RESEARCH ARTICLE

www.advmattechnol.de

Porous Platform Inks for Fast and High-Resolution 3D

Printing of Stationary Phases for Downstream Processing
Mariachiara Conti, Jodie A. Symington, James R. Pullen, Rok Mravljak, Aleš Podgornik,
and Simone Dimartino*

1. Introduction
Purifications of biologics can be improved using 3D printed stationary phases
with perfectly ordered morphology. However, limited spatial resolution and Bioprocessing industry rapidly grows
lack of porous materials have hindered application of additive manufacturing due to the need of improving exist-
ing products and developing new thera-
in bioprocessing. To bridge this gap, digital light processing and
pies. Currently, chromatographic purifi-
polymerization-induced phase separation are combined to fabricate platform cations mostly rely on traditional porous
materials with bed morphology at the micrometer scale, and porous network beads that suffer from suboptimal perfor-
in the nanometer scale. Four different porous inks are developed, 3D printed, mance due to the random macroporous
and characterized in terms of their rheological behaviour, polymerization structure of the packed bed.[1,2] In ad-
kinetics, and printing resolution. Rapid 3D printing (down to 1 h) is achieved dition, pore diameters in the range of
10–100 nm pose a major challenge to
at scale (up to 100 mL column) of porous supports (50% porosity) at high
processing large new-generation drugs
resolution (up to 50 μm for linear features and 200 μm for complex such as antibodies, mRNA-based vac-
geometries). 3D-printed gyroids are chemically functionalized with various ion cines, and viral vectors, which are char-
exchange ligands. These are successfully challenged for i) the separation of acterized by a broad size distribution
model proteins in dynamic conditions and ii) protein capture from a clarified (1–400 nm).[3] Additionally, chromato-
graphic separations represent up to 70%
cell harvest, demonstrating dynamic binding capacities between 5 and
of biomanufacturing costs.[4] Therefore,
16 mg mL−1 and up to 86% purity in a single run. This work introduces a the bioprocessing industry has an ur-
rapid and facile approach to 3D printing porous inks to fabricate perfectly gent need for new cost-effective pro-
ordered stationary phases for downstream processing. duction platforms that are able to cope
with the demand for faster, economical,
and flexible production of therapeutics.
3D printing is a precise, repeatable,
and customizable fabrication technique
M. Conti, S. Dimartino that can create any type of microstructure with finely controlled
Institute for Bioengineering features. Channel size and bed morphology can be easily ma-
School of Engineering
The University of Edinburgh
nipulated, providing a potential avenue for the production of
Colin McLaurin Road, The King’s Building, Edinburgh EH9 3DW, UK flexible purification devices with bespoke ordered structures for
E-mail: [email protected] the purification of novel therapeutics, comprising large macro-
J. A. Symington, J. R. Pullen molecules, biological nanoassemblies, or even stem cells.[5]
Fujifilm Diosynth Biotechnologies Currently, there are two major challenges in 3D printing adsor-
Teesside TS23 1LH, UK bents for the biomanufacturing industry: i) the lack of printable
R. Mravljak, A. Podgornik materials with suitable composition for bioseparations and ii) the
Department of Chemical Engineering and Technical Safety
Faculty of Chemistry and Chemical Technology trade-off between 3D printing resolution, printing time and build
University of Ljubljana size. For example, high-resolution printers (<1 μm) require 1 to
Ljubljana SI-1000, Slovenia 3 weeks to fabricate a 660 nL column.[6]
To overcome these issues, our group has developed
methacrylate-based porous inks that could be 3D printed to
yield fully functional chromatographic columns in one step
The ORCID identification number(s) for the author(s) of this article with static protein binding capacities similar to traditional
can be found under https://doi.org/10.1002/admt.202300801
chromatography resins.[5] Polymerization-induced phase sepa-
© 2023 The Authors. Advanced Materials Technologies published by
Wiley-VCH GmbH. This is an open access article under the terms of the
ration enabled creation of a porous network with porosity in the
Creative Commons Attribution-NonCommercial-NoDerivs License, submicron scale. These 3D-printed stationary phases can also be
which permits use and distribution in any medium, provided the original used as expanded bed adsorbers due to their ability to capture
work is properly cited, the use is non-commercial and no modifications proteins and viruses directly from solid-laden feedstocks with
or adaptations are made. purification factors close to those of traditional resins and recov-
DOI: 10.1002/admt.202300801 eries above 80%.[5,7] However, direct printing of functionalized

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (1 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

Figure 1. Schematic illustration of fabrication workflow of 3D printed parts. a) Ink formulation with selection of appropriate components and concen-
trations. b) Computer-aided design (CAD) of geometrical features. c) Layer-by-layer DLP printing with polymerization-induced phase separation. d) 3D
printed objects (gyroidal column and British telephone booth) with e) column channels in the sub-mm scale as defined by the CAD file, and f) displaying
porous network at the sub-μm scale created by the pore-forming agents.

porous supports using polymerization-induced phase separation troduced thanks to the presence of the epoxy-based monomer gly-
requires development of a bespoke ink formulation for each cidyl methacrylate (GMA) in the ink formulation, making this a
chromatographic modality, including time-consuming profiling platform material to flexibly manufacture 3D printed stationary
of its 3D printing behavior (e.g., polymerization kinetics, light phases for a range of applications in downstream processing.[12]
scattering issues, choice of photoinitiators and photoadsorbers, Here, gyroidal columns were 3D printed and functionalized with
etc.). This technique has recently been employed to 3D print cation and anion exchange ligands and challenged for the separa-
porous structures using 3D microprinting and digital light tion of model proteins as well as protein capture from a clarified
processing (DLP).[8,9] Most recently, 3D printed hierarchically supernatant.
micro-nanoporous cubes with 300 μm wall thicknesses demon-
strated 20 times higher adsorption capacity than the nonporous 2. Results and Discussion
one.[10]
Traditional chromatographic supports are generally manufac- In DLP 3D printing, monomers are incorporated into a poly-
tured in a two-step process: i) generation of the stationary phase meric network in a layer-by-layer manner, following a computer-
under standardized manufacturing operations and ii) chemical aided design (CAD) model. The chemical reaction occurring
functionalization to meet a specific chromatographic modality. while printing is a light-initiated polymerization, which is prop-
This procedure has been recently employed also in cellulose agated thanks to the presence of a photoinitiator in the ink for-
stationary phases fabricated using a 3D printed wax template mulation. The schematic of the fabrication process can be ob-
and functionalized with hydroxyapatite and diethylaminoethyl served in Figure 1. GMA was selected as the key monomer for all
for the purification of adenovirus and lentivirus from clarified ink compositions formulated in this work (Figure 1a). GMA is
cell lysates.[11] a bifunctional monomer with a methacrylate group suitable for
Herein, we describe the development of novel ink formula- photopolymerization and containing an epoxide ring for simple
tions for porous stationary phases. Our approach differs from follow-up functionalization to suit a range of chromatographic
previous works by demonstrating the possibility of finely tun- modalities.[13–15] Ethylene glycol dimethylacrylate (EDMA), a two-
ing polymerization chemistry for fast fabrication of porous mor- arm cross-linker with proven biocompatibility,[12] and dipentaery-
phologies with customizable chemistry to introduce various thritol pentaacrylate (SR399), a multi cross-linker with five re-
functional ligands for targeted applications. In particular, the active groups, were introduced in the formulation to improve
ink composition was adjusted to reach a trade-off between spa- mechanical stability.[16] Omnicure 819 was added as a photoini-
tial resolution, printing time, and porous characteristics. The se- tiator due to its absorbance spectrum matching the emission
lected ink enables fast 3D printing (up to 100 mL column in spectrum of our DLP printer (385 nm), with 1.5% wt. concen-
1 h) of ordered structures with resolution at the sub-mm-scale tration selected after preliminary material profiling (Figure S1,
pushed to unprecedented levels of 50 μm resolution and super- Supporting Information). Omnistab 326 (0.125% wt.) was se-
imposed porous morphology at the sub-μm-scale (mean pore di- lected as a photoabsorber as it matched the refractive index of
ameter ≈800 nm). A variety of functional chemistries can be in- the ink formulations.[17] The surface area of the printed parts was

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (2 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

Table 1. Composition and curing time of 3D printable inks formulated in this work. Relative ratio for porogens and multi cross-linker was of 60%, with
the remaining 40% unchanged for all materials (24% GMA and 16% EDMA).

Ink name Porogens Multi-cross-linker Reactive sites Curing time [s]

[wt%] [wt%] [mmol g−1 ]
Exposure timea) Gelation timeb) Induction timec)

Nonporous ink 0% 60% 7.8 ≈2 ≈1 < 20

Porous ink 30% 30% 30% 5.6 ≈4 ≈2 < 20
Porous ink 50% 50% 10% 4.0 ≈9 ≈3 < 20
Porous ink 60% 60% 0% 3.3 ≈30 ≈29 < 60
a) b) c)
Exposure time was obtained from ink calibration experiments (Figure 2a); Gelation time was derived from photorheology experiments (Figure 2b); Induction time was
defined as the time for onset of double bond conversion (Figure 2c).

increased by including an inert porogenic phase (80% wt. cyclo- monoliths (≤ 4 mmol g−1 [23] ) or membranes such as Sartobind
hexanol and 20% wt. dodecanol) to create an internal porosity. epoxy (≈1 mmol g−1 [24] ). The super cross-linker SR399 and poro-
This porogen composition was selected based on well-established gens made up the remaining 60% wt., as illustrated in Table 1,
studies to maximize separation efficiency and hydrodynamic with progressively increasing porogens content (from 0 to 60%)
properties in methacrylate chromatography media.[16,18,19] and decreasing amounts of the multi cross-linker (from 60% to
To demonstrate the working principle of 3D printing porous 0%). The viscosity of the porous inks decreased from 20 mPa
structures (Figure 1b,c) we designed a British telephone booth s to 10 mPa s as the porogens concentration increased (Figure
4 mm wide and 1 cm tall with slender window linings of 100 μm S2, Supporting Information), remaining within the appropriate
× 100 μm running across an overhang of 4 mm, as well as range for DLP printing (below 600 mPa s). The upper limit of
porous gyroidal columns having 300 μm wall and channel width. porogens content was set equal to 60% in porous ink 60%, cor-
Figure 1d shows that the CAD models were successfully and re- responding to the top range porosity levels encountered in com-
liably 3D printed (in this case using porous ink 50% in Table 1). mercial stationary phases such as packed beds and monoliths (45
The mechanical properties of the material can be inferred from to 65%).[25–27]
the slenderness ratio (SR), defined for a linear structure as the
ratio between the overall length and the least radius of gyration
of its cross-section, as beyond a certain value of SR the element 2.1. Assessment of Rapid 3D Printing of Porous Inks
is prone to buckling and/or cracking.[20] In this case, the slender-
ness ratio of the window linings was 70. 3D printed models with For successful 3D printing via DLP, an ink must have sufficient
similar SR have been reported to withstand critical loads up to cure depth Cd to ensure adequate incorporation between subse-
60 MPa.[20] In addition, mechanical stability studies performed quent layers. On the other hand, a too-high Cd value leads to over-
on methacrylate-based inks showed Young moduli ranging from curing and long print times. In this study, all models were 3D
0.11 to 4.65 MPa[17] and in good agreement with chromatogra- printed with 50 μm resolution in the z-direction, so we assumed
phy materials (≈ 8 MPa for Toyopearl beads[21] and between 6 50 μm as minimum threshold for Cd . Characterization of the cur-
to 22 MPa for CIM monoliths[22] ). Therefore, it is reasonable to ing behavior of all inks was carried out by irradiation of indepen-
expect that our porous 3D printed material presents similar me- dent discs at different exposure times (Figure 2a). A Cd of 50 μm
chanical properties as other 3D printing materials and is in line was measured after 30 s of UV light exposure when no multi-
with commercial chromatography materials such as agarose and functional cross-linkers were employed (porous ink 60%). This
methacrylates. Overall, the workflow shows that the combina- is well outside the acceptable range for fast DLP printing.[28] The
tion of DLP printing with polymerization-induced phase separa- addition of the porogens to the porous ink dilutes the amount
tion enables fabrication of large 3D models with sub-mm flow of methacrylate sites for the polymerization to propagate, with
through channels (Figure 1e) and superimposed nanoporosity porous ink 60% having only ≈3 mmol of reactive groups per
(Figure 1f). g of formulation (Table 1). Consequently, there is the need for
Polymerization-induced phase separation generates an inher- longer exposure times to generate the same amount of radicals
ent porous structure with significantly increased surface area for required in the photocuring process, with severe limitations on
biomolecule adsorption.[10] However, porogens introduction de- printing resolution. The introduction of cross-linker as small as
creases overall printability as it requires longer curing times, in 10% shortened significantly the exposure times needed to cure
turn increasing light scattering and lowering printing resolution. 50 μm by >70%, from 30 s for porous ink 60% down to 9 s for
This generally causes 3D-printed objects to deviate from the in- porous ink 50% (Figure 2a). Further reduction to 4 s exposure
tended CAD model.[17] Our approach to curbing print times and time was observed when the porogen content was decreased to
light scattering was to empirically tune the kinetics of the poly- 30%, while the nonporous ink, with the highest concentration of
merization by adjusting the relative proportions of the multi- reactive sites, was the fastest to polymerize, requiring ≈2 s to cure
cross-linker and of the porogenic mixture. The concentration of 50 μm of material.
GMA was chosen to obtain 3D printed columns with a theoretical The photopolymerization kinetics of the DLP inks was inferred
density of epoxy groups ranging from 2 to 9 mmol g−1 (0 and 60% in photorheology experiments, where the evolution of the IR fin-
respectively), in the range or above commercially available CIM gerprint, as well as of the elastic (G’) and loss (G’’) moduli during

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (3 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

Figure 2. Inks characterization. a) Cure depth as a function of cure time. Error bars are standard deviation of three independent experiments (N = 3).
b) Real-time photorheology under constant shear frequency of 10 Hz and constant shear rate of 1%; black dashed line at 60 s indicates when the UV
light is switched on. c) Trend of double bond conversion, XDB , during the course of the polymerization reaction. d) Logarithmic increase of the elastic
and loss moduli and final conversion of double bond for the four inks at different porogen content.

UV curing was captured in real-time (Figure 2b). The crossover cous moduli (indicative of the relative change in the rheological
point between the elastic and loss moduli rheologically defines properties before and after curing) of the four inks investigated.
the point of gelation of a material. The exposure times to achieve The trends obtained by the two independent experimental meth-
the gelation point for the tested inks are in good agreement with ods (rheology and Fourier Transfom Infrared, FTIR) are very well
the values obtained from the working curves (Table 1). Photorhe- aligned.
ology confirms that the introduction of porogens to the ink com- Screening of the four inks informed selection of porous ink
position produces a delay in the photopolymerization reaction 50% as the best material for further characterization and testing
(Figure 2b). in chromatographic applications, as it offered i) relatively high
The stretching band of the double bond at 1632 cm−1 was porogen content (50%) to maximize surface area, ii) platform
monitored during photoreology (Figure S3, Supporting Informa- chemistry with 6 mmol g−1 theoretical epoxy density, around
tion), and the double bond conversion was calculated according or above commercially available CIM discs (≤4 mmol g−1[23] ) or
to Equation 1 (Figure 2c). Porous ink 60% exhibited a sigmoidal membranes such as Sartobind epoxy (≈1 mmol g−1[24] ), and iii)
conversion curve with time, with an induction period of ≈60 s fast printing times of <30 min for a 100 mL column having 4 cm
(Table 1), and unsatisfactory conversion below 25% was reached diameter and 3 cm bed height. This presents a particularly ex-
only after >100 s exposure to UV light. The other three inks in- citing potential as it overcomes limitations of other 3D printing
stead showed much faster polymerization kinetics, with induc- methods having higher resolution such as two-photon polymer-
tion times below the instrument acquisition rate of 20 s (Table 1), ization, requiring ≈300–500 h to 3D print a 75 μm i.d. porous
and a steep rise of the conversion levels to achieve a plateau of column (660 nL column volume, CV).[6]
conversion in ≈40 s exposure to UV light. In particular, porous
ink 30% reached the highest double bond conversion of ≈50%,
followed by the non-porous ink and porous ink 50% displaying a 2.2. High-Resolution 3D Printing at the sub-mm Scale
similar conversion slightly above 40% in line with methacrylate-
based inks (from 20 to 60%[29,30] ). Figure 2d summarizes double To investigate the interrelation between exposure time, light scat-
bond conversion and relative log increase of the elastic and vis- tering, and printing resolution of the porous inks, resolution

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (4 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

Figure 3. a,b) CAD renderings (top row) compared to 3D printed models using nonporous ink (middle row) and porous ink 50% (bottom row). Ge-
ometries tested are resolution cube and Schoen’s gyroids with 300, 200, and 150 μm walls.

cubes (10 × 10 × 10 mm3 ) with decreasing wall thicknesses from ements, overhangs, intertwined channels, and rotational symme-
1000 to 50 μm in all directions and fixed windows thickness try. Schoen gyroids with wall thicknesses of 300, 200, and 150 μm
(1 mm) were designed and 3D printed using porous ink 50% and were designed and 3D printed (Figure 3b). Schoen gyroids fabri-
the nonporous ink as reference (Figure 3a). The nonporous ink cated with porous ink 50% presented a similar trend as for the
resulted in a transparent model, while a white translucent object results observed with the resolution cube, with negligible (3%)
was obtained as porogens were introduced to the ink composi- deviation between designed and measured wall thicknesses for
tion due to reduced transmission and scattering of visible light
on the irregular interfaces of the porous networks.[8,10,31] Both
inks demonstrated satisfactory print performance, with all walls
successfully 3D printed even at the designed 50 μm, a result of
particular interest given the relatively large amount of porogens
(50%). The fidelity of the 3D printed objects with respect to the
CAD models was determined by comparing the measured and
designed dimensions of the wall thicknesses (Figure 4). Excel-
lent dimensional accuracy was observed for the nonporous ink,
with <10% inaccuracy between measured and designed thick-
ness for studs of 75 μm or larger, while a certain degree of over-
curing was observed for the 50 μm (70% deviation), indicative
of the challenge of 3D printing features at the nominal resolu-
tion limit of the DLP printer used in this work. Porous ink 50%
showed good dimensional accuracy with deviations below 20%
for wall thicknesses between 1000 and 200 μm, and becoming
more prominent for wall thicknesses below 200 μm. This result
is associated with light scattering due to the pores formed dur-
ing polymerization-induced phase separation, exacerbated as the
feature size is reduced, leading to overpolymerization. The qual-
itative red line in Figure 4 is a visual aid to identify the limit of
3D printing features, with an asymptote at ≈150 μm indicative of
Figure 4. Comparison between theoretical (from CAD model) and mea-
the limit of printing accuracy for porous ink 50%. sured (from 3D printed model) wall thickness of resolution cubes (empty
The non-porous and porous ink 50% were further tested to squares) and 300, 200, and 150 μm Schoen gyroids (full circles) for non-
fabricate models with challenging features such as non-linear el- porous ink (black data) and porous ink 50% (red data).

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (5 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

the 300 μm gyroid and a reasonably low deviation of 20% for the carboxyl (IDA) groups. Elemental analysis (Table 2) and FTIR
200 μm gyroid (Figure 4, full circle). The 150 μm limit observed (Figure S5, Supporting Information) confirmed successful lig-
for the 3D printing of the resolution cube with porous ink 50% and immobilization to the 3D printed material for all ligand
was confirmed also for the 3D printing of the complex gyroidal chemistries investigated, with ligand densities in good agree-
models, with deviation between model and measured thickness ment with those published for other chromatographic GMA-
of up to 80%, and a number of blocked channels due to overpoly- based materials such as CIM discs and fibers[23,28,37,38]
merization. Overall, the flexibility of functionalization chemistry here
demonstrated indicates that the epoxy porous ink 50% can be
used as a platform material to suit a range of bioseparation re-
2.3. Porous Morphology at the sub-μm Scale
quirements and chemical modalities.
The porosity, pore size distribution, and permeability of the
porous network within the 3D printed phase, i.e., at the sub-μm
2.5. Porous Gyroidal Columns for Protein Separation and Capture
scale, were characterized through a range of experimental meth-
ods using discs (0.34 mL CV) 3D printed with porous ink 50%.
Gyroids with 300 μm wall thickness were 3D printed and tested
Scanning Electron Microscopy (SEM) imaging shows the char-
for the following chromatographic functionalization and charac-
acteristic globular porous structure of methacrylate monoliths
terization, safely above the limit for complex geometries observed
(Figure 1f), with globules forming an intricate network of pores,
for porous ink 50%.
both in the range of 1 μm (Figure S4a,b, Supporting Informa-
The 3D-printed columns were first tested with mixtures of
tion).
model proteins and then with an industrial complex feedstock.
The wet/dry mass method and the moment analysis returned
For these tests, we selected the QA and IDA functionalizations to
porosities of 60 ± 2% and 58 ± 4%, respectively, above the theo-
span across weak/strong and anion/cation exchange modalities.
retical 50% porogen phase in the ink. Similar results were con-
A QA gyroidal column was tested with a protein mixture
sistently recorded for other porous inks, with an apparent in-
of conalbumin (isoelectric point, pI, 6.24) and trypsin (pI 4.5).
crease in porosity due to a combination of molecular rearrange-
The observed elution for two subsequent runs is displayed in
ment during polymerization-induced phase separation,[32] caus-
Figure 5a. The elution step shows a first peak corresponding
ing a volume reduction in the polymeric backbone with respect
to conalbumin and a second peak for trypsin, as expected from
to the porogenic phase.[9]
the pIs of the two model proteins, with chromatographic res-
Data from pulse injection experiments (Figure S4d, Support-
olution of 1.2 ± 0.1. The binding capacity in dynamic condi-
ing Information) were also used to estimate the height equivalent
tions was 5.21 ± 1 mg mL−1 for conalbumin and 16.1 ± 0.1 mg
to a theoretical plate (HETP) associated with the porous network,
mL−1 for trypsin. The IDA gyroidal column was tested with a
resulting in an HETP of ≈150 μm, in the same order of magni-
protein mixture of cytochrome-C (pI 10) and lysozyme (pI 11)
tude as other 3D printed chromatography media.[17,33]
over three subsequent runs (Figure 5b). Proteins were eluted in
The permeability of 3D printed discs ranged between 2 and
the order predicted from their pIs, with first peak correspond-
25 × 10−15 m2 (Figure S4c, Supporting Information). These val-
ing to elution of less bound cytochrome-C, followed by the sec-
ues are within the range of commercial methacrylate monoliths
ond peak of the more strongly bound lysozyme. The runs had
(6.3 × 10−15 m2[34] ) as well as other 3D printed chromatography
dynamic binding capacity of 16 ± 2.1 mg mL−1 and 5 ± 0.8 mg
stationary phases previously published by our group (5.5 × 10−15
mL−1 for cytochrome-C and lysozyme, respectively. Results of
± 3.2 × 10−16 m2 [17] ).
protein separations here described demonstrate the material’s
ability to operate as stationary phase with good dynamic bind-
2.4. Surface Modification of Platform Material ing capacities, above previous results on 3D printed columns by
our group (8.0 ± 0.2 mg mL−1 )[17] and in line with other convec-
A key requirement for 3D-printed stationary phases is its ac- tive media such as monoliths (5–20 mg mL−1[22,39] ) or nanofibers
tive chemistry to enable functionalization for biomolecule sepa- and Sartobind membranes (5–7 mg mL−1 [40–42] ) with same
ration. The presence of the epoxy chemistry in models 3D printed functionalities.
with porous ink 50% was checked via comparison of its FTIR fin- Furthermore, QA gyroidal column was applied as a capture
gerprint with that of pure GMA and a model 3D printed with step in purification of DARPin off7 from a clarified supernatant.
a control ink that did not include GMA (Figure S5, Support- Successful isolation of DARPin off7 was achieved, although the
ing Information). In particular, the 3D printed model retains the elution peak was not well resolved due to the complex mixture of
characteristic vibration band of epoxy groups at 905 cm−1 after different interacting species (Figure 5c). This is a common occur-
photopolymerization,[35,36] which is instead lacking in the con- rence and can be attributed to mass transfer limitations through
trol model 3D printed with the ink not containing GMA. The the relatively thick 300 μm walls of the gyroidal structure. De-
conserved epoxy rings enable derivatization of 3D-printed porous spite this challenge, the column was able to produce high-purity
structures with a plethora of specific functional groups using the DARPin off7 of up to 86% in one single chromatographic step
broad range of ligand attachment techniques available for GMA (Figure 5d). In addition, future optimization of the porous ink,
and already in use in chromatography.[12] 3D printing settings, and the functionalization protocol have the
In this work, epoxy rings were converted into (strong and promise to increase binding capacity to levels comparable with
weak) anion and cation exchangers through introduction of qua- conventional porous beads (50–100 mg mL−1[43] with character-
ternary amine (QA) tertiary amine (TA), sulfonic acid (SP) and istics dimensions of 40 – 80 μm). For example, membranes with

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (6 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

Table 2. Summary of properties of models 3D printed with porous ink 50% and functionalized with different ligand chemistries. Ligand densities obtained
using CHNS elemental analysis.

Functionality Ligand Ligand density [mmol g−1 ]

This work Literature

Non functionalized Epoxy 6a) 4.2[ 23 ]

Strong anion exchanger QA 1.54 0.1–0.2[ 38]
Weak anion exchanger TA 0.69 0.1[ 35 ]
Strong cation exchanger SP 0.13 0.1–0.5[ 38]
Weak cation exchanger IDA 0.18 0.2–0.6[ 37]
a)
Theoretical ligand density based on the amount of GMA introduced into the ink formulation.[ 38 ]

Figure 5. Chromatograms of Schoen gyroidal columns (CV = 1.96 mL, 300 μm thick walls and channels) 3D printed using porous ink 50%. a) Elution
step resulting from 500 μL injection of 2 mg mL−1 conalbumin (5.7 ± 0.2 mL, 9.8 ± 0.1 mS cm−1 ) and of 4 mg mL−1 trypsin (15.1 ± 0.1 mL, 21.9 ±
0.1 mS cm−1 ) on QA functionalized columns. b) Elution step resulting from 500 μL injection of 4 mg mL−1 cytochrome-C (5.3 ± 0.02 mL, 7 ± 0.1 mS
cm−1 ) and 1 mg mL−1 lysozyme (17.3± 0.13 mL, 20.05 ± 0.7 mS cm−1 ) on IDA functionalized columns. c) Elution step resulting from the capture of
DARPin off7 from clarified supernatant. d) Purity and CE-SDS results of eluted fractions; first lane is molecular marker, with following lanes corresponding
to elution fractions from 13 to 20 CV.

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (7 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

improved grafting procedures exhibited up to five times the bind- added as photoabsorber and photoinitiator. Calibration of printing condi-
ing capacity increase for standard proteins.[37] tions was carried out by irradiating the inks at different exposure times
Overall, the advances in ink development presented in this and measuring the resulting curing depth.[5]
Rheological Behavior: The shear stress was measured at 25 °C with a
work enabled improvement in the printing resolution, from
HAAKE MARS Rheometer (Thermo Fisher Scientific, Hemel Hempstead,
500 μm down to 300 μm, and led to an increase of the dynamic UK) with a 25 mm diameter parallel plate mode in a rotation shear ramp
binding capacity, from ≈8 mg mL−1 to 16 mg mL−1 .[7] This shows test, with initial shear at a constant rate of 1 s−1 for 40 s, followed by a
that there is an opportunity for improvement in 3D printing in logarithmic gradient of the shear rate between 0.01 and 300 s−1 lasting
material science alongside 3D printing technology. 160 s. Photoreology was measured with a double plate configuration, with
a 1 mm gap under a constant shear frequency of 10 rad s−1 and a strain am-
plitude of 1% in the linear viscoelastic region. A UV transparent quartz top
plate allowed irradiation of UV light at 385 nm using an Omnilamp S1500
3. Conclusion (Jenton International, Whitchurch, UK). The light onset occurred after
60 s to allow for system equilibration. Loss and storage moduli, G’ and
This study aimed at developing a platform material suitable for G’’ respectively, were measured as a function of exposure time. Gel point
different chromatographic modalities that could be 3D printed at was derived from the intersection of G’ and G’’ and tan 𝛿.[44] Conversion
high resolution to produce pilot-scale columns in a few hours. of the double bond in the methacrylate monomers during polymerization,
Commercial deployment of such 3D-printed stationary phases is XDB , was calculated from the peak areas measured in the IR fingerprints
conditional to meeting critical constraints of build size, printing (Figure S3, Supporting Information) recorded during photoreology mea-
surements according to equation 1:
resolution, and time, but their interrelation is such that optimiza-
tion of one goes to the detriment of another. Therefore, a compro-
AtDB / Atref
mise among these criteria needs to be found. XDB = 1 − (1)
The work presented demonstrates that ordered hierarchically A0DB A0ref
porous stationary phases can be fabricated via DLP printing
and polymerization-induced phase separation. The 3D printed where superscripts 0 and t indicate the uncured ink and the ink at differ-
columns were designed with sub-mm scale features to create flow ent polymerization times, respectively, while subscripts DB and ref corre-
through channels (as dictated by the CAD model), with super- spond to the peak for the double bond (1635 cm−1 ) and a reference peak,
respectively.[45–47] The vibrational band at 1670–1780 cm−1 for the car-
imposed sub-𝜇m scale porosities generated during printing of bonyl group was selected as a reference as it remains unchanged during
the porous inks (to increase surface area for molecule adsorp- polymerization.
tion). The reactivity of the porous inks was tailored by adjusting Model Design and 3D Printing: All CAD models were created on
the composition leading to stationary phases 3D printed rapidly NetFabb Premium 2021 (Autodesk, San Rafael, CA, USA). All objects
at high resolution with reliable features down to 200 μm. We were printed with a layer thickness of 50 μm using a Soflex 350 DLP
demonstrated that the selected porous ink is a platform mate- printer (W2P, Vienna, Austria) with a 385 nm UV projector (intensity of
8 mW cm−2 ). Post-printing, unpolymerized material was washed out from
rial that can be flexibly tuned to fabricate ordered supports hav-
the porous 3D-printed object using organic solvents.
ing customizable morphologies, and with a range of chemical A UK telephone booth (4.6 mm × 4.6 mm × 13 mm) with 100 μm ×
functionalizations to suit bespoke applications such as liquid 100 μm × 4 mm window studs was designed to determine the ability to
chromatography. Porous gyroidal columns achieved separation fabricate porous structure from the μm- to cm-scale. After printing the
of model mixtures of proteins exhibiting dynamic capacities up UK telephone booth was dyed red with a commercially available food col-
to 16 mg m mL−1 , and successful protein capture from a com- orant to impart its iconic red colour. Resolution cubes (1 × 1 × 1 cm3 )
with decreasing linear struts from 1 mm to 50 μm in all (x, y, z) directions
plex supernatant with up to 86% purity. This is in the range of
(Figure 3a) were 3D printed to calibrate the resolution of the inks to gener-
commercial CIM monoliths and membranes.[40,41] ate simple linear geometries. Discs (12 mm diameter, 3 mm height) were
The advances shown here are a considerable step forward for 3D printed using 50% porous ink and inserted into a monolithic housing
the effective employment of 3D printed columns, and they must (Sartorius BIA Separations, Ajdovščina, Slovenia) to determine the per-
be coupled with a much-needed development of 3D printing tech- meability and internal porosity of the porous structures. Schoen gyroidal
nologies to further narrow down the resolution/speed/build size columns were designed to evaluate the printability of complex (non-linear)
gap, with special considerations of additive manufacturing of structures, as well as to test chromatographic separations. CAD models of
Schoen gyroids (50% external porosity) with equal sizes of walls and chan-
porous materials. In depth analysis of the design and impact of nels of 300 and 200 and 150 μm were generated and replicated in all direc-
the porous morphology resulting from a combination of the 3D tions, followed by an intersection with a cylindrical element (Figure 3b).
printing process and the ink formulation, also needs appropriate Gyroidal discs with 6.2 mm diameter and 5 mm height were 3D printed
investigation, so that the flexibility of this platform material can for assessment of ink printability. Chromatography experiments were per-
be fully harnessed in other fields than bioprocessing, such as an- formed using Schoen gyroidal column (300 μm walls, 50% external poros-
alytical HPLC and a range of other processes where solid-fluid ity) with 10 mm outer diameter and 25 mm height. A 400 μm outer wall
was designed around the gyroid for easier handling and adherence to the
contacting is required, e.g. extraction and adsorption processes.
column housing (10 mm internal diameter (i.d.), Astrea Bioseparation,
Cambridge, UK).
Characterization of Printing Fidelity and Porous Morphology: Printing fi-
4. Experimental Section delity at the sub-mm scale was assessed by comparing the CAD model and
measuring thicknesses of each wall of the resolution cubes and Schoen gy-
Ink Formulations: All formulations included 24% wt. GMA and roids using ImageJ (U.S. Institutes for Health, Bethesda, Maryland, USA).
16% wt. EDMA.[11] The remaining 60% contained different relative The internal porosity of 3D printed discs was obtained independently by
amounts of porogens and multi cross-linker (Table 1). A mixture of 80% wt. measuring the dimensions and masses of dry and wet discs, as well as
cyclohexanol and 20% wt. dodecanol were used as a porogenic agent. In all with the moment analysis by injecting 30 μL of 5% acetone in 1 M NaCl
compositions, 0.125% wt. Omnistab 326 and 1.5% wt. Omnirad 819 were at 1 mL min−1 and comparing the first moment with the one obtained for

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (8 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

the empty column housing. The HETP was calculated as defined by Carta School of Geosciences, the University of Edinburgh, for help with data ac-
et al.[26] Permeability was determined by measuring the pressure drops quisition for CHN analysis. Mariachiara Conti acknowledges the support
across 3D printed discs at different water flow rates (15 – 265 cm h−1 ) from Fujifilm Diosynth Biotechnologies UK, the Scottish Research Part-
and applying Darcy’s law. The pore size distribution of 3D printed objects nership in Engineering (SRPe-IDP/03), and the IBioIC “Ready for Indus-
was observed through SEM (S-4700, Hitachi, Tokyo, Japan) by measur- try” Ph.D. training program, funded by the Scottish Funding Council under
ing the diameter of 200 pores in SEM images from two different printing their Innovation Centre program. The authors also gratefully acknowledge
batches using ImageJ. Printed samples were dried in hexamethyldisilazane financial support from the Slovenian Research and Innovation Agency (Re-
and gold coated before imaging.[5] search Core Funding No. P1-0153 as well as project J7-2603). The authors
Chemical Functionalization and Characterization: The 3D printed ob- would like to thank Arkema Sartomer for their continuous support and for
jects were functionalized into QA, TA, SP, and carboxyl groups according providing samples.
to the literature.[38] The reactions in aqueous solutions were performed
with 100 mM trimethylammonium chloride (50% v/v) pH 11 at 8 °C for 4
days for QA, with 1.5 M diethylaminoethyl at room temperature for 3 days Conflict of Interest
for TA, with 1 M Na2 SO3 at 60 °C for 10 h and with 2 M iminodiacetic
acid pH of 12 at 60 °C for 3 days for IDA. The 3D printed columns were The authors declare no conflict of interest.
then washed extensively for three days with distilled water until neutral pH,
followed by hydrolysis of residual epoxy groups in 0.1 M H2 SO4 at 60 °C
overnight.[48] Data Availability Statement
Introduction of the functional chemistry on the surface of the 3D
printed models was characterized by attenuated total reflectance FTIR us- The data that support the findings of this study are available from the cor-
ing a Nicolet iS50 spectrometer with a smart OMNI-Sampler accessory responding author upon reasonable request.
(Thermo-Nicolet FTIR, Thermo Fisher Scientific, Waltham, MA).
The amounts of nitrogen and sulfur content for all functionalizations
were determined via CHNS analysis using a Thermo Fisher Scientific Flash Keywords
SMART 2000 Elemental Analyzer.
Chromatographic Characterization: All chromatographic characteriza- 3D printing, downstream processing, photopolymerization, porous inks,
tion was carried out in dynamic conditions using 3D-printed porous gy- protein captures
roidal columns (CAD model as described above). Protein separation ex-
periments were carried out on an AKTApure 25 (Cytiva, Amersham, UK). Received: May 19, 2023
A solution containing 4 mg mL−1 cytochrome C and 1 mg mL−1 lysozyme Revised: June 21, 2023
was loaded to the IDA cation exchanger column (CV = 1.96 mL), while a Published online: July 27, 2023
mixture of 4 mg mL−1 trypsin and 2 mg mL−1 conalbumin was injected to
the QA anion exchanger (CV = 1.96 mL). Unbound proteins were washed
out with 5 CV binding buffers (20 mM phosphate buffer pH 7.4) followed
by a post-load wash at 5% elution buffer (20 mM phosphate buffer pH 7.4,
[1] F. Gritti, M. F. Wahab, MJH Life Sciences 2018.
1 M NaCl). Proteins were recovered in gradient elution from 0% to 50%
over 15 CV for the anion exchanger, and at 10% isocratic over 3 CV fol- [2] A. E. Reising, S. Schlabach, V. Baranau, D. Stoeckel, U. Tallarek, J.
lowed by a gradient from 10% to 60% over 6 CV for the cation exchanger. Chromatogr. A 2017, 1513, 172.
Subsequently, a regeneration step consisting of 3 CV of 0.5 M NaOH hold [3] Viral Vectors for Gene Therapy: Methods and Protocols, (Eds.: O.-W.
for 30 min was performed. All runs were performed at 1 mL min−1 flow rate Merten, M. Al-Rubeai), Humana Press, Totowa, NJ, 2011.
for every step (≈2 min residence time) except for sample loading in which [4] C. De Luca, G. Lievore, D. Bozza, A. Buratti, A. Cavazzini, A. Ricci, M.
flow rate was 0.25 mL min−1 (≈8 min residence time). Protein capture Macis, W. Cabri, S. Felletti, M. Catani, Molecules 2021, 26, 4688.
was tested using an AKTAavant 25 (Cytiva, Amersham, UK) by injecting [5] C. Salmean, S. Dimartino, Chromatographia 2019, 82, 443.
the clarified supernatant containing the DARPin off7 protein onto a QA 3D [6] F. Matheuse, K. Vanmol, J. Van Erps, W. De Malsche, H. Ottevaere,
printed monolith and using 50 mM sodium phosphate pH 7.5 as equili- G. Desmet, J. Chromatogr. A 2022, 1663, 462763.
bration buffer (buffer A) and 50 mM sodium phosphate, 1 M sodium chlo- [7] U. Simon, L. C. T. Scorza, S. Teworte, A. J. McCormick, S. Dimartino,
ride, pH 7.5 as elution buffer (buffer B). In particular, 3D printed columns J. Sep. Sci. 2021, 44, 1078.
were charged with 3 CV buffer B and then equilibrated with 8 CV of buffer [8] F. Mayer, D. Ryklin, I. Wacker, R. Curticean, M. Čalkovský, A.
A before loading 1.4 mL of DARPin off7 via a 2 mL capillary loop. The Niemeyer, Z. Dong, P. A. Levkin, D. Gerthsen, R. R. Schröder, M.
columns were then washed with 3 CV of buffer A, before applying a lin- Wegener, Adv. Mater. 2020, 32, 2002044.
ear gradient from 0 to 100% buffer A to buffer B over 16 CV. The flow rate [9] D. G. Moore, L. Barbera, K. Masania, A. R. Studart, Nat. Mater. 2020,
during load was 0.49 mL min−1 (4 min residence time). All buffers were 19, 212.
degassed and prefiltered through a 0.2 μm sterile membrane using a vac-
[10] Z. Dong, H. Cui, H. Zhang, F. Wang, X. Zhan, F. Mayer, B. Nestler,
uum pump. Eluted fractions were analyzed using capillary electrophoresis
M. Wegener, P. A. Levkin, Nat. Commun. 2021, 12, 247.
gel (CE-SDS) on a LabChip GX Touch HT (PerkinElmer, Waltham, Mas-
[11] M. G. Moleirinho, S. Feast, A. S. Moreira, R. J. S. Silva, P. M. Alves, M.
sachusetts).
J. T. Carrondo, T. Huber, C. Fee, C. Peixoto, Sep. Purif. Technol. 2021,
254, 117681.
Supporting Information [12] E. G. Vlakh, T. B. Tennikova, J. Sep. Sci. 2007, 30, 2801.
[13] F. Svec, J. M. J. Fréchet, Biotechnol. Bioeng. 1995, 48, 476.
Supporting Information is available from the Wiley Online Library or from [14] D. J. Martina Vodopivec, Horst Schwinn, 1998.
the author. [15] F. Švec, J. Hradil, J. Čoupek, J. Kálal, Angew. Makromol. Chem. 1975,
48, 135.
[16] F. Svec, J. M. J. Frechet, Chem. Mater. 1995, 7, 707.
Acknowledgements [17] U. Simon, 3D Print Your Own Chromatography Column: Material De-
The authors would like to acknowledge Dr. Thomas Johnson at the Depart- velopment and Characterisation of Ion Exchange Monoliths for Pro-
ment of Biochemical Engineering, University College London, for provid- tein Purification, Doctoral dissertation, The University of Edinburgh,
ing the computed tomography scan in Figure 1e, and Joe Casillo from the 2021.

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (9 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmattechnol.de

[18] T. B. Tennikova, M. Náhůnek, F. Švec, J Liq Chromatogr 1991, 14, 2621. [35] H. Liu, P. V. Gurgel, R. G. Carbonell, J. Membr. Sci. 2015, 493,
[19] T. B. Tennikova, F. Svec, Hydrophobic Interaction and Reversed-Phase 349.
Modes, 1993, 646, 279. [36] A. Bruchet, V. Dugas, I. Laszak, C. Mariet, F. Goutelard, J. Randon, J.
[20] R. Torre, S. Brischetto, I. R. Dipietro, Addit. Manuf. 2021, 38, 101790. Biomed. Nanotechnol. 2011, 7, 415.
[21] E. Müller, Chem. Eng. Technol. 2005, 28, 1295. [37] S. M. Lemma, C. Boi, R. G. Carbonell, Membranes 2021, 11, 181.
[22] A. Podgornik, A. Savnik, J. Jančar, N. L. Krajnc, J. Chromatogr. A 2014, [38] R. Mravljak, O. Bizjak, P. Krajnc, M. Paljevac, A. Podgornik, J. Chro-
1333, 9. matogr. A 2021, 1652, 462077.
[23] F. Svec, T. B. Tennikova, Z. Deyl, Monolithic Materials, 1st Ed., Vol. 67, [39] S. Jerenec, M. Šimić, A. Savnik, A. Podgornik, M. Kolar, M. Turnšek,
Elsevier, Amsterdam 2003. P. Krajnc, React. Funct. Polym. 2014, 78, 32.
[24] C. Boi, S. Dimartino, S. Hofer, J. Horak, S. Williams, G. C. Sarti, W. [40] O. Hardick, S. Dods, B. Stevens, D. G. Bracewell, Biotechnol. Bioeng.
Lindner, J Chromatogr 2011, 879, 1633. 2013, 110, 1119.
[25] R. Hahn, M. Panzer, E. Hansen, J. Mollerup, A. Jungbauer, Sep. Sci. [41] S. Yamamoto, A. Kita, Food Bioprod. Process. 2006, 84, 72.
Technol. 2002, 37, 1545. [42] A. Podgornik, V. Smrekar, P. Krajnc, A. Štrancar, J. Chromatogr. A
[26] G. Carta, A. Jungbauer, Protein Chromatography: Process Development 2013, 1272, 50.
and Scale-Up, Wiley-VCH, Weinheim, 2010. [43] Preparative Chromatography, (Eds.: H. Schmidt-Traub, M. Schulte, A.
[27] I. Mihelič, T. Koloini, A. Podgornik, A. Štrancar, J High Resolut Chro- Seidel-Morgenstern), 1st Ed., Wiley, Hoboken, 2020.
matogr 2000, 23, 39. [44] B.-S. Chiou, S. R. Raghavan, S. A. Khan, Macromolecules 2001, 34,
[28] M. Caprioli, I. Roppolo, A. Chiappone, L. Larush, C. F. Pirri, S. 4526.
Magdassi, Nat. Commun. 2021, 12, 2462. [45] A. Champion, B. Metral, A.-S. Schuller, C. Croutxé-Barghorn, C. Ley,
[29] B. Steyrer, B. Busetti, G. Harakály, R. Liska, J. Stampfl, Addit. Manuf. L. Halbardier, X. Allonas, ChemPhotoChem 2021, 5, 839.
2018, 21, 209. [46] C. Noè, A. Cosola, C. Tonda-Turo, R. Sesana, C. Delprete, A.
[30] H. Lu, J. W. Stansbury, C. N. Bowman, J Dent Res 2005, 84, 822. Chiappone, M. Hakkarainen, M. Sangermano, Polymer 2022, 247,
[31] T. L. Grover, C. Allan Guymon, Polym. Chem. 2023, 14, 126. 124779.
[32] J.-S. Chen, S.-L. Tu, R.-Y. Tsay, J. Taiwan Inst. Chem. Eng. 2010, 41, 229. [47] C. Gorsche, R. Harikrishna, S. Baudis, P. Knaack, B. Husar, J. Laeuger,
[33] F. Gritti, D. Hlushkou, U. Tallarek, J. Chromatogr. A 2023, 1693, H. Hoffmann, R. Liska, Anal. Chem. 2017, 89, 4958.
463860. [48] J. Fan, C. Boi, S. M. Lemma, J. Lavoie, R. G. Carbonell, Membranes
[34] E. I. Trilisky, A. M. Lenhoff, Biotechnol. Bioeng. 2009, 104, 127. 2021, 11, 530.

Adv. Mater. Technol. 2023, 8, 2300801 2300801 (10 of 10) © 2023 The Authors. Advanced Materials Technologies published by Wiley-VCH GmbH