Robustness and Epistasis
in the C. elegans Vulval Signaling Network
Revealed by Pathway Dosage Modulation
Michalis Barkoulas,1,2,3Jeroen S. van Zon,4,5Josselin Milloz,2,6Alexander van Oudenaarden,4and Marie-Anne Fe ´lix1,2,3,*
1Institute of Biology of the Ecole Normale Supe ´rieure
2Centre National de la Recherche Scientifique, UMR 8197
3Institut National de la Sante ´ et de la Recherche Me ´dicale U1024
46 rue d’Ulm, 75230 Paris Cedex 05, France
4Department of Biology, Massachusetts Institute of Technology (MIT), 77 Massachusetts Avenue, Cambridge, MA 02139, USA
5Present address: FOM Institute AMOLF, Science Park 104, 1098 XG Amsterdam, The Netherlands
6Present address: FAS Center for Systems Biology, Harvard University, 52 Oxford Street, Cambridge, MA 02138, USA
Biological systems may perform reproducibly to
generate invariant outcomes, despite external or
internal noise. One example is the C. elegans vulva,
in which the final cell fate pattern is remarkably
robust. Although this system has been extensively
studied and the molecular network underlying cell
fate specification is well understood, very little is
known in quantitative terms. Here, through pathway
dosage modulation and single molecule fluores-
cence in situ hybridization, we show that the system
can tolerate a 4-fold variation in genetic dose of
the upstream signaling molecule LIN-3/epidermal
growth factor (EGF) without phenotypic change
in cell fate pattern. Furthermore, through tissue-
specific dosage perturbations of the EGF and Notch
pathways, we determine the first-appearing pattern-
ing errors. Finally, by combining different doses of
both pathways, we explore how quantitative path-
way interactions influence system behavior. Our
results highlight the feasibility and significance of
launching experimental studies of robustness and
Developmental patterning systems can operate with astonish-
ing precision and reproducibility akin to optimally engineered
machines (Csete and Doyle, 2002). The property of a system to
produce a relatively invariant output in the face of considerable
variation, be it genetic, stochastic, or environmental, is called
robustness (Wagner, 2005). The study of robustness has in-
creasingly attracted the attention of biologists, because of its
importance in understanding system behavior and evolution,
but also due to its plausible implications in human disease (Ki-
have seen an increase in theoretical studies addressing ques-
tions on robustness (Barkai and Leibler, 1997; von Dassow
et al., 2000; Meir et al., 2002; Siegal and Bergman, 2002; Ma
et al., 2006; Hoyos et al., 2011), experimental studies have re-
mained very scarce, especially in multicellular eukaryotes (Eldar
et al., 2002; Moriya et al., 2006; Gregor et al., 2007; Ansel et al.,
2008; Braendle and Fe ´lix, 2008; Levy and Siegal, 2008). This is
partly because of technical challenges in addressing the extent,
limits, and mechanisms of robustness. A key challenge is to
devise methods to precisely perturb a given system and then
quantify the perturbations together with their phenotypic effects
for the system (Masel and Siegal, 2009). To this end, the Caeno-
rhabditis elegans vulva offers an excellent opportunity, as it is
highly robust (Fe ´lix and Barkoulas, 2012) and readily amenable
to genetic experiments, including transgenic manipulations.
More importantly, both controlled dosage manipulations and
quantitative techniques with single cell resolution have lately
become available in C. elegans (Frøkjaer-Jensen et al., 2008;
Raj et al., 2008), paving the way for quantitative, perturbation-
driven studies of robustness.
The C. elegans vulva has become a ‘‘textbook’’ example of in-
tercellular communication driving animal organogenesis (Stern-
berg, 2005). Functionally, it is the egg-laying and copulatory
organ of the adult hermaphrodite that is specified during the
third larval stage (L3) of postembryonic development from a
row of six epidermal precursor cells (Pn.p cells, numbered
from P3.p to P8.p). Although these six Pn.p cells are all compe-
tent to adopt vulval fates, only three, P5.p–P7.p, are induced to
form the vulva, whereas the other three (P3.p, P4.p, and P8.p)
contribute to vulval tissues in situations when P5.p–P7.p fail to
do so. Vulval induction involves two major signaling pathways,
namely epidermal growth factor (EGF)-Ras-mitogen-activated
protein kinase (MAPK) and Notch, which act together to specify
three distinct cell fates (Sternberg and Horvitz, 1989; Sternberg,
2005) (Figure 1A). The master upstream inducer of vulval cell
fates is LIN-3, a molecule similar to EGF (used here interchange-
ably with LIN-3), which is secreted by the anchor cell (Hill and
Sternberg, 1992). As a consequence of LIN-3 secretion, the
Ras-MAPK pathway is activated in P6.p, leading to the acquisi-
to the concomitant production of the Notch ligands (Figure 1A).
64 Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc.
Subsequently, the 2?fate (depicted in red) is acquired by the
neighboring P5.p and P7.p through paracrine Notch lateral
signaling (Simske and Kim, 1995; Chen and Greenwald, 2004)
(Figure 1A). LIN-3 may also act as a short-scale morphogen
and directly promote, at low-dose, the 2?fate in P5.p and P7.p
(Zand et al., 2011). The remaining cells of the competence group
acquire the default 3?fate (depicted in yellow), and their daugh-
ters fuse with the hypodermis after one division (in about 50% of
the individuals, P3.p fuses to the hypodermis without dividing).
Therefore, the overall phenotypic space of this system is that
of six competent cells acquiring one of three possible cell fates
a good understanding of the molecular signaling network under-
lying vulval development. Nonetheless, even in such a well-
studied system, systematic quantitative studies have been
very limited, and very little is understood in quantitative terms
is of paramount importance to vulval induction, as reduction of
lin-3 levels results in hypoinduced vulval phenotypes (where
theaverage numberofinducedcells,alsocalledinduction index,
is less than three), whereas lin-3 overexpression leads to hyper-
induction phenotypes (more than three cells induced). We also
know that the system can tolerate some variation in lin-3 expres-
sion, given that heterozygous animals for a null mutation in lin-3
show a wild-type vulva cell fate pattern—that is, lin-3 null muta-
tions are recessive (Ferguson and Horvitz, 1985). However,
a basic question concerning the EGF pathway remains unex-
plored: what is the lin-3 dose range that allows the formation
of a correct wild-type cell fate pattern?
A second basic question concerns the role of the Notch
pathway in the vulval cell fate patterning network. LIN-12/Notch
is thought to be playing a dual role in the vulval precursor cells,
which is both to promote the 2?fate, but also to inhibit the 1?
fate in P5.p and P7.p (Kenyon, 1995) (Figure 1A). Primary fate
inhibition is implied by the lin-12 loss-of-function mutant pheno-
type with adjacent 1?fates and mediated by direct tar-
gets inhibiting the phosphorylation of MAPK, such as the
phosphatase lip-1 (Berset et al., 2001; Yoo et al., 2004).
Secondary fate promotion is supported by the excess of 2?fates
Although both LIN-12 functions are likely to be relevant for vulva
any, prevails. Thisismostly because LIN-12 isinvolved in anchor
cell specification and can thus indirectly influence EGF signaling
by altering the number of anchor cells in the somatic gonad.
For example, strong lin-12 loss-of-function mutants display two
to four anchor cells (Greenwald et al., 1983; Sternberg and Hor-
vitz, 1989) and thus show increased EGF signaling, whereas
even weak lin-12 mutations increase anchor cell number at a
low penetrance and potentially lin-3 expression (Sundaram and
Greenwald, 1993). The effect of altering the Notch pathway
specifically in the Pn.p cells thus remains unclear.
A third general question concerns the effects of variation in
two components of a network. Extensive epistatic analysis has
been performed in the vulva system, as in several other systems,
by developmental geneticists seeking the relative position of
genes in genetic pathways (Ferguson et al., 1987; Sternberg
and Han, 1998; Sternberg, 2005). This type of epistasis analysis
aims at defining upstream and downstream components in a
given pathway and therefore requires null alleles to fully elimi-
nate the activity of a gene product. On the other side, quantita-
tive and population genetic theory ignores molecular pathways,
but includes allelic variation at different loci. In this evolutionary
framework, epistasis or gene interactions correspond to the
nonadditive effect of combinations of alleles at two loci in a pop-
ulation (Phillips, 2008). Synthesizing laboratory and evolutionary
ular pathways is central in accounting for phenotypic variation
among individuals, be it stochastic, environmental, or genetic.
ways within a network can inform on quantitative states of the
Figure 1. Robustness and First Errors upon
LIN-3 Dose Variation
(A) A simplified cartoon depicting the wiring of the
vulval network as described in the introduction.
The Ras-MAPK pathway is depicted in blue, the
Notch pathway in red, and interactions of the two
pathways in orange. LIN-3 secretion from the
anchor cell is depicted in green. The 3?3?2?1?2?3?
wild-type pattern is presented on the upper right-
(B–E) Phenotypic characterization of the pertur-
bation lines that show the first-appearing pat-
terning errors when increasing (C) or decreasing
(D and E) the LIN-3 dose in comparison to the N2
reference wild-type stain (B). The results are
schematically presented in color-coded tables,
where each column represents a Pn.p cell (3–8)
and each line an animal (n = 100). Half fates
represent cases when the two daughters of a Pn.p
cell adopted different fates. Throughout this work,
P3.p was only scored for induction and not for
division (Pe ´nigault and Fe ´lix, 2011) and is repre-
sented in yellow (as P4.p) for simplicity.
See also Figure S1.
Quantitative Analysis of Vulva Patterning Network
Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc. 65
network (Gutie ´rrez, 2009; Corson and Siggia, 2012). Systematic
combinations ofalleles ofdifferent strengthshaveneverbeen at-
tempted, yet we argue that they are essential for a quantitative
understanding of a system and its variational characteristics.
Here we thus sought out to investigate quantitative aspects
of vulval development by experimentally perturbing the activity
of the two main pathways involved in vulval patterning. First,
we establish a LIN-3/EGF dose-response curve by performing
dosage perturbations and then quantifying the degree of per-
turbation and the resulting phenotypic cell fate pattern for
P3.p–P8.p. These experiments allow us to quantitatively define
robustness by characterizing the range of lin-3 dose variation
compatible with the correct cell fate pattern. Second, by per-
forming tissue-specific perturbations for the Notch pathway,
we identify the first-appearing patterning errors to changes of
its dose and contrast our tissue-specific LIN-12 downregulation
to the existing results using lin-12 mutants. Third, we combine
dosage perturbations of the two pathways and investigate how
the patterning systembehaves. Weuncoverbothsynthetic inter-
actions and unexpected, nonmonotonous system behavior. Our
approach highlights emerging opportunities for experimental,
quantitative studies of developmental robustness and network
behavior in model eukaryotes.
Robustness and First-Appearing Errors to EGF Dose
We started our experimental perturbations with the major
pathway involved in vulval induction, the LIN-3/EGF pathway.
To vary LIN-3 dose, weused a combination of transgenic manip-
ulations, a cis-regulatory mutation, and RNA interference (RNAi).
First, to increase the lin-3 dose compared to the reference strain
N2, we created lines with additional copies of lin-3 at defined
chromosomal positions, using the Mos1-mediated single copy
insertion (MosSCI) methodology (Frøkjaer-Jensen et al., 2008).
We observed that the addition of three extra copies of lin-3 per
haploid genome, as in strain JU2038 (Figures 1B and 1C), was
necessary to disturb the normal vulval cell fate pattern and led
to about 10% penetrant errors, whereas no defects were found
in animals carrying either one or two additional copies of lin-3
(Figures S1A–S1C available online). We determined that the
nature of the errors in JU2038 was mostly induction of P4.p,
while P5.p acquired a half or full 1?fate (Figures 1C and S1A).
To support this finding and obtain more perturbation lines, we
produced integrated multicopy transgenes, expressing different
levels of lin-3, by introducing a genomic lin-3 fragment under its
anchor cell specific enhancer (Hwang and Sternberg, 2004). To
obtain lines that represent the first errors to increasing LIN-3
dose, we created arrays with few lin-3 copies by diluting the
lin-3 transgene with carrier DNA in the injection mix. We were
able to recover both phenotypically silent lines for the vulva
phenocopies JU2038 (Figure S1A).
To decrease LIN-3 dose, we used lin-3 RNAi treatment by
feeding the worms with E. coli bacteria, producing double-
stranded RNAs corresponding to a lin-3 fragment. Such a
treatment in the N2 reference strain phenocopies strong loss-
of-function mutations in lin-3 (Figure S1A). To define the first
errors upon downregulation of lin-3, we diluted the double-
stranded RNA-producing bacteria with control bacteria to reach
concentrations that gave rise to low-penetrance errors, whereas
a treatment with even higher dilutions did not alter the cell fate
pattern (Figures 1D and S1A). We observed that the first errors
upon lin-3 downregulation were either loss of 2?fate for P5.p
and P7.p or loss of all induced fates (Figure 1D). We were unable
RNAi treatment, indicating that the transition from wild-type
pattern to loss of all induced fates is very steep.
lower than N2 by overexpressing a lin-3 genomic fragment con-
taining the lin-3(e1417) promoter mutation that substantially
diminishes lin-3 expression (Hwang and Sternberg, 2004) into
the lin-3(e1417) mutant background. We recovered rescued
aphenotypic lines with induction index of 3 (n > 100), but also
one line, JU1070 (Figure 1E), which showed concurrent loss of
2?fates and complete loss of induction, both at low penetrance,
thus supporting the RNAi results.
Taken together, using the above approaches, we obtained
a series of lines that displace the vulval induction index toward
the hypoinduced or hyperinduced side. Notably, the relationship
between variance and mean induction index is nonmonotonous,
with low variance at vulval induction indices of 0, 3, and 6 and
two peaks of high variance in between (Figure S1D). This pattern
of variance denotes the intrinsic robustness of the wild-type
system around index 3, where variance is zero. Perturbations
that shift the mean induction index also increase its variance
(Figure S1D),as theyresult in partial penetrant changes of induc-
tion for P5.p–P7.p at low egf doses and for P3.p, P4.p, and P8.p
at high doses. At the two extremes, i.e., in the absence of any
inductive signal or when the signal reaching the Pn.p cells is
saturated, none, or all six, competent cells acquire vulval fates
respectively, so variance is again zero.
Quantification of lin-3 Expression in the Anchor Cell
To determine the range of lin-3 dose variation to which the
system is robust, we sought to quantify the level of lin-3 expres-
sion in the perturbation lines, using a quantitative technique with
singlecellresolution, singlemolecule fluorescenceinsituhybrid-
ization (smFISH) (Raj et al., 2008, 2010). This technique has been
successfully used in C. elegans, for transcripts including lin-3
(Saffer et al., 2011), to localize single messenger RNA (mRNA)
molecules as quantifiable, diffraction-limited fluorescent spots.
We quantified lin-3 mRNA number during the L3 stage for all
EGF perturbation lines in synchronized populations (Figures
2A, 2B, and S2A–S2D). EGF perturbations are expected to alter
lin-3 expression specifically in the anchor cell; thus, we used
lag-2 as a marker to delimit the position and boundaries of this
cell (Figure 2A). lag-2 is also expressed in the distal tip cells of
the gonad (Figure S2D) and in P6.p when induction has occurred
(Figure 2A). Distal tip cell expression of lag-2 was useful to
estimate gonad length as a proxy for developmental stage and
thus control for variation in growth of individual worms within
In the N2 reference strain, the number of mRNAs first slightly
increases during development, but soon becomes fairly steady
(Figure S2B for gonad length 44–163 mm, mean = 26.6 ± 1.1
Quantitative Analysis of Vulva Patterning Network
66 Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc.
standard error, range z17–34 molecules, n = 20). Therefore, we
used this time window for our quantifications, which corre-
sponds to the time of induction of Pn.p cells until the first Pn.p
cell division. The average number of mRNAs was the lowest in
the lin-3(e1417) mutant (mean = 1.7 ± 0.4 mRNA molecules)
and reached the highest measurable values in lines that overex-
press multiple copies of lin-3, such as JU1107 (mean = 73.2 ±
1.82 mRNAs) (Figure 2B). We ordered the lines from lowest to
highest mean mRNA number, which largely correlated with the
phenotypic induction index ranking (Figures 2B and S2C).
Quantification of lin-3 mRNAs in lines with the first errors upon
of variation in lin-3 expression that the system can buffer. For
the hyperinduction robustness boundary, line JU2038 (induction
index 3.07) was found to present an average mRNA number of
55.7 ± 2.5 and line JU1105 (induction index 3.025) an average
of 49.5 ± 1.8 molecules. We observed that addition of increas-
ingly more lin-3 MosSCI copies increased linearly the number
of mRNA molecules, although each copy added fewer mRNAs
than the wild-type lin-3, perhaps due to the fact that our lin-3
fragment does not reconstitute the full lin-3 genomic locus (Fig-
ure S2E). Moreover, in keeping with the adjacent 1?fate pheno-
type in lines such as JU2038, we observed some P5.p cells ex-
pressing lag-2, while lag-2 expression is strictly confined to P6.p
in the N2 context (Figure S2F). For the hypoinduction boundary,
lin-3(e1417) heterozygous animals that are completely apheno-
typic for the vulva show 16.8 ± 1.26 lin-3 mRNAs (range 11–25)
on average, whereas dilutions of lin-3 RNAi that show the first
patterning errors were found to display an average of 14.4 ± 1
lin-3 mRNA molecules (range 5–27 molecules). Taken together,
we conclude that the vulva system is robust to genetic variation
in lin-3 expression between an average expression level of 15
and 50 lin-3 mRNAs. Any variation in lin-3 dose within these
boundary values is predicted to be phenotypically silent for
vulval patterning (Figure 2C). However, it is important to clarify
that what we measured here is a correlation between mean
lin-3 mRNAs in the anchor cell and patterning errors in the
Pn.p cells and not a correlation between lin-3 expression and
cell fate in individual animals. Since the first deviant patterns
occur at low penetrance, it is possible that these errors are asso-
ciated with some extreme values of the respective lin-3 mRNA
distributions (Figure S2C, compare minimum values of distribu-
tion between mild lin-3 RNAi or JU1070, which both show vulval
errors, to lin-3(e1417) heterozygotes, which do not).
First Errors to Tissue-Specific Notch Dosage
The second key signaling pathway involved in vulval cell fate
patterning is the LIN-12/Notch pathway, which is activated
downstream of the EGF pathway through the transcriptional
Figure 2. Quantificationoflin-3 Expressionin the Perturbation Lines
N2, and JU2038 (mfSi1; mfSi2) strains. Left panels show lag-2 signal, middle
panels lin-3, and right panels an overlay of both signals (lag-2 is in red and lin-3
(B) Quantification of the mean number of lin-3 mRNAs detected by smFISH
in all perturbation lines. Bars are color-coded to indicate the nature of the
perturbation: orange bars indicate RNAi experiments, red bars are used for
lines obtained by injecting the lin-3(e1417) sequence into the lin-3(e1417)
mutant, blue bars for lines obtained by injections of wild-type lin-3 into N2,
purple bars for lin-3 MosSCI lines, where the number of insertions per haploid
genome is presented within the bar, the gray bar for N2, and dashed bars for
lin-3(e1417) homozygous or heterozygous animals. Error bars represent
standard error of the mean and numbers above each bar the number of
(C) Graph showing the vulval induction index for each perturbation line as
a function of the number of detected lin-3 mRNAs. The robustness range to
lin-3 genetic dose variation is the area between the first errors on both sides of
the wild-type N2 dose. X error bars represent standard error of the mean
mRNA number and Y error bars standard error of the mean induction index.
See also Figure S2.
Quantitative Analysis of Vulva Patterning Network
Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc. 67
upregulation of Deltas (Chen and Greenwald, 2004). Similar to
the EGF pathway perturbations, we wanted to identify the first
deviant patterns upon Notch pathway activity variation, specifi-
cally within the Pn.p cells, without affecting anchor cell determi-
nation. To vary Notch pathway dose, we decided to tune the
expression of the Notch receptor LIN-12, as we found that
overexpression of Deltas from their endogenous regulatory
sequences does not perturb wild-type vulval cell fate patterning.
To decrease Notch dose in a tissue-specific way, we devel-
is based on tissue-specific rescue of the RNAi-deficient rde-1
mutant of C. elegans (Qadota et al., 2007). We used the lin-31
promoter, which is specifically expressed in the vulval precursor
cells (Tan et al., 1998), to drive expression of the wild-type
rde-1(+) sequence. RNAi then should function exclusively in
the Pn.p cells. Indeed, by introducing let-858::GFP and cdh-
3::GFP reporters and performing GFP RNAi, we verified that, in
the tissue-specific RNAi background, gene expression is down-
regulated only in the Pn.p cells, but not in other tissues, such as
the anchor cell or the seam cells (Figures S3A–S3D; Table S1).
The available bacterial clone for lin-12 RNAi by feeding in
C. elegans worked poorly in our hands, so we produced
a more efficient clone that can be used to phenocopy the
extra-anchor-cell phenotype of lin-12 mutants when used
non-tissue-specifically (Figures S3E and S3F). When we then
knocked down lin-12 expression, specifically in the vulval pre-
cursor cells, we observed mainly a loss of 2?fates for P5.p and
P7.p (Figures 3A–3C). By diluting the RNAi bacteria with control
bacteria, we showed that this phenotype represents the first-ap-
pearing errorupondecrease inNotchdose (Figure 3D).Thus, the
phenotype we observed for tissue-specific lin-12 downregula-
tion suggested that the promotion of 2?fates is the most impor-
tant role of the Notch pathway for Pn.p cell patterning, rather
than the inhibition of 1?fate, at least in the N2 background and
in standard culture conditions (Figure 3E).
To further test this idea, we compared the results of tissue-
specific lin-12 RNAi with RNAi performed in a non-tissue-
specific way in N2 (Figures 4A–4E). Similar to the tissue-specific
phenotypes, and in contrast to the lin-12 loss-of-function
mutants, lin-12 downregulation mostly led to loss of 2?fates
in N2 (Figure 4A). We wondered whether this result reflects
poor downregulation of lin-12, so we used an rrf-3 mutation to
increase the sensitivity to RNAi (Simmer et al., 2002). Notably,
when we performed lin-12 RNAi at the whole animal level in the
rrf-3(pk1426) background, we observed that the most frequent
phenotype was adjacent 1?fates for P5.p and P6.p (Figure 4B).
type was not observed when the same rrf-3 mutation was intro-
that some animals with adjacent 1?fates can always be ob-
served upon tissue-specific lin-12 downregulation, the inhibitory
role of LIN-12 on 1?fate cannot be merely associated with a
change in anchor cell number. We conclude that the major role
extent, to inhibit the 1?fate (Figure 3E).
To increase Notch pathway activity in the Pn.p cells, we ex-
under the Pn.p cell-specific promoter lin-31 (Struhl et al., 1993;
Tan et al., 1998). To obtain lines with a wide spectrum of pene-
trance of vulval defects, we used different dilutions of the injec-
tion mix (Figures 3F–3H). Consistent with a role of Notch in 2?
fate induction, we observed that the first errors to increasing
Notch signaling uniformly in the Pn.p cells were induction of 2?
fate for P3.p, P4.p, or P8.p (line JU2064, Figures 3F and 3G for
mild lin-12 RNAi
in Pn.p cells
Figure 3. Tissue-Specific Notch Perturba-
tions Define the First Patterning Errors
(A and B) Nomarski pictures of the L4 stage vulva
in wild-type (A) and lin-12 RNAi-treated animals
carrying a lin-31::rde-1 transgene in a rde-1
mutant background (JU2039) (B). Arrowhead in(B)
marks a P7.p cell that has not been induced,
whereas P5.p is half induced.
(C and D) Phenotypic scoring of P3.p–P8.p
fates in a representative Pn.p-specific lin-12
RNAi expression bacteria (n = 79) (C) or 3-fold
dilution with control bacteria (n = 83) (D). We
never observed similar defective vulval lineages
in animals grown on control, vector-only, bacteria
(n > 500).
(E) Cartoon depicting the prevailing function of
Notch within Pn.p cells to promote the 2?fate,
rather than inhibit the 1?fate.
(F) Nomarski picture of the L4 stage vulva in
JU2064 (mfIs76[lin-31::lin-12(intra)]); arrowhead
marks an ectopic 2?fate induction.
(G and H) Phenotypic scoring in two lines with
different levels of activated lin-12/Notch ex-
pression: JU2064 (mfIs76[lin-31::lin-12(intra)]), n =
162 (G) and JU2060 mfIs72[lin-31::lin-12(intra)],
n = 26 (H).
nodilution of the
(I) Cartoon summarizing the cell fate patterns observed when increasing or decreasing Notch signaling in the Pn.p cells and their comparison to lin-12 weak
loss-of-function and gain-of-function mutants. WT stands for wild-type N2 pattern.
See also Figure S3 and Table S1.
Quantitative Analysis of Vulva Patterning Network
68 Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc.
7% penetrant vulval defects). Higher Notch overexpression
resulted in all Pn.p cells but P6.p adopting the 2?fate (line
JU2060, Figure 3H for 100% penetrant vulval defects).
Therefore, our tissue-specific Notch perturbations highlight
loss or gain of 2?fates as the first responses to variation in
pathway dose, in contrast to the results previously obtained
using lin-12 mutant alleles (Figures 3E and 3I) (Sternberg and
Horvitz, 1989; Sundaram and Greenwald, 1993). We conclude
that the vulval patterning phenotypes of lin-12 mutants are
dictated by the changes in anchor cell specification (Figure 3I).
Quantitative Interactions between the EGF and Notch
We then sought out to investigate nonadditive interactions
between perturbations in EGF and Notch pathways by com-
We hypothesized that such a quantitative network analysis can
potentially reveal interactions that may have been missed out
from classical vulva genetics. Indeed, by assaying multiple com-
binations (Table S2), we were able to uncover two interesting
First, we found that the cell fate pattern phenotype of tissue-
specific lin-12 downregulation is modified qualitatively by the
addition of a single lin-3 MosSCI copy (mfSi3). This extra lin-3
copy increases the average lin-3 mRNA number in the anchor
cell by 30% and does not alter the vulval fate pattern (Figure 2).
combination of the extra lin-3 copy with tissue-specific lin-12
RNAi instead results in a synthetic phenotype, the frequent
transformation of 2?to 1?fate for P5.p and P7.p (Figure 5A).
This synthetic interaction could also be observed at very low,
almost aphenotypic, doses of lin-12 RNAi (Figure S4A). There-
fore, although the major role of LIN-12 in P5.p and P7.p is to
promote the 2?fate under wild-type conditions, the inhibition
is even mildly increased (Figure 5B).
To consolidate this finding using stable mutants instead of
RNAi, we used a mutation in the Notch coactivator osm-11,
which has been proposed to facilitate Notch signaling in the
Pn.p cells, but not in the anchor cell (Komatsu et al., 2008).
Consistent with this role, the osm-11(rt142) mutation, which
deletes the whole mature protein, phenocopies the tissue-
specific lin-12 knockdown by displaying loss of 2?fates for
P5.p and/or P7.p (Figure S4B; Table S2). In the osm-11(rt142)
mutant background, we observed that two extra copies of
lin-3, which are normally aphenotypic, can similarly result in
the adjacent 1?fate phenotype (Figure S4B). However, it is
mostly P7.p in this case that acquires the 1?fate and not P5.p,
consistent with the bias of osm-11 mutants to lose the 2?
fate more often in P7.p than P5.p (Figure S4B; Table S2). The
genetic interaction between osm-11(rt142) and the two extra
copies of lin-3 was also apparent at the level of mean induction
index (Figure S4C) and average number of vulval invaginations
Second, through our network analysis, we obtained evidence
for a synergistic interaction between EGF and Notch in P6.p
induction. EGF and Notch are supposed to act antagonistically
within a given vulval precursor cell, where both pathways mutu-
ally inhibit each other (Berset et al., 2001; Shaye and Greenwald,
2002; Yoo et al., 2004). However, these two pathways may also
act synergistically on 2?fate specification, as EGF at low dose
rrf-3(-); lin-31::rde-1(+); rde-1(-)
loss of2° fates
% of defective animals
Figure 4. The Role of Notch in the Pn.p Cells Is Mostly to Promote
(A–D) Phenotypic scoring of Pn.p fates of defective animals upon lin-12 RNAi
of N2, n = 42 (A), rrf-3(pk1426), n = 45 (B), lin-31::rde-1(+); rde-1(ne219), n = 46
(C), and rrf-3(pk1426); lin-31::rde-1(+); rde-1(ne219), n = 43 (D) animals. The
wild-type animals are not presented for simplicity.
(E) Quantification of the RNAi phenotypes divided in three classes: adjacent 1?
fates (blue), loss of 2?fates (yellow), and both phenotypes in the same indi-
vidual (green). Differences in the frequency of these classes between N2 and
rrf-3(pk1426) or rrf-3(pk1426) and rrf-3(pk1426); lin-31::rde-1(+); rde-1(ne219)
animals are significant with a chi-square test (p < 0.0001).
Quantitative Analysis of Vulva Patterning Network
Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc. 69
can promote 2?fate induction (Katz etal.,1995), althoughit isyet
unclear whether this is mediated by Notch signaling. Surpris-
ingly, we found that EGF and Notch can also act synergistically
on 1?fate specification, since the P6.p induction defects (1?to
3?fate transformation) conferred by the lin-3(e1417) mutation
were enhanced by the weak lin-12(e2621) mutation (Figure 6A).
Because this interaction could be explained by an effect of lin-
12(e2621) on EGF production from the anchor cell, we also
used our mfIs76[lin-31::lin-12(intra)] transgene that increases
Notch signaling specifically in the Pn.p cells. We observed that
the P6.p induction defects of lin-3(e1417) (Figure 6A) or lin-3
RNAi-treated animals (Figure S4E) were partially suppressed,
indicating that the Notch-mediated effect on 1?fate is likely
confined to Pn.p cells.
A way to interpret this interaction is that the Notch pathway at
low dose may be sufficient to promote the 1?fate independently
in a strain carrying the mfIs76[lin-31::lin-12(intra)] transgene and
the 1?fate marker egl-17::CFP. Contrary to the hypothesis, we
never observed gonad-independent induction (n = 20). egl-17::
CFP expression at the P6.px stage (after one division of Pn.p
between the two pathways in these conditions (Figure S4F).
We then hypothesized that the synergistic effect of the Notch
pathway on 1?fate induction and egl-17::CFP expression only
occurs when LIN-3 activity is impaired, as in lin-3(e1417)
mutants. In the lin-3(e1417) background, egl-17 expression
was hardly detectable in Pn.p cells because of low MAPK
signaling, but became detectable at the Pn.pxx stage (after the
second division round). At this stage, the lin-31::lin-12(intra)
transgene did not affect the egl-17::CFP level in P6.p (Fig-
ure S4G), yet resulted in ectopic egl-17::CFP expression in other
Pn.p cells (9/22 animals), usually P5.p and P7.p, albeit at low
levels (Figures 6B–6D and S4H). This ectopic expression was
never observed in lin-3(e1417) animals without the transgene
(n = 15) and rarely with the transgene alone (0/24 animals at
Pn.px stage and 1/15 animals at Pn.pxx). Therefore, low hyper-
activation of the Notch pathway in a lin-3 reduction-of-function
mutant background can trigger the ectopic expression of a 1?
fate marker. We conclude that EGF and Notch can also act in
a synergistic way within a cell, perhaps depending on their rela-
tive dose, when the EGF pathway is weak.
A Phase Diagram for Vulval Fates in Signaling Pathway
To collectively represent the results of our quantitative network
analysis, we draw the analogy to a phase diagram. These dia-
grams present all discrete phases of a system and their occur-
rence upon quantitative change of parameters, usually physical
parameters, such as temperature or pressure. By analogy,
such a phase diagram represents here the quantitative state of
thevulval cellfateoutputwhenthetwo major signaling pathways
are varied genetically (Figure 7).
In Figure 7, the most frequent error patterns are depicted as
a function of EGF and Notch pathway activities. In most cases,
several cell fate pattern states can coexist for a given combina-
tion of perturbations, due to stochastic variation in the isogenic
population. The two axes of the graph are not independent.
For example, given that production of the Notch ligands is
directly downstream of the EGF pathway, perturbations of lin-3
expression can have an effect on the lateral signal. However,
the reverse is not true: as expected, we observed by smFISH
that our tissue-specific Notch perturbations do not change
lin-3 expression in the anchor cell (data not shown).
The robustness of the wild-type pattern to stochastic variation
among individuals is gradually lost as the system accumulates
quantitative changes in EGF and Notch pathways. As mentioned
above, even a combination of aphenotypic perturbations in
both pathways can fall out of the robustness zone (mild EGF
increase in concert with mild Notch decrease). Similar dia-
grams of cell fate of single Pn.p cells upon quantitative perturba-
tions indicated that P6.p is more robust than other Pn.p cells
for the combinations of perturbations that we assayed (Figure
Moreover, the induction index does not necessarily vary
monotonously along horizontal or vertical lines, that is, when
either EGF or Notch activity is varied (Figure 7). For example, in
the presence of one extra copy of lin-3/egf in the genome, the
vulval induction index first decreases and then increases when
Notch activity is gradually increased, with a corresponding tran-
sition in the vulval fate pattern (Figure S5B).
Previous two-dimensional representations ofvulval cellfate as
resented the theoretical relationship between pathway activities
Pn.p-specific lin-12 RNAi
copy of lin-3
Figure 5. Lateral Inhibition Is Relevant When lin-3 Dose Is Mildly
(A) Comparison of phenotypic effects on P3.p–P8.p fates of Pn.p-specific lin-
12 RNAi performed in parallel in strains without and with a single extra copy of
lin-3 (mfSi3). Note the increased frequency of animals with adjacent 1?fates in
the combined treatment (n = 43), a phenotype that is very rare in lin-31::rde-
1(+); rde-1(?) animals (n = 41) and never observed in animals carrying only the
mfSi3 insertion (n = 100).
(B) Cartoon of our model depicting that the inhibition of 1?fate by LIN-12 is
biologically relevant in situations where lin-3 expression is mildly increased.
See also Figure S4 and Table S2.
Quantitative Analysis of Vulva Patterning Network
70 Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc.
in a given cell (i.e., variables, notparameters in the model) and its
cell fate (Giurumescu et al., 2006, 2009; Hoyos et al., 2011).
Instead, what we represent here is the relationship between
some upstream parameters, such as lin-3/egf synthesis rate
and lin-12/Notch synthesis and activation rate, which we could
experimentally control, and the fates of all six cells.
Here we performed a quantitative analysis of an intercellular
signaling network by modulating each signaling pathway in a
tissue-specific manner. This study significantly improves our
quantitative understanding of the system and its underlying
Robustness of the Vulval Cell Fate Pattern to lin-3 Dose
First, we demonstrated that genetic variation in lin-3/egf expres-
sion can be buffered by the system between an average number
of 15 and 50 lin-3 mRNA molecules at the time of induction. With
an average of 27 lin-3 mRNA molecules, the wild-type reference
strain is located midway, at a 2-fold distance of either boundary.
Given the high reproducibility of the C. elegans vulval cell fate
pattern in different environments or genetic backgrounds (Fe ´lix,
2007; Kiontke et al., 2007; Braendle and Fe ´lix, 2008), this finding
raises the question of whether the lack of errors reflects lack of
lin-3 expression variation outside this buffering zone. It will be
important to test this hypothesis by quantitative analysis of lin-
3 transcription in different growth conditions or in C. elegans
isolates other than N2. One possibility is that lin-3 transcription
is tightly regulated so that variation is low, and another is that
such variation is present, but effectively buffered downstream
in the system.
What is the mechanistic basis for the robustness of wild-type
patterning to lin-3 dose? We suggest that the buffering of varia-
tion in lin-3 expression by the system simply stems from the
spatial cellular context (the row of cells with P6.p being the
closest to the anchor cell) and the network topology (induction
by EGF of the 1?fate in P6.p and of 2?fate by lateral induction
in its neighbors), without the need to invoke robustness-confer-
ring genes or mechanisms (Fe ´lix and Barkoulas, 2012). By finely
varying lin-3 dose, we defined the first-appearing cell fate
pattern errors and thereby reveal the corresponding defective
mechanisms that allow and limit this robustness.
When mean lin-3 level is lowered, two types of error appear
concomitantly: (1) a 2?to 3?fate transformation, generally in
the outer daughters of P5.p and P7.p and (2) a loss of all induced
cell fates. Thus, the lowest activation level of the Ras pathway
allowing for 1?fate activation in P6.p also generally activates
the 2?fate in its neighbors—yet the signal is not always sufficient
in the outer daughter of P5.p/P7.p. In sum, normal patterning
requires a simple threshold of Ras-MAPK pathway activation in
P6.p: wild-type cell fates can then be patterned by MAPK-medi-
ated activation of the 1?fate in P6.p and lateral induction of 2?
When mean lin-3 level is elevated, the limit compatible with
wild-type patterning stems from the fact that EGF can act as
a long-range secreted signal. Intermediate EGF levels experi-
enced by P5.p and P7.p are permissible as they activate the 2?
fate. Higher egf doses may theoretically cause direct ectopic
induction of P4.p/P8.p or transform P5.p/P7.p to the 1?fate
(Giurumescu et al., 2009; Hoyos et al., 2011). We found that,
upon lin-3 overexpression, the first-appearing defect is the
occurrence of adjacent 1?fates, with ectopic induction of a 2?
fate next to the additional 1?fate (Figure 1). Thus, at high egf
levels, lateral inhibition of P5.p/P7.p is the limiting mechanism.
However, the loss of 2?fates upon tissue-specific downregula-
tion of LIN-12 indicates that lateral inhibition is not essential
under normal conditions, perhaps because the amount of EGF
reaching P5.p/P7.p is not sufficient for them to adopt the 1?fate.
We note that a similar first deviant pattern to lin-3 overexpres-
sion had previously been observed in C. briggsae (Fe ´lix, 2007),
but appears contradictory to a recent experiment in C. elegans,
where P4.p was induced in the absence of P5.p fate trans-
formation to the 1?fate (Hoyos et al., 2011). We explain this
discrepancy by the fact that the previous result was based on
overexpression of the lin-3(e1417) sequence, which is defective
mfIs76; lin-3(e1417); egl-17::CFP
P6.p not induced
gistic Way within the Pn.p Cells
(A) Graph showing percentage of P6.p induction in
double mutant combinations of lin-3(e1417) with
either lin-12(e2621) or lin-31::lin-12(intra) (mfIs76).
P6.p induction is 100% for lin-12(e2621) and
mfIs76 animals. Differences are significant by a
Fisher’s exact test (p = 0.0018 and p = 0.0039,
(B and C) Expression of egl-17::CFP is confined to
P6.p descendants at the Pn.pxx stage of mfIs76
(B) and lin-3(e1417) (C) animals. The left panel in
(B) is DIC (note P3.p induction), and the right panel
is the cyan fluorescent protein (CFP) signal.
(D) Ectopic egl-17::CFP expression in other Pn.p
cells is observed in mfIs76; lin-3(e1417) at the
Pn.pxx stage. White arrowheads point to weak
ectopic expression in P4.p, P5.p, and P7.p. The
upper panel is DIC, and the lower panel is CFP
See also Figure S4 and Table S2.
Quantitative Analysis of Vulva Patterning Network
Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc. 71
for anchor cell expression and thus likely produced a less
localized signal. Consistent with this, induction of P4.p before
the occurrence of adjacent 1?fates is also observed in back-
grounds that harbor multivulva mutations, such as lin-15 or let-
60(gf) (Sternberg, 1988, 2005; Sternberg and Horvitz, 1989;
Han et al., 1990; Sternberg and Han, 1998; Fe ´lix, 2012) or
upon heat-shock-mediated overexpression of lin-3 (Katz et al.,
1995). In all cases, the treatment results in either ectopic lin-3
overexpression (Saffer et al., 2011) or general ligand-indepen-
dent gain-of-activity in the Ras-MAKP kinase pathway rather
than an increase in lin-3 expression within the anchor cell.
Role of LIN-12/Notch Activity in Vulval Precursor Cells
Second, we re-evaluated the role of the Notch pathway in vulval
precursor cells using tissue-specific modulations of its activity.
This approach proved informative, since we observed dramati-
cally different cell fate pattern phenotypes upon tissue-specific
alteration compared to those of lin-12 mutants (Figure 3I). We
find that the main role of the Notch pathway in vulval precursor
cells is to induce the 2?fate, but that the system is poised close
to the threshold where the Notch pathway’s role transitions
from induction to inhibition. These findings are in keeping with
a previous report thatbroaddownregulation ofthe Notchligands
leads in the wild-type to some loss of 2?fates, but in a sensitized
background showing ectopic lin-3 expression to an increased
frequency of adjacent 1?fates (Chen and Greenwald, 2004).
These results challenge some textbook depictions of wild-
type vulval development where Notch acts as a lateral inhibitory
the major role of Notch in lateral induction rather than inhibition
does not contradict nor eliminate the fact that EGF is able to
reach other cells than P6.p in a graded manner. Importantly,
both our lin-3 dose-response curve and the switch in mode of
action of the Notch pathway are quantitative results, thus can
be readily used to challenge the existing mathematical models
of vulval cell fate patterning (Giurumescu et al., 2006; Hoyos
et al., 2011).
We conducted a quantitative network analysis by combining
dosage-perturbations in the two pathways and investigating
their cumulative effect. Even for a system like the C. elegans
vulva, where the genetics have been exhaustively carried out,
we did observe unexpected interactions, highlighting the signif-
icance of the approach. For example, we found that a single
extra copy of lin-3 can switch the Pn.p-specific function of
Notch from lateral induction to lateral inhibition. This system
feature had previously been missed, partly because of the
lack of tissue-specific Notch perturbations and partly due to
the absence of mild, aphenotypic EGF dosage-perturbations.
Importantly, this finding may have some evolutionary implica-
tions, as the distinction between lateral inhibition and induc-
tion for Notch relies on silent variation in EGF pathway levels.
Through analysis of EGF/Ras pathway activity reporter expres-
sion in different C. elegans isolates, we have previously revealed
such silent (cryptic) variation (Milloz et al., 2008). Therefore, the
Figure 7. A Vulval Fate Map
Cartoon depicting the most representative vulval fate pattern variants observed upon quantitative perturbation of the EGF and Notch pathways. Within each
rectangle,thefatepatterns arepresented indecreasingorderoffrequencyfromtoptobottom.Notchperturbationsaretissue-specificfor thePn.pcells,and lin-3
dosage perturbations affect lin-3 transcription in the anchor cell. Strong lin-12 overexpression refers to mfIs72, mild lin-12 overexpression to mfIs76, lin-3
reduction of function to lin-3(e1417), one extra lin-3 copy to mfSi3 or mfSi1, two extra copies to mfSi2 or mfSi1; mfSi3, three extra copies to mfSi1; mfSi2, and
stronger (>) lin-3 overexpression to mfIs54or mfIs55.In most cases, thefrequenciesof theseerrors are low, as the majority of the animals retainthe wild-type fate
pattern (with the exception of lines carrying the mfIs72 transgene, where errors are highly penetrant).
See also Figure S5 and Table S2.
Quantitative Analysis of Vulva Patterning Network
72 Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc.
relative mode of action of the Notch pathway in the Pn.p cells
may evolve, concomitant with variation in EGF pathway activity.
Another surprising result concerns the synergistic interaction
between EGF and Notch pathways in 1?fate induction of P6.p
at low LIN-3 levels. This interaction suggests that another layer
of robustness may be provided by Notch activating the 1?fate
in situations where LIN-3 activity is partially impaired. Although
the mechanistic basis of this interaction remains unclear, studies
in Drosophila have identified both positive and negative compo-
Notch activation can lead to opposing effects on EGF signaling,
even within the same developmental context (Krejcı ´ et al., 2009).
Consistent with this genetic interaction, we found ectopic ex-
pression of the 1?fate marker egl-17::CFP in the Pn.p cells of
lin-31::lin-12(intra); lin-3(e1417) animals. In addition, previous
studies have identified a link between the vulval precursor cell
cycle and the temporal window of their competence and com-
mitment to a particular cell fate (Ambros, 1999; Wang and Stern-
berg, 1999), indicating that the effects of these pathways along
the cell cycle may underlie some of the interactions observed
in phasespace. Specifically, Notch signaling has beenproposed
to maintain competence and thus prolong the developmental
window in which Pn.p cells can respond to morphogenetic sig-
nals, such as LIN-3 (Wang and Sternberg, 1999). This phenom-
enon could explain the partial rescue of P6.p induction defects
of lin-3(e1417) animals in the presence of a lin-31::lin-12(intra)
transgene. Itis notable that nonintuitive interactions in sensitized
backgrounds, including this very one, were also anticipated by
a recent geometric model of vulval cell specification (Corson
and Siggia, 2012), where a signal that pushes a cell in one direc-
tion can displace cell fate outcomes in different directions as
a result of nonlinear dynamic flow in the landscape defined by
More broadly, our network analysis highlights important
considerations when interpreting genetic results. In this system,
effects on the induction index or the average number of vulval
invaginations have been repeatedly used to address the func-
tion of newly identified regulators. Intuitively, factors found to
increase the induction index or the number of pseudovulvae of
a given background are believed to typify positive vulval regula-
tors. However, we demonstrated that, depending on the relative
dose of the EGF and Notch pathways, the same pathway can
increase the induction index both when its activity increases
and decreases. For example, in the presence of an extra copy
of lin-3 in the genome, both an increase and decrease of Notch
activity in the Pn.p cells will increase the induction index (Fig-
ure S5B). In this case, when Notch activity is reduced, P5.p
acquires the 1?fate and promotes the induction of P4.p by
secreting Notch ligands; when Notch activity is increased, addi-
tional cells adopt a 2?fate. The distinction thus becomes clear
when the cell fate pattern, and not only the induction index, is
taken into account. Another complication is that we occasionally
observed the same fate patterns upon Notch or EGF dosage
perturbations. For example, loss of 2?fates for P5.p and P7.p
can be observed with either mild decrease of EGF or Notch
and adjacent 1?fates for P5.p and P6.p with either strong
in Notch. This illustrates that the definition of a positive or nega-
tive vulval regulator and its assignment to a particular signaling
pathway can be ambiguous, unless precise scorings of actual
vulval fates are performed and quantitative context-dependent
aspects of interactions in the network are considered.
Within the Caenorhabditis genus, the vulval cell fate pattern is
morphologically invariant, yet it evolves by accumulating cryptic
variation (Fe ´lix, 2007; Milloz et al., 2008). Some of this variation
may concern quantitative evolution in the strength of EGF induc-
tive and Notch lateral signaling pathways. To this end, compar-
isons with the behavior of the C. elegans vulva upon quantitative
pathway change will likely give insights into the evolution of this
lin-3 Overexpression Strains and Genetics
lin-3 overexpression lines were created by amplifying a 5.2 kb genomic frag-
ment, as described in Hoyos et al. (2011). Extrachromosomal arrays were inte-
grated by g-ray irradiation and then backcrossed four times to N2. To create
the lin-3 insertions by MosSCI, the same 5.2 kb fragment was amplified using
primers lin-3AvrII and lin-3XhoI and then cloned into pCFJ151 (chromosome II
targeting vector) and pCFJ178 (chr. IV) (Frøkjaer-Jensen et al., 2008) as an
AvrII/XhoI fragment.Injectionsand recovery of insertions wasperformedusing
fied by PCR using primers flanking the insertion site, and copy number was
tested by pyrosequencing (see Supplemental Experimental Procedures).
lin-3(e1417)/+ heterozygotes were obtained by crossing lin-3(e1417)
mutants carrying a linked myo2::GFP insertion with a strain (BJ49) carrying
an ifb-2::CFP insertion on chromosome IV and selecting animals expressing
lin-12 Dosage Alteration
The lin-31::lin-12(intraDP)::unc-54 construct was built as described in Li and
DNA using primers lin-12intraBglII and lin-12intraNotI and then cloning the
product into pB253 (Tan et al., 1998) as a BglII/NotI fragment. We injected
this construct at two final concentrations in the injection mix: 20 ng/ml to obtain
mild overexpression lines (JU2064) and 75 ng/ml for strong overexpression
lines (JU2060). Extrachromosomal arrays were integrated by g-ray irradiation
and backcrossed five to seven times to N2.
To create the lin-31::rde-1(+)::unc-54 rescuing construct, the full rde-1
coding sequence was amplified from genomic N2 DNA using primers rde-
1F1 and rde-1R1 and then cloned into pGEM-T-easy (Promega). The rde-1
open reading frame was transferred from pGEM-T-easy to pB253 as a NotI
fragment. Insert direction and sequence fidelity were verified by sequencing.
mfIs70 represents a spontaneous integration event of this transgene into chro-
Microscopy and Phenotypic Characterization
To score vulval cell fates, animals were analyzed at the L4 stage by differential
interference contrast (DIC) microscopy, using a 1003 lens. Standard criteria
were used to assign fates (Katz et al., 1995; Fe ´lix, 2007), such as the final
number cells for each lineage, their relative position around the vulva, and
whether they are attached to the cuticle. See also Supplemental Experimental
Single Molecule Fluorescence In Situ Hybridization
Synchronized populations of L3 stage animals were prepared by bleaching
adults to recover their embryos. These embryos were washed with M9 and
plated on standard nematode growth medium plates, allowing them to hatch
and grow at 20?for 32–35 hr before fixation. smFISH was performed as
described before (Raj et al., 2008), using custom-made probes labeled with
Cy5 for lin-3 and Alexa-594 for lag-2. The oligonucleotide sequences used
for each probe can be found in the Supplemental Experimental Procedures.
From each animal, we acquired a Z-stack of 30 sections, using a Nikon Ti-e
inverted microscope or an upright Zeiss AxioImager M1, both equipped with
a Pixis 1024B camera (Princeton Instruments) and a Lumen 200 metal arc
Quantitative Analysis of Vulva Patterning Network
Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc. 73
et al., 2008), using a custom-made routine on Matlab, optimized for anchor
cell measurements. To minimize variation due to developmental stage, only
animals with gonad length >350 pixels (43.75 mm) were included in the quan-
tification. Quantification was not performed in strong lin-3 overexpression
lines, such as syIs1 (PS1123), because the lin-3 signal in the anchor cell could
not be resolved into individual fluorescent spots. Statistical comparisons were
performed using R and Graphpad Prism 5.
Supplemental Information includes five figures, two tables, and Supplemental
Experimental Procedures and can be found with this article online at http://dx.
We thank Christian Braendle for critical reading of the manuscript. This work
was funded by grants from the Association pour la Recherche sur le Cancer
(#1044) and the Agence Nationale de la Recherche (08 BLAN-0024) to
M.-A.F. We also acknowledge the support of the Bettencourt Foundation for
a Coup d’Elan (to M.-A.F.), EMBO for a short-term fellowship (to M.B.), and
Human Frontiers Science Program (to J.S.v.Z.). During the course of this
work, M.-A.F. was first a principal investigator at CNRS and is now a professor
at Ecole Normale Supe ´rieure.
Received: August 8, 2012
Revised: November 12, 2012
Accepted: December 3, 2012
Published: January 14, 2013
Ambros, V. (1999). Cell cycle-dependent sequencing of cell fate decisions
in Caenorhabditis elegans vulva precursor cells. Development 126, 1947–
Ansel, J., Bottin, H., Rodriguez-Beltran, C., Damon, C., Nagarajan, M.,
Fehrmann, S., Franc ¸ois, J., and Yvert, G. (2008). Cell-to-cell stochastic varia-
tion in gene expression is a complex genetic trait. PLoS Genet. 4, e1000049.
Barkai, N., and Leibler, S. (1997). Robustness in simple biochemical networks.
Nature 387, 913–917.
Berset, T., Hoier, E.F., Battu, G., Canevascini, S., and Hajnal, A. (2001). Notch
inhibition of RAS signaling through MAP kinase phosphatase LIP-1 during
C. elegans vulval development. Science 291, 1055–1058.
Braendle, C., and Fe ´lix, M.-A. (2008). Plasticity and errors of a robust develop-
mental system in different environments. Dev. Cell 15, 714–724.
Chen, N., and Greenwald, I. (2004). The lateral signal for LIN-12/Notch in
C. elegans vulval development comprises redundant secreted and transmem-
brane DSL proteins. Dev. Cell 6, 183–192.
Corson, F., and Siggia, E.D. (2012). Geometry, epistasis, and developmental
patterning. Proc. Natl. Acad. Sci. USA 109, 5568–5575.
Csete, M.E., and Doyle, J.C. (2002). Reverse engineering of biological
complexity. Science 295, 1664–1669.
Eldar, A., Dorfman, R., Weiss, D., Ashe, H., Shilo, B.Z., and Barkai, N. (2002).
Robustness of the BMP morphogen gradient in Drosophila embryonic
patterning. Nature 419, 304–308.
Fe ´lix, M.-A. (2007). Cryptic quantitative evolution of the vulva intercellular
signaling network in Caenorhabditis. Curr. Biol. 17, 103–114.
Fe ´lix, M.A. (2012). Caenorhabditis elegans vulval cell fate patterning. Phys.
Biol. 9, 045001.
Fe ´lix, M.A., and Barkoulas, M. (2012). Robustness and flexibility in nematode
vulva development. Trends Genet. 28, 185–195.
Ferguson, E.L., and Horvitz, H.R. (1985). Identification and characterization of
22 genes that affect the vulval cell lineages of the nematode Caenorhabditis
elegans. Genetics 110, 17–72.
Ferguson, E.L., Sternberg, P.W., and Horvitz, H.R. (1987). A genetic pathway
for the specification of the vulval cell lineages of Caenorhabditis elegans.
Nature 326, 259–267.
Frøkjaer-Jensen, C., Davis, M.W., Hopkins, C.E., Newman, B.J., Thummel,
J.M., Olesen, S.P., Grunnet, M., and Jorgensen, E.M. (2008). Single-copy
insertion of transgenes in Caenorhabditis elegans. Nat. Genet. 40, 1375–1383.
Giurumescu, C.A., Sternberg, P.W., and Asthagiri, A.R. (2006). Intercellular
coupling amplifies fate segregation during Caenorhabditis elegans vulval
development. Proc. Natl. Acad. Sci. USA 103, 1331–1336.
Giurumescu, C.A., Sternberg, P.W., and Asthagiri, A.R. (2009). Predicting
phenotypic diversity and the underlying quantitative molecular transitions.
PLoS Comput. Biol. 5, e1000354.
Greenwald, I.S., Sternberg, P.W., and Horvitz, H.R. (1983). The lin-12 locus
specifies cell fates in Caenorhabditis elegans. Cell 34, 435–444.
Gregor, T., Wieschaus, E.F., McGregor, A.P., Bialek, W., and Tank, D.W.
(2007). Stability and nuclear dynamics of the bicoid morphogen gradient.
Cell 130, 141–152.
Griffiths, A., Miller, J., Suzuki, D., Lewontin, R., and Gelbart, W. (2000). An
Introduction to Genetic Analysis, Seventh Edition (New York: W.H. Freeman).
Gutie ´rrez, J. (2009). A developmental systems perspective on epistasis:
computational exploration of mutational interactions in model developmental
regulatory networks. PLoS ONE 4, e6823.
switch between vulval and nonvulval cell fates in Caenorhabditis elegans.
Genetics 126, 899–913.
Hill, R.J., and Sternberg, P.W. (1992). The gene lin-3 encodes an inductive
signal for vulval development in C. elegans. Nature 358, 470–476.
Hoyos, E., Kim, K., Milloz, J., Barkoulas, M., Pe ´nigault, J.B., Munro, E., and
Fe ´lix, M.A. (2011). Quantitative variation in autocrine signaling and pathway
crosstalk in the Caenorhabditis vulval network. Curr. Biol. 21, 527–538.
Hwang, B.J., and Sternberg, P.W. (2004). A cell-specific enhancer that spec-
ifies lin-3 expression in the C. elegans anchor cell for vulval development.
Development 131, 143–151.
Katz, W.S., Hill, R.J., Clandinin, T.R., and Sternberg, P.W. (1995). Different
levels of the C. elegans growth factor LIN-3 promote distinct vulval precursor
fates. Cell 82, 297–307.
Kenyon, C. (1995). A perfect vulva every time: gradients and signaling
cascades in C. elegans. Cell 82, 171–174.
Kiontke, K., Barrie `re, A., Kolotuev, I., Podbilewicz, B., Sommer, R.J., Fitch,
D.H.A., and Fe ´lix, M.-A. (2007). Trends, stasis, and drift in the evolution of
nematode vulva development. Curr. Biol. 17, 1925–1937.
Kitano, H. (2004). Biological robustness. Nat. Rev. Genet. 5, 826–837.
Kitano, H. (2007). A robustness-based approach to systems-oriented drug
design. Nat. Rev. Drug Discov. 6, 202–210.
Komatsu, H., Chao, M.Y., Larkins-Ford, J., Corkins, M.E., Somers, G.A.,
Tucey, T., Dionne, H.M., White, J.Q., Wani, K., Boxem, M., and Hart, A.C.
(2008). OSM-11 facilitates LIN-12 Notch signaling during Caenorhabditis
elegans vulval development. PLoS Biol. 6, e196.
Krejcı ´, A., Bernard, F., Housden, B.E., Collins, S., and Bray, S.J. (2009). Direct
response to Notch activation: signaling crosstalk and incoherent logic. Sci.
Signal. 2, ra1.
Levy, S.F., and Siegal, M.L. (2008). Network hubs buffer environmental varia-
tion in Saccharomyces cerevisiae. PLoS Biol. 6, e264.
sion and timing of C. elegans vulval fate patterning. Curr. Biol. 20, 1875–1879.
Ma, W., Lai, L., Ouyang, Q., and Tang, C. (2006). Robustness and modular
design of the Drosophila segment polarity network. Mol. Syst. Biol. 2, 70.
Masel, J., and Siegal, M.L. (2009). Robustness: mechanisms and conse-
quences. Trends Genet. 25, 395–403.
Meir, E., von Dassow, G., Munro, E., and Odell, G.M. (2002). Robustness, flex-
ibility, and the role of lateral inhibition in the neurogenic network. Curr. Biol. 12,
Quantitative Analysis of Vulva Patterning Network
74 Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc.
the intercellular signaling network underlying a robust developmental system.
Genes Dev. 22, 3064–3075.
Moriya, H., Shimizu-Yoshida, Y., and Kitano, H. (2006). In vivo robustness
analysis of cell division cycle genes in Saccharomyces cerevisiae. PLoS
Genet. 2, e111.
Pe ´nigault, J.B., and Fe ´lix, M.A. (2011). High sensitivity of C. elegans vulval
precursor cells to the dose of posterior Wnts. Dev. Biol. 357, 428–438.
Phillips, P.C. (2008). Epistasis—the essential role of gene interactions in the
structure and evolution of genetic systems. Nat. Rev. Genet. 9, 855–867.
Qadota, H., Inoue, M., Hikita, T., Ko ¨ppen, M., Hardin, J.D., Amano, M.,
Moerman, D.G., and Kaibuchi, K. (2007). Establishment of a tissue-specific
RNAi system in C. elegans. Gene 400, 166–173.
Raj, A., van den Bogaard, P., Rifkin, S.A., van Oudenaarden, A., and Tyagi, S.
(2008). Imaging individual mRNA molecules using multiple singly labeled
probes. Nat. Methods 5, 877–879.
Raj, A., Rifkin, S.A., Andersen, E., and van Oudenaarden, A. (2010). Variability
in gene expression underlies incomplete penetrance. Nature 463, 913–918.
Saffer, A.M., Kim, D.H., van Oudenaarden, A., and Horvitz, H.R. (2011). The
Caenorhabditis elegans synthetic multivulva genes prevent ras pathway acti-
vation by tightly repressing global ectopic expression of lin-3 EGF. PLoS
Genet. 7, e1002418.
Shaye, D.D., and Greenwald, I. (2002). Endocytosis-mediated downregulation
of LIN-12/Notch upon Ras activation in Caenorhabditis elegans. Nature 420,
Siegal, M.L., and Bergman, A. (2002). Waddington’s canalization revisited:
developmental stability and evolution. Proc. Natl. Acad. Sci. USA 99, 10528–
Simmer, F., Tijsterman, M., Parrish, S., Koushika, S.P., Nonet, M.L., Fire, A.,
Ahringer, J., and Plasterk, R.H. (2002). Loss of the putative RNA-directed
RNA polymerase RRF-3 makes C. elegans hypersensitive to RNAi. Curr.
Biol. 12, 1317–1319.
elegans vulval induction. Nature 375, 142–146.
Sternberg, P.W. (1988). Lateral inhibition during vulval induction in
Caenorhabditis elegans. Nature 335, 551–554.
Sternberg, P.W. (2005). Vulval development. In Wormbook, The C. elegans
Research Community, ed., http://dx.doi.org/10.1895/wormbook.1.6.1, www.
lular signaling pathways specifies three cell fates during vulval induction in
C. elegans. Cell 58, 679–693.
Sternberg, P.W., and Han, M. (1998). Genetics of RAS signaling in C. elegans.
Trends Genet. 14, 466–472.
Struhl, G., Fitzgerald, K., and Greenwald, I. (1993). Intrinsic activity of the Lin-
12 and Notch intracellular domains in vivo. Cell 74, 331–345.
Sundaram, M., and Greenwald, I. (1993). Genetic and phenotypic studies
of hypomorphic lin-12 mutants in Caenorhabditis elegans. Genetics 135,
Tan, P.B., Lackner, M.R., and Kim, S.K. (1998). MAP kinase signaling speci-
ficity mediated by the LIN-1 Ets/LIN-31 WH transcription factor complex
during C. elegans vulval induction. Cell 93, 569–580.
von Dassow, G., Meir, E., Munro, E.M., and Odell, G.M. (2000). The segment
polarity network is a robust developmental module. Nature 406, 188–192.
Wagner, A. (2005). Robustness and Evolvability in Living Systems (Princeton,
NJ: Princeton University Press).
Wang, M., and Sternberg, P.W. (1999). Competence and commitment of
Caenorhabditis elegans vulval precursor cells. Dev. Biol. 212, 12–24.
Yoo, A.S., Bais, C.,and Greenwald, I.(2004). Crosstalk between the EGFR and
LIN-12/Notch pathways in C. elegans vulval development. Science 303,
Zand, T.P., Reiner, D.J., and Der, C.J. (2011). Ras effector switching promotes
divergent cell fates in C. elegans vulval patterning. Dev. Cell 20, 84–96.
Quantitative Analysis of Vulva Patterning Network
Developmental Cell 24, 64–75, January 14, 2013 ª2013 Elsevier Inc. 75