How the growth and patterning of developing tissues

are controlled and coordinated has been a long-standing
question in developmental biology. Morphogens are
secreted signalling molecules that play a part in these
processes by providing positional information to the cells
in the tissue and by acting as a trigger for tissue growth
(reviewed in REFS 1–3). Different morphogen concentra-
tions induce distinct target gene responses; therefore, in
a concentration gradient, the cellular response is posi-
tion dependent, resulting in tissue patterning. How
morphogen s regulate tissue growth is less clear.
An excellent model system for the study of morpho-
genetic growth regulation is the wing imaginal disc of
Drosophila melanogaster, in which tissue growth is
controlled by the morphogen Decapentaplegic (DPP)
4

(FIG. 1a,b). DPP is a bone morphogenetic protein (BMP)
homologue and triggers the activation of known wing
patterning genes in a concentration-dependent man-
ner
5–9
(FIG. 1b). Binding of DPP to its receptor Thickveins
(TKV) leads to the phosphorylation and activation of the
transcription factor MAD
10,11
. Phosphorylated MAD in
turn represses the expression of Brinker (BRK), which is
itself a transcriptional repressor
12–15
. Together, phospho-
rylated MAD and BRK regulate the expression of DPP
target genes
16
(FIG. 1c).
Although it is clear that DPP regulates growth
4
, for
a long time no direct growth-regulatory targets of DPP
were known. However, recently it was found that phos-
phorylated MAD interacts with the co-transcriptional
activator Yorkie to regulate the transcription of the
growth-promoting microRNA (miRNA) bantam (FIG. 1c).
Even so, absolute DPP signalling levels cannot directly
control proliferation; this is because in the disc cells are
exposed to different absolute DPP levels but proliferate
at the same rate
17–19
(BOX 1; FIG. 1b). Several models for
growth control by DPP have been proposed to address
this issue
4
.
These morphogenetic growth models fall into two
groups. In the first, DPP is permissive for growth: DPP
levels above a certain threshold promote growth in the
central region of the disc, whereas growth in the peri-
phery, where DPP levels are below the threshold, is regu-
lated by mechanical stress
20–22
or other growth factors
17,18
.
In the second, DPP is instructive for growth: to set their
growth rate or cell cycle length, cells are proposed to
measure a position-independent DPP gradient prop-
erty, such as spatial differences in DPP signalling levels
between neighbouring cells
23,24
or the increase in cellular
DPP signalling levels over time
24
.
Recent studies have provided new data on DPP gradi-
ent and wing disc growth properties and the relationship
between them
24
. Here, we review these data and discuss
how far they support the proposed morphogenetic
growth models.
Wing disc growth and the DPP gradient
The growth properties of the wing imaginal disc have
been well characterized (BOX 1). During the larval growth
period, cells undergo approximately ten divisions. Cells
initially proliferate quickly, but the proliferation rate
decreases as development proceeds. At any one time dur-
ing development, all of the cells, no matter where they
*Departments of
Biochemistry and
Molecular Cell Biology,
Faculty of Sciences,
University of Geneva,
30 Quai Ernest Ansermet,
1211 Geneva, Switzerland.

Max Planck Institute for the
Physics of Complex Systems,
Nöthnitzer Str. 38,
01187 Dresden, Germany.
Correspondence to M.G.‑G.
e‑mail: marcos.gonzalez@
unige.ch
doi:10.1038/nrm3169
Published online
18 August 2011
Imaginal disc
A flat epithelial pouch in
the larva that, during
metamorphosis in the pupa,
will give rise to an adult
structure of the fly.
Understanding morphogenetic
growth control — lessons from flies
*Ortrud Wartlick,

Peer Mumcu,

Frank Jülicher and *Marcos Gonzalez‑Gaitan
Abstract | Morphogens are secreted signalling molecules that control the patterning and
growth of developing organs. How morphogens regulate patterning is fairly well
understood; however, how they control growth is less clear. Four principal models have
been proposed to explain how the morphogenetic protein Decapentaplegic (DPP) controls
the growth of the wing imaginal disc in the fly. Recent studies in this model system have
provided a wealth of experimental data on growth and DPP gradient properties, as well as
on the interactions of DPP with other signalling pathways. These findings have allowed a
more precise formulation and evaluation of morphogenetic growth models. The insights
into growth control by the DPP gradient will also be useful for understanding other
morphogenetic growth systems.
REVI EWS
594 | SEPTEMBER 2011 | VOLUME 12 www.nature.com/reviews/molcellbio
REVI EWS
© 2011 Macmillan Publishers Limited. All rights reserved
0CVWTG4GXKGYU¦/QNGEWNCT%GNN$KQNQI[
Z
%
0
Disìunco lrom ìbo sourco λ
%
0
/o
.
D C
C
o
n
c
o
n
ì
r
u
ì
i
o
n
DPP E
TKV
MAD
MAD
MLDLA
P
MAD
MLDLA
P
MAD
Yorlio
Yorlio
P
DAD
8PK
8PK
DTM
&CFURCNV
DCPVCO
Nuclous
Cyìoµlusm
Coll mombruno
Anìorior Posìorior
Winq
Z
.
S
o
u
r
c
o
Amplitude
The maximum concentration of
a protein in the target region.
In the case of Decapentaplegic
(DPP), amplitude refers to the
concentration at the DPP
source boundary. Its value
depends on DPP production
and degradation rates,
as well as on the number
of DPP-producing cells
(the source width) and on
the rate of diffusion.
Decay length
A measure for the spatial range
of a protein gradient (how far
it reaches into the tissue):
the position at which the
concentration of the protein is
a fixed fraction of its amplitude
(C
0
/e; in which e is Euler’s
number). Its value depends on
how fast the molecules diffuse
and are degraded (less
degradation equals a higher
decay length).
Gradient scaling
If the gradient expands at
the same rate as the tissue,
it scales with tissue size; the
decay length of the gradient
is proportional to the width
of the tissue.
are in the tissue, divide at roughly the same rate
17,18,24,25
.
In other words, all cells contribute approximately equally
to the tissue: tissue proportions scale. This means that
the position of cell clones relative to the width of the tis-
sue remains approximately constant during development
(FIG. 2a). Similar observations have been made in other
morphogenetic growth systems, such as the vertebrate
limbs or the embryonic spinal chord (BOX 1). Below, we
discuss how the DPP concentration gradient changes
during wing disc growth.
DPP gradient scaling during wing disc growth. DPP is
expressed in a central stripe of cells (the DPP source)
and spreads in the tissue, forming a concentration gradi-
ent
6,7,26
(FIG. 1a). The DPP gradient can be characterized
by its amplitude (C
0
) and its decay length (λ)
27
(FIG. 1b).
Quantification of the amplitude and decay length of
the DPP gradient in the posterior compartment of the
wing disc showed that the DPP gradient expands during
disc growth
24
(FIG. 2a,b). In particular, the decay length
is proportional to the width of the compartment — the
gradient scales (FIG. 2c,d). Gradient scaling means that, as
the tissue grows, the range of the gradient grows propor-
tionately. Therefore, when the relative gradient profiles
are normalized to tissue size, gradients from all stages of
development have the same shape
24
(FIG. 2c).
An important implication of this invariant rela-
tive concentration profile is that, during development,
a particular relative position in the tissue always has
the same relative concentration of the morphogen
24

(C/ C
0
= constant; FIG. 2c). Because homogeneous growth
means that cells (and their lineages) do not change their
relative position during growth (FIG. 2a), gradient scaling
implies that a particular cell always experiences the
same relative DPP concentration as the tissue grows
24

(C
cell
/C
0
= constan t). This means that, for the observed
increase in the gradient amplitude in the growing tissue
(FIG. 2b,e), there is a corresponding increase of cellular
DPP concentration: C
cell
and C
0
are proportional in all
cells. In other words, the DPP signalling levels of all cells
in the tissue increase over time by the same percentage
as the gradient amplitude (FIG. 2e).
Gradient scaling is not only observed for the
DPP ligand concentration gradient, but also for the DPP
signalling gradient
24
. The transduction of the signal
involves successive nonlinear amplifications
24
. This means
that the spatial activity pattern of the pathway compo-
nents — for example, phosphorylated MAD, BRK and
DAD — is different to that of DPP ligand concentration
24
.
However, gradient scaling and the increase of the ampli-
tude are also observed for a downstream DPP signalling
‘readout’ (Dad tanscription)
24
. This is not trivial, and it will
be interesting to understand how the wiring of the signal
transduction network mediates the conservatio n of these
two properties at the end of the pathway.
Controlling the DPP gradient during wing disc growth.
Disc growth does not distort the DPP gradient because
the gradient renews itself faster than the tissue grows
24
.
Gradient scaling is therefore not a direct consequence
of growth. Instead, scaling is mainly due to a decrease in
the DPP degradation rate — empirically, when the num-
ber of cells in the tissue doubles, the DPP degradation
rate halves
24
. Both the increase of the gradient amplitude
C
0
and the scaling of the decay length λ can be explained
as a consequence of this decrease of the DPP degradation
Figure 1 | The Decapentaplegic gradient. a | The imaginal wing disc of Drosophila melanogaster. The compartment
boundary (green line), Decapentaplegic (DPP) source and gradient in the tissue (green; here the gradient is only shown
in the posterior target compartment) are shown. L indicates target tissue width, and x the distance from the source.
b | Magnified view of cells in the posterior compartment and the DPP gradient (top), with an intensity profile (bottom) as
a function of the distance from the source; the gradient is characterized by its amplitude (C
0
) and decay length (λ). Note
that growth is approximately homogeneous: two clones (blue and pink) at different positions in the gradient have grown
to the same size; however, because the clones are exposed to different DPP concentrations they express different target
genes (blue and pink). c | The DPP signalling pathway. The transcription factor MAD is phosphorylated and activated
following binding of DPP to its receptor Thickveins (TKV). MAD can then bind MEDEA and accumulates in the nucleus,
where it represses the transcription of the repressor brinker (brk). BRK and phosphorylated MAD regulate DPP target
genes, such as Dad and spalt. Phosphorylated MAD also interacts with the co-transcriptional activator Yorkie to regulate
the transcription of the growth-promoting microRNA bantam. MAD phosphorylation is negatively regulated by DAD.
REVI EWS
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 12 | SEPTEMBER 2011 | 595
FOCUS ON mORphOgENESI S
© 2011 Macmillan Publishers Limited. All rights reserved
0CVWTG4GXKGYU¦/QNGEWNCT%GNN$KQNQI[
Limb bud C
D F
E
0ρ#
20 40 60 80 100 120 140 160 180
10
1
10
-2
10
-1
10
0
10
2
10
3
10
4
10
S
10
6
T
i
s
s
u
o

u
r
o
u

(
#
)
C
o
l
l

n
u
m
b
o
r

(
0
)
20 40 60 80 100 120 140 160 180
T
i
s
s
u
o

q
r
o
v
ì
b

r
u
ì
o

(
I
#
)

P
r
o
l
i
l
o
r
u
ì
i
o
n

r
u
ì
o

(
I
0
)
Timo (bours)
Timo (bours)
I
0
ŮI
#
#

#
M
i
ì
o
ì
i
c

i
n
d
o
x
/
µ
r
o
l
i
l
o
r
u
ì
i
o
n

r
u
ì
o

(
I
)
Z/.
I
#
I
c
o
ll

Ů

I
#
µ
H
3

i
n
ì
o
n
s
i
ì
y
I
c
o
ll

Ů

I
#
0 1
Posiìion ulonq uxis
12
10
8
6
4
2
T
im
o
2S 22 18 14 11
Coll cyclo lonqìb (bours)
Anìorior
Posìorior
M
i
ì
o
ì
i
c

i
n
d
o
x

(
°
)
30
21
23
2S
21 20 10 18
rate during development (see Supplementary informa-
tion S1 (table)). A key question, therefore, is how tissu e
growth causes the decrease of the DPP degradation
rate. Molecules regulating morphogen degradation and
thereby expanding a gradient in response to growth
have been termed expanders
28
. Below, the mechanisms
by which an expander might control DPP degradation
are referred to as scaling mechanisms.
Two types of scaling mechanisms have been proposed
(FIG. 3). An expander mechanism based on expansion–
repression feedback was originally proposed by Ben-Zvi
and Barkai
28–30
. In this model, the expander antagonizes
DPP degradation. The expander is a long-lived, rapidly
diffusible molecule, the expression of which is repressed
above a certain DPP concentration level. When the imagi-
nal disc grows, cellular DPP levels at the edge of the disc
Box 1 | Tissue growth properties
The wing imaginal disc of
Drosophila melanogaster
displays growth properties
similar to those of many other
developmental systems. First,
like in the vertebrate limbs or
the embryonic spinal chord,
morphogen gradients seem
to control growth
4,79
. Second,
like in vertebrate systems, the
growth rate decreases during
development and is fairly
homogeneous in space
24,80–82
(some heterogeneities
are apparent, but they are
relatively small and confined
to the boundary regions).
This was described in the
classical paper by Hornbruch
and Wolpert in 1970 (REF. 82),
as well as more recently with
quantitative detail by the
Sharpe group
80
(see the
figure, part a, which shows
growth in a chick limb bud
(left) and a mouse limb
bud (right)).
However, vertebrate
limb systems grow as
mesenchymal masses of cells
(which are three-dimensional
(3D)), whereas imaginal discs
are epithelia (2D sheets); this
reduces the analysis of their
growth to a 2D problem. The
measurement of disc growth
properties is therefore simple.
The cell density (ρ) in disc
epithelia increases slightly
during development. Because
the increase in cell density is
much smaller than the
increase in cell number (N)
(threefold versus 1,000-fold
24,43
), the tissue area (A) is roughly proportional to the number of cells (see the figure, part b);
cell number = cell density × tissue area (N = ρA) As a result, the average cell doubling time is equal to the area doubling time
(see the figure, part c). The height of cells along the z axis increases slightly as the cell density increases
83,84
, so that overall the
average cell volume is approximately constant (O.W., unpublished observations). Indeed, cell growth and proliferation are
coordinated in wing disc cells — cells divide when they have doubled their volume
85–87
. Furthermore, apoptosis levels are low
and fairly uniform
88
. Therefore, the average cell growth rate and proliferation rate (g
N
) are roughly the same and correspond to
the effective tissue growth rate (g
A
). Finally, staining with 5‑bromodeoxyuridine (BrdU) and phosphorylated histone H3 (pH3)
(used to visualize proliferation and cell division) as well as clonal analyses suggest that growth is fairly homogeneous
17–19,24,25,50
.
This implies that all cells have the same growth rate (g
cell
), which is approximately equal to the average tissue growth rate g
A

(see the figure, part d) and can therefore be inferred directly from measurements of the disc area. L, tissue width; x, distance
from the morphogen source. Part a of the figure is modified, with permission, from REFS 80,82.
REVI EWS
596 | SEPTEMBER 2011 | VOLUME 12 www.nature.com/reviews/molcellbio
REVI EWS
© 2011 Macmillan Publishers Limited. All rights reserved
J
J
J
Anìorior Posìorior
Winq
S
o
u
r
c
o
Z
.
Z
Z
Z
.
1
.
2
.
3
20
SS
100
20
SS
100
20
SS
100
0
100
200
300
400
20
40
60
80
100
120
140
160
0
2000
4000
6000 1
0.8
0.6
0.4
0.2
0
0
0.2
0.4
0.6
0.8
1
20
40
60
80
100
120
140
160
C
o
n
c
o
n
ì
r
u
ì
i
o
n
%
/
%
0
/. - consìunì
%
0
%
coll
/%
0
%
coll
/%
0
.
%
coll
%
coll
Timo (bours)
Z Z/.
Timo (bours)
0 S0 100 1S0 200 2S0 300 3S0
0
20
40
60
80
100
120
Tissuo vidìb (.)
D
o
c
u
y

l
o
n
q
ì
b

(



)
C
o
n
c
o
n
ì
r
u
ì
i
o
n
(
l
o
q

s
c
u
l
o
)
Coll numbor (0)
(loq sculo)
2
14
2
10
2
11
2
12
2
13
2
0
2
8
2
1
α - consìunì
α
α
α
×2
×1.S
×1.S
×1.S
%
0
%
coll
%
coll
S
c
u
li
n
q
(


-
.
)
C D E
F G
0CVWTG4GXKGYU¦/QNGEWNCT%GNN$KQNQI[
Poluìivo µosiìion (Z/.) °
Poluìivo µosiìion (Z/.) °
Poluìivo µosiìion (Z/.) °
λ
λ
λ C
o
n
c
o
n
ì
r
u
ì
i
o
n
C
o
n
c
o
n
ì
r
u
ì
i
o
n
C
o
n
c
o
n
ì
r
u
ì
i
o
n
Anisotropic tissue growth
Directionally dependent
growth of a tissue: more growth
along one axis (for example,
horizontal) than another (for
example, vertical) in the plane
of the epithelium.
fall below the threshold and cells start expressing the
expander (FIG. 3a). The expander then diffuses in the tissu e
and decreases the DPP degradation rate, thus expanding
the DPP concentration gradient. This results in an increase
in DPP levels at the edge of the disc and repressio n of the
expander, accompanied by disc growth
28,30
.
The second type of scaling mechanism is based on
reducing the DPP degradation rate in response to cell
division events
24
. This can be achieved, for example, by
cell-autonomous dilution of a long-lived expander that
promotes DPP degradation and might also diffuse in the
tissue (FIG. 3b). In this expander-dilution mechanism, the
degradation rate is inversely proportional to the number
of cells and, by extension, the area of the tissue, so the
gradient scales with the square root of the tissue area
but not necessarily with the tissue width (FIG. 3b; see
Supplementary information S1 (table)). By contrast, in
the expansion–repression mechanism, the gradient does
not necessarily scale with tissue area but scales with the
width of the tissue in the direction of the gradient, as
repression of the expander directly depends on the DPP
gradient along this axis (FIG. 3a).
Therefore, anisotropic tissue growth could help dis-
tinguish between the two types of scaling mechanism
(see Supplementary information S1 (table)). Indeed,
analysis of DPP gradient scaling in different aniso-
tropic growth conditions showed that area scaling, but
not width scaling, was the same in the different experi-
mental conditions
24
, suggesting that gradient scaling
might be due to a dilution-type mechanism, rather than
expansion–repression.
So far, no molecular equivalent of an expander
has been identified. However, Pentagone (PENT), a
molecule that is conserved from flies to vertebrates
31
,
shows some of the features required of an expander:
it is secreted, diffusible and repressed by DPP signal-
ling (and therefore transcribed in regions of low DPP
concentration)
31
. PENT antagonizes DPP degradation
by interacting with the heparan sulphate proteoglycan
Division abnormally delayed (DALLY)
31
, which itself is
involved in DPP turn over and diffusion
32–35
. It has yet to
be determined whether DPP gradient scaling is modified
in pent mutants.
An obvious target for an expander in the dilution
mechanism would be DPP receptor levels: the expander
could affect the effective intracellular DPP degradatio n
rate by modulating DPP receptor levels through recep-
tor ubiquitylation and internalization or through
changes in endosomal dynamics. The TKV degradation
rate, like the DPP degradation rate, seems to decrease
Figure 2 | Decapentaplegic gradient properties during growth. a | The Decapentaplegic (DPP) gradient expands in
growing wing imaginal discs; homogeneous growth implies that tissue proportions, and therefore the relative position
(distance from the source/tissue width (x/L)) of cells and their lineage, stay constant during development (two clones, blue
and pink, are shown). b–e | Gradient scaling implies that the DPP concentration profiles at different times of development
(b) collapse onto the same shape when relative concentration (C/C
0
;

in which C
0
is the gradient amplitude) and relative
position (x/L) are considered (c). This implies that the ratio of gradient decay length to tissue width (λ/L) is constant during
development (in other words, λ is proportional to L (d)), and cells at certain relative positions (blue and pink) always see the
same relative DPP concentration (C
cell
/C
0
) (c). This, in turn, implies that DPP concentration experienced by a particular cell
(C
cell
) is proportional to C
0
. The relative increase in concentration over time is the same for all cells (e) and empirically
correlates with growth: whenever the cellular concentration increases by a percentage α, the number of cells doubles (e).
REVI EWS
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 12 | SEPTEMBER 2011 | 597
FOCUS ON mORphOgENESI S
© 2011 Macmillan Publishers Limited. All rights reserved
0CVWTG4GXKGYU¦/QNGEWNCT%GNN$KQNQI[
M# - consìunì
- ū(#)
M.
2
- consìunì
- .
M
i
.
i
M - 1/4 M
i
. - 2 .
i
Docuy by diluìion
ovinq ìo qrovìb
und coll divison
Morµboqon qrudionì oxµunsion Lxµundor
Lxµundor
Lxµundor ìrunscriµìion
C
D
M
i
#
i
M - 1/4 M
i
# - 4 #
i
Morµboqon qrudionì oxµunsion Lxµundor
Morµboqon qrudionì
λ
λ
during development
24
, so it would be interesting to
pursu e this line of investigation further.
Finally, it is unlikely that the scaling of the signalling
gradient happens only at the level of the ligand concen-
tration gradient. Indeed, in experiments in which both
endogenous dpp and its downstream target brk (FIG. 1c)
are mutated, an overall increase in the transcription
of the DPP target gene Dad was still observed during
development
24
. Increased transcription of Dad over time
even in the absence of DPP and the BRK repressor might
indicate that not only DPP degradation but also target
gene production is sensitive to tissue size. Alternative
explanations are that a second BMP-type morphogen,
GBB, can activate MAD and also contribute to Dad tran-
scription
36–39
, or that the mutations used here might not
represent complete loss of functions.
Growth control by the DPP gradient
The previous section considered how growth affects the
DPP gradient. But, how does the DPP gradient influ-
ence growth control in the imaginal disc? Proliferation is
approximately spatially uniform and decreases over time
(BOX 1). By contrast, DPP concentration and signalling
levels are graded in space and increase over time (FIG. 2).
Therefore, absolute DPP levels alone cannot control
homogeneous proliferation. However, cells could inter-
pret spatial differences
23,24
or temporal changes in DPP
concentration or signalling levels
24
to set their growth
rate or cell cycle length. Another possibility is that addi-
tional factors, such as mechanical stress or the presence
of other growth factors, act in concert with DPP to make
growth homogeneous (FIG. 4).
Below, we discuss the four major proposed morpho-
genetic growth control models: growth control by
mechanical feedback, growth control by complemen-
tary inhibition and growth control based on spatial dif-
ferences in DPP levels between neighbouring cells or on
temporal changes in DPP signalling.
Growth control based on mechanical feedback. In epi-
thelia, cells adjust their height and apical surface area in
response to mechanical stress (for example, compres-
sion by surrounding cells). In other words, high levels of
mechanical stress can physically limit tissue growth by
locally limiting cell growth
22,40
. The idea that mechanical
stress limits cell growth led to the formulation of models
in which DPP is merely permissive for growth (that is,
it promotes growth when its levels are above a certain
threshold) and mechanical stress acts as a long-range
signal to instruct cell growth rates
20–22
.
In a simple model based on mechanical feedback,
proliferation is stimulated by a central point source of
growth factors, generated by the intersection of line
sources of the two diffusible morphogens, DPP and
Wingless, in the centre of the disc
20–22
. Cells divide
above a certain threshold of growth factor concentra-
tion. Proliferation forces cells to move out of the centra l
growth factor zone into the periphery, where they can
divide easily even with low growth factor levels because
they experience low mechanical stress levels
20–22
.
Proliferation in the periphery then compresses cells in
the centre and limits their growth
20–22
. Thus, the build-
up of mechanical stress distributions in the tissue can
lead to a situation in which growth is homogeneous
even when there is a steep DPP gradient. As cell com-
pression, and thus mechanical stress, increases during
development, the effective tissue growth rate decreases
overall
20–22
(FIG. 4a).
Experimental support for the role of mechanical
stress in growth control comes from measurements of
cell compression in the centre of the disc
41,42
, which sug-
gest that central cell compression increases during devel-
opment
41
. Whether this increase in mechanical stress is
large enough to be limiting for cell growth is not clear.
Figure 3 | Possible mechanisms for morphogen gradient expansion and scaling.
a | Expansion–repression mechanism. The expander (pink) antagonizes morphogen
degradation. It is transcribed (orange) below a threshold morphogen concentration
(green); when tissue growth causes cells in the periphery to experience morphogen
levels below this threshold, they produce a long-lived expander that diffuses in the
tissue, reduces morphogen degradation rates (k) and thereby expands the morphogen
gradient to fit into the new tissue width (L). In this model, the initial degradation rate
k
i
 decreases in an inversely proportional manner to L
2
; this implies that the gradient
decay length (λ) is proportional to the tissue width L. This is known as width scaling.
b | Expander-dilution mechanism. The expander promotes degradation of the
morphogen. When the tissue grows, the expander is diluted in response to cell division
events, and, therefore, morphogen degradation rates (k) decrease and the morphogen
gradient expands. In this model, the degradation rate k is inversely proportional to the
tissue area A; this implies that the gradient decay length λ is proportional to the square
root of the tissue area A. This is known as area scaling (see also Supplementary
information S1 (table)).
REVI EWS
598 | SEPTEMBER 2011 | VOLUME 12 www.nature.com/reviews/molcellbio
REVI EWS
© 2011 Macmillan Publishers Limited. All rights reserved
0CVWTG4GXKGYU^/QNGEWNCT%GNN$KQNQI[
6KOG
&KXKUKQP
&KXKUKQP
+P[QWPIFKUE F E
C D
6
K
O
G
α
α
+PQNFFKUE
S
ì
r
o
ì
c
b
i
n
q
C
o
m
µ
r
o
s
s
i
o
n
luì ucìiviìy
G
r
o
v
ì
b

i
n
b
i
b
i
ì
i
o
n
Z
Wild ìyµo
G
r
o
v
ì
b

r
u
ì
o
Z
Z
8PK
DTMmuìunì HCVmuìunì
DPP
C
o
n
c
o
n
ì
r
u
ì
i
o
n
(
l
o
q

s
c
u
l
o
)

. . .
Timo
DPP
Ducbs
Coll division No division
Timo
?
In fact, area growth does not seem to be constrained by
the increase in cell compression: the tissue area increases
substantially even late in development, when the cell
density no longer changes
24,43
, indicating that mechani-
cal stress levels in the disc may not directly control or
significantly limit cell growth rates.
Furthermore, some assumptions made by mechani-
cal growth models do not seem to apply to the wing
imaginal disc. First, data suggest that Wingless does not
control growth in the wing disc
44,45
, so the growth factor
(DPP) is produced by a line source rather than a point
source. A line source causes different stress and pro-
liferation patterns than a point source because there is
no longer a peripheral ring of stretched, proliferating
cells that compresses the cells in the central growth fac-
tor region. As a result, proliferation would (probably)
not be uniform.
Another assumption in the original model was that
the growth factor gradient does not scale
21,22
. For a scal-
ing gradient such as DPP, the central growth factor
region would simply expand as the tissue grows. As a
consequence, cells would not move out of this zone into
the periphery, there would be no compression of cells
in the central region, and growth would not stop.
A mechanical model based on modified assumptions
might still provide an explanation for the growth prop-
erties of the wing imaginal disc. However, the specific
form of such a modified model is currently unclear. In
any case, a systematic experimental analysis of stress
distributions in the disc at different times of develop-
ment is needed to get a clearer picture of the possible
role of mechanical stresses as regulators of proliferation.
Growth control by complementary inhibition. Another
permissive growth model proposed that DPP allows
growth by suppressing BRK, an integral component of
the DPP pathway (FIG. 4b). BRK is a repressor of DPP
target genes and is repressed by DPP signalling; it is
therefore expressed as an ‘anti-gradient’ (REFS 12,13,15)
(FIG. 1c). The proposal for a permissive role of DPP in
growth through the repression of BRK was based on
the observation that, although tkv-mutant or Mad -
mutant cells do not grow, cells mutated for both TKV
and BRK or for both MAD and BRK do grow
12–14
. Thus,
high level s of BRK inhibit growth, and mutation of
BRK rescues the growth of cells lacking DPP signalling
upstream of BRK (FIG. 1c). Therefore, BRK plays a major
part in DPP signal transduction and the downstream
regulatio n of growth.
Consistent with these observations, wing discs in
which BRK is ubiquitously expressed do not grow, brk-
mutant discs overgrow in lateral regions (where BRK is
normally expressed) and, strikingly, wing discs that are
deficient in both DPP and BRK have a growth pheno-
type similar to brk mutants
25
. Because BRK acts as a
repressor in the DPP pathway, the output of the pathway
manifested in target gene expression was assumed to
be maximal in brk-mutant cells
4,25,46
. As a result, DPP
signalling levels were assumed to be spatially uniform
in a brk-mutant disc
4,25,46
. Thus, the observation that in
brk-mutant discs cells in lateral regions proliferate more
than cells in the centre (although they are supposedly
experiencing the same DPP signalling levels) implied
that growth in central and lateral regions of the disc had
to be regulated independently and that the growth role
of DPP in the centre is just permissive: the gradient is
irrelevant for growth in the centre
25
.
Figure 4 | Models for growth control by morphogen gradients. a | Mechanical
stress model: the proliferation of cells in the central region in response to a morphogen
(green) stretches the cells in the periphery, causing them to divide (low stress; yellow).
The proliferation of cells in the periphery compresses the cells in the centre, inhibiting
their growth. Growth factors (morphogens) and low stress (stretching) stimulate
division, whereas compression inhibits growth. This model assumes that gradient
scaling does not occur, and the shape of the gradient is irrelevant. b | Complementary
inhibitor model. In this model, the gradient is irrelevant. Decapentaplegic (DPP) allows
growth by repressing Brinker (BRK) in the central region of the disc; in addition, Fat
activity controls growth in a similar manner at the edge of the disc. The spatial profile
of the Fat activity gradient (top) has not been measured directly but is inferred from the
growth phenotype of fat mutants (bottom). This model requires that the DPP and Fat
activity gradients are established in an independent manner. x indicates distance from
the morphogen source. c | Spatial model. The relative slope of an exponential
morphogen concentration gradient could control cell division (top). For a scaling
gradient, the relative slope decreases as the tissue grows, consistent with decreasing
growth rates; disc cells (bottom) can sense relative spatial differences in DPP (green)
(and probably other factors) through the non-conventional myosin Dachs (yellow).
Large spatial differences in DPP levels lead to cell division. By contrast, cell division
would not occur when the DPP concentration gradient is uniform. d | Temporal model.
Growth is controlled by a percentage increase in morphogen concentration (of
α = 50%) over time (morphogen concentration is illustrated by the number of dots in
each cell). The time it takes to reach a 50% increase in concentration is equal to the cell
cycle length and is longer at late stages of development because the increase in
concentration becomes smaller as the tissue grows.
REVI EWS
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 12 | SEPTEMBER 2011 | 599
FOCUS ON mORphOgENESI S
© 2011 Macmillan Publishers Limited. All rights reserved
0CVWTG4GXKGYU¦/QNGEWNCT%GNN$KQNQI[
C9KNFV[RG DDTMFRROWVCPV
Wild ìyµo
Z/.
D
P
P

s
i
q
n
u
l
l
i
n
q

l
o
v
o
l
s
(
&
C
F

r
o
u
d
o
u
ì
)
E
DTM muìunì
Z/.
F
Timo Timo
DTMFRR muìunì
Z/.
G
Timo
S0 µm S0 µm
At first, it was hypothesized that growth in the periph-
ery is controlled by mechanical stress
20,21
(see above).
More recently, it was proposed that homogeneous
growth could be regulated by an additional growth
inhibitory gradient that is complementary to the inhibi-
tory BRK gradient
46
; specifically, the activity gradient of
Fat, which is involved in growth mediated by the Hippo
pathway (see below) (FIG. 4b). This proposal was based
on the growth phenotype of fat-mutant wing imaginal
discs, which is complementary to that of brk mutants
46

(FIG. 4b). Furthermore, overgrowth of imaginal discs with
increased DPP signalling levels could be suppressed by
increasing Fat activity
46
.
In the complementary inhibitor model, the fact that
BRK and Fat activity are graded in space is not relevant
for growth regulation, as long as the shape of the gra-
dients is perfectly complementary. However, for this
model it is crucial that the two gradients are independ-
ent
46,47
. If they are not — for example, if Fat activity is
downstream of DPP — then there is no complementary
inhibition, but instead growth depends only on DPP
signalling. Indeed, DPP signalling affects the expression
of Dachsous and Four-jointed, two proteins regulating
Fat activity, and the localization of Dachs, a protein
responding to Fat activity
48
. It is difficult to reconcile
these experimental findings with the complementary
inhibition model. Independent establishment of two
exactly complementary gradients is also problematic
because it is difficult to ensure robustness. In any case,
however, factors other than Fat could still regulate
growth in a manner complementary to BRK.
The key assumption of this model is that DPP signal-
ling levels and the expression of target genes is maximal
in brk-mutant cells and, therefore, spatially uniform in
brk-mutant discs, which overgrow
4,25,46
. In other words,
in the absence of a DPP signalling pattern in space
(a spatially uniform signal) and of changes in time (a
maximal signal), growth still occurs: thus, DPP is just
permissive for growth. However, although DPP target
genes (such as spalt and Dad) are ectopically expressed
in brk-mutant clones of cells and brk-mutant discs, their
expression is not spatially uniform
12,13,15,24,25
(FIG. 5). This
indicates that loss of repression by BRK does not lead to
maximum DPP signalling. Indeed, analysis of the Dad
enhancer revealed parallel inputs into the Dad regu-
latory regions from BRK and MAD, rather than BRK
alone
16
. Therefore, graded signalling in brk mutants
could, in principle, be due to BRK-independent signal-
ling mediated by MAD. This does not seem to be the
case, however, because in brk dpp double mutants, which
lack phosphorylated MAD, the expression of target
genes is also not spatially uniform
24,25
(FIG. 5). This reveals
that our understanding of the DPP signalling pathway is
incomplete, because our current model of the pathway
(FIG. 1c) would predict spatially and temporally uniform
DPP target gene expression in the absence of both BRK
and phosphorylated MAD.
Strikingly, although absolute levels of DPP output, as
monitored with a Dad transcriptional reporter, are much
lower in the brk dpp mutant (FIG. 5e), cells still proliferate,
indicating that the absolute output of the DPP signal-
ling pathway downstream of BRK cannot directly deter-
mine the growth rate (although one cannot completely
exclude the possibility that the low DPP signalling levels
are still above a permissive threshold). However, because
the DPP signalling output in brk dpp mutant conditions is
neither uniform in space nor constant in time
24,25
(FIG. 5),
growth of Mad brk mutant or tkv brk mutant cells
12–14
is
not necessarily uncovering a scenario in which DPP
is merely permissive for growth. The growth phenotypes
of these mutants might also be consistent with instruc-
tive growth models, in which cells measure either spatial
differences or temporal changes in DPP signalling levels
to set their growth rate or cell cycle length.
Growth control based on spatial differences in DPP
signallin g. One of the first instructive models proposed
that proliferation could be controlled by the local slope
of the DPP gradient: neighbouring cells measure this
slope through their differences in DPP levels, the slope
flattens during the growth phase, and proliferation stops
when the slope reaches a minimal steepness
23
(FIG. 4c).
This model was first proposed by Day and Lawrence
23
,
who assumed that there is a linear gradient that stretches
as the tissue grows. This would explain homogeneous
Figure 5 | The role of BRK in signal transduction and growth. a,b | Decapentaplegic
(DPP) target gene spalt (green) in a wild-type imaginal disc (a) and in a brinker (brk) and
dpp double mutant (b). Note that the expression pattern and range of spalt (white arrow)
is similar in both the wild type and the brk dpp mutant; note also that, although there is
ectopic expression, the expression of spalt is not spatially uniform in brk dpp double
mutants. This indicates that the output of the DPP signalling pathway downstream of BRK
is not uniform in space because otherwise spalt should be uniformly expressed in the
entire disc. c–e | DPP signalling levels (graphs) and growth phenotypes (insets) in wild-type
imaginal discs (c), brk mutants (d) and brk dpp double mutants (e) based on quantifications
of red fluorescent protein expression, driven by the Dad enhancer
24
. DPP signalling levels
are not spatially uniform and they increase over time; they are also more than tenfold
lower in brk dpp mutants than in brk mutants and wild-type cells (not drawn to scale here).
L indicates tissue width and x distance from the morphogen source. Images in parts a and
b are modified, with permission from REF. 25 © (2008) The Company of Biologists.
REVI EWS
600 | SEPTEMBER 2011 | VOLUME 12 www.nature.com/reviews/molcellbio
REVI EWS
© 2011 Macmillan Publishers Limited. All rights reserved
Spatial derivative
The spatial difference in
a quantity (for example,
concentration) across one
spatial unit (for example, a
cell); the spatial derivative is
denoted with a prime symbol,
for example dC / dx = C′.
Time derivative
The rate of change of a
quantity (for example,
concentration) over time; the
time derivative is denoted with
a dot symbol, for example
dC / dt = C
˙
.
growth (a single slope in a linear gradient) and the
slow-down of the growth rate (as the gradient becomes
shallowe r during growth, its slope becomes smaller).
Regeneration experiments in cockroaches supported
this hypothesis
49
. However, the model was challenged
by a mosaic analysis in D. melanogaster wing discs car-
rying clones of cells that express a constitutively active
form of TKV (TKV
QD
) and therefore have higher DPP
signalling levels than surrounding cells
19
. According to
the model, cells at the edge of such a clone would expe-
rience a big spatial difference in DPP signalling levels
and should therefore divide more. No extra proliferation
was observed at the border of clones
19
, and this observa-
tion was thought to refute the model. However, recent
experiments showed that there is a transient additional
proliferation effect on the cells at the border of clones,
observed only in a time-course experiment
50
. The tran-
sient extra proliferation of cells exposed to big spatial
differences in DPP signalling levels supports the idea
that spatial cues have a role in growth control.
The DPP gradient is not a linear gradient with a
constant slope in space, as had been assumed by Day
and Lawrence in the first spatial model; instead, it is
steeper close to the source and shallow away from it
27,28,51

(FIG. 1a). In other words, the slope of the DPP gradient
is actually position dependent and so cannot control
position-independent proliferation rates. However, the
exponential approximation of the DPP gradient
26,27,50

shows that relative spatial differences in DPP concen-
tration — that is, the percentage by which the DPP
concentration decreases across a cell’s surface, or C′/C
(the spatial derivative of C, normalized to C) — are, like
the growth rate, position independent. In an exponential
gradient, C′/C is the same for all the cells in the tissue at
a given time and depends only on one global parameter,
the decay length λ, which increases as the tissue grows
24

(C′/C = 1/λ; for details, see supplementary informa-
tion S1 (table)). Therefore, the value of C′/C is the same
for all cells, and, like their growth rate, C′/C decreases
during development.
One way to directly test whether growth is controlled
by spatial cues would be to create uniform DPP signal-
ling in space. Wing discs in which brk is mutated or DPP
is ubiquitously expressed overgrow
25
. It was thought that,
if spatial differences in DPP levels between cells were
the driving force of proliferation, there should have
been less growth. However, the overgrowth pheno-
type is not inconsistent with a spatial model because,
on closer inspection, the expression of DPP targets was
un expectedly not uniform in these conditions, indicating
a non-uniform DPP signalling profile
24,25
(FIG. 5). Indeed,
later quantifications showed that, even when DPP is
ubiquitously expressed, relative spatial differences in sig-
nalling levels between neighbouring cells still correlate
with their growth rates
24
. However, this correlation is less
clear in the brk mutant
24
. It is also important to note that
the parameters of the spatial model (for example, the
C′/C threshold below which cells stop dividing at the end
of the growth phase) differed between different tissues
and mutant conditions
24
. Nevertheless, the spatial model
remains a plausibl e mechanism for growth control.
Growth control based on temporal changes in DPP signal-
ling. An alternative to the spatial model proposes that cell
cycle length is determined by relative temporal changes in
cellular DPP levels; that is, by the percentage by which the
DPP concentration increases over time, or Ċ/C (the time
derivative of C, normalized to C) (FIG. 4d). Empirically, it
was found that when DPP signalling levels have increased
by a percentage α (in this case α = 50%) from the begin-
ning of the cell cycle, cells divide
24
(FIG. 2e). If the gradi-
ent scales, cellular DPP signalling levels increase by the
same percentage in all cells (FIG. 2e), and therefore, in a
temporal model, all cells have the same growth rate
24
(see
Supplementary information S1 (table)). If DPP signal-
ling levels increase quickly, α is reached quickly, and the
cell proliferation rate is high (corresponding to a short cell
cycle length). At the end of development, when DPP levels
increase more slowly (FIG. 2b), it takes cells a longer time
to reach α, and therefore the cell cycle is longer (FIG. 4d).
Support for the temporal model came from experi-
ments in which the velocity of the temporal increase of
DPP signalling was manipulated exogenously by drug-
mediated induction of the expression of TKV
QD
in clones
of cells. In this experiment, the DPP signalling levels in
clone cells increased faster than in wild-type cells; how-
ever, consistent with a temporal model, the cells divided
when DPP signalling levels had increased by α = 50%
24
.
How quickly cells attained this 50% increase depended
on the concentration of the drug that determined the
velocity of the increase in DPP signalling levels. It also
depended on how distant the clones were from the
source: clones far away from the source initially had lower
DPP signalling levels. Therefore, upon TKV
QD
expres-
sion, the relative increase in their DPP signalling levels
was bigger. Because of this, clones that were away from
the source were bigger than those closer to the source,
which is consistent with the temporal growth model
19,24
.
Interestingly, growth control based on the temporal
model can account for non-homogeneous proliferation
patterns observed in and around TKV
QD
clones and in
imaginal discs in which DPP signalling is ubiquitously
increased
24
. Under these conditions, the gradient no
longer scales, and, as a consequence, DPP signalling
levels increase faster or slower depending on position:
unlike in wild-type discs, Ċ/C is position dependent.
Importantly, cell proliferation rates are also position
dependent and correlate with the position-dependent
Ċ/C
24
. This further supports the temporal model as a
plausible mechanism for growth control.
Measurement of C′/C or Ċ/C by cells. In models in
which DPP is merely permissive for growth (FIG. 4a,b),
the relevance and output of DPP signalling is fairly easy
to understand: cells above a certain threshold of DPP
concentration would express DPP growth targets (for
example, the miRNA bantam) and divide. By contrast,
how cells could measure relative spatial differences or
temporal changes in DPP signalling levels is much less
intuitive. It is easy to show that these gradient properties
correlate with the growth rate, but how could cells actu-
ally measure relative changes in concentration in space
or over time?
REVI EWS
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 12 | SEPTEMBER 2011 | 601
FOCUS ON mORphOgENESI S
© 2011 Macmillan Publishers Limited. All rights reserved
0CVWTG4GXKGYU^/QNGEWNCT%GNN$KQNQI[
#PVGTKQT 2QUVGTKQT
9KPI
5
Q
W
T
E
G
Z
&22 C D
&CEJUQWU
&22
&CEJUQWU
&22
6-8
/#& /#&
/'&'#
2
2
&CEJUQWU
(CV
9CTVU
;QTMKG
;QTMKG
&CEJU
&CEJU
Hippo 9CTVU
0WENGWU
%[VQRNCUO
%GNNOGODTCPG
/#&
/'&'#
2
/#&
;QTMKG
2
$4-
$4-
DTM
&CFURCNV
DCPVCO
Planar cell polarity
A mechanism of cellular
organization, distinct from
apical–basal polarity, by which
cells acquire information about
their orientation within the
tissue in the plane of the
epithelium.
To measure relative spatial differences in DPP, cells
must compare signalling levels at different locations to
each other and with respect to their own signalling lev-
els. A related phenomenon might occur during planar
cell polarity (PCP), when local levels of PCP factors in
neighbouring cells can influence intracellular gradients
of such factors
52
. In the wing disc, cells do indeed seem
to be able to measure DPP signalling differences between
adjacent cells using the Fat–Hippo pathway
48
(FIG. 6).
This signalling pathway regulates organ size by control-
ling cell growth, proliferation and apoptosis
53
. The ligand
for Fat is the protocadherin Dachsous, which forms a
proximo–distal gradient in the disc
48,54,55
(FIG. 6a). Cells
respond to the Fat signalling gradient by differential
intracellular activation of the non-conventional myosin
Dachs
48,55–57
: Dachs accumulates at the cell–cell inter-
face that is facing the lower levels of Dachsous (FIG. 6b).
This also corresponds to the cell–cell interface at which
DPP (and Wingless) signalling levels are higher. When
the spatial pattern of DPP signalling was changed, for
example by generating clones of cells expressing TKV
QD

or BRK, Dachs relocalized to the new interface facing
higher DPP signalling levels
48
. The amount of Dachs at
the interface seems to be independent of absolute DPP
levels, suggesting that cells may sense relative spatial
difference s in DPP signalling levels
48
.
Dachs regulates the activity of Yorkie by inactivat-
ing Warts, a kinase that reduces Yorkie activity (FIG. 6b).
Yorkie regulates the expression of growth regulatory
targets, such as Cyclin E, D. melanogaster inhibitor of
apoptosis 1 and MYC, as well as the growth-promoting
miRNA bantam
53,58–65
(the last in cooperation with the
DPP transcription factor MAD
66
). Although the quanti-
tative relationships between the Fat–Hippo pathway and
the DPP gradient are still obscure, an intriguing possi-
bility is that this pathway could detect spatial disconti-
nuities in the DPP gradient and in response regulate the
elimination of these discontinuities by local upregula-
tion of cell proliferation or apoptosis, through titration
of active Dachs and Yorkie levels. Such a mechanism
might be responsible for regeneration in response to
tissue damage
67,68
, the perturbation of growth observed
in and around TKV
QD
- or BRK-expressing clones
50
,
and even the phenomenon of cell competition, when
slow-growing clones with low DPP signalling levels are
eliminate d from the tissue
69,70
.
Measurement of the relative temporal changes of sig-
nalling levels implies that adaptive responses are gener-
ated in the signalling pathway (reviewed in REF. 71).
An adaptive response allows cells to measure relative
increases of a signal, because the signalling system adapts
to ambient concentrations of the signal and increasing
Figure 6 | Interaction between Decapentaplegic and the Fat–Hippo pathway. a | Decapentaplegic (DPP; green) forms a
gradient along the anterior–posterior axis (here the gradient is only shown in the posterior compartment to the right).
Dachsous (blue) forms a gradient along the proximo–distal axis (it is high along the edges and decreases towards the centre).
x indicates distance from the morphogen source. b | The DPP pathway and the Fat signalling pathway responding to
Dachsous interact. High levels of Dachsous (right) activate the Fat signalling pathway, inhibiting Dachs and Yorkie (through
the kinase Warts). By contrast, on the side of the cell where Dachsous levels are low (left) — which can be caused by an
ectopic increase in DPP signalling levels in clones of cells — Dachs is active, leading to the activation of Yorkie. Together with
phosphorylated MAD, Yorkie regulates the transcription of the growth target bantam. The arrows shown do not indicate
regulatory pathways but outcomes. The dashed arrow indicates an indirect interation. BRK, Brinker; TKV, Thickveins.
REVI EWS
602 | SEPTEMBER 2011 | VOLUME 12 www.nature.com/reviews/molcellbio
REVI EWS
© 2011 Macmillan Publishers Limited. All rights reserved
concentrations are necessary to maintain the response
72
.
In other words, such signalling systems can sense an
increase with respect to previous levels of the signal.
Adaptive signalling pathways have been extensively
studied in other model systems in which relative changes
of a signal are measured, such as chemotaxis in bacteria
or amoebae and olfactory or gustatory transduction in
Caenorhabditis elegans
73–76
. Such relative changes of sig-
nal can be detected by one of two network motifs: the
incoherent feed-forward loop (IFFL)
77,78
and the negative
feedback loop with a buffering node (NFBLB)
78
. It will
be important to study if any of these motifs are present
in the wiring of the DPP pathway and relevant to DPP-
dependent growth control. Interestingly, it seems that
DPP, Yorkie and MYC could form a NFBLB
66
. Therefore,
cells might adapt to relative temporal changes in DPP
signalling levels through Fat signalling. Another possi-
bility, not discussed in more detail here, is that cells could
measure both C′/C and Ċ/C, or even relative temporal
changes of C′/C (see Supplementary informatio n S1
(table)), with a single sensor module
76
.
Conclusions and perspectives
Recent quantifications of growth and morphogen gra-
dient properties in the wing imaginal disc of D. mela-
nogaster have allowed us to formulate and discuss
current morphogenetic growth models in a slightly more
precise manner, although we still cannot favour one
model over another. Many components of the machinery
that control morphogenetic growth in this model system
are already known, but it will be interesting to identify
new (or old) factors that implement this control through
cell growth or cell cycle progression. However, given
the wealth of qualitative information we already have
in this model system, the emphasis might now be not
on identifying new components in the machinery that
controls morphogenetic growth, but on unravelling the
dynamics of the interactions of the known components
by quantitativ e experimental analysis.
Measurements of gradient properties so far have
been either qualitative in nature or focused on average
properties of cell populations. Long-term in vivo cul-
tures and movies will open the way for the description
and analysis of parameter distributions in individual
cells. It will be interesting to measure DPP concentra-
tion and signal transduction, Ċ/C and C′/C distribu-
tions and mechanical properties during individual cell
cycles concomitantly with cell volume increase, cell cycle
phase progression, growth regulatory protein levels and
orientation of cell division as a function of cell position
and developmental time. It is very likely that cells can
measure and use a range of inputs — absolute signalling
levels as well as spatial, temporal and mechanical cues
— to orchestrate patterning and growth. Measuring and
relating these properties to each other and to the global
gradient and growth properties of the disc will hope-
fully allow us to complete the picture of morphogenetic
growth control.
1. Wolpert, L. Positional information and patterning
revisited. J. Theor. Biol. 269, 359–365 (2011).
2. Lecuit, T. & Le Goff, L. Orchestrating size and
shape during morphogenesis. Nature 450, 189–192
(2007).
3. Schwank, G. & Basler, K. Regulation of organ growth
by morphogen gradients. Cold Spring Harb. Perspect.
Biol. 2, a001669 (2010).
4. Affolter, M. & Basler, K. The Decapentaplegic
morphogen gradient: from pattern formation to growth
regulation. Nature Rev. Genet. 8, 663–674 (2007).
5. Bollenbach, T. et al. Precision of the Dpp gradient.
Development 135, 1137–1146 (2008).
6. Entchev, E. V., Schwabedissen, A. & Gonzalez-Gaitan, M.
Gradient formation of the TGF-β homolog Dpp. Cell
103, 981–991 (2000).
7. Teleman, A. A. & Cohen, S. M. Dpp gradient formation
in the Drosophila wing imaginal disc. Cell 103,
971–980 (2000).
References 6 and 7 show the distribution of the
DPP ligand as a gradient by using DPP tagged to
green fluorescent protein.
8. Nellen, D., Burke, R., Struhl, G. & Basler, K. Direct
and long-range action of a DPP morphogen gradient.
Cell 85, 357–368 (1996).
9. Lecuit, T. et al. Two distinct mechanisms for long-range
patterning by Decapentaplegic in the Drosophila wing.
Nature 381, 387–393 (1996).
References 8 and 9 reveal that DPP acts as a
morphogen that directly activates target genes at
different distances from the source of production
in a DPP concentration-dependent manner.
10. Kim, J., Johnson, K., Chen, H. J., Carroll, S. &
Laughon, A. Drosophila Mad binds to DNA and
directly mediates activation of vestigial by
Decapentaplegic. Nature 388, 304–308 (1997).
11. Ruberte, E., Marty, T., Nellen, D., Affolter, M. &
Basler, K. An absolute requirement for both the type II
and type I receptors, Punt and Thick veins, for Dpp
signaling in vivo. Cell 80, 889–897 (1995).
12. Campbell, G. & Tomlinson, A. Transducing the
Dpp morphogen gradient in the wing of Drosophila:
regulation of Dpp targets by brinker. Cell 96,
553–562 (1999).
13. Jazwinska, A., Kirov, N., Wieschaus, E., Roth, S. &
Rushlow, C. The Drosophila gene brinker reveals
a novel mechanism of Dpp target gene regulation.
Cell 96, 563–573 (1999).
14. Marty, T., Muller, B., Basler, K. & Affolter, M.
Schnurri mediates Dpp-dependent repression of brinker
transcription. Nature Cell Biol. 2, 745–749 (2000).
15. Minami, M., Kinoshita, N., Kamoshida, Y., Tanimoto, H.
& Tabata, T. brinker is a target of Dpp in Drosophila
that negatively regulates Dpp-dependent genes.
Nature 398, 242–246 (1999).
16. Weiss, A. et al. A conserved activation element
in BMP signaling during Drosophila development.
Nature Struct. Mol. Biol. 17, 69–76 (2009).
17. Garcia-Bellido, A. & Merriam, J. R. Parameters of the
wing imaginal disc development of Drosophila
melanogaster. Dev. Biol. 24, 61–87 (1971).
18. Gonzalez-Gaitan, M., Capdevila, M. P. &
Garcia-Bellido, A. Cell proliferation patterns in the wing
imaginal disc of Drosophila. Mech. Dev. 46, 183–200
(1994).
19. Martin-Castellanos, C. & Edgar, B. A. A characterization
of the effects of Dpp signaling on cell growth and
proliferation in the Drosophila wing. Development
129, 1003–1013 (2002).
20. Aegerter-Wilmsen, T., Aegerter, C. M., Hafen, E. &
Basler, K. Model for the regulation of size in the
wing imaginal disc of Drosophila. Mech. Dev. 124,
318–326 (2007).
21. Hufnagel, L., Teleman, A. A., Rouault, H., Cohen, S. M.
& Shraiman, B. I. On the mechanism of wing size
determination in fly development. Proc. Natl Acad. Sci.
USA 104, 3835–3840 (2007).
22. Shraiman, B. I. Mechanical feedback as a possible
regulator of tissue growth. Proc. Natl Acad. Sci. USA
102, 3318–3323 (2005).
References 20–22 are the first theoretical
proposals of the role of mechanics on growth
control in the imaginal discs.
23. Day, S. J. & Lawrence, P. A. Measuring dimensions:
the regulation of size and shape. Development 127,
2977–2987 (2000).
24. Wartlick, O. et al. Dynamics of Dpp signaling and
proliferation control. Science 331, 1154–1159 (2011).
Analyses quantitatively both the growth of imaginal
discs and the properties of the DPP gradient,
showing that the gradient scales owing to the
regulation of the degradation rate of DPP and that
cells divide when they increase their signalling levels
by 50%.
25. Schwank, G., Restrepo, S. & Basler, K. Growth
regulation by Dpp: an essential role for Brinker
and a non-essential role for graded signaling levels.
Development 135, 4003–4013 (2008).
Questions whether the graded distribution of
DPP is relevant to growth control, based on the
assumption that brk mutants (or brk dpp double
mutants) show uniform DPP signalling. The authors
also propose that DPP controls growth permissively
in the centre of the discs, whereas growth in the
periphery is independent of the DPP gradient.
26. Tanimoto, H., Itoh, S., ten Dijke, P. & Tabata, T.
Hedgehog creates a gradient of DPP activity in
Drosophila wing imaginal discs. Mol. Cell 5, 59–71
(2000).
27. Kicheva, A. et al. Kinetics of morphogen gradient
formation. Science 315, 521–525 (2007).
28. Barkai, N. & Ben-Zvi, D. ‘Big frog, small frog’ —
maintaining proportions in embryonic development:
delivered on 2 July 2008 at the 33
rd
FEBS Congress in
Athens, Greece. FEBS J. 276, 1196–1207 (2009).
29. Ben-Zvi, D., Shilo, B. Z., Fainsod, A. & Barkai, N.
Scaling of the BMP activation gradient in Xenopus
embryos. Nature 453, 1205–1211 (2008).
The first theoretical analysis of a possible
mechanism of gradient scaling.
30. Ben-Zvi, D. & Barkai, N. Scaling of morphogen gradients
by an expansion-repression integral feedback control.
Proc. Natl Acad. Sci. USA 107, 6924–6929 (2010).
Proposes for the first time an expander mechanism
based on expansion–repression feedback.
31. Vuilleumier, R. et al. Control of Dpp morphogen
signalling by a secreted feedback regulator. Nature Cell
Biol. 12, 611–617 (2010).
The first report of the effects of PENT on the DPP
gradient. PENT displays properties akin to the
expander by expansion–repression feedback, as
proposed in reference 30.
REVI EWS
NATURE REVIEWS | MOLECULAR CELL BIOLOGY VOLUME 12 | SEPTEMBER 2011 | 603
FOCUS ON mORphOgENESI S
© 2011 Macmillan Publishers Limited. All rights reserved
32. Akiyama, T. et al. Dally regulates Dpp morphogen
gradient formation by stabilizing Dpp on the cell
surface. Dev. Biol. 313, 408–419 (2008).
33. Takeo, S., Akiyama, T., Firkus, C., Aigaki, T. &
Nakato, H. Expression of a secreted form of Dally,
a Drosophila glypican, induces overgrowth phenotype
by affecting action range of Hedgehog. Dev. Biol. 284,
204–218 (2005).
34. Han, C., Yan, D., Belenkaya, T. Y. & Lin, X. Drosophila
glypicans Dally and Dally-like shape the extracellular
Wingless morphogen gradient in the wing disc.
Development 132, 667–679 (2005).
35. Fujise, M. et al. Dally regulates Dpp morphogen
gradient formation in the Drosophila wing.
Development 130, 1515–1522 (2003).
36. Bangi, E. & Wharton, K. Dpp and Gbb exhibit different
effective ranges in the establishment of the BMP
activity gradient critical for Drosophila wing
patterning. Dev. Biol. 295, 178–193 (2006).
37. Ray, R. P. & Wharton, K. A. Context-dependent
relationships between the BMPs gbb and dpp during
development of the Drosophila wing imaginal disk.
Development 128, 3913–3925 (2001).
38. Wharton, K. A. et al. Genetic analysis of the bone
morphogenetic protein-related gene, gbb, identifies
multiple requirements during Drosophila development.
Genetics 152, 629–640 (1999).
39. Khalsa, O., Yoon, J. W., Torres-Schumann, S. &
Wharton, K. A. TGF-β/BMP superfamily members,
Gbb-60A and Dpp, cooperate to provide pattern
information and establish cell identity in the Drosophila
wing. Development 125, 2723–2734 (1998).
40. Basan, M., Risler, T., Joanny, J. F., Sastre-Garau, X. &
Prost, J. Homeostatic competition drives tumor growth
and metastasis nucleation. HFSP J. 3, 265–272 (2009).
41. Nienhaus, U., Aegerter-Wilmsen, T. & Aegerter, C. M.
Determination of mechanical stress distribution in
Drosophila wing discs using photoelasticity. Mech. Dev.
126, 942–949 (2009).
42. Schluck, T. & Aegerter, C. M. Photo-elastic properties
of the wing imaginal disc of Drosophila. Eur. Phys. J. E
Soft Matter 33, 111–115 (2010).
43. Bittig, T., Wartlick, O., Gonzalez-Gaitan, M. &
Julicher, F. Quantification of growth asymmetries in
developing epithelia. Eur. Phys. J. E Soft Matter 30,
93–99 (2009).
44. Couso, J. P., Bishop, S. A. & Martinez Arias, A.
The wingless signalling pathway and the patterning
of the wing margin in Drosophila. Development
120, 621–636 (1994).
45. Piddini, E. & Vincent, J. P. Interpretation of the
wingless gradient requires signaling-induced self-
inhibition. Cell 136, 296–307 (2009).
46. Schwank, G. et al. Antagonistic growth regulation by
Dpp and Fat drives uniform cell proliferation. Dev. Cell
20, 123–130 (2011).
Following along the lines of reference 25, this report
proposes that, whereas in the centre of the imaginal
disc growth control is mediated by permissive
DPP, in the periphery growth is mediated by
DPP-independent Fat signalling.
47. Serrano, N. & O’Farrell, P. H. Limb morphogenesis:
connections between patterning and growth. Curr. Biol.
7, R186–R195 (1997).
48. Rogulja, D., Rauskolb, C. & Irvine, K. D. Morphogen
control of wing growth through the Fat signaling
pathway. Dev. Cell 15, 309–321 (2008).
Shows that cells can measure the differences in DPP
signalling levels between adjacent cells, as
monitored by the localization of Dachs at the
interfaces between cells, where differences in DPP
signalling are maximal. Also shows that Fat
signalling is affected by DPP.
49. Bohn, H. Extent and properties of the regeneration
field in the larval legs of cockroaches (Leucophaea
maderae) III. Origin of the tissues and determination
of symmetry properties in the regenerates. J. Embryol.
Exp. Morphol. 32, 81–98 (1974).
50. Rogulja, D. & Irvine, K. D. Regulation of cell
proliferation by a morphogen gradient. Cell 123,
449–461 (2005).
Proposal that growth is controlled by the differences
in signalling between adjacent cells, based on an
assay consisting of the activation of DPP in clones of
cells.
51. Kruse, K., Pantazis, P., Bollenbach, T., Julicher, F. &
Gonzalez-Gaitan, M. Dpp gradient formation by
dynamin-dependent endocytosis: receptor trafficking
and the diffusion model. Development 131,
4843–4856 (2004).
52. Bayly, R. & Axelrod, J. D. Pointing in the right direction:
new developments in the field of planar cell polarity.
Nature Rev. Genet. 12, 385–391 (2011).
53. Reddy, B. V. & Irvine, K. D. The Fat and Warts signaling
pathways: new insights into their regulation,
mechanism and conservation. Development 135,
2827–2838 (2008).
54. Ma, D., Yang, C. H., McNeill, H., Simon, M. A. &
Axelrod, J. D. Fidelity in planar cell polarity signalling.
Nature 421, 543–547 (2003).
55. Yang, C. H., Axelrod, J. D. & Simon, M. A. Regulation
of Frizzled by Fat-like cadherins during planar polarity
signaling in the Drosophila compound eye. Cell 108,
675–688 (2002).
56. Cho, E. & Irvine, K. D. Action of Fat, Four-jointed,
Dachsous and Dachs in distal-to-proximal wing
signaling. Development 131, 4489–4500 (2004).
57. Mao, Y. et al. Dachs: an unconventional myosin that
functions downstream of Fat to regulate growth, affinity
and gene expression in Drosophila. Development 133,
2539–2551 (2006).
58. Bennett, F. C. & Harvey, K. F. Fat cadherin modulates
organ size in Drosophila via the Salvador/Warts/ Hippo
signaling pathway. Curr. Biol. 16, 2101–2110
(2006).
59. Cho, E. et al. Delineation of a Fat tumor suppressor
pathway. Nature Genet. 38, 1142–1150 (2006).
60. Huang, J., Wu, S., Barrera, J., Matthews, K. & Pan, D.
The Hippo signaling pathway coordinately regulates cell
proliferation and apoptosis by inactivating Yorkie, the
Drosophila homolog of YAP. Cell 122, 421–434
(2005).
61. Neto-Silva, R. M., de Beco, S. & Johnston, L. A.
Evidence for a growth-stabilizing regulatory feedback
mechanism between Myc and Yorkie, the Drosophila
homolog of Yap. Dev. Cell 19, 507–520 (2010).
62. Nolo, R., Morrison, C. M., Tao, C., Zhang, X. &
Halder, G. The bantam microRNA is a target of the
Hippo tumor-suppressor pathway. Curr. Biol. 16,
1895–1904 (2006).
63. Silva, E., Tsatskis, Y., Gardano, L., Tapon, N. &
McNeill, H. The tumor-suppressor gene fat controls
tissue growth upstream of expanded in the Hippo
signaling pathway. Curr. Biol. 16, 2081–2089 (2006).
64. Thompson, B. J. & Cohen, S. M. The Hippo pathway
regulates the bantam microRNA to control cell
proliferation and apoptosis in Drosophila. Cell 126,
767–774 (2006).
65. Willecke, M. et al. The Fat cadherin acts through the
Hippo tumor-suppressor pathway to regulate tissue
size. Curr. Biol. 16, 2090–2100 (2006).
66. Oh, H. & Irvine, K. D. Cooperative regulation of growth
by Yorkie and Mad through bantam. Dev. Cell 20,
109–122 (2011).
67. Grusche, F. A., Degoutin, J. L., Richardson, H. E. &
Harvey, K. F. The Salvador/Warts/Hippo pathway
controls regenerative tissue growth in Drosophila
melanogaster. Dev. Biol. 350, 255–266 (2011).
68. Sun, G. & Irvine, K. D. Regulation of Hippo signaling
by Jun kinase signaling during compensatory cell
proliferation and regeneration, and in neoplastic
tumors. Dev. Biol. 350, 139–151 (2011).
69. Moreno, E. & Basler, K. dMyc transforms cells into
super-competitors. Cell 117, 117–129 (2004).
70. Moreno, E., Basler, K. & Morata, G. Cells compete for
Decapentaplegic survival factor to prevent apoptosis in
Drosophila wing development. Nature 416, 755–759
(2002).
71. Alon, U. An Introduction to Systems Biology: Design
Principles of Biological Circuits (eds Etheridge, A. M.
et al.) (Chapman & Hall/CRC Press, Taylor and Francis
Group, London, 2007).
72. Shoval, O. et al. Fold-change detection and scalar
symmetry of sensory input fields. Proc. Natl Acad. Sci.
USA 107, 15995–16000 (2010).
73. Barkai, N. & Leibler, S. Robustness in simple
biochemical networks. Nature 387, 913–917
(1997).
74. Friedrich, B. M. & Julicher, F. Chemotaxis of sperm
cells. Proc. Natl Acad. Sci. USA 104, 13256–13261
(2007).
75. Bargmann, C. I. Chemosensation in C. elegans.
WormBook 25 Oct 2006 (doi:10.1895/
wormbook.1.123.1).
76. Levchenko, A. & Iglesias, P. A. Models of eukaryotic
gradient sensing: application to chemotaxis of
amoebae and neutrophils. Biophys. J. 82, 50–63
(2002).
77. Goentoro, L., Shoval, O., Kirschner, M. W. & Alon, U.
The incoherent feedforward loop can provide fold-
change detection in gene regulation. Mol. Cell 36,
894–899 (2009).
78. Ma, W., Trusina, A., El-Samad, H., Lim, W. A. &
Tang, C. Defining network topologies that can
achieve biochemical adaptation. Cell 138, 760–773
(2009).
79. Hopyan, S., Sharpe, J. & Yang, Y. Budding behaviors:
growth of the limb as a model of morphogenesis.
Dev. Dyn. 240, 1054–1062 (2011).
80. Boehm, B. et al. The role of spatially controlled cell
proliferation in limb bud morphogenesis. PLoS Biol.
8, e1000420 (2010).
81. Dobbing, J. & Sands, J. Quantitative growth and
development of human brain. Arch. Dis. Child. 48,
757–767 (1973).
82. Hornbruch, A. & Wolpert, L. Cell division in the early
growth and morphogenesis of the chick limb. Nature
226, 764–766 (1970).
83. McClure, K. D. & Schubiger, G. Developmental analysis
and squamous morphogenesis of the peripodial
epithelium in Drosophila imaginal discs. Development
132, 5033–5042 (2005).
84. Widmann, T. J. & Dahmann, C. Dpp signaling promotes
the cuboidal-to-columnar shape transition of
Drosophila wing disc epithelia by regulating Rho1.
J. Cell Sci. 122, 1362–1373 (2009).
85. Neufeld, T. P., de la Cruz, A. F., Johnston, L. A. &
Edgar, B. A. Coordination of growth and cell division
in the Drosophila wing. Cell 93, 1183–1193
(1998).
86. Thompson, B. J. Developmental control of cell growth
and division in Drosophila. Curr. Opin. Cell Biol. 22,
788–794 (2010).
87. Resino, J. & Garcia-Bellido, A. Drosophila genetic
variants that change cell size and rate of proliferation
affect cell communication and hence patterning.
Mech. Dev. 121, 351–364 (2004).
88. Milan, M., Campuzano, S. & Garcia-Bellido, A.
Developmental parameters of cell death in the wing
disc of Drosophila. Proc. Natl Acad. Sci. USA 94,
5691–5696 (1997).
Acknowledgements
The authors thank A. Genevet for comments on the manu-
script. This work was supported by the Max-Planck-
Gesellschaft, the Swiss National Science Foundation, grants
from the Swiss SystemsX.ch initiative and LipidX-2008/011,
and by an ERC advanced investigator grant to M.G.-G.
Competing interests statement
The authors declare no competing financial interests.
FURTHER INFORMATION
Frank Jülicher’s homepage: http://www.mpipks-dresden.mpg.
de/mpi-doc/julichergruppe
marcos gonzalez gaitan’s homepage: http://cms.unige.ch/
sciences/biochimie/-Marcos-Gonzalez-Gaitan-Lab-.html
SUPPLEMENTARY INFORMATION
See online article: S1 (table)
ALL LINKS ARE ACTIVE IN THE ONLINE PDF
REVI EWS
604 | SEPTEMBER 2011 | VOLUME 12 www.nature.com/reviews/molcellbio
REVI EWS
© 2011 Macmillan Publishers Limited. All rights reserved

 2a). 2c. with an intensity profile (bottom) as a function of the distance from the source. the decay length of the gradient is proportional to the width of the tissue. Disc growth does not distort the DPP gradient because the gradient renews itself faster than the tissue grows24.d). it scales with tissue size. the decay length is proportional to the width of the compartment — the gradient scales (FIG. and it will be interesting to understand how the wiring of the signal transduction network mediates the conservation of these two properties at the end of the pathway. Instead. for the observed increase in the gradient amplitude in the growing tissue (FIG. FIG. MAD phosphorylation is negatively regulated by DAD. divide at roughly the same rate17. gradients from all stages of development have the same shape24 (FIG. Decapentaplegic (DPP) source and gradient in the tissue (green. Below. because the clones are exposed to different DPP concentrations they express different target genes (blue and pink). Because homogeneous growth means that cells (and their lineages) do not change their relative position during growth (FIG. Controlling the DPP gradient during wing disc growth. forming a concentration gradient 6. and x the distance from the source. 2c). DPP gradient scaling during wing disc growth. there is a corresponding increase of cellular DPP concentration: Ccell and C0 are proportional in all cells.26 (FIG.F O C U S O N m O R p h OR EE N E W S g VI SI λ Figure 1 | The Decapentaplegic gradient.24. such as Dad and spalt. during development. This means that. Amplitude The maximum concentration of a protein in the target region. Its value depends on how fast the molecules diffuse and are degraded (less degradation equals a higher decay length).7. scaling is mainly due to a decrease in the DPP degradation rate — empirically.e). 2c). Both the increase of the gradient amplitude C0 and the scaling of the decay length λ can be explained as a consequence of this decrease of the DPP degradation VOLUME 12 | SEPTEMBER 2011 | 595 Gradient scaling If the gradient expands at the same rate as the tissue. 2e). such as the vertebrate limbs or the embryonic spinal chord (BOX 1). In particular. are in the tissue. a particular relative position in the tissue always has the same relative concentration of the morphogen24 (C/C0 = constant. 1b). 2a). amplitude refers to the concentration at the DPP source boundary. as the tissue grows. the DPP degradation rate halves24. In the case of Decapentaplegic (DPP). NATURE REVIEWS | MOLECULAR CELL BIOLOGY © 2011 Macmillan Publishers Limited. The transduction of the signal involves successive nonlinear amplifications24. 2b. Therefore. b | Magnified view of cells in the posterior compartment and the DPP gradient (top). gradient scaling Decay length A measure for the spatial range of a protein gradient (how far it reaches into the tissue): the position at which the concentration of the protein is a fixed fraction of its amplitude (C0/e. c | The DPP signalling pathway. BRK and phosphorylated MAD regulate DPP target genes. In other words. Gradient scaling is not only observed for the DPP ligand concentration gradient. implies that a particular cell always experiences the same relative DPP concentration as the tissue grows24 (Ccell/C0 = constant). Note that growth is approximately homogeneous: two clones (blue and pink) at different positions in the gradient have grown to the same size. all cells contribute approximately equally to the tissue: tissue proportions scale. the gradient is characterized by its amplitude (C0) and decay length (λ). Gradient scaling means that. the DPP signalling levels of all cells in the tissue increase over time by the same percentage as the gradient amplitude (FIG. The DPP gradient can be characterized by its amplitude (C0) and its decay length (λ)27 (FIG. DPP is expressed in a central stripe of cells (the DPP source) and spreads in the tissue. This means that the position of cell clones relative to the width of the tissue remains approximately constant during development (FIG.b). 1a).25. in which e is Euler’s number). here the gradient is only shown in the posterior target compartment) are shown. An important implication of this invariant relative concentration profile is that. Phosphorylated MAD also interacts with the co-transcriptional activator Yorkie to regulate the transcription of the growth-promoting microRNA bantam. L indicates target tissue width. when the number of cells in the tissue doubles. Quantification of the amplitude and decay length of the DPP gradient in the posterior compartment of the wing disc showed that the DPP gradient expands during disc growth24 (FIG. where it represses the transcription of the repressor brinker (brk). In other words. MAD can then bind MEDEA and accumulates in the nucleus. we discuss how the DPP concentration gradient changes during wing disc growth. BRK and DAD — is different to that of DPP ligand concentration24. Similar observations have been made in other morphogenetic growth systems. The transcription factor MAD is phosphorylated and activated following binding of DPP to its receptor Thickveins (TKV). the range of the gradient grows proportionately. However. gradient scaling and the increase of the amplitude are also observed for a downstream DPP signalling ‘readout’ (Dad tanscription)24. The compartment boundary (green line). All rights reserved . a | The imaginal wing disc of Drosophila melanogaster.18. when the relative gradient profiles are normalized to tissue size. however. as well as on the number of DPP-producing cells (the source width) and on the rate of diffusion. Gradient scaling is therefore not a direct consequence of growth. This means that the spatial activity pattern of the pathway components — for example. Its value depends on DPP production and degradation rates. This is not trivial. but also for the DPP signalling gradient 24. 2a. phosphorylated MAD.

the growth rate decreases during development and is fairly homogeneous in space24. distance from the morphogen source. The measurement of disc growth properties is therefore simple. The cell density (ρ) in disc epithelia increases slightly during development.50. rapidly diffusible molecule. A key question. When the imaginal disc grows. x. 3). the expander antagonizes DPP degradation. with permission.nature. so that overall the average cell volume is approximately constant (O. Because the increase in cell density is much smaller than the increase in cell number (N) (threefold versus 1. 82). like in vertebrate systems. but they are relatively small and confined to the boundary regions). part a. the tissue area (A) is roughly proportional to the number of cells (see the figure. Indeed. cell number = cell density × tissue area (N = ρA) As a result. from REFS 80. as well as more recently with quantitative detail by the Sharpe group80 (see the figure. staining with 5‑bromodeoxyuridine (BrdU) and phosphorylated histone H3 (pH3) (used to visualize proliferation and cell division) as well as clonal analyses suggest that growth is fairly homogeneous17–19. First.com/reviews/molcellbio © 2011 Macmillan Publishers Limited.43). Molecules regulating morphogen degradation and thereby expanding a gradient in response to growth have been termed expanders28. rate during development (see Supplementary information S1 (table)). cellular DPP levels at the edge of the disc (FIG.000-fold24. Part a of the figure is modified. the mechanisms by which an expander might control DPP degradation are referred to as scaling mechanisms. the average cell growth rate and proliferation rate (gN) are roughly the same and correspond to the effective tissue growth rate (gA). morphogen gradients seem to control growth4. 596 | SEPTEMBER 2011 | VOLUME 12 Two types of scaling mechanisms have been proposed An expander mechanism based on expansion– repression feedback was originally proposed by Ben-Zvi and Barkai28–30. whereas imaginal discs are epithelia (2D sheets).W. part d) and can therefore be inferred directly from measurements of the disc area. vertebrate limb systems grow as mesenchymal masses of cells (which are three-dimensional (3D)).. like in the vertebrate limbs or the embryonic spinal chord. L. This was described in the classical paper by Hornbruch ρ and Wolpert in 1970 (REF. the average cell doubling time is equal to the area doubling time (see the figure. tissue width. the expression of which is repressed above a certain DPP concentration level. unpublished observations). this reduces the analysis of their growth to a 2D problem. is how tissue growth causes the decrease of the DPP degradation rate. All rights reserved . This implies that all cells have the same growth rate (gcell). Furthermore.82.84. Finally. which is approximately equal to the average tissue growth rate gA (see the figure.79. therefore.25.REVIEWS Box 1 | Tissue growth properties The wing imaginal disc of Drosophila melanogaster displays growth properties similar to those of many other developmental systems. The expander is a long-lived.24. Below. which shows growth in a chick limb bud (left) and a mouse limb bud (right)). The height of cells along the z axis increases slightly as the cell density increases83. part b).80–82 (some heterogeneities are apparent. Therefore. cell growth and proliferation are coordinated in wing disc cells — cells divide when they have doubled their volume85–87. www. Second. In this model. apoptosis levels are low and fairly uniform88. part c). However.

suggesting that gradient scaling might be due to a dilution-type mechanism. However. The expander then diffuses in the tissue and decreases the DPP degradation rate. This. in the expansion–repression mechanism. no molecular equivalent of an expander has been identified. This implies that the ratio of gradient decay length to tissue width (λ/L) is constant during development (in other words. as repression of the expander directly depends on the DPP gradient along this axis (FIG.30. which itself is involved in DPP turnover and diffusion32–35. the degradation rate is inversely proportional to the number of cells and. the number of cells doubles (e). anisotropic tissue growth could help distinguish between the two types of scaling mechanism (see Supplementary information S1 (table)). implies that DPP concentration experienced by a particular cell (Ccell) is proportional to C0. Pentagone (PENT). accompanied by disc growth28. 3b. diffusible and repressed by DPP signalling (and therefore transcribed in regions of low DPP concentration)31. It has yet to be determined whether DPP gradient scaling is modified in pent mutants. λ is proportional to L (d)). seems to decrease VOLUME 12 | SEPTEMBER 2011 | 597 NATURE REVIEWS | MOLECULAR CELL BIOLOGY © 2011 Macmillan Publishers Limited. but not width scaling. the area of the tissue. so the gradient scales with the square root of the tissue area but not necessarily with the tissue width (FIG.F O C U S O N m O R p h OR EE N E W S g VI SI λ λ × × × λ α α α α × Figure 2 | Decapentaplegic gradient properties during growth. was the same in the different experimental conditions24. by extension. By contrast. and cells at certain relative positions (blue and pink) always see the same relative DPP concentration (Ccell/C0) (c). rather than expansion–repression. 3a). blue and pink. stay constant during development (two clones. So far. are shown). horizontal) than another (for example. thus expanding the DPP concentration gradient. 3b). An obvious target for an expander in the dilution mechanism would be DPP receptor levels: the expander could affect the effective intracellular DPP degradation rate by modulating DPP receptor levels through receptor ubiquitylation and internalization or through changes in endosomal dynamics. a molecule that is conserved from flies to vertebrates31. for example. Therefore. The TKV degradation rate. the gradient does not necessarily scale with tissue area but scales with the width of the tissue in the direction of the gradient. shows some of the features required of an expander: it is secreted. and therefore the relative position (distance from the source/tissue width (x/L)) of cells and their lineage. homogeneous growth implies that tissue proportions. 3a). The relative increase in concentration over time is the same for all cells (e) and empirically correlates with growth: whenever the cellular concentration increases by a percentage α. All rights reserved . b–e | Gradient scaling implies that the DPP concentration profiles at different times of development (b) collapse onto the same shape when relative concentration (C/C0. Indeed. in turn. In this expander-dilution mechanism. analysis of DPP gradient scaling in different anisotropic growth conditions showed that area scaling. fall below the threshold and cells start expressing the expander (FIG. in which C0 is the gradient amplitude) and relative position (x/L) are considered (c). The second type of scaling mechanism is based on reducing the DPP degradation rate in response to cell division events24. vertical) in the plane of the epithelium. by cell-autonomous dilution of a long-lived expander that promotes DPP degradation and might also diffuse in the tissue (FIG. Anisotropic tissue growth Directionally dependent growth of a tissue: more growth along one axis (for example. a | The Decapentaplegic (DPP) gradient expands in growing wing imaginal discs. like the DPP degradation rate. PENT antagonizes DPP degradation by interacting with the heparan sulphate proteoglycan Division abnormally delayed (DALLY)31. This can be achieved. This results in an increase in DPP levels at the edge of the disc and repression of the expander. see Supplementary information S1 (table)).

 4). The expander promotes degradation of the morphogen. Whether this increase in mechanical stress is large enough to be limiting for cell growth is not clear.42.40. such as mechanical stress or the presence of other growth factors. we discuss the four major proposed morphogenetic growth control models: growth control by mechanical feedback. reduces morphogen degradation rates (k) and thereby expands the morphogen gradient to fit into the new tissue width (L). The idea that mechanical stress limits cell growth led to the formulation of models in which DPP is merely permissive for growth (that is. Indeed. and. therefore. increases during development. it is unlikely that the scaling of the signalling gradient happens only at the level of the ligand concentration gradient. Increased transcription of Dad over time even in the absence of DPP and the BRK repressor might indicate that not only DPP degradation but also target gene production is sensitive to tissue size.nature. the expander is diluted in response to cell division events. Growth control by the DPP gradient The previous section considered how growth affects the DPP gradient. Experimental support for the role of mechanical stress in growth control comes from measurements of cell compression in the centre of the disc41. Proliferation in the periphery then compresses cells in the centre and limits their growth20–22. which suggest that central cell compression increases during development 41. absolute DPP levels alone cannot control homogeneous proliferation. It is transcribed (orange) below a threshold morphogen concentration (green). In this model. As cell compression. or that the mutations used here might not represent complete loss of functions. where they can divide easily even with low growth factor levels because they experience low mechanical stress levels 20–22. proliferation is stimulated by a central point source of growth factors. an overall increase in the transcription of the DPP target gene Dad was still observed during development24. generated by the intersection of line sources of the two diffusible morphogens.24 or temporal changes in DPP concentration or signalling levels24 to set their growth rate or cell cycle length. www. Thus. In this model. b | Expander-dilution mechanism. this implies that the gradient decay length (λ) is proportional to the tissue width L. the buildup of mechanical stress distributions in the tissue can lead to a situation in which growth is homogeneous even when there is a steep DPP gradient. Alternative explanations are that a second BMP-type morphogen. when tissue growth causes cells in the periphery to experience morphogen levels below this threshold. Proliferation forces cells to move out of the central growth factor zone into the periphery. DPP and Wingless. In epithelia. 2). However. The expander (pink) antagonizes morphogen degradation. In a simple model based on mechanical feedback. growth control by complementary inhibition and growth control based on spatial differences in DPP levels between neighbouring cells or on temporal changes in DPP signalling. in experiments in which both endogenous dpp and its downstream target brk (FIG. By contrast. this implies that the gradient decay length λ is proportional to the square root of the tissue area A. the degradation rate k is inversely proportional to the tissue area A. the initial degradation rate ki decreases in an inversely proportional manner to L2. how does the DPP gradient influence growth control in the imaginal disc? Proliferation is approximately spatially uniform and decreases over time (BOX 1). 4a). in the centre of the disc20–22. a | Expansion–repression mechanism. 598 | SEPTEMBER 2011 | VOLUME 12 © 2011 Macmillan Publishers Limited. cells could interpret spatial differences23. it promotes growth when its levels are above a certain threshold) and mechanical stress acts as a long-range signal to instruct cell growth rates20–22.com/reviews/molcellbio λ λ Figure 3 | Possible mechanisms for morphogen gradient expansion and scaling. so it would be interesting to pursue this line of investigation further. Cells divide above a certain threshold of growth factor concentration. the effective tissue growth rate decreases overall20–22 (FIG. can activate MAD and also contribute to Dad transcription36–39. high levels of mechanical stress can physically limit tissue growth by locally limiting cell growth22. Another possibility is that additional factors. they produce a long-lived expander that diffuses in the tissue. Therefore. When the tissue grows. DPP concentration and signalling levels are graded in space and increase over time (FIG. In other words. Growth control based on mechanical feedback. and thus mechanical stress. act in concert with DPP to make growth homogeneous (FIG. compression by surrounding cells). cells adjust their height and apical surface area in response to mechanical stress (for example. Below.REVIEWS during development 24. morphogen degradation rates (k) decrease and the morphogen gradient expands. Finally. This is known as area scaling (see also Supplementary information S1 (table)). But. This is known as width scaling. All rights reserved . 1c) are mutated. GBB.

wing discs in which BRK is ubiquitously expressed do not grow. yellow). Thus. the output of the pathway manifested in target gene expression was assumed to be maximal in brk-mutant cells4. The time it takes to reach a 50% increase in concentration is equal to the cell cycle length and is longer at late stages of development because the increase in concentration becomes smaller as the tissue grows. some assumptions made by mechanical growth models do not seem to apply to the wing imaginal disc. In fact. indicating that mechanical stress levels in the disc may not directly control or significantly limit cell growth rates. BRK is a repressor of DPP target genes and is repressed by DPP signalling.15) (FIG. proliferation would (probably) not be uniform. high levels of BRK inhibit growth. c | Spatial model. Another permissive growth model proposed that DPP allows growth by suppressing BRK. d | Temporal model. and growth would not stop. A mechanical model based on modified assumptions might still provide an explanation for the growth properties of the wing imaginal disc. A line source causes different stress and proliferation patterns than a point source because there is no longer a peripheral ring of stretched. a systematic experimental analysis of stress distributions in the disc at different times of development is needed to get a clearer picture of the possible role of mechanical stresses as regulators of proliferation. inhibiting their growth. disc cells (bottom) can sense relative spatial differences in DPP (green) (and probably other factors) through the non-conventional myosin Dachs (yellow). an integral component of the DPP pathway (FIG. whereas compression inhibits growth. In this model. cells would not move out of this zone into the periphery. the observation that in brk-mutant discs cells in lateral regions proliferate more than cells in the centre (although they are supposedly experiencing the same DPP signalling levels) implied that growth in central and lateral regions of the disc had to be regulated independently and that the growth role of DPP in the centre is just permissive: the gradient is irrelevant for growth in the centre25. First. As a result. the gradient is irrelevant. Therefore. Thus. b | Complementary inhibitor model.45. the specific form of such a modified model is currently unclear. proliferating cells that compresses the cells in the central growth factor region. VOLUME 12 | SEPTEMBER 2011 | 599 © 2011 Macmillan Publishers Limited. it is therefore expressed as an ‘anti-gradient’ (REFS 12. For a scaling gradient. in addition. This model assumes that gradient scaling does not occur. Growth factors (morphogens) and low stress (stretching) stimulate division. Decapentaplegic (DPP) allows growth by repressing Brinker (BRK) in the central region of the disc. Because BRK acts as a repressor in the DPP pathway. Growth control by complementary inhibition. DPP signalling levels were assumed to be spatially uniform in a brk-mutant disc4. the relative slope decreases as the tissue grows.46.22. The relative slope of an exponential morphogen concentration gradient could control cell division (top). Another assumption in the original model was that the growth factor gradient does not scale21. This model requires that the DPP and Fat activity gradients are established in an independent manner. The spatial profile of the Fat activity gradient (top) has not been measured directly but is inferred from the growth phenotype of fat mutants (bottom). brkmutant discs overgrow in lateral regions (where BRK is normally expressed) and.13. data suggest that Wingless does not control growth in the wing disc44. BRK plays a major part in DPP signal transduction and the downstream regulation of growth. area growth does not seem to be constrained by the increase in cell compression: the tissue area increases substantially even late in development. causing them to divide (low stress. x indicates distance from the morphogen source. wing discs that are deficient in both DPP and BRK have a growth phenotype similar to brk mutants25. By contrast. so the growth factor (DPP) is produced by a line source rather than a point source. 1c). the central growth factor region would simply expand as the tissue grows. 4b). In any case. when the cell density no longer changes24. although tkv-mutant or Madmutant cells do not grow. Large spatial differences in DPP levels lead to cell division.43. cells mutated for both TKV and BRK or for both MAD and BRK do grow 12–14. NATURE REVIEWS | MOLECULAR CELL BIOLOGY . Fat activity controls growth in a similar manner at the edge of the disc. 1c). Consistent with these observations. For a scaling gradient such as DPP. a | Mechanical stress model: the proliferation of cells in the central region in response to a morphogen (green) stretches the cells in the periphery. The proliferation of cells in the periphery compresses the cells in the centre. As a result. strikingly.25. and mutation of BRK rescues the growth of cells lacking DPP signalling upstream of BRK (FIG. consistent with decreasing growth rates.46. As a consequence.F O C U S O N m O R p h OR EE N E W S g VI SI Furthermore. All rights reserved α α Figure 4 | Models for growth control by morphogen gradients. The proposal for a permissive role of DPP in growth through the repression of BRK was based on the observation that.25. Growth is controlled by a percentage increase in morphogen concentration (of α = 50%) over time (morphogen concentration is illustrated by the number of dots in each cell). cell division would not occur when the DPP concentration gradient is uniform. However. there would be no compression of cells in the central region. and the shape of the gradient is irrelevant.

However. graded signalling in brk mutants could. This proposal was based on the growth phenotype of fat-mutant wing imaginal discs. overgrowth of imaginal discs with increased DPP signalling levels could be suppressed by increasing Fat activity 46. If they are not — for example.25 (FIG. Furthermore. factors other than Fat could still regulate growth in a manner complementary to BRK. it was hypothesized that growth in the periphery is controlled by mechanical stress20. This would explain homogeneous www. if Fat activity is downstream of DPP — then there is no complementary inhibition. their expression is not spatially uniform12. however. Note that the expression pattern and range of spalt (white arrow) is similar in both the wild type and the brk dpp mutant. One of the first instructive models proposed that proliferation could be controlled by the local slope of the DPP gradient: neighbouring cells measure this slope through their differences in DPP levels. the activity gradient of Fat. 25 © (2008) The Company of Biologists. because in brk dpp double mutants. which overgrow 4. 5). 5e). Growth control based on spatial differences in DPP signalling. in which cells measure either spatial differences or temporal changes in DPP signalling levels to set their growth rate or cell cycle length. the slope flattens during the growth phase. indicating that the absolute output of the DPP signalling pathway downstream of BRK cannot directly determine the growth rate (although one cannot completely exclude the possibility that the low DPP signalling levels are still above a permissive threshold).15.nature. which lack phosphorylated MAD. in the absence of a DPP signalling pattern in space (a spatially uniform signal) and of changes in time (a maximal signal). 4b). in principle. two proteins regulating Fat activity. brk mutants (d) and brk dpp double mutants (e) based on quantifications of red fluorescent protein expression.21 (see above). cells still proliferate.25 (FIG. and the localization of Dachs.47. a protein 600 | SEPTEMBER 2011 | VOLUME 12 . as monitored with a Dad transcriptional reporter. Strikingly. who assumed that there is a linear gradient that stretches as the tissue grows. This model was first proposed by Day and Lawrence23. c–e | DPP signalling levels (graphs) and growth phenotypes (insets) in wild-type imaginal discs (c). because the DPP signalling output in brk dpp mutant conditions is neither uniform in space nor constant in time24.46. the fact that BRK and Fat activity are graded in space is not relevant for growth regulation. This indicates that loss of repression by BRK does not lead to maximum DPP signalling.REVIEWS responding to Fat activity 48. Independent establishment of two exactly complementary gradients is also problematic because it is difficult to ensure robustness. driven by the Dad enhancer24. it was proposed that homogeneous growth could be regulated by an additional growth inhibitory gradient that is complementary to the inhibitory BRK gradient 46. DPP is just permissive for growth. However. analysis of the Dad enhancer revealed parallel inputs into the Dad regulatory regions from BRK and MAD.13. although there is ectopic expression. because our current model of the pathway (FIG. 5). growth of Mad brk mutant or tkv brk mutant cells12–14 is not necessarily uncovering a scenario in which DPP is merely permissive for growth. Therefore. It is difficult to reconcile these experimental findings with the complementary inhibition model. In any case. All rights reserved µ µ Figure 5 | The role of BRK in signal transduction and growth. the expression of spalt is not spatially uniform in brk dpp double mutants. but instead growth depends only on DPP signalling. This reveals that our understanding of the DPP signalling pathway is incomplete. Indeed.25. The growth phenotypes of these mutants might also be consistent with instructive growth models. they are also more than tenfold lower in brk dpp mutants than in brk mutants and wild-type cells (not drawn to scale here). This does not seem to be the case.b | Decapentaplegic (DPP) target gene spalt (green) in a wild-type imaginal disc (a) and in a brinker (brk) and dpp double mutant (b). as long as the shape of the gradients is perfectly complementary. However. be due to BRK-independent signalling mediated by MAD.24. Images in parts a and b are modified. 1c) would predict spatially and temporally uniform DPP target gene expression in the absence of both BRK and phosphorylated MAD.com/reviews/molcellbio © 2011 Macmillan Publishers Limited. In other words. In the complementary inhibitor model. note also that. growth still occurs: thus. although DPP target genes (such as spalt and Dad) are ectopically expressed in brk-mutant clones of cells and brk-mutant discs. 4b). More recently. DPP signalling levels are not spatially uniform and they increase over time. however. DPP signalling affects the expression of Dachsous and Four-jointed. spatially uniform in brk-mutant discs.25 (FIG. although absolute levels of DPP output. rather than BRK alone16. therefore. and proliferation stops when the slope reaches a minimal steepness23 (FIG. for this model it is crucial that the two gradients are independent 46. specifically. 4c). which is complementary to that of brk mutants46 (FIG. Indeed. are much lower in the brk dpp mutant (FIG. with permission from REF. a. L indicates tissue width and x distance from the morphogen source. At first. which is involved in growth mediated by the Hippo pathway (see below) (FIG. The key assumption of this model is that DPP signalling levels and the expression of target genes is maximal in brk-mutant cells and. the expression of target genes is also not spatially uniform24. This indicates that the output of the DPP signalling pathway downstream of BRK is not uniform in space because otherwise spalt should be uniformly expressed in the entire disc. 5).

the C′/C threshold below which cells stop dividing at the end of the growth phase) differed between different tissues and mutant conditions24.28. 2e). It is easy to show that these gradient properties correlate with the growth rate. It also depended on how distant the clones were from the source: clones far away from the source initially had lower DPP signalling levels. it was found that when DPP signalling levels have increased by a percentage α (in this case α = 50%) from the beginning of the cell cycle. the spatial derivative is denoted with a prime symbol. This further supports the temporal model as a plausible mechanism for growth control. and therefore. it takes cells a longer time to reach α. Regeneration experiments in cockroaches supported this hypothesis49. 2b). All rights reserved Spatial derivative The spatial difference in a quantity (for example. Because of this. cell proliferation rates are also position dependent and correlate with the position-dependent Ċ/C 24. NATURE REVIEWS | MOLECULAR CELL BIOLOGY . In other words. However. In this experiment. how cells could measure relative spatial differences or temporal changes in DPP signalling levels is much less intuitive. and the cell proliferation rate is high (corresponding to a short cell cycle length). It is also important to note that the parameters of the spatial model (for example. even when DPP is ubiquitously expressed. but how could cells actually measure relative changes in concentration in space or over time? VOLUME 12 | SEPTEMBER 2011 | 601 © 2011 Macmillan Publishers Limited. for details. and. It was thought that. instead. 1a). DPP signalling levels increase faster or slower depending on position: unlike in wild-type discs. the DPP signalling levels in clone cells increased faster than in wild-type cells. however. Support for the temporal model came from experiments in which the velocity of the temporal increase of DPP signalling was manipulated exogenously by drugmediated induction of the expression of TKVQD in clones of cells. the gradient no longer scales. by the percentage by which the DPP concentration increases over time. α is reached quickly. like their growth rate. One way to directly test whether growth is controlled by spatial cues would be to create uniform DPP signalling in space. melanogaster wing discs carrying clones of cells that express a constitutively active form of TKV (TKVQD) and therefore have higher DPP signalling levels than surrounding cells19. In models in which DPP is merely permissive for growth (FIG. the cells divided when DPP signalling levels had increased by α = 50%24. the slope of the DPP gradient is actually position dependent and so cannot control position-independent proliferation rates. concentration) across one spatial unit (for example. Nevertheless. Wing discs in which brk is mutated or DPP is ubiquitously expressed overgrow 25. cells at the edge of such a clone would experience a big spatial difference in DPP signalling levels and should therefore divide more. position independent. the exponential approximation of the DPP gradient 26. How quickly cells attained this 50% increase depended on the concentration of the drug that determined the velocity of the increase in DPP signalling levels. If the gradient scales. 4d). normalized to C) (FIG. Therefore. later quantifications showed that. upon TKVQD expression. on closer inspection.25 (FIG. indicating a non-uniform DPP signalling profile24.F O C U S O N m O R p h OR EE N E W S g VI SI growth (a single slope in a linear gradient) and the slow-down of the growth rate (as the gradient becomes shallower during growth. Interestingly. relative spatial differences in signalling levels between neighbouring cells still correlate with their growth rates24. However.27.24. Indeed.51 (FIG. consistent with a temporal model. or C′/C (the spatial derivative of C. the spatial model remains a plausible mechanism for growth control. as a consequence. normalized to C) — are. However. No extra proliferation was observed at the border of clones19. which increases as the tissue grows24 (C′/C = 1/λ. cellular DPP signalling levels increase by the same percentage in all cells (FIG.b). 4d). when DPP levels increase more slowly (FIG. C′/C is the same for all the cells in the tissue at a given time and depends only on one global parameter. However. the value of C′/C is the same for all cells. The transient extra proliferation of cells exposed to big spatial differences in DPP signalling levels supports the idea that spatial cues have a role in growth control. However. clones that were away from the source were bigger than those closer to the source. At the end of development. the decay length λ. recent experiments showed that there is a transient additional proliferation effect on the cells at the border of clones. If DPP signalling levels increase quickly. for example ˙ dC / dt = C. this correlation is less clear in the brk mutant 24. By contrast. 4a. or Ċ/C (the time derivative of C. In an exponential gradient. Therefore. the percentage by which the DPP concentration decreases across a cell’s surface. in a temporal model. as had been assumed by Day and Lawrence in the first spatial model. The DPP gradient is not a linear gradient with a constant slope in space. the overgrowth phenotype is not inconsistent with a spatial model because. concentration) over time. An alternative to the spatial model proposes that cell cycle length is determined by relative temporal changes in cellular DPP levels. growth control based on the temporal model can account for non-homogeneous proliferation patterns observed in and around TKVQD clones and in imaginal discs in which DPP signalling is ubiquitously increased24. if spatial differences in DPP levels between cells were the driving force of proliferation. the miRNA bantam) and divide. its slope becomes smaller). the model was challenged by a mosaic analysis in D. cells divide24 (FIG. Time derivative The rate of change of a quantity (for example. which is consistent with the temporal growth model19. 2e). Measurement of C′/C or Ċ/C by cells. and therefore the cell cycle is longer (FIG. and. a cell). it is steeper close to the source and shallow away from it27. all cells have the same growth rate24 (see Supplementary information S1 (table)). observed only in a time-course experiment 50. for example dC / dx = C′. the time derivative is denoted with a dot symbol. there should have been less growth. Growth control based on temporal changes in DPP signalling. like the growth rate. that is. Importantly. According to the model. the expression of DPP targets was unexpectedly not uniform in these conditions. see supplementary information S1 (table)). Empirically. the relative increase in their DPP signalling levels was bigger. the relevance and output of DPP signalling is fairly easy to understand: cells above a certain threshold of DPP concentration would express DPP growth targets (for example.50 shows that relative spatial differences in DPP concentration — that is. Ċ/C is position dependent. C′/C decreases during development. Under these conditions. and this observation was thought to refute the model. 5).

The arrows shown do not indicate regulatory pathways but outcomes. an intriguing possibility is that this pathway could detect spatial discontinuities in the DPP gradient and in response regulate the elimination of these discontinuities by local upregulation of cell proliferation or apoptosis. which forms a proximo–distal gradient in the disc48. Thickveins. as well as the growth-promoting miRNA bantam53. when local levels of PCP factors in neighbouring cells can influence intracellular gradients of such factors52. The amount of Dachs at the interface seems to be independent of absolute DPP levels. Measurement of the relative temporal changes of signalling levels implies that adaptive responses are generated in the signalling pathway (reviewed in REF. when slow-growing clones with low DPP signalling levels are eliminated from the tissue69. melanogaster inhibitor of apoptosis 1 and MYC. 6a).55 (FIG.REVIEWS Hippo Figure 6 | Interaction between Decapentaplegic and the Fat–Hippo pathway. Cells respond to the Fat signalling gradient by differential intracellular activation of the non-conventional myosin Dachs48. A related phenomenon might occur during planar cell polarity (PCP). green) forms a gradient along the anterior–posterior axis (here the gradient is only shown in the posterior compartment to the right). through titration of active Dachs and Yorkie levels.58–65 (the last in cooperation with the DPP transcription factor MAD66). TKV. proliferation and apoptosis53.or BRK-expressing clones50. cells must compare signalling levels at different locations to each other and with respect to their own signalling levels. Yorkie regulates the expression of growth regulatory targets. By contrast.70. This also corresponds to the cell–cell interface at which DPP (and Wingless) signalling levels are higher. Dachsous (blue) forms a gradient along the proximo–distal axis (it is high along the edges and decreases towards the centre). D. and even the phenomenon of cell competition. 71). a | Decapentaplegic (DPP. 6). a kinase that reduces Yorkie activity (FIG. Dachs relocalized to the new interface facing higher DPP signalling levels48.com/reviews/molcellbio 602 | SEPTEMBER 2011 | VOLUME 12 © 2011 Macmillan Publishers Limited. cells do indeed seem to be able to measure DPP signalling differences between adjacent cells using the Fat–Hippo pathway 48 (FIG. To measure relative spatial differences in DPP. by which cells acquire information about their orientation within the tissue in the plane of the epithelium. BRK. In the wing disc.55–57: Dachs accumulates at the cell–cell interface that is facing the lower levels of Dachsous (FIG. the perturbation of growth observed in and around TKVQD. b | The DPP pathway and the Fat signalling pathway responding to Dachsous interact. suggesting that cells may sense relative spatial differences in DPP signalling levels48. distinct from apical–basal polarity.54. All rights reserved . Brinker. Together with phosphorylated MAD. The ligand for Fat is the protocadherin Dachsous. When the spatial pattern of DPP signalling was changed. Such a mechanism might be responsible for regeneration in response to tissue damage67. Dachs regulates the activity of Yorkie by inactivating Warts. 6b). High levels of Dachsous (right) activate the Fat signalling pathway. leading to the activation of Yorkie. An adaptive response allows cells to measure relative increases of a signal.nature. Planar cell polarity A mechanism of cellular organization. for example by generating clones of cells expressing TKVQD or BRK. on the side of the cell where Dachsous levels are low (left) — which can be caused by an ectopic increase in DPP signalling levels in clones of cells — Dachs is active. inhibiting Dachs and Yorkie (through the kinase Warts).68. Yorkie regulates the transcription of the growth target bantam. The dashed arrow indicates an indirect interation. x indicates distance from the morphogen source. This signalling pathway regulates organ size by controlling cell growth. Although the quantitative relationships between the Fat–Hippo pathway and the DPP gradient are still obscure. because the signalling system adapts to ambient concentrations of the signal and increasing www. such as Cyclin E. 6b).

521–525 (2007). Nature Rev. & Affolter. Perspect. Lecuit. Mol. Dynamics of Dpp signaling and proliferation control. 2. B. Development 129. 25. K. B. E. K. as proposed in reference 30. Carroll. Biol. Development 135. & Basler. M. Development 127. Interestingly. Nature 450. 2. C. Gradient formation of the TGF-β homolog Dpp. & Lawrence. Aegerter. based on the assumption that brk mutants (or brk dpp double mutants) show uniform DPP signalling. J. A characterization of the effects of Dpp signaling on cell growth and proliferation in the Drosophila wing. Mechanical feedback as a possible regulator of tissue growth. Natl Acad. is that cells could measure both C′/C and Ċ/C. A. 17.. A. Muller. Affolter. Hafen. Cell 80. Garcia-Bellido. Schwabedissen. temporal and mechanical cues — to orchestrate patterning and growth. Cell 103. Two distinct mechanisms for long-range patterning by Decapentaplegic in the Drosophila wing. USA 102. M. K. NATURE REVIEWS | MOLECULAR CELL BIOLOGY © 2011 Macmillan Publishers Limited. Ruberte. V. Mech. & Cohen. A. brinker is a target of Dpp in Drosophila that negatively regulates Dpp-dependent genes. that control morphogenetic growth in this model system are already known. 189–192 (2007). P. Cell 5. R.. A. & Tabata. 1196–1207 (2009). not discussed in more detail here. A. H. H. S. Natl Acad. Dev. 28.. Regulation of organ growth by morphogen gradients... Parameters of the wing imaginal disc development of Drosophila melanogaster. S. A. et al. T. A. & Barkai. 1154–1159 (2011). 269. 15. Analyses quantitatively both the growth of imaginal discs and the properties of the DPP gradient. but it will be interesting to identify new (or old) factors that implement this control through cell growth or cell cycle progression. Sci. 4003–4013 (2008). T. Wartlick. 357–368 (1996). cell cycle phase progression. & Basler. Schwank. Gonzalez-Gaitan. 124. S. 6. Hedgehog creates a gradient of DPP activity in Drosophila wing imaginal discs. USA 104. Martin-Castellanos. D. Ċ/C and C′/C distributions and mechanical properties during individual cell cycles concomitantly with cell volume increase. Marty. G. References 6 and 7 show the distribution of the DPP ligand as a gradient by using DPP tagged to green fluorescent protein. 889–897 (1995). D. Teleman. Genet. Positional information and patterning revisited. 9. T. Biol. 6924–6929 (2010). Orchestrating size and shape during morphogenesis. Fainsod.. References 20–22 are the first theoretical proposals of the role of mechanics on growth control in the imaginal discs. for Dpp signaling in vivo. K. 46. et al. although we still cannot favour one model over another. 31. T. N. & Basler. In other words. Cold Spring Harb. Punt and Thick veins. & Barkai. The authors also propose that DPP controls growth permissively in the centre of the discs. L. Dev. Shilo. et al. 553–562 (1999). S. ten Dijke. 13. 2. A. Questions whether the graded distribution of DPP is relevant to growth control. Mech. Cell 96. M. H. Capdevila. A. 3835–3840 (2007). Model for the regulation of size in the wing imaginal disc of Drosophila. References 8 and 9 reveal that DPP acts as a morphogen that directly activates target genes at different distances from the source of production in a DPP concentration-dependent manner. E. Rouault.. whereas growth in the periphery is independent of the DPP gradient. Entchev. 1. Proc. 23. et al. Direct and long-range action of a DPP morphogen gradient. ‘Big frog. Growth regulation by Dpp: an essential role for Brinker and a non-essential role for graded signaling levels. Basler. such signalling systems can sense an increase with respect to previous levels of the signal. T. Cell 103. G. 304–308 (1997). Such relative changes of signal can be detected by one of two network motifs: the incoherent feed-forward loop (IFFL)77. S. N. & Basler.. Measuring dimensions: the regulation of size and shape. Affolter. 1205–1211 (2008). PENT displays properties akin to the expander by expansion–repression feedback.F O C U S O N m O R p h OR EE N E W S g VI SI concentrations are necessary to maintain the response72. & Garcia-Bellido. 5. Biol. T. Bollenbach. M.. I. & Rushlow. L. Struhl.. Yorkie and MYC could form a NFBLB66. Kicheva. with a single sensor module76. Scaling of the BMP activation gradient in Xenopus embryos. Ben-Zvi. Tanimoto. A conserved activation element in BMP signaling during Drosophila development. A. 8. Nature Cell Biol.. Measuring and relating these properties to each other and to the global gradient and growth properties of the disc will hopefully allow us to complete the picture of morphogenetic growth control. & Merriam. C. M. Ben-Zvi. Dpp gradient formation in the Drosophila wing imaginal disc. Kirov. Campbell.. 10. 318–326 (2007). Day. Vuilleumier. Proc. growth regulatory protein levels and orientation of cell division as a function of cell position and developmental time. but on unravelling the dynamics of the interactions of the known components by quantitative experimental analysis. Burke. Mol. M. R.. It will be interesting to measure DPP concentration and signal transduction. 12. Kamoshida. A. Long-term in vivo cultures and movies will open the way for the description and analysis of parameter distributions in individual cells. 19. Nature 381. 69–76 (2009). 4. The Decapentaplegic morphogen gradient: from pattern formation to growth regulation. Johnson.. 2977–2987 (2000). & Basler. 59–71 (2000). G. 24.. it seems that DPP. Theor. Wieschaus. & Tabata. Kinetics of morphogen gradient formation. Roth. Kim. Cell 85. Itoh. J. Schwank. Weiss. 61–87 (1971). a001669 (2010). However. Nature 453. Drosophila Mad binds to DNA and directly mediates activation of vestigial by Decapentaplegic. Adaptive signalling pathways have been extensively studied in other model systems in which relative changes of a signal are measured. Natl Acad. Barkai. Sci. H. such as chemotaxis in bacteria or amoebae and olfactory or gustatory transduction in Caenorhabditis elegans 73–76. 11. & Ben-Zvi. Science 315. B. 29. 12. et al... A.. 27. S. It is very likely that cells can measure and use a range of inputs — absolute signalling levels as well as spatial. 30. Sci. or even relative temporal changes of C′/C (see Supplementary information S1 (table)). C. A.. Many components of the machinery Wolpert. A.. The first report of the effects of PENT on the DPP gradient. Z. Scaling of morphogen gradients by an expansion-repression integral feedback control. 20. 8. Nellen. 663–674 (2007). N. P. 611–617 (2010). 276. 745–749 (2000). M. the emphasis might now be not on identifying new components in the machinery that controls morphogenetic growth. Cohen. Marty. Cell proliferation patterns in the wing imaginal disc of Drosophila. USA 107. It will be important to study if any of these motifs are present in the wiring of the DPP pathway and relevant to DPPdependent growth control. K. Shraiman. Greece. Nature 398. T. Proposes for the first time an expander mechanism based on expansion–repression feedback. Jazwinska. Teleman. M. 971–980 (2000). 7.78 and the negative feedback loop with a buffering node (NFBLB)78. O. & Shraiman. 21. Y. Transducing the Dpp morphogen gradient in the wing of Drosophila: regulation of Dpp targets by brinker. & Le Goff. Tanimoto. melanogaster have allowed us to formulate and discuss current morphogenetic growth models in a slightly more precise manner. 3. All rights reserved VOLUME 12 | SEPTEMBER 2011 | 603 . Control of Dpp morphogen signalling by a secreted feedback regulator. & Edgar. small frog’ — maintaining proportions in embryonic development: delivered on 2 July 2008 at the 33rd FEBS Congress in Athens. & Tomlinson. On the mechanism of wing size determination in fly development. J. & Gonzalez-Gaitan. T. Therefore. cells might adapt to relative temporal changes in DPP signalling levels through Fat signalling. Lecuit. P. L. Chen.. 3318–3323 (2005). Nature Struct. 183–200 (1994). A. 1137–1146 (2008). 18. Nature Cell Biol. 24. K. Nellen. Development 135. Cell 96. Dev.. J. B. 981–991 (2000). Hufnagel. Measurements of gradient properties so far have been either qualitative in nature or focused on average properties of cell populations. 17. 16. N. D. D. 563–573 (1999). B. R. Aegerter-Wilmsen. G. et al. Minami. Precision of the Dpp gradient. 22. The Drosophila gene brinker reveals a novel mechanism of Dpp target gene regulation. N. K. E. Science 331. Conclusions and perspectives Recent quantifications of growth and morphogen gradient properties in the wing imaginal disc of D. Schnurri mediates Dpp-dependent repression of brinker transcription. I. K. & Laughon. Restrepo. M.. Biol. Another possibility. Kinoshita. Nature 388. D. 14.... M. An absolute requirement for both the type II and type I receptors. & Basler. FEBS J. Proc. 359–365 (2011). showing that the gradient scales owing to the regulation of the degradation rate of DPP and that cells divide when they increase their signalling levels by 50%. E. 387–393 (1996). J. 26. S. 242–246 (1999).. The first theoretical analysis of a possible mechanism of gradient scaling. 1003–1013 (2002).. given the wealth of qualitative information we already have in this model system.

350. K. E Soft Matter 33. 81–98 (1974). Coordination of growth and cell division in the Drosophila wing. e1000420 (2010). Cell Sci. 667–679 (2005). B. Cell 93. D. 764–766 (1970). 5691–5696 (1997). 73. 47. M. Lim. Shoval. & Schubiger. Dev. Nature 416. Development 131. Regulation of Frizzled by Fat-like cadherins during planar polarity signaling in the Drosophila compound eye. C. Cell 108. Extent and properties of the regeneration field in the larval legs of cockroaches (Leucophaea maderae) III. Defining network topologies that can achieve biochemical adaptation.. Thompson. Axelrod. 13256–13261 (2007). Delineation of a Fat tumor suppressor pathway. L. J. El-Samad. et al. A. Exp. K. & Edgar. SUPPLEMENTARY INFORMATION See online article: S1 (table) ALL LINKS ARE ACTIVE IN THE ONLINE PDF 604 | SEPTEMBER 2011 | VOLUME 12 © 2011 Macmillan Publishers Limited. J. F. Dev. the Swiss National Science Foundation. 788–794 (2010). & Julicher. Dev. Cho.. 894–899 (2009).. 55.. 51. grants from the Swiss SystemsX. 50–63 (2002). Biol.. 38. Nature 226. Barkai.REVIEWS 32. 122. T. Bollenbach. & Axelrod. & Iglesias. Biol. Ma. G. 50. Dev. 621–636 (1994). 629–640 (1999). 43. P. D. 36. Dally regulates Dpp morphogen gradient formation in the Drosophila wing. a Drosophila glypican. C. L. Basan. Akiyama. C. McNeill. 755–759 (2002). Curr. Serrano. S. Cell 126. based on an assay consisting of the activation of DPP in clones of cells. F. Campuzano. S. E. A. Proc. Sastre-Garau. Antagonistic growth regulation by Dpp and Fat drives uniform cell proliferation. Mao. Richardson. & McNeill. Natl Acad. Bargmann. Expression of a secreted form of Dally. Biol. Torres-Schumann. 760–773 (2009).. & Prost. J.G. M. 54. 53. Neufeld. Chemotaxis of sperm cells. Y. P. N. T. X. & Morata.html 71. 76. 85. 309–321 (2008). 48. 86. C. Alon. M. The role of spatially controlled cell proliferation in limb bud morphogenesis. M. Also shows that Fat signalling is affected by DPP. as monitored by the localization of Dachs at the interfaces between cells. 385–391 (2011). H. 74. 1895–1904 (2006). 84. 37. Drosophila genetic variants that change cell size and rate of proliferation affect cell communication and hence patterning. Tsatskis.. 543–547 (2003). Genet. Action of Fat. & Gonzalez-Gaitan. Context-dependent relationships between the BMPs gbb and dpp during development of the Drosophila wing imaginal disk. & Basler. Models of eukaryotic gradient sensing: application to chemotaxis of amoebae and neutrophils. 117–129 (2004). 16. 33. Regulation of cell proliferation by a morphogen gradient. S. M. Natl Acad. O. & Wharton. Fold-change detection and scalar symmetry of sensory input fields. D. U. A. B. K. 2539–2551 (2006). M. K. M. W. H. 67. Thompson. J. 1054–1062 (2011). Cho. J. & Halder. 75. Cell 122. Budding behaviors: growth of the limb as a model of morphogenesis. The Fat and Warts signaling pathways: new insights into their regulation. A. Dis. Eur. Simon. Hornbruch. 295. Development 120. & Tang. Johnston. D. dMyc transforms cells into super-competitors. J. 126. Akiyama. M. H. Cell 123. & Garcia-Bellido. E. K. & Irvine. 240. 34. D. A. et al. T. Cells compete for Decapentaplegic survival factor to prevent apoptosis in Drosophila wing development. HFSP J. 265–272 (2009). 2827–2838 (2008). Mol. The Salvador/Warts/Hippo pathway controls regenerative tissue growth in Drosophila melanogaster. W. E. & Alon.. 82. Nolo. Sun. S. T. USA 94. & Lin. Dev. 139–151 (2011).-G. Degoutin. Four-jointed. 38. F. J. & Pan. et al. 58. T. Sci. The tumor-suppressor gene fat controls tissue growth upstream of expanded in the Hippo signaling pathway. Dev. 40. Drosophila glypicans Dally and Dally-like shape the extracellular Wingless morphogen gradient in the wing disc. & Aegerter. F. McClure. O. G. 39. Schluck. Interpretation of the wingless gradient requires signaling-induced selfinhibition.. L. & Yang.. R. S. The incoherent feedforward loop can provide foldchange detection in gene regulation. Reddy. and in neoplastic tumors. Dev. The wingless signalling pathway and the patterning of the wing margin in Drosophila. 88. K. Limb morphogenesis: connections between patterning and growth. E. Development 128. 757–767 (1973). gbb.. 57. Genetics 152. 70. N. 35. The Fat cadherin acts through the Hippo tumor-suppressor pathway to regulate tissue size. J. S. Takeo. F. Dev. A. Gardano. G. Nature Genet. Rauskolb. H. WormBook 25 Oct 2006 (doi:10. Bohn.. U. Proc. Development 125. Developmental analysis and squamous morphogenesis of the peripodial epithelium in Drosophila imaginal discs. Morphol.. 72. Neto-Silva. C. F. R. Ray. Fujise. W.. 41. Silva. Biol. & Johnston. et al. 22. Zhang. E. Matthews. C. the Drosophila homolog of YAP. London. K. Acknowledgements The authors thank A. A. Bishop. & Aegerter. 82.. D. C.1). Fat cadherin modulates organ size in Drosophila via the Salvador/Warts/Hippo signaling pathway. Boehm. C.. J. T. & Nakato. & Martinez Arias. L. Sci. A. Genetic analysis of the bone morphogenetic protein-related gene. Fidelity in planar cell polarity signalling. Schwank. Aigaki. Wartlick. A. Kirschner. cooperate to provide pattern information and establish cell identity in the Drosophila wing. de/mpi-doc/julichergruppe marcos gonzalez gaitan’s homepage: http://cms. H. whereas in the centre of the imaginal disc growth control is mediated by permissive DPP. Nature 387. H. D. R186–R195 (1997). D. S. Proposal that growth is controlled by the differences in signalling between adjacent cells. 48.. Bangi. Following along the lines of reference 25. M. where differences in DPP signalling are maximal. Tapon. Child. 64. et al. M.. Arch. & Sands. D. Bayly. 178–193 (2006). 59. Competing interests statement The authors declare no competing financial interests. 44. K. Homeostatic competition drives tumor growth and metastasis nucleation. J. E. Y. Yang. Pantazis... H. Morphogen control of wing growth through the Fat signaling pathway.. 1362–1373 (2009). Belenkaya. Cell 20. et al. A. A. N. Bennett. Kruse. 81. 2723–2734 (1998). A. et al. Dpp signaling promotes the cuboidal-to-columnar shape transition of Drosophila wing disc epithelia by regulating Rho1. & O’Farrell. 2081–2089 (2006). R. B. J. Y. 77. Mech. 675–688 (2002). Developmental control of cell growth and division in Drosophila. Gonzalez-Gaitan. & Vincent. & Axelrod. Milan. C. elegans. Shoval. Biol. The bantam microRNA is a target of the Hippo tumor-suppressor pathway. T. 942–949 (2009). T. C. Khalsa. Eur. Dev. A. Dpp and Gbb exhibit different effective ranges in the establishment of the BMP activity gradient critical for Drosophila wing patterning. Phys. J. A. FURTHER INFORMATION Frank Jülicher’s homepage: http://www. 83. P. 66. Dev. Cooperative regulation of growth by Yorkie and Mad through bantam.1. Development 133. A. E. J. Risler. Goentoro. & Irvine. Pointing in the right direction: new developments in the field of planar cell polarity. J. M. F.1895/ wormbook. 255–266 (2011). in the periphery growth is mediated by DPP-independent Fat signalling. & Irvine. 913–917 (1997). W. Han. M. Sci. 408–419 (2008). J. Cell 15. J. induces overgrowth phenotype by affecting action range of Hedgehog. O.. Biol. Basler. 1142–1150 (2006). Sharpe.. Development 132. Determination of mechanical stress distribution in Drosophila wing discs using photoelasticity. L. Joanny. Cell 117. 1183–1193 (1998). Tao.. Dachs: an unconventional myosin that functions downstream of Fat to regulate growth. Biol. The Hippo pathway regulates the bantam microRNA to control cell proliferation and apoptosis in Drosophila. Couso. 56. All rights reserved www. Nature Rev. C.mpipks-dresden. An Introduction to Systems Biology: Design Principles of Biological Circuits (eds Etheridge.. Dev. B. M. 2101–2110 (2006). & Harvey. et al. et al. Friedrich. Curr. 7. Development 135. Cell division in the early growth and morphogenesis of the chick limb. 767–774 (2006). Development 131. C. Gbb-60A and Dpp.. C.. Dachsous and Dachs in distal-to-proximal wing signaling. 2007). & Harvey. S. & Simon. D. Trusina. Development 132. I. K. Quantitative growth and development of human brain. Cell Biol. B. Dally regulates Dpp morphogen gradient formation by stabilizing Dpp on the cell surface. 5033–5042 (2005). 123–130 (2011). X. 69. Taylor and Francis Group. Curr. Grusche. P. The Hippo signaling pathway coordinately regulates cell proliferation and apoptosis by inactivating Yorkie. 32. this report proposes that. & Julicher. Willecke. Natl Acad. 8. K. Chemosensation in C. de Beco. T. Rogulja. H. 42. 3913–3925 (2001). Phys. Aegerter-Wilmsen. 60. G. Huang.ch initiative and LipidX-2008/011. J. Nature 421. Levchenko. Ma. 4843–4856 (2004).. 449–461 (2005). 45. J. 111–115 (2010).. identifies multiple requirements during Drosophila development. & Leibler. Dev. TGF-β/BMP superfamily members. D. Moreno. R. USA 107. 12. & Irvine. Dobbing. Evidence for a growth-stabilizing regulatory feedback mechanism between Myc and Yorkie. 65.. K. A. 204–218 (2005). U.. 121. 87. PLoS Biol. K. 109–122 (2011). 507–520 (2010).123. O. D. & Dahmann. Genevet for comments on the manuscript. 62. 284. Yoon. P. L. & Irvine. D. J. K. H. J... Bittig. & Wharton. Opin. P. V. Yang. Biol. Biophys. 16. 313. Cell 136. Resino.) (Chapman & Hall/CRC Press. D. K. H. Y. affinity and gene expression in Drosophila. Development 130.. K. 296–307 (2009). This work was supported by the Max-PlanckGesellschaft. 4489–4500 (2004). Robustness in simple biochemical networks.. 79. A. G. 16. Embryol.. S. T.ch/ sciences/biochimie/-Marcos-Gonzalez-Gaitan-Lab-. and by an ERC advanced investigator grant to M. 421–434 (2005). Wharton.unige. T. Firkus.. USA 104. K. & Garcia-Bellido. A. Mech. Hopyan. F. & Irvine. 52. mechanism and conservation. Nienhaus. M. Piddini. Origin of the tissues and determination of symmetry properties in the regenerates. Biol.. Curr. Quantification of growth asymmetries in developing epithelia. Dyn. Photo-elastic properties of the wing imaginal disc of Drosophila.mpg. Cell 36. A. 2090–2100 (2006). de la Cruz. P. Cell 20. 3.. & Wharton. D. & Wolpert. Dpp gradient formation by dynamin-dependent endocytosis: receptor trafficking and the diffusion model. 49. J. Morrison. K. et al.. 80. 16. 1515–1522 (2003). A. Curr. Barrera. & Cohen. M. X. 63. 93–99 (2009). Cell 138. M. Wu.nature. Curr. Developmental parameters of cell death in the wing disc of Drosophila. Widmann. 351–364 (2004). E. 46. Biol.. the Drosophila homolog of Yap. 68.. Rogulja. 15995–16000 (2010). J.. J. 78.. Shows that cells can measure the differences in DPP signalling levels between adjacent cells. B. Yan. 61. Moreno. Julicher. Proc. F. Cell 19.com/reviews/molcellbio . 350. Oh. E Soft Matter 30. Regulation of Hippo signaling by Jun kinase signaling during compensatory cell proliferation and regeneration.