TECHNIQUES

MicroCT for Developmental Biology:A Versatile Tool for High-Contrast 3D Imagingat Histological ResolutionsBrian D. Metscher*

Understanding developmental processes requires accurate visualization and parameterization of three-dimensional embryos. Tomographic imaging methods offer automatically aligned and calibratedvolumetric images, but the usefulness of X-ray CT imaging for developmental biology has been limited bythe low inherent contrast of embryonic tissues. Here, I demonstrate simple staining methods that allowhigh-contrast imaging of embryonic tissues at histological resolutions using a commercial microCT system.Quantitative comparisons of images of chick embryos treated with different contrast agents show that threevery simple methods using inorganic iodine and phosphotungstic acid produce overall contrast anddifferential tissue contrast for X-ray imaging at least as high as that obtained with osmium. The stains canbe used after any common fixation and after storage in aqueous or alcoholic media, and on a wide varietyof species. These methods establish microCT imaging as a useful tool for comparative developmentalstudies, embryo phenotyping, and quantitative modeling of development. Developmental Dynamics 238:632–640, 2009. © 2009 Wiley-Liss, Inc.

Key words: microCT; X-ray computed tomography; three-dimensional imaging; chick; embryonic development; contrastagents

Accepted 9 December 2008

INTRODUCTION

The need for three-dimensional (3D)representations of developing em-bryos is as old as embryology itself,and our modern genomic, functional,mutational, and quantitative ap-proaches to studying developmentalprocesses are no less dependent on ac-curate knowledge of normal and af-fected tissues and structures (Dickin-son, 2006). This requires size-calibrated 3D images that preservethe volumetric relationships withinthe original specimen. Confocal micro-scopic methods provide such imagesfor samples that are no thicker than a

few hundred microns and usually flu-orescence-labeled (Wanninger, 2007),and histological reconstructions fromserial sections are extremely labor-in-tensive and give uneven results. Epi-scopic (Weninger et al., 2006) andother block-face (Ewald et al., 2002;Tyszka et al., 2005) microscopic imag-ing techniques provide automaticallyaligned high-resolution serial-sliceimage stacks of embedded samples,but the sample is destroyed in the im-aging process. For embryo-size sam-ples, tomographic imaging methodsare the most natural candidates fornondestructive volumetric imaging.

X-ray microtomography (microCT)is already widely used in studies ofmineralized tissues (e.g., Neues et al.,2007; Stauber and Muller, 2008;Vasquez et al., 2008), and it has beeneffectively used in quantitative stud-ies of variation (Hallgrimsson et al.,2007) and of development (Parsons etal., 2008). The applications of othertomographic methods, particularlymicro-MRI (Jacobs et al., 2003; Ruf-fins et al., 2007) and optical projectiontomography (Kerwin et al., 2004;Sharpe, 2004), have been demon-strated convincingly, but the use ofX-ray–based tomographic imaging

Additional Supporting Information may be found in the online version of this article.Department of Theoretical Biology, University of Vienna, Vienna, Austria*Correspondence to: Brian D. Metscher, Department of Theoretical Biology, Althanstrasse 14, University of Vienna, Vienna1090, Austria. E-mail: brian.metscher@univie.ac.at

DOI 10.1002/dvdy.21857Published online 16 February 2009 in Wiley InterScience (www.interscience.wiley.com).

DEVELOPMENTAL DYNAMICS 238:632–640, 2009

© 2009 Wiley-Liss, Inc.

has been constrained by the low in-trinsic X-ray absorption of unmineral-ized tissue and the lack of establishedcontrast agents. Phase contrast en-hancement of X-ray images (Betz etal., 2007), while useable on unstainedmaterial, is much more useful for ob-jects with distinct edges, rather thanthe softer gradients characteristic ofembryos. Three-dimensional X-raymicroscopy has been used to imagesubcellular structures (Larabell andLe Gros, 2004; Le Gros et al., 2005),but resolutions of tens of nanometersrequire a synchrotron source and afield of view too small to be of generalutility to organismal developmentalresearch.

To date, the most successful con-trast stain used for microCT imagingof soft tissues and embryos is osmiumtetroxide (Johnson et al., 2006; Bent-ley et al., 2007). The K-shell energy ofosmium (73.9 keV) and its known tis-sue binding properties (Kiernan,1990) make osmium tetroxide a natu-ral candidate for an X-ray contraststain. And although the staining is al-ready available in any institution withan EM facility, its tissue penetrationis limited (Hayat, 1970), it does notwork well on tissues that have beenpreserved in alcohol, and osmium isvolatile, toxic, and expensive to pur-chase and to dispose of.

This work presents quantitativecomparisons of contrast in microCTimages of chick embryos treated withseveral very simple contrast stains. Toillustrate the usefulness of microCTfor developmental biology, high-con-trast, true-volume images of chick em-

Fig. 1.

Fig. 1. Comparison of several contrast stains.Virtual sections through the hindlimb buds ofstage 24 chick embryos. For illustration, eachvolume image was re-sliced arbitrarily along thehindlimb proximodistal axis to show, amongother structures, the spinal ganglia, spinalnerves, myogenic mesenchyme, and apical ec-todermal ridge. The images were processed forillustration only by rotating, cropping, and ad-justing the window and level values (brightnessand contrast). The reconstructed voxel size inthese images after 2 � 2 binning is 4.6 �m (5.1�m for the osmium and IKI/4F1G images), giv-ing a practical spatial resolution limit of approx-imately 12 �m. All slices are one voxel thick andare shown at the same scale. The 10% IKIimage shows saturated brightness in the de-scending aorta most likely due to intense stain-ing of blood trapped during fixation.

MicroCT IMAGING OF EMBRYOS 633

bryos are presented showing develop-ing structures and tissues at spatialresolutions comparable to low-powerlight micrographs of histological sec-tions.

RESULTS AND DISCUSSION

Soft-tissue contrasts produced by sev-eral stains are illustrated in Figure 1,and quantitative comparisons of theraw tomographic images are given inFigure 2. Because computed tomo-graphic CT values represent linear X-ray absorption coefficients (Kalender,2005), the voxel grayscale values canbe used as a linear measure of relativeabsorption at different points in thesample for a given X-ray source spec-trum. Using a simple common refer-ence material—the alcohol in whicheach sample was scanned—directlycomparable measures of absorptionfor differently stained tissues were ob-tained.

Unstained embryonic tissue in alco-hol gives discernable X-ray contrast,but the pixel brightness distributionsof the object and background mediumoverlap considerably (Fig. 2). All con-trast is due to intrinsic differences intissue density and composition, andoverall X-ray contrast is low, around30% above that of ethanol. Contrastwas even lower for unstained samplesin water (not shown), possibly becauseof the condensing effect of alcohol dehy-dration on proteins (Kiernan, 1990).

Gallocyanin-chromalum, a histolog-ical stain for cell nuclei (Presnell andSchreibman, 1997), appears to impartX-ray contrast to tissues according totheir native cell densities (Fig. 1). Itgave a clearer overall contrast be-tween object and background than un-stained tissue, but not enough differ-ence to allow complete exclusion ofbackground by raising the limit ofpixel values displayed as black (i.e.,the lower limit of the brightness win-dow; Fig. 2, histogram).

The two inorganic iodine stainsgave results similar to each other, andboth are extremely simple to prepareand use (Table 1). Penetration into tis-sues was rapid, and contrast impartedto various tissues is excellent. Tissuesstored in alcohol stained well with 1%iodine in absolute ethanol (I2E) ormethanol (I2M) after being dehy-drated to pure alcohol. Iodine potas-

sium iodide (IKI, one formulation ofLugol’s solution) works well on fixedtissues that are still in aqueous me-dium. A 10% dilution of IKI in waterwas found to work as well as pure IKIbut with less tissue shrinkage. IKI-stained samples could also be scannedin water (not shown), and the stainingwas stable for at least a few months insamples stored in ethanol.

Phosphotungstic acid staining isequally simple (Table 1), and also pro-duces excellent contrast among differ-ent tissues. The staining is stable forat least several months, but penetra-tion is much slower than iodine: chickembryos later than approximatelystage 24 required overnight or longer.PTA is especially suitable for Bouin- orformalin-fixed material stored in 70%alcohol. Chick embryos fixed in Glyo-Fixx (glyoxal-based) and stored inmethanol worked well also. Phospho-molybdic acid was tested on a few sam-ples and gave similar results to PTA,but this stain was not investigatedfully, because PTA is at least as widelyused in histology and EM labs (Hayat,1970).

All the stains imparted clear con-trast to neural tissues and to denser

aggregations such as premyogenicmesenchyme and the apical ectoder-mal ridge. Epithelia can be distin-guished from mesenchyme, and somedetail can be seen within the neuraltube and other structures.

Osmium staining was tested along-side PTA and IKI on chick embryosfixed in 3.7% formaldehyde, 1% glu-teraldehyde in phosphate buffer(4F1G; McDowell and Trump, 1976).This fixation gave excellent tissuepreservation with all three stains. Theother stains tested here gave X-rayimages at least as good as those ob-tained with osmium (Fig. 1), and theyare simpler and more versatile. Thefiner detail in the osmium specimen isdue in large part to the fixative (4F1Ginstead of formalin), which, in combi-nation with osmium tetroxide postfix-ation, resulted in less shrinkage thanthe other treatments.

To quantify the differences amongstaining methods, the ratio of meanobject level to mean ethanol back-ground level was used as a simplemeasure of separation of signal fromnoise (Fig. 2). Because all samples usethe same background medium, thisratio is also directly proportional to

Fig. 2. Quantitative measures of contrast for the staining methods illustrated in Figure 1. Foranalysis, raw reconstructed tomographic sections were used before any processing or re-slicing.A histogram for each stain shows the distribution of pixel grayscale values, on a relative linear scale.Each graph has been scaled to the same mean for the ethanol background peak (blue dashed line).The mean of the object pixel distribution after scaling is marked by a vertical red dashed line, andthe orange bar indicates the width of one standard deviation above and below the mean. (Mediansfor all distributions were nearly equal to the means.) The smaller peak to the left of the alcohol peakin some histograms represents the polypropylene in the sample holders, depending on how muchof the plastic appeared in the selected slice.

634 METSCHER

TA

BL

E1.

Use

ful

Co

ntr

ast

Sta

ins

for

So

ft-T

issu

eM

icro

CT

Con

tras

tst

ain

Gal

locy

anin

-ch

rom

alu

mIK

I(L

ugo

l’s)

iodi

ne

pota

ssiu

mio

dide

I2E

,I2

Mal

coh

olic

iodi

ne

PT

Aph

osph

otu

ngs

tic

acid

Osm

ium

tetr

oxid

e

Sam

ple

fixa

tion

a

and

stor

age

NB

F,

Bou

in’s

;st

ill

inaq

ueo

us

med

ium

NB

F,

4F1G

;st

ill

inaq

ueo

us

med

ium

An

yfi

xati

on;

eth

anol

-or

met

han

ol-

pres

erve

dsp

ecim

ens

NB

F,

Bou

in’s

,4F

1G;

aqu

eou

sor

alco

hol

icst

orag

e

4F1G

orot

her

EM

fix;

not

alco

hol

-deh

ydra

ted

Sta

inso

luti

onb

Gal

locy

anin

-ch

rom

alu

m5%

(w/v

)in

wat

er.

Boi

lto

diss

olve

,co

olan

dfi

lter

.S

tabl

efo

r1-

2w

eeks

.

1%(w

/v)

iodi

ne

met

al(I

2)

�2%

pota

ssiu

mio

dide

(KI)

inw

ater

.S

tabl

efo

rm

onth

s.D

ilu

teto

10%

inw

ater

just

befo

reu

se.

1%(w

/v)

iodi

ne

(I2)

in10

0%et

han

ol(I

2E)

orm

eth

anol

(I2M

).S

tabl

efo

rm

onth

s.

1%(w

/v)

PT

Ain

wat

er.

Mix

30%

PT

Aso

luti

on,

70%

abs.

eth

anol

Sta

ble

for

mon

ths.

1-2%

OsO

4in

phos

phat

ebu

ffer

Sta

inin

gpr

oced

ure

Was

hsa

mpl

esin

wat

er.

Sta

inov

ern

igh

t.W

ash

inw

ater

.D

ehyd

rate

toet

han

ol.

Sam

ples

inaq

ueo

us

solu

tion

.S

tain

30m

into

over

nig

ht.

Was

hin

wat

er.

Deh

ydra

teto

eth

anol

.

Sam

ples

in10

0%al

coh

ol.

Sta

in30

min

toov

ern

igh

t.W

ash

inal

coh

ol.

Sam

ples

in70

%et

han

ol.

Sta

in2

hr

toov

ern

igh

t.W

ash

in70

%et

han

ol.

Sca

nin

70-1

00%

eth

anol

.

Sta

in1-

2h

r.W

ash

inph

osph

ate

buff

er.

Deh

ydra

teto

eth

anol

.

Adv

anta

ges

Dem

onst

rate

sce

llde

nsi

ties

and

indi

vidu

alce

lls.

Pos

sibl

yu

sefu

las

cou

nte

rsta

in.

Rap

idan

dde

eppe

net

rati

on.

Exc

elle

nt

con

tras

t.C

hea

p,si

mpl

e,n

onto

xic.

Ver

sati

le,

robu

st.

Exc

elle

nt

con

tras

t.S

tabl

est

ain

.S

har

pti

ssu

edi

ffer

enti

atio

n.

Com

bin

esre

adil

yw

ith

EM

.H

igh

con

tras

t.C

anbe

use

don

resi

nbl

ocks

.

Lim

itat

ion

sL

owov

eral

lco

ntr

ast.

May

over

stai

nso

me

min

eral

ized

tiss

ues

.In

ten

sity

and

perm

anen

cem

ayva

ryw

ith

diff

eren

tsa

mpl

espr

epar

atio

ns.

Slo

wti

ssu

epe

net

rati

on,

lim

ited

toa

few

mil

lim

eter

s.

Tox

ic,

expe

nsi

vedi

spos

al.

Lim

ited

pen

etra

tion

.N

otu

sefu

lfo

ral

coh

ol-

stor

edti

ssu

es.

aF

ixat

ives

:NB

F:n

eutr

al-b

uff

ered

form

alin

(3.7

%fo

rmal

deh

yde

in0.

1Mph

osph

ate

buff

er,p

H7.

0);4

F1G

:3.7

%(o

r4%

)fo

rmal

deh

yde,

1%gl

ute

rald

ehyd

ein

0.1M

phos

phat

ebu

ffer

;G

lyo-

Fix

x(T

her

mo

Ele

ctro

nC

orp.

Sh

ando

nG

lyo-

Fix

x)or

Pre

fer

(An

atec

hL

td.)

:gl

yoxa

l-ba

sed

form

alin

subs

titu

te.

bS

tain

s:ga

lloc

yan

in-c

hro

mal

um

:C

hro

ma

1A45

8(P

ress

nel

lan

dS

chre

ibm

an,

1997

);io

din

e:S

igm

a20

,777

-2;

pota

ssiu

mio

dide

:S

igm

a20

7969

;ph

osph

otu

ngs

tic

acid

:M

alli

nck

rodt

2824

.

MicroCT IMAGING OF EMBRYOS 635

the ratio of stained tissue level to un-stained tissue level. The means mea-sure overall stained-tissue contrastfor this X-ray source (tungsten at 40kV), and the dispersion of object pixelvalues reflects the range of brightnesslevels among stained tissues. Becauseoutlying points in the brightness dis-tribution are important for differenti-ating among tissue features, a root-mean-square measure like standarddeviation is more appropriate herethan a linear nonparametric measuresuch as interquartile range. It is im-portant to note that this is a measureof the difference between two absorb-ing materials and not a measure ofimage background noise.

A desirable feature in a micrograph isthe possibility of setting the brightnesslevel and window so that background iseliminated. A simple nonparametricmeasure of object-background separa-tion is given by the overlap percentilebetween the two distributions (Fig. 2,last column). This is calculated simplyas the percentile score at which the left-tail percentile object distribution isequal to the right-tail percentile back-ground distribution. This measure is in-dependent of the number of pixels ineither sampled distribution, but it doesrequire separate sampling of the twodistributions. It gives the fraction ofpixels within the object that have a 50%or greater probability of having theirbrightness value due to absorption bythe intercalating medium rather thanobject. An image display window func-tion with its lower bound set at thispoint will eliminate all but that percent-age of background pixels, at a cost of thesame percentage of object pixels. Thismay or may not correspond to the opti-mum window for overall image quality,but it is a consistent quantitative mea-sure of object and medium separation inimages of an object in an absorbing me-dium. The precision with which theoverlap percentile can be estimated isdetermined by the percentile differ-ences between adjacent histogram bins(hence, the different precisions for theoverlap percentiles given in Fig. 2).

To demonstrate the usefulness ofthese methods in embryology re-search, normal chick embryos of sev-eral stages were stained and imagedat different magnifications (Figs. 3–7).At stage 12, the virtual sections andpartial volume renderings may be

more informative than the completeoutward view (Fig. 3, top). The invag-inating otic placode and neural crestare represented clearly with both IKIand gallocyanin staining (Figs. 3, 4),and the image resolution, even after2 � 2 pixel binning, is sufficient toshow individual cells.

The volume rendering of a stage 14chick head (Fig. 5 and Supp. Video,which is available online) shows theheart loop, the developing regions ofthe brain, the rhombomeres, the otic

vesicles, and the cranial nerves (Supp.Video S1, which is available online).The reconstructed images preservethe original shapes and positions oftissues and organs in the scannedsample. CT imaging avoids distortionsdue to realignment of separately im-aged sections, but any distortion orshrinkage due to sample fixation ordehydration must still be accountedfor just as in histology or any otherimaging method.

At stage 24, the placement of thecranial nerves is clearly shown (Fig. 6and Supp. Video), as are the posteriorspinal nerves and ganglia. As withmost tissue stains, PTA and iodinegive clearer contrast to differentiatedtissues than to embryonic mesen-chyme. Even so, the premyogenic mes-enchyme in the hindlimbs is demon-strated effectively by both stains(Figs. 1, 6).

Fig. 3. Stage 12 chick embryo stained with10% IKI (iodine potassium iodide). Top: A vol-ume rendering of the head shown at two differ-ent angles. Reconstructed cubic voxel size is3.5 �m after 2 � 2 binning. Center: Highermagnification scan of the same stage (0.97 �mvoxels), showing an oblique cutaway volumerendering. View is from anterior, and the virtualcut passes through the otic placode on the rightand posterior to it on the left, where neural crestcells can be seen. Bottom: A single virtual sec-tion through the same volume image. Scalebars � 100 �m.

Fig. 4. Stage 12 chick embryo stained withgallocyanin-chromalum. Individual cells areclearly distinguishable, although no specifictests were performed to confirm that the X-raycontrast is restricted to cell nuclei. Tissue thattook up less of the stain, especially undifferen-tiated mesenchyme, shows little or no X-raycontrast above the background and, therefore,appears as dark in a single (one voxel thick)section. Voxel size, 0.99 �m. Scale bar � 100�m.

636 METSCHER

The stains tested here are similar intheir overall contrast enhancement,and none of them is tissue-specific.PTA is acidic in solution, and may re-sult in decalcification or other chemi-cal changes to stained tissues; furthertests are needed to determine theseeffects. Iodine has been observed tooverstain some mineralized tissues,giving areas of saturation in X-ray im-ages. Both stains tend to bind heavilyto yolk, which can be a difficulty whenimaging embryos of some species.

The staining methods are not spe-cies-specific, and are predicted to giveuseful results for almost any model ornonmodel animal. A few samples ofmouse, Xenopus, and zebrafish em-bryos have been tested with PTA andI2E, and the preliminary results (notshown) indicate that the presentedstaining methods will give useful re-sults with other model species. In ad-dition to model species embryos, oneor another of these stains has beenused successfully on samples from no

fewer than 20 different species in fivechordate classes and four invertebratephyla. It is anticipated that these con-trast methods will find broad applica-tions in various areas of animal re-search.

By using very simple contrast stains,microCT can be used to make high-con-trast, high-resolution, true-3D imagesof embryonic and other unmineralizedtissues. Because the imaging is nonde-structive, it can be combined with otheranalysis methods such as transmissionelectron microscopy or histology, and itcan be used on rare or irreplaceablesamples, such as rare embryos or oldmuseum specimens. The methods pre-sented can be used for specimens in liq-uid media or embedded in resins, withno need for drying. Tomographic im-ages are inherently volumetric and au-tomatically aligned and size-calibrated,so they are directly useful for quantita-tive studies and theoretical modeling ofdevelopment. All of the embryos imagedfor this demonstration are of unper-turbed phenotypes, but the range of ap-plicability of these methods to studies ofnatural and artificial genetic variants,experimentally manipulated embryos,and different species are not difficult toenvisage.

EXPERIMENTALPROCEDURES

The self-contained X-ray microtomog-raphy system used in this work ismodel MicroXCT from Xradia Inc.,Concord, CA (www.xradia.com). Thisscanner uses a Hamamatsu L9421-02tungsten X-ray source with an anodevoltage between 20 and 90 kV, powerbetween 4 and 8W, and a spot size of 5to 8 �m. This scanner’s configurationallows fields of view from 5 mm downto less than 500 �m. For the demon-strations presented here, the beamwas filtered only by the beryllium win-dow of the X-ray source housing. TheX-ray projection image is formed on ascintillator crystal, made in-house byXradia. The optical emissions of thescintillator is then imaged by a Nikonmicroscope objective lens onto a CCDcamera cooled to �55°C to reducedark noise. The optical imaging of thescintillator allows a final magnifica-tion independent of the geometricmagnification of the X-ray projectionimaging, and a final image resolution

Fig. 5. Stage 14 chick embryo, PTA stained. Top: Volume rendering of the head and pharyngealregion, 4.9 �m voxels. The heart loop, otic vesicles, cranial nerves, and rhombomeres are clearlyvisible. Bottom: Region of somitogenesis, volume rendering and single horizontal virtual section.Posterior is toward the top of the figure. Voxel size, 0.99 �m. Scale bars: top � 500 �m, bottom �100 �m.

MicroCT IMAGING OF EMBRYOS 637

that is not limited by the X-ray sourcespot size. Several different optical ob-jective lenses allow selection of the fi-nal magnification, while adjustmentsto the source-sample and sample-de-tector distances can be used to changethe geometric magnification of thesample image on the scintillator.

Projection images are collected andtomographic slices reconstructed byXradia’s supplied software running ona dedicated Dell Precision 490 com-puter. The system automatically sub-tracts a background image from eachprojection, and the reconstruction pro-gram uses a ring-artifact-reductionutility. Projection images were col-lected as 1,024 � 1,024-pixel framesand binned 2 � 2 during reconstruc-tion, partly to reduce noise and partlyto reduce file size, resulting in imageblocks of 512 � 512 � 512 voxels (ac-tually slightly less due to clipping out-side the cone shape of the X-raybeam). The 3D images were stored as8-bit grayscale image stacks, requir-ing approximately 130 MB of diskspace if saved in TIFF format. It ispossible to reconstruct and use thefull-size images, but each volume im-age is over 1 GB in size and the com-puting can become cumbersome.

For the contrast comparisons in Fig-ures 1 and 2, a set of stage 24 chickembryos (Hamburger and Hamilton,1992) were fixed in 10% phosphate-buffered formalin and stained over-night in 10% IKI, 1% iodine in etha-nol, 0.3% PTA in 70% ethanol, or 5%gallocyanin-chromalum in water (de-tails are given in Table 1). Each stainwas washed out with its respectivesolvent, and all samples were dehy-drated to ethanol. Unstained sampleswere simply fixed and dehydrated toethanol for scanning. A second set ofembryos was fixed in 4F1G (3.7%formaldehyde � 1% gluteraldehyde in0.2 M phosphate buffer) overnight forstaining with osmium (1% osmium te-troxide in 0.05% phosphate buffer for2 hr), IKI, and PTA.

The scans used in this analysis weremade with a source voltage of 40 kV(40 keV maximum X-ray energy), witha 20-sec exposure every 0.25° over180° rotation. All samples werescanned in ethanol in 200-�l polypro-pylene pipette tips, using the �4 ob-jective lens in the MicroXCT, and allat the same source-sample and sam-

ple-detector distances (40 mm and 16mm, respectively). Comparable sliceswere taken directly from each recon-structed 3D image, before any furtherprocessing, for analysis. The signaland background of each image weresampled by outlining most of the alco-hol-filled space or the embryo tissueusing the lasso tool in ImageJ 1.40,and the object and background regionswere analyzed by using the histogramtool to obtain numerical data and

summary statistics for each pixelbrightness distribution. The distribu-tions were normalized to the samemean background value (which repre-sents a CT value for ethanol). The lo-cation and dispersion of CT values ofthe stained tissues could then be com-pared directly. Histograms of the ac-tual sampled data were produced us-ing Plot 0.997, and the final chart wasassembled with Adobe IllustratorCS3.

Fig. 6. Stage 24 chick embryos. Left: IKI-stained head showing brain and cranial nerves (9.5 �mvoxels). Bottom left: inferior view of the brain, eyes, and cranial nerves. Right: PTA-stained pelvicregion showing hindlimb buds, spinal nerves and ganglia, and mesonephros (4.6 �m voxels). Whilethese stains are not tissue-specific, they do allow clear differentiation among structures.

638 METSCHER

To illustrate the different contraststains (Fig. 1), virtual re-sliced imageswere processed from the scans usedfor the analysis presented in Figure 2.The display levels were adjusted inPhotoshop CS3 to show the contrastsand clarity obtainable with each stain.For all but the osmium scan, thesource-sample distance was 40 mmand the sample-detector distance was16 mm. The osmium sample was

scanned with the source at 60 mm togive a slightly larger field of view, be-cause the 4F1G-fixed, osmicated sam-ple shrank less than the formalin-fixed samples.

Scans of other chick embryos weremade similarly, but using combina-tions of source and detector distancesand objective magnifications appropri-ate to the different specimen sizes.Samples were scanned in ethanol, al-

though other liquids can be used. Tominimize the amount of fluid sur-rounding the specimen, each samplewas placed in either a 0.2-ml PCRtube or a micropipette tip (both arepolypropylene, with 200- to 300-�m-thick walls). For smaller embryos, amicropipette tip (200 or 1,000 �l) wasflame-sealed at the tip, filled with al-cohol, and agitated to remove bubbles.The sample was simply placed in thealcohol and allowed to sink, thenpressed gently into the conical cavityto brace it against rotation duringscanning. The sample was easily re-moved without damage by cutting offthe tip below it and washing it out.Larger samples were placed in a PCRtube full of alcohol and held in placeby a small piece of alcohol-soakedsponge. The tube could then be in-verted and mounted on a sampleholder (provided with the microCTscanner) with adhesive putty (e.g.,UHU Patafix).

Reconstructions of samples stainedwith heavy metals are sometimes im-proved by adjusting the beam-harden-ing constant used by the reconstruc-tion program. This factor helpsaccount for greater attenuation oflower energy (“softer”) X-rays as thebroadband beam traverses the sampleand can remove some “shadow” effectsin the reconstructed slices. The recon-structions of PTA-stained embryosshown in Figure 6 used a beam-hard-ening constant of 0.6 instead of thedefault value of 0.

Volume images were displayed us-ing the software provided by Xradia(TXM3DViewer), which allows export-ing of virtual section images as well asvolume projections and movies. Therendered volume images can be ad-justed in size, orientation, windowing,and transparency.

ACKNOWLEDGMENTSThe author thanks Prof. Gerd Mullerfor advice and guidance, and the Uni-versity of Vienna for supporting thisresearch.

REFERENCES

Bentley MD, Jorgensen SM, Lerman LO,Ritman EL, Romero JC. 2007. Visualiza-tion of three-dimensional nephron struc-ture with microcomputed tomography.Anat Rec (Hoboken) 290:277–283.

Fig. 7. Guide to some features visible in Figures 3–6. AER, apical ectodermal ridge; da, dorsalaorta; DM, derma-myotomes; drg, dorsal root ganglia; e, eye; HL, heart loop; mes, mesencephalon;mm, premyogenic mesenchyme; mn, mesonephros; n, notochord; NC, neural crest; NT, neuraltube; ph, pharynx; np, nasal placode; OtV, otic vesicle; Rh, rhombomeres; sn, spinal nerves; So,somite; y, yolk granules; II, optic nerve; III, oculomotor nerve; V, trigeminal nerve.

MicroCT IMAGING OF EMBRYOS 639

Betz O, Wegst U, Weide D, Heethoff M,Helfen L, Lee WK, Cloetens P. 2007. Im-aging applications of synchrotron X-rayphase-contrast microtomography in bio-logical morphology and biomaterials sci-ence. I. General aspects of the techniqueand its advantages in the analysis of mil-limetre-sized arthropod structure. J Mi-crosc 227:51–71.

Dickinson ME. 2006. Multimodal imagingof mouse development: tools for the post-genomic era. Dev Dyn 235:2386–2400.

Ewald AJ, McBride H, Reddington M,Fraser SE, Kerschmann R. 2002. Surfaceimaging microscopy, an automatedmethod for visualizing whole embryosamples in three dimensions at high res-olution. Dev Dyn 225:369–375.

Hallgrimsson B, Lieberman DE, Liu W,Ford-Hutchinson AF, Jirik FR. 2007.Epigenetic interactions and the struc-ture of phenotypic variation in the cra-nium. Evol Dev 9:76–91.

Hamburger V, Hamilton HL. 1992. A se-ries of normal stages in the developmentof the chick embryo. 1951. Dev Dyn 195:231–272.

Hayat MA. 1970. Principles and tech-niques of electron microscopy: biologicalapplications. New York: Van NostrandReinhold.

Jacobs RE, Papan C, Ruffins S, Tyszka JM,Fraser SE. 2003. MRI: volumetric imag-ing for vital imaging and atlas construc-tion. Nat Rev Mol Cell Biol Suppl:SS10–SS16.

Johnson JT, Hansen MS, Wu I, Healy LJ,Johnson CR, Jones GM, Capecchi MR,Keller C. 2006. Virtual histology oftransgenic mouse embryos for high-throughput phenotyping. PLoS Genet2:e61.

Kalender WA. 2005. Computed tomogra-phy: fundamentals, system technology,image quality, applications. Erlangen:Publicis Corporate Publishing. 304 p.

Kerwin J, Scott M, Sharpe J, Puelles L,Robson SC, Martinez-de-la-Torre M,Ferran JL, Feng G, Baldock R, StrachanT, Davidson D, Lindsay S. 2004. 3 di-mensional modelling of early humanbrain development using optical projec-tion tomography. BMC Neurosci 5:27.

Kiernan JA. 1990. Histological and histo-chemical methods: theory and practice.Oxford: Pergamon Press.

Larabell CA, Le Gros MA. 2004. X-ray to-mography generates 3-D reconstructionsof the yeast, saccharomyces cerevisiae,at 60-nm resolution. Mol Biol Cell 15:957–962.

Le Gros MA, McDermott G, Larabell CA.2005. X-ray tomography of whole cells.Curr Opin Struct Biol 15:593–600.

McDowell EM, Trump BF. 1976. Histologicfixatives suitable for diagnostic light andelectron microscopy. Arch Pathol LabMed 100:405–414.

Neues F, Goerlich R, Renn J, Beckmann F,Epple M. 2007. Skeletal deformations inmedaka (Oryzias latipes) visualized bysynchrotron radiation micro-computertomography (SRmicroCT). J Struct Biol160:236–240.

Parsons TE, Kristensen E, Hornung L,Diewert VM, Boyd SK, German RZ, Hall-grimsson B. 2008. Phenotypic variabilityand craniofacial dysmorphology: in-creased shape variance in a mouse modelfor cleft lip. J Anat 212:135–143.

Presnell JK, Schreibman MP. 1997. Huma-son’s animal tissue techniques. Balti-more: Johns Hopkins University Press.

Ruffins SW, Martin M, Keough L, TruongS, Fraser SE, Jacobs RE, Lansford R.

2007. Digital three-dimensional atlas ofquail development using high-resolutionMRI. ScientificWorldJournal 7:592–604.

Sharpe J. 2004. Optical projection tomogra-phy. Annu Rev Biomed Eng 6:209–228.

Stauber M, Muller R. 2008. Micro-com-puted tomography: a method for the non-destructive evaluation of the three-di-mensional structure of biologicalspecimens. In: Methods in molecular bi-ology, Vol. 455. Osteoporosis: methodsand protocols. Totowa, NJ: HumanaPress. p 273–292.

Tyszka JM, Ewald AJ, Wallingford JB,Fraser SE. 2005. New tools for visualiza-tion and analysis of morphogenesis inspherical embryos. Dev Dyn 234:974–983.

Wanninger A. 2007. The application of con-focal microscopy and 3D imaging softwarein functional, evolutionary, and develop-mental zoology: reconstructing myo- andneurogenesis in space and time. In: Men-dez-Vilas A, Dıaz J, editors. Modern re-search and educational topics in micros-copy. Badajoz, Spain: FORMATEX. p 353–361.

Weninger WJ, Geyer SH, Mohun TJ,Rasskin-Gutman D, Matsui T, Ribeiro I,Costa Lda F, Izpisua-Belmonte JC, Mul-ler GB. 2006. High-resolution episcopicmicroscopy: a rapid technique for highdetailed 3D analysis of gene activity inthe context of tissue architecture andmorphology. Anat Embryol (Berl) 211:213–221.

Vasquez SX, Hansen MS, Bahadur AN,Hockin MF, Kindlmann GL, Nevell L, WuIQ, Grunwald DJ, Weinstein DM, JonesGM, Johnson CR, Vandeberg JL, CapecchiMR, Keller C. 2008. Optimization of volu-metric computed tomography for skeletalanalysis of model genetic organisms. AnatRec (Hoboken) 291:475–487.

640 METSCHER