Download pdf - Generational Analysis Reveals that TGF-β1 Inhibits the Rate of Angiogenesis in Vivo by Selective Decrease in the Number of New Vessels

R

o

Microvascular Research 59, 221–232 (2000)doi:10.1006/mvre.1999.2213, available online at http://www.idealibrary.com on

Generational Analysis Reveals that TGF-b1 Inhibits theate of Angiogenesis in Vivo by Selective Decrease

in the Number of New Vessels

Patricia Parsons-Wingerter,* Katherine E. Elliott,† Andrew G. Farr,*Krishnan Radhakrishnan,‡ John I. Clark,* and E. Helene Sage*,§*Department of Biological Structure, School of Medicine, University of Washington, Seattle, Washington 98195;†3Com Corporation, 3535 128th Avenue, SE Suite 110, Bellevue, Washington 98006; ‡ICOMP, NASA GlennResearch Center, Cleveland, Ohio 44135; and §Department of Vascular Biology, The Hope Heart Institute,528 18th Avenue, Seattle, Washington 98122

Received May 3, 1999

E

Quantitative analysis of vascular generational branchingdemonstrated that transforming growth factor-b1 (TGF-b1), a multifunctional cytokine and angiogenic regulator,strongly inhibited angiogenesis in the arterial tree of thedeveloping quail chorioallantoic membrane (CAM) by in-hibition of the normal increase in the number of new,small vessels. The cytokine was applied uniformly in so-lution at embryonic day 7 (E7) to the CAMs of quailembryos cultured in petri dishes. After 24 h the rate ofarterial growth was inhibited by as much as 105% as afunction of increasing TGF-b1 concentration. Inhibitionf the rate of angiogenesis in the arterial tree by TGF-b1

relative to controls was measured in digital images bythree well-correlated, computerized methods. The firstcomputerized method, direct measurement by the com-puter code VESGEN of vascular morphological parame-ters according to branching generations G1 through G>5,revealed that TGF-b1 selectively inhibited the increase inthe number density of small vessels, Nv>5 (382 6 85 cm22

for specimens treated with 1 mg TGF-b1/CAM for 24 h,compared to 583 6 99 cm22 for controls), but did not
significantly affect other parameters such as average ves-sel length or vessel diameter. The second and third meth-ods, the fractal dimension (Df) and grid intersection (rv),
0026-2862/00 $35.00Copyright © 2000 by Academic PressAll rights of reproduction in any form reserved.

are statistical descriptors of spatial pattern and density.According to Df and rv, arterial density increased in con-trol specimens from 1.382 6 0.007 and 662 6 52 cm22 at

7 (0 h) to 1.439 6 0.013 and 884 6 55 cm22 at E8 (24 h),compared to 1.379 6 0.039 and 650 6 111 cm22 forspecimens treated with 1 mg TGF-b1/CAM for 24 h.TGF-b1 therefore regulates vascular pattern and the rateof angiogenesis in a unique “fingerprint” manner, as doother major angiogenic regulators that include VEGF,FGF-2 (bFGF), and angiostatin. TGF-b1 did not stimu-late angiogenesis significantly at low cytokine concentra-tions, which suggests that this quail CAM model of an-giogenesis is not associated with an inflammatoryresponse. © 2000 Academic Press

Key Words: TGF-b1; angiogenesis; fractal; chorioallan-toic membrane; CAM; quail; complexity; development.

INTRODUCTION

Transforming growth factor-b (TGF-b) is a pluripo-
tent cytokine that is critical to physiological andpathological angiogenesis, as well as to other aspectsof development and inflammation (Derynck and Feng,
221

K1t

hwh1

e

tc

1997; Heldin et al., 1997; Hoodless and Wrana, 1998;ingsley, 1994; Pepper, 1997; Roberts and Sporn,993). Confusion persists about the contextual regula-ory role of TGF-b, a 25-kDa protein that is expressed

in mammalian species as the three isoforms TGF-b1,TGF-b2, and TGF-b3. The major phenotypes of a tar-geted disruption of the murine TGF-b1 gene, defective

ematopoiesis and vasculogenesis in the yolk sac,ere accompanied by a prenatal lethality of 50% for

omozygous TGF-b1-null embryos (Dickson et al.,995). In vivo, exogeneously added TGF-b, and in par-

ticular TGF-b1, stimulates or inhibits angiogenesis inan assay-dependent manner; in vitro, TGF-b1 stimu-lates or inhibits many cellular activities, including pro-liferation, migration, differentiation, and apoptosis(Madri and Sankar, 1997; Pepper, 1997). In general,TGF-b1 inhibits the proliferation of endothelial cellsover a wide range of concentration (Pepper, 1997), aswas recently reported for bovine retinal endothelialcells (Yan and Sage, 1998). However, the relationshipbetween endogenous TGF-b1 activity and angiogenicinhibition remains unclear (Pepper, 1997).

In a review of quantitative assays of angiogenesis(Jain et al., 1997), it was concluded that no single assaywas convenient with respect to ease of performance,quantification, and cost. However, we recently de-scribed a novel assay of angiogenesis in the quailchorioallantoic membrane (CAM, Parsons et al., 1998)that is convenient with respect to all of these criteria.Angiogenic regulators are applied in solution easilyand uniformly to the entire surface of the CAM. As aquasi-two-dimensional (2D), transparent, highly vas-cularized membrane, the CAM is readily obtainedfrom embryonic culture, imaged by microscopy, andsubsequently analyzed for vascular pattern. Most im-portantly, major regulators of angiogenesis elicit “fin-gerprint” patterns of vascular response: that is, eachregulator induces a unique spatiotemporal pattern ofvascular morphology. Angiogenic regulators distin-guished to date by their specific perturbations of vas-cular pattern include the stimulators vascular endo-thelial growth factor (VEGF; Parsons et al., in progress)and fibroblast growth factor-2 (FGF-2 or bFGF), as

222

well as the inhibitors TGF-b1 and angiostatin (Parsonst al., 1998).

We describe here results that clarify the role of

oll

Copyright © 2000 by Academic PressAll rights of reproduction in any form reserved.

TGF-b1 as a regulator of angiogenesis in vivo whenadded exogenously to the quail CAM. By comparisonof generational analysis of vascular branching andfractal geometry, we demonstrate that TGF-b1 inhib-ited the rate of angiogenesis in the CAM arterial treeprimarily by inhibition of the normal increase in thenumber density of new, small vessels. Relative to con-trol specimens, the rate of angiogenesis decreased byas much as 105% as measured by vessel number den-sity (N v), the fractal dimension (D f), and grid inter-section (rv). Scaling analysis revealed that the normal,ongoing growth of preexisting vessels was not inhib-ited by TGF-b1.

METHODS

Embryonic culture, assay, mounting, imaging, andfractal/grid intersection analysis have been describedpreviously (Parsons et al., 1998).

Culture, Assay, and Mounting

Fertilized eggs of Japanese quail (Coturnix coturnixjaponica, Boyd’s Bird Co., Pullman, WA) were incu-bated at 37°C under ambient atmosphere, cracked atembryonic day 3 (E3; following incubation of eggs for56 h), and cultured further at 37°C in petri dishes(cross-sectional area of well 5 10 cm2). At E7 (follow-ing incubation for an additional 96 h), 0.5 ml pre-warmed ovalbumin (Ov)/PBS solution containing 0–2mg human recombinant TGF-b1 (R&D Systems, 240-B/CF, a generous gift from Marsha Whitney and JohnRanieri, Sulzer Carbomedics) was applied dropwise tothe surface of each CAM. If the diameter of an imma-ture avian epithelial cell is assumed to be 13 mm, anapplication of TGF-b1 at 0.5 mg/CAM (1 mg/ml) rep-resents a stimulus of 50 and 25 fg/cell at E7 and E9,respectively, when calculated for a CAM of knownsurface area en face (Parsons et al., 1998). The calcula-ion further assumes that the CAM consists of fiveellular layers: i.e., two epithelial bilayers comprised

Parsons-Wingerter et al.

f the chorion and allantois (a total of four cellularayers) separated by an intervening mesenchymalayer. In preliminary experiments TGF-b1 was applied

pAI

wTa

KqbtbavttT

at 2 mg/ml (1 mg/CAM) in PBS at 0, 50, 100, and 200mg/ml of Ov (Sigma, albumin chicken egg grade VII).As evaluated by visual inspection, vascular responseto TGF-b1 did not appear to vary as a function of Ovconcentration from 50 to 200 mg/ml; hence all exper-iments reported here were performed at a concentra-tion of 100 mg Ov/ml. TGF-b1 appeared to exert littleor no effect on vascular morphology in the absence ofcarrier protein (Ov). Following treatment with TGF-b1, the embryos were incubated further at 37°C for 24or 48 h, at which time they were fixed in 4% parafor-maldehyde/2% glutaraldehyde/PBS.

Each data point at E8 and E9 represents the meanand standard deviation (SD) of a minimum of quadru-plicate CAMs from at least two independent experi-ments (with the exception of triplicate CAMs fromembryos that were treated with 1 mg TGF-b1/CAMand fixed at E9). Six experiments were performed anda total of 73 CAM specimens were analyzed; a mini-mum of two Ov/PBS-treated control specimens perexperiment was analyzed for each time point. Imagesof additional specimens from the experiments pro-vided qualitative confirmation of the analytical re-sults. P values were calculated by a two-tailed, het-eroscedastic Student’s t test. Surface area of the CAMen face, although virtually circular, was calculated asan ellipse from measurement of the major and minoraxes (Parsons et al., 1998).

Imaging

Aldehyde fixation of the avian CAM results in highcontrast of the arterial tree due to retention of thedark-red erythrocytes (RBC) within the arteries duringdissection, but low contrast of the venous tree due toevacuation of RBC from the veins. Digital images(640 3 480 pixels) of arterial endpoint vessels from themiddle region of the CAM (Parsons et al., 1998) wereacquired in grayscale (0–255 intensity) at a total mag-nification of 103 and resolution of 13 mm/pixel with aNikon Microphot-SA microscope attached to a CCDcamera (DEI-470, Optronics). There was no increase invessel density at a total magnification of 203 in com-

TGF-b1 Inhibits Angiogenesis

arison with 103. The images were transferred withpple Video Player and further processed with NIH

mage software. A square region (480 3 480 pixels)

ctfi

as extracted from the original image, rescaled to aIFF image of 514 3 514 pixels (to support the fractalnalysis), binarized to black/white, and skeletonized.

Analysis by Fractal Dimension and Grid Intersection

The fractal dimension (D f), a statistical descriptor ofspace-filling pattern and density, necessarily variesfrom 1 to 2 for 2D binary fractal images. D f wasestimated for the skeletonized vessels of an imagewith a computer program (Parsons et al., 1998) imple-menting the method of box-counting (Bassingth-waighte et al., 1994; Kirchner et al., 1996), for which aleast-squares regression analysis of the data consis-tently produced a linear fit with a confidence (r 2) $

0.97. Grid intersection (rv), also a statistical measureof space-filling density (Kirchner et al., 1996; Rieder etal., 1995), was measured with a computer program(Parsons et al., 1998), in which the number of intersec-tions between skeletonized vessels and a superim-posed rectangular grid of 32 pixel spacing wascounted. We previously used a grid size of 64 pixels(Parsons et al., 1998), but have found that a grid size of32 pixels improves reproducibility (i.e., decreases vari-ation) among replicate specimens, due to improveddetection of vessel branching.

Generational Analysis of Vascular Branching

In 3D arterial and venous trees of the mammalianheart and lung, the strongest correlate of vessel gen-eration is vessel diameter, not vessel length or branch-ing angle (Gan et al., 1993; Kassab et al., 1994, 1993).

assab and coworkers also found that the most fre-uent branching event is not equivalent or symmetricranching (i.e., the branching of a parent vessel intowo equal daughter vessels) but rather nonequivalentranching (i.e., the branching of a smaller vessel frommuch larger vessel). For our generational analysis ofascular branching, we therefore classified vessels intoheir respective branching generations according tohe proportional decrease in vessel diameter (Fig. 1).he largest arterial tree was extracted from a pro-

223

essed binary image and the single parent vessel (alsohe vessel of largest diameter) was designated as therst generation (G1). If internal cross-sectional area is


a

L

(

stbbnimtt

F

CotGsubccm

conserved, as is usually assumed for branching vesselsdue to conservation of blood flow, the theoretical in-ner diameter of two equivalent daughter vesselsbranching from a parent vessel is 71% of the innerdiameter of the parent vessel. Vessels of G5 or greaterwere treated as a single generation (G$5), because oflimiting resolution for accurate measurement of vesseldiameter.

The parameters of vessel length density (L v), vesselarea density ( A v), and vessel number density (N v)were measured for each branching generation G1

through G$5, with a computer program (VESGEN)written by the first author, and normalized as de-scribed below. Average vessel diameter (D v) was cal-culated as D v 5 A v/L v. We use the symbols L v, A v, N v,nd D v to denote these parameters when they do not

refer to specific generations, whereas L v(all), L v1-4, andv$5, for example, denote L v with respect to specific

generations. VESGEN compares binary and skeleton-ized images of the isolated arterial tree using 8-pixelneighborhood connectivity and the binary morpho-logical operators of the Image Processing Toolbox ofMatlab software (Mathworks, Natick, MA). Only ves-sel segments were measured for G$5; this approxima-tion overestimates N v$5 by approximately 5%. A vesselsegment is defined here as the vessel length connect-ing two adjacent branchpoints. Binary and skeleton-ized images are direct representations of total vesselarea density ( A v(all)) and total vessel length densityL v(all)), respectively, for all branching generations G1

through G$5. Results for L v and A v in this study arepresented as the total (combined) length and area,respectively, for all vessels of designated generation(s)because, as discussed above, the individual values ofthese parameters per vessel vary greatly.

N v, L v, and A v (but not D v) were calculated as den-ity functions by normalization to the surface area ofhe CAM occupied by the extracted arterial tree. Theoundary of the normalizing area was determined byisection of the distance between the arterial tree andeighboring arteries (as shown by the solid black line

n Fig. 1). This function-based normalization (i.e., nor-

224

alization by the CAM area served by the arterialree) resulted in better reproducibility than an alterna-ive geometric normalization in which an irregular

bbs


polygon was fitted to the endpoints of the extractedarterial tree.

Calculation of Normal and Perturbed Rates ofAngiogenesis

We reported previously (Parsons et al., 1998) thatvascular development from E6 to E10 in the quailCAM of untreated specimens is a remarkably linear(i.e., first-order) function of time, according to well-correlated measurements of arterial density by D f andrv. Hence the average rate of angiogenesis (R A) iseasily estimated in control specimens by the time-dependent change in D f or rv. In the present study, R A

was calculated according to the change in D f and rv

IG. 1. Generational branching in the arterial tree of the quailAM. The major arterial tree was extracted from the binary imagef a control CAM specimen fixed at E8. A pseudocolor image of theree illustrates the classification of vessel generation (G1 through

$5) according to the relative decrease in vessel diameter (D v). Thechematic also depicts the area occupied by the arterial tree that wassed to normalize vascular density parameters such as vessel num-er (N v), vessel length (L v), and vascular area ( A v), measured by theomputer program VESGEN. The area of normalization (demar-ated by the black line enclosing the arterial tree) lies approximatelyidway between the arterial endpoints of the tree and the neigh-


oring arteries. In regions where vessels of the arterial tree extendeyond the edge of the image, the area of normalization was definedimply by the edge of the image.

R

Iit

M wasn of Rstanda

from E7 (Dtime 5 Dt 5 0 h) in untreated controlspecimens to E8 (Dt 5 24 h) or E9 (Dt 5 48 h) incontrol specimens treated with Ov/PBS or specimenstreated with TGF-b1 as follows: R A(D f) 5 DD f/Dtand R A(r v) 5 Dr v/Dt. Baseline values for R A in con-trol specimens (R Ac) were R Ac(D f)) 5 0.06/day and

Ac(r v) 5 222 cm2/day (or 124 cm2/day for a grid sizeof 64 pixels). These values of R Ac are essentially equiv-alent to those measured previously for untreated con-trols from E6 to E10 (Parsons et al., 1998).

Relative to control specimens, the percentagechange in R A induced by application of TGF-b1 after24 or 48 h was calculated as follows: Change in R A

(%) 5 {[(R A-R Ac)/R Ac] 3 100}. The advantage of thisrepresentation (i.e., percentage change in R A) is thatthe baseline value for controls is always 0%; therefore,stimulation or inhibition of the rate of angiogenesis isa positive or a negative quantity, respectively. R A-(L v(all)) and percentage change in R A(L v(all)) were calcu-lated for measurements of total vessel length densityL v(all) analogously to those for measurements of D f andrv.

RESULTS

Regulation of the Rate of Angiogenesis by TGF-b1

FIG. 2. TGF-b1 inhibits the rate of angiogenesis in the arterial treein response to TGF-b1 applied at E7 (0 h) from 0.025 to 1.0 mg/CA(A) the fractal dimension (D f) and (B) grid intersection (rv). Inhibitiofunction of increasing cytokine concentration. Error bars represent


As assessed by change in the rate of angiogenesis(R A) from measurements of D f and rv, R A declined by

as much as 105% relative to control specimens inresponse to application of TGF-b1 for 24 h (Fig. 2). (Asmeasured by D f, the probability (P value) that thedecrease in arterial density after 24 h is significantrelative to control specimens is greater than 97 and96% for treatments with 0.5 and 1.0 mg/CAM, respec-tively, and as measured by rv, greater than 96 and98%, respectively.) For concentrations of TGF-b1 $

0.05 mg/CAM, the percentage change in R A is a de-creasing and linear (i.e., first-order) function of thelogarithm of concentration (r 2 5 0.98 for D f and rv).nspection of representative binary and skeletonizedmages (Fig. 3) confirms qualitatively that specimensreated with TGF-b1 displayed decreased vessel den-

sity compared to untreated controls. After applicationof 0.5 and 1 mg TGF-b1/CAM for 48 h (Fig. 4), thedecrease in R A relative to control specimens declinedto approximately 25%. (P values for D f after 48 hrelative to control specimens are greater than 98 and81% for treatments with 0.5 and 1.0 mg/CAM, respec-tively, and for rv, greater than 97 and 82%).

Regulation of Vascular Morphology by TGF-b1

The generational analysis of vascular branching inisolated arterial trees (illustrated in Fig. 1) was appliedto specimens treated for 24 h with 0.5 and 1.0 mg

quail CAM. The percentage change in the rate of angiogenesis (R A)measured in skeletonized images of the arterial tree after 24 h by

A at concentrations greater than 0.05 mg TGF-b1/CAM was a directrd deviations.

225

of the

TGF-b1/CAM relative to control specimens at 0 and24 h. Five vessel generations (G1 through G$5) wereidentified in all images when vessels were classified


As dii pecime

b

tE

according to a generational decrease in D v of approx-imately 70%. Results for parameters of vessel numberdensity (N v), vessel length density (L v), vessel areadensity ( A v), and vessel diameter (D v) suggest thatTGF-b1 decreased the rate of angiogenesis (R A) in the

FIG. 3. Angiogenesis of the arterial tree is inhibited by TGF-b1.nhibited arterial growth at 0.5 and 1.0 mg/CAM relative to control s

magnification).

FIG. 4. Inhibition of the rate of angiogenesis by TGF-b1 is largelyy (A) the fractal dimension (D f) and (B) grid intersection (rv), the p

226

at E7 for 48 h from 0.025 to 1.0 mg/CAM declined in comparison to decreahat in Fig. 2. However, the scales for D f and rv are doubled because the trror bars represent standard deviations.


arterial tree principally by inhibition of the numberdensity of small vessels (N v$5), that is, by inhibition ofthe normal increase in the number of small vessels perunit surface area (Figs. 5 and 6).

The number density of larger vessels (N v1-4) did not

splayed in binary (A–D) and skeletonized images (E–H), TGF-b1ns when applied for 24 h at E7. (Photographed digitally at 103 total

me by 48 h. As measured in skeletonized images of the arterial treeage change in rate of angiogenesis (R A) induced by TGF-b1 applied


overcoercent

ses in R A at 24 h (Fig. 1). The scale for change in R A is identical toime interval for the calculation of R A has doubled from 24 to 48 h.

o

ectiond leng

vary significantly among the four populations (Fig.5A), but N v$5 was considerably less for specimenstreated with 1.0 mg TGF-b1/CAM for 24 h relative tocontrol specimens at both 0 and 24 h (Fig. 5B). Sur-prisingly, no significant increase in N v$5 was detectedfor control specimens from 0 to 24 h. Examination ofcontrol images (Fig. 3) suggests that the normal in-crease in vascular density from E7 (0 h) to E8 (24 h) as

FIG. 5. TGF-b1 inhibits the growth of new, small vessels, but doef vascular branching, (A) the number density of larger arteries (N

significantly in control specimens from E7 (0 h) to E8 (24 h). In specidecreased, relative to controls at 0 and 24 h. Images of untreated conin the area of the CAM and its vascular trees from E7 to E8. This remuch lower rate, in specimens treated with TGF-b1 relative to contthe untreated control specimen at E7 in Fig. 3A was rescaled 23 by aof the CAM from a specimen treated with 1.0 mg TGF-b1/CAM. Inspthe normal increase of arterial density (Figs. 2–4), vessel density an


measured by D f and rv (Fig. 2; see also Parsons et al.,1998) resulted more from increased L v$5 than fromincreased N v$5 (i.e., a large number of small, stubby

vessels are visible in control images at E7 that presum-ably elongate by E8). Therefore inhibition of N v$5 byTGF-b1 could be explained by two alternative hypoth-eses. That is, TGF-b1 may have decreased the densityof small vessels relative to that of control specimens at0 and 24 h either by (1) inhibition of the growth ofnew, small vessels within the expanding CAM, or by(2) the induction of vessel regression. To test these

nduce vessel regression. As measured by the generational analysisd (B) the number density of smaller arteries (N v$5) did not changereated with 1.0 mg TGF-b1/CAM, N v1-4 remained constant but N v$5

ecimens at E7 were rescaled twofold to mimic the twofold increaseanalysis demonstrates that from E7 to E8 N v$5 increased, but at a

cimens. Error bars represent standard deviations. (C) The image of414 by length) and is here compared with (D), the unrescaled imageof the two images suggests that, although TGF-b1 strongly inhibitedth increased somewhat following application of the cytokine.

227

s not i

v1-4) anmens ttrol spscaling

rol sperea (1.

hypotheses, we performed a rescaling analysis basedon normal developmental growth within the CAM,described as follows.


T1

Baattttpoatp1abn

T id incrl the lenC

During the developmental window of this assay (E6to E10), the quail CAM expands most rapidly from E7to E8 by a doubling of the CAM surface area in un-treated specimens from 4 to 8 cm2 (Parsons et al., 1998).

he area of the CAM at E8 in specimens treated with.0 mg TGF-b1/CAM for 24 h was 7.6 6 0.3 cm2,

compared to 7.9 6 0.3 cm2 in control specimens at E8.y measurements of the spatial growth of the entirerterial tree and the number of branching generations,s well as by visual inspection, we determined thathis doubling is an isotropic (i.e., spatially uniform),wofold (23) rescaling of the CAM and its vascula-ure. The area of the CAM is a second-order parame-er, directly proportional to the square of first-orderarameter(s) of characteristic length such as the axesf the elliptical CAM. Consequently both the lengthnd diameter of vessels of all generations (and hencehe total length of the major arterial tree) that areresent at E7 increase by a factor of approximately.414 (i.e., the square root of 2). There is a smalldditional contribution to the total length of the treey the growth of small, new vessels: that is, the total

FIG. 6. Only preexisting vessels elongate in the presence of TGF-barteries (L v1-4), increased in untreated control specimens from E7remained constant relative to controls at 0 and 24 h, and L v$5 did n

GF-b1/CAM, L v(all) did not increase relative to controls at 0 h, but dength density is only slightly larger than that required to maintainAM. Error bars represent standard deviations.

228

umber of branching generations (G(all)) throughoutthe entire tree increases from 7 or 8 at E7 to 8 or 9 at E8.

The area of the CAM imaged in this study (i.e., the


terminal branches at the center of the CAM originatingfrom the major arterial tree) was 0.41 cm2. Thus theimages, all acquired at 103 total magnification, repre-sent an absolute, fixed sampling area (Eulerian sam-pling) of vascular parameters at E7 and E8. We wishedto study the rescaling of the vasculature from E7 to E8by Lagrangian sampling, in which the same region ofthe test specimen is followed through time. However,we could not photograph or videotape the actualgrowth of the CAM from E7 to E8 because the vascu-lature of the yolk sac lies directly beneath the vascu-lature of the transparent CAM. Instead, for a quasi-Lagrangian sampling, we extracted half of the totalarea (specifically, the region at which the arteries ter-minated) from the images of untreated control speci-mens at E7 and rescaled this area twofold to mimic thegrowth of the CAM (accompanied by an expansion inlength of the image by a factor of 1.414; see Fig. 5C).The contribution of the preexisting vessels to theirsubsequent isotropic expansion was then measured.The rescaling expansion assumes that the tissue ex-pands by the growth of preexisting cells and/or by the

The total length density of smaller arteries (L v$5), but not of largerE8 (24 h). In specimens treated with 1.0 mg TGF-b1/CAM, L v1-4

ase relative to controls at 0 h. (B) In specimens treated with 1.0 mgease relative to rescaled controls at 0 h. This increase in total vesselgth of preexisting vessels within the expanding arterial tree of the


1. (A)(0 h) toot incre

uniform addition of new cells throughout the tissue,resulting in a uniformly distributed increase of thetissue. N v and L v were measured in the rescaled major

i

T

Easic

rLtc

(

arterial trees as in the original images. If the morphol-ogy of the vascular tree is indeed isotropic, the theo-retical decrease in density parameters N v and L v of therescaled images relative to the original images is 1/2(50%) and 1/1.414 (71%), respectively.

The number density of smaller vessels (N v$5) in therescaled images was measured as 241 6 31 cm22, adecrease of 54% relative to that of 529 6 53 cm22

measured in original (i. e., unrescaled) images of un-treated control specimens at E7 (Fig. 5B). N v$5 in spec-mens treated for 24 h with 1.0 mg TGF-b1/CAM was

382 6 85 cm22, compared to 583 6 99 cm22 for controlspecimens at E8 after 24 h. Relative to the rescaledvalue at E7 (0 h), N v$5 in specimens treated with 1.0 mg

GF-b1/CAM therefore increased by only 40% asmuch as the increase in N v$5 in control specimens at E8after 24 h. These results for N v$5 indicate that TGF-b1inhibited angiogenesis primarily by a decrease in thegrowth of new vessels, and not by the induction ofvessel regression. It is noteworthy that in control spec-imens the increase in N v(1-4) and N v$5 from E7 (0 h) to

8 (24 h) is modest (Figs. 5A and 5B), despite vigorousrterial growth during this time, as evidenced by vi-ual inspection (Figs. 3A, 3B, 3E, and 3F) and thencrease in D f and rv from 1.382 6 0.007 and 662 6 52m22 to 1.439 6 0.013 and 884 6 55 cm22, respectively

(Fig. 2).Total vessel length density for larger vessels (L v1-4)

did not vary significantly among the four populations(Fig. 6A). However, in contrast to N v$5 (Fig. 5B), L v$5

increased significantly in unrescaled control speci-mens from E7 (0 h) to E8 (24 h; Fig. 6A). The increasein L v$5 measured in unrescaled control specimens wasstrongly inhibited in specimens treated with 0.5 and1.0 mg TGF-b1/CAM (Fig. 6A). Proportional differ-ences between control specimens and specimenstreated with TGF-b1 at 0.5 and 1.0 mg/CAM werelargest for L v$5 (Fig. 6A; P values are 96 and 99%,espectively), but absolute differences were largest forv(all) (Fig. 6B; P values are 99% for both concentra-

ions). When calculated as for D f and rv, percentagehange in R A(L v(all)) in response to application of


TGF-b1 at 0.5 and 1.0 mg/CAM relative to unrescaledcontrol specimens was 70 and 84%, respectively. Thesevalues of percentage change in R A(L v(all)) are less than,

but of similar magnitude to, those of the percentagechange in R A(D f) and R A(rv) (Fig. 2).

L v(all) in the rescaled images of untreated controlspecimens (Fig. 6B) at E7 was 12.8 6 1.1 cm21 relativeto 17.0 6 0.8 cm21 in the original images, a rescalingvalue of 75% (compared to the theoretical value of71%). L v(all) in the images of control specimens at E8was 22.8 6 0.4 cm21. These results imply that, duringexpansion of the arterial tree from E7 to E8, approxi-mately 50% of the linear increase of the arterial treemeasured by L v(all) was required to support the scaledexpansion of preexisting vessels (i.e., the rescalingrequirement for L v(all) from E7 to E8 for preexistingvessels 5 (17.0 2 12.8) cm21 5 4.2 cm21, which isapproximately 50% of the increase of L v(all) from 12.8cm21 (the rescaled contribution) to 22.8 cm21 at E8).The remaining 50% of increase in L v(all) in control spec-imens from E7 to E8 was invested in a net increase ofarterial density that was also measured by D f and rv

(Fig. 2).In specimens treated with 1.0 mg TGF-b1/CAM,

L v(all) increased by only 14% above that required torescale preexisting vessels from E7 to E8 (Fig. 6B).Thus the degree of inhibition of total vessel lengthdensity, L v(all), by TGF-b1 correlates positively with theinhibition of N v$5 (Fig. 5B). Furthermore, the averagelength per vessel for G v$5 in specimens treated at E7with TGF-b1 at 0.5 and 1.0 mg/CAM for 24 h was300 6 48 mm and 306 6 38 mm, respectively, whichrepresents increases of only 13 and 15% above thatmeasured in control specimens of 273 6 56 mm at 24 hcompared to 210 6 35 mm at E7 or 0 h, Fig. 6B). Thus

inhibition of N v$5 (Fig. 5B), not inhibition of averagevessel length, is the primary morphological mecha-nism by which TGF-b1 inhibited angiogenesis in thequail CAM. The positively correlated results for N v,L v, and average vessel length in the rescaling analysissupport the hypothesis that TGF-b1 decreased N v$5

relative to the unrescaled control specimens at 0 h (E7)by inhibition of the normal increase in N v of new,small vessels, and do not support the alternative hy-pothesis that TGF-b1 decreased N v$5 by induction ofvessel regression.

229

The diameters of larger vessels (D v1-4) increased inspecimens treated with TGF-b1 by approximately 20%compared to control specimens, but the standard de-


1

ns

t

i

viations of these increases invariably overlapped (datanot shown). For example, relative to controls at 24 h,D v2 in specimens treated with TGF-b1 at 1 mg/CAMincreased from 182 6 26 to 234 6 39 mm and D v3, from43 6 13 to 163 6 20 mm, respectively. The total

number of pixels in the binary vascular trees, whichdirectly represents total vascular area ( A v(all)) in the 2Dplane of the CAM en face for G1 through G$5, also did

ot vary greatly among the four populations (data nothown). A v(all) was measured as 0.129 6 0.026 cm2 per

cm2 of CAM area in specimens treated with TGF-b1 at1 mg/CAM for 24 h, relative to 0.156 6 0.011 cm2 percm2 of CAM area in control specimens at 24 h.

DISCUSSION

We recently proposed that the measurement of abasal or normal rate of angiogenesis is a useful tool forthe quantitative description of angiogenic perturba-tion by both positive and negative regulators (Parsonset al., 1998). We now propose that rate-based mor-phometry can provide insight into regulatory mecha-nisms of angiogenesis by the analysis of perturbedvascular patterns induced by an angiogenic regulator.These vascular patterns are dynamically (i.e., spatio-temporally) distinct.

The correlation of direct measurements of vascularmorphology (L v, N v, A v, and D v) with fractal geome-ry (D f and rv) demonstrates that TGF-b1 decreased

the rate of angiogenesis in the quail CAM by inhibi-tion of the normal increase in numbers of new vessels.In turn, the inhibition or delay of increase in newvessels resulted in the inhibition of total (i.e., com-bined) vessel elongation, but not individual (i.e., av-erage), vessel elongation. The potent inhibition or de-lay of angiogenesis by TGF-b1 after 24 h was largelyovercome by 48 h. This early but transient perturba-tion of angiogenesis by TGF-b1 contrasts spatiotem-porally with the inhibition of angiogenesis by an-giostatin (Parsons et al., 1998), which significantly

230

decreased angiogenesis in the quail CAM after 48 hbut not after 24 h. Moreover, the application of an-giostatin resulted in a disorganized arterial morphol-


ogy of highly irregular spatial distribution, comparedto the inhibited but normal morphology of specimenstreated with TGF-b1. Striking differences in the re-sponse of regulatory kinetics and vascular morphol-ogy to these two angiogenic inhibitors suggest thatinhibition by TGF-b1 is mediated by different signal-ing mechanisms than is inhibition by angiostatin.

The results for A v(all) (the measure of vessel densityin a binary image) and for L v(all) (the measure of totalvessel length in a skeletonized image) agree with ourprevious observation in untreated CAMs that from E6to E10 the fractal dimension (D f) of binary imagesincreased less than D f of skeletonized images (Parsonset al., 1998). The stronger dependence of D f on increas-ng L v(all) than on increasing A v(all) further confirms that

increasing vascular penetrance of the host tissue dur-ing normal CAM development from E7 to E8 dependsprincipally on increasing vessel length density (L v),and to a lesser extent, on increasing vessel diameter(D v), vascular area density ( A v), and vessel numberdensity (N v, which is almost constant from E7 to E8).However, the primary mechanism of inhibition of vas-cular growth in the CAM by TGF-b1 from E7 to E8 isthe inhibition of N v, that is, the growth of new, smallvessels from preexisting vessels. A twofold increase invessel number density is required to maintain N v at aconstant value during the twofold expansion of theCAM area from E7 to E8. The cytokine does not inter-fere with the normal extension of preexisting vessels(i.e., increase in L v) within the expanding CAM.

We did not measure strong stimulation of angiogen-esis at lower concentrations of TGF-b1 (Figs. 2 and 4),as has been reported for other models of angiogenesisin vivo (Madri and Sankar, 1997; Pepper, 1997). Nota-bly, however, the slight increase in arterial density (asmeasured by D f and rv) in response to application of0.05 mg TGF-b1/CAM after 24 h (Fig. 2), as well as thelarge variability of response, was repeated indepen-dently in measurements of arterial density at this sameconcentration after 48 h (Fig. 4). Although standarddeviations for control specimens and specimenstreated at 0.05 mg TGF-b1/CAM overlap in Figs. 2 and


4, mean arterial density after 24 h in specimens treatedwith 0.05 mg TGF-b1/CAM is distinguished from con-trol specimens with a P value of $91% for D f and of

tt

thsr

a

0t

$99% for rv. Visual inspection of images from addi-ional unanalyzed specimens treated at lower concen-rations of the cytokine suggests that TGF-b1 may

have both stimulated and inhibited the rate of angio-genesis in a bimodal manner.

The fact that lower concentrations of TGF-b1 did notstrongly stimulate angiogenesis in the quail CAM in-dicates that the exogenous perturbation of angiogen-esis in this system is a noninflammatory model ofangiogenesis (Madri and Sankar, 1997; Pepper, 1997;Roberts et al., 1986). Stimulation (but not inhibition) ofangiogenesis in vivo by exogenous addition of TGF-b1is associated with inflammation, wound-healing andthe presence of infiltrating monocytes. Endothelial cellinvasion and capillary morphogenesis stimulated byVEGF and FGF-2 in vitro were potentiated by TGF-b1at low, but not at high concentrations (Pepper et al.,1993).

One hallmark of a complex, adaptive biological sys-tem such as the angiogenic CAM, as well as othercomplex physical systems (Radhakrishnan, 1991), issensitivity to some types of initial conditions: that is,small variation in initial parameters can result inhighly divergent, irreversible outcomes (Coffey, 1998;Science, 1999). Thus the exogenous application ofsmall amounts of TGF-b1 may amplify the normallysmall variation in endogenous signaling within indi-vidual CAMs into mutually exclusive programs ofaccelerated (stimulated) versus decelerated (inhibited)rates of angiogenesis, as is suggested by the largevariation in arterial density in response to low concen-trations of TGF-b1 measured at E8 and E9.

On the other hand, another characteristic of adap-ive complex systems that is more fundamental toomeostatic physiology is the ability to dampen atrong input signal, as evidenced in our results byecovery at E9 (48 h) from inhibition by TGF-b1, com-

pared to E8 (24 h). When R A is plotted as a function oftime (Fig. 7), we see that inhibitory concentrations ofTGF-b1 (i.e., 0.5 and 1.0 mg/CAM) resulted first in thedeceleration, and subsequently in the acceleration, ofR A (compared to no acceleration in control specimens


due to the constant value of R A). Since the primaryfunction of the CAM is the exchange of oxygen andcarbon dioxide through the eggshell, the acceleration

of R A from E8 to E9 in specimens treated with TGF-b1at E7 perhaps reflects an adaptive response caused byhypoxia that resulted from the relatively small capac-ity of the vascular bed at E8.

Major angiogenic regulators that have elicited spe-cific patterns of perturbed vascular morphology in thequail CAM include the stimulators VEGF (Parsons etal., in progress) and FGF-2 (Parsons et al., 1998), andthe inhibitors angiostatin and TGF-b1. With the quailCAM model, dominant regulatory mechanisms can beinferred from spatiotemporal alterations in vascularpattern. Thereafter these inferences can be studied inother models of angiogenesis that test specifically theproposed mechanisms. The CAM model of angiogen-esis thereby supports the study of emergent biologicalproperties (Baish and Jain, 1998; Coffey, 1998): that is,the definition of fundamentally important, nonlinear

FIG. 7. The normal rate of angiogenesis first decelerates, but lateraccelerates, in response to TGF-b1. The rate of angiogenesis (R A)calculated by the change in the fractal dimension (R A 5 DD f/Dt)for control specimens is a constant function of time. Its acceleration(a) is therefore zero (a 5 DR A/Dt 5 0 h22). However, in responseto the application of 0.5 and 1.0 mg TGF-b1/CAM, R A first deceler-ted from 0 h to 24 h (a 5 20.052/24 h2 and 20.059/24 h2, respec-

tively) and then accelerated from 24 to 48 h (a 5 0.135/24 h2 and.146/24 h2, respectively). The second-order polynomials fitted tohe temporal rate of change in R A in response to TGF-b1 are hypo-

thetical (it seems unlikely that these changes would be linear). Theacceleration of R A from 24 to 48 h in specimens treated with TGF-b1may have been driven by hypoxia resulting from the low vasculardensity measured at E8 (Fig. 1).

231

properties of highly cooperative systems such as an-giogenic vasculature that cannot be revealed by reduc-tionist studies alone.


B

C

D

D

G

P

R

R

R

R

S

Y

ACKNOWLEDGMENTS

We thank Marsha Whitney and John Ranieri, Sulzer Carbomedics,Austin, TX, for their generous gift of TGF-b1. We are grateful toundergraduate students Adam Chu, Van Thuy Le, Piotr Ochenkowski,and Minh Nguyen for image processing/analysis, and to ElaineRaines, Anita Hendrickson, Alanna Ruddell, and Amy Bradshaw forhelpful discussion and critical reading of the manuscript. This researchwas supported by Sulzer Carbomedics, by National Science Founda-tion Grant EEC-9529161 to the University of Washington EngineeredBiomaterials (UWEB), by National Institutes of Health Grants EY 04542and GM 40711, and by a Venture Fund award from the NationalAeronautics and Space Administration Glenn Research Center.

REFERENCES

Baish, J. W., and Jain, R. K. (1998). Cancer, angiogenesis and fractals[letter]. Nat. Med. 4, 984.

assingthwaighte, J. B., Liebovitch, L. S., and West, B. J. (1994).Fractal Physiology. Oxford Univ. Press, New York.

offey, D. S. (1998). Self-organization, complexity and chaos: Thenew biology for medicine. Nat. Med. 4, 882–885.

erynck, R., and Feng, X. H. (1997). TGF-beta receptor signaling.Biochim. Biophys. Acta. 1333, F105–F150.

ickson, M. C., Martin, J. S., Cousins, F. M., Kulkarni, A. B., Karls-son, S., and Akhurst, R. J. (1995). Defective haematopoiesis andvasculogenesis in transforming growth factor-beta1 knock outmice. Development 121, 1845–1854.

an, R. Z., Tian, Y., Yen, R. T., and Kassab, G. S. (1993). Morphometryof the dog pulmonary venous tree. J. Appl. Physiol. 75, 432–440.

Heldin, C. H., Miyazono, K., and ten Dijke, P. (1997). TGF-betasignaling from cell membrane to nucleus through SMAD pro-teins. Nature 390, 465–471.

Hoodless, P. A., and Wrana, J. L. (1998). Mechanism and function ofsignaling by the TGF beta superfamily. Curr. Top. Microbiol. Im-munol. 228, 235–272.

Jain, R. K., Schlenger, K., Hockel, M., and Yuan, F. (1997). Quanti-tative angiogenesis assays: Progress and problems. Nat. Med. 3,1203–1208.

Kassab, G. S., Lin, D. H., and Fung, Y. C. (1994). Morphometry of pigcoronary venous system. Am. J. Physiol. 267, H2100–H2113.

232

Kassab, G. S., Rider, C. A., Tang, N. J., and Fung, Y. C. (1993).Morphometry of pig coronary arterial trees. Am. J. Physiol. 265,H350–H365.


Kingsley, D. M. (1994). The TGF-beta superfamily: New members,new receptors, and new genetic tests of function in differentorganisms. Genes Dev. 8, 133–146.

Kirchner, L. M., Schmidt, S. P., and Gruber, B. S. (1996). Quantitationof angiogenesis in the chick chorioallantoic membrane modelusing fractal analysis. Microvasc. Res. 51, 2–14.

Madri, J. A., and Sankar, S. (1997). The biphasic effects of transform-ing growth factors beta in angiogenesis. In “Tumor Angiogene-sis” (R. Bicknell, C. E. Lewis, and N. Ferrara, Eds.), pp. 239–249.Oxford Univ. Press, New York.

Parsons-Wingerter, P., Le, V. T., Albarran, B., and Farr, A. G., Vesseldensity and vessel diameter increase as independent responses tostimulation by VEGF165. In progress.

Parsons-Wingerter, P., Lwai, B., Yang, M. C., Elliott, K. E., Mila-ninia, A., Redlitz, A., Clark, J. I., and Sage, E. H. (1998). A novelassay of angiogenesis in the quail chorioallantoic membrane:Stimulation by bFGF and inhibition by angiostatin accordingto fractal dimension and grid intersection. Microvasc. Res. 55,201–214.

Pepper, M. S. (1997). Transforming growth factor-beta: Vasculogen-esis, angiogenesis, and vessel wall integrity. Cytokine Growth Fac-tor Rev. 8, 21–43.

epper, M. S., Vassalli, J. D., Orci, L., and Montesano, R. (1993).Biphasic effect of transforming growth factor-beta 1 on in vitroangiogenesis. Exp. Cell Res. 204, 356–363.

adhakrishnan, K. (1991). Combustion Kinetics and SensitivityAnalysis Computations. In “Numerical Approaches to Combus-tion” (E. S. Oran and J. P. Boris, Eds.), pp. 83–128. AIAA, Wash-ington, D.C.

ieder, M. J., O’Drobinak, D. M., and Greene, A. S. (1995). Acomputerized method for determination of microvascular den-sity. Microvasc. Res. 49, 180–189.

oberts, A. B., and Sporn, M. B. (1993). Physiological actions andclinical applications of transforming growth factor-beta (TGF-beta). Growth Factors 8, 1–9.

oberts, A. B., Sporn, M. B., Assoian, R. K., Smith, J. M., Roche, N. S.,Wakefield, L. M., Heine, U. I., Liotta, L. A., Falanga, V., Kehrl,J. H., and Fauci, A. S. (1986). Transforming growth factor typebeta: Rapid induction of fibrosis and angiogenesis in vivo andstimulation of collagen formation in vitro. Proc. Natl. Acad. Sci.USA 83, 4167–4171.

cience. (1999). Complex Systems (Special Section). Science 284, 79–109.

an, Q., and Sage, E. H. (1998). Transforming growth factor-beta1


induces apoptotic cell death in cultured retinal endothelial cellsbut not pericytes: Association with decreased expression ofp21waf1/cip1. J. Cell Biochem. 70, 70–83.