Lab on a Chip

View Article Online

PAPER View Journal

Micro-dissected tumor tissues on chip: an ex vivo

Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

method for drug testing and personalized

Cite this: DOI: 10.1039/c5lc01108f
therapy†
M. Astolfi,bc B. Péant,c M. A. Lateef,c N. Rousset,a J. Kendall-Dupont,c E. Carmona,c
F. Monet,a F. Saad,cd D. Provencher,ce A.-M. Mes-Massoncf and T. Gervais*abc

In cancer research and personalized medicine, new tissue culture models are needed to better predict the
response of patients to therapies. With a concern for the small volume of tissue typically obtained through
a biopsy, we describe a method to reproducibly section live tumor tissue to submillimeter sizes. These
micro-dissected tissues (MDTs) share with spheroids the advantages of being easily manipulated on-chip
and kept alive for periods extending over one week, while being biologically relevant for numerous assays.
At dimensions below ~420 μm in diameter, as suggested by a simple metabolite transport model and con-
firmed experimentally, continuous perfusion is not required to keep samples alive, considerably simplifying
the technical challenges. For the long-term culture of MDTs, we describe a simple microfluidic platform
that can reliably trap samples in a low shear stress environment. We report the analysis of MDT viability for
eight different types of tissues (four mouse xenografts derived from human cancer cell lines, three from
ovarian and prostate cancer patients, and one from a patient with benign prostatic hyperplasia) analyzed by
Received 16th September 2015, both confocal microscopy and flow cytometry over an 8-day incubation period. Finally, we provide a proof
Accepted 2nd December 2015
of principle for chemosensitivity testing of human tissue from a cancer patient performed using the
described MDT chip method. This technology has the potential to improve treatment success rates by
DOI: 10.1039/c5lc01108f
identifying potential responders earlier during the course of treatment and providing opportunities for
www.rsc.org/loc direct drug testing on patient tissues in early drug development stages.

Introduction fail to correctly predict clinical outcomes in patients.2 In addi-

tion, with a growing awareness that treatment response can
Oncology drugs have a very low success rate in clinical trials be highly context dependent, there is a lack of proper patient
and less than 7% of the drugs entering clinical trials are ulti- stratification to prospectively identify subgroups of patients
mately approved by the Food and Drug Administration (FDA), that will most likely benefit from new treatments.3 Therefore,
according to a recent United States based survey conducted generating more relevant models that could be used for pre-
between 2003 and 2011.1 This low performance points to po- clinical testing or as a patient-specific drug-testing tool to
tential weaknesses in the current drug development pipeline. guide the choice of a therapy would be of great interest to the
One obstacle may be that currently used preclinical models cancer research community.
There is growing evidence that tissue tridimensionality,
cellular composition, and micro-environment play an impor-
a
Department of Engineering Physics, Polytechnique Montréal, Montreal, QC, tant role in cancer development and response to therapy. For
Canada. E-mail: thomas.gervais@polytmtl.ca; Tel: +1 514 340 4711, ext. 3752 these reasons, researchers need more sophisticated tissue
b
Institute of Biomedical Engineering, Polytechnique Montréal, Montreal, QC, based culture models in order to mimic critical features not
Canada
c
represented in the traditional monolayer cultures.4–6 Even
Institut du cancer de Montréal and Centre de recherche du Centre hospitalier de
l'Université de Montréal (CRCHUM), Montreal, QC, Canada
animal models, mainly genetically modified mice and cancer
d
Department of Surgery, Faculty of Medicine, Université de Montréal, Montreal, cell line xenografts grown in immunodeficient mice, have lim-
QC, Canada itations related to inherent biological features3 as well as cost
e
Division of Gynecologic Oncology, Department of Obstetrics-Gynecology, and time considerations. The field of tissue engineering
Université de Montréal, Montreal, QC, Canada
f
strives to create new 3D models that reconstruct some of the
Department of Medicine, Faculty of Medicine, Université de Montréal, Montreal,
QC, Canada
important characteristics found in physiological tissues, but
† Electronic supplementary information (ESI) available. See DOI: 10.1039/ matching the complexity of in vivo tissues remains an impor-
c5lc01108f tant challenge.

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

Recent attention has focused on the use of spheroids that tissues die prematurely, leaving little or no time to test a ther-
provide 3D models that begin to bridge the gap between apy and obtain a relevant readout. One approach has been to
monolayer cultures and tissues, as evidenced by several cut the tissue into thin (250–500 μm) slices of relatively large
parameters including gene expression studies.7 Spheroids (~4 mm) diameter,14,15 with the smaller dimension facilitat-
can be reproducibly mass-produced from a number of ing the transport of nutrients to the center of the tissue. Such
established cell lines, although not all cell lines have the samples have been cultured in microfluidic devices under
potential to form spheroids.8 Due to their small size, they are continuous perfusion,16,17 but their large format makes them
compatible with microfluidic approaches and there are an challenging to process in microsystems and the cumbersome
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

increasing number of miniaturized devices specifically fluidic connections to a perfusion system reduce the number
designed to study spheroids.9–13 However, spheroids have a of independent tissue samples that can be assayed in paral-
limited ability to mimic the complex tissue architecture and lel. To circumvent these problems, we propose a new tissue
cell composition of human tumors and they do not repro- culture method combining the high biological relevance of
duce the unique characteristics of specific patients' cancers. patient-derived tumor slices with the manipulation simplicity
At their present level of development, spheroids provide in microsystems of small spheroid-sized tissue samples. This
limited advantages in the growing field of personalized novel tissue culture method can potentially be applied
medicine. throughout the drug life cycle from preclinical testing to clin-
Employing patient-derived organotypic ex vivo cultures ical patient response prediction guiding the selection of an
may provide a better model for the empirical testing of thera- optimal treatment regimen.
peutics. However, this approach has presented some signifi- Conceptually, our proposed approach to enable treatment
cant hurdles, including the maintenance of viability over a response assays on patient tissue (Fig. 1) first involves cutting
sufficient number of days for different analytical purposes. In limited amounts of a patient tumor, obtained through sur-
the absence of a functional vasculature, ex vivo primary gery or biopsy, into individual submillimeter-sized tissue

Fig. 1 Proposed approach to treatment selection using micro-dissected tumors on chip. 1) Tumor tissue is extracted from the patient, either
through surgery or biopsy, and cut to multiple individual micro-dissected tumor samples (MDTs). 2) The MDTs are then loaded, trapped, and incu-
bated within a microfluidic device composed of several channels, each able to trap five MDTs. 3) One or several selected drug candidates to be
assessed are applied to MDTs in independent channels while other MDTs are kept as a non-treated controls. 4) After incubation with the drugs,
the MDTs are analyzed by detecting live and dead cells using confocal microscopy or by measuring other drug-response parameters, and the
results are compared to those of the non-treated controls. 5) Results are then interpreted to identify non responders to treatment and obtain use-
ful information to elaborate a personalized treatment strategy.

Lab Chip This journal is © The Royal Society of Chemistry 2015

 View Article Online

Lab on a Chip Paper

sections. These micro-dissected tumor/tissue samples (MDTs) of O2 in non-perfused medium by taking advantage of the
are then loaded into a specialized microfluidic platform in PDMS material property.
which small volumes of one or several drug candidates are In a seminal paper on anoxia in human tumors,
tested directly on patient tissue in independent channels. Thomlinson and Gray21 used a simple O2 consumption
The chemoresponse of the tissue is then evaluated and com- model to explain the formation of necrotic cores in cylindri-
pared to non-treated controls in order to generate drug- cal non-vascularized human lung tumors and to obtain a crit-
response data specific to each patient. ical tissue thickness above which anoxia is induced. Others
To begin to address the challenges of this approach, we have contributed to the development of models to character-
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

first derived theoretical evidence that sectioning tissue to ize the distribution of nutrients and waste in spherical
submilllimeter dimensions helps to maintain viability ex vivo samples,22–26 especially to study spheroids. In order to justify
by ensuring adequate oxygenation throughout the tissue, the pertinence of dissecting the tumor samples to submilli-
even without continuous perfusion. Based in part on 3D meter sizes and to determine optimal dimensions to avoid
numerical simulations, we designed and created a simple complete depletion of O2 in the center of MDTs, we derived a
microfluidic platform to trap and culture MDTs while simple diffusion–reaction model of O2 consumption for non-
shielding them from excessive shear stress. After refining the vascularized spherical tissue samples. Our model supposes
methodology to generate spheroid-sized MDTs, we experi- that O2 is consumed at a constant rate (zero-order reaction
mentally validated that tissue viability is preserved within our kinetics), following previous studies,21–23,26 despite knowing
microsystem for at least eight days without perfusion, with that cells can modulate their consumption as a function of
regular medium replacement, using MDTs produced from the available O2 concentration.24,27,28 This simple approach
different types of tissues: xenografts derived from four differ- guarantees a lower bound on the maximum tissue diameter
ent human cancer cell lines and four ex vivo tissues from before O2 is depleted. Indeed, all other reaction kinetics
patients. Finally, a proof-of-principle assay was performed models of higher order (1st order, Michaelis–Menten, etc.)
using ovarian cancer tissue obtained from a patient to dem- would set reduced consumption rates in the presence of
onstrate that the procedure can generate patient-specific drug lower O2 concentrations and consequently derive a higher
response data of potentially high clinical value. critical diameter before hypoxia sets in.
In non-perfused conditions, the distribution of O2 inside
and around the tissue follows a diffusive process. Within the
Theory tissue, O2 is additionally consumed by the cells. The equation
describing these mass transfer phenomena is:
Analysis of critical tissue size to avoid anoxia
Our goal is to prepare tissue sections that are large enough (1)
to mimic naturally occurring gradients of nutrients, waste
and signaling molecules while also being small enough to
maintain high viability throughout the tissue without risking where C is the concentration of O2, t is the time, D is the dif-
anoxia in the center. As a non-polar molecule, oxygen (O2) fusion constant of O2 and q is the volumetric O2 consump-
dissolves only to low concentrations in medium (Table 1), tion rate. The 3D space is partitioned into two subdomains of
which explains in part why most groups have chosen to con- live tissue and medium surrounding it, with a continuity
tinuously perfuse their tissue samples.16–19 However, polydi- boundary condition linking both of them. The analysis was
methylsiloxane (PDMS) is a gas-permeable polymer that is further simplified by placing the MDT in an infinite aqueous
often used to fabricate microsystems. As demonstrated exper- medium rather than in a PDMS device, and assuming spheri-
imentally using spheroids,20 if the tissue samples are suffi- cal symmetry. We also excluded the necrotic core subdomain
ciently small, it becomes possible to maintain adequate levels as it is unnecessary for the purpose of the demonstration

Table 1 Description of the variables for the calculation of oxygen (O2) concentration in a tissue sample and in the surrounding medium

Variable Description Value References

−3
Cmax Maximum dissolution concentration of O2 in water at 37 °C and at 0.2 atm 0.20 mol m Henry's Law; Sander29
(partial pressure of O2 in a cell incubator)
Q = q/ρ Maximum O2 uptake rate per cell 7.37 × 10−17 mol s−1 Average from
literature30–34
ρ Density of cells in cancerous tissue 2.76 × 1014 cells per m3 Experimental value
DT O2 diffusivity constant in cancerous tissue at 37 °C 1.83 × 10−9 m2 s−1 Average from
literature23,33–36
DM O2 diffusivity constant in water at 37 °C 2.62 × 10−9 m2 s−1 Han & Bartels37
q Maximum O2 uptake rate per volume of tissue Calculated (Q × ρ) N/A
RC Critical radius of a non-perfused spherical tissue section Calculated (eqn (2)) N/A
RCP Critical radius of a perfused spherical tissue section Calculated (eqn (3)) N/A

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

outlined here. All these simplifications again yield a lower Materials and methods
boundary on the critical diameter because O2 in PDMS has Fabrication of a microfluidic incubation device for MDTs
both a greater permeability and diffusivity than in water
(ESI† Table S1). The other boundary conditions are zero con- Each microfluidic platform is composed of two PDMS rep-
centration at the center of the tissue (C (r = 0)) and maximum licas obtained from micromachined master molds. The bot-
dissolution concentration of O2 in medium at infinity (C (r → tom PDMS layer forms five open channels with a 600 μm-
∞ = Cmax)). Outside the tissue, the O2 consumption term q wide square cross-section, each containing five 600 μm-wide
falls to zero. Assuming steady state (∂C/∂t = 0) and solving square-bottom microfluidic wells of 500 μm in height. The
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

the subdomain differential equations for radius r, the follow- top layer is composed of 3 mm diameter inlet holes and
ing expression of the non-perfused critical diameter (2RC) is 2 mm diameter outlet holes, and closes the upper side of the
found (see ESI† for details): channels once assembled with the bottom layer (ESI† Fig.
S1). The molds were carved out of polyIJmethyl methacrylate)
blocks using 3.57 mm and 1 mm diameter end mills con-
trolled by a computerized numerical control machine. Liquid
(2)
PDMS (Sylgard® 184 silicone elastomer kit, Dow Corning,
Midland, USA) prepared at a base polymer to curing agent
the critical diameter depends on the diffusivity constant of mass ratio of 10 : 1 was poured into each mold, degassed,
O2 in the tissue (DT) and in the medium (DM), on the maxi- and cooked at 80 °C for 1.5 hours. The platform was assem-
mum dissolution concentration of O2 in aqueous medium bled by plasma-bonding the two PDMS layers together and by
(Cmax), and on the volumetric O2 consumption rate by the tis- fitting hollow nylon cylinders (#91145A138, McMaster-Carr,
sue (q = ρ Q) where ρ is the density of cells in tissue and Q is Elmhurst, USA) into the inlet holes of the top layer to form a
a constant O2 consumption rate per cell. larger inlet reservoir (ESI† Fig. S1).
Using parameters found in the literature28,29,36,37 or mea-
sured experimentally for tumor tissue and aqueous solutions Production of human prostate and ovarian cancer xenografts
(Table 1), a critical diameter equal to 424 μm was calculated in mice
according to eqn (2). By keeping all dimensions of a tissue Four different human carcinoma cell lines derived from pros-
sample below this critical point, perfusion thus becomes tate cancer tumors (22Rv1 and PC3, ATCC, Manassas, USA)
unnecessary to maintain sufficient O2 levels. and ovarian cancer tumors (TOV112D) or ascites (OV90)38
Under continuous perfusion, fresh oxygenated medium is were used to produce mouse xenografts. Cell suspensions
constantly supplied by convection at the surface of the tissue. were obtained after amplification in 2D cultures and mixed
In mass transport terms, the effect of perfusion is equivalent with Matrigel® (BD Biosciences, Franklin Lakes, USA) before
to an increase of the apparent diffusion constant of O2 in the being subcutaneously injected into severely combined immu-
medium. In the extreme case of infinitely fast perfusion nodeficient NOD SCID male mice (Charles River Develop-
around the tissue, mass transfer occurs across an infinitely ment, Burlington, USA) for 22Rv1 and PC3, or female mice
thin boundary layer, which is equivalent in our model to set- for TOV112D and OV90 cells lines. Solid tumors were formed
ting medium diffusion constant to infinity (DM → ∞). Setting and harvested after growth periods varying from 21 to 70 days
this condition in eqn (2) yields an expression for the critical depending on the cell line injected. All protocols involving ani-
diameter (2RCP) in perfectly perfused conditions: mals were reviewed and approved by the Comité institutionnel
de protection des animaux (CIPA) at the CRCHUM.

(3) Patient tissues

Patient tissue specimens were collected from patients follow-
thus, we conclude that perfusion would only increase the crit- ing informed consent from the Centre hospitalier de
ical diameter slightly, i.e. by a factor up to ~ in the cur- l'Université de Montréal (CHUM), Division of Gynecologic
Oncology and the Uro-Oncology Service. They were kept on
rent example.
ice until the sectioning procedure was initiated within three
These simple calculations that only take O2 into account
hours. This part of the study involving human samples was
were used as a guideline to select optimal MDT size in our
approved by both institutional ethics committees: the Comité
experimental setup. According to these calculations, perfu-
d'éthique de la recherche du CHUM (CÉR-CHUM) and the
sion could be avoided altogether without any risk of anoxia if
Comité d'éthique de la recherche de l'École Polytechnique de
MDTs have diameters of d < 424 μm. Other nutrients such
Montréal.
as glucose are also essential to the metabolic activity of tissue
sections. To complete the analysis, the consumption of
oxygen and glucose in the PDMS microsystems was therefore Tissue sectioning procedure
studied using 3D numerical simulations (see Results To produce the MDTs, thin tumor tissue fragments (~1 mm
section). by 5 mm) were cut using a scalpel and embedded into 3.7%

Lab Chip This journal is © The Royal Society of Chemistry 2015

 View Article Online

Lab on a Chip Paper

low melting point (LMP) agarose (Thermo Fisher Scientific, the tissue was provided from a patient with high-grade serous
Waltham, USA) supplemented with 10% fetal bovine serum ovarian cancer, carboplatin (Hospira, Lake Forest, IL USA) at
(FBS, BD Biosciences), kept liquid at 40–45 °C. The agarose a concentration of 350 μM was applied directly within the
was solidified on ice for at least 30 minutes, thereby creating device. Treatment was initiated one day following the sur-
a supporting structure around the embedded tissue. A tradi- gery, renewed after one day of incubation, and removed one
tional vibratome (The Vibratome Company, St. Louis, USA) day later (i.e. three days after surgery). Two independent
was used to produce 300 μm-thick tissue slices, inside a 15 °C channels were treated with carboplatin and three were kept
bath containing Hank's Buffered Saline Solution (HBSS, as non-treated controls.
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

#311-516-CL, Wisent Inc., Saint-Bruno-de-Montarville, Canada)

supplemented with 10% FBS, 55 mg L−1 gentamicin (Wisent On-chip live-tissue imaging by confocal microscopy and
Inc.), and 600 μg L−1 amphotericin B (Wisent Inc.). The pro- image processing
duced slices were kept in the same solution and finally fur-
Confocal microscopy was employed as an endpoint assay to
ther cut into disk-like MDTs using a 500 μm diameter tissue
measure the viability of samples. Multiple systems were thus
punch (Zivic Instruments, Pittsburgh, USA). The final product
analyzed at different time points to reflect the evolution of
was a cylindrical MDT of approximately 300 μm in height
cell viability over the whole 8-day incubation period. Dual
and 380 μm in diameter.
fluorescent staining of the MDTs was performed at different
time points using CellTracker™ Green CMFDA (CTG,
Device operation for culturing MDTs Thermo Fisher Scientific) labelling viable cells and
The microfluidic systems were first treated overnight with a propidium iodide (PI, Sigma-Aldrich) labelling the nucleic
10 mg mL−1 solution of triblock copolymer (Pluronic® F-108, acids of dead cells. Solutions of HBSS containing either CTG
Sigma-Aldrich, St. Louis, USA) in order to reduce cell adhe- (5 μM) alone or a combination of both CTG (5 μM) and PI
sion to the PDMS surfaces.39 Air bubbles were then removed (1.5 μM) were added to the systems sequentially by following
from the channels using 100% ethanol, and the devices were the medium replacement procedure. MDTs were first incu-
sterilized by applying 70% ethanol for 15 minutes. The chan- bated one hour with the CTG solution, and then an addi-
nels were thoroughly rinsed with non-supplemented sterile tional 30 minutes with both dyes. After replacing the dye
HBSS and placed in a humidified cell incubator (37 °C, 5% solutions with HBSS, the samples were imaged.
CO2, 95% ambient air). To minimize autofluorescence, the upper thresholds for
Once the MDTs were produced, the devices were removed the microscope settings (laser power and detector gains) were
from the incubator and placed under a stereoscope for the determined by first imaging non-labelled MDTs incubated
MDT loading procedure. With the inlet filled with HBSS, five under the same conditions within the microfluidic system
MDTs were collected using a 20 μL micropipette and allowed and ensuring minimal signal detection. The platforms were
to sediment to the bottom of the inlet inside the micro- held with a microscope slide holder over the 20X dry objec-
reservoir. Flow was induced within the channel by aspirating tive of an inverted Leica TCS SP5 II confocal microscope
liquid from the outlet using a micropipette and MDTs were (Leica Microsystems, Wetzlar, Germany). The labelled MDTs
carried inside the channels with the flow. By either aspirating were then imaged directly through the thin PDMS layer
or ejecting liquid from the outlet, the MDTs were steered underneath the platform in 10 μm-spaced optical slices. CTG
within the channel. Flow was stopped for 1–2 seconds when and PI were independently excited with the 488 and 561 nm
an MDT was positioned above an empty well, allowing it to laser lines and their fluorescence signal was collected in the
sediment to the bottom. Once all five MDTs loaded, the load- wavelength ranges 500–550 nm and 600–700 nm, respectively.
ing medium was rinsed two times by adding culture medium Maximum projections, which are 2D representations of the
to the channels: either OSE (#316-031-CL, Wisent Inc.) for 3D z-stack acquisitions, were computed for each MDT.
ovarian cancer or RPMI 1640 (# 350-045-CL, Wisent Inc.) for A custom-built MATLAB (MathWorks, Natick, USA) algo-
prostate cancer MDTs, both supplemented with 10% FBS, rithm was used to analyze each focal image of the acquired
gentamicin and amphotericin B and warmed to 37 °C. With MDT z-stacks based on the ratio of the area occupied by
about 20 μL of medium left in the inlet, the loaded devices CTG-labelled cells over the total area of both CTG- and PI-
were kept within a perforated and humidified box, inside the labelled cells (see details in ESI†). The reported MDT diame-
cell incubator. ters were calculated from confocal microscopy projection
One, three, six, and eight days following MDT production, images as the average of two perpendicular (horizontal and
medium was replaced in each system. All collected media vertical) diameter measurements per MDT.
was kept at 4 °C for further analysis in flow cytometry
experiments. Off-chip analysis of dissociated cells by flow cytometry
Fluorescence-activated cell sorting (FACS) was also used as
Chemotherapy treatment of patient-derived MDTs an end-point assay to measure the survival of individual cells
One set of patient MDTs was used to test the constituting the MDTs after their incubation in the platform.
chemosensitivity of the tissue to standard treatment. Since The MDTs were first labelled within the microsystems using

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

the apoptotic fluorescent dyes annexin V (3 : 100 dilution) Device characterization with 3D numerical simulations
and 7AAD (5 : 100 dilution) (PE Annexin V Apoptosis Detec- The finite element method with the commercial COMSOL
tion Kit I, BD Biosciences). The relatively high surface rough- Multiphysics® software was used to model the device and
ness of our PDMS replicas lead to reduced plasma bonding simulate both convective flow and diffusion. The device
strength, which was exploited to extract MDTs from the model was drawn to scale using the built-in COMSOL geome-
microsystems for off-chip analysis. Sharp scissors were used try drawing tools. The parameters applied to the models can
to initiate the separation of the two PDMS layers which were be found in the ESI† Table S1. Convective flow was defined
then peeled apart without affecting the position of MDTs in using the Navier–Stokes equations for incompressible flow
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

their sedimentation traps. All five MDTs from a single system while diffusion was defined using the reaction–diffusion
were individually pipetted out of their wells and pooled in equation with Michaelis–Menten kinetics. These convection
the same tube for analysis, together with all the medium frac- and diffusion differential equations were solved using,
tions collected at previous time points which were also respectively, the built-in steady-state fully coupled solver for
labelled thus ensuring that all cells were analyzed. The five laminar flow and the built-in transient segregated solver. A
MDTs from a single channel were dissociated into single cells user-defined mesh was necessary to reach a sufficiently accu-
by incubating them for 15 minutes at 37 °C in 400 μL of a rate result of flow and diffusion around smaller MDTs. A
saline solution supplemented with 0.25 mg mL−1 collagenase parametric sweep of inlet flow and MDT dimensions was
IV (#LS004209, Worthington Biochemical Corp., Lakewood, done to fully characterize operating conditions of the device.
USA). Samples were rinsed twice and reconstituted in 500 μL Maximum shear stress on the MDT, lift forces on the MDT
of buffer. Prior to the analysis by the flow cytometer (LSR- and minimum metabolite concentration in the MDT were
Fortessa, BD Biosciences), cell suspensions were passed probed for each solved parameter with the built-in compo-
through a 35 μm cell strainer (#352235, Corning Inc., nent coupling functions.
Corning, USA). Some MDTs were submitted to the same treat-
ment, but without the staining step. They were used, together
with positive controls, to set the detector levels and thresh- Results
olds in the annexin V and 7AAD fluorescent channels. The
data from each acquisition was analyzed using FlowJo (FlowJo Microfluidic device design
LLC, Ashland, USA) by gating the cell population in the front Channel configuration. Each 100% PDMS platform is
scatter/side scatter (FSC/SSC) graph, removing doublets, and made up of two fluidic levels: the top level where the samples
associating each cell to one of three populations according to circulate through channels to their traps and the bottom level
its fluorescent labelling: early apoptotic cells (annexin composed of square-bottom wells where the samples sedi-
V-stained only), late apoptotic or dead cells (double stained ment (Fig. 2A–C). Five independent channels fit on a 2.5 cm
with annexin V and 7AAD), and live cells (non-stained). by 7.5 cm surface (equivalent to that of a standard glass
slide), making it possible to trap up to 25 individual MDTs
and to submit them to five different treatment conditions.
Culture of large tissue fragments The distance between the traps ensures that each MDT has
For comparison, xenograft tissue was also cut to larger sec- access to a maximum amount of nutrients in non-perfused
tions of approximately 8 mm3, named large tissue fragments conditions and the total channel volume of about 30 μL
(LTFs). Nine to ten LTFs were cultured in a 60 mm Petri dish makes it possible to control the samples using a 20 μL micro-
(#83.3901, Sarstedt, Nümbrecht, Germany) in 12 mL of the pipette. Each channel is laid out in a serpentine fashion and
same medium as the one used for MDTs (as defined above). can be viewed entirely in the field of view of a low magnifica-
After 30 minutes of incubation, medium was changed and tion stereoscope. The channel cross-section is 600 μm by
samples were kept in a cell incubator (37 °C, 5% CO2). 600 μm. The gravitational square-bottom traps are 600 μm in
Medium was changed at the same intervals as for MDTs. At width by 500 μm in height. These dimensions have been
specific time points, each LTF (or a group of five MDTs) was tested to accommodate disk-like tissue samples with diame-
dissociated into single cells by incubating it for 15 minutes ters of 381 ± 47 μm (mean (μ) ± standard deviation (σ),
at 37 °C in 400 μL of a saline solution supplemented with Fig. 2H) by 300 μm in height.
1 mg mL−1 of collagenase crude (#C9407, Sigma-Aldrich) and Loading of MDTs. With average sphericities of 0.87, cylin-
0.25 mg mL−1 of collagenase type 1A (#C9891, Sigma-Aldrich). drical MDTs of reproducible sizes were obtained by produc-
The cell suspensions were then rinsed twice, reconstituted in ing 300 μm-thick tissue slices using a vibratome and by fur-
100 μL of binding buffer, and stained with the apoptotic fluo- ther micro-dissecting the slices using a biopsy punch
rescent dyes annexin V (3 μL) and 7AAD (5 μL) (PE Annexin V (Fig. 2D). Tissue diameter fluctuations could in part be due
Apoptosis Detection Kit I, BD Biosciences). Volumes were to variable tissue elasticity. For the MDTs to enter the chan-
finally brought up to 500 μL before the samples were ana- nels, they were deposited at the bottom of the inlet, inside
lyzed by flow cytometry, as detailed above. MDTs used for the microreservoir, and flow was induced by aspirating fluid
this comparison were prepared for FACS analysis under the from the outlet (Fig. 2E). Samples located in the micro-
same conditions as described in this section for LTFs. reservoir (Fig. 2E, inset) were submitted to a higher flow

Lab Chip This journal is © The Royal Society of Chemistry 2015

 View Article Online

Lab on a Chip Paper

Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

Fig. 2 Microfluidic device design and loading of MDTs. A) Top view of a microfluidic device showing five independent channels (black contours)
of 78 mm in length, each containing five equally-spaced microfluidic traps (red); dimensions in mm. B) Side view schematic of a MDT in the loading
channel near a sedimentation trap; dimensions in mm. C) Picture of an assembled device made of two PDMS layers and five inlet reservoirs. D)
MDT micro-dissection technique: a scalpel is used to form thin tissue fragments (a) that are then embedded in a LMP agarose matrix (b and c), a
vibratome is used to produce 300 μm-thick slices (d), and the tissue slices are further cut into disk-like samples using a 500 μm biopsy punch (e).
E) Procedure to load MDTs into the channel using a micropipette to induce fluid flow and position MDTs above traps where they sediment. F)
Top-view picture of a microfluidic device loaded with a MDT in each of the five square traps; scale bar: 2 mm. G) Close-up view of traps loaded
with MDTs; scale bars: 100 μm. H) Diameter distribution of 519 MDTs; average: μ = 381 μm, standard deviation : σ = 47 μm.

velocity than if they were positioned elsewhere in the inlet, square wells. Two conditions must be met for gravitational
which facilitated their entry into the microchannels. To trap trapping to succeed: the MDT needs a positive differential
the MDTs, flow was again induced in the desired direction. density compared to the surrounding medium and the sedi-
Under a stereoscope, a user observed the tissue samples cir- mentation time into the trap has to be shorter than the trav-
culating within the device and manually positioned them elling time over the trap. The equation describing the steady-
above their respective wells where they were trapped by sedi- state velocity at which a sphere sediments in an infinite fluid
mentation (Fig. 2E). This method was very effective at compartment (vinf) is derived from Newton's second law by
avoiding multiple MDTs getting trapped in the same well considering the drag force opposed to the gravity force
since the user precisely controlled the flow with the micropi- exerted on a sphere:40
pette, accelerating, decelerating, or reversing it as needed.
Complete loading of a channel with five samples, as shown (4)
in Fig. 2F, was generally accomplished in less than a minute
(ESI† Video S1). where Δρ is the particle to medium differential density, g is
Gravitational trapping mechanism. The MDTs, once posi- the gravitational acceleration, d is the particle diameter, and
tioned above a trap, are gravity-driven into the microfluidic η is the fluid viscosity. The differential density (Δρ) of MDTs

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

of known sizes was determined experimentally with eqn (4) and considering a trap of finite depth, MDTs sediment to the
by timing MDT sedimentation in a large tube with diameter bottom of the wells – over a total distance of 500 μm – in less
dtube ≫ d. An average experimental relative density of 21 ± than two seconds.
6 kg m−3 was found (ESI† Table S2). Larger spheres of same As shown in Fig. 3B, under a steady flow rate, MDTs are
density sediment more rapidly (vsed ∝ d 2), which explains subjected to drag forces pinning them in the upstream por-
why tissue samples (d ~ 380 μm) would settle about 1500 tion of the well and exerting a lift force upon them. The mag-
times faster than single cells (d ~ 10 μm) in an infinite com- nitude and direction of the resulting force is a function of
partment, making this trapping method more efficient for tumor, channel, and well dimensions.44 Their position at the
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

particles of larger diameter. bottom of the wells is stable unless the net vertical hydrody-
However, since the width (w) of the microfluidic traps is namic force (lift) on the MDT exceeds gravitational trapping,
comparable to the diameter of the MDTs, the walls signifi- thus ejecting it from the well. In general, exceeding critical
cantly slow down the sedimentation process by adding a drag flow rates may lead to either ejection of the MDT or high
component to the movement of the particle.41–43 This effec- shear stresses likely to cause cellular damage. Fig. 3C shows
tively reduces the sedimentation velocity of MDTs by a factor that, for tissue sizes within our experimental distribution
of 6 to 16 depending on MDT size based on our numerical (Fig. 2H), MDTs are ejected from the trap before being
simulations which consider a falling sphere in a bottomless subjected to high shear stresses. Relatively high flow rates
long square trap (see Fig. 3A inset). This retardation factor are necessary to reach these critical values since flow penetra-
(vsed/vinf) can be precisely accounted for using a quadratic fit tion in the trap is minimal, causing the fluid velocity around
(Fig. 3A). Nonetheless, taking into account these wall effects the sample to remain under 10% of that in the channel.

Fig. 3 Device characterization using 3D numerical simulations. A) Wall effect velocity correction ratios for MDTs of different dimensions
sedimenting in square cross-section traps of infinite length (inset). Quadratic curve fit: vsed/vinf = 1.15 (d/w)2 – 2.13 (d/w) + 1. B) Streamlines, gravi-
tational force on MDTs (red arrow – negligible for (b)), hydrodynamic force on MDTs (black arrow) and fluid velocity colormap in m/s at flow rates
(Q) provoking particle lift (a) and inducing shear stress superior to 1 Pa (b). C) Critical flow rate leading to MDT ejection from the wells (blue dashed
line) or inducing shear stress on MDT surface superior to 1 Pa (green solid line) for different MDT diameters. D) 3D simulations of O2 and glucose
consumption by an average sized MDT (381 μm) showing a cross-sectional view of the distribution of O2 at steady state (a), glucose after 24 h (b)
and glucose after 48 h (c) through the system. E) Metabolite concentration along the y-axis (as defined in D) inside a trapped MDT of average size
(381 μm) normalized to initial O2 and glucose concentration in culture medium. Michaelis–Menten (Km) constants are shown for O2 (grey dotted
line) and glucose (grey dashed line). F) Time (t) before glucose concentration [glucose] falls below Km for different MDT diameters.

Lab Chip This journal is © The Royal Society of Chemistry 2015

 View Article Online

Lab on a Chip Paper

Mass transport characterization in non-perfused conditions Long-term survival of non-treated xenograft-derived MDTs
Diffusion is the main transport mechanism in vivo between As a first step to demonstrate the effectiveness of our systems
capillaries and the surrounding tissue. Similarly, within the to trap and maintain live MDTs, we assessed the survival of
microfluidic platform, diffusion between the channel and non-treated MDTs obtained from four different types of
traps allows the tissue samples to access nutrients, drugs and mouse xenografts derived from two human prostate cancer
other reagents provided through the medium, and to dispose (22Rv1 and PC3) and two human ovarian cancer (OV90 and
of their cellular waste. When fresh medium is added through TOV112D) cell lines. Over a period of 1 to 8 days, viability
the channels, only the layers of fluid near the channel are was evaluated either by measuring the fluorescence of cells
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

renewed instantly whereas the solution within the traps labelled with viability and death fluorophores (CTG and PI
mostly recirculates. Gradually, nutrients from the channel respectively) through confocal microscopy or by dissociating
diffuse into the traps and waste molecules diffuse out. Since MDTs into single cells for FACS analysis of apoptosis (Fig. 4).
diffusion times depend on the size of the molecules, small Comparing MDTs to a larger format of tissue samples
molecules such as ions diffuse rapidly in and out of the traps (LTFs), our results obtained by FACS demonstrate that MDTs
whereas larger molecules such as glucose take more time maintain significantly higher viability over several days
(~3.5 minutes). (Fig. 4B). Considering that LTFs and MDTs were cultured in
Since the O2 saturation level in medium is low, an alter- similar volumetric medium to tissue ratios, this supports our
nate influx of O2 is necessary to ensure that non-perfused hypothesis that sectioning tissue to dimensions below the
samples have enough O2 for normal cell metabolism. In our critical diameter for adequate tissue oxygenation, as defined
case, the gas-permeable PDMS walls of the system provide in the theory section, is important to maintain high tissue
this influx. In steady-state, our simulations show that O2 survival under untreated conditions. We performed some pre-
levels are ample within our microfluidic device to maintain liminary cell proliferation experiments by flow cytometry
samples viable (Fig. 3D and E). (data not shown) indicating that there is cell proliferation
However, other nutrients have no external influx and cellu- ex vivo on chip over a period of 8 days, which might explain
lar waste accumulates through the system. Consequently, the viability recovery trend observed in Fig. 4B.
medium needs to be changed frequently in non-perfused Based on our confocal microscopy results, the xenograft
conditions. Glucose, for example, is continuously consumed MDTs remain highly viable throughout the experimental
by the MDTs and may be depleted if the system is left period, with average viabilities above 60% for all types of
unattended for too long. According to our simulations, we xenografts. Some types of MDTs show a trend of increasing
show that although glucose is abundant after 24 hours, its con- viability from day 1 to day 8. We suspect that in vitro cell pro-
centration falls below the Michaelis–Menten constant (Km) liferation post dissection might contribute to this effect.
after 35 hours for average-sized MDTs (Fig. 3F). This Km value Some dead cells might also slowly shed from the tissue, leav-
indicates the concentration at which the Michaelis–Menten ing a higher proportion of live cells in the remaining MDT.
uptake of a nutrient is reduced to half its maximum rate. As Given certain limitations of confocal microscopy with
this rate keeps dropping due to the finite amount of glucose, MDTs, in particular the imaging depth limited to about
cells slowly transition from proliferating to quiescent or dying 50 μm, the survival of the tumor tissue samples was also
states.45 Since the cell-cycling process is slow (taking up to studied using flow cytometry after labelling the MDTs with
1 day),46 changing the medium at 2 to 3 days interval is neces- annexin V and 7AAD, and dissociating them into single cells.
sary to avoid glucose deprivation possibly leading to cell death. As shown in Fig. 4C, initial viability was low for PC3 and
Fig. 3F also shows that smaller MDTs take a longer time highly variable for TOV112D, but higher viability values were
than larger ones to deplete a similar amount of glucose regained over time. This reduced cell survival was probably
within the tissue. The inversely proportional scale between due to the stress induced by the tissue sectioning steps. We
time to reach Km and MDT volume can be derived by solving speculate that certain types of xenografts might be more sen-
eqn (1) for glucose transport (see ESI† for details). sitive to mechanical stress and might take longer to recover
As shown in Fig. 3D and E, due to tissue tridimensional- from the sectioning procedure. Nonetheless, for all types of
ity, natural gradients of nutrients and waste exist within the xenografts, viabilities above 50% and sometimes as high as
MDTs from the outer surface to the center. The device asym- 85% were consistently measured after only 3 days of incuba-
metry along the y axis from the bottom of the trap to the top tion and were maintained up to day 8.
of the channel causes a decentralization of the minimum Differences between the two viability analysis techniques
nutrient concentration position within the tissue. This mini- might be explained by the fact that FACS provides a more
mum is slightly displaced towards the bottom of the sample complete representation of all the cells within MDTs, includ-
for glucose, since glucose comes from the channel. In con- ing centrally-located cells that are not detected with the
trast, it is displaced towards the top for O2, since the diffu- employed confocal microscopy technique and cells washed
sion constant of O2 through water is slightly inferior to that out of the microsystems that are not accounted for by confo-
through PDMS (ESI† Table S1) and therefore, the more abun- cal microscopy. In addition, early apoptotic events are mea-
dant source of O2 is from the bottom of the device. sured by FACS and thus the two techniques will not yield

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

Fig. 4 Long-term viability of non-treated MDTs analyzed by confocal microscopy and by flow cytometry. A) Maximum projection images of con-
focal optical slices showing examples of MDTs labelled with viability dyes – CTG (live, green) and PI (dead, red) – and corresponding viability score
calculated by the automatic image segmentation algorithm. B) Live cell fraction as a function of time, measured by FACS, for LTFs (dashed line)
compared to MDTs (solid line) formed from different xenografts pooled in the same graph. Error bars: standard error of the mean for four indepen-
dent experiments. *p-value < 0.05 for the t-test comparing values for LTFs to those for MDTs. C) MDT survival over an 8 day incubation period for
four types of xenografts, analyzed by confocal microscopy (black lines: viable) and by flow cytometry (bar graphs, green: viable, grey: early apopto-
sis, red: dead or late apoptosis). The results for an overall total of 427 MDTs from 26 xenografts are represented. Error bars: standard error of the
mean for at least two independent experiments.

identical results and should therefore be regarded as tissue sample (Patient #4). Due to the small amounts of tis-
complementary. sue available, only one type of assay was performed on these
Overall, the MDT viability results from both confocal samples at two or three different time points. Confocal
microscopy and flow cytometry suggest that the MDTs' inte- microscopy viability results are shown in Fig. 5A and B. A
grity is preserved within the current device design, as shown majority of live cells were detected for all types of tissues
by equal or increasing viability over the analysis period of up throughout the incubation period extending for up to eight
to 8 days, across all types of xenografts tested. The viability of days. Different tissue morphologies and staining patterns
the xenograft MDTs was deemed sufficient for extended were observed in MDTs between patients, but also in MDTs
experimentation such as chemotherapeutic testing. from a same patient (Fig. 5B). Only a pale staining was accom-
plished in many of the MDTs from Patient #4, which could be
attributed to a slower metabolism of this benign tissue.
Long-term survival of non-treated MDTs derived from These tests confirm that different types of patient tissues
patients can be sectioned into MDTs and maintained alive for several
With an ultimate objective of performing personalized assays days within our microfluidic platform by following the same
on ex vivo primary tissue from patients, we produced MDTs procedure that was optimized using mouse xenograft tissue.
from patient tissue and evaluated their survival within our
chip. We had access to a small portion of tissue from four
different patients who had undergone surgery. Different types Chemosensitivity of patient ovarian cancer MDTs to
of patient tissue were thus obtained: two ovarian cancer tis- carboplatin
sue samples (labelled Patient #1 and #2), one prostate cancer Some MDTs from one of the ovarian cancer patients was
tissue sample (Patient #3), and one non-cancer prostate additionally treated with a chemotherapy. For this particular

Lab Chip This journal is © The Royal Society of Chemistry 2015

 View Article Online

Lab on a Chip Paper

Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

Fig. 5 Viability study of MDTs from four cancer patients under non-treated conditions and study of the survival of MDTs from one ovarian cancer
patient after exposure to a 48 hour carboplatin treatment within the chip. A) Average viability score obtained by confocal microscopy for different
types of non-treated patient MDTs: two ovarian cancer (OvCa) tissue samples (circle markers, blue and grey curves), one prostate cancer (PCa) tis-
sue sample (square markers, red curve) and one benign prostatic hyperplasia (BPH) tissue sample (square markers, green curve). **Result for MDTs
stained and imaged a second time. B) Examples of confocal microscopy maximum projection images for each type of tissue (as defined in A) at dif-
ferent time points. Scale bar: 100 μm. **MDT stained and imaged for a second time. C) Average viability score obtained by confocal microscopy
for non-treated MDTs and for the treated ovarian cancer MDTs from Patient #2. Error bars: standard error of the mean across all MDTs exposed to
a same condition. *p-value = 0.014 for the t-test comparing results for treated MDTs to those for control MDTs. D) Representative confocal
microscopy maximum projection images of control MDTs compared to carboplatin-treated ones. Scale bar: 100 μm.

patient (Patient #2 in Fig. 5), a total of 25 MDTs were loaded samples and using other analysis techniques that may better
into a 5-channel device. Three of these channels were used as capture the effects of chemotherapies.
non-treated controls while the two remaining channels were
treated for 48 hours (from day 1 to day 3) with carboplatin at Discussion
a concentration of 350 μM, which is equivalent to the maxi-
mum theoretical blood concentration of the drug in a normal Our micro-sectioning technique produces viable submillime-
patient treated with a 360 mg m−2 dose47 (see ESI† for details ter tissue sections of reproducible dimensions (Fig. 2). It has
of the calculation). At day 3, after removing the chemother- been thoroughly validated using a total of 24 xenografts
apy, significantly lower viability was detected in the treated formed from different types of human prostate and ovarian
channels compared to the controls (Fig. 5C and D). Some cancer cell lines (Fig. 4) and four different ex vivo tissues
MDTs appeared to be less affected by the treatment (Fig. 5D, from patients (Fig. 5). The technique has also been validated
carboplatin-treated, 3rd MDT), and some regions within a with four additional types of mouse xenografts (data not
single MDT also seemed to respond differently (Fig. 5D, shown). About 300 MDTs could typically be produced from a
carboplatin-treated, 1st and 2nd MDTs), which might be 0.1 cm3 xenograft tissue fragment and loaded into micro-
attributed to a variable chemoresponse of different cell sub- fluidic platforms in less than six hours by a team of three
populations within the tumor tissue, as ovarian tumors are people. Our approach differs from the technique developed
known to exhibit high intra-tumoral heterogeneity.48,49 by Jahnke et al.50 to form tumor micro-fragments since our
As shown by using carboplatin in this experiment, the MDTs are immediately captured in a microfluidic device after
platform allows for treatments to be administered to the sectioning rather than being incubated on a gyratory shaker
MDTs via microchannels, and the effects of such treatments for 48 hours and incubated in 48-well plates afterwards. In
could be assessed through confocal microscopy. These steps our setup, the use of a vibratome together with a biopsy
still need to be validated with a larger pool of patient punch offers better control on tissue size. MDTs can also be

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

exposed to treatments earlier, leaving less time for undesired designed to capture preformed spheroids11,12 could also be
cell adaptation to the in vitro culture conditions. employed with MDTs, but we found that their resistive trap-
With all dimensions below 500 μm, MDTs present a num- ping mechanism was sensitive to perturbations provoked by
ber of advantages that can be exploited in cancer research. normal handling of the platform, causing samples to be
Their size range is close to the dimensions of the viable por- ejected from their traps. Trapping by sedimentation in
tion of cylindrical human lung tumors21 and just slightly square-bottom wells, as described here, offers superior sam-
below the calculated critical diameter (d < 424 μm) to avoid ple stability while also shielding the samples from excessive
anoxia in spherical samples (see Theory section) such that shear stress (Fig. 3A–C) and preserving their spatial orienta-
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

MDTs have access to enough nutrients to remain mostly via- tion for imaging purposes.
ble under untreated conditions while also mimicking natural Other groups have cultured pieces or slices of primary tis-
gradients of nutrients and waste occurring in tumors sue in large compartments which were continuously perfused
(Fig. 3D–F). Their 3D structure reproduces several aspects of via microfluidic channels.16,18,19 Since MDTs do not need to
a tumor which are associated to reduced effectiveness of drug be perfused, our systems are simpler to operate and more
therapy, such as the limited delivery of drugs to their center, conditions could be tested in parallel on small amounts of
and varying microenvironments which may affect the metab- tissue. An elegant approach has also been proposed by Chang
olism and proliferation of tumor cells.51 Their relatively small et al.17 to expose different regions of a single tissue slice to
size also facilitates their manipulation within microsystems multiple conditions using only one pump. However, in their
and eliminates the necessity to continuously perfuse the tis- current design, tissue samples need to be imaged off-chip
sue. Compared to patient-derived tissue slices, a greater num- and their approach is not directly compatible with off-chip
ber of MDTs can be generated from small amounts of patient FACS, generally considered the gold-standard in cell analysis.
tissue. However we did note in some instances that the We have shown, using a high-grade serous ovarian cancer
mechanical properties of the human tissue made them less patient tissue sample, that a positive response to therapy
amenable to this approach than xenografts. The micro- could be measured using our approach (Fig. 5C and D). Inter-
dissection process also consistently yielded fewer MDTs com- estingly, during clinical follow-up, the patient who received
pared to xenografts of the same size, leaving room for carboplatin-taxol adjuvant chemotherapy (additional to a
improvement in future studies. neoadjuvant regimen prior to surgery) was identified as sensi-
MDTs thus hold the potential to test multiple conditions tive to treatment, so the positive response measured in vitro
in parallel on rare tissue biopsies, with controls in close prox- within the microfluidic chip is concordant with the clinical
imity to the drug-tested regions. A sampling effect is inevita- response of the patient.
bly associated to our procedure, the same way it is inherent The microfluidic platform we developed could be used in
to any technique, such as core-needle biopsies52 or tumor several types of assays spanning from fundamental research
micro-arrays (TMAs),53 aiming to maximize the amount of to clinical drug testing. The tissue within the platform could
information obtained from small volumes of tissue. This be imaged directly through the transparent PDMS bottom
effect is nonetheless offset in our systems by the analysis of layer or through a modified version of the device comprising
five MDTs per channel, each MDT representing a different a coverslip window, which enables any microscopy technique
sub-region of the initial specimen. to be employed to investigate tumor behavior. Medium frac-
MDTs share several aspects with spheroids, such as a 3D tions could also be withdrawn from the outlet of the chan-
structure, size, and ease of manipulation on chip. Although nels to detect detached cells or components secreted by the
spheroids could be generated in larger numbers using auto- tumors. The platform could be used not only to measure the
mated methods, primary cells only rarely aggregate as spher- effects of chemotherapy on patient tissue, but also to study
oids, making the method an unlikely candidate for personal- the effects of other treatment strategies on 3D tissue in con-
ized medicine. Recent research has shown that spheroids ditions that are closer to the in vivo setup.
could be formed from patient colorectal and urothelial pri-
mary cancerous tissue with very high success rates,54,55 but Conclusions
some cellular fractions need to be discarded through the pro-
cess and it remains uncertain whether the technique will be Our method leads to whole new possibilities for the study of
as successful for other types of tumors. By keeping the extra- patient tissue within microfluidic systems. By preserving the
cellular matrix intact, we expect MDTs to better preserve all cell composition and organization of the original tissue while
cellular populations present in the original tumor, including also mimicking the micro-environment of tumors, MDTs
stromal, immune, and heterogeneous subgroups of cancer- trade off some simplicity for increased biological relevance
ous cells. Also, our technique is not dependent on the ability compared to the most common in vitro tumor models: mono-
of cells to self-aggregate, which makes it promising for all layer cell cultures and spheroids. Our microfluidic platform
types of solid tumors and other non-cancerous tissues. is nonetheless operated using simple instruments typically
Manipulating and tracking MDTs in a multiwell plate or found in any cell biology laboratory, which allows scientists
in large reservoirs typically used for organotypic tissue slices to concentrate on the complex biology of the cancerous tissue
would be impractical. Some microfluidic systems that were and leaves place for on-chip integration of additional

Lab Chip This journal is © The Royal Society of Chemistry 2015

 View Article Online

Lab on a Chip Paper

components. Our highly personalized technique can provide 10 L. Y. Wu, D. Di Carlo and L. P. Lee, Biomed. Microdevices,
information on the drug-response profile of each patient and 2008, 10, 197–202.
shows potential to improve the results of clinical trials by 11 T. Das, L. Meunier, L. Barbe, D. Provencher, O. Guenat, T.
associating patients to the best test group, and to increase Gervais and A.-M. Mes-Masson, Biomicrofluidics, 2013, 7,
patient survival and quality of life by prescribing the most 011805.
efficient treatment earlier during the course of the disease. 12 J. Ruppen, L. Cortes-Dericks, E. Marconi, G. Karoubi, R. A.
Schmid, R. Peng, T. M. Marti and O. T. Guenat, Lab Chip,
Acknowledgements 2014, 14, 1198–1205.
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

13 B. Patra, Y. H. Chen, C. C. Peng, S. C. Lin, C. H. Lee and

This research was funded by the Québec Consortium for Drug Y. C. Tung, Biomicrofluidics, 2013, 7, 054114.
Discovery (CQDM), Prostate Cancer Canada (PCC), the Natu- 14 V. Vaira, G. Fedele, S. Pyne, E. Fasoli, G. Zadra, D. Bailey, E.
ral Sciences and Engineering Research Council of Canada Snyder, A. Faversani, G. Coggi, R. Flavin, S. Bosari and M.
(NSERC) and the Canadian Cancer Society Research Institute Loda, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 8352–8356.
(CCSRI). A.-M. Mes-Masson, F. Saad and D. Provencher are 15 I. A. M. de Graaf, P. Olinga, M. H. de Jager, M. T. Merema,
researchers of the CRCHUM, which receives support from the R. de Kanter, E. G. van de Kerkhof and G. M. M. Groothuis,
Fond de recherche du Québec – Santé (FRQS). M. Astolfi Nat. Protoc., 2010, 5, 1540–1551.
acknowledges graduate scholarships from NSERC and the 16 P. M. Van Midwoud, G. M. M. Groothuis, M. T. Merema and
Fonds de recherche du Québec – Nature et technologies E. Verpoorte, Biotechnol. Bioeng., 2010, 105, 184–194.
(FRQ-NT). M. A. Lateef was supported through a grant from 17 T. C. Chang, A. M. Mikheev, W. Huynh, R. J. Monnat, R. C.
the Canadian Uro-Oncologist Group (CUOG). The ovarian and Rostomily and A. Folch, Lab Chip, 2014, 14, 4540–4551.
prostate tumor banking was supported by the “Banque de 18 S. M. Hattersley, C. E. Dyer, J. Greenman and S. J. Haswell,
tissus et de données of the Réseau de recherche sur le cancer” of Lab Chip, 2008, 8, 1842–1846.
the FRQS affiliated with the Canadian Tumor Repository Net- 19 L.-T. Cheah, Y.-H. Dou, A.-M. L. Seymour, C. E. Dyer, S. J.
work (CTRNet). We would like to thank the uro-oncologists Haswell, J. D. Wadhawan and J. Greenman, Lab Chip,
and the gynecologic oncologists of the CHUM for providing 2010, 10, 2720–2726.
human specimens. We also thank the staff from the animal 20 T. Anada, J. Fukuda, Y. Sai and O. Suzuki, Biomaterials,
facility, in particular Kim Leclerc Désaulniers and Maxime 2012, 33, 8430–8441.
Thiffault, for technical support as well as Liliane Meunier, 21 R. Thomlinson and L. Gray, Br. J. Cancer, 1955, 9, 539–549.
Hubert Fleury, and Dr. Philip Wong for useful discussions. 22 I. F. Tannock, Br. J. Radiol., 1972, 45, 512–524.
23 W. Mueller-Klieser, Biophys. J., 1984, 46, 343–348.
Notes and references 24 K. Groebe and W. Mueller-Klieser, Int. J. Radiat. Oncol., Biol.,
Phys., 1996, 34, 395–401.
1 M. Hay, D. W. Thomas, J. L. Craighead, C. Economides and 25 Y. Jiang, J. Pjesivac-Grbovic, C. Cantrell and J. P. Freyer,
J. Rosenthal, Nat. Biotechnol., 2014, 32, 40–51. Biophys. J., 2005, 89, 3884–3894.
2 C. G. Begley and L. M. Ellis, Nature, 2012, 483, 531–533. 26 D. R. Grimes, C. Kelly, K. Bloch and M. Partridge, J. R. Soc.,
3 E. H. Rubin and D. G. Gilliland, Nat. Rev. Clin. Oncol., Interface, 2014, 11, 20131124.
2012, 9, 215–222. 27 R. Ivanova and G. Simeonov, Comput. Math. with Appl.,
4 L. G. Griffith and M. A. Swartz, Nat. Rev. Mol. Cell Biol., 2012, 64, 322–336.
2006, 7, 211–224. 28 P. Vaupel, H. P. Fortmeyer, S. Runkel and F. Kallinowski,
5 F. Pampaloni, E. G. Reynaud and E. H. K. Stelzer, Nat. Rev. Cancer Res., 1987, 47, 3496–3503.
Mol. Cell Biol., 2007, 8, 839–845. 29 R. Sander, Atmos. Chem. Phys. Discuss., 2014, 14,
6 S. Vaughan, J. I. Coward, R. C. Bast, A. Berchuck, J. S. Berek, 29615–30521.
J. D. Brenton, G. Coukos, C. C. Crum, R. Drapkin, D. 30 J. J. Casciari, S. V. Sotirchos and R. M. Sutherland, J. Cell.
Etemadmoghadam, M. Friedlander, H. Gabra, S. B. Kaye, Physiol., 1992, 151, 386–394.
C. J. Lord, E. Lengyel, D. A. Levine, I. A. McNeish, U. Menon, 31 C. J. Koch and J. E. Biaglow, Br. J. Cancer, Suppl., 1978, 3,
G. B. Mills, K. P. Nephew, A. M. Oza, A. K. Sood, E. A. 163–167.
Stronach, H. Walczak, D. D. Bowtell and F. R. Balkwill, Nat. 32 J. P. Freyer and R. M. Sutherland, J. Cell. Physiol., 1985, 124,
Rev. Cancer, 2011, 11, 719–725. 516–524.
7 M. Zietarska, C. M. Maugard, A. Filali-Mouhim, M. Alam- 33 A. Bertuzzi, A. Fasano, A. Gandolfi and C. Sinisgalli, J. Theor.
Fahmy, P. N. Tonin, D. M. Provencher and A.-M. Mes- Biol., 2010, 262, 142–150.
Masson, Mol. Carcinog., 2007, 46, 872–885. 34 R. Venkatasubramanian, M. A. Henson and N. S. Forbes,
8 J. Friedrich, C. Seidel, R. Ebner and L. A. Kunz-Schughart, J. Theor. Biol., 2008, 253, 98–117.
Nat. Protoc., 2009, 4, 309–324. 35 W. Mueller-Klieser, J. P. Freyer and R. M. Sutherland, Br. J.
9 A. Y. Hsiao, Y. S. Torisawa, Y. C. Tung, S. Sud, R. S. Cancer, 1986, 53, 345–353.
Taichman, K. J. Pienta and S. Takayama, Biomaterials, 36 J. Grote, R. Süsskind and P. Vaupel, Pflugers Archiv.—Eur. J.
2009, 30, 3020–3027. Physiol, 1977, 372, 37–42.

This journal is © The Royal Society of Chemistry 2015 Lab Chip

 View Article Online

Paper Lab on a Chip

37 P. Han and D. M. Bartels, J. Phys. Chem., 1996, 100, 5597–5602. 48 L. Khalique, A. Ayhan, M. Weale, I. Jacobs, S. Ramus and S.
38 D. M. Provencher, H. Lounis, L. Champoux, M. Tétrault, Gayther, J. Pathol., 2007, 211, 286–295.
E. N. Manderson, J. C. Wang, P. Eydoux, R. Savoie, P. N. 49 S. L. Cooke, C. K. Y. Ng, N. Melnyk, M. J. Garcia, T.
Tonin and A. M. Mes-Masson, In Vitro Cell. Dev. Biol.: Anim., Hardcastle, J. Temple, S. Langdon, D. Huntsman and J. D.
2000, 36, 357–361. Brenton, Oncogene, 2010, 29, 4905–4913.
39 M. Amiji and K. Park, Biomaterials, 1992, 13, 682–692. 50 H.-G. Jahnke, S. Poenick, J. Maschke, M. Kendler, J. C.
40 H. Lamb, Hydrodynamics, Cambridge University Press, 6th Simon and A. A. Robitzki, Cancer Res., 2014, 74,
ed., 1994. 6408–6418.
Published on 02 December 2015. Downloaded by UNIVERSITY OF OTAGO on 14/12/2015 03:44:24.

41 A. W. Francis, J. Appl. Phys., 1933, 4, 403–406. 51 O. Trédan, C. M. Galmarini, K. Patel and I. F. Tannock,
42 W. L. Haberman and R. M. Sayre, Motion of fluid spheres in J. Natl. Cancer Inst., 2007, 99, 1441–1454.
stationary and moving liquids inside cylindrical tubes, 1958. 52 C. Öbek, T. Doanca, S. Erdal, S. Erdoǧan and H. Durak,
43 J. Markiewicz and W. R. Matson, Am. J. Mod. Phys., 2013, 2, 98. J. Urol., 2012, 187, 2051–2055.
44 M. Khabiry, B. G. Chung, M. J. Hancock, H. C. 53 M. Ilyas, H. Grabsch, I. O. Ellis, C. Womack, R. Brown, D.
Soundararajan, Y. Du, D. Cropek, W. G. Lee and A. U. Berney, D. Fennell, M. Salto-Tellez, M. Jenkins, G. Landberg,
Khademhosseini, Small, 2009, 5, 1186–1194. R. Byers, D. Treanor, D. Harrison, A. R. Green, G. Ball and P.
45 R. Venkatasubramanian, M. A. Henson and N. S. Forbes, Hamilton, Histopathology, 2013, 62, 827–839.
J. Theor. Biol., 2006, 242, 440–453. 54 J. Kondo, H. Endo, H. Okuyama, O. Ishikawa, H. Iishi, M.
46 J. P. Ward and J. R. King, IMA J. Math Appl. Med. Biol., Tsujii, M. Ohue and M. Inoue, Proc. Natl. Acad. Sci. U. S. A.,
1997, 14, 39–69. 2011, 108, 6235–6240.
47 R. R. Barakat, A. Berchuck, M. Markman and M. E. Randall, 55 H. Okuyama, T. Yoshida, H. Endo, M. Nakayama, N.
Principles and Practice of Gynecologic Oncology, Wolters Nonomura, K. Nishimura and M. Inoue, J. Urol., 2013, 190,
Kluwer, 6th ed., 2013. 302–310.

Lab Chip This journal is © The Royal Society of Chemistry 2015