A ROLE FOR FIBRONECTIN IN CELL SORTINGdicular to the interface between the two members of the pair....

y. Cell Sd. 69, 179-197 (1984) 179Printed in Great Britain © The Company of Biologists Limited 1984

A ROLE FOR FIBRONECTIN IN CELL SORTING

PETER B. ARMSTRONG* AND MARGARET T. ARMSTRONGDepartment of Zoology, University of California, Davis, California 95616, U.SA.

SUMMARY

A useful approach to the investigation of embryonic morphogenesis is the study of the factors thatcontrol cell movement in cell aggregates in organ culture. Previous studies, in which aggregates ofembryonic chick heart ventricle tissue were paired in organ culture, supported the hypothesis thatthe associative behaviour is dominated by the mesenchymal cell (at the stages used the ventricle iscomposed of approximately 25 % mesenchyme (Mes) and 75 % myocyte tissue (My)) by virtue ofthis cell's ability to establish a pericellular matrix rich in fibronectin. In aggregate pairs, theaggregate types that develop a fibronectin-rich matrix rapidly are spread over by the aggregate typesthat are less able to deposit fibronectin in the matrix. In sorting conditions, Mes sorts to the surfaceof My. This is explained as a consequence of a requirement that Mes have access to a componentin the serum fraction of the culture medium for deposition of fibronectin in the matrix. It is proposedthat the factor penetrates to a shallow depth in aggregates, limiting the establishment of afibronectin-rich matrix to superficially located Mes. As fibronectin appears in the matrix, Mesbecomes more cohesive than My, allowing it to exclude myocytes and establish itself as a pure tissuethat increases in volume as mesenchyme cells migrating within the interior contact the surface zone,becoming immobilized and also activated to secrete fibronectin. The analysis presented includes anexperimental investigation of the different elements of this hypothesis and also explores some of thepredictions of the hypothesis.

INTRODUCTION

At the morphological level, embryonic and larval development is characterized bythe establishment of increasingly complex structural interrelations of tissues andorgans until the final form of the adult is realized. One of the major tasks facingdevelopmental biologists is to provide an accounting at cellular and molecular levelsof the factors that establish during development, and that stabilize in the adult, theorderly structure of the tissues and organs.

It has long been presumed that adhesive interactions of cell with cell, and cell withextracellular matrix, play important roles in regulating embryonic morphogenesis andin stabilizing final structure (Armstrong, 1977, 1984; Bellairs, Curtis & Dunn, 1982;Letourneau, Ray & Bernfield, 1981; Trinkaus, 1976, 1984). One way to investigatethis supposition is to study the consequences, on the morphogenetic performance ofembryonic cells, of alterations of the character and/or presentation of putative ad-hesive molecules at the cell surface or in the extracellular matrix. One way to conductthis kind of experimental analysis is to study the morphogenetic performance ofaggregates of cells maintained in organ culture. When two or more tissues are com-bined as attached aggregates, the cells of one tissue spread over the surface of thesecond in a fashion reminiscent of spreading morphogenetic movements in vivo

•Author for correspondence.

180 P. B. Armstrong and M. T. Armstrong

(for a review see Armstrong, 1982). When the cells of two or more tissues are coculturedas randomized cell reaggregates, the cells sort out into homogeneous tissue domainsthat are organized with respect to each other in a reproducible pattern dependent onthe identity of the tissues (Armstrong, 1970, 1971; Holtfreter, 1944; Moscona, 1957;Steinberg, 1963, 1970; Townes & Holtfreter, 1955). With these systems, one cancontrol the cell types present and their initial spatial relationship and, in favourablecases, the presence or absence of specific macromolecules in the extracellular matrix.By determining the consequences on cell sorting and tissue spreading of such altera-tions, it is possible to examine experimentally the roles of various factors in morpho-genesis and tissue stability.

A recent study of chick embryonic heart ventricle has demonstrated that the spread-ing behaviour in culture is, at the cellular level, dependent on activities of the cardiacmesenchyme (Mes)*, and, at the molecular level, on the ability of Mes, but notmyocyte tissue (My), to elaborate an extracellular matrix rich in the matrix protein,fibronectin (Armstrong, 1978, 1980, 1982; Armstrong & Armstrong, 1981). Inaggregate pairs, either the member with the higher content of Mes or the one thatcontains higher quantities of fibronectin in the extracellular matrix is reproducibly theaggregate that becomes enveloped. The spreading of one aggregate over its partnerinvolves extensive deformation of the tissue that assumes the superficial position. Itappears that the relative cohesiveness of the two tissues in an aggregate pair deter-mines spreading behaviour, with the less-cohesive member (e.g. the aggregate typeless resistant to the deformation of tissue structure required for an initially sphericalaggregate to spread out into a tissue sheet) enveloping the more cohesive member(Steinberg, 1963, 1970, 1978a,6). The observation that increased fibronectin levelscorrelates with an internal position following spreading is consistent with thisanalysis, since fibronectin has been identified as being capable of promoting adhesionof cells to artificial surfaces (AH, Mautner, Langa & Hynes, 1977; Hook et al. 1977;Rovasio et al. 1983; Yamada, Yamada & Pastan, 1976; Zetter, Martin, Birdwell &Gospodarowicz, 1978), to structural elements of the extracellular matrix (e.g.collagen) (Chiquet, Eppenberger & Turner, 1981; Erickson & Turley, 1983; Gold &Pearlstein, 1980; Grinnell, 1978; Grinnell &Minter, 1978;Klebe, 1974; Kleinmann,Klebe&Martin, 1981; Pearlstein, 1976), and possibly to other cells (Chen, Gallimore& McDougall, 1976; Hedman, Vaheri & Wartiovaara, 1978; Saison, Van Leuven,Cassiman & Van Den Berghe, 1983; Yamada, Yamada & Pastan, 1975).

The present study represents an extension of this analysis to include the sorting-outbehaviour of heart tissue. We observed that Mes sorted out to the surface of mixedmesenchyme-myocyte aggregates (Mes+My). The analysis of this phenomenon isdeveloped around the following hypothesis:

(1) Cultured Mes, but not My, can deposit fibronectin in the extracellular matrix.Previously reported immunocytochemical studies have demonstrated this: My

•The two principal cell types in embryonic heart ventricle are the myocyte (approximately70-80% of the cells) and the mesenchymal cell (approximately 20-30%) (DeHaan, 1967).Mesothelium, endothelium and nerve are present as minor cell types.

Fibronectin and cell sorting 181

aggregates contain very little fibronectin in comparison to Mes aggregates (Armstrong& Armstrong, 1981).

(2) The freshly trypsinized mesenchymal cells included in the initial Mes+Myaggregate possess little surface fibronectin, presumably as a result of exposure of thecells to trypsin (fibronectin is notably proteinase-sensitive: Blumberg & Robbins,1975; Hynes, 1973; Yamada & Weston, 1974, 1975).

(3) The establishment of a fibronectin-rich matrix renders Mes more cohesivethan My.

(4) Deposition of fibronectin in the matrix surrounding cultured Mes is dependenton exposure to a factor in the serum fraction of the culture medium.

(5) This serum factor penetrates the aggregate to a shallow depth only.Under the precepts of this hypothesis, sorting out of the mixed aggregate is en-

visaged as involving deposition of fibronectin by surface-located Mes in response tothe serum factor and a progressive entrapment of Mes cells at the surface as Mes cellsmigrating in the fibronectin-poor interior encounter this surface layer and becomeentrapped. Sorting, as envisaged by this explanation, has a positive feedback charac-ter, since it is expected that as these mesenchymal cells accumulate in the surfacelayers, they too become activated to secrete fibronectin. The progressive accumula-tion of Mes at the surface segregates the My tissue to the interior. In the analysis thatfollows, each part of this hypothesis is evaluated experimentally. In addition, theresults of an investigation of some of the predictions of the hypothesis are reported.

MATERIALS AND METHODS

Cell cultureIn all cases, the culture medium was Dulbecco-modified Eagles medium (Gibco) with various

additions, the gas phase was 95 % air/5 % CO2 and the culture temperature was 37 °C. When serumwas included, it was depleted of plasma fibronectin by passage over a gelatin-Sepharose 4B affinitycolumn (Engvall & Ruoslahti, 1977). Ten-day chick embryo heart ventricle was disaggregated in1 mg/ml trypsin (Difco 1: 250) and freed of a majority of cardiac mesenchyme (Mes) by culture ofthe resulting cell suspension in tissue culture-grade plastic Petri dishes (Falcon Plastics) for 1-5 h.During this period the fibroblasts attach to the dish whereas the myocytes (My) do not and the Mycan be removed when the medium is decanted (Armstrong & Armstrong, 1978, 1979). Mes wasprepared by reincubating the myocyte-free cultures until the cells grew to confluency. The cells wereremoved from the dish by incubation in 1 mg/ml trypsin for lOmin (37°C). When required, thecells were labelled with [3H]thymidine by including this in the culture dish (1 /iCi/ml, 2Ci/mmol)during growth of the monolayer to confluency. In situations in which the cells were to be culturedfurther in the absence of serum, trypsinization was followed by exposure of the cell suspension tosoybean trypsin inhibitor at 1 mg/ml (15min, room temperature).

Aggregates were prepared by centrifuging cell suspensions in 16-ml screw-cap test tubes followedby stationary incubation at 37 °C for 1-4 h. The coherent pellets were chopped into smaller pieces,which were cultured further in shaker flask culture. When cells labelled with [3H]thymidine werepresent, unlabelled thymidine (100/JM) was included in the culture medium at all steps in theprocedure, to reduce transfer of labelled thymidine from labelled cells to unlabelled cells.

Cell sorting and aggregate fusionRandom reaggregates were prepared by centrifuging into pellets suspensions of cells containing

the two cell types of interest at the desired ratio. Usually, one cell type was prelabelled with[3H]thymidine. Centrifugation was performed from a small volume (approximately 100-200 /A) to


prevent stratification of cells in the pellet as a function of specific gravity. After 1—4 h of incubationat 37 °C, the pellets were chopped into smaller pieces and the cells were allowed to sort out in shaker-flask culture. The pairing of aggregates for aggregate fusion studies was achieved by placing oneaggregate of each of the two types to be studied in a 20/d drop of culture medium standing on thebottom of a 60 mm bacteriologic plastic Petri dish and then inverting the dish so that the standingdrops of culture medium were converted to hanging drops and the two aggregates in each drop werebrought into contact at the bottom of the drop. Usually, 10-20 aggregate pairs were produced ineach dish and a given trial might involve 10—100 pairs. The dishes were incubated in a humidifiedCO2 incubator at 37 °C until the pairs of aggregates were firmly adherent to each other (2—4 h), afterwhich the aggregate pairs were cultured in shaker flasks.

Histology and radioautographyAggregates were washed thoroughly in several changes of phosphate-buffered saline (PBS) and

fixed for 0-5 h in freshly prepared 4% (w/v) paraformaldehyde in PBS. The aggregates were thenwashed in PBS, exposed for lOmin to PBS modified to contain 1 M-NaCl, 0-1 M-glycine and 0-5 M-histamine (G-H), then stored in PBS+1 mM-NaN3 at 4°C. Aggregates were embedded in paraffin(exposure to melted paraffin was 40min) and sectioned at 6/mi. In fragment-fusion studies, theaggregate pairs were oriented in the paraffin block so that the plane of section would pass perpen-dicular to the interface between the two members of the pair. Radioautography was accomplishedby coating deparaffinated sections with a 1:1 dilution of NTB-2 liquid emulsion (Kodak) followedby incubation for 3 weeks at 4°C in a light-tight box containing desiccant. Autoradiograms weredeveloped in a 1: 3 dilution of Dektol (2min, 17°C) and were stained through the emulsion withHarris' haematoxylin and eosin Y (Kopriwa & Leblond, 1962).

ImmunocytochemistryParaformaldehyde-fixed specimens destined for immunofluorescence cytochemistry were ex-

posed to G-H for 15 min at room temperature, washed, and incubated in sequence (with wash stepsbetween each reagent) with a 1: 5 dilution of normal rabbit serum, a 1: 50 dilution of affinity-purifiedgoat anti-chick embryo cellular fibronectin antiserum (Yamada, 1978), and a 1:16 dilution offluorescein isothyocyanate (FITC)-labelled rabbit anti-goat immunoglobulin G (IgG) antiserum(Miles). Control preparations were prepared similarly but with the substitution of preimmune goatserum for the goat anti-fibronectin antiserum. Coverslips were mounted with PBS/glycerol (1:9)containing 1 mM-phenylenediamine (Johnson & Nogueira Araujo, 1981) to reduce photobleachingof the fluorochrome. Slides were viewed with a Zeiss WL microscope equipped with a Zeissepifluorescence condenser, a SOW high-pressure mercury lamp and a fluorescein filter set.Photomicroscopy was performed on Ektachrome 400 film. Standardization was achieved by controlof section thickness and staining protocol, restriction of the time of exposure of the preparation tothe illuminating beam before photography to less than S s and the choice of constant exposure times(120s with a Plan 16/0-32 objective and 30s with a Planapo 40/1-0 objective). Indirect immuno-peroxidase staining followed the outline described above but with modifications dictated by the useof the Vector ABC biotin-avidin kit. When staining tissue sections, it was found necessary to controlpH during the incubation with the ABC reagent (biotin-avidin-horseradish peroxidase complex)by buffering with 0-15M-HEPES. Living specimens (monolayer cultures or cell aggregates) wereexposed to the primary antiserum when it was necessary to stain only extracellular fibronectin. Inthese situations exposure to primary antiserum was for 30 min at 4°C. The specimens were thenwashed, fixed and processed as described above. When aggregates were used exposure to the secondantibody was conducted on tissue sections.

RESULTS

In order to distinguish Mes from My in mixed aggregates, the Mes was labelled with[3H]thymidine before combination with My and was subsequently identified in tissuesections by radioautography. Reaggregates containing Mes and My at ratios of 20: 80,25: 75 and 50: 50 showed sorting of the Mes to the surface (Fig. 1). The disposition


Figs 1-2 for legend see p. 185.


Figs 3-4


of cells in the initial reaggregate showed an apparently random intermingling of thetwo cell types (Fig. 2).

Stimulation by a serum factor of extracellular fibronectin deposition

As required by the hypothesis described in the Introduction, freshly trypsinizedmesenchymal cells were almost completely devoid of surface fibronectin (Fig. 3A).They did contain large quantities of antigen in the cytoplasm (Fig. 3B). Evidencefor a serum factor that promotes fibronectin deposition in the matrix was providedby comparing the localization of fibronectin in Mes maintained in dense monolayerculture in DMEM + 5 mg/ml bovine serum albumin (BSA) with that in the presenceof DMEM+4-5 mg/ml BSA+1 % fibronectin-depleted chicken serum. In thepresence of serum, fibronectin-containing fibrils associated with cells wereprominent (Fig. 4A). In the absence of serum, such strands were present but wereconsistently less-abundant and smaller (Fig. 4B). In the presence of serum, ex-tracellular fibronectin-containing fibrils are present by 3-5 h of culture (Fig. 5). Asimilar dependence of fibronectin deposition on serum has been observed in Mesaggregates in organ culture (Fig. 6). These observations are interpreted as support-ing the proposition that fibronectin deposition in the extracellular matrix is depen-dent on exposure of Mes to a factor(s) in the serum. Similar observations have been

Fig. 1. Autoradiogram showing the sorting of [3H]thymidine-labelled heart mesen-chymal cells (Mes) to the surface of aggregates containing Mes + cardiac myocytes (My).The aggregates were prepared by centrifuging a mixed suspension of trypsin-dissociatedcells into a compact pellet from a small volume of culture medium. The pellet was thencut into smaller pieces that were maintained in suspension culture in DMEM + 10%fibronectin-depleted chicken serum, B. A camera lucida drawing of the sectionphotographed in A, in which the labelled Mes nuclei are indicated by filled profiles andthe unlabelled My nuclei by open profiles. In this section, 27% of the cells are mesen-chymal, 75% of the cells in the surface layers are mesenchymal, and 90% of the cells inthe interior are myocytes; 2 days in culture. X 131.

Fig. 2. No detectable sorting of [3H]thymidine-labelled Mes from unlabelled My hasoccurred in mixed Mes + My aggregates at times when extensive fibronectin secretion byMes has already occurred; 5 h in culture. Culture medium, DMEM + 10% fibronectin-depleted chicken serum. X124.

Fig. 3. Mes cells are nearly devoid of surface fibronectin immediately following trypsiniz-ation (A) but contain considerable fibronectin in the cytoplasm (B). In this preparation,freshly trypsinized cells were attached to coverglasses derivatized with poly-L-lysine.Surface fibronectin (A) was revealed by exposing living cells to goat anti-fibronectin (4°C,30min) followed by FITC-rabbit anti-goat IgG. Cytoplasmic fibronectin (B) wasrevealed by treating attached cells with 1 mg/ml Triton X-100 before exposure to the firstantiserum. X564.

Fig. 4. Indirect immunofluorescent localization of fibronectin in culture of Mes platedat high density onto glass. Only extracellular fibronectin is visualized since exposure tothe primary antiserum was performed on living cultures (4°C, 0-5 h). Cells cultured inDMEM + 4-5 mg/ml B S A + 1 % fibronectin-depleted chicken serum (A) elaboratefibronectin-containing fibres in an extracellular matrix associated with the cells. Thisfibrillar matrix is sparse in cells cultured in DMEM +5 mg/ml BSA (B). The cell densityis equivalent to that in A. Diffuse fluorescence is present in the latter situation on areasof the coverglass not occupied by cells; 20h in culture. X580.

7 CEL69

P. B. Armstrong and M. T. Armstrong

Figs 5-6


reported by others (Chiquet, Eppenberg & Turner, 1981; Puri, Chiquet & Turner,1979).

Two observations haVe been made that are consistent with the notion that onlythose Mes cells situated close to the surface of the aggregate engage in the depositionof fibronectin into the extracellular matrix. (1) The scattered Mes cells located deepto the surface'of Mes+My aggregates lacked demonstrable pericellular fibronectin(Fig. 7). (2) Large quantities of extracellular fibronectin are restricted to the super-ficial regions of Mes aggregates cultured for 1-3 days in DMEM+10% fibronectin-depleted chicken serum. Much less fibronectin was demonstrable in the interiors ofsuch aggregates (Fig. 8).

Inversion of Mes+My aggregates bisected after the completion of sorting

The hypothesis under investigation proposes that the layer of Mes at the surface ofthe sorted-out Mes+My aggregate is more cohesive than the internal My tissue. Thisis supported by the observation that My aggregates envelope My+Mes aggregates(Fig. 9A) and Mes aggregates (Fig. 9B). In spreading situations of this sort, it isexpected that the less-cohesive tissue is the one that executes the spreading behaviourbecause it changes more readily from the compact spherical shape (the shape of theaggregate at the time of aggregate fusion) to the annular sheet of superficially locatedtissue (the form adopted at the completion of spreading).

Fig. 5. Indirect immunocytochemical detection of pericellular matrix fibronectin. Livingmonolayers (A) or aggregates (B) were exposed to goat anti-fibronectin (0-5 h, 4°C),washed and fixed in 4% freshly prepared paraformaldehyde. Aggregates were embeddedin paraffin and sectioned before exposure to the second antibody. Culture medium wasDMEM + 10% fibronectin-depleted chicken serum. In A and B extracellular fibronectinis present even after a short residence in culture (3 h, A; 5 h, B). In B, the fibronectin onlyat the very surface of the aggregate is revealed since there appears to be little penetrationof the primary antiserum beyond the surface of the living aggregate under the exposureconditions used in this study, A, X517; B, X536.Fig. 6. Indirect immunofluorescent localization of fibronectin in sections of Mes aggre-gates. Surface fibronectin only is revealed since exposure to the primary antiserum was con-ducted with living, intact aggregates. The aggregate in A had been cultured in DMEM +10% fibronectin-depleted chicken serum; the aggregate in B, in DMEM + 1 mg/ml BSA;21 h in culture. X 585.

Fig. 7. Indirect immunoperoxidase staining of fibronectin in a section of a Mes+Myaggregate in which the [3H]thymidine-labelled Mes cells have been visualized by radio-autography. The surface layers of the aggregate, which are composed predominantely ofMes, show an abundance of fibronectin whereas isolated Mes cells scattered about withinthe interior of the aggregate do not show detectable quantities of fibronectin (arrows); 1day in culture in DMEM +10% fibronectin-depleted chicken serum. X131.

Fig. 8. Indirect immunofluorescent localization of fibronectin in a section of a Mesaggregate that had been cultured for 3 days in DMEM + 10% chicken serum. The surfacelayers of the aggregate contain an abundance of fibronectin whereas the level of fibronectinin the interior is lower. X206.

Fig. 9. Spreading behaviour of paired aggregates. My aggregates spread over the surfaceof: A. 0-8 My + 0-2Mes aggregate (2 days of culture); B, Mes aggregate (26 h culture). Inboth cases, the My aggregate is unlabelled. The plane of section is approximately perpen-dicular to the plane of original contact between the two aggregates, A, X114; B, X178.

P. B. Armstrong and M. T. Armstrong

Figs 7-9 for legend see p. 187.

Tab

le 1

. Eff

ect

of c

ultu

re c

ondi

tions

on

spre

adin

g be

havi

our

Ext

ent

of s

prea

ding

-

-

--

1

Iden

tity

of

aggr

egat

e L

abel

led

over

unl

abel

ledb

U

nlab

elle

d ov

er l

abel

ledC

P

Day

sin

<p

A-,

A

Exp

erim

ent

Lab

elle

d8

Unl

abel

led8

ap

posi

tion

S

tron

gd

Mod

erat

ed

Slig

htd

Non

ed

hg

ht

Mod

erat

e ~

tro

ni

'One

po

pula

tion

of

aggr

egat

es w

as l

abel

led

wit

h [3

~]t

hy

mid

ine,

the

othe

r po

pula

tion

was

unl

abel

led,

and

ide

ntif

icat

ion

was

acc

ompl

ishe

d by

--

-

-

-

auto

radi

ogra

phy

of t

issu

e se

ctio

ns.

bT

he

labe

lled

tis

sue

spre

ad o

ver

the

unla

bell

ed t

issu

e.

'The

unl

abel

led

tiss

ue s

prea

d ov

er t

he l

abel

led

tiss

ue.

*Ext

ent o

f sp

read

ing:

Str

ong,

gre

ater

tha

n 2 o

f th

e su

rfac

e of

the

inte

rior

agg

rega

te w

as c

over

ed b

y th

e en

velo

ping

agg

rega

te;

Mod

erat

e, b

etw

een

4 and

4 of

the

sur

face

of

the

inte

rior

agg

rega

te w

as c

over

ed;

Sli

ght,

less

tha

n 4 o

f th

e su

rfac

e w

as c

over

ed;

Non

e, a

fla

t in

terf

ace

exis

ted

betw

een

the

two

aggr

egat

es.

'Tis

sue

type

(M

es =

mes

ench

yme;

0.7

5 M

y + 0

.25

Mes

= m

ixed

agg

rega

tes

cont

aini

ng 7

5% m

yocy

tes

and

25%

mes

ench

ymal

cel

ls)

and

cond

itio

ns

of cu

ltur

e of

th

e ag

greg

ates

bef

ore

fusi

on (

BS

A,

aggr

egat

es c

ultu

red

in D

ME

M +

5 mg/

ml

BS

A;

S,

aggr

egat

es c

ultu

red

in D

ME

M +

10%

fi

bron

ecti

n-de

plet

ed c

hick

en s

erum

).

'Num

ber

of a

ggre

gate

pai

rs.

- 00

\O


If the fibronectin matrix of superficially located Mes tissue, once established, isrelatively long-lived, then the arrangement of Mes surrounding My that is establishedby sorting is presumably quasi-stable and would be expected to persist only as longas the shell of Mes is intact without major areas of discontinuity that allow exposureof the interior domain of My to the surface of the aggregate. This prediction was testedby allowing [3H]thymidine-labelled Mes to sort out to the surface of Mes+Myaggregates in DMEM+10% chicken serum. After 1 or 2 days in culture, theaggregates were bisected with fine knives. A portion of the cut surface is then occupiedby My. The bisected aggregates were cultured for 1 or 2 days in DMEM + 5 mg/mlBSA to prevent further deposition of fibronectin and then fixed and sectioned. Asignificant fraction (51 of 76, or 67 %) of the bisected aggregates inverted, with Mytissue spreading from its interior location to envelop the Mes (Fig. 10). Thisbehaviour was not observed in control Mes+My aggregates that had been treated inan identical fashion but were not bisected.

Modulation of sorting and spreading behaviour by removal of serum from the culturemedium

The hypothesis under investigation proposes that the superficial sorting of Mes inMes+My aggregates is dependent on a serum factor-induced deposition of fibronectininto the pericellular matrix of Mes tissue. The observation that fibronectin depositionwas reduced by the omission of serum from the culture medium suggested that itshould be possible to suppress sorting by removal of serum. This indeed is what wasobserved: sorting was essentially complete by 1 or 2 days in medium containing 10 %chicken serum (Fig. 1), but was undetectable in medium from which the serum hadbeen omitted (Fig. 11).

The hypothesis also proposes that strong adhesiveness of Mes tissue is dependenton the elaboration of the fibronectin-rich extracellular matrix. If true, then Mesaggregates that have little fibronectin in the pericellular matrix should spread overMes aggregates whose pericellular matrix is rich in fibronectin. This has been ob-served: Mes aggregates prepared in DMEM+5 mg/ml BSA spread over Mes

Fig. 10..Following microsurgical bisection of sorted-out Mes+My aggregates, the Myspreads from its internal position to spread over the surface of the Mes tissue. The Mesis identified by [3H]thymidine label in this radioautogram; 1 day after bisection. X161.

Fig. 11. Sorting of [3H]thymidine-labelled Mes from unlabelled My is abortive if con-ducted under conditions in which fibronectin deposition is suppressed (culture inDMEM + 5 mg/ml BSA); 2 days in culture. The aggregate shown in Fig. 1 was preparedfrom the same batch of random reaggregates, but it was cultured in DMEM + 10%fibronectin-depleted chicken serum, B. A camera lucida drawing of the sectionphotographed in A, with the labelled (Mes) nuclei indicated by filled profiles and theunlabelled (My) nuclei represented by open profiles. In this section, 27% of the cells arelabelled and 38% of the cells at the surface of the aggregate are labelled; 2 days in culture.X127.Fig. 12. Spreading of [3H]thymidine-labelled aggregate of Mes prepared inDMEM + 5 mg/ml BSA over an unlabelled Mes aggregate prepared in DMEM+10%fibronectin-depleted chicken serum; 2 days in culture. X194.


IItI-

<f:

10 12

11A

Figs 10-12


aggregates prepared in DMEM+10% fibronectin-depleted chicken serum (Table 1,experiments 1 and 2; Fig. 12). The extent of spreading of Mes+My aggregates overMes aggregates was less if the Mes aggregates were prepared before fusion inserum-free medium than if the Mes aggregates were prepared in medium containingserum (Table 1, experiments 3 and 4). The fused aggregates were cultured inDMEM + 5 mg/ml BSA. Thus it appears possible to modulate the sorting and spread-ing behaviour of heart mesenchyme by manipulating the composition of the culturemedium (e.g. by omitting serum). Although the effects of culture in the absence ofserum are possibly manyfold, one documented effect is a reduction in the ability ofMes to elaborate a fibronectin-rich extracellular matrix (Figs 4, 6). The deprivationof serum has no detectable effect on the pattern or amount of collagen deposited inMy+Mes or Mes aggregates, as judged by the reticulin staining procedure, or on thedeposition of glycosaminoglycans, as judged by Alcian blue staining (data not shown).Based on observations reported previously (Armstrong, 1980, 1982; Armstrong &Armstrong, 1981), it is suggested that alteration of the fibronectin content of Mestissue is the cause of alteration in the sorting and spreading behaviour that results fromthe omission of serum from the culture medium.

DISCUSSION

In an earlier series of reports (Armstrong, 1978, 1980a, 1982; Armstrong & Arm-strong, 1981), evidence was presented that supports the hypothesis that the spreadingbehaviour of heart cell aggregates is dominated by a minor cell type, the cardiacmesenchyme cell, which exerts its influence by the deposition of cellular fibronectininto the extracellular matrix. Aggregates that rapidly develop a fibronectin-richmatrix segregate to an internal position when cultured in contact with aggregates thatcontain, at the time of apposition, lower levels of fibronectin. Addition ofmesenchyme-derived fibronectin to aggregates composed of cells that are unable tosecrete fibronectin (e.g. My aggregates) conferred behavioural characteristics similarto those of Mes+My aggregates (e.g. My aggregates pretreated with an extract of Mesthat was enriched in fibronectin were enveloped by control My aggregates; Arm-strong, 1980, 1982). The present report extends these ideas to provide an explanationfor the sorting-out behaviour of mixed My+Mes reaggregates. The explanation em-phasizes the importance of the composition of the extracellular matrix in the pattern-ing of tissues in organ-cultured tissue aggregates. This contrasts with previous ideasthat events of sorting and spreading are controlled solely by direct cell—cell adhesiveinteractions (Steinberg, 1963, 1964), but is consistent with a more general macro-scopic formulation of the differential adhesion hypothesis (Phillips & Steinberg,1969; Phillips & Davis, 1978; Steinberg, 1978a,6).

One important feature of this analysis is that it can account for the lack of identityin the arrangement of My and Mes produced by sorting out of random reaggregatesor by spreading of fused aggregate pairs. In the first situation Mes occupies thesurface, in the second My. In most tissue combinations reported in the literature,sorting and spreading result in an identical final pattern (Fig. 13). The importance of

tissue dissociation

Fibronectin and cell sorting

•' reaggregation

cell suspension

disorderedreaggregate

193

TISSUE A TISSUE B

tissue dissection FINAL TISSUEARRANGEMENT

tissuefragments

aggregatepair

tissuespreading

Fig. 13. Configurational equilibrium. For most binary combinations of tissues, an identi-cal final arrangement of the two tissues results from sorting out of disordered aggregatesprepared by allowing a mixed suspension of dissociated cells to reaggregate (upper half offigure) or by the tissue-spreading that follows apposition in organ culture of homogeneousaggregates of the same two tissues (lower half). Most frequently, the final arrangement isone in which one tissue completely envelopes the second. This behaviour is consistent withthe differential adhesion hypothesis, which proposes that the final arrangement of tissuesis governed by their relative cohesive strengths and is, thus, independent of the initialorganization of the binary aggregate.

this was recognized by M. Steinberg in the formulation of the differential adhesionhypothesis as a way to account for sorting and spreading (Steinberg, 1963, 1970,\97%a,b) and represents one of the principal elements of experimental support for thathypothesis. The present study suggests that the exception represented by thebehaviour of aggregates containing Mes and My tissues can be reconciled with thedifferential adhesion hypothesis. In fragment fusion, the Mes aggregate has at thetime of fusion been maintained in culture for a period adequate to establish afibronectin-rich matrix and is, as a consequence, more cohesive than the My aggregateand is enveloped. In sorting-out, the mesenchymal cells in the initial aggregate lackappreciable quantities of surface fibronectin and tissue segregation is viewed as occur-ring concomitant with the establishment of a fibronectin-containing matrix, selectively,by those mesenchymal cells that find themselves close to the surface of the Mes+Myreaggregate.

A useful feature of the hypothesis presented in this report is the explanation thatit suggests for certain aspects of cell sorting behaviour that were not previouslyunderstood. Especially interesting are a class of phenomena collectively labelledposition reversal. These include the observation that ventricle tissue from 2-day chickembryos spreads over a variety of other tissues; whereas these same tissues spread over5-day embryonic ventricle (Lesseps, 1973; Lesseps & Brown, 1974), pigmentedretinal epithelium (PRE) usually spreads over ventricle tissue (the reverse being


observed when reaggregates only a few hours old are apposed; Armstrong & Nieder-man, 1972) but sorts internally to ventricle (Armstrong, 1970), and ventricle tissuemaintained in culture for 0-5 day is enveloped by reaggregates of dissociated ventriclebut, itself, envelops ventricle tissue maintained in culture for 2-5 days (Phillips,Wiseman & Steinberg, 1977; Wiseman, Steinberg & Phillips, 1972). These, and othersimilar phenomena, have in common the use of heart tissue as one partner in a pairing ofaggregates or in cell sorting and, based on observations reported in the present study,can be accounted for by the suggestion that heart tissue is characterized by segregationto the surface of a strongly cohesive layer of Mes under conditions of organ culture. Theeffect of age of heart tissue (Lesseps, 1973; Lesseps & Brown, 1974) is explainable bythe absence of mesenchyme in the 2-day heart and its presence by 5 days (Fitzharris,1981;Fitzharris&Markwald, 1982; Kinsella& Fitzharris, 1980;Manasek, 1970,1979;Markwald, Fitzharris & Manasek, 1977; Patten, Kramer&Barry, 1948; Thompson &Fitzharris, 1979). The effects of trypsinization and time of residency in culture(Phillipser al. 1977; Wiseman et al. 1972) are explainable on the basis of an increase inthe extent and fibronectin content of the superficial layer of Mes with prolonged culture.The differences in position following spreading and sorting of ventricle tissue and PRE(Armstrong, 1970; Armstrong & Niederman, 1972) can be explained if it is suggestedthat, as in Mes+My aggregates, the most cohesive tissue, the Mes, sorts to the surfaceof the aggregate and forces into the interior the other less-cohesive cell types (which, inthe case of PRE—ventricle combinations would be My and PRE). These suggestions arecurrently under investigation using culture in the absence or presence of serum tomanipulate the degree of fibronectin deposition in Mes.

The study described here is an attempt to provide a coherent explanation for theunderlying basis of cell sorting in heart tissue. The evidence implicates fibronectin asplaying an essential role in the process. Whether the analysis can be generalized totissues other than heart remains a subject for further investigation. Nicol & Garrod(1982) have demonstrated that fibronectin is associated with limb mesenchymal tissuethat has sorted out in two-dimensional culture from a variety of epithelial tissues (but,interestingly, with the mesenchymal tissue forming the discontinuous and, by impli-cation, less-cohesive phase). It will be interesting to determine if fibronectindeposition is required for sorting in this instance. It may prove that extracellularmatrix components are involved in sorting only of aggregates containing mesen-chymal tissues. The sorting behaviour of epithelial tissues appears to be regulated byintegral membrane proteins (the cell adhesion molecules (CAMs); Edelman, 1983;Rutishauser, Thiery, Brackenbury & Edelman, 1978; and proteins of cell junctions,Overton, 1977). Epithelial tissues are organized with extensive cell—cell contact andwith the extracellular matrix being limited to the basal lamina. Mesenchymal tissuesare organized with relatively little direct cell-cell contact and with the extracellularmatrix forming the continuous phase of the tissue. It is possible that these differencesin tissue organization impose differences on the character of the molecular speciesgoverning sorting behaviour, with integral membrane proteins serving as recognitionmolecules for epithelia and extracellular matrix components playing this role formesenchymal tissues.


We thank Drs K. M. Yamada and D. R. Garrod for providing anti-fibronectin antisera and DrC. A. Erickson for reading the manuscript. This study was supported by grant no. PCM 80-24181from the National Science Foundation and grant no. 1 RO1 GM30O62-01 from the NationalInstitute of General Medical Services.

REFERENCES

ALI, I. U., MAUTNER, V., LANGA, R. &HYNES, R. 0 . (1977). Restoration of normal morphology,adhesion and cytoskeleton in transformed cells by addition of a transformation-sensitive protein.Cell 11, 115-126.

ARMSTRONG, M. T. & ARMSTRONG, P. B. (1978). Cell motility in fibroblast aggregates. J. Cell Set.33, 37-52.

ARMSTRONG, M. T. & ARMSTRONG, P. B. (1979). The effects of antimicrotubule agents on cellmotility in fibroblast aggregates. Expl Cell Res. 120, 359-364.

ARMSTRONG, P. B. (1970). A fine structural study of adhesive cell junctions in heterotypic cellaggregates. J . Cell Biol. 47, 197-210.

ARMSTRONG, P. B. (1971). Light and electron microscope studies of cell sorting in combinationsof chick embryo neural retina and retinal pigment epithelium. Wilhelm Roux Arch. EnttuMech.Org. 168, 125-141.

ARMSTRONG, P. B. (1977). Cellular positional stability and intercellular invasion. BioSci. 27,803-809.

ARMSTRONG, P. B. (1978). Modulation of tissue affinities of cardiac myocyte aggregates by mesen-chyme. Devi Biol. 64, 60-72.

ARMSTRONG, P. B. (1980). Ability of a cell-surface protein produced by fibroblasts to modify tissueaffinity behavior of cardiac myocytes. J. Cell Sci. 44, 263-271.

ARMSTRONG, P. B. (1982). Role of the extracellular matrix in the expression of tissue affinities. InCell Behaviour (ed. R. Bellairs, A. S. G. Curtis & G. A. Dunn), pp. 203-224. CambridgeUniversity Press.

ARMSTRONG, P. B. (1984). Invasiveness of non-malignant cells. In Invasion (ed. M. Mareel & K.Caiman), pp. 126—167. Oxford University Press.

ARMSTRONG, P. B. & ARMSTRONG, M. T. (1981). Immunofluorescent histological studies of therole of fibronectin in the expression of the associative preferences of embryonic tissues..7. Cell Sci.50, 121-133.

ARMSTRONG, P. B. & NIEDERMAN, R. (1972). Reversal of tissue position after cell sorting. DeviBiol. 28, 518-527.

BELLAIRS, R., CURTIS, A. & DUNN, G. (1982). Cell Behaviour. Cambridge University Press.BLUMBERG, P. M. & ROBBINS, P. W. (1975). Effect of proteases on activation of resting chick

embryo fibroblasts and on cell surface proteins. Cell 6, 137-147.CHEN, L. B., GALLIMORE, P. H. & MCDOUGALL, J. K. (1976). Correlation between tumor

induction and the large external transformation sensitive protein on the cell surface. Proe. natn.Acad. Sci. U.SA. 73, 3570-3574.

CHIQUET, M., EPPENBERGER, H. M. & TURNER, D. C. (1981). Muscle morphogenesis: evidencefor an organizing function of exogenous fibronectin. Devi Biol. 88, 220—235.

DEHAAN, R. L. (1967). Regulation of spontaneous activity and growth of embryonic heart cells intissue culture. Devi Biol. 16, 216-249.

EDELMAN, G. M. (1983). Cell adhesion molecules. Science, N.Y. 219, 450-457.ENGVALL, E. & RUOSLAHTI, E. (1977). Binding of soluble form of fibroblast surface protein,

fibronectin, to collagen. Int. J. Cancer 20, 1-5.ERICKSON, C. A. & TURLEY, E. A. (1983). Substrata formed by combinations of extracellular

matrix components alter neural crest cell motility in vitro. J. Cell Sci. 61, 299-323.FITZHARRIS, T. P. (1981). Endocardial shape change in the truncus during cushion formation. In

Mechanisms of Cardiac Morphogenesis and Teratogenesis, Perspectives in CardiovascularResearch, vol. 5 (ed. T. Pexider), pp. 227-235. New York: Raven Press.

FITZHARRIS, T. P. & MARKWALD, R. R. (1982). Cellular migration through the cardiac jellymatrix: a stereoanalysis by high-voltage electron microscopy. Devi Biol. 92, 315-329.


GOLD, L. I. & PEARLSTEIN, E. (1980). Fibronectin-collagen binding and requirement during celladhesion. Biochem. jf. 186, 551-559.

GRINNELL, F. (1978). Cellular adhesiveness and extracellular substrata./w«.7?et;.Cy/o/. 53,65-144.GRINNELL, F. & MINTNER, D. (1978). Attachment and spreading of baby hamster kidney cells to

collagen substrata: effects of cold-insoluble globulin. Proc. natn. Acad. Set. U.SA. 75,4408-4412.

HEDMAN, K., VAHERI, A. & WARTIOVAARA, J. (1978). External fibronectin of cultured humanfibroblasts is predominantly a matrix protein. J . Cell Biol. 76, 748-760.

HOLTFRETER, J. (1944). Experimental studies on the development of the pronephros. Rev. Can.Biol. 3, 220-250.

HOOK, M., RUBIN, K., OLDBERG, A., OBRINK, B. & VAHERI, A. (1977). Cold-insoluble globulinmediates the adhesion of rat liver cells to plastic petri dishes. Biochem. biopkys. Res. Commun.79, 726-733.

HYNES, R. 0 . (1973). Alteration of cell-surface proteins by viral transformation and by proteolysis.Proc. natn. Acad. Set. U.SA. 70, 3170-3174.

JOHNSON, G. D. & NOGUEIRA ARAUJO, G. M. DE C. (1981). A simple method of reducing thefading of immunofluorescence during microscopy. J. Immun. Meth. 43, 349-350.

KINSELLA, M. G. & FrrzHARRis, T. P. (1980). Origin of cushion tissue in the developing chickheart: cinematographic recordings of in situ formation. Science, N.Y. 207, 1359-1360.

KLEBE, R. J. (1974). Isolation of a collagen dependent cell attachment factor. Nature, Land. 250,248-251.

KLEINMAN, H. K., KLEBE, R. J. & MARTIN, G. R. (1981). Role of collagenous matrices in theadhesion and growth of cells. J. Cell Biol. 88, 473-485.

KOPRIWA, B. M. & LEBLOND, C. P. (1962). Improvements in the coating technique of radio-autography. J . Histochem. Cytochem. 10, 269-284.

LESSEPS, R. J. (1973). Developmental change in morphogenetic properties: Embryonic chick hearttissue and cells segregate from other tissues in age-dependent patterns. J . exp. Zool. 185,159-168.

LESSEPS, R. J. & BROWN, S. A. (1974). Further evidence for a developmental change in morpho-genetic properties of embryonic chick heart cells. J. exp. Zool. 187, 261-266.

LETOURNEAU, P. C , RAY, P. N. & BERNFIELD, M. R. (1980). The regulation of cell behavior bycell adhesion. In Biological Regulation and Development, vol. 2 (ed. R. F. Goldburger), pp.339-376. New York: Plenum Press.

MANASEK, F. J. (1970). Histogenesis of the embryonic myocardium. Atn.J. Cardiol. 25, 149-168.MANASEK, F. J. (1979). Organization, interactions, and environment of heart cells during myocar-

dial ontogeny. In Handbook of Physiology, section 2, vol. 1, Tne Cardiovascular System: TheHeart (ed. R. M. Berne), pp. 29-42. Bethesda, MD: American Physiological Soc.

MARKWALD, R. R., FITZHARRIS, T. P. & MANASEK, F. J. (1977). Structural development ofendocardial cushions. Am.J. Anat. 148, 85-120.

MOSCONA, A. (1957). Development in vitro of chimeric aggregates of dissociated embryonic chickand mouse cells. Proc. natn. Acad. Sri. U.SA. 43, 184-194.

NICOL, A. & GARROD, D. R. (1982). Fibronectin, intercellular junctions and the sorting-out ofchick embryonic tissue cells in monolayer. J. Cell Sri. 54, 357-372.

OVERTON, J. (1977). Formation of junctions and cell sorting in aggregates of chick and mouse cells.Devi Biol. 55, 103-116.

PATTEN, B. M., KRAMER, T. C. & BARRY, A. (1948). Valvular action in the embryonic chick heartby localized apposition of endocardial masses. Anat. Rec. 102, 299—311.

PEARLSTEIN, E. (1976). Plasma membrane glycoprotein which mediates adhesion of fibroblasts tocollagen. Nature, Land. 262, 497-500.

PHILLIPS, H. M. & DAVIS, G. S. (1978). Liquid-tissue mechanics in amphibian gastrulation:germ-layer assembly in Rana pipiens. Am. Zool. 18, 81-93.

PHILLIPS, H. M. & STEINBERG, M. S. (1969). Equilibrium measurements of embryonic chick celladhesiveness. I. Shape equilibrium in centrifuged fields. Proc. natn. Acad. Sri. U.SA. 64,121-127.

PHILLIPS, H. M., WISEMAN, L. L. & STEINBERG, M. S. (1977). Self vs. nonself in tissue assembly.Correlated changes in recognition behavior and tissue cohesiveness. Devi Biol. 57, 150-159.

PURJ, E. C , CHIQUET, M. & TURNER, D. C. (1979). Fibronectin-myoblast fusion in suspensioncultures. Biochem. biophys. Res. Commun. 90, 883-889.


ROVASIO, R. A., DELOUVEE, A., YAMADA, K. M., TIMPL, R. & THIERY, J.-P. (1983). Neural

crest migration: requirements for exogenous fibronectin and high cell density. J. Cell Biol. 96,462-473.

RUTISHAUSER, U., THIERY, J. P., BRACKENBURY, R. & EDELMAN, G. (1978). Adhesion amongneural cells of the chick embryo. III. Relationship of the surface molecule CAM to cell adhesionin the development of histotypic patterns. J'. Cell Biol. 79, 371-381.

SAISON, M., VAN LEUVEN, F., CASSIMAN, J.-J. & VAN DEN BERGHE, H. (1983). Evidence for

a role of fibronectin and other cell surface components in a cell—cell layer adhesion of humanfibroblasts. Expl Cell Res. 143, 237-245.

STEINBERG, M. S. (1963). Reconstitution of tissues by dissociated cells. Science, N.Y. 141,401-408.

STEINBERG, M. S. (1964). The problems of adhesive selectivity in cellular interactions. In CellularMembranes in Development (ed. M. Locke), pp. 321-366. New York: Academic Press.

STEINBERG, M. S. (1970). Does differential adhesion govern self-assembly processes in histo-genesis? Equilibrium configurations and the emergence of a hierarchy among populations ofembryonic cells. J . exp. Zool. 173, 395-434.

STEINBERG, M. S. (1978a). Cell-cell recognition in multicellular assembly: levels of specificity. InCell-Cell Recognition, Symp. Soc. exp. Biol. vol. 32 (ed. A. S. G. Curtis), pp. 25-49. CambridgeUniversity Press.

STEINBERG, M. S. (19786). Specific cell ligands and the differential adhesion hypothesis: how dothey fit together? In Specificity oj'Embryological Interactions, Receptors and Recognition, ser. B.,vol. 4 (ed. D. R. Garrod), pp. 99-130. London: Chapman & Hall.

THOMPSON, R. P. & FITZHARRIS, T. P. (1979). Morphogenesis of the truncus ateriosus of the chickembryo heart: the formation and migration of mesenchymal tissue. Am.J.Anat. 154, 545-556.

TOWNES, P. & HOLTFRETER, J. (1955). Directed movements and selective adhesion of embryonicamphibian cells. J. exp. Zool. 128, 53-120.

TRINKAUS, J. P. (1976). On the mechanism of metazoan cell movement. In The Cell Surface inAnimal Embryogenesis and Development (ed. G. Poste & G. L. Nicholson), pp. 225-329. NewYork: North-Holland.

TRINKAUS, J. P. (1984). Cells Into Organs, 2nd edn. Englewood Cliffs, N.J.: Prentice-Hall.WISEMAN, L. L., STEINBERG, M. S. & PHILLIPS, H. M. (1972). Experimental modulation of

intercellular cohesiveness: reversal of tissues assembly patterns. Devi Biol. 28, 498-517.YAMADA, K. M. (1978). Immunological characterization of a major transformation-sensitive

fibroblast cell surface glycoprotein. Localization, redistribution, and role in cell shape. J. CellBiol. 78, 520-541.

YAMADA, K. M. & WESTON, J. A. (1974). Isolation of a major cell surface glycoprotein fromfibroblasts. PTVC. natn. Acad. Sci. U.SA. 71, 3492-34%.

YAMADA, K. M. & WESTON, J. A. (1975). The synthesis, turnover, and artificial restoration of amajor cell surface glycoprotein. Cell 5, 75-81.

YAMADA, K. M., YAMADA, S. S. & PASTAN, I. (1975). The major cell surface glycoprotein ofchick embryo fibroblasts is an agglutinin. Proc. natn. Acad. Sci. U.SA. 72, 3158—3162.

YAMADA, K. M., YAMADA, S. S. & PASTAN, I. (1976). Cell surface protein partially restoresmorphology, adhesiveness, and contact inhibition of movement to transformed fibroblasts. Proc.natn. Acad. Sci. U.SA. 73, 1217-1221.

ZETTER, B. R., MARTIN, G. R., BIRDWELL, C. R. & GOSPODAROWICZ, D. (1978). Role of the highmolecular weight glycoprotein in cellular morphology, adhesion, and differentiation. Ann. N.Y.Acad. Sci. 312, 299-316.

(Received 8 February J984-Accepted 29 February 1984)

A ROLE FOR FIBRONECTIN IN CELL SORTINGdicular to the interface between the two members of the pair....

Documents

Transcript of A ROLE FOR FIBRONECTIN IN CELL SORTINGdicular to the interface between the two members of the pair....