Content uploaded by Kimberly Homan
Author content
All content in this area was uploaded by Kimberly Homan on Apr 11, 2016
Content may be subject to copyright.
Three-dimensional bioprinting of thick
vascularized tissues
David B. Kolesky
a,1
, Kimberly A. Homan
a,1
, Mark A. Skylar-Scott
a,1
, and Jennifer A. Lewis
a,2
a
School of Engineering and Applied Sciences, Wyss Institute for Biologically Inspired Engineering, Harvard University, Cambridge, MA 02138
Edited by Kristi S. Anseth, Howard Hughes Medical Institute, University of Colorado Boulder, Boulder, CO, and approved February 2, 2016 (received for review
October 28, 2015)
The advancement of tissue and, ultimately, organ engineering
requires the ability to pattern human tissues composed of cells,
extracellular matrix, and vasculature with controlled microenviron-
ments that can be sustained over prolonged time periods. To date,
bioprinting methods have yielded thin tissues that only survive for
short durations. To improve their physiological relevance, we report a
method for bioprinting 3D cell-laden, vascularized tissues that exceed
1 cm in thickness and can be perfused on chip for long time periods
(>6 wk). Specifically, we integrate parenchyma, stroma, and endothe-
lium into a single thick tissue by coprinting multiple inks composed of
human mesenchymal stem cells (hMSCs) and human neonatal dermal
fibroblasts (hNDFs) within a customized extracellular matrix alongside
embedded vasculature, which is subsequently lined with human um-
bilical vein endothelial cells (HUVECs). These thick vascularized tissues
are actively perfused with growth factors to differentiate hMSCs to-
ward an osteogenic lineage in situ. This longitudinal study of emer-
gent biological phenomena in complex microenvironments represents
a foundational step in human tissue generation.
bioprinting
|
stem cells
|
vasculature
|
tissues
|
biomaterials
The ability to manufacture human tissues that replicate the
essential spatial (1), mechanochemical (2, 3), and temporal
aspects of biological tissues (4) would enable myriad applica-
tions, including 3D cell culture (5), drug screening (6, 7), disease
modeling (8), and tissue repair and regeneration (9, 10). Three-
dimensional bioprinting is an emerging approach for creating
complex tissue architectures (10, 11), including those with em-
bedded vasculature (12–15), that may address the unmet needs
of tissue manufacturing. Recently, Miller et al. (15) reported an
elegant method for creating vascularized tissues, in which a
sacrificial carbohydrate glass is printed at elevated temperature
(>100 °C), protectively coated, and then removed, before in-
troducing a homogeneous cell-laden matrix. Kolesky et al. (14)
developed an alternate approach, in which multiple cell-laden, fu-
gitive (vasculature), and extracellular matrix (ECM) inks are
coprinted under ambient conditions. However, in both cases, the
inability to directly perfuse these vascularized tissues limited
their thickness (1–2 mm) and culture times (<14 d). Here, we
report a route for creating thick vascularized tissues (≥1cm)
within 3D perfusion chips that provides unprecedented control
over tissue composition, architecture, and microenvironment
over several weeks (>6 wk). This longitudinal study of emergent
biological phenomena in complex microenvironments repre-
sents a foundational step in human tissue generation.
Central to the fabrication of thick vascularized tissues is the design
of biological, fugitive, and elastomeric inks for multimaterial 3D
bioprinting. To satisfy the concomitant requirements of process-
ability, heterogeneous integration, biocompatibility, and long-term
stability, we first developed printable cell-laden inks and castable
ECM based on a gelatin and fibrinogen blend (16). Specifically,
these materials form a gelatin–fibrin matrix cross-linked by a dual-
enzymatic, thrombin and transglutaminase (TG), strategy (Fig. 1
and SI Appendix,Fig.S1). The cell-laden inks must facilitate printing
of self-supporting filamentary features under ambient conditions
as well as subsequent infilling of the printed tissue architectures by
casting without dissolving or distorting the patterned construct (Fig.
1A). The thermally reversible gelation of the gelatin–fibrinogen
network enables its use in both printing and casting, where gel and
fluid states are required, respectively (SI Appendix,Fig.S2).
Thrombin is used to rapidly polymerize fibrinogen (17), whereas TG
is a slow-acting Ca
2+
-dependent enzymatic cross-linker that imparts
the mechanical and thermal stability (18) needed for long-term
perfusion. Notably, the cell-laden ink does not contain either enzyme
to prevent polymerization during printing. However, the castable
matrix contains both thrombin and TG, which diffuse into adjacent
printed filaments, forming a continuous, interpenetrating polymer
network, in which the native fibrillar structure of fibrin is preserved
(SI Appendix,Fig.S3). Importantly, our approach allows arbitrarily
thick tissues to be fabricated, because the matrix does not require
UV curing (19), which has a low penetration depth in tissue (20) and
can be readily expanded to other biomaterials, including fibrin and
hyaluronic acid (SI Appendix,Fig.S4).
The gelatin–fibrin matrix supports multiple cell types of in-
terest to both 2D and 3D culture conditions, including human
umbilical vein endothelial cells (HUVECs), human neonatal
dermal fibroblasts (HNDFs), and human bone marrow-derived
mesenchymal stem cells (hMSCs) (Fig. 1 B–Dand SI Appendix,
Fig. S5). We find that endothelial cells express vascular endo-
thelial-cadherin (VE-Cad) (Fig. 1B), and HNDFs (Fig. 1C) and
hMSCs (Fig. 1D) proliferate and spread on this matrix surface
and in bulk. Moreover, the printed cell viability can be as high as
95%, depending on how gelatin is processed before ink formu-
lation. At higher processing temperatures, the average molecular
weight of gelatin is reduced from 69 kDa at 70 °C to 32 kDa at
95 °C processing, resulting in softer gels with lower viscosity,
Significance
Current tissue manufacturing methods fail to recapitulate the
geometry, complexity, and longevity of human tissues. We
report a multimaterial 3D bioprinting method that enables the
creation of thick human tissues (>1 cm) replete with an engi-
neered extracellular matrix, embedded vasculature, and mul-
tiple cell types. These 3D vascularized tissues can be actively
perfused with growth factors for long durations (>6 wk) to
promote differentiation of human mesenchymal stem cells to-
ward an osteogenic lineage in situ. The ability to construct and
perfuse 3D tissues that integrate parenchyma, stroma, and
endothelium is a foundational step toward creating human
tissues for ex vivo and in vivo applications.
Author contributions: D.B.K., K.A.H., M.A.S.-S., and J.A.L. designed research; D.B.K., K.A.H., and
M.A.S.-S. performed research; D.B.K., K.A.H., M.A.S.-S., and J.A.L. analyzed data; and D.B.K. and
J.A.L. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
1
D.B.K., K.A.H., and M.A.S.-S. contributed equally to this work.
2
To whom correspondence should be addressed. Email: jalewis@seas.harvard.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1521342113/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1521342113 PNAS Early Edition
|
1of6
ENGINEERING
shear yield stress, and shear elastic modulus. These cell-laden
inks can be printed with ease and accommodate cell densities
ranging from 0.1 million per mL to 10 million cells per mL (Fig.
1Eand SI Appendix, Fig. S6). Upon printing, hMSCs within this
soft gelatin–fibrinogen matrix continue to spread, proliferate,
and contract into dense, cellular architectures that align along
the printing direction (Fig. 1F), likely arising due to cellular
confinement (21) and contraction via the Poisson effect (22).
To construct thick, vascularized tissues within 3D perfusion
chips, we coprinted cell-laden, fugitive, and silicone inks (Fig. 1
Hand I). First, the silicone ink is printed on a glass substrate and
cured to create customized perfusion chips (Movie S1 and SI
Appendix, Fig. S1). Next, the cell-laden and fugitive inks are
printed on chip, and then encapsulated with the castable ECM
(Fig. 1 J–Land Movie S2). The fugitive ink, which defines the
embedded vascular network, is composed of a triblock copolymer
[i.e., polyethylene oxide (PEO)–polypropylene oxide (PPO)–PEO].
This ink can be removed from the fabricated tissue upon cooling
to roughly 4 °C, where it undergoes a gel-to-fluid transition
(14, 23). This process yields a pervasive network of inter-
connected channels, which are then lined with HUVECs. The
resulting vascularized tissues are perfused via their embedded
vasculature on chip over long time periods using an external pump
(Movie S3) that generates smooth flow over a wide range of flow
rates (24).
To demonstrate the formation of stable vasculature, we prin-
ted a simple tissue construct composed of two parallel channels
embedded within a fibroblast cell-laden matrix (Fig. 2). The
channels are lined with HUVECs, perfused with 1:1 ratio of
endothelial growth media (EGM-2 Bullet kit) and HNDF growth
media [DMEM plus 10% (vol/vol) FBS], and subsequently form
a confluent monolayer that lines each blood vessel (Fig. 2A). The
medium is preincubated for 5 h in the incubator at 37 °C and 5%
CO
2
and replaced every other day. Importantly, after 6 wk of
active perfusion, these endothelial cells maintain endothelial
phenotype and remain confluent, characterized by expression of
CD31, von Willebrand factor (vWF), and VE-Cad (Fig. 2 Band
C). The cross-sectional view of a representative vessel reveals
lumen formation (Fig. 2Dand Movie S4). Confirming the barrier
function of the endothelium, we measured a fivefold reduction
in the diffusional permeability compared with unlined (bare)
channels (Fig. 2Eand SI Appendix, Fig. S7). Stromal HNDFs
residing within the surrounding matrix exhibit cell spreading and
proliferative phenotypes localized to regions within ∼1mmof
Vascular ink Cell ink
Fibrinogen / Fibrin
Gelatin
Printed Cells
Thrombin
Pluronic F-127
Transglutaminase
Cell media
Endothelial cells
AA’
Section A-A’
1 cm
10
2
10
3
10
4
0
20
40
60
80
100
70 75 80 85 90 95
Plateau
Modulus
Viability
(iv)
LK
J
IH
G
E
D
hBM-MSCs
Actin
DAPI
CB
(iii)
(ii)
(i)
A
b
print
cast
evacuate
perfuse
b
HUVECs
VE-Cadherin
DAPI
hBM-MSCs
Actin
HNDFs
Smooth Muscle Actin
DAPI
BM-M
Alkaline
phosphotase
F
Fig. 1. Three-dimensional vascularized tissue fabrication. (A) Schematic illustration of the tissue manufacturing process. (i) Fugitive (vascular) ink, which contains
pluronic and thrombin, and cell-laden inks, which contain gelatin, fibrinogen, and cells, are printed within a 3D perfusion chip. (ii) ECM material, which contains
gelatin, fibrinogen, cells, thrombin, and TG, is then cast over the printed inks. After casting, thrombin induces fibrinogen cleavage and rapid polymerization into
fibrin in both the cast matrix, and through diffusion, in the printed cell ink. Similarly, TG diffuses from the molten casting matrix and slowly cross-links the gelatin
and fibrin. (iii) Upon cooling, the fugitive ink liquefies and is evacuated, leaving behind a pervasive vascular network, which is (iv) endothelialized and perfused
viaanexternalpump.(B) HUVECs growing on top of the matrix in 2D, (C) HNDFs growing inside the matrix in 3D, and (D) hMSCs growing on top of the matrix in
2D. (Scale bar: 50 μm.) (Eand F) Images of printed hMSC-laden ink prepared using gelatin preprocessed at 95 °C before ink formation (E) as printed and (F)after
3 d in the 3D printed filament where actin (green) and nuclei (blue) are stained. (G) Gelatin preprocessing temperature affects the plateau modulus and cell viability
after printing. Higher temperatures lead to lower modulus and higher HNDF viability postprinting. (H) Photographs of interpenetrated sacrificial (red) and cell
inks (green) as printed on chip. (Scale bar: 2 mm.) (I) Top-down bright-field image of sacrificial and cell inks. (Scale bar: 50 μm.). (J–L) Photograph of a printed tissue
construct housed within a perfusion chamber (J) and corresponding cross-sections (Kand L). (Scale bars: 5 mm.)
2of6
|
www.pnas.org/cgi/doi/10.1073/pnas.1521342113 Kolesky et al.
the vasculature (Fig. 2Fand SI Appendix, Fig. S8); cells further
away from these regions become quiescent likely due to an in-
sufficient nutrient supply. As cell density increases, their viability
rapidly decreases at distances beyond 1 mm from the embedded
blood vessels (e.g., only 5% of the cells remain viable at 7 mm).
Clearly, the perfusable vasculature is critical to support living
tissues thicker than 1 mm over long time periods.
To explore emergent phenomena in complex microenvironments,
we created a heterogeneous tissue architecture (>1cmthickand
10 cm
3
in volume) by printing a hMSC-laden ink into a 3D lattice
geometry along with intervening in- and out-of-plane (vertical)
features composed of fugitive ink, which ultimately transform into a
branched vascular network lined with HUVECs. After printing, the
remaining interstitial space is infilled with an HNDF-laden ECM
(Fig. 3A) to form a connective tissue that both supports and binds to
the printed stem cell-laden and vascular features. In this example,
fibroblasts serve as model cells that surround the heterogeneously
patterned stem cells and vascular network. These model cells could
be replaced with either support cells (e.g., immune cells or peri-
cytes) or tissue-specific cells (e.g., hepatocytes, neurons, or islets) in
future embodiments. The embedded vascular network is designed
with a single inlet and outlet that provides an interface between the
printed tissue and the perfusion chip. This network is symmetrically
branched to ensure uniform perfusion throughout the tissue, in-
cluding deep within its core. In addition to providing transport of
nutrients, oxygen, and waste materials, the perfused vasculature is
used to deliver specific differentiation factors to the tissue in a more
uniform manner than bulk delivery methods, in which cells at the
core of the tissue are starved of factors (25). This versatile
platform (Fig. 3A) is used to precisely control growth and dif-
ferentiation of the printed hMSCs. Moreover, both the printed
cellular architecture and embedded vascular network are visible
macroscopically with this thick tissue (Fig. 3B).
To develop a dense osteogenic tissue, we transvascularly de-
livered growth media to the tissue during an initial proliferation
phase (6 d) followed by an osteogenic differentiation mixture that is
perfused for several weeks. Our optimized mixture is composed of
BMP-2, ascorbic acid, and glycerophosphate, to promote mineral
deposition and alkaline phosphatase (AP) expression (SI Appendix,
Fig. S9). To assess tissue maturation, changes in cell function and
matrix composition are observed over time. In good agreement with
prior studies (21), we find that AP expression in hMSCs occurs
within 3 d, whereas mineral deposition does not become noticeable
until 14 d, which coincides with visible collagen-1 deposition by
hMSCs (SI Appendix,Fig.S9)(21).Fig.3Cshows an avascular
tissue produced with comparable hMSC density, in which positive
alizarin stains are only observed within a few hundred microns of
the tissue surface. By contrast, the thick vascularized tissue stains
positive in hMSC regions deep within its core after 30 d of osteo-
genic differentiation by perfusion. We characterized the mineral
deposits, which consist of particulates ∼20–200 nm in size, using
SEM/energy-dispersive X-ray spectroscopy (EDS) analysis. Calcium
A
EF
B
D
C
Fig. 2. Three-dimensional vascularized tissues remain stable during long-term perfusion. (A) Schematic depicting a single HUVEC-lined vascular channel
supporting a fibroblast cell-laden matrix and housed within a 3D perfusion chip. (Band C) Confocal microscopy image of the vascular network after 42 d,
CD-31 (red), vWF (blue), and VE-Cadherin (magenta). (Scale bars: 100 μm.) (D) Long-term perfusion of HUVEC-lined (red) vascular network supporting HNDF-
laden (green) matrix shown by top-down (Left) and cross-sectional confocal microscopy at 45 d (Right). (Scale bar: 100 μm.) (E) Quantification of barrier
properties imparted by endothelial lining of channels, demonstrated by reduced diffusional permeability of FITC-dextran. (F) GFP-HNDF distribution within
the 3D matrix shown by fluorescent intensity as a function of distance from vasculature.
Kolesky et al. PNAS Early Edition
|
3of6
ENGINEERING
and phosphorous peaks are only observed for vascularized tissues, not
the avascular control (SI Appendix,Fig.S9Eand F). The phenotype
of hMSCs varies across the printed filamentary features: cells are
close-packed, compacted, and exhibit a high degree of mineraliza-
tion within the filament core, whereas those in the periphery are
more elongated and exhibit less mineralization. We observe that
subpopulations of HNDFs and hMSCs migrate from their initial
patterned geometry toward the vascular channels and wrap cir-
cumferentially around each channel (Fig. 3D). After 30 d, the
printed hMSCs express osteocalcin within the tissue, and osteocalcin
expression is proportional to distance from the nearest vessel (Fig.
3E). Furthermore, we find that collagen deposition is localized
A
D
E
B
C
F
G
H
I
Fig. 3. Osteogenic differentiation of thick vascularized tissue. (A) Schematic depicting the geometry of the printed heterogeneous tissue within the customized
perfusion chip, whereinthe branched vascular architecture pervades hMSCs that are printed into a 3D lattice architecture, and HNDFs are cast within an ECM that
fills the interstitial space. (B) Photographs of a printed tissue construct within and removed from the customized perfusion chip. (C) Comparative cross-sections of
avascular tissue (Left) and vascularized tissue (Right) after 30 d of osteogenicmedia perfusion with alizarin red stain showing location of calcium phosphate. (Scale
bar: 5 mm.) (D) Confocal microscopy image through a cross-section of 1-cm-thick vascularized osteogenic tissue construct after 30 d of active perfusion and in situ
differentiation. (Scale bar: 1.5 mm.) (E) Osteocalcin intensity across the thick tissue sample inside the red lines shown in C.(F) High-resolution image showing
osteocalcin (purple) localized within hMSCs, and they appear to take on symmetric osteoblast-like morphologies. (Scale bar: 100 μm.) After 30 d (Gand H), thick
tissue constructs are stained for collagen-I (yellow), which appears to be localized near hMSCs. (Scale bars: 200 μm.) (I) Alizarin red is used to stain calcium
phosphate deposition, and fast blue is used to stain AP, indicating tissue maturation and differentiation over time. (Scale bar: 200 μm.)
4of6
|
www.pnas.org/cgi/doi/10.1073/pnas.1521342113 Kolesky et al.
within printed filaments and around the circumference of the
vasculature (Fig. 3 F–Hand SI Appendix, Fig. S9).
In summary, thick, vascularized human tissues with programmable
cellular heterogeneity that are capable of long-term (>6-wk) perfu-
sion on chip have been fabricated by multimaterial 3D bioprinting.
The ability to recapitulate physiologically relevant, 3D tissue mi-
croenvironments enables the exploration of emergent biological
phenomena, as demonstrated by observations of in situ development
of hMSCs within tissues containing a pervasive, perfusable, endo-
thelialized vascular network. Our 3D tissue manufacturing platform
opens new avenues for fabricating and investigating human tissues
for both ex vivo and in vivo applications.
Methods
Solution Preparation. Ink and matrix precursor solutions are prepared before
printing the tissue engineered constructs. A 15 wt/vol% gelatin solution (type A;
300 bloom from porcine skin; Sigma) is produced by warming in DPBS (1×
Dulbecco’s PBS without calcium and magnesium) to 70 °C (unless otherwise noted)
and adding gelatin powder to the solution while vigorously stirring for 12 h at
70 °C (unless otherwise noted), and then the pH is adjusted to 7.5 using 1 M
NaOH. The warm gelatin solution is sterile filtered and stored at 4 °C in aliquots
for later use (<3 mo). Fibrinogen solution (50 mg·mL
−1
) is produced by dis-
solving lyophilized bovine blood plasma protein (Millipore) at 37 °C in sterile
DPBS without calcium and magnesium. The solution is held at 37 °C for 45 min
to allow complete dissolution. The TG solution (60 mg·mL
−1
) is prepared by
dissolving lyophilized powder (Moo Glue) in DPBS without calcium and mag-
nesium and gently mixing for 20 s. The solution is then placed at 37 °C for 20 min
and sterile filtered before use. A 250 mM CaCl
2
stock solution is prepared by
dissolving CaCl
2
powder in DPBS without calcium and magnesium (Corning). To
prepare stock solution of thrombin, lyophilized thrombin (Sigma-Aldrich) is
reconstituted at 500 U·mL
−1
using sterile DPBS and stored at −20 °C. The
thrombin aliquots are thawed immediately before use.
Matrix Formulations. The solutions are mixed together at 37 °C to achieve a final
concentration of 10 mg·mL
−1
fibrinogen, 7.5 wt% gelatin, 2.5 mM CaCl
2
,and
0.2 wt% TG. For printing, we use 1 wt% TG to account for diffusion and dilution
into printed cell filaments. The equilibration time before mixing with thrombin
(at a ratio of 500:1) determines optical clarity (SI Appendix,Fig.S3). After mixing,
the matrix must be quickly cast, as rapid polymerization ensues. Native fibrin
matrix is created by the same procedure without gelatin and TG (SI Appendix,Fig.
S4). Alternatively, hyaluronic acid methacrylate can be synthesized and used (26).
Ink Formulations. A silicone ink, composed of a two-part silicone elastomer (SE
1700; Dow Chemical) with a 10:1 base to catalyst (by weight), is used to create
customized perfusion chips. It is homogenized using a mixer (2,000 speed; AE-310;
Thinky Corporation) and printed within 2 h of mixing. A fugitive ink, composed of
38 wt% Pluronic F127 (Sigma) and 100 U·mL
−1
thrombin in deionized, ultra-
filtrated water, is used to print the vasculature. A stock solution (40% Pluronic
F127) is homogenized using a Thinky mixer and subsequently stored at 4 °C.
Before use, 2,000 U·mL
−1
thrombin solution is added to ink at a ratio of 1:20,
homogenized, loaded into a syringe (EFD, Inc.) at 4 °C, and centrifuged to
remove any air bubbles. All inks are printed at room temperature.
A cell-laden ink, composed of 7.5 wt/vol% gelatin and 10 mg·mL
−1
fibrinogen, is
prepared for printing. Ink stiffness is tuned by varying the gelatin-processing
temperature (70–95 °C) (SI Appendix,Fig.S6). This ink is prepared similarly to the
matrix, but without TG and thrombin. Upon printing, cross-linking is achieved by
diffusion of these enzymes from the surrounding matrix. To disperse cells in the
ink, the fibrinogen–gelatin blend is held at 37 °C, and then cell suspensions are
introduced via gentle pipetting. After mixing, the ink is held at 4 ° C for 15 min to
drive thermal gelation of the gelatin phase. Next, the ink is warmed to room
temperature for at least 15 min, where it can be immediately printed for up to 2 h.
Fibrinogen–Fluorophore Conjugation. To visualize the fibrin network in printed
filaments and the cast matrix (SI Appendix,Fig.S3), fibrinogen is conjugated to
two fluorophores. Specifically, 1 g of bovine fibrinogen is dissolved in 100 mL of
50 mM borate buffer, pH 8.5 (Thermo Scientific), to form a 10 mg·mL
−1
solution.
N-Hydroxysuccinimide, conjugated with either fluorescein or rhodamine, is
added at a 10:1 molar ratio of dye/fibrinogen. After reacting for 2 h at room
temperature, the labeled fibrinogen is separated from unconjugated dye by
dialysis using 10-kDa MWCO dialysis tubing in a 2-L bath against PBS for 3 d,
changing the PBS in the bath twice daily. After dialysis is complete, the fluo-
rescently conjugated fibrinogen is frozen at −80 °C, lyophilized, and stored at
−20 °C before use.
Rheological Characterization. Ink rheology is measured using a controlled stress
rheometer (DHR-3; TA Instruments) with a 40-mm diameter, 2° cone and plate
geometry. The shear storage (G’)andloss(G’’) moduli are measured at a fre-
quency of 1 Hz and an oscillatory strain (γ) of 0.01. Temperature sweeps are
performed using a Peltier plate over the range from −5to40°C.Samplesare
equilibrated for 5 min before testing and for 1 min at each subsequent tem-
perature to minimize thermal gradients throughout the sample. Time sweeps
are conducted by rapidly placing a premixed solution onto the temperature-
controlled Peltier plate held at 37 or 22 °C, unless otherwise noted.
Cell Culture and Maintenance. hMSCs (Rooster Bio) are cultured in Booster
Media (Rooster Bio) and are not used beyond two passages. Green fluo-
rescent protein-expressing HNDFs (GFP-HNDFs) (Angio-Proteomie) are cul-
tured in Dulbecco’s modified Eagle medium containing high glucose and
sodium pyruvate (DMEM) (GlutaMAX; Gibco) and supplemented with 10%
FBS (Gemini Bio-Products). Primary red fluorescent protein-expressing HUVECs
(RFP-HUVECs) (Angio-Proteomie) are cultured in EGM-2 media (complete
EGM-2 BulletKit; Lonza). GFP-HNDFs and RFP HUVECs are not used beyond
the 15th and 9th passages, respectively.
Three-Dimensional Tissue Fabrication on Perfusable Chips. All vascularized tis-
sues are created on a custom-designed multimaterial 3D bioprinter equipped with
four independently addressable print heads mounted onto a three-axis, motion-
controlled gantry with build volume of 725 ×650 ×125 mm (AGB 10000; Aer-
otech). Each ink is housed in a syringe equipped with a leur-locked nozzle of
varying size (i.e., 100-μmto410-μm diameter) (EFD, Inc.). Inks are deposited by
applying air pressure (800 Ultra dispensing system; EFD, Inc.), ranging from 10 to
140 psi, corresponding to print speeds from 1 mm·s
−1
to 5 cm·s
−1
.
To manufacture the customized perfusion chips, the silicone ink is loaded into a
10-mL syringe, centrifuged to remove air bubbles, and deposited through a ta-
pered 410-μm nozzle. The gasket design is created using custom MATLAB soft-
ware and the structures are printed onto 50 ×75-mm glass slides. After printing,
the chips are cured at 80 °C in an oven for >1 h and stored at room temperature.
To produce thick vascularized tissues, multiple inks are sequentially coprinted
within the customized perfusion chips. To form a base layer, a thin film of gelatin–
fibrin matrix, containing 0.1 wt% TG, is cast onto the base of the perfusion chip
and allowed to dry. Next, the fugitive Pluronic F127 and cell-laden inks are
printed onto the surface using 200-μm straight and tapered nozzles, respectively.
After printing, stainless metal tubes are fed through the guide channels of the
perfusion chip and pushed into physical contact with printed vertical pillars of
the fugitive ink positioned at the inlet and outlet of each device (SI Appendix,
Fig. S1,andMovie S2). Before encapsulation, TG is added to the molten 37 ° C
gelatin–fibrin matrix solution and preincubated for 2–20 min depending on the
desired matrix transparency (SI Appendix,Fig.S3). To form a cell-laden matrix,
the molten 37 °C gelatin–fibrin matrix is first mixed with HNDF-GFP cells and
then mixed with thrombin. Next, this matrix is cast around the printed tissue,
where it undergoes rapid gelation due to thrombin activity. The 3D tissue chips
are stored at 37 °C for 1 h before cooling to 4 °C to liquefy and remove the
printed fugitive ink, which is flushed through the device using cold cell media,
leaving behind open conduits.
The 3D perfusion chips are loaded onto a machined stainless-steel base, and
a thick acrylic lid is placedon top. The lid and baseare clamped together by four
screws, forming a seal around the silicone 3D printed gasket top. Next, sterile
two-stopperistaltic tubing(PharMed BPT) is filled with mediaand connected to
the outlet of a sterile filter that is attached to a 10-mL syringe (EFD Nordson),
which servesas a media reservoir. Media that has been equilibratingfor >6hin
an incubator at 37 °C, 5% CO
2
is added to the media reservoir, and by means
of gravity, is allowed to flow through the filter and peristaltic tubing, until all
of the air is displaced, before connecting the peristaltic tubing to the inlet of
each perfusion chip. Hose pinch-off clamps are added at the inlet and outlet of
the perfusion chip to prevent uncontrolled flow when disconnected from the
peristaltic pump, which can damage the endothelium or introduce air bubbles
to the vasculature. The media reservoir is allowed to equilibrate with atmo-
spheric pressure at all times by means of a sterile filter connecting the in-
cubator environment with the reservoir.
Endothelialization of Vascular Networks. With the peristaltic tubing removed
from the chip outlet, 50–500 μL of HUVEC suspensions (1 ×10
7
cells per mL)
are injected via pipette to fill the vascular network. The silicone tubing is
then replaced, and both the outlet and inlet pinch-clamp are sealed. The
perfusion chip is incubated at 37 °C to facilitate cell adhesion to the channels
under zero-flow conditions. After 30 min, the chip is flipped 180° to facili-
tate cell adhesion to the other side of the channel, and achieve circumfer-
ential seeding of cells in the channel. Finally, the cells are further incubated
for between 5 h and overnight at 37 °C before commencing active perfusion.
Kolesky et al. PNAS Early Edition
|
5of6
ENGINEERING
Active Perfusion. After endothelial cell seeding, the peristaltic tubing is
affixed to a 24-channel peristaltic pump (Ismatec), after which the hose
clamps are removed. For single vascular channels, the perfusion rate is set at
13 μL·min
−1
, whereas for thick vascularized tissues, it is set at 27 μL·min
−1
.
Cell Viability Assay. Cell viability is determined postprinting by printing inks
with 2 ×10
6
cells per mL for each condition. Printed cell-laden filaments (2 ×
10
6
cells per mL for each condition) are deposited onto a glass substrate and
then stained using calcein-AM (“live”;1μL·mL
−1
; Invitrogen) and ethidium
homodimer (“dead”;4μL·mL
−1
; Invitrogen) for 20 min before confocal im-
aging (n=3 unique samples, imaged n=10 times). To assess cell viability,
live tissue is removed from the perfusion chip, cross-sectioned, and stained
using the same staining protocol. Live and dead cell counts are obtained
using the 3D objects counter plugin in ImageJ software. The results are av-
eraged and SDs determined for each sample.
Imaging and Analysis. Photographs and videos of tissue fabrication are ac-
quired using a DSLR camera (Canon EOS, 5D Mark II; Canon). Fluorescent dyes
are used to improve visualization of Pluronic F127 (Red, Risk Reactor) and
gelatin–fibrin ink (Fluorescein; Sigma-Aldrich). Printed tissue structures are
imaged using a Keyence Zoom (VHX-2000; Keyence), an inverted fluores-
cence (Axiovert 40 CFL; Zeiss), and an upright confocal microscope (LSM710;
Zeiss). ImageJ is used to generate composite microscopy images by com-
bining fluorescent channels. Three-dimensional rendering and visualization
of confocal stacks are performed in Imaris 7.6.4, Bitplane Scientific Software,
and ImageJ software. Cell counting is performed using semiautomated na-
tive algorithms in Imaris and ImageJ counting and tracking algorithms.
Immunostaining. Immunostaining and confocal microscopy are used to assess
the 3D vascularized tissues. Printed tissues are first washed with PBS via
perfusion for several minutes. Next, 10% buffered formalin is perfused
through the 3D tissue for 10–15 min. The tissue is removed from the per-
fusion chip and bathed in 10% buffered formalin. A 2-h fixation time is
required for a 1-cm-thick tissue. The 3D tissues are then washed in PBS for
several hours and blocked overnight using 1 wt% BSA in PBS. Primary an-
tibodies to the cell protein or biomarker of interest are incubated with the
constructs for 2 d in a solution of 0.5 wt% BSA and 0.125 wt% Triton X-100
(SI Appendix, Table S1). Removal of unbound primary antibodies is accom-
plished using a wash step against a solution of PBS or 0.5 wt% BSA and
0.125 wt% Triton X-100 in PBS for 1 d. Secondary antibodies are incubated
with the constructs for 1 d at the dilutions listed in SI Appendix, Table S1,in
a solution of 0.5 wt% BSA and 0.125 wt% Triton X-100 in PBS. Samples are
counterstained with NucBlue or ActinGreen for 2 h and then washed for
1 d in PBS before imaging. Confocal microscopy is performed using an up-
right Zeiss LSM 710 with water-immersion objectives ranging from 10×to
40×using spectral lasers at 405-, 488-, 514-, 561-, and 633-nm wavelengths.
Image recons tructions of zstacks are performed in ImageJ using the z-project
function with the maximum pixel intensity setting. Three-dimensional image
reconstructions are performed using Imaris software.
hMSC Staining. Fast Blue (Sigma-Aldrich) and alizarin red (SigmaFast; Sigma-
Aldrich) are used to visualize AP activity and calcium deposition. One tablet of Fast
Blue is dissolved in 10 mL of deionized (DI) water. This solution is stored in the
dark and used within 2 h. Cells are washed using 0.05% Tween 20 in DPBS
without calcium and magnesium and fixed as described above. The samples are
then covered with Fast Blue solution and incubated in the dark for 5–10 min and
washed using PBS-Tween buffer. To assess mineralization, 2% alizarin red so-
lution is dissolved in DI water, mixed vigorously, filtered, and used within 24 h.
Samples are equilibrated in DI water and incubated with alizarin red solution for
a few minutes, then the staining solution is removed, and samples are washed
three times in DI water or until background dye is unobservable. Representative
slices of both avascular and vascularized, thick tissues are digested using 2 wt%
Collagenase I in PBS without Ca
2+
,Mg
2+
at 37 °C for >24 h. The resulting solu-
tions are filtered using a 0.2-μm sterile filter and rinsed with DI water. SEM/EDS is
used to carry out elemental analysis on harvested mineral particulates.
FITC-Dextran Permeability Testing. To assess barrier function of the printed
vasculature, diffusional permeability was quantified by perfusing culture media
in the vascular channel, while alive, containing 25 μg/mL FITC-conjugated
70-kDa dextran (FITC-Dex; Sigma product 46945) at a rate of 20 μL·min
−1
for
3minand1μL·min
−1
thereafter for ∼33 min. The diffusion pattern of FITC-Dex
was detected using a wide-field fluorescent microscope (Zeiss Axiovert 40 CFL).
Fluorescence images were captured before perfusion and every 3–5 min after
for 33 min. Diffusional permeability of FITC-Dex is calculated by quantifying
changes of fluorescence intensity over time using the following equation:
Pd=1
I1−IbI2−I1
td
4.
P
d
is the diffusional permeability coefficient, I
1
is the average intensity at an
initial time point, I
2
is an average intensity after some time (t,∼30 min), I
b
is
background intensity (before introducing FITC-Dex), and dis the channel
diameter (27). The measurements are performed on embedded channels
with and without endothelium (n=3).
ACKNOWLEDGMENTS. We thank Donald Ingber, David Mooney, and Christopher
Hinojosa for useful discussions; Jessica Herrmann, Humphrey Obuobi, Hayley
Price, Nicole Black, Tom Ferrante, and Oktay Uzun for their experimental
assistance; and Lori K. Sanders for help with photography and videography.
This work was supported by NSF Early-concept Grants for Exploratory Research
(EAGER) Award Division of Civil, Mechanical and Manufacturing Innovation
(CMMI)-1548261 and by the Wyss Institute for Biologically Inspired Engineering.
1. Nelson CM, Vanduijn MM, Inman JL, Fletcher DA, Bissell MJ (2006) Tissue geometry
determines sites of mammary branching morphogenesis in organotypic cultures.
Science 314(5797):298–300.
2. Lee K, Silva EA, Mooney DJ (2011) Growth factor delivery-based tissue engineering: General
approaches and a review of recent developments. J R Soc Interface 8(55):153–170.
3. Ingber DE (2003) Mechanobiology and diseases of mechanotransduction. Ann Med
35(8):564–577.
4. Abbott RD, Kaplan DL (2015) Strategies for improving the physiological relevance of
human engineered tissues. Trends Biotechnol 33(7):401–407.
5. Pampaloni F, Reynaud EG, Stelzer EHK (2007) The third dimension bridges the gap
between cell culture and live tissue. Nat Rev Mol Cell Biol 8(10):839–845.
6. Huh D, Hamilton GA, Ingber DE (2011) From 3D cell culture to organs-on-chips. Trends
Cell Biol 21(12):745–754.
7. Bhatia SN, Ingber DE (2014) Microfluidic organs-on-chips. Nat Biotechnol 32(8):760–772.
8. Lee GY, Kenny PA, Lee EH, Bissell MJ (2007) Three-dimensional culture models of
normal and malignant breast epithelial cells. Nat Methods 4(4):359–365.
9. Langer R, Vacanti JP (1993) Tissue engineering. Science 260(5110):920–926.
10. Murphy SV, Atala A (2014) 3D bioprinting of tissues and organs. Nat Biotechnol 32(8):
773–785.
11. Atala A, Kasper FK, Mikos AG (2012) Engineering complex tissues. Sci Transl Med
4(160):160rv12.
12. Cui X, Boland T (2009) Human microvasculature fabrication using thermal inkjet
printing technology. Biomaterials 30(31):6221–6227.
13. Norotte C, Marga FS, Niklason LE, Forgacs G (2009) Scaffold-free vascular tissue en-
gineering using bioprinting. Biomaterials 30(30):5910–5917.
14. Kolesky DB, et al. (2014) 3D bioprinting of vascularized, heterogeneous cell-laden
tissue constructs. Adv Mater 26(19):3124–3130.
15. Miller JS, et al. (2012) Rapid casting of patterned vascular networks for perfusable
engineered three-dimensional tissues. Nat Mater 11(9):768–774.
16. Lee KY, Mooney DJ (2001) Hydrogels for tissue engineering. Chem Rev 101(7):
1869–1879.
17. Mosesson MW (1998) Fibrinogen structure and fibrin clot assembly. Semin Thromb
Hemost 24(2):169–174.
18. Chen R-N, Ho H-O, Sheu M-T (2005) Characterization of collagen matrices crosslinked
using microbial transglutaminase. Biomaterials 26(20):4229–4235.
19. Stoien JD, Wang RJ (1974) Effect of near-ultraviolet and visible light on mammalian
cells in culture II. Formation of toxic photoproducts in tissue culture medium by
blacklight. Proc Natl Acad Sci USA 71(10):3961–3965.
20. Jayakumar MKG, Idris NM, Zhang Y (2012) Remote activation of biomolecules in deep
tissues using near-infrared-to-UV upconversion nanotransducers. Proc Natl Acad Sci
USA 109(22):8483–8488.
21. Klumpers DD, Zhao X, Mooney DJ, Smit TH (2013) Cell mediated contraction in 3D
cell-matrix constructs leads to spatially regulated osteogenic differentiation. Integr
Biol (Camb) 5(9):1174–1183.
22. Oster GF, Murray JD, Harris AK (1983) Mechanical aspects of mesenchymal morpho-
genesis. J Embryol Exp Morphol 78:83–125.
23. Wu W, DeConinck A, Lewis JA (2011) Omnidirectional printing of 3D microvascular
networks. Adv Mater 23(24):H178–H183.
24. Giulitti S, Magrofuoco E, Prevedello L, Elvassore N (2013) Optimal periodic perfusion
strategy for robust long-term microfluidic cell culture. Lab Chip 13(22):4430–4441.
25. Griffith LG, Swartz MA (2006) Capturing complex 3D tissue physiology in vitro. Nat
Rev Mol Cell Biol 7(3):211–224.
26. Burdick JA, Chung C, Jia X, Randolph MA, Langer R (2005) Controlled degradation and
mechanical behavior of photopolymerized hyaluronic acid networks. Biomacromolecules
6(1):386–391.
27. Price G, Tien J (2011) Methods in Molecular Biology, ed Khademhosseini A (Humana,
Totowa, NJ).
6of6
|
www.pnas.org/cgi/doi/10.1073/pnas.1521342113 Kolesky et al.
