Dynamic Article LinksC<Lab on a Chip

Cite this: Lab Chip, 2011, 11, 1815

www.rsc.org/loc PAPER

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

Zebrafish embryo development in a microfluidic flow-through system†

Eric M. Wielhouwer,‡a Shaukat Ali,‡a Abdulrahman Al-Afandi,a Marko T. Blom,b Marinus B. Olde Riekerink,b

Christian Poelma,c Jerry Westerweel,c Johannes Oonk,b Elwin X. Vrouwe,b Wilfred Buesink,b

Harald G. J. vanMil,a Jonathan Chicken,d Ronny van ’t Oeverb and Michael K. Richardson*a

Received 24th September 2010, Accepted 17th March 2011

DOI: 10.1039/c0lc00443j

The zebrafish embryo is a small, cheap, whole-animal model which may replace rodents in some areas

of research. Unfortunately, zebrafish embryos are commonly cultured in microtitre plates using cell-

culture protocols with static buffer replacement. Such protocols are highly invasive, consume large

quantities of reagents and do not readily permit high-quality imaging. Zebrafish and rodent embryos

have previously been cultured in static microfluidic drops, and zebrafish embryos have also been raised

in a prototype polydimethylsiloxane setup in a Petri dish. Other than this, no animal embryo has ever

been shown to undergo embryonic development in a microfluidic flow-through system. We have

developed and prototyped a specialized lab-on-a-chip made from bonded layers of borosilicate glass.

We find that zebrafish embryos can develop in the chip for 5 days, with continuous buffer flow at

pressures of 0.005–0.04 MPa. Phenotypic effects were seen, but these were scored subjectively as

‘minor’. Survival rates of 100% could be reached with buffer flows of 2 mL per well per min. High-

quality imaging was possible. An acute ethanol exposure test in the chip replicated the same assay

performed in microtitre plates. More than 100 embryos could be cultured in an area, excluding

infrastructure, smaller than a credit card. We discuss how biochip technology, coupled with zebrafish

larvae, could allow biological research to be conducted in massive, parallel experiments, at high speed

and low cost.

Introduction

In biological and biomedical research, there is an unmet need for

low-cost, high-throughput in vitro animal models.1,2 Whole

animal models are valuable because they provide data that may

be extrapolated to humans; they also allow complex organismal

functions (e.g. behavior and development) to be studied.3 On the

other hand, in vitro models (e.g. cell and tissue culture) offer the

advantages of low cost, of being less prone to legal and ethical

restrictions and of having the ability to be scaled-up to high-

throughput (1000–10 000 assays per day4) or ultra high

throughput (100 000 assays per day5).

Zebrafish (Danio rerio) embryos have been proposed as an in

vitro animal model which could bridge the gap between simple

assays based on cell culture, and biological validation in whole

aInstitute of Biology, Leiden University, Sylvius Laboratory, Sylviusweg72, 2333, BE, Leiden, The Netherlands. E-mail: richardsonmk@biology.leidenuniv.nlbMicronit Microfluidics BV, Enschede, The NetherlandscLaboratory for Aero & Hydrodynamics, Delft University of Technology,The NetherlandsdFLIR Systems LTD, Nottingham, UK

† Electronic supplementary information (ESI) available. See DOI:10.1039/c0lc00443j

‡ These authors contributed equally.

This journal is ª The Royal Society of Chemistry 2011

animals such as rodents.1 The zebrafish embryo has a small size,

low-cost, rapid development and can be raised easily in large

numbers. It also has a transparent body, whichmakes it relatively

easy to collect numerous data points using high-quality imaging

(including the fluorescence imaging of transgenic lines6). This

species is therefore suitable not only for high throughput screens,

but also for high content screens. Indeed it is already beginning to

be used in large-scale screens and assays.6–8 Several zebrafish-

embryo assays can help to predict drug safety in humans,9,10 and

zebrafish disease models have been developed.11,12

Unfortunately, the ambitions that biologists have for zebrafish

embryos have outstripped the available culture protocols, bor-

rowed as they are from traditional cell culture. Thus, zebrafish

embryos are commonly raised for assay purposes in plastic

microtitre plates or Petri dishes.7,13 In Table 1 we give examples

of assay protocols in zebrafish studies. Typically, the buffer is

refreshed periodically (‘static renewal’) or not at all (‘static non-

renewal’).14 Periodic aspiration and replacement of the buffer is

extremely invasive, causing stress to zebrafish embryos and

requiring enormous care in order to avoid embryos being

damaged or sucked up. Another issue is that static replacement

regimes may not be ideal for the zebrafish, a species which nor-

mally breeds in slow-flowing waters.15 The imaging of embryos in

a microtitre plate is distorted, not only by the depth of the buffer

filling the well, but by the curved meniscus of the buffer which

Lab Chip, 2011, 11, 1815–1824 | 1815

Table 1 A selection of assays to give an indication of protocols currently used in zebrafish research. In this case we list ethanol teratogenicity assays toshow the wide variation in setup used for testing a single reagent. Note: dpf ¼ days post fertilisation; hpf ¼ hours post fertilisation

Durationof exposure

Stageof exposure Plate format Ref.

Acute (1 h) 3–4 month Aquarium (15 l) 33Acute (2 h) 1 dpf Petri dish, 60 per dish, tank, 20 per tank 34Acute (3 h) 256 cells, high, dome/30% epiboly, germ-ring Petri dishes or glass beakers 27Acute (1 h) 4 month Tank 35Acute (1 h) 6 dpf 96-Well plate 30Acute (20 min) 7 dpf 10 per chamber 8 � 6 � 2 cm 25Chronic 6–24, 12–24, 24–36, 48–60, 60–72 hpf Petri dish 36Chronic (2 weeks) Young adult 5-Gal aquarium 37Chronic 6–24, 12–24, 24–36, 48–60, 60–72 hpf Petri dish 38Chronic 1 dpf Aquarium 39Chronic (3 d) 1 dpf 6-Well plate, 10 per well 40Chronic (3 d) and acute (4 h) 2 dpf 6-Well plate 41Chronic (6 h) 1 dpf Petri dish 42Chronic (ca. 20 h) 1 dpf Petri dish 43Chronic (6 d) 1 dpf 24-Well plate, 10 per well 44Chronic (ca. 20 h) 1 dpf 5 ml (format not specified) 45Chronic (ca. 20 h) 1 dpf Petri dishes or glass beakers 46Chronic (ca. 24 h) 1 dpf Glass beaker 47

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

interferes with phase contrast and bright-field microscopy. A

final problem with traditional, static replacement regimes is the

possible bolus effect resulting from the intermittent replacement

of buffer and its contained drug.

An example of static non-renewal culture of zebrafish embryos

is their successful growth inside Teflon� tubing, each embryo

being isolated in a drop of buffer.16 Chronic exposure to drugs is

possible in such a system, but the embryo is not accessible during

the experiment. Furthermore, culture in Teflon� tubing involves

distortion of the image because of the curved surfaces, and does

not provide continuous buffer refreshment. Another study in

which the microfluidic culture of embryos is described (mouse, in

that case) used static culture in droplets with no fluidic flow.17

Another approach was developed by a student team and

reported in an educational-themed issue of the journal Zebrafish.

Unfortunately, no relevant biological data were given in that

paper, although the authors claim that the zebrafish could

survive for a few days in their single-well PDMS (poly-

dimethylsiloxane) open set-up in a Petri dish.18 While this is

a very intriguing report, the ‘biorector’ described does not appear

to meet the criteria of a high throughput, microfluidic

technology.

We believe that true microfluidic technology could provide

a quantum leap forward in this field by providing non-invasive

culture conditions and high-quality imaging, as well as the ability

to access the embryo at any stage of development (this not being

possible, for example, with the culture of embryos inside lengths

of tubing). For the purposes of our study, true microfluidic

technology involves: continuous flow-through (‘dynamic

renewal’) of pressurised buffer; the embryos being continuously

accessible and isolated in parallel arrays to prevent cross-

contamination; the use of small culture volumes to save valuable

compounds; and the capability of high quality, real-time imaging

of the embryo.

Microfluidics have been used for sorting nematode (Caeno-

rhabditis elegans) embryos into 96-well plates, and for holding

nematode larvae.19 Unfortunately, no animal embryo has been

shown to be capable of developing in a true microfluidic

1816 | Lab Chip, 2011, 11, 1815–1824

environment. Our aim here is to determine whether zebrafish

embryos can indeed complete normal organogenesis under such

conditions.

Results and discussion

Properties of the biochip

The microfluidic chip that we have designed and tested here

(Fig. 1a) is made of three layers of bonded borosilicate glass, with

an array of wells connected in parallel by channels. The

temperature of the wells can be controlled by water flowing

through in-built heating channels (Fig. 1b and c). Thermal

imaging showed that the temperature difference between the

extreme ends of any row (Fig. 1b and c) ranged from 0.08–

0.53 �C at a flow rate of heating water of 1 ml per min per chip.

Several prototypes were tested, having wells of either 1.5 mm,

1.67 mm, 1.83 mm or 2 mm inner diameter; the four well sizes

were in either round or square shape, giving eight biochip

prototypes in total. For comparison, we determined that the

diameter of a pre-hatching zebrafish embryo (2–4 somite stage)

averages 806 mm (SD, 31) or 1237 mm (SD, 24) including the

chorion (N¼ 15). Therefore the embryo fits comfortably into the

well, and can swim around after hatching (Movies S1 and S2†),

but is confined within a much smaller physical area than is the

case with 96-well plates.

Fluidic flow

We next looked at how buffer circulated in a well containing an

embryo. Obstruction of the buffer flow by the embryo, or by

debris (including the chorion shed at hatching), is prevented by

having 3 inlet and 3 outlet channels per well (Fig. 2a). Cross-

contamination between wells (e.g. by chemicals or infectious

agents) is ameliorated by the parallel arrangement of wells

(Fig. 1a). The accumulation of air bubbles under the lid is pre-

vented by having the outlet channels positioned high on the wall

(Fig. 2b and c). In Movie S1†, several air bubbles can be seen to

spontaneously shrink and disappear during operation of the chip.

This journal is ª The Royal Society of Chemistry 2011

Fig. 1 Layout and thermal properties of the biochip. (a) An example of

a 32-well chip. The pre-heated water circulating in the temperature

channels (tc) is completely isolated from the buffer in which the embryos

develop. Buffer enters the chip via inlet (bi) and outlet (bo) ports, and

circulates through a parallel array of feeder channels. These in turn are

connected to the wells (w) by smaller channels, each of which is connected

to the well by three terminal branches, so each well has a total of 6

connections (3 inflow, 3 outflow). Scale bar ¼ 1 cm. (b) Thermographic

recording made with a FLIR SC5600-M large format infrared camera of

a 24-well chip running in a room with 20 �C ambient temperature.

Heating water was circulated via thermal inlet (ti) channels and the

common thermal outlet (to) channel. The temperature of heating water

entering at ti1 was 26.06 �C; ti 2, 26.20 �C; ti 3, 25.04 �C; ti4, 25.11 �C. (c)Temperature profile from (b), showing that a stable temperature can be

maintained in different wells of the chip. There difference along each row

(i.e. between wells 1 and 6) averages 0.18 �C. A stable temperature can be

maintained in the chip even though the surrounding air is at room

temperature.

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

We tested fluidic properties of the wells using micro-PIV

techniques.20,21 The flow regime over the range of flow rates of

1.0–10.0 mL per well per min, was laminar, with no vortices or

turbulent motion (Fig. 2d–i). This is predictable, given that the

dimensionless Reynold’s number (Re ¼ rVD/h) is significantly

smaller than unity, indicating that viscous forces dominate over

any inertial effects. Interestingly, it can be seen in Movie S3† that

twitching movements of the embryo caused periodic rearrange-

ment of the flow pattern of buffer in the well.

Embryo survival in the biochip

We next wanted to determine whether zebrafish embryos could

survive and develop in a microfluidic environment because this

This journal is ª The Royal Society of Chemistry 2011

has never been shown for any animal embryo. As controls, we

cultured zebrafish embryos in conventional 96-well microtitre

plates with periodic (24 h) static renewal of buffer. We dispensed

8 hpf zebrafish embryos into the biochip, 1 per well. We plated

them into the chip at 8 h, because this avoids the mortality-wave

seen in earlier stages.22

We first looked at survival of embryos after 5 days of culture.

We chose 5 days as the cut-off point in order to conform to local

ethical requirements. However, at five days, most of the organs

are developed and the larva already shows a complex behav-

ioural repertoire.

In the 96-well plate controls, there was 100% survival at 5 d. In

the biochip, 5 d survival of the zebrafish was strongly influenced

by the buffer flow-rate (Fig. 3). The average 5 d survival from all

experiments at each flow rate was as follows: 36.7% survival at

0.5 mL per well per min; 53.9% at 1.0; 88.3% at 2.0; 87.5% at 4.0;

and 71.1% at 6.0 mL per well per min. Survival in the biochip at

2.0–4.0 mL per well per min was significantly higher (p < 0.01

using one-way ANOVA) than at 0.5 mL per well per min.

Considering single runs with individual chips, survival ranged

from 19.0% (one chip run at 0.5 mL per well per min) to 100%

(two chips run at 2 mL per well per min). In the biochip, well size

and shape had no effect on survival (biochip wells of 1.67 mm or

1.83 mm inner diameters, round or square, were examined; two-

way ANOVA). Interestingly, the hatching of embryos in the

biochip (Fig. 3b) was suppressed at higher flow rates.

To see whether oxygen shortage could explain the poor

survival at 0.5 mL per well per min, we forcibly aerated the buffer

reservoir using an aquarium air-stone. The average 5 d survival

at 0.5 mL per well per min with no extra aeration was 36.7%; at

the same flow rate, with forced aeration of the buffer, survival

only slightly increased to 43.8%. Thus, forced aeration was not

able to raise the survival rates to those seen at 2.0 or 4.0 mL per

well per min, implying that some deleterious effect other than

oxygen depletion is at work when low flow rates are used.

Phenotypic screening of embryos

Surviving embryos at 5 d were fixed, stained and screened

morphologically (Fig. 3c and 4). The phenotypes scored are listed

in Table 2, and the data given in Table 3. Embryos grown in the

biochip had a shorter average body length (3.63 mm) compared

to controls in 96-well plates (3.84 mm) (see Fig. 5a). This

difference was highly significant (p > 0.001 using an ANCOVA

linear regression model). Greater embryo-length was obtained

for flow rates of 1, 4 and 6 mL per well per min respectively

compared to 0.5 mL per well per min, the difference being very

significant (p < 0.05, df 467; Box–Cox analysis followed by

a factorial ANOVA performed on the squared embryo-length

against buffer flow as an explanatory factor).

To look for signs of stress, we examined the phenotype of

melanin-containing pigment cells (melanocytes) at 5 days.

Pigment dispersion is known to occur in teleosts exposed to

stressors.23–25 We performed a cubic transformation of the ratio

between the number of embryos with contracted pigment

pattern, and the total number of embryos, and a two-way

ANOVA for factors FLOWRATE andWELL-TYPE. No effect

was found for FLOW RATE, but a significant effect was

observed for WELL-TYPE, the ratio being 0.96 for 96-well

Lab Chip, 2011, 11, 1815–1824 | 1817

Fig. 2 Fluidic flow in the biochip. (a) Zebrafish embryo in a 2.0 mm (inner diameter) well to show the positioning and relative size of an embryo in the

well (photographed without a lid, and therefore without buffer flow, so as to give an unobstructed view of the connections); tc, temperature channel; bi,

buffer inlet, bo, buffer outlet. The central, darker ring (r) in the picture is the sandblasted wall of the well. (b) Schematic drawing showing the levels of the

z-planes and x� y coordinates, and their relation to the inlet channels (blue) and outlet channels (orange). (c) Schematic diagram of flow pattern around

and over the embryo. (d–i) micro-PIV recordings of flow patterns in a fluidic flow experiment using micro-PIV techniques.20,21 Polystyrene particles (1.3

mm diameter, containing rhodamine and coated with PEG) were used. The flow rate used in this experiment was 1 mL per well per min and the inner

diameter of the well was 1.5 mm. (d, e, and f) schematic interpretations of the actual recordings (g, h, and i), respectively. Red arrows in g, h, and i

indicate the most rapid flow, blue arrows the slowest, and blue dots represent flow parallel to the z-axis. bi, position of buffer inlet; bo, position of buffer

outlet; r, wall of well.

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

plates versus 0.64 for biochips (p < 0.01, df ¼ 8). In summary,

there is a significantly higher incidence of a putative stress

phenotype (dispersed pigment granules) in embryos grown in the

biochip.

Other phenotypic effects were screened for, and we found mild

yolk sac oedema (compare Fig. 4b and c) much more commonly

in the biochip embryos than in controls (p < 0.001, generalized

linear model). However, there was no significant difference in the

incidence between biochip and 96-well culture (p > 0.05) of other

malformations (cardiac oedema, pectoral fin hypoplasia, bran-

chial arch hypoplasia, hypoplasia of Meckel’s cartilage, bent tail,

and bent primary axis).

Acute ethanol exposure test

To see whether we could introduce bioactive compounds into the

wells during an experiment, we introduced 10% ethanol into the

inflow stream of the biochip for 1 h at prim-16 stage. Ethanol was

chosen because of several reasons: it produces a strong

1818 | Lab Chip, 2011, 11, 1815–1824

phenotypic effect in the zebrafish embryo;12,26 it passes easily

through the chorion;27 and it is widely used as a carrier solvent28

in biomedical research. We scored the embryos at 5 d for

a variety of phenotypic abnormalities (Table 2). Then, we clas-

sified each embryo as to whether it was mild, moderate or severe

in terms of its clustering of phenotypic abnormalities, using the

criteria in Table 4. In control (vehicle only) embryos there was

a very low incidence of severely abnormal embryos (Table 5). But

with ethanol exposure, the percentage of severely abnormal

embryos increased in both the 96 well plate (to 85%) and the

biochip (to 65%).

Implications of the current findings

It is not clear what causes the minor phenotypic effects (e.g. yolk

sac oedema, slight reduction in body length) in biochip-raised

embryos. Possible explanations could be physical features of the

biochip itself, such as the small size of the wells. Fig. 4a and

Movie S2† show that there is sufficient room for the embryos to

This journal is ª The Royal Society of Chemistry 2011

Fig. 3 Data on survival and hatching percentage of embryos cultured in the biochip. (a) 5 d survival in the biochip, plotted as a function of buffer flow

rate. The results from 4 different chip designs (each with wells of a particular size and shape) are shown. For each data point, N ¼ 32 (embryos, except

those for the biochip experiments with a flow rate of 1 mL per well per min, where N ¼ 64). For each experiment, a control was run simultaneously in

a 96-well plate, placed next to the biochip and using embryos of the same stage and from the same mating. For each control, N ¼ 32 embryos. (b)

Hatching rate as a percentage of surviving embryos at different time points and at different flow rates of buffer (in mL per min per well). Higher flow rates

in the biochip tend to delay or suppress hatching, relative to controls. Each error bar represents� SEM (standard error of the mean) ofN¼ 48 embryos,

that is, 4 replicate chips, with 12 embryos each, per flow rate; for the 96-well plate controls, N ¼ 160 embryos, consisting of 5 replicates each with 32

embryos. (c) Incidence of malformations in surviving embryos at 5 days in 96 well plates and biochips (each line representing a different flow rate). The

number of embryos affected by various malformations is shown (note that some embryos had multiple malformations and therefore the numbers do not

add up to 100%). Note also that the 96-well plates had daily static replacement regimes and so the different ‘flow-rates’ indicated for 96-well plates are

simply referring to the biochip run for which they were the control. Key: pf, pectoral fin hypoplasia; bt, bent tail; ys, yolk sac oedema; pc, pericardial

oedema; Mc, Meckel’s cartilage hypoplasia; normal, none of the other malformations in this series were seen; ba, branchial arch hypoplasia; bc,

curvature of body axis.

This journal is ª The Royal Society of Chemistry 2011 Lab Chip, 2011, 11, 1815–1824 | 1819

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

Fig. 4 Embryos cultured in the biochip. (a) Consecutive photos of the same embryo developing in the same well (1.8 mm internal diameter) of a 32 well

biochip with a buffer flow of 2 mL per min per well (note that by 4 days, the embryo had swum into a different focal plane in the upper part of the well).

Each picture is framed by a circular hole in the metal clamp that holds the lid in place. Notice that between 1 dpf and 2 dpf, the embryo has changed

position within the chorion. (b and c) Two embryos grown in the biochip and fixed at 5 days (Alcian blue staining and clearing); ys ¼ yolk sac; (b)

morphologically normal embryo; (c) embryo with mild yolk sac oedema. Scale bar in a ¼ 1 mm; scale bar in b and c ¼ 500 mm.

Table 3 The effects of flow rate in the biochip on survival and pheno-types of embryos at 5 dpf. For each flow rate, a 96-well plate control withstatic replacement of buffer was established. Biochip data for each flowrate are pooled from chip versions with differences in well diameter. Thisis because the statistical analysis revealed no significant effect of well sizeor shape on survival (see main text)

Flow rate/mL permin Well

Total Dead LostaSurvivors(5 dpf)b Normal Abnormalc

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

lie straight and swim around until at least 4 days of culture.

Nonetheless, the embryo has approached the limits of the well at

5 d and this is perhaps an explanation for the increase in the

putative stress phenotype (dispersed melanocytes) seen in the

biochip embryos at that time.

The ethanol exposure experiment shows that compounds can

be introduced into the buffer stream at will and can produce an

effect on embryo development. The percentage incidence of

severe abnormalities after ethanol exposure in the biochip was

less than that after the same test conducted in a 96-well plate. A

possible explanation is that the specific gravity of ethanol is lower

than that of water, resulting in a failure of the ethanol to enter the

lower part of the well where the embryo is located.

What are the potential advantages and limitations of micro-

fluidic embryo culture? Our study suggests several advantages for

Table 2 Phenotype analysis. Description of the seven categories used toscore larval phenotype at 5 dpf

Larval phenotype Criteria

1. Normal Absence of any of the phenotypeslisted below:

2. Pigmentation Dispersed melanocytes3. Heart Presence of pericardial oedema4. Yolk Presence of yolk sac (vitelline) oedema5.Meckel’s cartilage Meckel’s cartilage grossly hypoplastic, missing

or unfused in midline. These effects may beunilateral or bilateral

6.Branchial arches One or more cartilages of the branchialskeleton hypoplastic or missing

7. Pectoral fins One or both pectoral fins hypoplastic or missing

1820 | Lab Chip, 2011, 11, 1815–1824

medical and life sciences research. In the field of high-

throughput, high content research (i.e. when very large numbers

of samples are screened, and a large number of data points is

collected) a biochip could represent a major advance over

conventional microtitre plates. First, the biochip concentrates

embryos into a small physical area, because the wells are both

per well format N N (%) N (%) N (%) N (%) N (%)

0.5 96-well 32 0 (0) 0 (0) 32 (100) 29 (91) 3 (9)Chips 128 81 (63) 5 (4) 42 (33) 15 (36) 27 (64)

1.0 96-well 32 0 (0) 0 (0) 32 (100) 31 (97) 1 (3)Chips 128 92 (72) 3 (2) 33 (26) 11 (33) 22 (67)

2.0 96-well 32 0 (0) 1 (3) 31 (97) 24 (77) 7 (23)Chips 128 15 (12) 12 (9) 101 (79) 37 (37) 64 (63)

4.0 96-well 32 0 (0) 2 (6) 30 (94) 23 (77) 7 (23)Chips 128 16 (13) 8 (6) 104 (81) 17 (16) 87 (84)

6.0 96-well 32 0 (0) 3 (9) 29 (91) 20 (69) 9 (31)Chips 128 37 (29) 10 (8) 81 (63) 20 (25) 61 (75)

a ‘Lost’ indicates that embryos were lost during processing (mostlythrough aspiration during pipetting of buffer or other reagents).b Phenotype at 5 dpf was classified as normal or abnormal according tothe criteria in Table 2. c Abnormal embryos were classified as mild,moderate or severe according to the criteria listed in Table 4.

This journal is ª The Royal Society of Chemistry 2011

Fig. 5 Further characterisation of the phenotypes of embryos grown in the biochip versus controls grown in 96-well plates. (a) Chart of body length at 5

dpf for surviving embryos grown in the biochip (pooled data from all biochip version 1 models) versus their respective 96-well plate controls (which had

static volumes with no buffer flow). The abscissa gives different flow rates per well in the biochip. The number inside the base of the bars ¼ N embryos.

(b) Relative incidence among 5 dpf survivors of dispersed versus contracted melanocyte morphology.

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

small and close together. This facilitates parallel imaging of

multiple wells at high resolution. It also reduces the ‘seek’ time

needed to locate embryos, either manually or automatically.

Furthermore, there is no water meniscus in the biochip to

distort the image, and the walls are frosted, avoiding mirroring

artefacts. Finally, the biochip is made of optical-quality glass,

and the depth of fluid covering the embryos is much less than that

in a 96-well plate. As shown in Movie S4† it is possible to obtain

high quality imaging of zebrafish lines in the chip.

With respect to buffer replacement, the great advantage of

a microfluidic flow-through system is that there is no repeated

invasion and disruption of the embryos environment by draining

This journal is ª The Royal Society of Chemistry 2011

and refilling. This is crucial because zebrafish embryos are

sensitive to handling stress.29

The small volume of the biochip wells (ca. 10 ml) means that

some types of experiment, particularly acute drug exposures, will

consume far less precious reagent or drug than the same exper-

iment in microtitre plates (for example, 250 ml is typically used

per well). This makes the biochip promising for drug discovery

where only small quantities of compound are available, or where

the compound is very expensive.

The most obvious limitation of microfluidic systems is size; at

some point, the embryos will simply outgrow the wells. This is

not, of course, a problem in short-term assays up to 4 or 5 d. It

Lab Chip, 2011, 11, 1815–1824 | 1821

Table 4 Severity scale used to express the degree to which individualembryos were phenotypically abnormal. Branchial and Meckel’s carti-lage defects were excluded from the ‘moderate’ category becausecraniofacial defects are typical of severe ethanol teratogenicity in theclinical situation.48

Severity Criteria

Mild An individual embryohad a single abnormality(of any type listed in Table 2)

Moderate An individual embryo had any twodefects, excluding branchial and Meckel’scartilage abnormalities (i.e. the embryoshowed two from categories 2–4,or 7, in Table 2)

Severe An individual embryo had alteration of thebranchial arches and/or Meckel’s cartilagecombined with one or more of otherdefects (2–7 in Table 2)

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

should also be noted that the small well size of the biochip does

not necessarily mean a saving of test compound—if chronic

exposure is used for the full 5 days of embryo culture. For

example, with a buffer flow rate of 2 mL per well per min, the

biochip consumes 14.4 ml of buffer per embryo over 5 days. This

compares with 2.35 ml per embryo in a 96-well plate for the same

time period, assuming daily buffer replacement (see Methods for

buffer refreshment protocol). In such cases, conventional 96-well

zebrafish assays, which have now a high level of automation,13

remain a valuable alternative.

In conclusion, we show that zebrafish embryos can develop

normally in complete isolation with constant buffer flow and

strictly controlled environmental parameters. Crucially, the

zebrafish embryos can undergo normal organogenesis in a pres-

surized fluid stream. This is important because pressurized flow

distinguishes microfluidics systems from conventional culture

systems. Problems that need to be solved include the rather high

occurrence of mild yolk sac oedema in the biochip embryos.

Ultimately, it is possible that the microfluidic culture of verte-

brate embryos could become a bridge between conventional cell

culture, and whole animal models. Such a bridge is especially

important, given the need to find alternatives to mammalian

whole-animal experiments; and the need to find new medicines

by means of more efficient assays.

Table 5 General outcomes of ethanol treatment. Overview of total numbabnormalities at 5 dpf, and the degree of severity of those abnormalities

Morphology (5 dpf)a

Well format TreatmentTotal Dead Lostc SurviN N (%) N (%) N (%

96-well Veh 96 2 (2) 26 (27) 68 (71EtOH 96 70 (73) 6 (6) 20 (21

Chip (1.5 mm, square) Veh 64 13 (20) 12 (19) 39 (61EtOH 64 23 (36) 4 (6) 37 (58

a Morphology at 5 dpf was classified as normal or abnormal according to theor severe according to the criteria listed in Table 4. c ‘Lost’ indicates that epipetting of buffer or other reagents).

1822 | Lab Chip, 2011, 11, 1815–1824

Experimental

Biochip design

The biochip prototypes were specially fabricated for this study by

Micronit Microfluidics (Enschede, The Netherlands). They

consists of three layers of bonded borosilicate glass into which an

array of channels and wells is introduced by etching and powder-

blasting. A range of prototypes was tested, with slight variations

in design (round or square cross-sections, and well sizes of 1.5–

2.0 mm inner diameter). Temperature channels permit the

circulation of pre-warmed water through the chip. The open

wells can be closed by a sandwich of a silicon polymer sheet, and

a layer of glass, the whole assembly being compressed in a holder

to make it watertight.

Embryo preparation

Adult zebrafish (Danio rerio) were maintained in aquaria at 26 �Cunder a cycle of 14 h light, 10 h dark. Eggs were obtained by

random pair-wise mating. The eggs were transferred to 9 cm Petri

dishes containing 0.1� Hanks’ Balanced Salt Solution30 (0.1�HBSS) at pH 7.46, and periodically rinsed to remove debris and

dead embryos. The HBSS did not contain methylene blue or

antibiotics.

Experimental setup

All biochip culture experiments were carried out at 28.0 � 0.5 �Cunder a 14 h/10 h light/dark cycle. Embryos of 8 h post-

fertilisation (hpf) were loaded, one per well, with intact chorion,

into the biochip with residual buffer. Then, the lid was sealed (see

above), the flow of buffer (0.1� HBSS) was initiated, and the

setup ran continuously until the embryos had reached 5 dpf. Inlet

and outlet channels of the wells were fed with the buffer from

a high-performance liquid chromatography (HPLC) pump,

without a de-gasser. The biochips were connected to the pump in

parallel by means of phenyl/methyl deactivated capillary tubing

(150 mm inner diameter and 375 mm outer diameter; BGB Ana-

lytik AG: Schlossboeckelheim), and cross interconnectors.31 The

buffer reservoir was not actively aerated but had a loose-fitting

foil cover. Buffer was not recirculated.

er embryos treated, survival at 5 dpf, the presence of morphological

Severity of abnormalityb

vors (5 dpf) Normal Abnormal Mild Moderate Severe) N (%) N (%) N (%) N (%) N (%)

) 35 (51) 33 (49) 20 (61) 11 (33) 2 (6)) 0 (0) 20 (100) 3 (15) 0 (0) 17 (85)) 11 (28) 28 (72) 9 (32) 15 (54) 4 (14)) 5 (14) 32 (86) 1 (3) 7 (22) 24 (75)

criteria in Table 2. b Abnormal embryos were classified as mild, moderatembryos were lost during processing (mostly through aspiration during

This journal is ª The Royal Society of Chemistry 2011

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

Controls

To compare performance of the biochipwith conventional culture

conditions we made control cultures using zebrafish embryos

cultured in conventional 96-well microtitre plates. Embryos of

8 hpf, randomly selected from the same batches used for the

biochip experiments, were established in 96-well microtitre plates

(Costar 3599, Corning Inc., NY). One embryo was placed in each

well, with 250 mL 0.1� HBSS. The buffer was replenished every

24 h. The daily refreshment of buffer was done by replacing

175 mL, three times. The buffer, temperature and light cycle were

the same as described above for the biochip experiments.

Acute ethanol exposure

Embryos were established in the biochip (1.5 mm internal well

diameter, 1.0 mL per well per min flow rate), and in 96-well plates

(controls) as described above, except that the embryos used were

all at the prim-16 stage32 (approximately 1.5 dpf). This is because

preliminary data indicated that this was an ethanol-sensitive

stage. Embryos were exposed for 1 h to 10% ethanol (1.64 M in

0.1� HBSS) or buffer only as controls. The ethanol was high

purity, medical grade (Emprove� ethanol, Cat. No. 100971,

Merck KGaA, Darmstadt, Germany). Exposure to ethanol in

the biochip was accomplished by temporarily disconnecting the

normal buffer reservoir, and connecting in its place, for 1 h, the

10% ethanol/buffer reservoir.

Embryos were not dechorionated since the chorion is known

to be completely permeable to ethanol.27 The ethanol exposure

was followed by rinsing with fresh 0.1� HBSS (4 rinses in the

case of 96-well cultures). All embryos were then further cultured

as described above, until five days.

Analysis of embryos and statistical analysis design

Total body length was measured from the rostral margin of

Meckel’s cartilage in Alcian blue stained embryos, to the caudal

extremity of the caudal fin fold. Statistical analysis for Fig. 3c

was made in R. The need for data transformations was assessed

by post-diagnostic and Box–Cox analysis of the model. Count

data were analysed as a generalized linear model whereas

continuous data as ANCOVA model.

Fixation and staining of embryos

At 5 dpf, embryos were fixed overnight in 4% paraformaldehyde

and stained with Alcian blue as follows: embryos were rinsed 5

times in distilled water and dehydrated in a graded series of

ethanol (25%, 50%, and 70%) for 5 minutes each. They were then

rinsed in acid alcohol (1% concentrated hydrochloric acid in 70%

ethanol) for 10 minutes and placed in filtered Alcian blue solu-

tion (0.03%Alcian blue in acid alcohol) overnight. Embryos were

then differentiated in acid alcohol for 1 h, washed 2 � 30 min in

distilled water. Finally, they were cleared and stored in 100%

glycerol.

Conclusions

We show for the first time that an animal embryo can develop in

a microfluidic environment. Zebrafish embryos grown in the

This journal is ª The Royal Society of Chemistry 2011

biochip commonly showed minor phenotypic effects—including

the possible stress phenotype of dispersed melanocytes.

However, there was no increase in gross malformations. Our

scoring of the former as ‘minor’ defects is of course subjective,

and does not rule out the possibility of significant but undetected

effects. There was a strong, non-linear relationship between

buffer flow-rate and 5 d embryo survival in the biochip, such that

the optimal flow rate was in the range 2.0–4.0 mL per well per

min. Survival rates at 5 d reached 100% in two chip runs at 2.0 mL

per well per min. These results could lead to a new generation of

assays for the pharmaceutical industry based on the low-cost,

microfluidic culture of zebrafish embryos.

Acknowledgements

We thank Jan de Sonneville and Dr Maxim E. Kuil for helpful

discussions on imaging, microfluidics applications and for

comments on the manuscript; Ewie de Kuyper, Arjen C. Geluk,

Jeroen Mesman, Frits (M.W.) van Tol, and Tim A. P. Sanders

helped design and build prototype biochip holders; Peter Steen-

bergen and Ulrike Nehrdich for culturing zebrafish embryos;

Edwin Heida for scientific illustration; Nathan D. Lawson and

Brant M. Weinstein for transgenic Fli1 eGFP zebrafish; Michael

Richardson gratefully acknowledges the financial support of the

Smart Mix Program of the Netherlands Ministry of Economic

Affairs and the Netherlands Ministry of Education, Culture and

Science.

Reference

1 G. J. Lieschke and P. D. Currie, Animal models of human disease:zebrafish swim into view, Nat. Rev. Genet., 2007, 8, 353–367.

2 J. Bull and B. Levin, Perspectives: microbiology. Mice are not furryPetri dishes, Science, 2000, 287, 1409–1410.

3 D. M. Barnes, Tight money squeezes out animal models, Science,1986, 232, 309–311.

4 A. S. Verkman, Drug discovery in academia, Am. J. Physiol. CellPhysiol., 2004, 286, C465–C474.

5 A. Dove, Drug screening—beyond the bottleneck, Nat. Biotechnol.,1999, 17, 859–863.

6 R. Dahm and R. Geisler, Learning from small fry: the zebrafish asa genetic model organism for aquaculture fish species, Mar.Biotechnol., 2006, 8, 329–345.

7 A. Zebrafish, Practical Approach, Oxford University Press, NewYork, 2002.

8 F. Emran, J. Rihel and J. E. Dowling, A behavioral assay to measureresponsiveness of zebrafish to changes in light intensities, J. Vis. Exp.,2008.

9 W. S. Redfern, G. Waldron, M. J. Winter, P. Butler, M. Holbrook,R. Wallis and J. P. Valentin, Zebrafish assays as early safetypharmacology screens: paradigm shift or red herring?, J. Pharmacol.Toxicol. Methods, 2008, 58, 110–117.

10 S. Berghmans, P. Butler, P. Goldsmith, G. Waldron, I. Gardner,Z. Golder, F. M. Richards, G. Kimber, A. Roach, W. Alderton andA. Fleming, Zebrafish based assays for the assessment of cardiac,visual and gut function—potential safety screens for early drugdiscovery, J. Pharmacol. Toxicol. Methods, 2008, 58, 59–68.

11 S. A. Brittijn, S. J. Duivesteijn, M. Belmamoune, L. F. Bertens,W. Bitter, J. D. de Bruijn, D. L. Champagne, E. Cuppen, G. Flik,C. M. Vandenbroucke-Grauls, R. A. Janssen, J. I. de, E. R. deKloet, A. Kros, A. H. Meijer, J. R. Metz, A. M. van der Sar,M. J. Schaaf, S. Schulte-Merker, H. P. Spaink, P. P. Tak,F. J. Verbeek, M. J. Vervoordeldonk, F. J. Vonk, F. Witte,H. Yuan and M. K. Richardson, Zebrafish development andregeneration: new tools for biomedical research, Int. J. Dev. Biol.,2009, 53, 835–850.

Lab Chip, 2011, 11, 1815–1824 | 1823

Dow

nloa

ded

by R

ijksu

nive

rsite

it L

eide

n on

11

June

201

1Pu

blis

hed

on 1

4 A

pril

2011

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0044

3JView Online

12 R. L. Tanguay and M. J. Reimers, Analysis of ethanol developmentaltoxicity in zebrafish, Methods Mol. Biol., 2008, 447, 63–74.

13 C. Pardo-Martin, T. Y. Chang, B. K. Koo, C. L. Gilleland,S. C. Wasserman and M. F. Yanik, High-throughput in vivovertebrate screening, Nat. Methods, 2010, 7, 634–636.

14 Methods for Measuring the Acute Toxicity of Effluents and ReceivingWaters to Freshwater and Marine Organisms (EPA-821-R-02-012),United States Environmental Protection Agency, Washington, DC,2002.

15 R. Spence, G. Gerlach, C. Lawrence and C. Smith, The behaviour andecology of the zebrafish, Danio rerio, Biol. Rev. Cambridge Philos.Soc., 2008, 83, 13–34.

16 A. Funfak, A. Brosing, M. Brand and J. M. Kohler, Micro fluidsegment technique for screening and development studies on Daniorerio embryos, Lab Chip, 2007, 7, 1132–1138.

17 J. Melin, A. Lee, K. Foygel, D. E. Leong, S. R. Quake andM. W. Yao, In vitro embryo culture in defined, sub-microlitervolumes, Dev. Dyn., 2009, 238, 950–955.

18 Y. C. Shen, D. Li, A. Al-Shoaibi, T. Bersano-Begey, H. Chen, S. Ali,B. Flak, C. Perrin, M. Winslow, H. Shah, P. Ramamurthy,R. H. Schmedlen, S. Takayama and K. F. Barald, A student teamin a University of Michigan biomedical engineering design courseconstructs a microfluidic bioreactor for studies of zebrafishdevelopment, Zebrafish, 2009, 6, 201–213.

19 C. B. Rohde, F. Zeng, R. Gonzalez-Rubio, M. Angel andM. F. Yanik, Microfluidic system for on-chip high-throughputwhole-animal sorting and screening at subcellular resolution, Proc.Natl. Acad. Sci. U. S. A., 2007, 104, 13891–13895.

20 C. Poelma, K. Van der Heiden, B. P. Hierck, R. E. Poelmann andJ. Westerweel, Measurements of the wall shear stress distribution inthe outflow tract of an embryonic chicken heart, J. R. Soc.,Interface, 2010, 7, 91–103.

21 S. T. Wereley and C. D. Meinhart, Recent advances in micro-particleimage velocimetry, Annu. Rev. Fluid Mech., 2010, 42, 557–576.

22 B. Fraysse, R. Mons and J. Garric, Development of a zebrafish 4-dayembryo-larval bioassay to assess toxicity of chemicals, Ecotoxicol.Environ. Saf., 2006, 63, 253–267.

23 E. Hoglund, P. H. Balm and S. Winberg, Skin darkening, a potentialsocial signal in subordinate arctic charr (Salvelinus alpinus): theregulatory role of brain monoamines and pro-opiomelanocortin-derived peptides, J. Exp. Biol., 2000, 203, 1711–1721.

24 J. Peng, M. Wagle, T. Mueller, P. Mathur, B. L. Lockwood,S. Bretaud and S. Guo, Ethanol-modulated camouflage responsescreen in zebrafish uncovers a novel role for cAMP andextracellular signal-regulated kinase signaling in behavioralsensitivity to ethanol, J. Neurosci., 2009, 29, 8408–8418.

25 B. Lockwood, S. Bjerke, K. Kobayashi and S. Guo, Acute effects ofalcohol on larval zebrafish: a genetic system for large-scale screening,Pharmacol., Biochem. Behav., 2004, 77, 647–654.

26 M. J. Reimers, A. R. Flockton and R. L. Tanguay, Ethanol- andacetaldehyde-mediated developmental toxicity in zebrafish,Neurotoxicol. Teratol., 2004, 26, 769–781.

27 P. Blader andU. Strahle, Ethanol impairs migration of the prechordalplate in the zebrafish embryo, Dev. Biol., 1998, 201, 185–201.

28 A. Hallare, K. Nagel, H. R. Kohler and R. Triebskorn, Comparativeembryotoxicity and proteotoxicity of three carrier solvents tozebrafish (Danio rerio) embryos, Ecotoxicol. Environ. Saf., 2006, 63,378–388.

29 D. L. Champagne, C. C. Hoefnagels, R. E. de Kloet andM. K. Richardson, Translating rodent behavioral repertoire tozebrafish (Danio rerio): relevance for stress research, Behav. BrainRes., 2010, 332–342.

1824 | Lab Chip, 2011, 11, 1815–1824

30 R. C. Macphail, J. Brooks, D. L. Hunter, B. Padnos, T. D. Irons andS. Padilla, Locomotion in larval zebrafish: influence of time of day,lighting and ethanol, Neurotoxicology, 2009, 30, 52–58.

31 J.C. Stachowiak,E.E. Shugard,B. P.Mosier,R.F.Renzi, P. F.Caton,S. M. Ferko, J. L. Van de Vreugde, D. D. Yee, B. L. Haroldsen andV. A. VanderNoot, Autonomous microfluidic sample preparationsystem for protein profile-based detection of aerosolized bacterialcells and spores, Anal. Chem., 2007, 79, 5763–5770.

32 C. B. Kimmel, W. W. Ballard, S. R. Kimmel, B. Ullmann andT. F. Schilling, Stages of embryonic development of the zebrafish,Dev. Dyn., 1995, 203, 253–310.

33 R. Gerlai, M. Lahav, S. Guo and A. Rosenthal, Drinks like a fish:zebra fish (Danio rerio) as a behavior genetic model to study alcoholeffects, Pharmacol., Biochem. Behav., 2000, 67, 773–782.

34 Y. Fernandes and R. Gerlai, Long-term behavioral changes inresponse to early developmental exposure to ethanol in zebrafish,Alcohol.: Clin. Exp. Res., 2009, 601–609.

35 R. Gerlai, F. Ahmad and S. Prajapati, Differences in acute alcohol-induced behavioral responses among zebrafish populations,Alcohol.: Clin. Exp. Res., 2008, 32, 1763–1773.

36 J. Bilotta, S. Saszik, C. M. Givin, H. R. Hardesty andS. E. Sutherland, Effects of embryonic exposure to ethanol onzebrafish visual function, Neurotoxicol. Teratol., 2002, 24, 759–766.

37 C. A. Dlugos and R. A. Rabin, Ethanol effects on three strains ofzebrafish: model system for genetic investigations, Pharmacol.,Biochem. Behav., 2003, 74, 471–480.

38 J. Bilotta, J. A. Barnett, L. Hancock and S. Saszik, Ethanol exposurealters zebrafish development: a novel model of fetal alcoholsyndrome, Neurotoxicol. Teratol., 2004, 26, 737–743.

39 Z. Lele, S. Engel and P. H. Krone, hsp47 and hsp70 gene expression isdifferentially regulated in a stress- and tissue-specific manner inzebrafish embryos, Dev. Genet., 1997, 21, 123–133.

40 C. A. Dlugos and R. A. Rabin, Ocular deficits associated with alcoholexposure during zebrafish development, J. Comp. Neurol., 2007, 502,497–506.

41 J. I. Matsui, A. L. Egana, T. R. Sponholtz, A. R. Adolph andJ. E. Dowling, Effects of ethanol on photoreceptors and visualfunction in developing zebrafish, Invest. Ophthalmol. Visual Sci.,2006, 47, 4589–4597.

42 Y. X. Li, H. T. Yang, M. Zdanowicz, J. K. Sicklick, Y. Qi, T. J. CampandA.M.Diehl, Fetal alcohol exposure impairsHedgehog cholesterolmodification and signaling, Lab. Invest., 2007, 87, 231–240.

43 F. J. Arenzana, M. J. Carvan, III, J. Aijon, R. Sanchez-Gonzalez,R. Arevalo and A. Porteros, Teratogenic effects of ethanol exposureon zebrafish visual system development, Neurotoxicol. Teratol.,2006, 28, 342–348.

44 M. J. Carvan, III, E. Loucks, D. N. Weber and F. E. Williams,Ethanol effects on the developing zebrafish: neurobehavior andskeletal morphogenesis, Neurotoxicol. Teratol., 2004, 26, 757–768.

45 E. J. Loucks and S. C. Ahlgren, Deciphering the role of Shh signalingin axial defects produced by ethanol exposure,Birth Defects Res., PartA, 2009.

46 E. J. Loucks, T. Schwend and S. C. Ahlgren, Molecular changesassociated with teratogen-induced cyclopia, Birth Defects Res., PartA, 2007, 79, 642–651.

47 B. Kashyap, L. C. Frederickson andD. L. Stenkamp,Mechanisms forpersistent microphthalmia following ethanol exposure during retinalneurogenesis in zebrafish embryos, Vis. Neurosci., 2007, 24, 409–421.

48 A. Rostand, M. Kaminski, N. Lelong, P. Dehaene, I. Delestret,C. Klein-Bertrand, D. Querleu and G. Crepin, Alcohol use inpregnancy, craniofacial features, and fetal growth, J. Epidemiol.Community Health, 1990, 44, 302–306.

This journal is ª The Royal Society of Chemistry 2011