Oncogene (1998) 17, 1109 ± 1118
 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00
http://www.stockton-press.co.uk/onc
The C. elegans MDL-1 and MXL-1 proteins can functionally substitute for
vertebrate MAD and MAX
Je€rey Yuan1,2, Rebecca S Tirabassi1, Andrew B Bush1 and Michael D Cole1,2
1
Department of Molecular Biology, Princeton University, Princeton, New Jersey 08544-1014, USA
The genes of the myc/max/mad family play an important
role in controlling cell proliferation and di€erentiation.
We have identi®ed the ®rst homologues of the mad and
max genes in the nematode C. elegans, which we have
named mdl-1 and mxl-1 respectively. Like the vertebrate
MAD proteins, MDL-1 binds an E-box DNA sequence
(CACGTG) when dimerized with MXL-1. However,
unlike vertebrate MAX, MXL-1 can not form homodimers and bind to DNA alone. Promoter fusions to a
GFP reporter suggest that these genes are coexpressed in
posterior intestinal and post-mitotic neuronal cells during
larval development. The coexpression in the posterior
intestinal cells occurs before their ®nal division at the
end of the L1 stage and persists afterwards, demonstrating that mad and max expression can be correlated
directly to the cell cycle state of an individual cell type.
These data also show that mxl-1 is an obligate partner
for mdl-1 in vivo and in vitro and indicate that these
genes may play an important role in post-embryonic
development. Finally, MDL-1 can suppress activated cMYC/RAS-induced focus formation in a rat embryo
®broblast transformation assay. Like the vertebrate
MAD protein, MDL-1 activity in suppressing transformation is dependent on a functional SIN3 interaction
domain.
Keywords: C. elegans; MAD; MAX
Introduction
The control of cell proliferation and di€erentiation is a
fundamentally important process for any metazoan. In
the nematode, Caenorhabditis elegans, proliferation and
di€erentiation form a tightly controlled program since
the adult hermaphrodite has precisely 959 somatic
nuclei (Wood, 1988; Sulston and Horvitz, 1977;
Sulston et al., 1983). Very few genetic loci have been
uncovered that a€ect the somatic proliferation program
in C. elegans. One example is the semidominant gain of
function allele of the heterochronic gene lin-14 which
causes a limited number of cell lineages, such as the T
lineage, to repeat its L1 larval stage division pattern
during the L2 larval stage. This leads to the production
of 11 daughter cells where normally there are only four
daughter cells (Ambros and Horvitz, 1984; Ruvkun
and Giusto, 1989). Another example is the cul-1 gene.
Correspondence: MD Cole
2
Current address: Merck & Co, Inc, Department of Bioinformatics,
PO Box 2000, RY80A-1,Rahway, New Jersey, 07065, USA
Received 7 November 1997; revised 6 April 1998; accepted 7 April
1998
Loss of cul-1 activity results in increased G1 CYCLIN/
CYCLIN DEPENDENT KINASE complex activity
and hence promotes the G1 to S cell cycle progression.
This acceleration of G1 into S causes several somatic
lineages to proliferate abnormally (Kipreos et al.,
1996). Nevertheless, none of the known mutations
result in uncontrolled somatic proliferation and at
most they induce only one or two extra rounds of postembryonic proliferation in a few somatic lineages. This
lack of somatic proliferation mutants may re¯ect both
an inadequate survey for such mutations and/or a
molecular mechanism that prevents uncontrolled
proliferation.
In contrast to the dearth of genes a€ecting general
cell proliferation in C. elegans, many genes exist in
vertebrates that a€ect proliferation in a wide range of
cell lineages, and a large number of these genes, known
as oncogenes, contribute to the growth of cancer cells.
Prominent among these are the genes of the myc
family, which include c-myc, N-myc and L-myc, and
these in turn are members of a larger family that
includes max, mad, mnt and rox. The myc genes are
required for cell proliferation. Furthermore, when
overexpressed they can block di€erentiation or induce
apoptosis (Evan and Littlewood, 1993; Henriksson and
LuÈscher, 1996). On the other hand, mad genes are
induced in di€erentiating cells and can antagonize myc
action when overexpressed (Ayer et al., 1993; Cerni et
al., 1995; Koskinen et al., 1995; Lahoz et al., 1994;
Zervos et al., 1993). Recently, two new genes mnt and
rox, have also been shown to act as transcriptional
repressors possibly to antagonize myc action (Hurlin et
al., 1997; Meroni et al., 1997). Individually MYC,
MAD, MNT, or ROX are unable to form strong
homodimers, and DNA binding by each requires
heterodimerization with MAX. This dimerization is
mediated by the helix ± loop ± helix leucine-zipper
domain. In contrast, MAX can form homodimers
and bind to DNA alone, although its DNA binding
activity is suppressed in vivo by CASEIN KINASE II
phosphorylation (Berberich and Cole, 1992; Berberich
et al., 1992; Blackwood and Eisenman, 1991; Koskinen
et al., 1994).
The vertebrate mad family currently consists of four
genes, mad1, mad3, mad4, and mxi1 which can all
induce cell di€erentiation and suppress cell proliferation (Ayer et al., 1993; Hurlin et al., 1995; Zervos et
al., 1993). In contrast to the weak but reproducible
transcriptional activation by the myc genes, the MAD
protein subfamily forms a complex with the transcriptional repressor SIN3 to suppress transcription (Ayer
et al., 1995; Schreiber-Agus et al., 1995). Deletions in
the mxi1 gene have been found in human prostate
tumors, raising the possibility that these genes could
function as tumor suppressors (Eagle et al., 1995).
C elegans MAD and MAX proteins
J Yuan et al
1110
Despite the knowledge that the mad genes a€ect cell
proliferation and di€erentiation when ectopically
expressed, little is known of their role in normal cell
physiology or the genes that they control. Compounding this problem has been the lack of any homologues
in genetically amenable invertebrate organisms. In this
paper we describe the identi®cation and initial
characterization of mad and max homologues in the
nematode C. elegans.
Results
Identi®cation of mdl-1 and mxl-1 genes
We identi®ed a mad-like gene in C. elegans by
searching through the partially sequenced C. elegans
genome using the tblastn search program (Altschul et
al., 1991). From this search, a putative open reading
frame that is similar to human mad1 was identi®ed on
cosmid R03E9 which resides on chromosome X. To
determine if the predicted ORF encoded a transcribed
gene, we assayed both a cDNA library and genomic
DNA by PCR using two oligonucleotide primers that
¯ank the potential basic helix ± loop ± helix leucine
zipper (B-HLH-LZ) domain. From the genomic
sequence, the predicted ORF contains a putative
290 bp intron which interrupts the predicted B-HLHLZ domain. Therefore, the PCR product from genomic
DNA should di€er from the PCR product from the
cDNA library by 290 bp. This di€erence was observed,
indicating that the predicted ORF does encode a
transcribed gene (data not shown). Furthermore, the
ampli®ed PCR fragments did match the genomic
sequence. The ampli®ed PCR fragment was then used
to probe a C elegans cDNA library. A full length clone
was isolated which encodes a 281 amino acid protein
that matches the predicted gene (Figure 1a). Since the
gene is most similar to vertebrate mad, we named it
mdl-1 (mad-like gene 1).
The assignment of MDL-1 as a member of the mad
family is based on ®ve criteria: (1) The protein is 60%
identical and 72% similar to human MAD1 within the
basic helix ± loop ± helix domain. (2) The overall
organization of the MDL-1 protein is similar to the
vertebrate proteins. Like mammalian MAD family
proteins, the B-HLH-LZ domain of MDL-1 is in the
center of the protein, which is in contrast to the Cterminal location of the B-HLH-LZ domain in MYC
or encompassing most of the protein in MAX. (3)
There are several residues in the B-HLH-LZ domain
that are speci®c to the MAD family. These conserved
Figure 1 Primary sequences of MDL-1 and MXL-1 and sequence alignments to vertebrate homologues. (a) A multiple sequence
alignment of MDL-1 and MXL-1 to members of the MAD and MAX protein families. Human, Xenopus, and Drosophila MAXs
were used for comparison. For the MAD family, Human MAD1, MXI1, Mouse MAD3, MAD4 and MNT and ROX were used for
comparison. The primary sequences of MDL-1 and MXL-1 are titled in bold. Known and potential CASEIN KINASE II
phosphorylation sites are boxed, as are the B-HLH-LZ domain and the potential and known SIN3 interaction domain. (b) An
unrooted phylogram of the aligned proteins is shown
C elegans MAD and MAX proteins
J Yuan et al
residues include the cysteine at position 114, the TTL
motif at position 132, the HI motif at position 143, and
the EQ motif at position 164 (Figure 1a). (4) The C.
elegans mdl-1 gene contains an amphipathic a-helix in
the extreme N-terminus analogous to the segment in
the mammalian genes that mediates binding to the
SIN3 repressor. The consensus sequence for SIN3
binding is N(L/I)**LL*AA*(L/Y)L(E/D) which is
similar to the MDL-1 sequence NLGHLLTAARLLD
from amino acids 6 ± 18 (Figure 1a). A predicted SIN3
homologue was identi®ed using the tblastn computer
program (Altschul et al., 1991) on cosmid F02E9 of C.
elegans chromosome I. (5) The multiple sequence
alignment shown in Figure 1a of all the MAD/MNT/
MAX family members emphasizes that MDL-1 belongs
to the MAD family given its size and organization and
not to the MAX or MNT/ROX family (Figure 1a).
Furthermore, a phylogram plot of the evolutionary
relationships between MDL-1 and other family
members further emphasizes that it is a member of
the MAD family and not of the MAX or MNT/ROX
family (Figure 1b).
Since all known MAD proteins heterodimerize with
MAX, we asked whether the C. elegans genome
encoded a MAX-like protein with which MDL-1
could dimerize. A fusion protein that contained only
the MDL-1 B-HLH-LZ domain was produced and
used in a protein interaction screen. Screening of a
C. elegans mixed stage cDNA expression library
yielded three related phage isolates that expressed
MDL-1 interacting proteins and each contained
nearly identical 450 bp inserts. The inserts contained
a novel max-like gene which was named mxl-1 (maxlike gene 1). This gene encodes a 123 amino acid
long polypeptide which is 64% identical and 74%
similar to Drosophila MAX and 62% identical and
74% similar to Human MAX within the B-HLH
domain (Figure 1a). Conservation is looser within the
LZ domain especially when comparing the invertebrate MAX proteins to the vertebrate MAX proteins
(Figure 1a). Besides containing a similar B-HLH
domain, MXL-1 has a potential CASEIN KINASE
II phosphorylation site at amino acid position 5
(Figure 1a). Furthermore, like its homologues, the BHLH-LZ domain of MXL-1 is near the N-terminus
of the protein, although the protein is small enough
that this B-HLH-LZ domain encompasses nearly the
entire protein (Figure 1a). An alignment and
phylogram of MXL-1 with other members of the
MAX and MAD/MNT/ROX family clearly indicates
that MXL-1 is related to the MAX family (Figure
1b). Finally, the MXL-1 cDNA clone was used to
obtain a genomic clone which was ®ngerprinted to
cosmid C08E12, which allowed us to map mxl-1 to
chromosome V.
MDL-1 and MXL-1 are obligate dimerization partners
In vertebrates, the MAX protein is a common
dimerization partner for both the MYC and MAD
family of proteins and we anticipated that the same
might hold true for C. elegans. Since MXL-1 was
isolated using MDL-1 protein as a probe, it was clear
that these two proteins were likely to form stable
heterodimers, but we wanted to analyse their DNA
binding and dimerization potential further. The
vertebrate MAD/MAX heterodimers and MAX homodimers bind to the DNA E-box motif, CACGTG
(Hurlin et al., 1995; Ayer et al., 1995), so we tested
MDL-1 and MXL-1 to determine if they have similar
activity. Not surprisingly, the MDL-1 protein alone (BHLH-LZ domain; amino acids 89 ± 191) had no DNA
Figure 2 MDL-1 and MXL-1 heterodimers bind DNA in an electrophoretic mobility shift assay. (a) Bacterially puri®ed (His)6tagged MXL-1 and truncated MDL-1 protein were assayed either individually or together for DNA binding activity with 32P-labeled
MYC consensus (GACCACGTGGTC) oligonucleotide probe in an EMSA. (b) Truncated MDL-1 protein was assayed for DNA
binding activity with 32P-labeled MYC consensus oligonucleotide in the presence of either truncated mouse c-MYC (amino acids
182-439) or mouse MAX. A control of mouse MYC with mouse MAX was included to demonstrate the MAX homodimer and
MYC/MAX heterodimer shifts. 100 ng, 50 ng and 10 ng of MDL-1 were used. (c) EMSA analysis of MXL-1 protein with mouse
MAX or MYC. All assays, unless otherwise indicated, used *10 ng of mouse MAX, *20 ng of mouse MYC, *10 ± 50 ng of
MDL-1, *10 ng of MXL-1, and 1 ng of 32P-labeled MYC consensus oligonucleotide
1111
C elegans MAD and MAX proteins
J Yuan et al
1112
binding activity in an electrophoretic mobility shift
assay (EMSA), implying that the protein does not form
homodimers which is similar to vertebrate MAD
(Figure 2a and b). When the MDL-1 binding reaction
was supplemented with full length MXL-1 protein, a
complex formed that bound tightly to the vertebrate
MYC/MAX consensus sequence oligonucleotide (Figure 2a). Surprisingly, we found that the dimerization
behavior of the full length MXL-1 protein di€ered
signi®cantly from its other homologues, including
Drosophila MAX. While mouse MAX readily forms
homodimers (Figure 2b and c), MXL-1 will not
dimerize (Figure 2c), even at high concentrations
(4100 ng) (data not shown).
We also tested whether the C elegans proteins could
interact with the vertebrate MYC and MAX proteins.
MDL-1 could dimerize with mouse MAX and bind to
the CACGTG sequence in an EMSA (Figure 2b),
although the DNA binding activity of this interspecies
heterodimer was much weaker than the intraspecies
MDL-1/MXL-1 heterodimer. None of the C. elegans
proteins could interact with mouse c-MYC (Figure 2b
and c).
interactions and the presence of four attractive
interactions. This analysis also supports the observation that MXL-1 will not form homodimers since
the two aspartic acid to glutamic acid interactions
and the two lysine to arginine interactions will
overwhelm the two attractive glutamic acid to lysine
interactions (Figure 4). Similarly, the predictions also
support the observation that MDL-1 will not form
homodimers.
Temporal expression of MDL-1 and MXL-1
The ligand binding assays and electrophoretic
mobility assays demonstrate that MDL-1 and MXL1 are dimerization partners. However, it was
important to demonstrate that these genes have
similar, or at least overlapping, temporal and spatial
Analysis of protein-protein interactions in the absence of
DNA binding
Since the MDL-1/MXL-1 heterodimer bound to
DNA but neither protein alone was able to do so,
the simplest explanation is that neither homodimer is
stable in the EMSA. However, this result does not
exclude the possibility that MDL-1 and/or MXL-1
proteins can homodimerize, yet not bind to the
CACGTG motif o€ered. The original isolation of
MXL-1 by screening with MDL-1 protein implies
that DNA binding is not necessary for the formation
of heterodimers, as is true for most, if not all,
proteins in the B-HLH-LZ and B-Zip families.
Therefore, to study the interactions further, ligand
blotting analysis was performed. Puri®ed truncated
MDL-1 protein and full length MXL-1 protein were
electrophoresed on an SDS-polyacrylamide gel
(Figure 3a) and transferred onto nylon membranes.
The membranes were probed with either 32P-labeled
MXL-1 protein or 32P-labeled MDL-1 protein. These
experiments revealed that 32P-labeled MXL-1 will only
associate with MDL-1 (Figure 3b) and conversely 32Plabeled MDL-1 will only associate with MXL-1
(Figure 3c). Neither protein interacted with itself or
the proteins within the molecular weight standards
lane. These results support the EMSA data that
MDL-1 and MXL-1 can form heterodimers even in
the absence of DNA, and furthermore, that neither
protein can homodimerize.
Amati and coworkers have examined which amino
acids within the leucine zipper domain of c-MYC
and MAX are involved in dimerization (Amati et
al., 1993). Their work indicates that amino acids at
the 1 and 4 position of each leucine repeat are
important for the formation of dimers. These amino
acids have the potential to form electrostatic
interactions or repulsions with amino acids in the
leucine zipper domain of a potential dimerization
partner (Figure 4). From this predictive analysis, the
observed dimerization between MDL-1 and MXL-1
is not surprising, given the lack of repulsive
Figure 3 Ligand blotting analysis indicates MDL-1 and MXL-1
will form heterodimers but not homodimers. (a) Coomassie blue
stain of puri®ed MDL-1 and MXL-1 proteins electrophoresed on
a SDS polyacrylamide gel. The MDL-1 protein is truncated
(amino acids 89 ± 191), the MXL-1 protein is full length. Both
proteins have His6 and the RRASV motif fused at the amino
terminus (Armand et al., 1994). (b) An equivalent gel to panel a
was electrophoresed and transfered onto nylon membranes and
probed with *50 ng of 32P-labeled MXL-1 protein. The
molecular weight standard markers were also transferred as a
control. (c) Same as panel b but probed with *50 ng of 32Plabeled MDL-1 protein
Figure 4 Predicted interactions within the leucine zipper of
MDL-1 and MXL-1 during homodimer and heterodimer
formation. The leucine repeats and the amino acids predicted to
form salt bridges are indicated. Attractive interactions are
indicated by arrows and repulsions are indication by bars
C elegans MAD and MAX proteins
J Yuan et al
expression. To examine the temporal expression of
mdl-1 and mxl-1 at di€erent stages of development,
we used RT ± PCR analysis on polyA-containing
mRNA from staged nematodes. Both genes are
trans-spliced to the SL1 leader RNA sequence
(Figure 5 and data not shown), which is transcribed
from a separate locus in the genome and added posttranscriptionally. The trans-spliced leader sequence is
a common feature of many C. elegans transcripts
(Krause and Hirsh, 1987). The presence of the SL1
leader RNA sequence greatly simpli®ed the RT ± PCR
analysis, since any contaminating genomic DNA
would not be ampli®ed. PCR reactions using primers
against SL1 and either mdl-1 or mxl-1 led to the
ampli®cation of bands that were consistent with the
predicted full length sequence (arrows in Figure 5),
which con®rms the structure of the genes. From these
experiments, it is clear that mdl-1 and mxl-1 have
similar temporal expression patterns. Both genes are
expressed weakly in L1 animals and strongly
throughout the remainder of larval development.
Neither gene is detectable during embryogenesis
(Figure 5a). To control for the integrity of the
RNA and uniformity of the PCR reaction conditions,
we used RT ± PCR in parallel reactions to amplify
the actin gene, act-3, which is also trans-spliced to
the SL1 leader (Figure 5b) (Krause and Hirsh, 1987).
The act-3 product was quantitated by including
[32P]dCTP in the reaction and counting the PCR
product, after ®rst determining that the reaction was
still within a linear range (data not shown). Since
actin is expressed equally throughout development,
the actin signal was used to estimate the relative
levels of mdl-1 and mxl-1 and the values were
arbitrarily normalized to the L3 larval level (Figure
5c). However, it should be noted that this analysis
will not detect expression during late embryogenesis
or late L1 stage, due to the manner in which the
synchronized worms were harvested. The embryonic
worm pool consists primarily of early embryos that
have not been laid by their mothers. Similarly, the
L1 pool will consist of young L1 animals that have
hatched into a foodless environment and these
animals have arrested at the L1 diapause. The
remaining stages were more heterogenous. Therefore,
the RT ± PCR analysis is only a coarse semiquantitative gauge of the temporal expression of these
genes.
Figure 5 Reverse Transcriptase Polymerase Chain Reaction analysis of polyA selected mRNA from staged nematodes. 16 ng of
polyA selected RNA from each stage was primed with oligo dT (23mer) primer. No reverse transcriptase reactions and a no RNA
reaction (lane c) were performed for all stages as controls but only shown for the act-3/SL1 reactions. The mdl-1/SL1 and mxl-1/SL1
PCR reactions were electrophoresed on the same 4% polyacrylamide gel. The arrows indicate the ampli®ed band that corresponds
to the predicted size. The act-3/SL1 reactions were ampli®ed with [32P]dCTP and then electrophoresed on a separate 4%
polyacrylamide gel from the mdl-1/mxl-1 reactions. The bands were quantitated by Phosphoimager analysis and the amounts were
compared to the L3 reaction product. The molecular weight markers are pBR322 digested with Msp I
1113
C elegans MAD and MAX proteins
J Yuan et al
1114
MDL-1 is expressed in post-mitotic neurons and
intestinal cells
If MDL-1 and MXL-1 were to form heterodimers in
vivo, then the spatial expression of these two proteins
should also overlap. Therefore, to understand the
expression pattern of mdl-1 and mxl-1, the promoter
of each gene was fused to the green ¯uorescent
protein (GFP) (Chal®e et al., 1994). DNA beginning
at the translational start site and extending 5 kb
upstream of the mdl-1 coding region was fused to
GFP, and this construct was coinjected with the unc119(+) gene to create a transgenic line in an unc-119
(e2498) genetic background. Transgenic animals were
scored for the rescue of the unc-119 (e2498) mutant
phenotype (Fire, 1986). While the possibility exists
that we have not recapitulated the complete expression pattern of the endogenous gene, the expression
pattern was consistent with the RT ± PCR analysis
described above. mdl-1 promoter-driven GFP expression was observed in all larval stages, but most
strongly from late L1 through L4 (Figure 6). There
was strong late L1 expression in the posterior eight
intestinal cells just prior to their division near the end
of the L1 stage (Figure 6c). Furthermore, certain
intestinal cells expressed the transgene more strongly
than others (Figure 6c). In the example shown, the
posterior most pair of intestinal cells expresses the
transgene at a signi®cantly higher level than the more
anterior cells. In other animals, di€erent posterior
intestinal cells express the mdl-1::GFP transgene at a
higher level than their neighboring intestinal cells
(data not shown). However, in every animal,
expression always began at the midpoint of the
animal and ended at the posterior. No expression
was observed in the anterior intestinal cells. After the
L1 stage, the mdl-1::GFP transgene continues to be
expressed in the intestinal cells, although only in some
of the divided binucleated posterior intestinal cells
(Figure 6b). Robust expression was also observed in
many neurons during all larval stages, such as the
CEPDR and CEPVR neurons. These neurons were
readily identi®ed since the transgene expression was
strong enough to outline the axons and dendrites
(Figure 6a). Similarly, the posterior pharyngeal bulb
muscle cells also expressed the transgene (Figure 6a).
From the position and morphology of the cells, the
ventral cord motoneurons appear to express the
transgene in all larval stages (Figure 6b and c).
Expression has also been observed in several
hypodermal cells and body wall muscles, although
their identities have not been veri®ed. Expression of
the mdl-1::GFP transgene was also observed in the Z1
and Z4 cells in the L1 larva, but not in their
daughters (Figure 6d). Since the transgenes are not
integrated into the chromosome in these strains but
are carried as stable extrachromosomal arrays, some
of the variability in the expression pattern may be
due to mosaicism of the array during development.
Nevertheless, the results outlined are from the
aggregate observation of more than 20 animals from
three independent transgenic lines.
The MXL-1 expression pattern is coincidental with
MDL-1 in L1 premitotic intestinal cells
Figure 6 Expression pattern of mdl-1 and mxl-1 from GFP
promoter fusion transgenes. The mdl-1::GFP expression patterns
are shown in panels a ± d. The mxl-1::GFP expression patterns are
shown in panels e ± f. Panels b, d and e were imaged with a 206
or 406 objective on a Leica Axiophot. The remaining panels (a, c
and e) were imaged with a BioRad MRC600 confocal microscope
with a 406 objective. The arrows point to intestinal nuclei. The
arrowheads points to neuronal nuclei. The bars indicate
motoneurons. (a) The mdl-1::GFP reporter construct expressed
in the CEPDR and CEPVR neurons and the posterior pharyngeal
bulb in an L4 animal. (b) The mdl-1::GFP reporter construct
expressed in the nuclei of two binucleate ventral posterior
intestinal cells and in at least four motoneuron nuclei of an L2
animal at 206. The inset is the bright ®eld view of the same
animal. Posterior end is pointed downwards. (c) The posterior
third of a late stage L1 animal. (d) The indicated somatic gonad
cells Z1 and Z4 of a late stage L1 animal express the mdl-1::GFP
reporter construct. (e) The bright ®eld view of a late stage L1
animal at 206. (f) The expression pattern of the mxl-1::GFP
reporter construct in the same animal as in panel f at 406
To determine whether mxl-1 is spatially coexpressed
with mdl-1, a mxl-1::GFP transgene was created.
Again, 5 kb of DNA sequence directly upstream of
the mxl-1 gene was fused to GFP and was injected into
an unc-119(e2498) genetic background. Fluorescence
microscopy of two independent lines showed that
expression of the GFP reporter was strongest in the
posterior intestinal cells, just prior to division in late
L1 (Figure 6f). As with the mdl-1::GFP transgene, the
mxl-1::GFP transgene expression begins at the midpoint of the animal and continues to the posterior most
intestinal cells. The intestinal expression persists into
the L2 stage, although at a reduced level compared to
that in the late L1 premitotic intestinal cells (data not
shown). There was also weak expression in many anal
and head neurons during the L2 ± L4 stages (data not
shown). This expression data is again consistent with
the RT ± PCR analysis. Coexpression of mdl-1 and mxl1 in the posterior intestinal cells supports the
biochemical data indicating that these two proteins
interact.
Suppression of activated C-MYC/RAS induced foci by
MDL-1 requires an intact SIN3 interaction domain
Overexpression of MDL-1 or of an antimorphic
version of MXL-1 (basic region deleted) in C. elegans
C elegans MAD and MAX proteins
J Yuan et al
failed to suppress growth or a€ect any lineage (data
not shown), therefore a heterologous approach to
studying the function of these genes was taken. Earlier
studies have demonstrated that vertebrate MAD
proteins have growth suppressive activities as demonstrated by their ability to suppress activated c-MYC/
RAS-induced cell transformation in a REF focus assay
(Schreiber-Agus et al., 1994; Lahoz et al., 1994).
Furthermore, this activity is dependent upon the
presence of an intact SIN3 interaction domain (SID).
Disruption or deletion of this domain will relieve the
suppression of cell transformation. SIN3 has been
recently shown to interact with histone deacetylase, a
known transcriptional repressor (Alland et al., 1997;
Hassig et al., 1997; Heinzel et al., 1997; Kadosh and
Struhl, 1997; Laherty et al., 1997; Nagy et al., 1997;
Zhang et al., 1997).
To test whether MDL-1 has biological functions
similar to its vertebrate homologue, cotransfection
experiments were conducted with c-myc and H-ras
oncogenes into REFs. Full length MDL-1 suppressed
the ability of activated RAS and c-MYC to induce
cell transformation to a level that was approximately
25% of the vector control (Figure 7b). Cotransfection
of REFs with both MXL-1 and MDL-1 did not
further increase the level of the c-MYC/RAS
suppression. Deletion of the SID from MDL-1
(DMDL-1) destroyed its ability to suppress activated
c-MYC/RAS-induced cell transformation (Figure 7a
and b), which parallels the activity of vertebrate MAD
in this assay. The MXL-1 protein gave a partial
suppression of cell transformation (to 55% of control
levels). This can be attributed to the possibility that it
dimerizes weakly to endogenous c-MYC but does not
bind to DNA, thereby preventing c-MYC from
inducing cell transformation. This partial suppression
by MXL-1 is lost when the DMDL-1 protein is
coexpressed (Figure 7b). In this scenario, MXL-1 may
bind to DMDL-1 with a higher anity than it would
to c-MYC and hence reverse any a€ect on c-MYC/
RAS-induced cell transformation. The ability of
MDL-1 or DMDL-1 to function in REFS without
MXL-1 con®rms the electrophoretic mobility shift
data shown in Figure 2, where MDL-1 can also
dimerize with vertebrate MAX and bind DNA.
Finally, in the absence of c-MYC, MDL-1, DMDL1, MXL-1, or vector alone had no e€ect on REF
monolayers with or without H-RAS coexpression
(data not shown).
Discussion
Relationship of the C elegans mad/max-like genes to
their vertebrate homologues
We demonstrate in this paper that the nematode C
elegans has at least two genes that are homologues of
the vertebrate myc/mad/max gene family. One of
these genes, mxl-1, encodes a B-HLH-LZ protein
structurally similar to vertebrate MAX, while the
other gene, mdl-1, encodes a B-HLH-LZ protein
similar to vertebrate MAD. These two genes appear
to be obligate partners since the encoded proteins
needed to form heterodimers to bind DNA and their
temporal and spatial expression are largely coincident, suggesting that each gene requires the other to
e€ect its function. In contrast, vertebrate MAX may
have some di€erent functions in the cell since it can
homodimerize. What function might the mxl-1 gene
have by itself in C. elegans, if its sole purpose is to
dimerize with mdl-1? Why not evolve an mdl-1 that
can form homodimers? It remains possible that mxl-1
has other dimerization partners than mdl-1, that it
may have some function as a monomer, or that it
homodimerizes under conditions di€erent than we
provided in vitro. Searches for alternative partners for
mxl-1 have not yet yielded any genes other than mdl1 (data not shown).
The function of mad-like genes
Figure 7 C elegans MDL-1 suppresses rat embryo ®broblast
transformation. (a) The ®rst 40 amino acids of the wild type
MDL-1 is shown on top and the corresponding region of DMDL1 with a deletion of the SIN3 interaction domain is shown below.
(b) REFs were transfected with 1 mg each of activated H-ras and
c-myc plasmids and supplemented with 1 mg of each of the
varying constructs. The number of foci with c-myc, H-ras, and
vector was set to 100%. The data are the average of three
independent trials in which each gene combination contained 3 ± 4
plates
Among members of the MYC/MAD/MAX network
of interacting proteins a key question that emerges is
how each protein contributes to the growth and
di€erentiation of individual cells. In some cell lines,
there is a switch from MYC/MAX heterodimers to
MAD/MAX heterodimers as cells withdraw from the
cell cycle and terminally di€erentiate (Ayer and
Eisenman, 1993). This result has led to the
proposition that induction of MAD antagonizes
MYC activity and accelerates cell cycle withdrawal.
1115
C elegans MAD and MAX proteins
J Yuan et al
1116
In support of this, MDL-1 can suppress activated cMYC/RAS-induced cell transformation in rat embryo
®broblasts. This suppression requires the C. elegans
domain similar to vertebrate SID and the DNA
binding domain to function in vertebrate cells.
The temporal and spatial expression studies of
mdl-1 also support a role for mad-like genes to
function in di€erentiated post-mitotic cells from C.
elegans to vertebrates. Like the mouse mad1, mad3
and mad4 genes, mdl-1 is expressed abundantly in
post-mitotic neurons, and high levels of expression
were also observed in posterior intestinal cells and
cells of the pharyngeal muscle. Nevertheless, mdl-1
expression is not exclusively restricted to differentiated cells, since it is also observed in a few
premitotic cells. Thus, mdl-1 shares some of the
properties of the vertebrate mxi1 gene which is
expressed in both pre- and post-mitotic cells
(Blackwood and Eisenman, 1991; Eagle et al.,
1995) (Figure 6).
In particular, it is interesting to consider a role
for mdl-1 in the posterior intestinal cells. Why is
mdl-1 expressed exclusively in these cells and not
their anterior siblings? The posterior intestinal cells
split from their anterior siblings during the ®rst cell
division of the E (intestinal) lineage just before
gastrulation. This division would result in the
posterior cells having a di€erent set of inductive
signals from the anterior cells, which may result in
inducing cell speci®c factors required during larval
development for mdl-1/mxl-1 expression. The major
di€erence between the anterior and posterior
intestinal cells is that the posterior 6 ± 8 cells divide
during the L1 stage, whereas the anterior cells never
undergo nuclear division but endoreduplicate their
chromosomes. Therefore, we can speculate that mdl1 may be required for activities speci®c to the
di€erentiated posterior intestinal cells or have a
speci®c dual role in both proliferation and differentiation of these cells.
Are there C. elegans myc-like genes?
We have been unable to ®nd any additional B-HLHLZ proteins through interaction screening with MDL-1
and MXL-1. Furthermore, no C. elegans myc-like gene
has been identi®ed in the 490% of the genome that
has been sequenced. Another max-like gene, mxl-2 has
been identi®ed in the genome whose gene product will
not interact with either MDL-1 or MXL-1 (Yuan and
Cole, manuscript in preparation). This raises two
possibilities, the ®rst of which is that C. elegans myc
is underrepresented in the cDNA libraries we screened
and resides in the last 10% of the genome that has not
been sequenced. The other possible scenario is that C.
elegans does not have a myc-like gene, although
Drosophila myc and max genes have been identi®ed
(Gallant et al., 1996; Schreiber-Agus et al., 1997).
Perhaps C. elegans does not require a myc-like gene
because it has a very restricted proliferation program.
Therefore, it becomes particularly important to
determine the null phenotypes of mdl-1 and mxl-1 to
resolve their developmental functions and hence gain
further insight into their function in C. elegans and
also understand the function of their vertebrate
counterparts.
Materials and methods
Isolation of mdl-1 cDNA
The human MAX protein was used as the query sequence
to initiate a remote tblastn (Altschul et al., 1991) search of
the C. elegans genomic sequence database (http://
genome.wustl.edu/gsc/programs/sysadmin/C.elegans_blast_
server.html). A region of high similarity was identi®ed
on cosmid R03E9 from positions *17000 to *17500.
From this sequence, two oligonucleotide primers, cml-1-1
(GTGGGATCCGCTAGAAATCTTCGAAGCACAG) and
cml-1-2
(GGGAAGCTTTGGCAAGCTTGGCTAGTTGCTTG) were synthesized. The cml-1-1 primer begins at
position 16976, just 5' of the basic domain while the cml-12 primer is located just 3' of the leucine zipper domain at
position 17576. Using these primers, 35 cycles of PCR were
performed on a diluted C. elegans mixed stage lgt11 cDNA
expression library (courtesy of A Fire). The reaction
conditions were 948C - 1', 578C - 1', 728C - 1' using Taq
DNA polymerase (Boehringer Mannheim) in an Ericomp
Thermal Cycler. The reaction product was sequenced and
random primer labeled with 32P-dCTP and Klenow DNA
polymerase (New England Biolabs). The labeled DNA was
then used to probe the same C. elegans cDNA library. Two
clones were isolated. The inserts were sequenced by
dideoxy sequencing using the Sequenase Version 2.0
sequencing kit (Amersham). The Genbank database
accession number for MDL-1 is U82968.
Preparation of MDL-1 and MXL-1 fusion proteins
A (His)6 fusion protein that contains the B-HLH-LZ
domain of MDL-1 was produced by subcloning the above
ampli®ed PCR fragment into the pHis6HMK vector. The
fusion protein was isolated and puri®ed as described
(Armand et al., 1994). A full length (His)6-MXL-1 fusion
protein was prepared identically to the (His)6-MDL-1
protein. The purity of each fusion protein was assayed by
SDS-polyacrylamide gel electrophoresis (Laemmli, 1970).
Isolation of the mxl-1 gene
P-labeled MDL-1 protein was prepared and used to
screen a C. elegans mixed stage lgt11 cDNA expression
library induced with IPTG as described (Armand et al.,
1994). Five positive plaques were puri®ed after screening
through *200 000 colonies. Phage DNA was prepared and
the inserts were isolated by PCR with lgt11forward and
lgt11reverse primers (New England Biolabs). The ampli®ed
DNA was then sequenced by the dideoxy method as
described above. A second PCR ampli®cation product was
generated using the lgt11reverse primer and the CeMAX1
primer (GAGGATCCAAGATGTCTGACATGAGT). The
ampli®cation product was inserted into the pHis6HMK
vector for (His)6 fusion protein production. The PCR
ampli®cation conditions were 948C - 1', 528C - 1', 728C - 1'
for 35 cycles. The Genbank database accession number for
MXL-1 is U82967.
32
Multiple sequence alignment
The multiple sequence alignments of the MDL-1/MXL-1
proteins to their homologues were performed using the
Clustalw multiple sequence alignment computer program.
The MDL-1/MAD/MNT/ROX proteins were ®rst aligned
as a cluster using the GONNET protein matrix for
pairwise alignments with a gap opening penalty of 1.0
followed by using the IDENTITY protein matrix for
multiple alignments, again with a gap opening penalty of
1.0. The default value was used for the other parameters.
The MAX/MXL-1 proteins were also aligned using the
same method. The two alignments were then aligned to
C elegans MAD and MAX proteins
J Yuan et al
each other again using Clustalw using the same parameters.
Empirically, this method provided the best overall
alignment. The guide ®le from the Clustalw program was
used to generate the phylogram using the Treeview
program.
Electrophoretic mobility shift assays
Electrophoretic mobility shift assays were performed using
1 ml of puri®ed protein in a 20 ml reaction volume. The
reaction consists of 4 ml of 56 binding bu€er
(56=100 mM Tris HCl, pH 7.5, 5 mM DTT, 5 mM
EDTA, 25% Glycerol, 250 mM NaCl), 10 ng ± 150 ng of
puri®ed protein, 1 ml of double stranded 32P-labeled MYC/
MAX consensus binding site DNA (GATCCTGACGACCACGTGGTCTTACGCTAG) (0.2 ng ± 1 ng of DNA),
1 ml of 1 mg/ml Salmon Sperm DNA, and 12 ml of dH2O.
The binding reaction was ®rst incubated at 378C for 10'
and then at room temperature for 20'. The reaction was
loaded onto a 0.256TBE 4% acrylamide gel and
electrophoresed in 0.256TBE running bu€er. The gel
was then dried and autoradiographed.
Ligand blotting assays
Ligand blotting assays were performed according to
methods as described (Worman et al., 1988) with the
following modi®cations. The MDL-1 and MXL-1 proteins
were labeled with g-32P-ATP as previously described
(Armand et al., 1994). The proteins were not boiled
before loading on to a 12% SDS ± PAGE. The proteins
were transferred on to a 0.45 mm Nytran Plus membrane
(Schleicher and Schuell) and blocked with blocking bu€er
that contained 5% nonfat milk and allowed to renature
overnight at room temperature.
Preparation of mRNA from staged nematodes
Synchronized embryonic, L1, L2, L3, and L4 stage C
elegans were prepared on 15 cm NGM plates as described
(Wood, 1988). A nonsynchronized population was also
prepared. Total RNA was isolated using Trizol reagent
(Gibco ± BRL). Brie¯y, worms were washed o€ plates with
DEPC treated dH2O into sterile 15 ml Falcon Tubes. The
pellet of worms were washed twice with dH2O. To every
100 ml of packed worms, 1 ml of Trizol reagent was added
and vigorously vortexed. The mixture was incubated at
room temperature for 7.5' and spun at 14 k r.p.m. for 10'
at 48C. 200 ml of CHCl3 was added to each ml of
supernatant and vortexed for 15'' and then incubated at
room temperature for 3'. The mixture was recentrifuged at
14 k r.p.m. for 15' at 48C and the aqueous upper layer was
removed and 500 ml of isopropanol was added per ml of
supernatant to precipitate the RNA. RNA pellet was
resuspended in 20 ml of DEPC treated dH2O and stored at
7708C until use. mRNA was isolated using the Oligotex
beads according to manufacturer's protocol (Qiagen) and
quantitated by O.D. 260 nm readings.
RT ± PCR analysis
cDNA was generated from each stage using *16 ng of
mRNA and an oligo-dT primer and reverse transcriptase
(Gibco ± BRL) in a 25 ml reaction. A no reverse transcriptase control was performed for each stage. A no RNA
control was also included. PCR ampli®cation was
performed using 1 ml of the RT reaction and the SL1
primer (GGTTTAATTACCCAAGTTTGAG) and the
following downstream primers: cml-1-2 for mdl-1, cml-3-2
(CATTATTGAG-CATGCATAGACGG) for mxl-1, and
act-3-1 ( ) for act-3. The data presented is only
semiquantitative and the actin controls were performed at
the same time as the other samples. PCR ampli®ction was
performed using the following conditions: 2' 948C hot start
followed by 40 cycles of 948C - 30'', 528C - 1', 728C -1'. The
reaction products were visualized by Ethidium Bromide
staining on a 4% acrylamide gel.
Transgenic strains
The transgenic mxl-1::GFP and mdl-1::GFP strains were
generated by microinjection as described (Fire, 1986).
Reporter gene plasmids were injected at 25 mg/ml for
mdl-1::GFP and 50 mg/ml for mxl-1::GFP. The 50 ± 100 mg/
ml of the plasmid pDP#MM016B which rescues the unc119(e2498) mutation was coinjected with the reporter
plasmids into unc-119(e2498) animals (Maduro and
Pilgrim, 1995).
The reporter gene plasmids were constructed in the
following manner. For the mdl-1::GFP construct, the
promoter was ampli®ed from 100 ng of genomic DNA by
long distance PCR using KlenTaq (Clontech) and the
oligonucleotide
primers, cml-1-3 (GTAGTAGCAATGCTTCCATGCTTG) and cml-1-4 (GAGTTGCTGTTCCATTGCAAAC) which give a *5.5 kb product using the
following conditions: 5 cycles of 948C - 30'', 608C - 30'', 688C - 6'
followed by 30 cycles of 948C - 30'', 528C - 30'', 688C - 6'. The
reaction was then re-ampli®ed with the cml-1-3 and cml-1-7
(GGAGAGCTCCGCAAACGGGGGTAACTTATTGAA)
primers using the same conditions. The cml-1-3/cml-1-7
product was inserted into the GFP reporter construct,
pPD95.67 which has the unc-54 3' untranslated region
(courtesy of A Fire). The mxl-1::GFP construct was
constructed in a similar manner to mdl-1::GFP, except that
the promoter was ampli®ed ®rst with the primers cml-3-3
(GTTACCC-CAATATTGCAT) and cml-3-4 (GGCGACGGTCTCTACACGGTCCAC) using the same conditions as
described above except for a 9 min 688C polymerization step.
The ampli®ed product was then used in a second PCR reaction
(the annealing step was omitted) utilizing the cml-3-4 and cml3-6 (CATTCTAGACATCTGAGGTTTATTTTCCATTAGGGAAGTGAACA) primers. This produces a 6.5 kb DNA
fragment when inserted into the GFP reporter construct
pPD95.69 (courtesy of A Fire).
Construction of DMDL-1
A DELTA MDL-1 oligonucleotide primer (ATGGAACAGCAAATTGGCGCTCTTGACATTTCTTCCC) that
deleted the predicted SIN3 interaction domain and began
at the translation initiation start site was synthesized. A
full length DMDL-1 gene was ampli®ed by PCR using the
DELTA MDL-1 primer and cml-1-7 primer (GGAGAGCTCCGCAAACGGGGGTAACTTATTGAA). The
ampli®ed fragment was subcloned into pGEM-T (Promega) and sequenced to verify that there were no PCR
induced mutations.
Transfections and rat embryo ®broblast foci assays
A full length clone of MXL-1 was inserted into the
pLbSGH vector, which contains a Moloney sarcoma virus
LTR promoter, a rabbit b-globin second intron, and a
bovine growth hormone 3'UTR and poly adenylation
sequence. A full length clone of MDL-1 and DMDL-1
were also inserted into pLbSGH. Rat embryo ®broblasts
(REFs) were maintainted in DMEM supplemented with
10% fetal calf serum (Irvine Scienti®c) and 1 mg/ml
pyruvic acid. REFs were seeded at 26105 cells in a
10 cm dish and transfected 24 h later with activated RAS
and c-MYC expression vectors (Brough et al., 1995). The
presence of transformed foci were scored 16 days after
transfection.
1117
C elegans MAD and MAX proteins
J Yuan et al
1118
Acknowledgements
We would like to thank Dr Joseph Goodhouse for help
with the confocal imaging, Drs Alan Coulson and Ratna
Shownkeen for the genomic ®ngerprinting, and members of
the Cole lab for helpful discussions. We would also like to
thank Dr A Fire for the lambda library and GFP plasmids,
Dr M Maduso for the unc-119 strains and plasmids, and
Dr J Schwarzbauer for critical reading of the manuscript
and for use of the microinjection apparatus. This work was
supported by NIH grant CA55248 to MDC.
References
Alland L, Muhle R, Hou Jr H, Potes J, Chin L, SchreiberAgus H and DePinho RA. (1997). Nature, 387, 49 ± 55.
Altschul SF, Gish W, Miller W, Myers EW and Lipman P.
(1991). J. Mol. Biol., 215, 403 ± 410.
Amati B, Brooks MW, Levy N, Littlewood TD, Evans GI
and Land H. (1993). Cell, 72, 233 ± 245.
Ambros V and Horvitz HR. (1984). Science, 226, 409 ± 416.
Armand P, Knapp AC, Hirsch AJ, Wieschaus EJ and Cole
MD. (1994). Mol. Cell. Biol., 13, 383 ± 390.
Ayer DE and Eisenman RN. (1993). Genes Dev., 7, 2110 ±
2119.
Ayer DE, Kretzner L and Eisenman RN. (1993). Cell, 72,
211 ± 222.
Ayer DE, Lawrence QA and Eisenman RN. (1995). Cell, 80,
767 ± 776.
Berberich SJ and Cole MD. (1992). Genes Dev., 6, 166 ± 176.
Berberich S, Hyde-DeRuyscher N, Espenshade P and Cole
M. (1992). Oncogene, 7, 775 ± 779.
Blackwood EM and Eisenman RN. (1991). Science, 251,
1211 ± 1217.
Brough D, Hofmann T, Ellwood K, Townley R and Cole M.
(1995). Mol. Cell Biol., 15, 1536 ± 1544.
Cerni C, Bousset K, Seelos C, Burkhardt H, Henriksson M
and LuÈscher B. (1995). Oncogene, 11, 587 ± 596.
Chal®e M, Tu Y, Euskirchen G, Ward WW and Prasher DC.
(1994). Science, 263, 802 ± 805.
Eagle LR, Yin X, Brothman AR, Williams BJ, Atkin NB and
Prochownik EV. (1995). Nat. Genet., 9, 249 ± 255.
Evan G and Littlewood TD. (1993). Curr. Opin. Genet. Dev.,
3, 44 ± 49.
Fire A. (1986). EMBO J., 5, 2673 ± 2680.
Gallant P, Shiio Y, Cheng PF, Parkhurst SM and Eisenman
RN. (1996). Science, 274, 1523 ± 1527.
Hassig CA, Fleischer TC, Billin AN, Schreiber SL and Ayer
DE. (1997) Cell, 89, 341 ± 347.
Heinzel T, Lavinsky RM, Mullen TM, Soderstrom M,
Laherty CD, Torchia J, Yang WM, Brard G, Ngo SD,
Davie JR, Seto E, Eisenman RN, Rose DW, Glass CK and
Rosenfeld MG. (1997). Nature, 387, 43 ± 48.
Henriksson M and LuÈscher B. (1996). Adv. Cancer Res., 68,
109 ± 182.
Hurlin PJ, QueÂva C, Koskinen PJ, SteingrõÂ msson E, Ayer
DE, Copeland NG, Jenkins NA and Eisenman RN.
(1995). EMBO J., 14, 5646 ± 5659.
Hurlin PJ, Queva C and Eisenman RN. (1997). Genes Dev.,
11, 44 ± 58.
Kadosh D and Struhl K. (1997). Cell, 89, 365 ± 371.
Kipreos ET, Lander LE, Wing JP, He WW and Hedgecock
EM. (1996). Cell, 85, 829 ± 839.
Koskinen PJ, Ayer DE and Eisenman RN. (1995). Cell
Growth Di€er., 6, 623 ± 629.
Koskinen PJ, VaÈstrik I, MaÈkelaÈ TP, Eisenman RN and
Alitalo K. (1994). Cell Growth Di€er., 5, 313 ± 320.
Krause M and Hirsch D. (1987). Cell, 49, 753 ± 761.
Laemmli U. (1970). Nature, 227, 680 ± 685.
Laherty C, Yang W-M, Sun J-M, Davie J, Seto E and
Eisenmann R. (1997). Cell, 89, 349 ± 356.
Lahoz EG, Xu L, Schreiber-Agus N and DePinho RA.
(1994). Proc. Natl. Acad. Sci. USA, 91, 5503 ± 5507.
Maduro M and Pilgrim D. (1995). Genetics, 141, 977 ± 988.
Meroni G, Reymond A, Alcalay M, Borsani G, Tanigami A,
Tonlorenzi R, Nigro CL, Messali S, Zollo M, Ledbetter
DH, Brent R, Ballabio A and Carrozzo R. (1997). EMBO
J., 16, 2892 ± 2906.
Nagy L, Kao HY, Chakravarti D, Lin RJ, Hassig CA, Ayer
DE, Schreiber SL and Evans RM. (1997). Cell, 89, 373 ±
380.
Ruvkun G and Giusto J. (1989). Nature, 338, 313 ± 319.
Schreiber-Agus N, Chin L, Chen K, Torres R, Thomson CT,
Sacchettini JC and DePinho RA. (1994). Oncogene, 9,
3167 ± 3177.
Schreiber-Agus N, Chin L, Chen K, Torres R, Rao G, Guida
P, Skoultchi AI and DePinho RA. (1995). Cell, 80, 777 ±
786.
Schreiber-Agus N, Stein D, Chen K, Goltz JS, Stevens L and
DePinho RA. (1997). Proc. Natl. Acad. Sci. USA, 94,
1235 ± 1240.
Sulston JE and Horvitz HR. (1997). Dev. Biol., 56, 110 ± 156.
Sulston JE, Shierenberg E, White JG and Thomson JN.
(1983). Dev. Biol., 100, 64 ± 119.
Wood W. (1988). The Nematode Caenorhabditis elegans Cold
Spring Harbor Laboratory, Cold Spring Harbor, NY.
Worman HJ, Yuan J, Blobel G and Georgatos SD. (1988).
Proc. Natl. Acad. Sci. USA, 85, 8531 ± 8534.
Zervos A, Gyuris J and Brent R. (1993). Cell, 72, 223 ± 232.
Zhang Y, Iratni R, Erdjument-Bromage H, Tempst P and
Reinberg D. (1997). Cell, 89, 357 ± 364.