st December 2023, PLInk-based organ-on-a-chip devices in 384-well plate format and produced 3420 individual devices
Accepted 12th April 2024
within an 8 h print run. We used the devices to co-culture two spheroids separated by a vascular barrier
over 5 days and observed endothelial sprouting, cellular reorganization, and migration. LCD 3D printing
DOI: 10.1039/d3lc01125a
together with tailored inks pave the way for democratizing access to high-resolution manufacturing of
[Link]/loc ready-to-use microfluidic and organ-on-a-chip devices by anyone, anywhere.

Introduction cleanroom as multiple replicates from a single

microfabricated mold could be made in a common research
Microfluidics, through miniaturization in micrometer-sized lab, and greatly accelerating the adoption and dissemination
vessels and microchannels, can reduce the fluid volumes of microfluidics for primarily research applications.3 More
required for analysis and synthesis to microliters and less, recently, direct manufacturing methods have been introduced
form the foundation for lab-on-a-chip devices, and are including subtractive ones such as laser ablation4 or
amenable to automation.1,2 However, wider adoption of micromilling,5 but offer limited relief due to drawbacks such
microfluidics and lab-on-a-chip devices in diagnostics, as the need for expensive equipment, technical expertise, or
synthesis, and research is slowed by complex fabrication provide limited resolution.
processes. Microfluidics emerged as a field thanks to Additive manufacturing, and in particular 3D
cleanroom microfabrication inherited from the stereolithography (SLA) printing characterized by layer-by-layer
semiconductor industry relying on photolithography and UV patterning and photopolymerization of successive layers in
using silicon or glass microfabrication methods which are a photocurable ink to build up a 3D printed object, has
dependent on capital cost intensive equipment. Soft received considerable attention thanks to its affordability,
lithography methods helped relieve the dependency on the high-resolution, and ease-of-use.6,7 A layer is exposed to a
digital pattern that solidifies the ink within a defined layer
thickness; the layer then rises to allow uncured ink to fill the
void, followed by digital photopolymerization of the new layer,
a
Biomedical Engineering Department, McGill University, Montreal, QC, Canada
b
and the process repeats iteratively. In an effort to clarify the
Victor Phillip Dahdaleh Institute of Genomic Medicine, McGill University,
Montreal, QC, Canada
terminology, we distinguish three methods of SLA and
† Electronic supplementary information (ESI) available. See DOI: [Link] strategies to selectively expose ink within the layer: (i) laser
10.1039/d3lc01125a scanning SLA operating with a galvanometer, (ii) digital light

2774 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

processing (DLP-SLA) that relies on a digital micromirror device been resolved for microfluidics which require small pixel
and an optical system for projecting a pattern, and most size, and hence small build areas, but come at the cost of
recently (iii) masked SLA using a liquid crystal display (LCD) limited manufacturing throughput.
3D printer where collimated light is directed through an LCD LCD photopolymerization 3D printers retail for as little as
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

screen that digitally renders the design and photopolymerizes ∼150–600 USD, with pixel numbers of 4 K (>8 M pixels), 8 K
ink atop the LCD. (>33 M pixels), and up to 12 K (>58 M pixels), and pixel size
Laser SLA gained popularity thanks to high-resolution of 18–50 μm, thus outperforming DLP 3D printers both in
prototyping on a large print bed (335 × 200 × 300 mm3) for terms of number of pixels and affordability. LCD 3D printers
microchannels ranging between 250–500 μm with 30–140 μm utilize an array of discrete light-emitting diodes (LEDs) that
laser spot sizes.8 Low-force SLA using a flexible vat reduces can now be mounted at high density (i.e., chip-on-board,
the adhesion force between formed layers and the bottom of COB) and that are collimated by an optical system (e.g., COB
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

the vat for intricate microfeature formation (e.g., separation lens and Fresnel lens) then pass through an LCD screen to
membranes).9 Additionally, many materials used in laser SLA reach the vat bottom. The number of pixels has been growing
are biocompatible,10 but have largely been limited to exponentially, and with a range of pixel sizes that extend to
commercial inks with proprietary formulations. Further, the smaller dimensions, thus offering both higher density and
single spot photopolymerization process with one or two larger print areas, and the capacity to print high resolution
lasers increases build times, especially for microfluidic structures such microfluidics on large print beds. However,
devices that are generally blocks of solid ink with few voids in a recent study, Caplins et al. report illumination non-
that constitute the channels. uniformity due to variable irradiance and spectral differences
DLP 3D printing became widely adopted for microfluidics in discrete LEDs resulting in inconsistent prints.20
thanks to rapid and high-resolution fabrication with reported Furthermore, the 50% transmittance loss of LCD screens by
microchannels as small as 18 × 20 μm2, 3D printer pixel sizes the crossed polarizers further reduces the irradiance of LCD
ranging from 2–40 μm, and an illumination wavelength 3D printers (2–3 mW cm−2) compared to their DLP
between 365–405 nm that can be used to photocure a wide counterparts (5–100 mW cm−2). Printing more voxels per time
range of materials.11–13 The availability of open-source requires higher irradiance as the rate of printing for a given
printers, online design repositories (e.g., Thingiverse, ink is limited by the power density of the light source.21
GrabCAD, Printables), tailored workflows (e.g., print–pause– Lastly, LCD screens degrade rapidly at low wavelengths and
print for multimaterial designs),14 and custom ink are thus limited to >400 nm illumination, which reduces
formulations further increase the potential. The development material selection and ink efficiency.12,20,22,23 Prior work has
of open-source inks such as those based on poly(ethylene shown success in leveraging LCD 3D printing for microfluidic
glycol) diacrylate (PEGDA) for DLP 3D printing benefit from master mold fabrication,24–27 but the potential for
known compositions, which could help evaluate the impact throughput manufacturing on large build plates and direct
of leachable and washable cytotoxic photosensitive LCD 3D printing of open and embedded microchannels has
components, and can be tailored and optimized for high- not been shown.
resolution embedded 3D printing, enhanced mechanical Here, we present high-resolution fabrication of embedded
properties, low viscosity for fast printing, as well as for low and open microfluidic devices using low-cost LCD 3D
protein adsorption and cytocompatibility.7,15,16 printing with a custom formulated low-viscosity PEGDA-
Altogether, high-precision 3D printers, custom inks, and based ink that cures using low irradiance and minimizes the
direct 3D printable designs enables digital manufacturing, i.e., effect of illumination variability on curing depth. The lateral
the seamless and automated fabrication from digital file to and vertical resolution of open and embedded structures are
final product with minimal post-processing. Digital characterized using a series of test structures, and showcases
manufacturing of microfluidic components has been possible high fidelity and dimensionally accurate printing down to a
early on, and now extended to the fabrication of fully resolution in the tens of micrometers. The biocompatibility
functional systems based on capillary flow.7 Indeed, as capillary of the ink is validated based on an ISO standard for cell
microfluidics can operate without peripherals,17 and complex toxicity. Three microfluidic devices are manufactured by LCD
fluidic algorithms could be structurally encoded into so-called 3D printing and characterized: (1) a microfluidic mixer
capillaric circuits (CCs),18,19 our group showed digital previously made by laser micromachining, (2) membrane
manufacturing of functional systems in the form of CCs. microvalves commonly made by replica molding, and (3) CCs
Thanks to custom intrinsically hydrophilic inks, ready-to-use previously made by DLP 3D printing. Finally, we demonstrate
CCs systems, could thus be printed using DLP 3D printers. LCD 3D printing for microfluidic organ-on-a-chip (OoC)
However, the capital cost of common research-grade devices and for mass production. An OoC for co-culture of
microfluidic DLP 3D printers (∼15–30 K USD) constitute a spheroids separated by an endothelial barrier was designed,
significant entry barrier for many potential users. printed and tested within 2 weeks. OoC devices are further
Furthermore, while the pixel numbers have increased, with used to illustrate large area printing in a well plate format,
many printers culminating at 3840 × 2160 ≅ 8 M pixels, the and for mass production by 3D printing thousands of OoC
trade-off between print resolution and build area has not devices in a single run.

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2775
 View Article Online

Paper Lab on a Chip

Materials and methods file for slicing in a third-party software, CHITUBOX, at a layer
Materials thickness of 20 or 50 μm. The slices were uploaded to the
Elegoo Mars 3 Pro, Elegoo Mars 4 Ultra, Elegoo Saturn 2,
Ink materials. Poly(ethylene glycol) diacrylate (PEGDA)-250 or Elegoo Saturn 3 Ultra (ELEGOO, Shenzhen, China) masked
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

(Cat. #475629, lot #MKCS0146, Sigma-Aldrich, Oakville, stereolithography LCD 3D printers with a 405 nm light source.
Ontario, Canada), diphenyl(2,4,6-trimethylbenzoyl)phosphine Print settings for all the devices presented here are given in
oxide (TPO) (Cat. #415952, lot #MKCK2346, Sigma-Aldrich, Table S1.† The printed chips were washed on the build plate to
Oakville, Ontario, Canada), 2-isopropylthioxanthone (ITX) remove excess uncured resin with IPA and dried with
(Cat. #I067825G, lot #ZNNQE-KT, TCI America, Portland, compressed air or nitrogen, followed by 1 min of UV curing
Oregon, United States), pentaerythritol tetraacrylate (PETTA) (CureZone, Creative CADWorks, Concord, Ontario, Canada).
(Cat. #408263, lot #MKCR5556, Sigma-Aldrich, Oakville, Embedded devices were ready for use following UV curing;
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Ontario, Canada). meanwhile, open channel CCs were sealed with a pressure
Other chemicals. 3-(Trimethoxysilyl)propyl methacrylate adhesive tape (9795R microfluidic tape, 3M, Perth, Ontario,
(Cat. #M6514, lot #SHBG7600V, Sigma-Aldrich, Oakville, Canada) to encapsulate the microchannels.
Ontario, Canada), fluorescein sodium salt (Cat. #46960, lot
#2082530, Sigma-Aldrich, Oakville, Ontario, Canada),
FTIR–ATR spectroscopy
isopropyl alcohol (IPA) (Fisher Scientific, Saint-Laurent,
Quebec, Canada). Functional group characterization was done using FTIR–ATR
Immunoassay. Purified mouse monoclonal IgG anti- spectroscopy (Nicolet 6700/Smart iTR, Thermo Scientific) on
human interferon-γ capture antibody (Cat. #MAB2852, lot both uncured and cured PLInk samples. The measurement
#FIO1022021, R&D Systems, Minneapolis, Minnesota, United was performed on 8 μL of uncured PLInk followed by curing
States), biotinylated affinity purified goat IgG anti-human the sample with a 405 nm illumination wavelength and
interferon-γ detection antibody (Cat. #BAF285, lot grinding into powder format, then loading 500 mg of the
#ZX2721071, R&D Systems, Minneapolis, Minnesota, United powder to measure the spectra of the cured sample.
States), recombinant human interferon-γ protein (Cat. #285- Functional groups were identified based on peak positions to
IF, lot #RAX2422031, R&D Systems, Minneapolis, Minnesota, monitor the difference after photopolymerization.
United States), Pierce streptavidin poly-horseradish
peroxidase (pHRP) (Cat. #21140, lot #XJ360080, Thermo Cure depth, light penetration depth, and absorbance
Fisher Scientific, Waltham, Massachusetts, United States), measurements
SIGMAFAST 3,3′-diaminobenzidine tablets (Cat. #D4293, lot To determine the penetration depth of light, 50 × 75 × 1 mm3
#SLCG5357, Sigma-Aldrich, Oakville, Ontario, Canada), glass slides were first cleaned with IPA, then silanized via
bovine serum albumin (BSA) (Cat. #001-000-162, lot #162191, liquid phase deposition by immersing a glass slides in a
Jackson ImmunoResearch Labs, West Grove, Pennsylvania, solution of 2% 3-(trimethoxysilyl)propyl methacrylate
United States), BSA-biotin (Cat. #A8549, Sigma-Aldrich, prepared in toluene for at least 2 h or overnight. The slides
Oakville, Ontario, Canada), Tween 20 (Cat. #P7949, lot were then cleaned in fresh toluene and dried with
#SLBX0835, Sigma-Aldrich, Oakville, Ontario, Canada). compressed nitrogen. The treated glass slides were placed
All assay reagents were prepared using 1× phosphate- directly on the 3D printer LCD screen; with the UV
buffered saline (PBS) (pH ∼ 7.4) supplemented with 0.05% illumination on, the power intensity was read through the
Tween 20 and 5% BSA. All other solutions were prepared glass using a UV light meter with a 405 nm probe (Model
using water from a Milli-Q system (resistivity: 18 MΩ cm; 222, G&R Labs, Santa Carla, California, United States) to be
Millipore). 2.23 mW cm−2. Then, 8 μL of uncured ink was placed on the
glass slide, and the UV light was illuminated at different
Ink preparation exposure times and repeated for each ink formulation.
The 3D printing ink was based on a low molecular weight Following exposure, the glass was cleaned with IPA to remove
PEGDA-250 supplemented with 0.5% (wt/wt) diphenyl(2,4,6- excess uncured ink, dried with compressed nitrogen, and the
trimethylbenzoyl)phosphine oxide (TPO) photoinitiator, 1.5% cure depth of the formulation was measured using a stylus
(wt/wt) 2-isopropylthioxanthone (ITX) photoabsorber, and 2% profilometer (DektakXT, Bruker, Billerica, Massachusetts,
(wt/wt) pentaerythritol tetraacrylate (PETTA) crosslinker. The United States) that was configured to measure using a 12.5
reagents were mixed in a 500 mL amber glass bottle under μm probe radius with a 3 mg force to scan a 7 mm region in
magnetic stir for at least 2 h before use and stored at room 20 s. The cure depths were recorded in Vision64 and the
temperature thereafter. average height was measured according to the International
Organization for Standardization (ISO) 4287 protocol after
2-point leveling to record the baseline.
3D printing of microfluidic chips The light absorbance of each photocurable ink was
The microfluidic chips were designed either in AutoCAD measured using a spectrophotometer (NanoDrop@ND-1000,
(Autodesk) or Fusion 360 (Autodesk), then exported as an STL NanoDrop Technologies, Wilmington, Delaware, United

2776 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

States). A blank reading was performed using MilliQ water, Fluidic demonstrations
followed by recording the light absorption spectra with 2 μL To visually assess the fluid flow in the microfluidic chips, a
of ink solution at a 0.1 mm path length. 2% solution of food dye in MilliQ water was prepared and
loaded in the chips. For the micromixer, a 10 μM solution of
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

3D printer emission spectra measurement fluorescein was prepared in MilliQ water. In the case of the
The emission spectrum of the 3D printer was measured using ELISA-chips, the devices were assessed with a solution of 2%
an extended range (200–1000 nm) charge-coupled device food dye in 1× PBS buffer containing 0.05% Tween 20.
spectrometer (CCS200, Thorlabs) via a 200 μm-diameter
bifurcated optical fiber (BFY200HF2, Thorlabs). Spectral data Fluorescent imaging through microfluidic chips
was recorded with a 500 μs integration time and analyzed in To facilitate microscopy imaging of fluorescent solutions in
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

the ThorSpectra software to obtain the emission spectra and the chips, micromixer devices were mounted to a glass slide
relative intensities. by UV photopolymerizing a drop of uncured resin between
the chips and a plain glass slide (25 × 75 × 1 mm3) for 40 s.
Viscosity measurement The device was printed with cylindrical ports connected to a
The viscosity of 3D printing inks was measured using a programmable syringe pump (Kd Scientific KDS250) via
vibrating rod viscometer (SV-10, A&D Company, Limited). A Tygon E-3603 tubing to flow solutions into the micromixer at
50 mL sample was loaded into the viscometer receptacle; the known flow rates. Fluorescent images were acquired using a
sensing rods were lowered into the receptacle until they were Nikon Ti2 inverted fluorescence microscope using NIS
fully immersed in the inks, then the rods were set to vibrate elements. Flow profiles were analyzed in ImageJ2 Ver. 2.9.0/
at a frequency of 30 Hz to measure the resistance to flow. All 1.53t (public domain software, National Institute of Health,
measurements were done at 21 °C. Bethesda, Maryland, United States).

Cell culture and cytotoxicity assay Flow rate measurements through the microvalve

To assess the cytotoxicity of the ink formulation, a To assess the functionality of the microvalve, black dyed
cytocompatibility assay was performed in compliance with water was flown continuously through the flow channel inlet
ISO 10993-5:2009 standards. The cells used in this study were and collected in a beaker at the outlet. Meanwhile, a pressure
kindly provided by Dr. Arnold Hayer of McGill University,28 gauge (MA059, MEASUREMAN) and an air pressure regulator
and they were grown and passaged according to ATCC's (850-AC, ControlAir Inc.) were used to control the air pressure
recommendations and cultured in EGM-2 media. Briefly, 8 × from a compressed air source directed at the control channel.
3 mm2 (diameter × thickness) rings were 3D printed and The pressure regulator was used to adjust the control
washed for 72 h with 70% ethanol with daily refresh of pressure and the liquid collected in the outlet beaker was
ethanol and then washed with PBS for 48 h to remove any massed after a known collection time (i.e., 10 s) on a digital
unreacted photoactive components. The rings were then co- analytical balance (XS204, Mettler-Toledo) to determine the
cultured with mCherry-labelled human umbilical vein flow rate. The air pressure in the control channel was
endothelial cells (HUVECs) seeded at a density of 100 000 increased in ∼3 kPa increments until the outlet flow rate
cells per well in a 24-well plate. Quantitative cell viability neared 0 μL s−1, indicating that the valve was closed.
measurements were performed every 24 h over a total of 72 h
using the PrestoBlue™ cell viability reagent. HUVECs seeded Contact angle measurement
at an identical density were cultured alongside the ring co- The static contact angle was measured by placing 2 μL of
culture as a control and used to establish 100% cell viability either water, PBST 0.05%, or PBST 0.1% supplemented with
for each time point. Both the control and co-culture 2% liquid food dye on a 10 × 10 × 2 mm3 3D printed part
conditions were imaged every 24 h over a total of 72 h using and then photographed from the side using a digital camera
a Ti2 inverted microscope and analyzed using NIS-Element (Panasonic Lumix DMC-GH3K) with a macro lens ([Link]
(Nikon, Japan) for all biological replicates. Digital ED 60 mm F2.8 Macro). The images were imported to
ImageJ where the Contact Angle plugin was used to measure
Numerical simulation of concentration fields in the micromixer the static contact angle.
The concentration field of the micromixer was calculated by
the finite element method using COMSOL Multiphysics v.5.6 ELISA-chip nitrocellulose assay preparation
(COMSOL, Inc., Burlington, Massachusetts, United States). The The assay was designed based on lateral flow nitrocellulose
diffusion coefficient of fluorescein (4.25 × 10−10 m2 s−1) was membranes (Vivid 120, no. VIV1202503R; Pall Corporation,
applied to solve the steady-state concentration field of Port Washington, USA) that were cut to 3 × 12 mm2 (width ×
fluorescein at a flow rate of 0.1 mL min−1. The concentration length) with a pointed base using a film cutter (Cameo 3,
field was sliced into cross-sections to obtain the splitting and Silhouette Portrait, Lindon, USA). The membranes were then
recombining stream profiles along the length of a mixing unit. spotted using an inkjet spotter (sciFLEXARRAYER SX,

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2777
 View Article Online

Paper Lab on a Chip

Scienion) with a 2.5 × 1 mm2 (width × length) test and was again incubated at 37 °C for an additional 1 h for
control line spaced 5 mm apart. The test line was spotted gelation. Finally, the media reservoirs were each loaded with
with 100 μg mL−1 of anti-human IFN-γ antibody in a 0.22 μm ∼100 μL of EGM-2 media and the device was imaged daily.
filtered 1X PBS buffer by programming the release of 350 pL
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

droplets in a 25 × 4 line array; spotting was done over 40 Videos and image processing
passes, wherein each pass covered alternating positions on
3D images of the microfluidic devices were obtained by
the line array to allow for spots to dry between passes.
micro-computed tomography (μCT) (SkyScan 1172, Bruker,
Similarly, the control line was spotted with 50 μg mL−1 of
Kontich, Belgium) at a pixel size of 8 μm. Images were
BSA-biotin with 8 passes covering alternating positions for
reconstructed using CT Analyzer (CTAn v.1.18, Bruker,
each pass. The spotted membranes were dried at 37 °C for 1
Kontich, Belgium) and orthogonal projections were visualized
h, then blocked in a solution of 0.1% Tween 20 in 1× PBS
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

and measured in ImageJ2 Ver. 2.9.0/1.53t (public domain

supplemented with 5% BSA by dipping and wetting the
software, National Institute of Health, Bethesda, Maryland,
membranes in a tray containing excess blocking buffer
United States). 3D microscopy images were taken with a
placed on an orbital shaker at 100 rpm for 1 h at room
stereomicroscope (SteREO Discovery.V20, Zeiss, Baden-
temperature. The blocked membranes were left to dry at 37
Württemberg, Germany). Videos and images were recorded to
°C for 1 h, followed by overnight storage in an air-tight
characterize flow using dyed water on either a Panasonic
container with desiccant at 4 °C, then used within 48 h of
Lumix DMC-GH3K with a macro lens ([Link] Digital ED 60
protein spotting. The nitrocellulose assay strip was mounted
mm F2.8 Macro) or Sony α7R III camera also with a macro
onto the chip and sandwiched between 4 absorbent pads
lens (Sony FE 90 mm F2.8 Macro G OSS Lens). Assay
(Electrophoresis and Blotting Paper, Grade 320, Ahlstrom-
nitrocellulose membranes were imaged using a flatbed
Munksjo Chromatography) laser cut to be 10 × 2.4 mm2
scanner (Epson Perfection V600) with the SilverFast 8
(width × length). The drainage channel of the chip was
software at 600 dpi in a 48 bit RGB format, then imported to
mounted with a 1.5 × 3.5 mm2 (width × length) glass fiber
ImageJ2 for 16 bit grayscale colorimetric line intensity
(G041 SureWick, Millipore Sigma), then both the glass fiber
readouts. The readouts were normalized to rescale the
and nitrocellulose strips were clamped into place using a
colorimetric intensity from 0–65 535 gray values to 0–1
custom 3D printed compressive clip.
relative signal intensities.

Assay protocol
Results and discussion
The assay solutions for IFN-γ detection were prepared in a
wash and diluent buffer of 0.05% Tween 20 in 1× PBS Fig. 1 shows the process flow including a low-cost LCD 3D
supplemented with 5% BSA. Recombinant human IFN-γ printer, a custom PEGDA-based ink (PLInk) optimized for LCD
protein was spiked in the buffer at concentrations of 0, 100, 3D printing, some microfluidic devices fabricated in this study,
101, 102, 103, 104, 105, and 106 pg mL−1, followed by and the device design that closes the rapid prototyping cycle.
preparing reagent solutions including anti-human IFN-
gamma biotinylated antibody at 1 μg mL−1 and streptavidin- Design of ink for LCD 3D printing
poly-horseradish peroxidase (pHRP) at 25 μg mL−1. To
Embedded microfluidic channels are designed as narrow voids
prepare the assay substrate solution, SIGMAFAST™ DAB
in a block of solid ink. To create these voids, the design of an
(3,3′-diaminobenzidine) tablets were dissolved in 5 mL Milli-
ink formulation consists of monomers, light-responsive
Q water, then 0.22 μm filtered prior to running the assay.
additives (i.e., photoinitiator to catalyze the reaction,
photoabsorbers to absorb excess energy), and crosslinkers.
Organ-on-a-chip seeding and maintenance Polymerization must proceed efficiently layer-by-layer, i.e.,
Spheroids were formed by seeding and culturing ∼10 000 within the defined thickness of each layer while both avoiding
cells in a 96-well ultra-low attachment plate. Then, a five-day- under-polymerization of the current layer, and over-
old MDA-MB-231 breast cancer spheroid stained with 5-(and- polymerization of uncured ink in voids of the preceding layers.
6)-carboxyfluorescein diacetate succinimidyl ester (CFDA-SE) The design of PLInk was based on our prior ink
(Invitrogen, Massachusetts, United States) and a five-day-old formulations for 385 nm DLP 3D printing,7,16 and adapted
IMR-90 lung fibroblast spheroid stained with cell tracker for LCD-based photopolymerization by considering the light
deep red (Invitrogen, Massachusetts, United States) was heterogeneity, low irradiance, and 405 nm illumination
embedded in ∼5 μL of 50% Matrigel DMEM solution and wavelength. Based on our prior inks, PEGDA-250 was selected
placed into the cell seeding chambers of the organ-on-a-chip as the monomer due to its low viscosity, low protein
device. The device was incubated at 37 °C for 30 min for adsorption, inherent cytocompatibility, and compatibility
gelation, then the central cell seeding chamber was loaded with solvents such isopropyl alcohol for efficient removal of
with a 50% Matrigel EGM-2 solution containing ∼200 000 uncured ink in embedded microchannels. Diphenyl(2,4,6-
mCherry-labelled HUVECs and placed on ice while the trimethylbenzoyl)phosphine oxide (TPO) was selected again
channels filled via capillary flow. Following filling, the device as the photoinitiator due to its low cytotoxicity and an

2778 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Fig. 1 Low-cost LCD 3D printing of microfluidic devices and the rapid prototyping cycle. The workflow including (1) manufacturing on low-cost
LCD 3D printers using a custom PEGDA-based ink (PLInk) optimized for LCD 3D printing, (2) directly manufactured microfluidic devices that can
be tested and characterized, and (3) inform design improvement for the next iteration. Pixel sizes of 18–50 μm and print areas of up to 218 × 122
mm2 afford high print resolution over large areas. Photoinks optimized for LCD 3D printing with reduced sensitivity to light heterogeneity and low
viscosity enable the direct manufacture of microfluidic chips including open and embedded microchannels with a lateral resolution <100 μm and
vertical features as thin as 22 μm in <45 min. Scale bars = 500 μm.

activation peak between 380–425 nm, as well as reactivity (discussed further below). Each of these ink
2-isopropylthioxanthone (ITX) as the photoabsorber due to its components individually met suitability for a 405 nm
broad absorbance peak between 350–425 nm, and known illumination source, Fig. 2a.
optical transparency, unlike other photoabsorbers such We confirmed photocuring by Fourier transform infrared–
2-nitrophenyl phenyl sulfide (NPS), Sudan-1, or UV absorbing attenuated total reflectance (FTIR–ATR) spectroscopy on
dyes with poor cytocompatibility and yellow-orange tints. The uncured and 405 nm cured PLInk samples. A broader peak at
concordance between the activation range of TPO and 1200 cm−1 was observed for the cured ink, consistent with
absorbance range of ITX allowed us to model our ink design carbon–carbon bond formation between adjacent PEGDA and
with the assumption that the absorbance remains consistent PETTA acrylate groups, ESI† Fig. S2.
over the photopolymerizable region, even where the PLInk Next, to mitigate the effects of light inhomogeneity, we
absorbance spectrum cuts off before the tail end of the 3D sought to characterize the photopolymerization of the ink as
printer emission spectrum, ESI† Fig. S1. In a case where the function of total energy dosage and varying ITX
photoinitiator range extended beyond the region attenuated photoabsorber concentration from 0 to 1.5%; the latter being
by the photoabsorber and within the 3D printer emission the maximal concentration at which ITX could readily be
spectrum, a polychromatic configuration that accounts for a dissolved. The fabrication of embedded microchannels, i.e.,
changing absorbance could be considered.12,29 Finally, due to voids, is predicated on precise control and understanding of
the low irradiance of LCD 3D printers, we added penta- the (measured) cure depth, Cd, to both avoid cross-linking of
erythritol tetraacrylate (PETTA) crosslinker to increase uncured ink trapped inside the microchannel while ensuring

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2779
 View Article Online

Paper Lab on a Chip

This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Fig. 2 PLInk optimization and characterization for LCD 3D printing of surface and embedded microstructures and microfluidics. (a) Formulation of
PLInk containing PEGDA-250, TPO, ITX and PETTA. (b) Jacob's working curve showing the cure depth as a function of the energy dose for 1.5%
ITX and yielding Dp = 56.5 μm and Ec = 2.31 mJ cm−2. (c) Dp and Ec values derived as in (b) for different concentrations of ITX showing a plateau
for ITX > 0.75%. (d) Pillar printed area compared to nominal area of 0.2 × 0.2 mm2 as function of crosslinker concentration. Data shows mean ±
standard deviation (STD) of four replicates. (e) 3D printed pillars showing features as small as ∼3 × 3 pixels (nominal printer pixel size = 28.5 × 28.5
μm2). Scale bar = 100 μm. (f) Monolithic circular microchannels with a corresponding μCT scan with channel cross-sections radii of ∼125 μm (left)
and ∼110 μm (right) (nominal printer pixel size = 35 × 35 μm2). Scale bar = 250 μm. (g) μCT image of a 22 μm thick embedded membrane (nominal
printer pixel size = 18 × 18 μm2). Scale bar = 500 μm. (h) Cell viability at different time points of HUVECs expressing actin-mCherry co-cultured
with 3D printed PLInk rings. Data shows mean ± STD of three biological replicates. Scale bar = 100 μm.

curing of the working layer. Cd is experimentally measurable The LCD 3D printer irradiance was measured to be 2.23 mW
and varies as function of the total energy, E, according to cm−2. The thickness of polymerized ink for 1.5% ITX as
Jacob's working curve:12,22,30,31 function of energy dosage was fitted with Jacob's curve to
  derive both Dp = 56.5 μm and Ec = 2.31 mJ cm−2, Fig. 2b. The
E
Cd ¼ Dp ln same experiment was repeated for varying concentrations of
Ec
ITX, and the resulting Dp and Ec derived, Fig. 2c. As expected,
where Dp is the penetration depth of light at which the light Dp decreased with increasing ITX concentration.
intensity of incident light is reduced by a factor 1/e derived Interestingly, Ec also decreased with increasing
from Beer–Lambert's law,22 and Ec is the critical energy photoabsorber, suggesting that ITX contributes not only to
dosage corresponding to the minimum required energy to light adsorption, but also to more effective
initiate photopolymerization. E is simply te, the exposure photopolymerization of the ink. We also observed that the
time multiplied by P, the irradiance: slope was flattened for higher ITX concentration, meaning
that the variation in the thickness of photopolymerized ink
E = te × P as a result of light illumination inhomogeneity would be

2780 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

minimized. Hence, high ITX concentrations were optimal for suggests favourability for intricate microchannel fabrication,
LCD 3D printing. While we observed a lower plateau in both ESI† Section S1 and Fig. S4.
Dp and Ec once ITX concentrations reached 0.75%, we chose To assess suitability for microfluidic device fabrication, we
1.5% as the optimal condition to minimize susceptibility to evaluated the resolution of the designed PLInk formulation
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

variations in ITX concentration. by printing open channels with decreasing size and were able
To balance precision, material sensitivity and print speed, to print features as small as ∼75 × 75 μm2 with a 35 μm pixel
and while considering printer pixel size, we set the print layer size LCD 3D printer, ESI† Fig. S5. We performed μCT scans
thickness (and model slicing) to 20 or 50 μm. This satisfied of the device to evaluate the printing accuracy; we measured
the requirement for printing embedded microchannels of the printed open channel size and found it to be within 2.8%
slice layer thickness = 0.3–1 × Dp formulated by Nordin and of the nominal dimension.
colleagues.22 PLInk also allowed for rapid To assess our ability to 3D print embedded
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

photopolymerization with an exposure time te of 1.3–1.8 s microchannels, we similarly evaluated the printing of
and 3.5 s for a Cd of ∼20 μm and ∼50 μm, respectively. As an progressively smaller rectangular and circular channels that
example of the benefits of lower Dp, for a change in energy were embedded a depth at least ten times greater than the
dosage of 7–9 mJ cm−2, the layer thickness variation with Dp. Embedded rectangular channels down to ∼170 × 220 μm2
1.5% and 0.02% ITX would be ∼20 μm and ∼80 μm, (width [W] × height [H]) were printed using a 35 μm pixel size
respectively, ESI† Fig. S3. Commercial inks typically favor a LCD 3D printer, ESI† Fig. S6. The smallest rectangular
high Dp (>179 μm),32 which has the advantage of printing embedded conduits were within 2.7% of their nominal size.
thicker layer slices and faster print times, but are inadequate We found that high aspect ratio (H/W > 1) channels were
for printing embedded microchannels and susceptible to limited by the pixel resolution of the LCD screen, i.e.,
variable cured thickness with a non-uniform light source. typically 3–4 pixels because of scattering, non-parallel
To improve printing fidelity, we supplemented the PLInk illumination, and possible photoinitiator diffusion.35,36
formulation with PETTA with four additional acrylate groups Meanwhile, the height of low aspect ratio microchannels (H/
to increase the availability of polymerizable groups and speed W < 1) was limited by the optical penetration (the shortest
up the formation of an interconnected polymer network. We embedded channel ∼2.3 × Dp).12 Circular conduits are
empirically adjusted the PETTA concentration by measuring notably of interest to minimize capillary edge flow (also
the printed area of 0.2 × 0.2 mm2 pillars with a 28.5 μm pixel called filaments),7 and embedded conduits with circular
size LCD 3D printer. Incomplete photopolymerization was cross-section and radius as small as ∼110 μm were printed
visualized by tracking underfilling of the nominal pillar with a dimensional accuracy within 1.5% of the nominal
shape and by the distortion or bending of the pillars.33 The dimension, Fig. 2f and S7.† A shallow Dp also benefits the
PETTA concentration was increased until the nominal XY printing of thin embedded membranes due to fine control
pillar area matched the 3D printed design, which was over the cured thickness and a sharp transition between
achieved at a value of 2%, Fig. 2d. Pillar printing confirmed cured and uncured layers. Vertical embedded channels
suitable mechanical stability of the print without collapse designed with a series of ever thinner membranes were 3D
and good dimensional accuracy, as illustrated with an array printed down to a thickness as low as ∼22 μm within a single
of ∼3 × 3 pillars, Fig. 2e. exposure to demonstrate free-standing membrane
fabrication, Fig. 2g and S8.†
Further, we evaluated the cytocompatibility of the ink by
PLInk performance characterization co-culturing 3D printed PLInk with human umbilical vein
PLInk is based on PEGDA-250 with a comparatively low endothelial cells (HUVECs) according to the ISO 10993-5:2009
viscosity of ∼16 mPa s, thus making it suitable for standard for implantable medical devices. A primary cell line
microchannel fabrication. Indeed, following photoexposure was selected due to specific but rigorous culturing conditions
of a layer, the retraction of a relatively flat print attached to for cells with high sensitivity to their environment and a
the vat bottom will create suction force; next, ink needs to limited passage number. We 3D printed 8 × 3 mm2 (diameter
flow into the growing gap, and immediately flow out of the × thickness) rings and thoroughly washed any unreacted
gap as the print is lowered back onto the vat bottom to photoactive elements (details in the Methods), then co-
expose the next layer, all of which would benefit from a low incubated the PLInk rings with cells in a single well with
viscosity ink. Notably, despite its high viscosity (∼700 mPa s shared media for 72 h.15 After 72 h, we found >80% cell
at 25 °C),34 the addition of PETTA at low concentrations did viability, meeting the threshold for a cytocompatible material
not impact the native viscosity of PEGDA-250, and thus vastly and demonstrating suitability for cell culture microfluidic
outperformed commercial inks (∼200–500 mPa s) in this device fabrication, Fig. 2h.
respect, ESI† Table S2. The low viscosity also facilitates In summary, the optimized and low viscosity PLInk
printing of fine features as it reduced the risks of mechanical formulation for LCD 3D printing was found to be suitable for
failure caused by suction and adhesion to the vat bottom. high-resolution and dimensionally accurate printing of
Coupling low viscosity and low Dp, the cured PLInk was smooth structures including posts, open and embedded
smooth with a surface roughness of ∼500 nm, which microchannels, embedded membranes, and to be

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2781
 View Article Online

Paper Lab on a Chip

cytocompatible, making it amenable for a broad range of that interweaved, merged into a large conduit of 900 × 900
applications, and notably in microfluidics as explored below. μm2, and then split again, and so on. The shaped pillars
measured 203 μm (≅ 6 pixels, nominal 3D printer pixel size
LCD 3D printing of embedded microfluidic mixer = 35 × 35 μm2) along its longest dimension and 306 μm (≅
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

9 pixels) along its widest dimension, Fig. 3b. The pillar

We re-designed a recently published micromixer extended across the full height of the mixer (45 layers of 20
implementing the Baker's transformation and made by μm each), which necessitated a sufficiently high energy
direct laser micromachining of two separate parts for direct dosage to fully crosslink the pillar as well as overhanging,
LCD 3D printing.37 In the original design, the micromixer suspended structures and preserve their integrity during the
was assembled from two substrate devices with open build plate movements, while at the same time preventing
conduits made by direct laser machining, chemical wet photopolymerization of PLInk in the embedded weaving
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

etching, and bonding, thus forming a closed, conduits. μCT scans of the 3D printed device confirmed the
interconnected weaving flow path. For LCD 3D printing, the fidelity and integrity of the pillar that narrowed down to
micromixer was designed as a single digital model of an ∼45–72 μm (≅ 1–2 pixels) in width before recombining the
embedded micromixer including (i) overhanging wedges to fluidic streams, and of the internal overhangs and the
split the fluidic streams horizontally and progressively, and sharp edges that split, guide, and merge the fluidic
(ii) an embedded pillar to create an interface before streams, Fig. 3c(i). A finite element method numerical
vertically recombining the two fluidic streams, Fig. 3a. The simulation that solved the steady-state concentration field
micromixer was designed with 310 × 310 μm2 cross-sections of two fluidic streams illustrated the importance of the

Fig. 3 Embedded microfluidic mixer. (a) Schematic representation of two fluidic channels combined into a single mixing unit. (b)
Stereomicroscope image of four 3D printed mixing units showing overhangs and pillar formation in an embedded device. Scale bar = 500 μm. (c)
(i) μCT cross-sections of the microchannel with (ii) a corresponding numerical simulation showing the mixing principle based on horizontal stream
splitting and vertical stream recombining. Scale bar = 500 μm. (d) Stereomicroscope image of a single mixing unit showing yellow and blue fluidic
streams split into ever thinning striations by the microarchitecture. Arrow shows the direction of flow. Scale bar = 300 μm. (e) Mixing of 10 μM
fluorescein with clear water at a flow rate ≅ 0.1, 1, and 10 mL min−1, corresponding to a Reynold's number ≅ 1.85, 18.5, and 185, respectively.
Arrow shows the direction of flow. Scale bar = 500 μm.

2782 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

microarchitecture to horizontally split and vertically the mixing efficiency did not decrease, but instead improved
recombine the flows for cross-sections along the length of again, which we attribute to inertial effects and recirculation.
the mixing unit, Fig. 3c(ii). The mixing performance was concordant with the laser-
Owing to the transparency of the device, mixing could be manufactured mixer and the Baker's transformation
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

visually tracked through the entire height of the channel. principle.37,38 Across three replicate devices, we quantified
Using water with yellow and blue dyes allowed for visual the mixing efficiency to be 92–99%, confirming the
tracking of the mixing and the observation of striations as successful printing and operation of the 3D printed device,
the streams folded and recombined within the micromixer, ESI† Fig. S10, Section S2. The micromixer illustrates the
Fig. 3d. We further assessed the mixing performance with a potential of LCD 3D printing for producing complex
fluorescent dye (10 μM fluorescein) in one of the streams and embedded structures that are not easily manufactured by
tracked the fluorescence intensity along the length of the more traditional micromachining methods.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

mixer by fluorescence microscopy. The progression from two

separate streams to complete mixing was visible from the
intensity profile that progressed from a step function to a LCD 3D printing of an embedded membrane microvalve
flat, homogeneous distribution of the dye across the width of Next, we 3D printed an embedded membrane microvalve.
the micromixer, Fig. 3e and S9.† When investigated over a Elastomeric microvalves made of polydimethylsiloxane
range of laminar flow rates (0.01–10 mL min−1), we observed (PDMS) and manufactured by replica molding have been
the efficiency of mixing decreased with increasing flow rates, widely used and adopted for microfluidics.39,40 Recently,
as expected because the time for diffusive mixing decreases. direct manufacturing of embedded free-standing membranes
Interestingly, we observed that for flow rates >1 mL min−1 by 3D printing has been demonstrated using DLP 3D

Fig. 4 Embedded membrane microvalve. (a) Microvalve schematic and (b) photograph showing flow channel, control channel, a ∼43 μm thick
membrane, ridged valve seat forming a separation wall with a 100 μm gap to the membrane at the centre (nominal 3D printer pixel size = 28.5
× 28.5 μm2). Actuation of the membrane by pressurization in the control channel leads to deflection onto the valve seat and closure of the
flow channel. (c) Orthogonal μCT views of the 3D printed membrane and valve seat. Scale bars = 500 μm. (d) Top view images of the open
and closed valve with schematics showing cross-sections of the valve according to the labels in (b). With an open valve shown on the left, the
black water flows through the channel, while for a closed valve shown on the right, the pneumatically deflected membrane is sealed onto the
valve seat and stops black water flow. Scale bars = 500 μm. (e) Flow rate in the flow channel as a function of the pressure in the control
channel showing the gradual closing of the valve and flow stop at ∼41 kPa. Data points are measurement collected from three different
devices. Line is a guide to the eye.

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2783
 View Article Online

Paper Lab on a Chip

printers.30,41 We LCD 3D printed an embedded membrane for self-filling, and function thanks to a controlled,
with a valve seat modelled based on existing ones comprising moderate hydrophilicity. We previously developed
a 40 μm-thick membrane with a diameter of 1.7 mm and hydrophilic inks for DLP 3D printing of functional CCs with
∼100 μm above a 500 μm-wide ridged valve seat. An embedded channels and with contact angles with water
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

embedded control channel overlaid orthogonally above the ∼35° owing to the use of hydrophilic acrylic or methacrylic
membrane and the valve seat in the flow channel was used acid additives.7 The contact angle with water of PLInk was
for membrane actuation by pressurization, Fig. 4a and b. All ∼65–70°, which while being moderately hydrophilic, was
the channels were 3D printed with a unique inlet and outlet too high for reliable capillary self-filling, ESI† Fig. S12a. The
to facilitate precursor ink removal and avoid post-processing photopolymeriziation of acrylic or methacrylic acid groups
fabrication steps. The valve seat in the form of a thin curved competes with crosslinking by PEGDA acrylate groups, and
ridge improved printability compared to a solid ‘bowl’ shape thus requires higher light energy doses, which would lead
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

that might lead to incomplete ink removal while providing to much higher exposure times for low irradiance LCD 3D
reliable valve closure upon actuation. μCT images of the valve printers. Previously, plasma activation had also been used,
revealed a fully released, free-standing membrane, Fig. 4c, but depends on access to a plasma chamber, and only
ESI† Video S1. The measured thickness on the μCT images provides temporarily hydrophilicity for select materials.18,19
(with 8 μm pixel resolution) was ∼43 μm, closely matching Hence, instead of increasing the surface energy of the
the design. microchannels, we opted to reduce the surface tension of
Water spiked with a black dye was flown through the the aqueous solutions by adding surfactants (i.e., Tween 20)
microvalve to visually assess whether the valve was open (i.e., and thereby reducing the contact angle to as low as ∼46°
flow channel junction was visually black) or closed (i.e., with 0.05% Tween 20 and ∼31° with 0.1% Tween 20, thus
junction visually clear). The valve was designed to be open at meeting the requirements for CC operation, ESI† Fig. S12b
rest, and as the compressed air pressure was increased in the and c. Considering that the use of surfactants in
control channel, the membrane deflected to form a seal with immunoassays is common to reduce non-specific binding,
the valve seat, interrupting the flow of the black water, their addition to the solutions does not compromise the
Fig. 4d. suitability of CCs for typical biological applications.
The mechanical properties of 3D printed PLInk were To illustrate the reliability of LCD 3D printing, we
assessed by tensile testing yielding a Young's modulus of 68 designed a CC with a microfluidic chain reaction (MCR)18
± 3 MPa, ESI† Fig. S11. Compared to elastomeric membranes, implementing an ELISA-on-a-chip akin to the ones made
PLInk's Young's modulus was ∼10× higher than PDMS; previously using DLP 3D printing of open microchannels
therefore, a thin (∼40–50 μm), 1.7 mm diameter membrane followed by sealing with a hydrophobic pressure adhesive
was predicted to deflect ∼100 μm at a control pressure of transparent cover.18,19 The ELISA-chip was developed for a
∼45 kPa to seal the valve, ESI† Section S3. The control new target, with adjusted geometries for LCD 3D printing,
pressure was increased incrementally while the flow was and importantly with a reduced time-to-result while
monitored and flow stop observed at ∼41 kPa, Fig. 4e. The maintaining high sensitivity, Fig. 5a. The target was
experimental valve closing pressure was thus in good interferon (IFN)-γ, a cytokine critical to the immune response
agreement with the prediction, and the variation could be against a wide range of infections,42 and which is notably
attributed to imprecision in the gap between the membrane used in the IFN-γ release assays as a biomarker for
and the valve, in the thickness of the membrane, or tuberculosis infection.43,44 The microfluidic assay was based
incomplete curing of the membrane that might make it more on a classical ELISA sandwich immunoassay using a capture
pliable. Overall, both the reproducibility of the closing antibody, a biotinylated detection antibody, and a
pressure across all valve replicates, and the agreement to streptavidin-enzyme conjugate (poly-horseradish peroxidase,
theory were consistent. While we did not assess the durability pHRP). While in conventional well-plate ELISAs soluble
under cyclical stress loading, the durability of 3D printed substrates are used, for on-chip applications with a
membranes based on low molecular weight PEGDA inks was nitrocellulose membrane and under active flow conditions,
demonstrated by Folch and colleagues,41 suggesting that the precipitating substrates are required for localized
PLInk membrane will also be suitable, or could be made accumulation of the enzymatically oxidized substrate, such as
suitable, for cyclical loading. These results indicate that LCD 3,3′-diaminobenzidine tetrahydrochloride (DAB), in the
3D printing can be used for making thin, compliant, and presence of pHRP and hydrogen peroxide, Fig. 5b. A
mechanically actuated embedded elements such as nitrocellulose membrane spotted with an anti-IFN-γ capture
membrane microvalves. antibody was connected to the ELISA-chip that encoded an
8-step assay for automated, sequential flow of wash buffers
and reagents. As in the DLP 3D printed ELISA-chip design,
LCD 3D printing of an ELISA-on-a-chip capillaric circuit – an functions for on-chip aliquoting were integrated to facilitate
ELISA-chip the operations for untrained users, Fig. 5c.
CCs operate by structurally encoding fluidic operations The lower limit of detection of the previous ELISA-chip19
using capillary valves for fluidic operation and capillary flow outperformed rapid tests (e.g., lateral flow assays), but the

2784 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Fig. 5 LCD 3D printed microfluidic ELISA chip. (a) ELISA-on-a-chip designed for LCD 3D printing with structurally encoded sequential delivery of
reagents to autonomously perform an assay and coupled with a built-in chip-to-assay connection. (b) ELISA workflow showing the sandwich
immunoassay designed for the detection of IFN-γ by sequentially delivering assay reagents and wash buffer to a nitrocellulose membrane pre-
spotted with anti-human IFN-γ capture antibody; (1) sample containing IFN-γ, (2) biotinylated anti-human IFN-γ detection antibody, (3)
streptavidin-conjugated enzyme pHRP, and (4) enzyme substrate in the presence of hydrogen peroxide to generate the colorimetric readout. (c)
Autonomous CC workflow for on-chip reagent aliquoting by metering the correct reagent volumes and drainage of the excess, followed by (d)
MCR-based sequential delivery of assay reagents with wash steps in between. Arrows show the direction of flow. Scale bar = 5 mm. (e) Binding
curve of the on-chip assay for the detection of IFN-γ with a limit of detection of 12 pg mL−1 (CV: 6.8%) across triplicate chips for each tested
concentration point; line shows a 4-point logistic fit.

assay time was longer at 1 h 15 min. Thus, we sought to our previous ELISA-chip design that also had a glass fiber
reduce the assay time for the LCD 3D printed ELISA-chip. conjugate pad mounted the nitrocellulose and served both as
The incubation times were structurally encoded by the a fluidic connection to the chip and an immediate capillary
volume of reagents that flowed over the test zone (see pump to wick reagents over the nitrocellulose, the glass fiber
discussion on assay optimization below for further details), was considered a source of analyte loss due to protein
the capillary pressure of the pump (i.e., absorbent pad and adsorption over the assay run time. To remedy these
glass fiber conjugate pad backing the nitrocellulose limitations, we connected the nitrocellulose membrane to the
membrane), and the flow resistance of the functional ELISA-chip directly. Without the glass fiber, the chip-to-assay
connections that linked each reservoir to the main channel. connection was re-designed as a gradual opening with a weak
The capillary pressure coming from an absorbent pad stop valve designed to break when the liquid front arrived at
backing the nitrocellulose membrane was the same as a the end of the channel and wetted the nitrocellulose
single pump was used to wick all the reagents. Compared to membrane; pre-wetting with buffer bridged the ELISA-chip's

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2785
 View Article Online

Paper Lab on a Chip

liquid interface with the absorbent pad, and facilitated a (25%), and pHRP (24%), ESI† Table S4. Altogether, this
connection to the capillary pump that subsequently began to indicated that a to reduce assay time while preserving the
wick the reagents over the nitrocellulose assay test zone. sensitivity, capture antibody spotting density needed to be
Finally, to adjust the flow rate, we increased the functional increased. We kept the reagent volumes relatively low, i.e.,
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

connection cross-sections to 200 × 200 μm2 across the entire sample volume was 75 μL and took ∼14 min to flow, and
chip. These changes reduced reagent loss and provided a ensured that all the reagent were being delivered to the
suitable flow speed for consistent fluidic performance, which nitrocellulose membrane with no losses on a connecting
allowed us to reduce reagent volumes and the time-to-result glass fiber; meanwhile, we increased the spotting density of
to 48 min, Fig. 5d, ESI† Video S2. capture antibody from our original ELISA-chip by nearly 10-
We evaluated the flow reproducibility of the new LCD 3D fold, resulting in 0.7 μL of 100 μg mL−1 capture antibody
printed ELISA-chip by timing each of the sequential steps in spotted on a thin 3 × 1 mm2 (width × length) line on the
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

three replicate chips, Table 1. The flow of sample, which nitrocellulose membrane. Taking the relative contribution of
contains only limited concentration of analyte is the most each parameter into consideration, the optimal IFN-γ assay
critical step when considering assay reproducibility and LOD, involved flowing 45 μL (at a fixed flow rate and fixed time) of
and the one that necessitated high reproducibility. Other the detection antibody at 1 μg mL−1 and of pHRP at 25 μg
steps, such as detection antibody and enzyme are provided in mL−1. The assay wash steps volume and time were minimized
excess concentration and hence variation of flow time is not to reduce assay run time while preventing pre-mixing of
expected to significantly affect the assay result. Likewise, reagents in the main channel.
precise incubation time for wash steps are not as critical as Following both optimization of the fluidic performance
long as reagents are flowed and flushed across the and the nitrocellulose assay, we evaluated the ELISA-chip
nitrocellulose membrane. The comparatively high variability over a wide concentration range of IFN-γ and achieved a limit
for the DAB incubation time could arise as a result of the of detection as low as 12 pg mL−1. With a 6.8% CV, our LCD
precipitate formed on the test strip, especially at higher 3D printed ELISA-chip showed consistent performance,
concentrations of IFN-γ, which could affect the flow Fig. 5e and S13.† Using PLInk and based on the cost of
properties of the strip. research-grade materials, assay reagents and assembly
The assay portion of the ELISA-chip was optimized using a components of the nitrocellulose assay, the ELISA-chip costs
design of experiments approach, which enabled the <4 USD per device, ESI† Table S5. The low capital cost and
optimization of multiple assay parameters simultaneously low material cost enable affordable fabrication of
since the optimal concentration of one parameter would autonomous ELISA-chip devices globally, especially in low-
dictate the optimal of another in a classical sandwich and middle-income countries with limited access to
immunoassay, and served to establish the relative traditional manufacturing and a high incidence of infectious
contribution of each parameter.45 We evaluated a capture disease. These results indicate the suitability of low-cost LCD
antibody spotting concentration of 50, 100 and 200 μg mL−1 3D printing for the fabrication of ready-to-use CC chips that
and both a detection antibody and pHRP concentration of 1, automate complete assays with lab-grade accuracy and short
5, and 25 μg mL−1 at a fixed sample concentration of 100 ng time-to-result.
mL−1. Using the Taguchi method for design of experiments,46
the selection led to nine experiments to determine
significantly impacting assay factors, ESI† Table S3. From the LCD 3D printing of organ-on-a-chip devices
results, we evaluated the significance of each factor using We 3D printed an organ-on-a-chip (OoC) device that included
analysis of variance and found that the capture antibody embedded microchannels. Microfabricated OoC devices have
concentration was a significant parameter ( p < 0.05) for the largely been fabricated by soft lithography for manufacturing
assay performance, and the weighted contribution of the high resolution microstructures; replica molded devices
capture antibody concentration was found to be 47%, which fabricated with PDMS offer biocompatibility, tunable
was higher than the other factors, i.e., detection antibody elasticity, and gas permeability.47 Recently, DLP 3D printed
OoC and cell culture devices have been introduced with
complex 3D architectures not feasible by replica molding.16,48
Table 1 LCD 3D printed ELISA-chip steps, reagent, volume, and timing However, the biocompatibility of photoinks including the
of automated assay photopolymer and additives such as the photoinitiator and
Reagent Vol. [μL] Time ± STD [s] photoabsorber must be validated, and in some case poorly
cytocompatible phototoxins can be leached out for
1 Sample (IFN-γ) 75 875 ± 16 (CV: 1.9%)
2 Wash buffer, PBST 0.05% + 5% BSA 15 152 ± 21 (CV: 14.2%) applications requiring cell seeding, cell reorganization and
3 Biotinylated detection antibody 45 513 ± 21 (CV: 3.9%) migration.11,15,29
4 Wash buffer 15 144 ± 19 (CV: 12.8%) In the case of PLInk, biocompatibility was assessed after
5 Streptavidin-pHRP 45 518 ± 34 (CV: 6.5%)
washing and sterilizing the 3D printed devices in 70%
6 Wash buffer 15 146 ± 17 (CV: 17.5%)
7 Enzyme substrate DAB 45 504 ± 52 (CV: 10.3%) ethanol and PBS for 5 days before cell seeding (see PLInk
8 Wash buffer 5 79 ± 25 (CV: 31.6%) performance metrics above and Materials and methods for more

2786 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

details). Then, to evaluate its potential for an LCD 3D printed observed, Fig. 6e and S16,† indicating that cells can be
cell culture device, we developed a new multi-OoC spheroid introduced in LCD 3D printed microchannels for multi-day
design to monitor the interaction between two organ studies and they begin self-assembling into a micro-
compartments. In vivo, organ–organ segregation and physiological system. While the current study is a proof-of-
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

communication are maintained by endothelial barriers and concept of an LCD 3D printed OoC device, future iterations
vascular flow, respectively. To mimic native physiology, we would benefit from optimizing the ink for cell culture
designed two shallow spheroid seeding compartments that applications, and from optimizing the valve and central
were separated with a 400 μm wide vascular barrier channel chamber geometry to ensure connectivity of endothelial
in the middle. To maintain interconnectivity between sprouts with the spheroids.
spheroid compartments and the vascular barrier, we added
capillary stop valves designed as parallel embedded LCD 3D printers and digital mass manufacturing
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

microchannels that preserve compartmentalization during We explored the potential for LCD 3D printers to mass
seeding subsequently forming an open conduit between the produce OoC devices in a stacked array of 19 × 18 × 10 (X × Y
two reservoirs and the central vascular barrier, Fig. 6a and b. × Z) devices connected by short breakable supporting struts
To increase the surface area of crosstalk, capillary stop valves providing mechanical stability during printing. The OoCs
with a 200 × 200 μm2 square cross-section were stacked as a were printed with a 50 μm layer thickness which decreased
3 × 4 array between the vascular barrier and each of the two print times. 3420 OoC devices were manufactured in an 8 h
organ compartments on either side; thus, each OoC included print run, Fig. 6f, and individual devices could be retrieved
24 embedded stop valves. A cross-section visualized by by breaking the struts. Given the vertical print range, we
stereomicroscopy shows the 3D printed structures, Fig. 6c. foresee that it would be possible to make >10 000 OoC
The spacing and dimensions of the OoC devices were devices in 24 h on a single 3D printer without user
designed to match the overall footprint of an industry-standard intervention. In consideration of the low capital cost of ∼500
384-well plate with 4 inlets per OoC, i.e., two for seeding the USD for a 12K LCD 3D printer, the possibility to start
spheroids and one to seed the endothelial cells and one for air printing immediately upon receipt of a digital design file,
to exit, hence 192 OoC units and 768 inlets in total per well and the minimal user intervention needed, the use of
plate. The large build area of the LCD 3D printer could multiple such 3D printers could be attractive for on-demand
accommodate up to three OoC plates that could be printed mass production applications.
within <1.5 h, Fig. 6d. To validate the fluidic operation, the
wells were loaded with gelatin solutions spiked with red and Conclusion
green dye to mimic an extracellular matrix loaded into the
seeding chamber, which were found to be effectively We presented the use of low-cost photopolymerization LCD
compartmentalized with 100% yield on three separate plates, 3D printing for the fabrication of microfluidic devices using
ESI† Fig. S14. The cost based on research-grade materials used PLInk, optimized for rapid polymerization under low
for PLInk synthesis of an individual OoC device and of a 384- irradiance, 405 nm illumination, and reliable printing of
well plate OoC are estimated at ∼0.10 USD and ∼20 USD, embedded microchannels (and thin membranes) despite
respectively, ESI† Table S6. Cost savings could readily be illumination inhomogeneity. The effect of ITX photoabsorber
achieved by increasing the void space on the plate, as is concentration on ink photopolymerization and notably Dp
common for injection molded well plates. and Ec was characterized for 20 μm-layer-by-layer printing.
To demonstrate cell culture compatibility of the multi- Posts with lateral resolution of 75 μm, embedded membranes
OoC design, the side chambers of the OoC device were 22 μm thin, and embedded microchannels with rectangular
seeded with a Matrigel-embedded five-day-old IMR-90 lung cross-sections of 170 × 220 μm2 and round cross-sections
fibroblast spheroid on the left and a Matrigel-embedded five- with 110 μm radius were 3D printed. Further, we
day-old MDA-MB-231 breast cancer spheroid on the right. demonstrated that microfluidic devices previously made by
The central channel was loaded with HUVEC-embedded other methods, such as laser machining, replica molding and
Matrigel solution to form the vascular barrier and the DLP 3D printing, can now be fabricated using LCD 3D
remainder of the device was loaded with media. To printing, including an embedded micromixer, a membrane
demonstrate reproducible seeding of the devices, three microvalve, and an ELISA-chip for IFN-γ detection. We also
devices 3D printed separately were seeded and imaged, 3D printed an OoC device and demonstrated high
showing rounded spheroids isolated in their respective throughput manufacturing by fabricating a 384-well plate
chambers and the endothelial cells in the middle channel, format OoC within 1.5 h. Finally, we 3D printed 3420 OoC
ESI† Fig. S15. Then, one OoC device was monitored for a devices in an 8 h print run, and considering the vertical print
time course of 5 days, revealing maturation of the model with capability, we anticipate the possibility of printing >10 000
endothelial cell tightening and reorganizing into a vascular OoC devices in a 24 h print run with just one 3D printer,
structure. Initial sprouting of the endothelial cells towards demonstrating the potential for on-demand mass production.
the spheroids and some migration of the breast cancer cells We note that the OoC devices and plates were
from the spheroid towards the vascular barrier were conceptualized, designed and manufactured within 2 weeks

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2787
 View Article Online

Paper Lab on a Chip

This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Fig. 6 LCD 3D printed OoC device. (a) Design of the OoC device with two organ compartments seeded with a lung fibroblast spheroid (red) and
breast cancer spheroid (green) and separated by a vascular barrier channel seeded with endothelial cells (orange). (b) Image of 3D printed OoC
device loaded with dyed gel mimicking Matrigel. (c) Stereomicroscope images of OoC device cross-section showing the 200 × 200 μm2 embedded
multi-layer stop valves. Scale bar = 500 μm. (d) 384-well plate configuration of the OoC device printed in one run. The inset shows red and green
dyed gel solutions confined by the capillary valve to their respective compartments. The yield was 100% across 3 plates with each 192 OoC
devices. (e) Maximal intensity projection of confocal fluorescent microscopy z-stack over ∼200 μm with 10 μm increments featuring an OoC
seeded with a five-day-old IMR-90 lung fibroblast spheroid (red) and a five-day-old MDA-MB-231 breast cancer spheroid (green), separated by a
vascular barrier channel with HUVEC-embedded gel (orange), imaged at day 0 and day 5. The 3D printed stop-valve features are outlined in white
dashed lines. Scale bar = 300 μm. (f) Mass manufacturing of OoC devices connected by breakable struts printed as an array of 19 × 18 × 10 = 3420
devices (X × Y × Z). Inset shows a zoomed-in view of stacked OoC devices. Scale bar = 5 mm.

2788 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024
 View Article Online

Lab on a Chip Paper

following reviewers' comments, which illustrates the potential users. The combination of low-cost, high-resolution
advantages and potential of 3D LCD printing (and PLInk) for 3D printers, and readily 3D printable designs enable the
rapidly exploring new ideas and concepts for microfluidic realization of low-cost and distributed digital manufacturing.
and OoC devices.
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

Future work could explore the concurrent variation of TPO Data availability
photoinitiator, ITX photoabsorber, and PETTA crosslinker
concentration to better understand their interplay with 3D design files are uploaded to Thingiverse and Printables
regards to Dp, Ec, and printing accuracy, and choose the ([Link] and
optimal mixture based on a specific application and criteria. [Link]
Photoactive components that better match LCD 3D printer Data not presented in the article or ESI† material will be
light spectrum, especially at higher wavelength between available upon request.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

∼420–450 nm could help improve photocuring efficiency and

reduce exposure time; however, increased light adsorption at Author contributions
higher wavelength is expected to come at the expense of
V. K. designed the ink formulation, V. K. and H. S.
yellow-orange tinted devices compared to visually overall
characterized the ink and assessed printability for LCD 3D
neutral and transparent PLInk. In addition, mapping the
printing. M. L. S. performed the biocompatibility assay. H.
light heterogeneity of the LCD 3D printers by the end user,
S., G. K., and Y. M. printed and characterized the
and the tools to do that, would open the door to digital
micromixer. G. K. and H. S. printed and characterized the
correction of the illumination heterogeneity by programming
microvalve. A. S.-K. contributed to the embedded sequential
the 3D printer, and further improve the resolution achievable
delivery chip. H. S. designed and optimized the ELISA-chips.
both with commercial and custom photoinks. Finally, low-
H. S. and M. L. S. designed and optimized the organ-on-a-
cost commercial inks that are easily accessible to the end-
chip devices. H. S. prepared the figures and analyzed the
user (e.g., water-washable inks), but which are often viscous
data. H. S., V. K., and D. J. prepared the initial draft of the
and suffer from overly large Dp for high-resolution
manuscript, with feedback from all the authors. H. S. and
microfluidic 3D printing, might be improved simply by
D. J. edited and reviewed the manuscript drafts. D. J.
supplementing them with additives. The addition of (i)
provided funding, and conceptualized, supervised, and
solvents such as PEGDA-250 to reduce the viscosity, (ii)
administered the project.
photoabsorbers to reduce the light penetration, and (iii)
photoiniators to reduce the exposure time could all be
explored. The biocompatibility of commercial inks needs to Conflicts of interest
be evaluated and possibly improved by testing different
The authors have no conflicts of interest to declare.
washing and leaching conditions towards removing residual
toxic components from the cured parts.
We may expect that 3D stereolithography printer
Acknowledgements
manufacturers driven by technological advances in LCDs (and We acknowledge Yongjun Xiao for assistance with μCT at the
light engines), and market pressure, will continue to increase Centre for Bone and Periodontal Research, Mohsen Ketabi
pixel numbers and concomitantly reduce pixel size, all while and Gwénaël Chamoulaud for assistance in profilometry,
preserving the affordability of LCD 3D printers, which will viscometry, and FTIR–ATR at the NanoQAM Research Center,
further increase their appeal and adoption. We foresee that and Lucie Riffard for assistance in tensile testing at the
some of the greatest opportunities lie in improving the McGill Structures and Composite Materials Laboratory. H. S.
photoinks for LCD (and more generally stereolithography) 3D acknowledges a CGS-M CIHR, an FRQNT master's research
printing, which are in their infancy. While here we showed the scholarship and a McGill BME recruitment award. V. K.
application of LCD 3D printing to microfluidics that were acknowledges an FRQNT doctoral research scholarship. G. K.
primarily designed based on prior manufacturing technologies, acknowledges an FRQNT doctoral research scholarship and a
opportunities arise to re-design and ideate microfluidic systems McGill BME recruitment award. Y. M. acknowledges a McGill
and OoC devices that leverage the strength of LCD and BME recruitment award. A. S.-K. acknowledges FRQNT
stereolithography 3D printing. postdoctoral fellowship. D. J. acknowledges support from a
Digital manufacturing by LCD 3D printing is as simple as Canada Research Chair in Bioengineering.
downloading a file and printing it, thus circumventing the
need for specialized machinery and advanced training, while References
enabling customizability and rapid design iterations by the
end user. The advent of low-cost and easy-to-use 3D printers 1 G. M. Whitesides, Nature, 2006, 442, 368–373.
compared to traditional manufacturing methods enables the 2 E. K. Sackmann, A. L. Fulton and D. J. Beebe, Nature,
fabrication of open and embedded microscopic features by 2014, 507, 181–189.
anyone, anywhere, thus democratizing access to high- 3 G. M. Whitesides and A. D. Stroock, Phys. Today, 2001, 54,
resolution fabrication and reducing the entry barrier for many 42–48.

This journal is © The Royal Society of Chemistry 2024 Lab Chip, 2024, 24, 2774–2790 | 2789
 View Article Online

Paper Lab on a Chip

4 G. Trotta, A. Volpe, A. Ancona and I. Fassi, J. Manuf. Process., 27 A. Vedhanayagam, M. Golfetto, J. L. Ram and A. S. Basu,
2018, 35, 107–117. Micromachines, 2023, 14, 1519.
5 D. J. Guckenberger, T. E. De Groot, A. M. Wan, D. J. Beebe 28 A. Hayer, L. Shao, M. Chung, L.-M. Joubert, H. W. Yang, F.-C.
and E. W. Young, Lab Chip, 2015, 15, 2364–2378. Tsai, A. Bisaria, E. Betzig and T. Meyer, Nat. Cell Biol.,
This article is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported Licence.

6 A. Naderi, N. Bhattacharjee and A. Folch, Annu. Rev. Biomed. 2016, 18, 1311–1323.
Eng., 2019, 21, 325–364. 29 M. Boaks, C. Roper, M. Viglione, K. Hooper, A. T. Woolley, K. A.
7 V. Karamzadeh, A. Sohrabi-Kashani, M. Shen and D. Christensen and G. P. Nordin, Micromachines, 2023, 14, 1589.
Juncker, Adv. Mater., 2023, 35, 2303867. 30 H. Gong, A. T. Woolley and G. P. Nordin, Lab Chip, 2016, 16,
8 B. Heidt, R. Rogosic, S. Bonni, J. Passariello-Jansen, D. 2450–2458.
Dimech, J. W. Lowdon, R. Arreguin-Campos, E. Steen 31 P. F. Jacobs, Rapid prototyping & manufacturing: fundamentals of
Redeker, K. Eersels, H. Diliën, B. van Grinsven and T. J. stereolithography, Society of Manufacturing Engineers, 1992.
Open Access Article. Published on 19 April 2024. Downloaded on 7/31/2024 [Link] PM.

Cleij, Phys. Status Solidi A, 2020, 217, 1900935. 32 Y. Xu, F. Qi, H. Mao, S. Li, Y. Zhu, J. Gong, L. Wang, N.
9 Z.-X. Low, Y. T. Chua, B. M. Ray, D. Mattia, I. S. Metcalfe and Malmstadt and Y. Chen, Nat. Commun., 2022, 13, 918.
D. A. Patterson, J. Membr. Sci., 2017, 523, 596–613. 33 R. Paul, Y. Zhao, D. Coster, X. Qin, K. Islam, Y. Wu and Y.
10 B. Carnero, C. Bao-Varela, A. I. Gómez-Varela, E. Álvarez and Liu, Nat. Commun., 2023, 14, 4520.
M. T. Flores-Arias, Mater. Sci. Eng., C, 2021, 129, 112388. 34 N. Rekowska, J. Huling, A. Brietzke, D. Arbeiter, T. Eickner,
11 A. P. Kuo, N. Bhattacharjee, Y. S. Lee, K. Castro, Y. T. Kim J. Konasch, A. Riess, R. Mau, H. Seitz and N. Grabow,
and A. Folch, Adv. Mater. Technol., 2019, 4, 1800395. Pharmaceutics, 2022, 14, 628.
12 H. Gong, B. P. Bickham, A. T. Woolley and G. P. Nordin, Lab 35 S. M. Montgomery, F. Demoly, K. Zhou and H. J. Qi, Adv.
Chip, 2017, 17, 2899–2909. Funct. Mater., 2023, 33, 2213252.
13 J. L. Saorin, M. D. Diaz-Alemán, J. De la Torre-Cantero, C. 36 S. M. Montgomery, C. M. Hamel, J. Skovran and H. J. Qi,
Meier and I. Pérez Conesa, Appl. Sci., 2021, 11, 3197. Extreme Mech. Lett., 2022, 53, 101714.
14 Y. T. Kim, A. Ahmadianyazdi and A. Folch, Nat. Protoc., 37 J. Qi, W. Li, W. Chu, J. Yu, M. Wu, Y. Liang, D. Yin, P. Wang,
2023, 18, 1243–1259. Z. Wang, M. Wang and Y. Cheng, Micromachines, 2020, 11,
15 C. Warr, J. C. Valdoz, B. P. Bickham, C. J. Knight, N. A. 213.
Franks, N. Chartrand, P. M. Van Ry, K. A. Christensen, G. P. 38 W. Ruijin, L. Beiqi, S. Dongdong and Z. Zefei, Sens.
Nordin and A. D. Cook, ACS Appl. Bio Mater., 2020, 3, Actuators, B, 2017, 249, 395–404.
2239–2244. 39 N. Li, C. H. Hsu and A. Folch, Electrophoresis, 2005, 26,
16 V. Karamzadeh, M. Shen, F. Lussier and D. Juncker, bioRxiv, 3758–3764.
2023, preprint, DOI: 10.1101/2023.11.27.568937. 40 B. Mosadegh, C.-H. Kuo, Y.-C. Tung, Y.-s. Torisawa, T.
17 A. Olanrewaju, M. Beaugrand, M. Yafia and D. Juncker, Lab Bersano-Begey, H. Tavana and S. Takayama, Nat. Phys.,
Chip, 2018, 18, 2323–2347. 2010, 6, 433–437.
18 M. Yafia, O. Ymbern, A. O. Olanrewaju, A. Parandakh, A. 41 Y.-S. Lee, N. Bhattacharjee and A. Folch, Lab Chip, 2018, 18,
Sohrabi Kashani, J. Renault, Z. Jin, G. Kim, A. Ng and D. 1207–1214.
Juncker, Nature, 2022, 605, 464–469. 42 G. Kak, M. Raza and B. K. Tiwari, Biomol. Concepts, 2018, 9,
19 A. Parandakh, O. Ymbern, W. Jogia, J. Renault, A. Ng and D. 64–79.
Juncker, Lab Chip, 2023, 23, 1547–1560. 43 M. Pai, C. M. Denkinger, S. V. Kik, M. X. Rangaka, A.
20 B. W. Caplins, C. I. Higgins, T. J. Kolibaba, U. Arp, C. C. Zwerling, O. Oxlade, J. Z. Metcalfe, A. Cattamanchi, D. W.
Miller, D. L. Poster, C. J. Zarobila, Y. Zong and J. P. Killgore, Dowdy and K. Dheda, Clin. Microbiol. Rev., 2014, 27,
Addit. Manuf., 2023, 62, 103381. 3–20.
21 P. Somers, A. Münchinger, S. Maruo, C. Moser, X. Xu and M. 44 M. Pai and M. Behr, Microbiol. Spectrum, 2016, 4(5), DOI:
Wegener, Nat. Rev. Phys., 2024, 6(2), 99–113. 10.1128/microbiolspec.TBTB2-0023-2016.
22 H. Gong, M. Beauchamp, S. Perry, A. T. Woolley and G. P. 45 W. Luo, M. Pla-Roca and D. Juncker, Anal. Chem., 2011, 83,
Nordin, RSC Adv., 2015, 5, 106621–106632. 5767–5774.
23 N. Guselnikov, P. Lazarev, M. Paukshto and P. Yeh, J. Soc. 46 G. Taguchi, S. Chowdhury and Y. Wu, Taguchi's quality
Inf. Disp., 2005, 13, 339–348. engineering handbook, 2004.
24 I. Ahmed, K. Sullivan and A. Priye, Biosensors, 2022, 12, 652. 47 D. Huh, B. D. Matthews, A. Mammoto, M. Montoya-Zavala,
25 V. G. Colin, T. Travers, D. Gindre, R. Barillé and M. H. Y. Hsin and D. E. Ingber, Science, 2010, 328, 1662–1668.
Loumaigne, Int. J. Adv. Manuf. Tech., 2021, 116, 537–544. 48 I. Salmon, S. Grebenyuk, A. R. Abdel Fattah, G. Rustandi, T.
26 M. G. Mohamed, H. Kumar, Z. Wang, N. Martin, B. Mills Pilkington, C. Verfaillie and A. Ranga, Lab Chip, 2022, 22,
and K. Kim, J. Manuf. Mater. Process., 2019, 3, 26. 1615–1629.

2790 | Lab Chip, 2024, 24, 2774–2790 This journal is © The Royal Society of Chemistry 2024