/ . Embryol. exp. Morph. Vol. 39, pp. 169-182, 1977
Printed in Great Britain
169
Control of anteroposterior pattern in the axolotl
forelimb by a smoothly graded signal
By J. M. W. SLACK 1
From the Department of Biology as Applied to Medicine,
The Middlesex Hospital Medical School, London
SUMMARY
(1) It is shown that the number of cartilage elements in an experimentally produced
reduplicated limb depends on the width of competent tissue between pieces of flank tissue.
(2) Seventy well formed reduplications were examined on the assumption that the difference
in the number of elements between them results from small differences in graft position.
(3) All the reduplications are symmetrical along their entire length.
(4) All possess the most posterior structures at both edges with other elements arranged in
between in the correct serial order.
(5) The existence of vestigial and of branched elements in the midline suggest that the
cases examined come from a potentially continuous series.
(6) All these characteristics can readily be explained by assuming the existence of a
continuous U-shaped anteroposterior gradient, and of threshold responses by the cells which
together control the pattern of cell differentiation.
(7) It is suggested that the gradient 'deepens' as more distal levels of the limb are laid down.
(8) Most but not all of the cases can be arranged in a single series in which proximal and
distal levels expand coordinately.
(9) Possible mechanisms for the establishment and maintenance of the gradient are
discussed.
In the accompanying paper (Slack, 1977), it is suggested that the normal
anteroposterior pattern of the amphibian limb arises as a consequence of an
interaction between the prospective limb and the flank. It has been shown that
that part of the limb rudiment which is adjacent to flank tissue in the embryo
becomes a posterior edge. So far we have considered neither the mechanism of
this 'posterior determination', nor the mechanism of limb individuation, that
is the process whereby cells in different anteroposterior positions differentiate
to form a characteristic pattern of muscles and cartilages.
In principle, the individuation process could occur by various mechanisms.
There might be a sequential induction of territories across the limb-bud, of
which the first was induced by the flank tissue. There might be a concentration
gradient of a morphogen established across the bud, with threshold responses
by the cells to divide the field into different territories (Crick, 1970; Wolpert,
1
Author's address: Imperial Cancer Research Fund, Mill Hill Laboratory, Burtonhole
Lane, London NW7 IAD, U.K.
170
d!2
n
cl
c2
r
II P2
II PI
II MC
d3
d4
Fig. 1. Nomenclature of cartilage elements in the normal axolotl forelimb.
H, humerus; R, radius; U, ulna; r, radiale; i, intermedium; u, ulnare; cl, c2,
centralia; dl2, d3, d4, distal carpals; I-IV, digits; MC, metacarpals; PI, P2, P3,
1 st, 2nd, 3rd phalanges.
1971; Gierer & Meinhardt, 1972; Lewis, Slack & Wolpert, 1977). There might
be a chemical wave front which passes slowly across the field and in conjunction
with an intracellular oscillator divides the field into territories (Cooke & Zeeman,
1976). There might be a complex chemical signal having several concentration
peaks, each one of which determined the same cell type, for example cartilage
(Turing, 1952; Wilby & Ede, 1975). Whichever mechanism in fact operates, if
we knew which it was we would also know the nature of the 'posterior determination' arising from contact between limb rudiment and flank tissue.
In this paper a comparative anatomical study of a large number of double
posterior reduplications is presented. It is argued that the various features of
this series are all consistent with the hypothesis that the morphogenetic signal
which underlies the reduplications is a U-shaped gradient, while they are
difficult to explain on the basis of other models for the individuation process.
By extension, it is proposed that in normal development the anteroposterior
pattern arises in response to a single monotonic gradient across the limb-bud.
The relationship between this gradient and the posterior determination in the
limb rudiment is discussed, as is the comparison with what is known of limb
development in other classes of vertebrate.
MATERIALS AND METHODS
Many of the reduplicated limbs examined in this paper were produced in
experiments described in the other paper of the series. Seventy specimens were
selected for examination on the basis of (i) clear staining of all cartilages (ii)
absence of 'spurs' i.e. extra digits or cartilages sticking out in funny places (iii)
Pattern formation in axolotl forelimb
171
Table 1
Posterior limb
A
Level
Cases
4
4/3
3
3/2
2
6
6
8
41
6
Suplimb (all
normal) pressed
6
2
2
0
0
6
2
2
0
0
Reduplicated Normal
0
4
6
28
0
0
0
0
13
6
Mean D
—
15
17
32
—
absence of dorsoventral reduplication of parts of the limb. In other respects
they are fully representative of all the reduplications obtained.
All the cases were examined as whole mounts stained in alcian green to reveal
the pattern of cartilage elements. The nomenclature of cartilages follows
Romer (1962) and their disposition in the normal limb is shown in Fig. 1.
The series letters refer to types of experiment: C, E, H, J, O, U, P, N are all
grafts of flank strips to the anterior margin of the limb rudiment, A are a.p.
inversions of the limb rudiment, K and L are implantations of prospective limb
strips into the flank and R are dorsoventral inversions of the limb rudiment,
five cases of which surprisingly yielded perfect double posterior reduplications.
These five may be examples of rotational regulation described by Nicholas
(1924).
RESULTS AND DISCUSSION
Position of grafts and the gradient model
The whole of the argument which follows is based on the premise that the
different individual cases of reduplication arise from the same mechanism, and
that differences in the number of cartilage elements are consequent on small
differences in the position of the graft. The experimental evidence for this comes
from implantation of flank strips at different levels in the host, presented in the
previous paper. Let us now look at the character of the resulting limbs in this
experiment: remembering that there can be anterior and posterior limbs formed
if the rudiment is divided (Table 1). The symbol D represents the total number
of cartilage elements in a reduplicated limb.
With the graft in its most posterior position (level 4) the posterior limb is
always suppressed, that is it appears only as a bud or small knob and forms no
limb structures. When the graft is placed a little further anterior (level 4/3 or
3), the posterior limb is reduplicated with a low D value. When it is further
anterior (level 3/2) reduplications are obtained with a high D value. At the most
anterior level (level 2) the grafts no longer induce reduplications, probably
because they are no longer in contact with the prospective limb tissue.
The important conclusion from this experiment is that small differences in
172
J. M. W. SLACK
A
* c2
11 Thresholds
P
B
„ ..
Position
A
C
Fig. 2. (A) Model for the specification of a particular element in the normal limb.
(B) suppression of the same element in a narrow reduplication. (C) representation by
a double posterior fusion in the midline. (D) representation by two copies in a
relation of mirror symmetry in a wide reduplication.
graft position, for example from level 3 to 3/2, can make a big difference to the
number of cartilage elements in the reduplication.
For clarity, I shall invert the usual order of a scientific paper and state the
conclusion at this stage. Suppose that the flank defines where the high points
of an anteroposterior gradient will be thereby determining the overall polarity.
The gradient then determines the pattern. In normal development, each position
has a different value and a given element will appear between positions corresponding to its upper and lower response thresholds (Fig. 2). In an experimental case where flank tissue is present on both sides of the limb rudiment
the gradient will be U-shaped. We may assume that the further apart the ends
are the deeper is the centre of the U. So for a particular element Fig. 2 shows
that a shallow gradient will suppress it, a deeper one will evoke it as a double
posterior structure in the centre of the limb, and a still deeper one will evoke it
in two copies. For the whole limb we may see that a very shallow gradient will
suppress all the elements, a medium gradient will suppress some anterior
elements, and a very deep gradient will evoke a complete double limb.
This model accounts for all features of the experiment in Table 1. We may
expect of all reduplications produced in this way that they possess the most
posterior elements and always preserve the correct serial order of elements (a
'correct serial order' is one in which an element has only the neighbour(s)
which it would have in the normal pattern). They may however contain more
Pattern formation in axolotl forelimb
173
Fig. 3. Whole mounts of reduplications showing variation in the total number of
elements. (A) case P7 x 50. (B) U6 x 25. (C) U15 x 40. (D) E26 x 32 (E) E36 x 32.
(F) J21 x 40.
(The magnifications refer to the 35 mm negatives of the photographs.)
or less elements in total depending on the depth of the central part of the
gradient. We may expect something else if the signal for pattern formation is
a real chemical gradient, and the model is more than a high level abstraction.
This is the presence of intermediate forms, in which elements are found in
vestigial forms in the centre of the limb, or where they are caught in the act of
dividing into two.
Serial order and continuity
In Fig. 3 are shown whole mounts of six of the most commonly occurring
types of reduplication, and they illustrate clearly the features to be expected
12
•
EMB 39
174
J. M. W. SLACK
Table 2
Ranking of Reduplications
Description*
D
Cases
All elements dupl.
Branched H
H fused
R & r fused
cl fused
IMC branched
IMC fused
1P1 fused
IP2 fused
r, cl lost
R half lost
R vestigial
Rlost
dl2 fused, i fused, IMC & IP
vestigia]
c2 fusing IMC lost
IP lost
c2 fused
48
47*
47
45
N4, A49
1/2, A51
E44, J21
C20, Ull, A45, E36, J23
II MC branched
11 MC fused
11 PI fused, II P2 branched
II P2 fused
11 MC half lost
Fusion of d3 and d4 on both sides
II MC lost
11 PI half lost
II PI and II P2 lost, III PI fused
U fusing
U fused, III MC, III P2, III P3 fused
3 carpals lost
'Thinner' limb structures have been
* Nomenclature given in Fig. 1.
311
43*
43
42
41
39
38*
38
38
35
33*
32*
32
31
29*
29
28*
26*
26
25*
22*
19
16
E40, 2/1
E31
U4
O22, N13, 2/4
Cl
J16
E26
C31
E35
U20, J30, C37
O7
C7, C29, C32, C33, C35, H35
H8, El7
C33, U15
E38, 2/3
N26
J69, A39, U6, H6, C9, C27, 1/5
A52
C21
U14
C36, C38
O18
U5, U9, U12 P7, C34, E20
P10
obtained but they are hard to interpret.
from the gradient model. All are reduplicated along their entire length; there is
no point on the proximodistal axis at which the reduplication commences.
Close inspection of central elements such as the humerus, the ulna of P7, or
the radius of E36 shows that they are not normal but consist of two posterior
halves. This may also be proved experimentally because reduplicated limbs
amputated through mid-humerus usually regenerate reduplicates and not normal
limbs as has been incorrectly reported in the past (Takaya, 1941). (These
experiments will be described in detail elsewhere.)
Although the total number of elements is very variable, the D values being
respectively 19, 29, 32, 38, 45 and 47, the most posterior elements are always
present and the remainder are arranged in the correct serial order.
Fig. 4. Whole mounts with vestigial elements. (A) C38 x 40. Distal half of IIPl floats freely in soft tissue, ulna is in a condition of incipient
division. (B) E35 x 32. Vestigial traces of IMC and IP. (C) J16 x 40. Distal portion of radius.
(Magnifications refer to the 35 mm negatives of the published photographs.)
B
I*
X
3"
176
J. M. W. SLACK
Since each specimen results from a separate graft, there is no question of
actual transformation between types except in the case of regeneration. But we
can ask whether the range of possible structures is, in principle, continuous or
discrete. If it is discrete then there would be a finite number of types differing
by at least one element at a time. If it is continuous then it should be possible
to find cases in the available sample which shade into one another.
Among the 70 cases, there were two sub series of six which strongly suggest
that the possible range is continuous. These are 018, C38, U14, C21, A52 &
J69 and U20, E35, C31, E26, J16 & Cl (see Table 2 and Fig. 4). In the first sub
series the central ulna is in a state of incipient division in C38 and 11 phalanges
of digit II appear distally floating in the midst of soft tissue. In U14 the remainder of phalange IIP1 has appeared and the ulna is fully divided. In A52
the metacarpal of digit II is partially visible and in Cl we reach a common
type illustrated in Fig. 3B. In the second sub-series, the carpal C2 is in a state
of incipient division in U20, vestigial portions of digit I appear in a floating
distal position in U20 and E35. The radius appears as a vestigial trace at its
distal end in E26 and then more and more of it is present in J16 and Cl.
Continuity is also suggested by the occasional observation of branched
elements. A branched humerus was found in cases A51, 1/3 and the pleurodele
shown in Fig. 5 A. A branched metacarpal IMC in E40 and 2/1, IIMC in C23
and U15 and IIP2 in N26 were also observed. (All these are central elements
and the branching proceeds distally.)
There are a few cases of proximal branching, all of IIIMC (Fig. 5C). But
none of these come from limbs which are strictly symmetrical, the one illustrated
may for example really be a four-fingered reduplicate which was disrupted
during digit formation. Because they are not symmetrical they were not included
in the series of 70 under analysis.
Distal deepening and the ranking principle
In many respects the reduplications seem to represent a 'deeper' U-shaped
gradient at their distal than at their proximal ends. For example, all 8 digits are
fully formed in cases where the humerus is not even branched (Fig. 3E, F);
branching between cartilage elements is common at all levels and always
proceeds distally (e.g. Fig. 5B). We have seen in the previous paragraph that
digits I and II and the radius can be found as distal but not as proximal
vestigial traces.
Of course we do not know whether this is a property of the gradient or of the
thresholds. But since continuous and coordinated changes of thresholds both
between and within elements represents a vastly more complex situation, I am
inclined to favour the view that the gradient becomes progressively deeper
distally. This is already postulated to occur for the polarizing region of the
chick limb-bud (Tickle, Summerbill & Wolpert, 1975), and is quite natural for
Fig. 5. Whole mounts with branched structures. (A) M2 x 32. Branched humerus. (B) E38 x 32. Branching of
digit II between elements. (C) O20 x 40. Proximal branching of IIIMC.
(Magnifications refer to the 35 mm negatives of the photographs.)
B
2
5s
B
Fig. 6. Whole mounts showing exceptions to the ranking principle. (A) Rl x 40. (B) L64 x 40. (C) O22R x 32.
(Magnifications'refer to the 35 mm negatives of the phctographs.)
A
r
oo
Pattern formation in axclotl forelimb
179
Proximal
Distal
Level of
gradient
Post. •*•
•*- A n t .
Fig. 7. Diagrammatic representation of the nesting of gradients suggested by the
ranking principle, (a) Gradient for narrow reduplication, (b) gradient for wide
reduplication.
diffusion or reaction-diffusion systems if the tip becomes wider as the more
distal primordia are laid down.
It is of some interest to know whether the whole sample of 70 reduplicates
can be arranged in a single series according to the value of D, such that every
proximodistal level becomes expanded in concert. If so then a limb with six
digits will always have two bones in the lower arm, not one or three. Another
way of putting this is to say that the entire structure of the reduplicate could be
denned by a single parameter: D. It turns out that there are some exceptions to
this ranking principle, but that all of them belong to the experimental series L,
K and R which differ rather from the others as described in Materials and
Methods. If these series are left out, the remaining 60 reduplicates can be
ranked as shown in Table 2.
Two of the exceptions are shown in Fig. 6 A, B. RI is anomalous because it
has a complete division of the ulna but only seven carpals and three digits;
L64 is anomalous because it has a radius but the intermedium is undivided and
digit I vestigial. In Fig. 6C is shown an artificially created exception. Case 022
which had D = 41 was amputated through the wrist and regenerated a * narrower'
reduplicated hand on a 'wider' stump. This is typical but not invariable
behaviour during regeneration.
Before the exceptions were discovered, it seemed that the ranking principle
might reflect a heritable positional value, recorded in the limb cells in the tail-
180
J. M. W. SLACK
bud stage and subsequently interpreted using different rules as different
proximodistal levels were laid down. But this cannot explain the exceptions,
and it is also not consistent with the hypothesis of a progressive deepening of the
gradient. Now it seems preferable to assume that the ranking principle reflects
a 'nesting' of gradients, as depicted in Fig. 7. In series L and R where the normal
growth pattern of the bud may have been altered, the gradient surfaces would
cross over those of the ranked set.
Possible mechanisms
The existence of a U-shaped gradient underlying the experimental production
of reduplications was suggested for the chick limb-bud by Tickle et ah (1975).
Together with the evidence presented here such a mechanism is now strongly
supported. It is a small step to assume that a monotonic gradient controls the
formation of pattern during normal development as shown in Fig. 2. In amniote
limb-buds there is supposed to exist a zone of polarizing activity, ZPA (Saunders
& Gasseling, 1968; Saunders, 1972; McCabe & Parker, 1976; Tickle, Shellswell,
Crawley & Wolpert, 1976) which is assumed by Tickle et ah to be the source of
the gradient.
The question now becomes: what is the relationship between the limb-flank
interaction described in the previous paper and the anteroposterior gradient?
There seem to be three possibilities:
(i) The active zone of the flank is the rudiment for a ZPA. At a certain stage
of development it begins to secrete the morphogen, which is destroyed by the
cells of the limb rudiment. Such processes will lead to the formation of an
exponential concentration gradient of the morphogen across the limb rudiment.
(ii) The flank induces a ZPA out of the limb rudiment. The ZPA cells subsequently differ from the rest of the limb-bud and are specialized to secrete the
morphogen.
(iii) There are no specialized ZPA cells but the limb rudiment is capable of
self organization of a chemical concentration gradient or ' dissipative structure'
by means of an autocatalytic reaction-diffusion system (Prigogine & Nicholis,
1971; Gierer & Meinhardt, 1972). The flank produces no specific signal but
because it provides a slightly different environment from the other surrounding
tissues the highest concentration of the morphogenetically active chemical
species is always found posteriorly.
None of these possibilities for the origin of the gradient can be categorically
excluded at the moment, either for the amphibia or the amniote species. The
first is less likely because it requires that a part of the flank becomes incorporated
into the posterior edge of the limb-bud. But the limb-bud when it appears is a
very small condensation of cells at level 4, somewhat anterior to the histological
limb-flank boundary. Also the experiments in which reduplications were induced
with marked grafts seem to show that the graft need not contribute to the tissue
of the bud.
Pattern formation in axolotl forelimb
181
The best evidence for a differentiated ZPA would be the discovery of some
histological or biochemical marker for the cells, even if it had no connexion
with their morphogenetic activity. This has not so far been found even in the
carefully studied chick limb-bud.
There seem to me to be two arguments in favour of the dissipative structure,
(possibility iii), although neither is conclusive. One characteristic of such
chemical systems is that they exhibit a number of stable spatial patterns and
that the same pattern can be reached from a number of starting points. When
one does experiments on the amphibian limb it is very striking that most
manipulations lead to normal or double posterior limbs and that despite the
vagaries of grafting experiments, partial, intermediate or chaotic structures are
very rare. The fact that neighbouring configurations regulate towards the
same final states is graphically illustrated by experiments of Nicholas (1924)
on the rotation of limb discs. Rotations of less than 90 ° or more than 270 °
gave buds which reverted to normal orientation in the course of their growth,
whereas rotations of 135 ° or 235 ° gave buds which grew towards the typical
reduplication arising from the 180 ° rotation.
The second argument arises from the fact that the ZPA of the chick limb-bud
can be removed without affecting the pattern formation in the remaining
anterior half bud (Fallon & Crosby, 1975), also a ZPA grafted anteriorly and
removed after 10 h still produces a reduplication (Tickle et ah 1975). A dissipative structure could account for this signal memory combined with the
progressive distal deepening of the gradient, while a model with a differentiated
ZPA could not.
Probably the best positive test for a dissipative structure would be one
relying on the fact that each 'mode' has a larger minimum size for stability
than the previous mode (Hiernaux, personal comm.). In other words there will
be a minimum size for the formation of a U-shaped gradient which is larger
than the minimum size for the normal monotonic gradient. So if increasingly
narrow strips of limb rudiment were grafted into the mid flank, there might be
a critical width below which they no longer yielded reduplications but instead
normal limbs or portions of normal limbs.
This work was supported by the Medical Research Council.
REFERENCES
J. & ZEEMAN, E. (1976). A clock and wavefront model for control of the number of
repeated structures during animal morphogenesis. J. theor. Biol. 58, 455-476.
CRICK, F. H. C. (1970). Diffusion in embryogenesis. Nature, Lond. 225, 420-422.
FALLON, J. F. & CROSBY, G. M. (1975). Normal development of the chick wing following
removal of the polarizing zone. /. exp. Zool. 193, 449-455.
GIERER, A. & MEINHARDT, H. (1972). A theory of biological pattern formation. Kybernetik
12, 30-39.
LEWIS, J., SLACK, J. & WOLPERT, L. (1977). Thresholds and precision in development.
J. theor. Biol. (in press.)
COOKE,
182
J. M. W. SLACK
J. A. & PARKER, B. W. (1976). Polarizing activity in the developing limb of the
Syrian hamster. /. exp. Zool. 195, 311-317.
NICHOLAS, J. S. (1924). Regulation of posture in the forelimb of Ambystoma punctatum.
J. exp. Zool. 40, 113-159.
PRIGOGINE, I. & NICHOLIS, G. (1971). Biological order, structure and instabilities. Q. Rev.
Biophys. 4, 107-148.
ROMER, A. S. (1962). The Vertebrate Body. 3rd ed. Philadelphia: W. B. Saunders.
SAUNDERS, J. W. (1972). Developmental control of three dimensional polarity in the avian
limb. Ann. N.Y. Acad. Sci. 193, 29-41.
SAUNDERS, J. W. & GASSELING, M. T. (1968). Ectodermal-mesenchymal interactions in the
origin of limb symmetry. In Epithelial-Mesenchymal Interactions (ed. R. Fleischmajer &
R. E. Billingham), pp. 78-98. Baltimore: Williams & Wilkins.
SLACK, J. M. W. (1977). Determination of anteroposterior polarity in the axolotl forelimb
by an interaction between limb and flank rudiments. /. Embryol. exp. Morph. 39 151-168.
TAKAYA, H. (1941). Experimental study on limb asymmetry. Annotnes. zool. jap. 20, (pt. 4
suppl.), 181-279.
TICKLE, C, SUMMERBELL, D. & WOLPERT, L. (1975). Positional signalling and specification
of digits in chick limb morphogenesis. Nature, Lond. 254, 199-202.
TICKLE, C, SHELLSWELL, G., CRAWLEY, A. & WOLPERT, L. (1976). Positional signalling by
mouse limb polarizing region in the chick wing bud. Nature, Lond. 259, 396-397.
TURING, A. M. (1952). The chemical basis of morphogenesis. Phil. Trans. R. Soc. Ser. B 237,
37-72.
WOLPERT, L. (1971). Positional information and pattern formation. Curr. Top. Devi Biol. 6,
183-224.
WILBY, O. K. & EDE, D. A. (1975). A model generating the pattern of cartilage skeletal
elements in the embryonic chick limb. /. theor. Biol. 52, 199-217.
MCCABE,
{Received 10 September 1976, revised 24 January 1977)