A major difference between serum and fibronectin in the divalent cation
requirement for adhesion and spreading of BHK21 cells
J. G. EDWARDS, R. T. ROBSON* and G. CAMPBELL
Department of Cell Biology, University nf Glasgotv, Glasgow G12 SQQ, UK
•Present address: Central Toxicology Laboratory, ICI pic, Alderley Park, Macclesficld, Cheshire SK10 4TJ, UK
Summary
Adhesion and spreading of BHK21 cells on
adsorbed bovine and foetal bovine serum require
addition to the medium of a divalent cation.
Divalent cations are effective in the order
M n 2 + > C o z + > M g 2 + > C a 2 + , with Ca2+ ineffective below 10~4M. On purified fibronectin, however, no added divalent cation is required, since
the requirement is largely met by adventitious
Caz+ (circa 10~5M) in nominally divalent cationfree saline. In such background Ca 2+ , adhesion
and spreading on fibronectin are only slightly
slower than in optimal Mg 2+ , and appear identical, morphologically, and in sensitivity to cytochalasin D. Cells also spread on fibronectin in
response to Mg2"*", when Ca2+ is buffered below
10~6M, showing that external Ca2+ is not needed
as a source for an increase in internal Ca 2+ . Cells
can be induced to spread on serum in low Ca2+
by substantially increasing the fibronectin concentration; this supports other evidence that at
its unsupplemented concentration, fibronectin
contributes little to the spreading of these cells on
serum.
The Ca2+ requirement for spreading on preparations of vitronectin (serum spreading factor),
partially purified from bovine serum, is similar
to that on whole serum. Thus the difference in
divalent cation requirement between serum and
fibronectin may arise because, on serum, the
dominant protein responsible for induction of
spreading is vitronectin rather than fibronectin.
We argue that such a difference in ion requirement between these surfaces points to a site of
action of the ions at the cell surface, perhaps
directly on binding of the adsorbed proteins by
their receptors, or in receptor-specific transduction events, rather than via ion fluxes, on cytoskeletal responses common to different surfaces.
Models are discussed to account for why the
divalent cation requirement should differ for
interaction of cells with different substrateadsorbed proteins.
Introduction
review). Considerable progress has been made in the
identification and characterization of specific matrix
proteins, such as fibronectin, laminin and serum
spreading factor (vitronectin), which when adsorbed or
otherwise bound to artificial substrata induce adhesion
and rapid spreading; and recently in the identification
of specific plasma membrane proteins that may function as receptors for the extracellular proteins (Damsky
et al. 1984; Pytela et al. 1985fl). Before these specific
spreading-inducing molecules were known, adhesion
was much studied on surfaces coated with a mixture of
proteins adsorbed from serum, such as might be
encountered by cells in the usual serum-supplemented
culture media. It was commonly found that, to adhere
The adhesion of cells cultured from animal tissues to
non-cellular substrates has attracted considerable attention over many years, because it seems to reproduce
outside the animal the normal interaction of cells with
various forms of extracellular matrix (Grinnell, 1978;
Yamada, 1983; Ruoslahti et al. 1985, for reviews).
Interest in the phenomenon is heightened by awareness
that the cell spreading that accompanies adhesion
depends on the active participation of cytoplasmic
filaments, so that the interaction must involve some
form of signal transduction from cell surface to interior
(Rabinovitch & DeStefano, 19736; Vasiliev, 1985, for
Journal of Cell Science 87, 657-665 (1987)
Printed in Great Britain © The Company of Biologists Limited 1987
Key words: fibronectin, adhesion, spreading.
657
and spread on such surfaces, cells require extracellular
Ca 2+ or Mg 2 + at the millimolar concentrations characteristic of extracellular fluids (Takeichi & Okada, 1972;
and briefly reviewed by Rabinovitch & DeStefano,
19736; Grinnell, 1978). Although in culture media the
requirement is met by Mg 2 + and/or Ca 2 + , certain nonphysiological ions, notably Mn 2 + , are effective at much
lower concentrations (Garvin, 1968; Stenn & Core,
1986), and indeed in some cases Mn 2 + induces a
response where Ca 2+ and Mg 2 + are inactive (Rabinovitch & DeStefano, 1973a). The part played by these
ions in adhesion and spreading remains a matter of
debate. Clarification of their role seems important for
understanding the mechanism by which matrix proteins induce adhesion and spreading. We have therefore re-investigated the divalent cation requirement for
adhesion and spreading of BHK21 cells by comparing
the response of these cells on adsorbed serum with that
on plasma fibronectin.
We were surprised to find that although, on serum,
Mg 2 + or Ca 2+ must be added at millimolar concentration (much as reported in earlier studies of many cell
types), on plasma fibronectin no such requirement is
evident. It emerged that, on the latter surface, both
adhesion and spreading are almost fully induced by
adventitious Ca 2 + , present at some 100-fold lower
concentration in nominally divalent cation-free media.
We suggest that such a difference in divalent cation
requirement between different inducing surfaces has
important implications for interpretation of the role of
external divalent cations.
Materials and methods
Cells
BHK21, clone 13 cells (Stoker & Macpherson, 1964) were
grown as attached cultures on plastic, in Glasgow-modified
Eagle's MEM with foetal bovine serum and tryptose phosphate broth (8:1:1, by vol. = EFT). Cells were periodically
checked for mycoplasma by fluorescence microscopy after
staining with Hoechst 33258 (Russell el al. 1975), and serially
passaged for not more than 6 weeks (normally 4).
Balanced salines
The balanced saline (HH), similar to Hanks', but buffered
with Hepesat pH7-4, contained 140mM-NaCl, S-4mM-KCl,
l-27niM-CaCl2, 098mM-MgCl 2 , 556mM-D-glucose, lOmMHcpes. HS denotes the same medium minus divalent cations.
Fibronectin
Bovine plasma fibronectin was purified from calf serum by
affinity chromatography on gelatin-Sepharose (Engvall &
Ruoslahti, 1977), and used to coat glass by dilution directly
into HH from solution (1 mgrnl"') in 8M-urea.
Serum spreading factor
Serum spreading factor was partially purified from calf serum
previously depleted of fibronectin on gelatin-Sepharose, by
658
J. G. Edwards et al.
adsorption on a glass bead column, and elution with 0 6 MKHCO3, O2M-K2CO3, as described by Barnes & Silnutzer
(1983) for human serum. Eluted material was dialysed
against HS. Though not purified to homogeneity, this
material had higher specific activity in spreading than purified fibronectin. Since the low concentrations of this preparation necessary to induce spreading were insufficient to
block access of cells to the naked glass, in experiments with
this material coverslips coated with either fibronectin or the
vitronectin-enriched fraction were further incubated with a
blocking solution of haemoglobin (500f*gml~' for 30 min),
a protein chosen for its homogeneity and inactivity in
spreading.
Calcium
Background Ca + concentrations in buffered salines were
determined by addition of Quin 2, ligand 3b described by
Tsien (1980) (typically 15jiM), and measurement of the
absorption at 265 nm. Free Ca 2+ in Ca 2 + /EGTA mixtures
was calculated using the equation of Bremel & Weber (1975)
and apparent p/Q = 7-4 for EGTA at pH 7-4.
Adhesion assay
To measure adhesion to protein-coated surfaces we determined the proportion of radiolabelled cells that were retained
on coated glass coverslips after stationary incubation followed
by a standard rinsing procedure. For labelling with
[ 2 P]phosphate, cells were grown for 1 day in medium minus
tryptose phosphate (EF), containing 2-8MBqml~ 1 carrierfree 32P-labelled inorganic phosphate (Amersham International), and chased by a further day's growth in unlabelled
EFT. For chromate labelling (found to be more convenient
yet equally reliable) 51Cr-labelled sodium chromate
(13-22GBq per mg Cr) was added (340MBq per 5ml
growth medium on a culture of area 25 cm2) 18 h before using
the cells.
Cells were removed from culture surfaces using a low
trypsin-EDTA procedure previously described (Edwards el
al. 1975), aspirated in EFT, collected by centrifugation and
rinsed twice by resuspension in HS. (Cells prepared by this
procedure spread as fast and as extensively on fibronectin and
serum as cells resuspended by the use of EDTA alone, and
moreover their spreading was not inhibited by cycloheximide. Thus, despite evidence that more extensive exposure
of cells to trypsin in the presence of EDTA can inactivate
spreading on fibronectin (Oppenheimer-Marks & Grinnell,
1984), this resuspension procedure does not detectably deplete the receptor function of the cells.) Cells were finally
resuspended in HS, and divalent cations added from 100-fold
concentrates just before cells were dispensed onto coverslips;
13 mm glass coverslips were cleaned by immersion for
10-15 min in 1:1 (v/v) nitric/sulphuric acids, rinsed with
water and coated with serum by immersion for 15 nun in 15 %
(v/v) serum in HS, or in the appropriate concentration of
fibronectin or serum spreading factor also diluted in HS, then
rinsed with HS. To allow rapid processing of numerous
samples, sometimes after short incubation times, coverslips
were placed immediately before use in rows supported on two
parallel glass rods, diameter 4-5 mm. Several pairs of such
rods were glued onto a board, the whole of which could be
covered between manipulations by a perspex lid to maintain
humidity. Dispensing of cells, incubation and washing of
coverslips were carried out in a room maintained at 37CC. A
0-1 ml sample of cell suspension containing 104 cells in the
appropriate medium was pipetted onto each eoverslip, on a
30-s cycle between samples. After the appropriate incubation, coverslips were picked up using forceps, dipped into
the first wash in HS, and further washed by immersion, in a
standard position and orientation, in HS stirred by a magnetic bar. The washed samples were then transferred directly
to vials for liquid scintillation counting (for 32P) or gamma
counting (for s l Cr). The percentage of cells attaching was
calculated as lOOXctsmin"1 on the coverslip/cell-bound
ctsmin" 1 in an 0-1 ml sample of cell suspension. Less than
10% of either radiolabel was released from cells to the
medium.
Cell spreading
To determine the extent of cell spreading we measured the
projected area of digitized images of fixed and stained cells.
Coverslips (22 mm square) were prepared as for adhesion
assays, but to obtain evenly distributed cells coverslips were
placed in 35 mm tissue-culture plastic dishes and 5X 104 cells
in 2ml suspension were added. After incubation at 37°C,
cells were fixed by addition of lml 4 % formaldehyde in
phosphate-buffered saline, pH7-2, rinsed with stain solvent,
water/methanol/acetic acid, 50:50:7 (by vol.), stained with
freshly filtered Kenacid Blue (0-1 % in that solvent), rinsed
twice with water, and mounted in Gurr's Clearmount.
Images were obtained using a 25X objective of a Leitz
Ortholux microscope equipped with a Hamamatsu Vidicon
C1000 camera, with M1438 gain expansion and zero offset
module. The images were digitized for input to the screen
memory of a BBC microcomputer using a Data Harvest video
interface (Data Harvest, Leighton Buzzard, UK). The
captured images were in three shades plus background at a
resolution of 160x256. The areas of 100 or 200 cells per
eoverslip were measured by counting for each cell the total of
8-connected pixels darker than background using specially
written software. Grey shade detection was standardized by
measurement of a reference field of cells. With the 25 X
objective, 1 pixel corresponded to 2-1 ^m 2 . In conditions
where induction of spreading is very low (e.g. on fibronectin
in Ca 2+ below 10~ 6 M) the spreading assay can over-estimate
the (very small) increase in mean spread area, as a result of
selective loss of rounded cells. This error in no way vitiates
our interpretation.
Actin staining
The distribution of F-actin in spreading cells was visualized
by staining with a fluorescent conjugate of phallotoxin using a
method based on that of Barak et al. (1980). Cells were fixed
with 4 % formaldehyde in PBS, rinsed in this buffer,
permeabilized in acetone for 5 min at —20°C, and stained
with 2 JIM (=Z-3figml~i)
TRITC-Phalloidin (Sigma),
rinsed twice with PBS, and viewed and photographed under
epi-illumination with a 50X objective in a Vickers M41
Photoplan microscope.
Results
We have used two different assays in this work:
adhesion of cells measured as the proportion of (radiolabelled) cells not removed by a standard rinse after a
period in stationary contact with the surface, and
spreading measured as mean increase in projected
area/cell. On both serum and fibronectin, some cells
remain adherent in the adhesion assay that arc not
extensively spread, but the two phenomena responded
largely in parallel to effects of divalent cations, giving
the impression that both are limited by a single divalent
cation-dependent step, as noted previously by Rabinovitch & DeStefano (1973c) and Stenn & Core (1986).
Comparison of the divalent cation requirements for
adhesion to serum and to fibronectin
Fig. 1 shows a typical time course of adhesion of cells to
serum-coated glass, showing, at the earliest times, the
often-reported superiority of Mn 2 + over M g 2 + . To
obtain a measure sensitive to differences in initial rate,
we chose 20 min as a standard time at which to compare
the effects of various divalent cations and the omission
of all. Fig. 2 summarizes data from a series of such
experiments with four ions of interest. The ion effective at lowest concentration is clearly M n z + , followed
by Co 2 + , with the physiological ions Ca 2 + and Mg 2 +
inducing comparable adhesion only at millimolar concentrations. The fall-off in activity of Ca 2+ at 10~2M is
sometimes observed also in spreading assays. The
important feature of these experiments in the present
context, however, is the extremely low background
of adhesion without added divalent cation (1-8%
60
,40
=3 20
U
10
20
Time (min)
Fig. 1. Time course of divalent cation-induced adhesion to
serum-coated glass coverslips. 32P-labelled cells were
diluted into HS with appropriate divalent cation, incubated
at 37 °C on glass coverslips and rinsed at various times.
( • ) 0-1 mM-Mn 2+ ; (O) l-0mM-Mg 2+ ; duplicate coverslips
for each time; + shows value (1 (± 0-7) %) for cells with
no divalent cation added, summarized from 11
experiments.
Adhesion and spreading on serum and
fibronectin
659
Mn2+
1r
•
•
•
80 0-5
60 0"
=t
lr
Co2+
40
I!
1
•
20
0-5
_•-.
u
•o
re
J2
..Q-r-P—-rfl
£ 80
U
60
c
Mg2+
1 x;
•a
40
a
= 0-51—
r
Ir
20
10
2+
Ca
0-5
0"
3
0
10"5
10~4
10"3
Divalent cation added (M)
10"
Fig. 2. Dose-response of adhesion to serum-coated glass,
for four divalent cations. Data are expressed for each
experiment relative to 0-1 mM-Mn2+ to compensate for
variation (±20%) in this (optimal) attachment between
different cell suspensions. Minimum of five separate
experiments with separately prepared cell suspensions at
each value, with error bar ±1 S.D.
(±1-3%)), in 11 experiments, of the value in 0-1 mMMn 2 + ).
When the adsorbed serum was replaced with purified
fibronectin, however, a totally different result was
obtained. In the experiment shown in Fig. 3, the same
cell suspension (either without added divalent cation,
or with fully inducing Mg 2 + ) was compared for adhesion on serum and fibronectin. On serum, as
expected from the data of Fig. 2, in the absence of
added divalent cations, adhesion was extremely low,
whereas on fibronectin it was high and similar to that of
the Mg 2+ -containing suspension.
Spreading on fibronectin without added divalent
cations
Not only do cells adhere to fibronectin in the absence of
added divalent cations, they also spread extensively.
Quantification by means of digitized images showed
that spreading in IIS is somewhat less extensive than
when 10mM (fully inducing) Mg 2 + is added. In six
experiments, the mean area/cell on fibronectin, coated
660
jf. G. Edwards et al.
20 30
Time (min)
40
Fig. 3. Comparison of fibronectin and serum for divalent
cation-dependence of cell adhesion. 3lCr-labelled cells were
incubated on glass coverslips precoated with bovine plasma
fibronectin, 25 fig ml"1 (•), or 15% calf serum (O). In the
upper figure, no divalent cations were added. In the lower,
incubation was in the presence of 10mM-IVIg2+ added just
prior to incubation.
at 25 fig ml"', in HS was 69% (±5%) of that in
HS+10mM-Mg 2+ . A typical distribution of areas is
shown in Fig. 6. However, the morphology of both
early and late stages is indistinguishable by both phasecontrast (not shown) and fluorescence microscopy of
cells stained for F-actin. Cells both with and without
added divalent, stained after incubation for 25 min,
showed prominent actin-containing peripheral ruffles,
and after 60 min both exhibited similar stress fibres
(Fig. 4). Spreading is believed to require the active
participation of intracellular filament systems, in particular of actin (Vasiliev, 1985) and it has been suggested that external divalent cations may be required
(perhaps via effects on ion fluxes) to establish internal
ionic conditions favourable to such events (Rabinovitch & DeStefano, 1973o). We therefore tested the
sensitivity of spreading, with and without Mg 2 + , to the
inhibitor of actin polymerization, cytochalasin D.
Spreading was equally sensitive under both conditions,
being completely inhibited at 1 fig ml""1 (data not
shown), so that spreading in HS is not a residual
'passive' spreading, but is just as dependent on cytoskeletal events as when Mg 2+ -activated.
Role of background Ca2+
The fact that cells spread on fibronectin in the absence
of added divalent cations could mean either that such
A
0-6
0-5 •
0-4 0-3
0-2 -
01
•
-F3—i
300
600
0
300
Cell area (pixels)
Fig. 5. Cell spreading on fibronectin: comparison of area
distributions on low Ca 2+ and on optimal Mg 2 + . Areas of
200 cells were measured after 90 min spreading on glass
pre-coated with fibronectin (25//grill" 1 ). A. Suppression of
spreading when free Ca 2+ is reduced to 035X 10~6M in US
containing 0'4mM-EGTA, 036mM-Ca 2+ . The peak at
about 100 pixels is of rounded cells. Mean area, 90;/m 2 .
B. Optimal spreading induced by lOmM-Mg2"1" in US,
[Ca 2 + ]=5Xl0~ 6 M. Mean area, 400//m 2 .
C - D . Intermediate spreading induced by background Ca 2+
in HS prepared in quartz redistilled (C) or normal cellculture grade (D) water, respectively. C. Mean area:
145/<m2, [Ca 2+ ] = 5xl0~ 6 M; D, mean area: 275/<m2,
Fig. 4. Actin display in cells spreading on fibronectin
without added divalent cation. Cells were stained with
fluorescent phallotoxin after 25 min (A), or 60min (B), in
the absence of added divalent cation. With or without
added divalent cation, staining at the earlier time is
predominantly of membrane ruffles, whereas later a stress
fibre pattern predominates.
spreading does not require an external divalent cation,
or that any such requirement can be met by divalent
cations, presumably Ca 2 + , present as an impurity in
the nominally divalent cation-free medium. In the
latter case, the difference between serum and fibronectin would arise because, on serum, higher concentrations are required. By titration against Quin 2, we
found that the free Ca 2+ concentration in HS was
approximately 10~5 M. Some of this was contributed by
impurities in the HS solutes, but by using quartzredistilled water to prepare HS we were able to reduce
this concentration to about 5xl0~ 6 M, a concentration
that supported less spreading than normal HS, but did
not suppress it entirely (Fig. 5). It is well known that
chelators such as EGTA cause cell rounding, and we
found this to be true of cells on fibronectin as well as on
serum. By using EGTA-Ca mixtures as ion-buffers,
we found that spreading on fibronectin is almost
completely inhibited when external Ca 2+ is reduced
below 10~6M (Fig. 5). Since EGTA binds Ca 2+ much
more strongly than Mg2"1", it was possible to use EGTA
to show that Mg also activates spreading on fibronectin when free Ca 2+ is below 10~6M. (Fig. 6).
Adhesion and spreading on serum and fibronectin
661
n5
I
O
''
400
0
f-0-8
!o-6
200
« 0-4
S-0-
o.
c
K
0 10
6
10"' 10
10"
10-
(M)
Fig. 6. Induction of spreading by Mg 2+ in presence of
buffered low Ca 2 + . Cells spread for 90min on glass precoated with fibronectin (25 fig ml" 1 ) in the presence of
0-4 mM-EGTA/0-34 mM-Ca2+ (free Ca 2+ = 0-22X 10"6 M)
and indicated concentration of Mg 2+ - (O, • ) Separate
experiments.
Added fibronectin
Fig. 7. Cells spreading on low serum with added
fibronectin. Cells spread for 90min on glass pre-coated
with 0-5 % (v/v) calf serum containing additional
fibronectin. ( • ) With added divalent cation (10mM-Mg 2+ );
(O) HS, no divalent cation added.
Mixed films versus purified proteins
We considered the possibility that the difference in
divalent cation requirement between serum and fibronectin might result from our use of a purified protein
on the one hand, and a complex mixture with very high
total protein concentration on the other, rather than on
the identity of the specific proteins inducing spreading
in each case. To investigate this possibility, we used a
concentration of serum (0-5%) that was sufficient
completely to block spreading in HS on naked glass,
and then added increasing amounts of fibronectin to
this serum before using the mixture to coat glass. As
shown in Fig. 7, even in this mixed protein environment the fibronectin was able to induce spreading in
the absence of added divalent cations. It is also of
interest that such spreading was appreciable only when
extra fibronectin was added along with the serum. This
662
J. G. Edwards et al.
10"6 10-5
[Ca2+]
_J
10"
io- 2
(M)
Fig. 8. Comparison of Ca 2+ requirement for spreading on
fibronectin and (partially purified) vitronectin. The lowest
Ca 2+ was in EGTA/Ca 2+ buffer, free Ca 2+ =
0-22 X 10"6M; 10~5 data in HS without added divalent
cation; higher by addition. ( # ) Fibronectin, 25/tgml" 1 ;
(O) vitronectin-enriched serum fraction, 7/igml~'. After
coating with the active proteins, all coverslips were further
incubated with a blocking solution of haemoglobin
(500/igmr 1 ).
suggests that the fibronectin already present in serum
contributes little to the spreading of BHK cells, in
agreement with the finding of Knox (1984). We also
adopted the converse approach, of replacing serum
with partially purified vitronectin. Although this preparation was not purified to homogeneity, it was specifically depleted of fibronectin and, most importantly,
was some 50-fold enriched in spreading activity relative
to whole serum. It therefore allowed us to obtain
spreading comparable with that induced by fibronectin
in response to a similar (indeed lower) coating concentration. In view of the relatively low protein concentration used, all coverslips in these experiments were
incubated with haemoglobin (500ftgml~'), in order to
block access to the naked glass. There was no spreading
on haemoglobin alone (not shown). Fig. 8 shows that,
as with whole serum, cells did not spread on glass
coated in this way in background ( 1 0 ~ 5 M ) Ca 2 + , but
did respond to higher concentrations, so that the
displacement of the dose—response relative to fibronectin is preserved.
Discussion
Many investigators of divalent cation effects on adhesion and spreading of tissue cells have argued (from
the patterns of ion specificity, and the low concentrations of ions such as Mn 2 + that are sometimes
effective) that a specific divalent cation-binding site in a
protein (such as in a divalent cation-activated enzyme,
or an ion channel) is likely to be involved (Garvin,
1968; Rabinovitch & DeStefano, 1973a; Klebe et al.
1977; Stenn & Core, 1986). Whether this site is
intracellular or extracellular, and where it acts in the
Fig. 9. Alternative models to account for activation of cell spreading by divalent cations (M 2 + ). RGD denotes the ArgGly-Asp cell—recognition tnpeptide (in the one-letter code) in the matrix protein. For discussion see the text.
sequence of causes and effects have remained a matter
of conjecture.
We believe our results strongly indicate an external
site, perhaps in the receptor for the matrix protein. Our
argument is based on the assumption that, whatever
the mechanism by which different substrate-adsorbed
proteins induce active spreading, it is most likely that
their effects converge to activate the cellular locomotory apparatus through a common intracytoplasmic
pathway. Although there are many ways in which such
internal events could be affected by external divalent
cations (such as via intracompartmental ion fluxes, and
the activity of enzymes related to cytoskeletal rearrangement), effects on such distal events would not
be expected to differ markedly in divalent cation
requirement between different spreading-inducing
proteins.
One candidate for such an external divalent cationbinding site could be a Ca 2+ channel in the plasma
membrane, but our result with EGTA/Mg + shows
clearly that on fibronectin, Ca 2+ acts just as one
divalent cation among others, and not in its specific
message role. This argument in no way detracts from
the possibility that changes in internal Ca 2 + , via
release from internal stores, may be important in the
triggering of spreading, especially in view of the
observation by Kruskal et al. (1986) of a Ca 2+ transient
during the spreading of polymorphonuclear leucocytes.
It may be of some importance that spreading, presumably dependent on the operation of some components
of the cellular locomotory machinery, can be shown to
be independent of external Ca 2 + . Two other phenomena involving cell extension have also been shown to
continue when external Ca 2+ is reduced even to
nanomolar concentration by EGTA: a turning response of cultured chick neurites towards nerve growth
factor (Gundersen & Barrett, 1980), and acquisition
of polarity by neutrophils in suspension (Haston &
Shields, 1986). A contrary observation, by Cooper &
Schliwa (1986), is that the Ca 2+ channel antagonists
verapamil and nitrendipine inhibit lamellipod formation and cause rounding of fish keratinocytes. It seems
desirable to examine the response of spreading of these
cells to depletion of external Ca 2 + , since the channel
antagonists may possibly inhibit events other than
influx across the plasma membrane.
Some workers have reported observations that can be
interpreted as supporting an action of divalent cations
on spreading as opposed to adhesion, a view that might
seem contrary to a site of action at a receptor for matrix
glycoproteins. For example, Rabinovitch & DeStefano
(1973a) found that Mn 2 + stimulates spreading of
sarcoma I cells on clean glass, a surface to which they
adhered strongly in divalent cation-free conditions.
Such observations do not necessarily mean that the
divalent cation directly affects cytoskeletal, rather than
surface events. They may equally be accommodated by
a model in which a matrix-receptor, in the presence of
an appropriate divalent cation, triggers the locomotory
apparatus via an unknown transduction mechanism.
Attention is currently focused on the importance of
the tripeptide Arg-Gly-Asp in the interaction between
the cell-binding domains of proteins such as fibronectin
and serum spreading factor (vitronectin) with their
cellular receptors (see Ruoslahti & Pierschbacher,
1986, for a review). Three distinct ways in which
divalent cations could regulate this interaction are
shown schematically in Fig. 9. (1) The divalent cation
binds to the receptor and generates a conformation
favourable to binding of Arg-Gly-Asp at a second,
distinct site. (2) The divalent cation binds to the matrix
protein, again producing a conformation change, in
Adhesion and spreading on serum and
fibronectin
663
this case causing appropriate presentation of Arg-GlyAsp to the receptor. (3) The divalent cation itself forms
part of the Arg-Gly-Asp binding site, perhaps through
an interaction with the oxygen of the aspartic acid
residue.
Any of these models could account for the difference
in metal ion activating-concentrations we have observed between different inducing surfaces. For model
(1), BHK21 cells may have two different divalent
cation-activated receptors for the two proteins, fibronectin and vitronectin, as reported for another cell line,
by Pytela et al. (19856), with different Ca z+ affinities.
For BHK21 cells there is independent evidence pointing to the existence of a divalent cation binding site in a
fibronectin receptor, since both Oppenheimer-Marks &
Grinnell (1984) and Akiyama & Yamada (1985) have
reported that Ca 2+ protects the activity of this receptor
against proteolysis. This suggests a possible parallel
with the divalent cation-activated, divalent cationprotected cell-cell adhesion molecule uvomorulin
(Hyafil et al. 1981). Model (2) simply presupposes
different divalent-cation affinities for sites in the
spreading-inducing proteins themselves. Divalent
cation effects on Arg-Gly-Asp binding, via binding of
the cation directly to the matrix protein, have been
suggested as a possibility in the interaction of thrombospondin with platelets, in the light of the conserved
calcium-binding sequences found near the Arg-GlyAsp tripeptide in this protein (Lawler & Hynes, 1986).
Model (3) is reminiscent in some respects of the type of
interaction exploited in chelate chromatography of
proteins (Porath & Olin, 1983). It has the singular
advantage that one could view the metal ion binding at
the receptor site, rather than binding of Arg-Gly-Asp,
as the crucial event for activation of spreading. Thus
the effect of binding of the tripeptide would be to
increase the affinity of this site for an activating
divalent cation. Such a model could provide an explanation for the unexplained ability of Mn 2 + to induce
spreading in the absence of proteins containing the
Arg-Gly-Asp tripeptide (Grinnell, 1984; Stenn &
Core, 1986). It would be necessary to assume that
different proteins presenting Arg-Gly-Asp shift the
divalent cation affinity to different extents. In this
respect it is worth noting that the Arg-Gly-Asp-Ser
sequence in fibronectin offers an extra metal coordinating possibility in the serine oxygen, which is
lacking in the Arg-Gly-Asp-Val sequence of vitronectin
(Suzuki et al. 1985), a difference that could provide a
simple explanation for the difference in the calcium
activation that we have found.
An alternative point of action of the divalent cations,
which would equally accommodate our results, is that
there are ligand-specific events in the unknown transduction mechanism by which ligand-binding induces
spreading. Variant cells selected for defective response
664
J. G. Edwards et al.
to fibronectin, such as those we have isolated (Edwards
et al. 1985), should be useful in the analysis of the level
at which divalent cations act.
Regardless of the site of divalent cation action, our
results indicate that the concentrations of different
divalent cations required to activate spreading may
prove to be characteristic of the inducing protein even
when its environment of other adsorbed proteins
varies. If so, information on divalent cation activation
by different purified matrix proteins may in future be
useful for understanding which proteins are important
for spreading of various cells, when induced in vitro by
complex mixtures such as serum or extracellular
matrix.
The data on adhesion to serum-coated surfaces were
obtained as part of an earlier project supported by the Cancer
Research Campaign. Spreading on fibronectin in HS was
first noticed by Pauline Mann, then an Honours student in
Molecular Biology. Tony Lawrence encouraged us to explore
the implications of model (3). We thank Jim Bryce for help
with some experiments, Andrew Hart for excellent technical
support, and Chris Edwards for the area measurement
software.
References
AKIYAMA, S. K. & YAMADA, K. M. (1985). The interaction
of plasma fibronectin with fibroblastic cells in
suspension, jf. biol. Chem. 260, 4492-4500.
BARAK, L. S., YOKUM, R. R., NOTHNAGEL, E. A. & WEBB,
W. W. (1980). Fluorescence staining of the actin
cytoskeleton in living cells with 7-nitrobenz-2-oxa-l,3diazole-phallacidin. Proc. natn. Acad. Sci. U.SA. 77,
980-984.
BARNES, D. W. & SILNUTZER, J. (1983). Isolation of
human serum spreading factor, jf. biol. Client. 25,
12 548-12 552.
BREMEL, R. D. & WEBER, A. (1975). Calcium binding to
rabbit skeletal myosin under physiological conditions.
Biochim. biophys. Ada 376, 366-374.
COOPER, M. S. & SCHLIWA, M. (1986). Motility of
cultured fish epidermal cells in the presence and absence
of direct current electric fields. J. Cell Biol. 102,
1384-1399.
DAMSKY, C. H., KNUDSEN, K. A. & BUCK, C. A. (1984).
Integral membrane proteins in cell-cell and
cell-substratum adhesion. In The Biology of
Glycopmteins (ed. R. J. Ivatt), pp. 1-64. New York:
Plenum.
EDWARDS, J. G., CAMPBELL, J. A., ROBSON, R. T . &
VlCKER, M. G. (1975). Trypsm'zed BHK21 cells
aggregate in the presence of metabolic inhibitors and in
the absence of divalent cations. J'. Cell Sci. 19, 653-667.
EDWARDS, J. G., CAMPBELL, G. & SALAMA, N. (1985).
Variants of polyoma-transformed BHK21 cells
unresponsive to fibronectin. Cell Biol. int. Rep. 9,
737-745.
ENGVALL, E. & RUOSLAHTI, E. (1977). Binding of soluble
form of fibroblast surface protein, fibronectin, to
collagen. Inl.jf. Cancer 20, 1-5.
GARVIN, J. E. (1968). Effects of divalent cations on
adhesiveness of rat polymorphonuclear neutrophils in
vitro, jf. cell. Physiol. 72, 197-212.
GKINNELL, F. (1978). Cellular adhesiveness and
extracellular substrata. Int. Rev. Cytol. 53, 65-144.
GRINNELL, F. (1984). Manganese-dependent
cell-substratum adhesion. J. Cell Sci. 65, 61-72.
GUNDERSEN, R. W. & BARRETT, J. N. (1980).
Characterization of the turning response of dorsal root
neurites toward nerve growth factor. J. Cell Biol. 87,
546-554.
HASTON, W. S. & SHIELDS, J. M. (1986). Signal
transduction in human neutrophil leucocytes: effects of
external Na + and Ca 2+ on cell polarity, jf. Cell Sci. 82,
249-261.
HYAFIL, F., BABINET, C. & JACOB, F. (1981). Cell-cell
interactions in early embryogenesis: a molecular
approach to the role of calcium. Cell 26, 447-454.
KLEBE, R. J., HALL, J. R., ROSENBERGER, P. & DICKEY,
W. D. (1977). Cell attachment to collagen: the ionic
requirements. Expl Cell Res. 110, 419-425.
KNOX, P. (1984). Kinetics of cell spreading in the presence
of different concentrations of serum or fibronectindepleted serum. J. Cell Sci. 71, 51-60.
KRUSKAL, B. A., SHAK, S. & MAXFIELD, F. R. (1986).
Spreading of human neutrophils is immediately preceded
by a large increase in cytoplasmic free calcium. Proc.
natn. Acacl. Sci. U.SA. 83, 2919-2923.
LAWLER, J. & HYNES, R. O. (1986). The structure of
human thrombospondin, an adhesive glycoprotein with
multiple calcium-binding sites and homologies with
several different proteins. J . Cell Biol. 103, 1635-1648.
OPPENHEIMER-MARKS, N. & GRINNELL, F. (1984). Calcium
ions protect cell-substratum adhesion receptors against
proteolysis. Expl Cell Res. 152, 467-475.
PORATH, J. & OLIN, B. (1983). Immobilized metal ion
affinity adsorption and immobilized metal ion affinity
chromatography of biomaterials. Serum protein affinities
for gel-immobilized iron and nickel ions. Biochemistry
22, 1621-1630.
PVTELA, R . , PlERSCHBACHER, M . D . & RUOSLAHTI, E .
(1985fl). Identification and isolation of a 140 kd cellsurface glycoprotein with properties expected of a
fibronectin receptor. Cell 40, 191-198.
PYTELA, R . , PlERSCHBACHER, M . D . & RUOSLAHTI, E .
(19856). A 125/115-kDa cell surface receptor for
vitronectin interacts with the arginine-glycine-aspartic
acid adhesion sequence derived from fibronectin. Proc.
natn. Acad. Sci. i'.SA. 82, 5766-5770.
RABINOVITCH, M. & DESTEFANO, M. J. (1973a).
Manganese stimulates adhesion and spreading of mouse
sarcoma I ascites cells. J . Cell Biol. 59, 165-176.
RABINOVITCH, M. & DESTEFANO, M. J. (19736).
Macrophage spreading in vitro. I. lnducers of spreading.
Expl Cell Res. 77, 323-334.
RABINOVITCH, M. & DESTEFANO, M. J. (1973r).
Macrophage spreading;;/ vitro. II. Manganese and other
metals as inducers or as co-factors for induced spreading.
Expl Cell Res. 79, 423-430.
RUOSLAHTI, E., HAYMAN, E. G. & PIERSCHBACHER, M. D.
(1985). Extracellular matrices and cell adhesion.
Arteriosclemsis 5, 581-594.
RUOSLAHTI, E. & PIERSCHBACHER, M. D. (1986). Arg-Gly-
Asp: a versatile cell recognition signal. Cell 35, 421-431.
RUSSELL, W. C , NEWMAN, C. & WILLIAMSON, D. H.
(1975). A simple cytochemical technique for
demonstration of DNA in cells infected with
mycoplasmas and viruses. Xatnre, Land. 253, 461-462.
STENN, K. S. & CORE, N. G. (1986). Cation dependence
of guinea pig epidermal cell spreading. In vitm Cell devl
Biol. 22, 217-222.
STOKER, M. & MACPHERSON, I. (1964). Syrian hamster
fibroblast cell line BHK21 and its derivatives. Xatnre,
Lond. 203, 1355-1357.
SUZUKI, S., OLDBERG, A., HAYMAN, E. G.,
PIERSCHBACHER, M. D. & RUOSLAHTI, E. (1985).
Complete amino acid sequence of human vitronectin
deduced from cDNA. Similarity of attachment sites in
vitronectin and fibronectin. EMBOJ. 4, 2519-2524.
TAKEICHI, M. & OKADA, T . S. (1972). Roles of
magnesium and calcium ions in cell-to-substratum
adhesion. Expl Cell Res. 74, 51-60.
TSIEN, R. Y. (1980). New calcium indicators and buffers
with high selectivity against magnesium and protons:
design, synthesis and properties of prototype structures.
Biochemistry 19, 2396-2404.
VASILIEV, J. M. (1985). Spreading of non-transformed and
transformed cells. Biochim. biophys. Ada 780, 21-65.
YAMADA, K. M. (1983). Cell surface interactions with
extracellular materials. A. Rev. Biochem. 52, 761-799.
(Received 23 February 19S7 -Accepted 25 March 19S7)
Adhesion and spreading on serum and fibronectin 665