brief communications
© 2015 Nature America, Inc. All rights reserved.
Functional differentiation of human pluripotent stem cells on a chip Giovanni G Giobbe1,2,5, Federica Michielin1,2,5, Camilla Luni1,2, Stefano Giulitti1,2, Sebastian Martewicz1,2, Sirio Dupont3, Annarosa Floreani4 & Nicola Elvassore1,2 Microengineering human “organs-on-chips” remains an open challenge. Here, we describe a robust microfluidics-based approach for the differentiation of human pluripotent stem cells directly on a chip. Extrinsic signal modulation, achieved through optimal frequency of medium delivery, can be used as a parameter for improved germ layer specification and cell differentiation. Human cardiomyocytes and hepatocytes derived on chips showed functional phenotypes and responses to temporally defined drug treatments.
The development of human organs-on-chips, in which microscale engineering technologies enable the recapitulation of the organ microenvironment, offers a unique opportunity to study human physiology and pathophysiology1,2. Recently, successful examples of organs-on-chips have been demonstrated3–6 based on primary animal and, in a few cases, human cells. Human embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs) grown in culture have the potential to give rise to any fetal and adult cell type. Developing direct organogenesis-on-a-chip from human pluripotent stem cells (hPSCs) could overcome the limited availability of human primary cells. Furthermore, hPSC-derived cells show intrinsic and unexpected levels of emergent self-organization for generating highly ordered structures and tissues7,8. The intrinsic properties of microfluidics allow spatiotemporal control of the cell culture microenvironment and enable the biomimetic scale-down of developmental processes, from germ layer specification and phenotypic differentiation to tissue morphogenesis9. Recent studies on mouse ESC cultures using microfluidic systems highlight the importance of accurate regulation of the soluble microenvironment, in terms of endogenous and exogenous extrinsic factors, to maintain pluripotency and self-renewal10,11. We and others have reported hESC culture on a chip, modulating cell culture conditions to promote pluripotency maintenance and self-renewal, and germ layer specification12–14.
Here we report the first, to our knowledge, integrated functional hPSC differentiation on a chip. We explored whether we could control hPSC expansion, selective germ layer commitment and differentiation into functional tissue-specific cells through a multistage microfluidic-based process. This would require the optimal delivery of exogenous factors and removal of endogenous cell-secreted factors which affect transcriptional activity15. We hypothesized that periodic medium delivery with stage-dependent frequency (number of cycles of medium changes per day; referred to as ‘perfusion frequency’ or ‘f’) is an effective method for modulating soluble microenvironments and, consequently, for controlling stem cell niche specification in vitro (Fig. 1a). Clearly, low perfusion frequencies allow endogenous factors to accumulate within the channel. Conversely, high frequencies promote a sustained supply of exogenous factors but lead to continuous washout of endogenous secreted factors (Fig. 1b), as qualitatively described by a computational model (Supplementary Fig. 1 and Supplementary Note 1). We observed that discontinuous periodic medium delivery, unlike continuous microfluidic flow perfusion, yielded homogeneity along the channels in terms of colony size, cell density and pluripotency marker expression for both hESCs and hiPSCs (Supplementary Figs. 2 and 3). This method also allowed proper glucose supply and removal of waste products such as ammonia and lactate (Supplementary Fig. 4). We first investigated the optimal perfusion frequency for hPSC expansion using an automated microfluidic experimental setup (Fig. 1c) in which we observed homogeneous growth and the expression of pluripotency markers along the microfluidic channel (Fig. 1d,e). Perfusion frequency–dependent expression levels of the pluripotency markers OCT4, NANOG and DNMT3B were observed in HES2 cells after 6 d of culture (Fig. 1f), with marked differences among frequencies for OCT4 and NANOG. Specifically, significantly higher OCT4 expression was observed for f = 2 d−1 (P ≤ 0.01), compared to all other frequencies or to continuous perfusion (CP), and higher NANOG expression for f = 2 d−1, compared to the other frequencies or CP (P ≤ 0.05). To further test the correlation between optimal perfusion frequency and the accumulation of endogenous cell-secreted factors, we used HES2-conditioned medium collected in an f = 2 d−1 experiment to culture HES2 cells at a higher perfusion frequency. We observed a significant threefold increase in OCT4 expression when conditioned, instead of fresh, medium was used at f = 8 d−1 (P ≤ 0.01) (Fig. 1g). Interestingly, the conditioned medium collected at f = 2 d−1 contained higher levels of TGF-β ligands, as demonstrated by its ability to activate SMAD2–SMAD3 transcriptional activity
1Department of Industrial Engineering, University of Padova, Padova, Italy. 2Venetian Institute of Molecular Medicine, Padova, Italy. 3Department of Molecular Medicine,
University of Padova, Padova, Italy. 4Department of Surgery, Oncology and Gastroenterology, University of Padova, Padova, Italy. 5These authors contributed equally to this work. Correspondence should be addressed to N.E. ([email protected]). Received 5 March 2014; accepted 10 March 2015; published online 1 June 2015; doi:10.1038/nmeth.3411
nature methods | ADVANCE ONLINE PUBLICATION |
brief communications b
Tr Gt
. V
Ex 0
f2
f1
f3
t t
t
t
t
TRA-1-60
hiPSCs
hiPSCs
hESCs
hESCs
SOX2
OCT4
SOX2
TRA-1-81
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0
OCT4 NANOG DNMT3B
*
ic
at
St
f=
1
d
5
=
1.
d
f=
2
d
f=
3
d
f=
4
d
f=
8
d
P
C
(measured using the synthetic CAGA12-luciferase reporter in HaCaT cells16), as compared to medium collected at f = 8 d−1 or under static conditions (no change of medium during the 24 h growth period) (Fig. 1h, Supplementary Fig. 5 and Online Methods). Consistently, the expression of NANOG, a known direct transcriptional target of SMAD2–SMAD3 in hESCs17, was upregulated in the f = 2 d−1 condition when compared to higher perfusion frequency or static conditions (Fig. 1i). These data suggested that endogenous TGF-β, activin and nodal ligands were released, or accumulated to optimal levels, in an f-dependent manner to promote pluripotency. We next investigated whether perfusion frequency could also be used to direct early germ layer commitment upon spontaneous differentiation (Fig. 2a and Supplementary Fig. 6). QRT-PCR analysis on static and low perfusion frequency (f = 1 d−1 and f = 2 d−1) cultures showed mainly ectoderm marker expression (TUBB3 and OTX2), whereas mesoderm (T, GATA4 and ACTA2) and endoderm (AFP, EOMES and FOXA2) marker expression were enhanced at a higher frequency (f = 8 d−1). This observation is consistent with the hypothesis that high accumulation of extrinsic endogenous factors promotes ectoderm differentiation while inhibiting meso-endoderm specification (Supplementary Fig. 7)10,15. | ADVANCE ONLINE PUBLICATION | nature methods
Relative NANOG expression
Luciferase activity (a.u.)
f Figure 1 | hPSC culture on a chip. g h i (a) Hypothetical model of the signaling 1.8 1.6 OCT4 pathway describing the synergistic ** 1.6 ** 2,000 1.4 NANOG ** 1.4 effects of extrinsic exogenous (Ex) 1.2 * 1.2 1,500 1.0 * and endogenous (En) factors, which 1.0 0.8 0.8 1,000 affect gene transcript (Gt) expression * 0.6 0.6 0.4 mediated by target transcription 0.4 500 0.2 0.2 factors (Tr). N( f ) is the perfusion 0 0 0 –1 –1 –1 + + – – – – – – – – + – – – + TGF-β TGF-β frequency-dependent molar flux. f = 8 d f=2d f=8d – + – + – + – + – + SB431542 conditioned Reactions (1) and (2) describe the Static f = 8 d–1f = 2 d–1 Fresh medium Static f = 8 d–1f = 2 d–1 coupling of factors to membrane receptors. Reaction (3) is the basal activation of Tr. Reactions (4) and (5) describe production of Gt and En, respectively. (b) A schematic of perfusion frequency ( f )-dependent temporal profiles of normalized Ex (red lines) and En (black lines) factor concentrations, which lead to different levels of mean concentrations over time, ‘t’ (straight lines). (c) Medium delivery in the microfluidic platform is controlled by a multiport delivery system. Flow rate (V), perfusion frequency ( fn), temporal delay between cycles and total number of cycles are set independently and automatically controlled. (d) Micrographs showing hESCs and hiPSCs, both expanded up to 6 d with f = 2 d−1 (inset). Scale bars, 300 µm. (e) Immunofluorescence analyses of pluripotency markers on the same cells as in (d). Scale bar, 50 µm. (f) qRT-PCR analysis of pluripotency markers in HES2 cells cultured under the indicated conditions. CP, continuous perfusion. Data are normalized to GAPDH and shown as mean ± s.e.m. (n = 6); ANOVA *P ≤ 0.05, **P ≤ 0.01. (g) qRT-PCR analysis of OCT4 and NANOG from HES2 cells cultured with fresh or conditioned medium at the indicated perfusion frequencies. Data are normalized to GAPDH and shown as mean ± s.e.m. (n = 6); t-tests **P ≤ 0.01. (h) Luciferase assay (based on CAGA12 SMAD2–SMAD3 cells) measuring the presence of TGF-β, activin and nodal ligands in conditioned medium from HES2 cells cultured under the indicated perfusion conditions. Data are normalized to total protein content (Bradford assay) and shown as mean ± s.e.m. (n = 6); t-tests **P ≤ 0.01. (i) NANOG expression by qRT-PCR analysis of HES2 cells. Data normalized to GAPDH and shown as mean ± s.e.m. (n = 6) t-tests *P ≤ 0.05. See statistical section of Online Methods for definition of n. Relative gene expression
© 2015 Nature America, Inc. All rights reserved.
Waste
**
–1
f OCT4
–1
e
Relative gene expression
d
Multi-port delivery system
Automatic control
f
–1
En
En
–1
Ex
OUT NEn (f)
–1
(3)
(1)
(2) (5) (4)
NEx (f)
Concentration
OUT
IN
NE x (f)
c 1
–1
a
To further test whether selective induction of the three germ layers could be achieved, we induced hPSCs to form ectodermal, mesodermal or endodermal cells by growing them in specific differentiation medium (containing germ layer–specific cytokines and factors) with optimal perfusion frequencies. As an example, the perfusion frequency optimization for microfluidic induction of mesodermal cells is shown in Figure 2b. Similarly, perfusion frequencies of 1, 2 or 3 d−1 were found to be optimal for the homogeneous differentiation of ectoderm, mesoderm and endoderm, respectively, as measured by qRT-PCR and immunofluorescence analyses (Fig. 2c,d and Supplementary Fig. 8). For endoderm commitment, we also observed significantly higher expression of the endoderm marker AFP when HES2 cells were differentiated under microfluidic conditions compared to conventional static differentiation (P ≤ 0.05). Consistently, optimized microfluidic conditions led to a significant threefold increase in SOX17+FOXA2+ cells (in 3 out of 6 conditions) compared to Petri dish culture (Supplementary Fig. 9). Collectively, our findings indicate that perfusion frequency is an additional parameter that can enhance germ layer specification, even when specific differentiation protocols are used. We then examined whether our strategy of controlled perfusion frequency could be used to achieve functional differentiation of cells of interest on chips. We first focused on cardiac cells (Fig. 3,
Ex factors
TUBB3 OTX2 T GATA4 ACTA2 AFP EOMES FOXA2
0.6 0.4 0.2 0 Petri dish
f = 1 d–1
TUBB3 T AFP
f = 2 d–1
f = 8 d–1
*
0.8 0.6 0.4 0.2 0 Ectoderm Mesoderm Endoderm Ectoderm Mesoderm Endoderm
Static
Supplementary Fig. 10 and Online Methods). We typically obtained 5,000–10,000 cardiac cells per channel after 14 d of cardiac differentiation (Fig. 3a) from HES2 and ADHF#1 cells. We observed spontaneous contractile activity from both hESC- and hiPSC-derived cardiomyocytes (Supplementary Videos 1–3). Overall, 65 ± 11% of cells were positive for cardiac troponin T (CTNT), and these also exhibited defined cardiac sarcomeric organization (Fig. 3b). hPSCderived cardiomyocytes showed spontaneous calcium transients (Fig. 3c) and excitation-contraction coupling. We also observed functional responses to 0.5 µM verapamil, with cells showing reduced calcium release after L-type channel inhibition, and to 10 mM caffeine, with cells showing increased cytosolic calcium after ryanodine channel activation (Fig. 3c). We next applied our methodology to derive hepatocytes on chips from hESCs and hiPSCs (Fig. 3d) using the previously endoderm-optimized conditions (Supplementary Fig. 9). After 14 d of differentiation we observed characteristic polygonalshaped hepatocyte-like cells, including binucleate cells, on the chip. These cells were positive for the hepatic markers albumin and cytochrome P450-3A (CYP-3A) (Fig. 3e), and showed albumin secretion, glycogen storage and indocyanine green digestion (Fig. 3f). Our optimized perfusion conditions yielded a 40% greater albumin secretion at the end-point of differentiation and a greater percentage of ALB and CYP-3A positive cells (Fig. 3g,h) in comparison to cells differentiated under static conditions. We observed robust differentiation of eight hiPSC lines obtained by different methods using the microfluidic strategy we defined (Supplementary Figs. 11–13 and Online Methods). We also used our approach to derive hepatocyte-like cells on chips from genetically corrected α1-antitrypsin–deficient hiPSCs18 (Supplementary Fig. 11), opening up new possibilities for disease modeling on chips and for patient-specific drug design. To test whether differentiated cells on a chip could be used to monitor cellular drug response, we performed a series of experiments monitoring acetaminophen-induced cytotoxicity. We chose
Microfluidic
**
1.0
**
0.8 0.6 0.4 0.2 0 f = 1 d–1
d
f = 2 d–1
f = 3 d–1 f = 12 d–1
Markers Mesoderm Ectoderm
0.8
1.0
T
Specific differentiation protocol
Concentration
En factors
1.0
c
b
High f
Relative gene expression
Intermediate f
Low f
Ectoderm β-III TUB
Mesoderm BRA
Endoderm AFP
Endoderm
a
Relative gene expression
Figure 2 | Germ layer induction of hESCs-ona-chip. (a) Top, schematic of the relationship between the perfusion frequency, f, and the concentrations of En and Ex factors. Bottom, qRT-PCR analysis for the indicated markers for ectoderm (TUBB3 and OTX2), mesoderm (T, GATA4 and ACTA2) and endoderm (AFP, EOMES and FOXA2) from spontaneously differentiated HES2 cells on chips, at the indicated perfusion frequencies. Phase contrast images showing the cell morphology for each differentiation condition is shown beneath the graph. Scale bar 50 µm. Data normalized to GAPDH and shown as mean ± s.d. (n = 6). (b) qRT-PCR analysis of T expression at different perfusion frequencies for mesoderm-induced HES2 cells. Data normalized to GAPDH and shown as mean ± s.d. (n = 6); ANOVA **P ≤ 0.01. (c) qRT-PCR analysis for the indicated markers of the three germ layers in HES2 cells differentiated under the indicated conditions. Data normalized to GAPDH and shown as mean ± s.d. (n = 6); t-tests *P ≤ 0.05. (d) Immunofluorescence analysis of germ layer-committed HES2 cells. Scale bar, 50 µm. See statistical section of Online Methods for definition of n.
Relative gene expression
© 2015 Nature America, Inc. All rights reserved.
brief communications
acetaminophen because it is mainly metabolized in the liver through CYP-3A activity19 and causes hepatocyte toxicity at high doses (Supplementary Fig. 14). We compared cytotoxicity on hESC-derived hepatocyte-like cells in static culture and in microfluidic culture at f = 2 d−1, for the same acetaminophen concentration (25 mM). Because of the small volume of medium in microfluidic chip, the overall amount of acetaminophen per cell was 10 times lower in this system. Despite this, we observed 75% of dead cells in the micro-channels compared to 20% in the static control for a 24-h exposure (Fig. 3i). This result is consistent with higher expression of CYP-3A in cells differentiated under microfluidic conditions (Fig. 3h and Supplementary Fig. 15). The tight temporal control of medium delivery achievable by microfluidic technology makes it particularly suitable to test acetaminophen dosage (Fig. 3j). Experiments on hESC-derived hepatic cells show that, even at low doses, repeated drug administration (4 times per day for 3 h each) had a significantly higher cytotoxic impact than application of a single high dose, regardless of the higher overall amount of drug in the single-administration case20. In conclusion, we have derived functional tissue-specific cells via hPSC differentiation directly on a chip through a robust multistage microfluidic technology, and demonstrate the derivation of functional hepatocyte-like cells that may prove suitable for highthroughput drug testing. Accurate spatiotemporal control of the soluble microenvironment around cells is a critical aspect of this methodology and was achieved through regulation of periodic perfusion frequencies. This technology opens up new possibilities for generating human organs-on-chip from hPSCs. Methods Methods and any associated references are available in the online version of the paper. Note: Any Supplementary Information and Source Data files are available in the online version of the paper.
nature methods | ADVANCE ONLINE PUBLICATION |
brief communications
2s
D4–D18 Cardiomyocyte CTNT
80
Positive cells (%)
Brachy
Static Microfluidic
100
0.5 µM verapamil
10 mM caffeine
Fluo-4 (F/F0)
D0–D3 Mesoderm
h
c
Hoechst CTNT
Fluo-4 (F/F0)
b
OCT4 NANOG
Fluo-4 (F/F0)
a
60
**
*
40 20 0
35
PAS staining – glycogen
9
0.8 Dead cells fraction
25 20 15 10 5
AM
12
Static Microfluidic
15
**
18 PM
6 3
**
21 24
0.6
CYP-3A
Midnight
**
0.4 0.2
j
4
0
d1
d6
d1
0 d1
2
6h clearance
d1 6
0
d6
0h uptake
ALB
Noon
1.0
**
30
ICG
CK-18
i
+ ctr Static Microfluidic
G
ALB CYP-3A
Hoechst CYP-3A
g
0 Ctr
1
5
10
25
Acetaminophen conc. [mM] Noon
Noon Noon Noon Figure 3 | Cardiac and hepatic cell differentiation on a chip. 12 12 12 12 9 15 9 9 15 15 9 15 (a) Microfluidic cardiac differentiation protocol. (b) Immunofluorescence 18 PM 6 AM 6 18 PM AM 6 18 PM AM 18 PM AM 6 of hESC-derived cardiomyocytes (CMs) showing cardiac troponin T (CTNT) 3 21 organization. Scale bars, 15 µm. (c) Spontaneous calcium transients in 3 21 3 3 21 21 24 24 24 24 hESC-derived CMs recorded with Fluo-4, under control conditions (left) Midnight Midnight Midnight Midnight ** or in the presence of verapamil (right, top) or caffeine (right, bottom). 0.3 0.3 0.3 0.3 (d) Microfluidic hepatic differentiation protocol. (e) Immunofluorescence * ** 0.2 0.2 0.2 0.2 of hESC-derived hepatocyte-like cells showing albumin (ALB) and CYP-3A expression. Scale bars, 15 µm. (f) Periodic acid–Schiff stain (PAS) 0.1 0.1 0.1 0.1 of hepatic cells; nuclei are stained in blue, pink-violet colored cells show 0 0 0 0 glycogen storage (upper panel). Lower panels, hepatic cells at 0 (left) Ctr 1 5 10 25 Ctr 1 5 10 25 Ctr 1 5 10 25 Ctr 1 5 10 25 and 6 h (right) after staining with indocyanine green (ICG). Colorant is Acetaminophen conc. [mM] absorbed and metabolized in 6 h by functional hepatocyte-like cells. Scale bars, 150 µm. (g) ELISA assay for secreted human albumin (ng/ml/day per 2 × 105 cells) from hESC-derived hepatocyte-like cells derived under the indicated conditions. Positive control (ctr) HepG2 cell line. Data are shown as mean ± s.d. (n = 6); t-tests **P ≤ 0.01. (h) Quantification of CK-18-, ALB- or CYP-3A-positive cells differentiated with the protocol reported in Supplementary Figure 10. Data are shown as mean ± s.d. (n = 10); t-tests *P ≤ 0.05, **P ≤ 0.01. (i) Fraction of dead cells upon treatment of hESC-derived hepatocyte-like cells with increasing acetaminophen concentrations in static or microfluidic conditions. Ctr, cells treated with DMSO only. Data are shown as mean ± s.d. (n = 6); ANOVA **P ≤ 0.01. (j) Fractions of dead cells upon treatment of hESC-derived hepatocyte-like cells with the indicated doses of acetaminophen, under the indicated temporal administration protocols. Ctr, cells treated with DMSO only. Data are shown as mean ± s.d. (n = 6); 2-way ANOVA *P ≤ 0.05, **P ≤ 0.01. See statistical section of Online Methods for definition of n.
Dead cells fraction
© 2015 Nature America, Inc. All rights reserved.
D9–D14 Hepatocytelike cell
Hoechst ALB
f
Albumin production
AFP
e
ep
OCT4 NANOG D0–D3 Definitive endoderm SOX17 FOXA2 D4–D8 Hepatic endoderm
H
d
Acknowledgments This research was supported by Progetti di Eccellenza Cassa di Risparmio di Padova e Rovigo (CARIPARO), Progetti di Eccellenza Giovani Ricercatori of Ministero della Salute and Progetto Strategico Università di Padova. We acknowledge Miltenyi Biotec for kindly providing the mRNA reprogramming kit. We thank L.Vallier (Cambridge University) and the Cambridge NIhR BRC hIPSC core facility for providing us A1ATD and corrected hiPS cell lines, and M. Oshimura (Department of Biomedical Science, Institute of Regenerative Medicine and Biofunction, Tottori University, Japan.) for the ADHF#1 hiPS cell line. AUTHOR CONTRIBUTIONS G.G.G., F.M. and N.E. designed the research; G.G.G. and F.M. performed the experiments; S.M. performed cardiac functional tests; S.G. helped in microfluidic platform set-up and microfluidic cell culture; S.G. and C.L performed reprogramming experiments; S.D. helped in TGF-β experiments; A.F. supervised hepatic differentiation experiments; N.E. coordinated the project; G.G.G., F.M. and N.E. wrote the manuscript. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. Reprints and permissions information is available online at http://www.nature. com/reprints/index.html.
| ADVANCE ONLINE PUBLICATION | nature methods
1. Huh, D., Torisawa, Y., Hamilton, G.A., Kim, H.J. & Ingber, D.E. Lab Chip 12, 2156–2164 (2012). 2. Ghaemmaghami, A.M., Hancock, M.J., Harrington, H., Kaji, H. & Khademhosseini, A. Drug Discov. Today 17, 173–181 (2012). 3. Huh, D. et al. Science 328, 1662–1668 (2010). 4. Lee, S.-A. et al. Lab Chip 13, 3529–3537 (2013). 5. Jang, K.-J. et al. Integr. Biol. 5, 1119–1129 (2013). 6. Grosberg, A., Alford, P.W., McCain, M.L. & Parker, K.K. Lab Chip 11, 4165–4173 (2011). 7. Sasai, Y. Nature 493, 318–326 (2013). 8. Eiraku, M. et al. Nature 472, 51–56 (2011). 9. Discher, D.E., Mooney, D.J. & Zandstra, P.W. Science 324, 1673–1677 (2009). 10. Przybyla, L.M. & Voldman, J. Proc. Natl. Acad. Sci. USA 109, 835–840 (2012). 11. Moledina, F. et al. Proc. Natl. Acad. Sci. USA 109, 3264–3269 (2012). 12. Figallo, E. et al. Lab Chip 7, 710–719 (2007). 13. Villa-Diaz, L.G. et al. Lab Chip 9, 1749–1755 (2009). 14. Korin, N., Bransky, A., Dinnar, U. & Levenberg, S. in Proc. SPIE 6416, 64160N (2006). 15. Giobbe, G.G. et al. Biotechnol. Bioeng. 109, 3119–3132 (2012). 16. Inui, M. et al. Nat. Cell Biol. 13, 1368–1375 (2011). 17. Xu, R.-H. et al. Cell Stem Cell 3, 196–206 (2008). 18. Yusa, K. et al. Nature 478, 391–394 (2011). 19. Manyike, P.T., Kharasch, E.D., Kalhorn, T.F. & Slattery, J.T. Clin. Pharmacol. Ther. 67, 275–282 (2000). 20. Watkins, P.B. et al. J. Am. Med. Assoc. 296, 87–93 (2006).
© 2015 Nature America, Inc. All rights reserved.
ONLINE METHODS Microfluidic device fabrication and functionalization. Each microfluidic device, containing ten independent channels 18 mm long and 1.5 mm wide, was fabricated by standard softlithographic techniques21. The master was photo-lithographically patterned through SU8-2100 negative photoresist (MicroChem) to obtain a final thickness of 200 µm, according to the manufacturer’s indications. A premixed 10:1 mixture of polydimethylsiloxane (PDMS) pre-polymer and curing agent solutions (Sylgard 184 kit; Dow Corning) was cast on the silicon wafer and cured at 70 °C for 2 h. The PDMS mold was cut, peeled off and punched with a 21G stainless steel needle (Small Part Inc.) to obtain inlet and outlet holes. The PDMS mold was assembled and sealed to a 50 × 75 mm cleaned glass slide by plasma bonding. Ten independent medium reservoirs (each with a 70 µL capacity), one for each channel, were obtained by sealing an additional PDMS block to the top of the device by plasma bonding. The assembled device was cleaned with isopropanol (Sigma-Aldrich) and sterilized by autoclaving. Microfluidic channels were filled with 20% Matrigel Reduced Factor (MRF; BD Biosciences) and incubated overnight at 4 °C. Before cell seeding, the protein solution was aspirated and the microfluidic device was incubated at 37 °C and a 5% CO2 atmosphere for 1 h. Cellular reprogramming. Send#1 hiPSCs were derived through Sendai virus-mediated reprogramming. Human fibroblasts were plated in 24-well plates and cultured to 70% confluency in DMEM supplemented with 10% FBS (Life Technologies). CytoTune-iPS Sendai Reprogramming Kit (Life Technologies), which is based on replication-incompetent Sendai virus, was used to deliver the four transcription factors Oct4, Sox2, Klf4 and c-Myc to induce reprogramming to iPSCs according to the manufacturer’s instructions. After a 24 h transduction, cells were fed every other day for 7 days. Transduced cells were transferred onto inactivated MEF feeder cells at a density of 1 × 104 cells in a 6-well plate and expanded as reported below. mRNA#1, #2, #3, #4 hiPSCs were generated through mRNA-mediated reprogramming following the protocol reported in Warren et al.22. Briefly, human foreskin BJ fibroblasts (Miltenyi Biotec) were seeded at 10 cells/mm2 and reprogrammed on inactivated human newborn foreskin fibroblast feeders (NuFF-RQ; AMS Biotechnology) in 6-well tissue culture plates. Pluriton Reprogramming Medium (Stemgent) was used supplemented with B18R (200 ng/ml; eBioscience cat# 34-8185) to limit interferon signaling. Daily mRNA transfections were performed for 12 days with the following modified mRNAs: Oct4, Sox2, Klf4, c-Myc, Nanog, Lin28, and nuclear GFP (Miltenyi Biotec). Two days after the last transfection, hiPSC colonies were picked and passaged on MEF-coated 12-well plates and expanded as reported below. Pluripotent cell lines induced in our laboratory (Send#1, mRNA#1, #2, #3, #4) have been characterized by conventional methods (immuno-staining and qRT-PCR for pluripotency markers expression, normal karyotype analysis, embryoid bodies formation). Cell culture and integration in microfluidic devices. HiPSCs ADHF#1 (kindly provided by M. Oshimura)23, hESCs HES2 (obtained from National Stem Cell Bank, Madison, WI) and Send#1 were cultured on mitomycin C-treated mouse embryonic feeder fibroblasts (MEF; Chemicon) in DMEM doi:10.1038/nmeth.3411
F-12 (Life Technologies) supplemented with 20% KSR (Life Technologies), 10% MEF-conditioned medium (only for hES2), 20 ng/ml or 10 ng/ml of basic fibroblast growth factor (bFGF; Life Technologies) for hESCs or hiPSCs respectively, 0.1 mM β-mercaptoethanol (Life Technologies), 1% non-essential amino acids (Life Technologies) and 1% penicillin-streptomycin (Life Technologies). hPSCs were passaged to new feeder cells using CTK solution (0.25% trypsin – collagenase IV – Ca 2+) for hiPSCs or 0.25% trypsin for hESCs (Life Technologies). HES2 cells were also adapted to feeder-free culture conditions using TeSR-E8 (Stemcell Technologies). mRNA#1, #2, #3, #4 lines were cultured in StemMACS iPS-Brew XF medium (Miltenyi Biotec). Alpha-1 antitrypsin-deficient A1ATZ/Z and the corrected A1ATR/R hiPSC cells18,24 (kindly provided by L. Vallier) were cultured in TeSR-E8. Feeder-free cells were cultured in multiwell plates coated with 0.5% MRF and passaged with 0.5 mM EDTA (Life Technologies). All cell lines have been tested for mycoplasma contamination. hPSCs were integrated within microfluidic devices by injecting cell suspensions into each pre-coated channel and incubating the devices overnight, at 37 °C and 5% CO2 without perfusion, to allow cell adhesion. We used the appropriate cell concentrations to achieve 70% confluency at 24 h after seeding. An 11-port pump (Cavro XLP pump, TECAN) was used to independently deliver the medium from the reservoirs into each microfluidic channel. Tygon tubing (0.5 inch inner diameter; Cole-Parmer), 21G stainless-steel needles and a polypropylene Luer (Microtest) were used to connect the microfluidic chip to the pump. Discontinuous medium delivery with defined temporal frequencies was achieved by automatically controlling the multichannel syringe pump through LabView 8.2 (National Instruments). Every medium change was performed using a flow rate of 6 µl/min for 2 min, which corresponds to perfusion with twice the channel volume. For pluripotency maintenance, hPSCs were cultured with f = 2 d−1. For conditioned medium-based experiments, HES2 cells were cultured in non-optimal frequency conditions ( f = 8 d−1) with conditioned medium ( f = 8 d−1 conditioned) collected from HES2 cultured in optimal conditions ( f = 2 d−1) and compared to cells cultured with fresh medium ( f = 8 d−1). hPSC differentiation into germ layers. For spontaneous differentiation, no exogenous factors or cytokines were used. Basal medium was prepared with KO-DMEM supplemented with 20% FBS, 1% NEAA, 0.5% L-glutamine and 1% penicillin-streptomycin (all from Life Technologies). Ectoderm differentiation was induced with a 1:1 ratio of DMEM F-12 and Neurobasal medium (Life Technologies), supplemented with 1% B27-supplement (Life Technologies), 1% N2-supplement (Life Technologies) and 0.1 mM β-mercaptoethanol25,26. Mesoderm differentiation was induced with StemPro-34 (Life Technologies) supplemented with 2 mM L-glutamine (Life Technologies), 200 ng/ml transferrin, 0.5 mM ascorbic acid (Sigma-Aldrich), 0.3 ng/ml Activin-A (R&D) and 3 ng/ml BMP-4 (R&D)27. Endoderm differentiation was induced with RPMI 1640 (Life Technologies) containing 1% B27supplement (RPMI-B27), 1 mM sodium butyrate (NaB; SigmaAldrich), 100 ng/ml Activin-A and 50 ng/ml Wnt3a (R&D)28 for the first 2 days. Medium was changed to KO-DMEM with 20% KSR, 1 mM L-glutamine, 1% NEAA, 1% DMSO, 0.1 mM β-mercaptoethanol and 1% penicillin-streptomycin for another 2–3 days. hPSCs were differentiated for 4–5 days. nature methods
© 2015 Nature America, Inc. All rights reserved.
Cardiac and hepatic specification. For cardiac differentiation, cells were cultured in mTeSR1 (Stemcell Technologies) in the microfluidic platform until confluent. The medium was switched to RPMI-B27 without insulin with 12 µM CHIR 99021 (Tocris) for 24 h. Medium was changed to RPMI-B27 without insulin for another 2 days. At day 3, the medium was changed to RPMI-B27 without insulin and supplemented with 10 µM IWP-4 (Stemgent) for an additional 2 days. At day 5 medium was changed to RPMI-B27 without insulin for 2 days. Beginning at day 7 the cells were cultured in RPMI-B27 until further experimental procedures29. Spontaneously beating cells were observed starting from day 12–15. For hepatocyte-like cell differentiation, endoderm was induced with RPMI-B27, 1 mM NaB, 100 ng/ml Activin-A, 50 ng/ml Wnt3a and 1% penicillin-streptomycin for 1 day and then changed to the same medium with reduced (0.5 mM NaB) for 2 more days. Medium was changed to KO-DMEM with 20% KSR, 1 mM L-glutamine, 1% NEAA, 0.1 mM β-mercaptoethanol, 1% DMSO (Sigma-Aldrich) and 1% penicillin-streptomycin for 6 days. Hepatic-like cells were matured with L15 medium (SigmaAldrich) supplemented with 8.3% FBS, 8.3% tryptose phosphate broth (Life Technologies), 10 µM hydrocortisone 21-hemisuccinate, 1 µM insulin (all from Sigma-Aldrich), 2 mM L-glutamine, 1% penicillin-streptomycin,10 ng/ml hepatocyte growth factor and 20 ng/ml oncostatin M (both from R&D) for 6 d30. For adult specification and maturation, cells were cultured with f = 2 d−1 maturation medium change. The functional differentiation of cardiac and hepatic cells was performed in 30 and 50 independent microfluidic experiments, respectively. HepG2 cell culture. HepG2 cells were cultured in 0.6% gelatin-coated glass slides in 24-well plates for static culture and gelatin-coated micro-channels for microfluidic culture. Cells were expanded in DMEM supplemented with 10% heat-inactivated FBS, 1% NEAA and 1% penicillin-streptomycin. Glucose and metabolite analyses of spent media. Media glucose concentration was measured with FreeStyle Lite glucometer and strips (Abbott). Enzymatic detection of ammonia and L(+)-Lactate concentrations in exhaust media were performed with the Ammonia Assay Kit (AA0100, Sigma-Aldrich) and Lactate assay kit (MAK064, Sigma-Aldrich), respectively, following the manufacturer’s instructions. Luciferase assay. CAGA12 SMAD2–SMAD3 reporter HaCaT cells were cultured in DMEM supplemented with 10% FBS16,31. For luciferase assays, cells were plated in 12-well plates at 70% confluency and incubated overnight in DMEM without serum and then treated with media containing recombinant TGF-β1 (Peprotech) or TGF-β receptor inhibitor SB431542 (Peprotech) or to conditioned medium for 8 h. In the last case, media were heat-treated for 5 min at 95 °C in order to activate endogenous TGF-β1. Luciferase expression was detected as described in Inui et al.16. Data were normalized on total protein content, determined through Bradford assay. To obtain conditioned media, HES2 cells adapted in feederfree culture conditions using TeSR-E8 medium, were plated in microfluidic channels or in Petri dishes at 70–90% confluency and cultured in TeSR-E7 medium, which has the same composition nature methods
of TeSR-E8, but with no exogenous recombinant TGF-β1. After 24 h the conditioned media were collected from Petri dishes and microfluidic channels at two different frequencies, f = 8 d−1 or f = 2 d−1, and heat-treated to induce latent TGF-β activation32. The amount of the TGF-β, activin and nodal ligands was indirectly quantified by exposing CAGA12-luciferase HaCaT cells to the collected media. f = 2 d−1 medium contained higher amounts of TGF-β ligands compared to f = 8 d−1 or Petri dish-grown cells. Fresh E8 and E7 media (with no TGF-β1) were used as positive and negative controls, respectively. Addition of the smallmolecule TGF-β receptor inhibitor SB431542 confirmed that the inductions were caused by extracellular factors in the medium. Functional tests. HPSC-derived cardiomyocytes were analyzed by confocal microscopy (Leica SP5) for calcium transients during caffeine or verapamil treatment33. HPSC-derived hepatocyte-like cells were stained with PAS staining kit (Sigma-Aldrich), according to the manufacturer’s instructions, for glycogen storage analysis. Indocyanine green test, ICG, (Sigma-Aldrich) was performed by incubating live cells for 15 min with dye solution, according to manufacturer’s instructions. Green cells were photographed. After 6 h, the same fields were photographed in order to see complete clearance of ICG by functional hepatocyte-like cells. Albumin production and secretion was assessed by performing ELISA assays (Immunology Consultants Laboratory, Inc.), following manufacturer’s instructions, on spent media (10 µL from each micro-channel). Albumin levels were detected at 450 nm using an Infinite F2000 PRO (Tecan) plate reader. Cytotoxicity experiments were performed using the hepatotoxic drug, acetaminophen, on HepG2 cells and hESC-derived hepatocytelike cells. Acetaminophen BioXtra ≥99.0% (Sigma-Aldrich) was dissolved in DMSO to a 5 M stock solution. Acetaminophen stock solution was diluted in fresh HepG2 culture medium or hESC-derived hepatocytes’ medium at final concentrations of 50, 25, 12.5, 10, 5, 1 and 0.5 mM for drug stimulation. As controls, untreated cells were cultured in their standard culture media or in the same media containing 1% DMSO (mock control). Cells were treated for 24 h at f = 2 d−1 or with multiple 3 h administrations using a multiple-port microfluidic switch34. For nuclear area calculation, treated cells were fixed and stained with Hoechst, and fluorescence images were taken at 80X magnification. For cytotoxic effects on cell morphology and integrity, treated cells were fixed and stained with phalloidin for F-actin or anti-albumin antibody. Cell viability assays (LIVE/DEAD Kit, Life Technologies) were performed on cells after 24 h to quantify dead cells. Cells were washed with PBS; incubated with 4 µM ethidium homodimer-1 (stains dead cells red), 4 µM calcein AM (stains live cells green) and 4 µM Hoechst (stains cell nuclei blue) for 45 min at room temperature; washed with PBS; and analyzed by fluorescence microscopy. Cytotoxicity was expressed as fraction of dead cells over total cells per 20 image. Immunofluorescence. Immunofluorescence analyses were performed on 4% paraformaldehyde-fixed cells for 15 min. Blocking and permeabilization were performed in blocking buffer (5% heat-inactivated FBS and 0.1% TritonX-100 (Sigma-Aldrich) in PBS -/- 1×) for 1 h. Depending on the antibody, cells were stained for 1 h at room temperature or overnight at 4 °C using primary antibodies diluted in blocking buffer as follows: Oct4 doi:10.1038/nmeth.3411
© 2015 Nature America, Inc. All rights reserved.
1:200 (sc-5279 Santa Cruz), Sox-2 1:200 (AB5603 Millipore), Tra-1-60 1:200 (MAB4360 Millipore), Tra-1-81 1:200 (MAB4381 Millipore), β-III tubulin 1:500 (T3952 Sigma-Aldrich), α-fetoprotein 1:250 (A8452 Sigma-Aldrich), Brachyury 1:100 (ab20680 Abcam), cardiac troponin T 1:100 (MS-295-P Thermo Scientific), albumin 1:100 (MAB1455 R&D), CYP-3A 1:150 (GTX117120 Genetex) and cytokeratin-18 1:150 (GTX105624S Genetex). For SOX17 (AF1924, R&D) and FOXA2/HNF3b (D56D6, Cell Signaling) staining, cells were incubated in 5% horse serum in 0.3% Triton X-100 blocking solution for 1 h at room temperature. Primary antibodies were diluted 1:400 and 1:20, respectively, and incubated in 1% BSA in 0.3% TritonX-100 overnight at 4 °C. Immunostaining was done by incubating secondary antibodies (Alexa Fluor 488 or 594 (Life Technologies) and Cy3 (Jackson ImmunoLab)) or DAPI (for nuclear staining) for 1 h at 37 °C. Pictures were taken on a Leica DMI6000 B microscope. Gene expression analysis. For RNA extraction, cells were lysed with 0.5 ml of TRIzol (Life Technologies), treated with 0.1 ml chloroform (Sigma-Aldrich) for 3 min and centrifuged at 12,500 g for 15 min at 4 °C. Aqueous supernatant was collected and diluted 1:1 with 70% ethanol. Total RNA was extracted from solution using RNeasy Mini Kit (Qiagen), following the manufacturer’s instructions, and quantified using a NanoDrop spectrophotometer. RNA was used with A(260/280) nm = 2.0 ± 0.1, A(260/230) nm = 2.0 ± 0.1 and A(320) nm ≈0.05. RNA retro-transcription was performed using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies), according to the manufacturer’s instructions. The qRT-PCR was performed with TaqMan gene expression assay probes (Life Technologies) according to the manufacturer’s instructions. The following genes were used: GAPDH (glyceraldehyde 3-phosphate dehydrogenase), POU5F1 (OCT4), NANOG, DNMT3B (DNA cytosine-5-methyltransferase 3 beta), TUBB3 (beta-III tubulin), OTX2 (orthodenticle homeobox 2), T (BRACHYURY-T), GATA4, ACTA2 (alpha-smooth muscle actin), AFP (alpha-fetoprotein), EOMES, FOXA2 and SOX17. Reactions were performed on an ABI Prism 7000 machine and results were analyzed with ABI Prism 7000 SDS software. GAPDH
doi:10.1038/nmeth.3411
expression was used to normalize Ct values for gene expression, and data were shown as relative fold change to control cells (static condition), using the Livak method35. For Figure 2a, the samples with highest gene expression were used as control for each gene and ∆∆Ct values were re-scaled for graphical representation. Statistical analysis. For statistical analyses, single pairwise comparisons were analyzed using Student’s t-test with P ≤ 0.05 (*) or P ≤ 0.01 (**) indicating significance. Multiple comparisons were performed by one-way ANOVA with Tukey post-test, with P ≤ 0.05 (*) or P ≤ 0.01 (**) indicating significance. Throughout the text, n = 6 indicates the number of replicates, referring to a combination of experiments performed in 2 different chips (with different batches of cells) as well as in 3 different independent channels within the same chip (using the same batch of cells). The robustness of our conclusions on perfusion frequency to different sources of variability (between and within independent biological replicates) was also verified by crossed-nested ANOVA on selected marker expression, using Minitab 17 statistical software (Supplementary Note 2).
21. Luni, C., Michielin, F., Barzon, L., Calabrò, V. & Elvassore, N. Biophys. J. 104, 934–942 (2013). 22. Warren, L. et al. Cell Stem Cell 7, 618–630 (2010). 23. Zatti, S. et al. Mol. Ther. Methods Clin. Dev. 1, 1–9 (2014). 24. Rashid, S.T. et al. J. Clin. Invest. 120, 3127–3136 (2010). 25. Chambers, S.M. et al. Nat. Biotechnol. 27, 275–280 (2009). 26. Camnasio, S. et al. Neurobiol. Dis. 46, 41–51 (2012). 27. Kouskoff, V., Lacaud, G., Schwantz, S., Fehling, H.J. & Keller, G. Proc. Natl. Acad. Sci. USA 102, 13170–13175 (2005). 28. Hay, D.C. et al. Proc. Natl. Acad. Sci. USA 105, 12301–12306 (2008). 29. Lian, X. et al. Proc. Natl. Acad. Sci. USA 109, E1848–E1857 (2012). 30. Hay, D.C. et al. Stem Cells 26, 894–902 (2008). 31. Levy, L. et al. Mol. Cell. Biol. 27, 6068–6083 (2007). 32. Slager, H.G., Freund, E., Buiting, A.M.J., Feijen, A. & Mummery, C.L. J. Cell. Physiol. 156, 247–256 (1993). 33. Martewicz, S. et al. Integr. Biol. (Camb) 4, 153–164 (2012). 34. Zambon, A., Zoso, A., Luni, C., Frommer, W.B. & Elvassore, N. Integr. Biol. (Camb) 6, 277–288 (2014). 35. Livak, K.J. & Schmittgen, T.D. Methods 25, 402–408 (2001).
nature methods
© 2015 Nature America, Inc. All rights reserved.
Functional differentiation of human pluripotent stem cells on a chip Giovanni G Giobbe1,2,5, Federica Michielin1,2,5, Camilla Luni1,2, Stefano Giulitti1,2, Sebastian Martewicz1,2, Sirio Dupont3, Annarosa Floreani4 & Nicola Elvassore1,2 Microengineering human “organs-on-chips” remains an open challenge. Here, we describe a robust microfluidics-based approach for the differentiation of human pluripotent stem cells directly on a chip. Extrinsic signal modulation, achieved through optimal frequency of medium delivery, can be used as a parameter for improved germ layer specification and cell differentiation. Human cardiomyocytes and hepatocytes derived on chips showed functional phenotypes and responses to temporally defined drug treatments.
The development of human organs-on-chips, in which microscale engineering technologies enable the recapitulation of the organ microenvironment, offers a unique opportunity to study human physiology and pathophysiology1,2. Recently, successful examples of organs-on-chips have been demonstrated3–6 based on primary animal and, in a few cases, human cells. Human embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs) grown in culture have the potential to give rise to any fetal and adult cell type. Developing direct organogenesis-on-a-chip from human pluripotent stem cells (hPSCs) could overcome the limited availability of human primary cells. Furthermore, hPSC-derived cells show intrinsic and unexpected levels of emergent self-organization for generating highly ordered structures and tissues7,8. The intrinsic properties of microfluidics allow spatiotemporal control of the cell culture microenvironment and enable the biomimetic scale-down of developmental processes, from germ layer specification and phenotypic differentiation to tissue morphogenesis9. Recent studies on mouse ESC cultures using microfluidic systems highlight the importance of accurate regulation of the soluble microenvironment, in terms of endogenous and exogenous extrinsic factors, to maintain pluripotency and self-renewal10,11. We and others have reported hESC culture on a chip, modulating cell culture conditions to promote pluripotency maintenance and self-renewal, and germ layer specification12–14.
Here we report the first, to our knowledge, integrated functional hPSC differentiation on a chip. We explored whether we could control hPSC expansion, selective germ layer commitment and differentiation into functional tissue-specific cells through a multistage microfluidic-based process. This would require the optimal delivery of exogenous factors and removal of endogenous cell-secreted factors which affect transcriptional activity15. We hypothesized that periodic medium delivery with stage-dependent frequency (number of cycles of medium changes per day; referred to as ‘perfusion frequency’ or ‘f’) is an effective method for modulating soluble microenvironments and, consequently, for controlling stem cell niche specification in vitro (Fig. 1a). Clearly, low perfusion frequencies allow endogenous factors to accumulate within the channel. Conversely, high frequencies promote a sustained supply of exogenous factors but lead to continuous washout of endogenous secreted factors (Fig. 1b), as qualitatively described by a computational model (Supplementary Fig. 1 and Supplementary Note 1). We observed that discontinuous periodic medium delivery, unlike continuous microfluidic flow perfusion, yielded homogeneity along the channels in terms of colony size, cell density and pluripotency marker expression for both hESCs and hiPSCs (Supplementary Figs. 2 and 3). This method also allowed proper glucose supply and removal of waste products such as ammonia and lactate (Supplementary Fig. 4). We first investigated the optimal perfusion frequency for hPSC expansion using an automated microfluidic experimental setup (Fig. 1c) in which we observed homogeneous growth and the expression of pluripotency markers along the microfluidic channel (Fig. 1d,e). Perfusion frequency–dependent expression levels of the pluripotency markers OCT4, NANOG and DNMT3B were observed in HES2 cells after 6 d of culture (Fig. 1f), with marked differences among frequencies for OCT4 and NANOG. Specifically, significantly higher OCT4 expression was observed for f = 2 d−1 (P ≤ 0.01), compared to all other frequencies or to continuous perfusion (CP), and higher NANOG expression for f = 2 d−1, compared to the other frequencies or CP (P ≤ 0.05). To further test the correlation between optimal perfusion frequency and the accumulation of endogenous cell-secreted factors, we used HES2-conditioned medium collected in an f = 2 d−1 experiment to culture HES2 cells at a higher perfusion frequency. We observed a significant threefold increase in OCT4 expression when conditioned, instead of fresh, medium was used at f = 8 d−1 (P ≤ 0.01) (Fig. 1g). Interestingly, the conditioned medium collected at f = 2 d−1 contained higher levels of TGF-β ligands, as demonstrated by its ability to activate SMAD2–SMAD3 transcriptional activity
1Department of Industrial Engineering, University of Padova, Padova, Italy. 2Venetian Institute of Molecular Medicine, Padova, Italy. 3Department of Molecular Medicine,
University of Padova, Padova, Italy. 4Department of Surgery, Oncology and Gastroenterology, University of Padova, Padova, Italy. 5These authors contributed equally to this work. Correspondence should be addressed to N.E. ([email protected]). Received 5 March 2014; accepted 10 March 2015; published online 1 June 2015; doi:10.1038/nmeth.3411
nature methods | ADVANCE ONLINE PUBLICATION |
brief communications b
Tr Gt
. V
Ex 0
f2
f1
f3
t t
t
t
t
TRA-1-60
hiPSCs
hiPSCs
hESCs
hESCs
SOX2
OCT4
SOX2
TRA-1-81
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0
OCT4 NANOG DNMT3B
*
ic
at
St
f=
1
d
5
=
1.
d
f=
2
d
f=
3
d
f=
4
d
f=
8
d
P
C
(measured using the synthetic CAGA12-luciferase reporter in HaCaT cells16), as compared to medium collected at f = 8 d−1 or under static conditions (no change of medium during the 24 h growth period) (Fig. 1h, Supplementary Fig. 5 and Online Methods). Consistently, the expression of NANOG, a known direct transcriptional target of SMAD2–SMAD3 in hESCs17, was upregulated in the f = 2 d−1 condition when compared to higher perfusion frequency or static conditions (Fig. 1i). These data suggested that endogenous TGF-β, activin and nodal ligands were released, or accumulated to optimal levels, in an f-dependent manner to promote pluripotency. We next investigated whether perfusion frequency could also be used to direct early germ layer commitment upon spontaneous differentiation (Fig. 2a and Supplementary Fig. 6). QRT-PCR analysis on static and low perfusion frequency (f = 1 d−1 and f = 2 d−1) cultures showed mainly ectoderm marker expression (TUBB3 and OTX2), whereas mesoderm (T, GATA4 and ACTA2) and endoderm (AFP, EOMES and FOXA2) marker expression were enhanced at a higher frequency (f = 8 d−1). This observation is consistent with the hypothesis that high accumulation of extrinsic endogenous factors promotes ectoderm differentiation while inhibiting meso-endoderm specification (Supplementary Fig. 7)10,15. | ADVANCE ONLINE PUBLICATION | nature methods
Relative NANOG expression
Luciferase activity (a.u.)
f Figure 1 | hPSC culture on a chip. g h i (a) Hypothetical model of the signaling 1.8 1.6 OCT4 pathway describing the synergistic ** 1.6 ** 2,000 1.4 NANOG ** 1.4 effects of extrinsic exogenous (Ex) 1.2 * 1.2 1,500 1.0 * and endogenous (En) factors, which 1.0 0.8 0.8 1,000 affect gene transcript (Gt) expression * 0.6 0.6 0.4 mediated by target transcription 0.4 500 0.2 0.2 factors (Tr). N( f ) is the perfusion 0 0 0 –1 –1 –1 + + – – – – – – – – + – – – + TGF-β TGF-β frequency-dependent molar flux. f = 8 d f=2d f=8d – + – + – + – + – + SB431542 conditioned Reactions (1) and (2) describe the Static f = 8 d–1f = 2 d–1 Fresh medium Static f = 8 d–1f = 2 d–1 coupling of factors to membrane receptors. Reaction (3) is the basal activation of Tr. Reactions (4) and (5) describe production of Gt and En, respectively. (b) A schematic of perfusion frequency ( f )-dependent temporal profiles of normalized Ex (red lines) and En (black lines) factor concentrations, which lead to different levels of mean concentrations over time, ‘t’ (straight lines). (c) Medium delivery in the microfluidic platform is controlled by a multiport delivery system. Flow rate (V), perfusion frequency ( fn), temporal delay between cycles and total number of cycles are set independently and automatically controlled. (d) Micrographs showing hESCs and hiPSCs, both expanded up to 6 d with f = 2 d−1 (inset). Scale bars, 300 µm. (e) Immunofluorescence analyses of pluripotency markers on the same cells as in (d). Scale bar, 50 µm. (f) qRT-PCR analysis of pluripotency markers in HES2 cells cultured under the indicated conditions. CP, continuous perfusion. Data are normalized to GAPDH and shown as mean ± s.e.m. (n = 6); ANOVA *P ≤ 0.05, **P ≤ 0.01. (g) qRT-PCR analysis of OCT4 and NANOG from HES2 cells cultured with fresh or conditioned medium at the indicated perfusion frequencies. Data are normalized to GAPDH and shown as mean ± s.e.m. (n = 6); t-tests **P ≤ 0.01. (h) Luciferase assay (based on CAGA12 SMAD2–SMAD3 cells) measuring the presence of TGF-β, activin and nodal ligands in conditioned medium from HES2 cells cultured under the indicated perfusion conditions. Data are normalized to total protein content (Bradford assay) and shown as mean ± s.e.m. (n = 6); t-tests **P ≤ 0.01. (i) NANOG expression by qRT-PCR analysis of HES2 cells. Data normalized to GAPDH and shown as mean ± s.e.m. (n = 6) t-tests *P ≤ 0.05. See statistical section of Online Methods for definition of n. Relative gene expression
© 2015 Nature America, Inc. All rights reserved.
Waste
**
–1
f OCT4
–1
e
Relative gene expression
d
Multi-port delivery system
Automatic control
f
–1
En
En
–1
Ex
OUT NEn (f)
–1
(3)
(1)
(2) (5) (4)
NEx (f)
Concentration
OUT
IN
NE x (f)
c 1
–1
a
To further test whether selective induction of the three germ layers could be achieved, we induced hPSCs to form ectodermal, mesodermal or endodermal cells by growing them in specific differentiation medium (containing germ layer–specific cytokines and factors) with optimal perfusion frequencies. As an example, the perfusion frequency optimization for microfluidic induction of mesodermal cells is shown in Figure 2b. Similarly, perfusion frequencies of 1, 2 or 3 d−1 were found to be optimal for the homogeneous differentiation of ectoderm, mesoderm and endoderm, respectively, as measured by qRT-PCR and immunofluorescence analyses (Fig. 2c,d and Supplementary Fig. 8). For endoderm commitment, we also observed significantly higher expression of the endoderm marker AFP when HES2 cells were differentiated under microfluidic conditions compared to conventional static differentiation (P ≤ 0.05). Consistently, optimized microfluidic conditions led to a significant threefold increase in SOX17+FOXA2+ cells (in 3 out of 6 conditions) compared to Petri dish culture (Supplementary Fig. 9). Collectively, our findings indicate that perfusion frequency is an additional parameter that can enhance germ layer specification, even when specific differentiation protocols are used. We then examined whether our strategy of controlled perfusion frequency could be used to achieve functional differentiation of cells of interest on chips. We first focused on cardiac cells (Fig. 3,
Ex factors
TUBB3 OTX2 T GATA4 ACTA2 AFP EOMES FOXA2
0.6 0.4 0.2 0 Petri dish
f = 1 d–1
TUBB3 T AFP
f = 2 d–1
f = 8 d–1
*
0.8 0.6 0.4 0.2 0 Ectoderm Mesoderm Endoderm Ectoderm Mesoderm Endoderm
Static
Supplementary Fig. 10 and Online Methods). We typically obtained 5,000–10,000 cardiac cells per channel after 14 d of cardiac differentiation (Fig. 3a) from HES2 and ADHF#1 cells. We observed spontaneous contractile activity from both hESC- and hiPSC-derived cardiomyocytes (Supplementary Videos 1–3). Overall, 65 ± 11% of cells were positive for cardiac troponin T (CTNT), and these also exhibited defined cardiac sarcomeric organization (Fig. 3b). hPSCderived cardiomyocytes showed spontaneous calcium transients (Fig. 3c) and excitation-contraction coupling. We also observed functional responses to 0.5 µM verapamil, with cells showing reduced calcium release after L-type channel inhibition, and to 10 mM caffeine, with cells showing increased cytosolic calcium after ryanodine channel activation (Fig. 3c). We next applied our methodology to derive hepatocytes on chips from hESCs and hiPSCs (Fig. 3d) using the previously endoderm-optimized conditions (Supplementary Fig. 9). After 14 d of differentiation we observed characteristic polygonalshaped hepatocyte-like cells, including binucleate cells, on the chip. These cells were positive for the hepatic markers albumin and cytochrome P450-3A (CYP-3A) (Fig. 3e), and showed albumin secretion, glycogen storage and indocyanine green digestion (Fig. 3f). Our optimized perfusion conditions yielded a 40% greater albumin secretion at the end-point of differentiation and a greater percentage of ALB and CYP-3A positive cells (Fig. 3g,h) in comparison to cells differentiated under static conditions. We observed robust differentiation of eight hiPSC lines obtained by different methods using the microfluidic strategy we defined (Supplementary Figs. 11–13 and Online Methods). We also used our approach to derive hepatocyte-like cells on chips from genetically corrected α1-antitrypsin–deficient hiPSCs18 (Supplementary Fig. 11), opening up new possibilities for disease modeling on chips and for patient-specific drug design. To test whether differentiated cells on a chip could be used to monitor cellular drug response, we performed a series of experiments monitoring acetaminophen-induced cytotoxicity. We chose
Microfluidic
**
1.0
**
0.8 0.6 0.4 0.2 0 f = 1 d–1
d
f = 2 d–1
f = 3 d–1 f = 12 d–1
Markers Mesoderm Ectoderm
0.8
1.0
T
Specific differentiation protocol
Concentration
En factors
1.0
c
b
High f
Relative gene expression
Intermediate f
Low f
Ectoderm β-III TUB
Mesoderm BRA
Endoderm AFP
Endoderm
a
Relative gene expression
Figure 2 | Germ layer induction of hESCs-ona-chip. (a) Top, schematic of the relationship between the perfusion frequency, f, and the concentrations of En and Ex factors. Bottom, qRT-PCR analysis for the indicated markers for ectoderm (TUBB3 and OTX2), mesoderm (T, GATA4 and ACTA2) and endoderm (AFP, EOMES and FOXA2) from spontaneously differentiated HES2 cells on chips, at the indicated perfusion frequencies. Phase contrast images showing the cell morphology for each differentiation condition is shown beneath the graph. Scale bar 50 µm. Data normalized to GAPDH and shown as mean ± s.d. (n = 6). (b) qRT-PCR analysis of T expression at different perfusion frequencies for mesoderm-induced HES2 cells. Data normalized to GAPDH and shown as mean ± s.d. (n = 6); ANOVA **P ≤ 0.01. (c) qRT-PCR analysis for the indicated markers of the three germ layers in HES2 cells differentiated under the indicated conditions. Data normalized to GAPDH and shown as mean ± s.d. (n = 6); t-tests *P ≤ 0.05. (d) Immunofluorescence analysis of germ layer-committed HES2 cells. Scale bar, 50 µm. See statistical section of Online Methods for definition of n.
Relative gene expression
© 2015 Nature America, Inc. All rights reserved.
brief communications
acetaminophen because it is mainly metabolized in the liver through CYP-3A activity19 and causes hepatocyte toxicity at high doses (Supplementary Fig. 14). We compared cytotoxicity on hESC-derived hepatocyte-like cells in static culture and in microfluidic culture at f = 2 d−1, for the same acetaminophen concentration (25 mM). Because of the small volume of medium in microfluidic chip, the overall amount of acetaminophen per cell was 10 times lower in this system. Despite this, we observed 75% of dead cells in the micro-channels compared to 20% in the static control for a 24-h exposure (Fig. 3i). This result is consistent with higher expression of CYP-3A in cells differentiated under microfluidic conditions (Fig. 3h and Supplementary Fig. 15). The tight temporal control of medium delivery achievable by microfluidic technology makes it particularly suitable to test acetaminophen dosage (Fig. 3j). Experiments on hESC-derived hepatic cells show that, even at low doses, repeated drug administration (4 times per day for 3 h each) had a significantly higher cytotoxic impact than application of a single high dose, regardless of the higher overall amount of drug in the single-administration case20. In conclusion, we have derived functional tissue-specific cells via hPSC differentiation directly on a chip through a robust multistage microfluidic technology, and demonstrate the derivation of functional hepatocyte-like cells that may prove suitable for highthroughput drug testing. Accurate spatiotemporal control of the soluble microenvironment around cells is a critical aspect of this methodology and was achieved through regulation of periodic perfusion frequencies. This technology opens up new possibilities for generating human organs-on-chip from hPSCs. Methods Methods and any associated references are available in the online version of the paper. Note: Any Supplementary Information and Source Data files are available in the online version of the paper.
nature methods | ADVANCE ONLINE PUBLICATION |
brief communications
2s
D4–D18 Cardiomyocyte CTNT
80
Positive cells (%)
Brachy
Static Microfluidic
100
0.5 µM verapamil
10 mM caffeine
Fluo-4 (F/F0)
D0–D3 Mesoderm
h
c
Hoechst CTNT
Fluo-4 (F/F0)
b
OCT4 NANOG
Fluo-4 (F/F0)
a
60
**
*
40 20 0
35
PAS staining – glycogen
9
0.8 Dead cells fraction
25 20 15 10 5
AM
12
Static Microfluidic
15
**
18 PM
6 3
**
21 24
0.6
CYP-3A
Midnight
**
0.4 0.2
j
4
0
d1
d6
d1
0 d1
2
6h clearance
d1 6
0
d6
0h uptake
ALB
Noon
1.0
**
30
ICG
CK-18
i
+ ctr Static Microfluidic
G
ALB CYP-3A
Hoechst CYP-3A
g
0 Ctr
1
5
10
25
Acetaminophen conc. [mM] Noon
Noon Noon Noon Figure 3 | Cardiac and hepatic cell differentiation on a chip. 12 12 12 12 9 15 9 9 15 15 9 15 (a) Microfluidic cardiac differentiation protocol. (b) Immunofluorescence 18 PM 6 AM 6 18 PM AM 6 18 PM AM 18 PM AM 6 of hESC-derived cardiomyocytes (CMs) showing cardiac troponin T (CTNT) 3 21 organization. Scale bars, 15 µm. (c) Spontaneous calcium transients in 3 21 3 3 21 21 24 24 24 24 hESC-derived CMs recorded with Fluo-4, under control conditions (left) Midnight Midnight Midnight Midnight ** or in the presence of verapamil (right, top) or caffeine (right, bottom). 0.3 0.3 0.3 0.3 (d) Microfluidic hepatic differentiation protocol. (e) Immunofluorescence * ** 0.2 0.2 0.2 0.2 of hESC-derived hepatocyte-like cells showing albumin (ALB) and CYP-3A expression. Scale bars, 15 µm. (f) Periodic acid–Schiff stain (PAS) 0.1 0.1 0.1 0.1 of hepatic cells; nuclei are stained in blue, pink-violet colored cells show 0 0 0 0 glycogen storage (upper panel). Lower panels, hepatic cells at 0 (left) Ctr 1 5 10 25 Ctr 1 5 10 25 Ctr 1 5 10 25 Ctr 1 5 10 25 and 6 h (right) after staining with indocyanine green (ICG). Colorant is Acetaminophen conc. [mM] absorbed and metabolized in 6 h by functional hepatocyte-like cells. Scale bars, 150 µm. (g) ELISA assay for secreted human albumin (ng/ml/day per 2 × 105 cells) from hESC-derived hepatocyte-like cells derived under the indicated conditions. Positive control (ctr) HepG2 cell line. Data are shown as mean ± s.d. (n = 6); t-tests **P ≤ 0.01. (h) Quantification of CK-18-, ALB- or CYP-3A-positive cells differentiated with the protocol reported in Supplementary Figure 10. Data are shown as mean ± s.d. (n = 10); t-tests *P ≤ 0.05, **P ≤ 0.01. (i) Fraction of dead cells upon treatment of hESC-derived hepatocyte-like cells with increasing acetaminophen concentrations in static or microfluidic conditions. Ctr, cells treated with DMSO only. Data are shown as mean ± s.d. (n = 6); ANOVA **P ≤ 0.01. (j) Fractions of dead cells upon treatment of hESC-derived hepatocyte-like cells with the indicated doses of acetaminophen, under the indicated temporal administration protocols. Ctr, cells treated with DMSO only. Data are shown as mean ± s.d. (n = 6); 2-way ANOVA *P ≤ 0.05, **P ≤ 0.01. See statistical section of Online Methods for definition of n.
Dead cells fraction
© 2015 Nature America, Inc. All rights reserved.
D9–D14 Hepatocytelike cell
Hoechst ALB
f
Albumin production
AFP
e
ep
OCT4 NANOG D0–D3 Definitive endoderm SOX17 FOXA2 D4–D8 Hepatic endoderm
H
d
Acknowledgments This research was supported by Progetti di Eccellenza Cassa di Risparmio di Padova e Rovigo (CARIPARO), Progetti di Eccellenza Giovani Ricercatori of Ministero della Salute and Progetto Strategico Università di Padova. We acknowledge Miltenyi Biotec for kindly providing the mRNA reprogramming kit. We thank L.Vallier (Cambridge University) and the Cambridge NIhR BRC hIPSC core facility for providing us A1ATD and corrected hiPS cell lines, and M. Oshimura (Department of Biomedical Science, Institute of Regenerative Medicine and Biofunction, Tottori University, Japan.) for the ADHF#1 hiPS cell line. AUTHOR CONTRIBUTIONS G.G.G., F.M. and N.E. designed the research; G.G.G. and F.M. performed the experiments; S.M. performed cardiac functional tests; S.G. helped in microfluidic platform set-up and microfluidic cell culture; S.G. and C.L performed reprogramming experiments; S.D. helped in TGF-β experiments; A.F. supervised hepatic differentiation experiments; N.E. coordinated the project; G.G.G., F.M. and N.E. wrote the manuscript. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. Reprints and permissions information is available online at http://www.nature. com/reprints/index.html.
| ADVANCE ONLINE PUBLICATION | nature methods
1. Huh, D., Torisawa, Y., Hamilton, G.A., Kim, H.J. & Ingber, D.E. Lab Chip 12, 2156–2164 (2012). 2. Ghaemmaghami, A.M., Hancock, M.J., Harrington, H., Kaji, H. & Khademhosseini, A. Drug Discov. Today 17, 173–181 (2012). 3. Huh, D. et al. Science 328, 1662–1668 (2010). 4. Lee, S.-A. et al. Lab Chip 13, 3529–3537 (2013). 5. Jang, K.-J. et al. Integr. Biol. 5, 1119–1129 (2013). 6. Grosberg, A., Alford, P.W., McCain, M.L. & Parker, K.K. Lab Chip 11, 4165–4173 (2011). 7. Sasai, Y. Nature 493, 318–326 (2013). 8. Eiraku, M. et al. Nature 472, 51–56 (2011). 9. Discher, D.E., Mooney, D.J. & Zandstra, P.W. Science 324, 1673–1677 (2009). 10. Przybyla, L.M. & Voldman, J. Proc. Natl. Acad. Sci. USA 109, 835–840 (2012). 11. Moledina, F. et al. Proc. Natl. Acad. Sci. USA 109, 3264–3269 (2012). 12. Figallo, E. et al. Lab Chip 7, 710–719 (2007). 13. Villa-Diaz, L.G. et al. Lab Chip 9, 1749–1755 (2009). 14. Korin, N., Bransky, A., Dinnar, U. & Levenberg, S. in Proc. SPIE 6416, 64160N (2006). 15. Giobbe, G.G. et al. Biotechnol. Bioeng. 109, 3119–3132 (2012). 16. Inui, M. et al. Nat. Cell Biol. 13, 1368–1375 (2011). 17. Xu, R.-H. et al. Cell Stem Cell 3, 196–206 (2008). 18. Yusa, K. et al. Nature 478, 391–394 (2011). 19. Manyike, P.T., Kharasch, E.D., Kalhorn, T.F. & Slattery, J.T. Clin. Pharmacol. Ther. 67, 275–282 (2000). 20. Watkins, P.B. et al. J. Am. Med. Assoc. 296, 87–93 (2006).
© 2015 Nature America, Inc. All rights reserved.
ONLINE METHODS Microfluidic device fabrication and functionalization. Each microfluidic device, containing ten independent channels 18 mm long and 1.5 mm wide, was fabricated by standard softlithographic techniques21. The master was photo-lithographically patterned through SU8-2100 negative photoresist (MicroChem) to obtain a final thickness of 200 µm, according to the manufacturer’s indications. A premixed 10:1 mixture of polydimethylsiloxane (PDMS) pre-polymer and curing agent solutions (Sylgard 184 kit; Dow Corning) was cast on the silicon wafer and cured at 70 °C for 2 h. The PDMS mold was cut, peeled off and punched with a 21G stainless steel needle (Small Part Inc.) to obtain inlet and outlet holes. The PDMS mold was assembled and sealed to a 50 × 75 mm cleaned glass slide by plasma bonding. Ten independent medium reservoirs (each with a 70 µL capacity), one for each channel, were obtained by sealing an additional PDMS block to the top of the device by plasma bonding. The assembled device was cleaned with isopropanol (Sigma-Aldrich) and sterilized by autoclaving. Microfluidic channels were filled with 20% Matrigel Reduced Factor (MRF; BD Biosciences) and incubated overnight at 4 °C. Before cell seeding, the protein solution was aspirated and the microfluidic device was incubated at 37 °C and a 5% CO2 atmosphere for 1 h. Cellular reprogramming. Send#1 hiPSCs were derived through Sendai virus-mediated reprogramming. Human fibroblasts were plated in 24-well plates and cultured to 70% confluency in DMEM supplemented with 10% FBS (Life Technologies). CytoTune-iPS Sendai Reprogramming Kit (Life Technologies), which is based on replication-incompetent Sendai virus, was used to deliver the four transcription factors Oct4, Sox2, Klf4 and c-Myc to induce reprogramming to iPSCs according to the manufacturer’s instructions. After a 24 h transduction, cells were fed every other day for 7 days. Transduced cells were transferred onto inactivated MEF feeder cells at a density of 1 × 104 cells in a 6-well plate and expanded as reported below. mRNA#1, #2, #3, #4 hiPSCs were generated through mRNA-mediated reprogramming following the protocol reported in Warren et al.22. Briefly, human foreskin BJ fibroblasts (Miltenyi Biotec) were seeded at 10 cells/mm2 and reprogrammed on inactivated human newborn foreskin fibroblast feeders (NuFF-RQ; AMS Biotechnology) in 6-well tissue culture plates. Pluriton Reprogramming Medium (Stemgent) was used supplemented with B18R (200 ng/ml; eBioscience cat# 34-8185) to limit interferon signaling. Daily mRNA transfections were performed for 12 days with the following modified mRNAs: Oct4, Sox2, Klf4, c-Myc, Nanog, Lin28, and nuclear GFP (Miltenyi Biotec). Two days after the last transfection, hiPSC colonies were picked and passaged on MEF-coated 12-well plates and expanded as reported below. Pluripotent cell lines induced in our laboratory (Send#1, mRNA#1, #2, #3, #4) have been characterized by conventional methods (immuno-staining and qRT-PCR for pluripotency markers expression, normal karyotype analysis, embryoid bodies formation). Cell culture and integration in microfluidic devices. HiPSCs ADHF#1 (kindly provided by M. Oshimura)23, hESCs HES2 (obtained from National Stem Cell Bank, Madison, WI) and Send#1 were cultured on mitomycin C-treated mouse embryonic feeder fibroblasts (MEF; Chemicon) in DMEM doi:10.1038/nmeth.3411
F-12 (Life Technologies) supplemented with 20% KSR (Life Technologies), 10% MEF-conditioned medium (only for hES2), 20 ng/ml or 10 ng/ml of basic fibroblast growth factor (bFGF; Life Technologies) for hESCs or hiPSCs respectively, 0.1 mM β-mercaptoethanol (Life Technologies), 1% non-essential amino acids (Life Technologies) and 1% penicillin-streptomycin (Life Technologies). hPSCs were passaged to new feeder cells using CTK solution (0.25% trypsin – collagenase IV – Ca 2+) for hiPSCs or 0.25% trypsin for hESCs (Life Technologies). HES2 cells were also adapted to feeder-free culture conditions using TeSR-E8 (Stemcell Technologies). mRNA#1, #2, #3, #4 lines were cultured in StemMACS iPS-Brew XF medium (Miltenyi Biotec). Alpha-1 antitrypsin-deficient A1ATZ/Z and the corrected A1ATR/R hiPSC cells18,24 (kindly provided by L. Vallier) were cultured in TeSR-E8. Feeder-free cells were cultured in multiwell plates coated with 0.5% MRF and passaged with 0.5 mM EDTA (Life Technologies). All cell lines have been tested for mycoplasma contamination. hPSCs were integrated within microfluidic devices by injecting cell suspensions into each pre-coated channel and incubating the devices overnight, at 37 °C and 5% CO2 without perfusion, to allow cell adhesion. We used the appropriate cell concentrations to achieve 70% confluency at 24 h after seeding. An 11-port pump (Cavro XLP pump, TECAN) was used to independently deliver the medium from the reservoirs into each microfluidic channel. Tygon tubing (0.5 inch inner diameter; Cole-Parmer), 21G stainless-steel needles and a polypropylene Luer (Microtest) were used to connect the microfluidic chip to the pump. Discontinuous medium delivery with defined temporal frequencies was achieved by automatically controlling the multichannel syringe pump through LabView 8.2 (National Instruments). Every medium change was performed using a flow rate of 6 µl/min for 2 min, which corresponds to perfusion with twice the channel volume. For pluripotency maintenance, hPSCs were cultured with f = 2 d−1. For conditioned medium-based experiments, HES2 cells were cultured in non-optimal frequency conditions ( f = 8 d−1) with conditioned medium ( f = 8 d−1 conditioned) collected from HES2 cultured in optimal conditions ( f = 2 d−1) and compared to cells cultured with fresh medium ( f = 8 d−1). hPSC differentiation into germ layers. For spontaneous differentiation, no exogenous factors or cytokines were used. Basal medium was prepared with KO-DMEM supplemented with 20% FBS, 1% NEAA, 0.5% L-glutamine and 1% penicillin-streptomycin (all from Life Technologies). Ectoderm differentiation was induced with a 1:1 ratio of DMEM F-12 and Neurobasal medium (Life Technologies), supplemented with 1% B27-supplement (Life Technologies), 1% N2-supplement (Life Technologies) and 0.1 mM β-mercaptoethanol25,26. Mesoderm differentiation was induced with StemPro-34 (Life Technologies) supplemented with 2 mM L-glutamine (Life Technologies), 200 ng/ml transferrin, 0.5 mM ascorbic acid (Sigma-Aldrich), 0.3 ng/ml Activin-A (R&D) and 3 ng/ml BMP-4 (R&D)27. Endoderm differentiation was induced with RPMI 1640 (Life Technologies) containing 1% B27supplement (RPMI-B27), 1 mM sodium butyrate (NaB; SigmaAldrich), 100 ng/ml Activin-A and 50 ng/ml Wnt3a (R&D)28 for the first 2 days. Medium was changed to KO-DMEM with 20% KSR, 1 mM L-glutamine, 1% NEAA, 1% DMSO, 0.1 mM β-mercaptoethanol and 1% penicillin-streptomycin for another 2–3 days. hPSCs were differentiated for 4–5 days. nature methods
© 2015 Nature America, Inc. All rights reserved.
Cardiac and hepatic specification. For cardiac differentiation, cells were cultured in mTeSR1 (Stemcell Technologies) in the microfluidic platform until confluent. The medium was switched to RPMI-B27 without insulin with 12 µM CHIR 99021 (Tocris) for 24 h. Medium was changed to RPMI-B27 without insulin for another 2 days. At day 3, the medium was changed to RPMI-B27 without insulin and supplemented with 10 µM IWP-4 (Stemgent) for an additional 2 days. At day 5 medium was changed to RPMI-B27 without insulin for 2 days. Beginning at day 7 the cells were cultured in RPMI-B27 until further experimental procedures29. Spontaneously beating cells were observed starting from day 12–15. For hepatocyte-like cell differentiation, endoderm was induced with RPMI-B27, 1 mM NaB, 100 ng/ml Activin-A, 50 ng/ml Wnt3a and 1% penicillin-streptomycin for 1 day and then changed to the same medium with reduced (0.5 mM NaB) for 2 more days. Medium was changed to KO-DMEM with 20% KSR, 1 mM L-glutamine, 1% NEAA, 0.1 mM β-mercaptoethanol, 1% DMSO (Sigma-Aldrich) and 1% penicillin-streptomycin for 6 days. Hepatic-like cells were matured with L15 medium (SigmaAldrich) supplemented with 8.3% FBS, 8.3% tryptose phosphate broth (Life Technologies), 10 µM hydrocortisone 21-hemisuccinate, 1 µM insulin (all from Sigma-Aldrich), 2 mM L-glutamine, 1% penicillin-streptomycin,10 ng/ml hepatocyte growth factor and 20 ng/ml oncostatin M (both from R&D) for 6 d30. For adult specification and maturation, cells were cultured with f = 2 d−1 maturation medium change. The functional differentiation of cardiac and hepatic cells was performed in 30 and 50 independent microfluidic experiments, respectively. HepG2 cell culture. HepG2 cells were cultured in 0.6% gelatin-coated glass slides in 24-well plates for static culture and gelatin-coated micro-channels for microfluidic culture. Cells were expanded in DMEM supplemented with 10% heat-inactivated FBS, 1% NEAA and 1% penicillin-streptomycin. Glucose and metabolite analyses of spent media. Media glucose concentration was measured with FreeStyle Lite glucometer and strips (Abbott). Enzymatic detection of ammonia and L(+)-Lactate concentrations in exhaust media were performed with the Ammonia Assay Kit (AA0100, Sigma-Aldrich) and Lactate assay kit (MAK064, Sigma-Aldrich), respectively, following the manufacturer’s instructions. Luciferase assay. CAGA12 SMAD2–SMAD3 reporter HaCaT cells were cultured in DMEM supplemented with 10% FBS16,31. For luciferase assays, cells were plated in 12-well plates at 70% confluency and incubated overnight in DMEM without serum and then treated with media containing recombinant TGF-β1 (Peprotech) or TGF-β receptor inhibitor SB431542 (Peprotech) or to conditioned medium for 8 h. In the last case, media were heat-treated for 5 min at 95 °C in order to activate endogenous TGF-β1. Luciferase expression was detected as described in Inui et al.16. Data were normalized on total protein content, determined through Bradford assay. To obtain conditioned media, HES2 cells adapted in feederfree culture conditions using TeSR-E8 medium, were plated in microfluidic channels or in Petri dishes at 70–90% confluency and cultured in TeSR-E7 medium, which has the same composition nature methods
of TeSR-E8, but with no exogenous recombinant TGF-β1. After 24 h the conditioned media were collected from Petri dishes and microfluidic channels at two different frequencies, f = 8 d−1 or f = 2 d−1, and heat-treated to induce latent TGF-β activation32. The amount of the TGF-β, activin and nodal ligands was indirectly quantified by exposing CAGA12-luciferase HaCaT cells to the collected media. f = 2 d−1 medium contained higher amounts of TGF-β ligands compared to f = 8 d−1 or Petri dish-grown cells. Fresh E8 and E7 media (with no TGF-β1) were used as positive and negative controls, respectively. Addition of the smallmolecule TGF-β receptor inhibitor SB431542 confirmed that the inductions were caused by extracellular factors in the medium. Functional tests. HPSC-derived cardiomyocytes were analyzed by confocal microscopy (Leica SP5) for calcium transients during caffeine or verapamil treatment33. HPSC-derived hepatocyte-like cells were stained with PAS staining kit (Sigma-Aldrich), according to the manufacturer’s instructions, for glycogen storage analysis. Indocyanine green test, ICG, (Sigma-Aldrich) was performed by incubating live cells for 15 min with dye solution, according to manufacturer’s instructions. Green cells were photographed. After 6 h, the same fields were photographed in order to see complete clearance of ICG by functional hepatocyte-like cells. Albumin production and secretion was assessed by performing ELISA assays (Immunology Consultants Laboratory, Inc.), following manufacturer’s instructions, on spent media (10 µL from each micro-channel). Albumin levels were detected at 450 nm using an Infinite F2000 PRO (Tecan) plate reader. Cytotoxicity experiments were performed using the hepatotoxic drug, acetaminophen, on HepG2 cells and hESC-derived hepatocytelike cells. Acetaminophen BioXtra ≥99.0% (Sigma-Aldrich) was dissolved in DMSO to a 5 M stock solution. Acetaminophen stock solution was diluted in fresh HepG2 culture medium or hESC-derived hepatocytes’ medium at final concentrations of 50, 25, 12.5, 10, 5, 1 and 0.5 mM for drug stimulation. As controls, untreated cells were cultured in their standard culture media or in the same media containing 1% DMSO (mock control). Cells were treated for 24 h at f = 2 d−1 or with multiple 3 h administrations using a multiple-port microfluidic switch34. For nuclear area calculation, treated cells were fixed and stained with Hoechst, and fluorescence images were taken at 80X magnification. For cytotoxic effects on cell morphology and integrity, treated cells were fixed and stained with phalloidin for F-actin or anti-albumin antibody. Cell viability assays (LIVE/DEAD Kit, Life Technologies) were performed on cells after 24 h to quantify dead cells. Cells were washed with PBS; incubated with 4 µM ethidium homodimer-1 (stains dead cells red), 4 µM calcein AM (stains live cells green) and 4 µM Hoechst (stains cell nuclei blue) for 45 min at room temperature; washed with PBS; and analyzed by fluorescence microscopy. Cytotoxicity was expressed as fraction of dead cells over total cells per 20 image. Immunofluorescence. Immunofluorescence analyses were performed on 4% paraformaldehyde-fixed cells for 15 min. Blocking and permeabilization were performed in blocking buffer (5% heat-inactivated FBS and 0.1% TritonX-100 (Sigma-Aldrich) in PBS -/- 1×) for 1 h. Depending on the antibody, cells were stained for 1 h at room temperature or overnight at 4 °C using primary antibodies diluted in blocking buffer as follows: Oct4 doi:10.1038/nmeth.3411
© 2015 Nature America, Inc. All rights reserved.
1:200 (sc-5279 Santa Cruz), Sox-2 1:200 (AB5603 Millipore), Tra-1-60 1:200 (MAB4360 Millipore), Tra-1-81 1:200 (MAB4381 Millipore), β-III tubulin 1:500 (T3952 Sigma-Aldrich), α-fetoprotein 1:250 (A8452 Sigma-Aldrich), Brachyury 1:100 (ab20680 Abcam), cardiac troponin T 1:100 (MS-295-P Thermo Scientific), albumin 1:100 (MAB1455 R&D), CYP-3A 1:150 (GTX117120 Genetex) and cytokeratin-18 1:150 (GTX105624S Genetex). For SOX17 (AF1924, R&D) and FOXA2/HNF3b (D56D6, Cell Signaling) staining, cells were incubated in 5% horse serum in 0.3% Triton X-100 blocking solution for 1 h at room temperature. Primary antibodies were diluted 1:400 and 1:20, respectively, and incubated in 1% BSA in 0.3% TritonX-100 overnight at 4 °C. Immunostaining was done by incubating secondary antibodies (Alexa Fluor 488 or 594 (Life Technologies) and Cy3 (Jackson ImmunoLab)) or DAPI (for nuclear staining) for 1 h at 37 °C. Pictures were taken on a Leica DMI6000 B microscope. Gene expression analysis. For RNA extraction, cells were lysed with 0.5 ml of TRIzol (Life Technologies), treated with 0.1 ml chloroform (Sigma-Aldrich) for 3 min and centrifuged at 12,500 g for 15 min at 4 °C. Aqueous supernatant was collected and diluted 1:1 with 70% ethanol. Total RNA was extracted from solution using RNeasy Mini Kit (Qiagen), following the manufacturer’s instructions, and quantified using a NanoDrop spectrophotometer. RNA was used with A(260/280) nm = 2.0 ± 0.1, A(260/230) nm = 2.0 ± 0.1 and A(320) nm ≈0.05. RNA retro-transcription was performed using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies), according to the manufacturer’s instructions. The qRT-PCR was performed with TaqMan gene expression assay probes (Life Technologies) according to the manufacturer’s instructions. The following genes were used: GAPDH (glyceraldehyde 3-phosphate dehydrogenase), POU5F1 (OCT4), NANOG, DNMT3B (DNA cytosine-5-methyltransferase 3 beta), TUBB3 (beta-III tubulin), OTX2 (orthodenticle homeobox 2), T (BRACHYURY-T), GATA4, ACTA2 (alpha-smooth muscle actin), AFP (alpha-fetoprotein), EOMES, FOXA2 and SOX17. Reactions were performed on an ABI Prism 7000 machine and results were analyzed with ABI Prism 7000 SDS software. GAPDH
doi:10.1038/nmeth.3411
expression was used to normalize Ct values for gene expression, and data were shown as relative fold change to control cells (static condition), using the Livak method35. For Figure 2a, the samples with highest gene expression were used as control for each gene and ∆∆Ct values were re-scaled for graphical representation. Statistical analysis. For statistical analyses, single pairwise comparisons were analyzed using Student’s t-test with P ≤ 0.05 (*) or P ≤ 0.01 (**) indicating significance. Multiple comparisons were performed by one-way ANOVA with Tukey post-test, with P ≤ 0.05 (*) or P ≤ 0.01 (**) indicating significance. Throughout the text, n = 6 indicates the number of replicates, referring to a combination of experiments performed in 2 different chips (with different batches of cells) as well as in 3 different independent channels within the same chip (using the same batch of cells). The robustness of our conclusions on perfusion frequency to different sources of variability (between and within independent biological replicates) was also verified by crossed-nested ANOVA on selected marker expression, using Minitab 17 statistical software (Supplementary Note 2).
21. Luni, C., Michielin, F., Barzon, L., Calabrò, V. & Elvassore, N. Biophys. J. 104, 934–942 (2013). 22. Warren, L. et al. Cell Stem Cell 7, 618–630 (2010). 23. Zatti, S. et al. Mol. Ther. Methods Clin. Dev. 1, 1–9 (2014). 24. Rashid, S.T. et al. J. Clin. Invest. 120, 3127–3136 (2010). 25. Chambers, S.M. et al. Nat. Biotechnol. 27, 275–280 (2009). 26. Camnasio, S. et al. Neurobiol. Dis. 46, 41–51 (2012). 27. Kouskoff, V., Lacaud, G., Schwantz, S., Fehling, H.J. & Keller, G. Proc. Natl. Acad. Sci. USA 102, 13170–13175 (2005). 28. Hay, D.C. et al. Proc. Natl. Acad. Sci. USA 105, 12301–12306 (2008). 29. Lian, X. et al. Proc. Natl. Acad. Sci. USA 109, E1848–E1857 (2012). 30. Hay, D.C. et al. Stem Cells 26, 894–902 (2008). 31. Levy, L. et al. Mol. Cell. Biol. 27, 6068–6083 (2007). 32. Slager, H.G., Freund, E., Buiting, A.M.J., Feijen, A. & Mummery, C.L. J. Cell. Physiol. 156, 247–256 (1993). 33. Martewicz, S. et al. Integr. Biol. (Camb) 4, 153–164 (2012). 34. Zambon, A., Zoso, A., Luni, C., Frommer, W.B. & Elvassore, N. Integr. Biol. (Camb) 6, 277–288 (2014). 35. Livak, K.J. & Schmittgen, T.D. Methods 25, 402–408 (2001).
nature methods
