(Received for publication, April 4, 1997, and in revised form, June 10, 1997)
From the Hubrecht Laboratory, Netherlands Institute
for Developmental Biology, Utrecht, The Netherlands and the
§ Institute of Biochemistry and Molecular Biology,
Albert-Ludwigs-Universität, Freiburg, Germany
The RelA subunit of NF-B and the
glucocorticoid receptor mutually repress each others transcriptional
activity, thus providing a mechanism for immunosuppression. Deletion
analysis of the glucocorticoid receptor has shown that the DNA binding
domain and the ligand binding domain are essential components for
repression. Here, we show by deletions and point mutations that both
the Rel homology domain and the transactivation domains of RelA are
required for repression of the transcriptional activity of the
glucocorticoid receptor in intact cells. However, only the Rel homology
domain of RelA was found to associate with the glucocorticoid receptor in vitro. RelA mutants, not able to repress glucocorticoid
receptor activity, but still able to dimerize, behaved as transdominant inhibitors of the repressive activity of wild type RelA. Furthermore, we show that the 13 S E1A protein is able to interfere with the transrepressive activity of RelA. We propose that negative
cross-talk between the glucocorticoid receptor and RelA is due to
direct interaction via the Rel homology domain of RelA and the DNA
binding domain of the glucocorticoid receptor in combination with
interference by the transactivation domains of RelA with the
transcriptional activity of the glucocorticoid receptor.
The NF-B/Rel family of transcription factors regulates the
expression of a variety of genes involved in immune and inflammatory responses. Presently, five members of the NF-
B/Rel family have been
identified in mammals including NF-
B1, NF-
B2, RelA, c-Rel, and
RelB. These proteins share homology in their 300-amino acid N-terminal
regions. This region of sequence similarity, which is designated the
Rel homology domain (RHD)1,
functions in DNA binding, dimerization, and interaction with I
B (1,
2). NF-
B was originally identified as a heterodimer of NF-
B1 and
RelA (3, 4), but a variety of other homo- and heterodimers have been
described. NF-
B is present in an inactive form in the cytoplasm,
associated to an inhibitor protein, I
B. Upon exposure of the cells
to inflammatory cytokines, like tumor necrosis factor-
and
interleukin-1, or lipopolysaccharide, UV radiation, or viral infection,
I
B becomes phosphorylated, ubiquitinated, and subsequently degraded
(2). As a result, NF-
B is translocated to the nucleus, where it
binds to specific DNA sequences and activates transcription.
Transactivation functions have been located in the C-terminal regions of RelA (5), c-Rel (6, 7), and in both the C-terminal and N-terminal region of RelB (8, 9). RelA contains at least two strong transactivation domains (TADs) within its C terminus; activation domain TA1, consisting of the 30 C-terminal amino acids, and TA2, located within the 90 amino acids next to TA1 (5, 10). At the N-terminal part of TA2, a mini-leucine zipper motif is present composed of three leucines arranged in a heptad repeat (5). Both TADs contain a common sequence motif (11).
Cross-talk between transcription factors of distinct families is an
important phenomenon in regulating gene transcription and has recently
become the subject of intensive investigation. NF-B, and
particularly RelA, has been shown to interact functionally and
physically with numerous other transcription factors, including members
of the AP-1 family, resulting in enhanced biological activity of these
transcription factors (12). Previously, we and others have reported
that steroid receptors, including the glucocorticoid receptor (GR)
(13-15), the estrogen receptor (16), the progesterone receptor (17),
and the androgen receptor (18), are able to inhibit NF-
B activity
and can physically interact with NF-
B proteins in vitro.
Since RelA represses ligand-dependent activation of steroid
receptor-regulated promoters, a mutually inactive complex formed either
by a direct protein-protein interaction of the receptor and RelA or via
a third partner has been proposed (13-18).
Steroid receptors belong to the superfamily of steroid/thyroid hormone receptors, and their modular structure, consisting of a DNA binding domain (DBD) and a ligand binding domain (LBD), is highly conserved. Transactivation domains have been located N-terminally to the DBD, designated AF-1, and within the LBD, named AF-2. Whereas AF-1 is a hormone-independent activation domain, AF-2 functions hormone dependently (19). Recently, several cofactors interacting with AF-2 have been described to inhibit (corepressors) or enhance (coactivators) transcription by nuclear receptors (20).
The domains involved in interaction between steroid receptors and
NF-B have not been mapped in much detail. For steroid receptors, analysis of deletion mutants has revealed that both the DBD and the LBD
are necessary for repression of NF-
B activity (13, 16, 17). So far,
the domain(s) of NF-
B involved in inhibition of steroid receptor
activity have not been determined. Therefore, we investigated the
importance of different regions in RelA in repressing GR activity. Our
results show that both the RHD and the TADs of RelA are required for
repression of hormone-dependent activation of GR, while
only the RHD and not the TADs were found to directly interact with GR
in vitro. Furthermore, cotransfection of the cofactor 13 S
E1A resulted in a decrease in repressive activity of RelA in COS-1
cells. These data suggest that the mutual repression between GR and
RelA is due to complex formation via the RHD of RelA, while the TADs of
RelA, able to interact with cofactors, are required to repress the
transcriptional activity of GR.
Dexamethasone was obtained
from Sigma. Polyclonal antibodies against the N-terminal domain of RelA
(SC-109) and against the C terminus of RelA (SC-372) were purchased
from Santa Cruz Biotechnology (Santa Cruz, CA). Polyclonal antibody
against IB
was from Upstate Biotechnology, Inc. (Lake Placid,
NY).
Monkey COS-1 cells
and human 293 embryonal kidney cells were obtained from American Type
Culture Collection (Rockville, MD). Cells were cultured in a 1:1
mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium
(Life Technologies, Inc.), buffered with bicarbonate and supplemented
with 7.5% fetal calf serum from Integro (Linz, Austria). For transient
transfections, the cells were cultured in six-well tissue culture
plates. Cells were transfected using calcium-phosphate coprecipitation
with 2 µg of luciferase reporter, 3 µg of PDMlacZ, and the
indicated amount of expression plasmids. pBluescript SK
plasmid was added to obtain a total amount of 10 µg of DNA/well. After 16 h, the medium was refreshed and hormone was added. Cells were harvested 24 h later and assayed for luciferase activity using the luciferase reporter gene assay kit (Packard). Values were
corrected for transfection efficiency by measuring
-galactosidase activity (21).
Details about the construction of the clones
presented in this report can be obtained from the authors upon request.
The luciferase reporter plasmid containing three NF-B sites from the
ICAM-1 promoter and the reporter plasmid 2 × GREtkluc have been
described elsewhere (22, 23). The CMV4 expression vectors containing human RelA and GR, the 12 S E1A, 13 S E1A, and RXR
expression plasmids, and glutathione S-transferase (GST)-NF-
B1 and
-RAR
have been described previously (13, 17, 24). CBP and p300 were
kind gifts from Drs. R. H. Goodman (Portland, OR) and R. Eckner
(Boston, MA).
Proteins were synthesized in vitro using the TnT-coupled rabbit reticulocyte lysate system (Promega) in the presence of [35S]methionine according to the manufacturer's description. GST fusion proteins were expressed in Escherichia coli BL21(plysS). Expression and purification with glutathione-coated beads (Pharmacia) was performed as described previously (17). The fusion proteins loaded on Sepharose beads were subsequently incubated with in vitro synthesized proteins in 20 mM Tris, pH 7.5, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 1 µg/ml leupeptin (NETN) for 1 h at room temperature. Beads were washed four times with NETN, resuspended in sample buffer, and analyzed by SDS-polyacrylamide gel electrophoresis.
Electrophoretic Mobility Shift Assay293 cells were grown
in 10-cm dishes and transfected as described above with 20 µg of
expression plasmid and 20 µg of both plasmids when combinations are
used. Cells were harvested in Dignam C (20 mM Hepes, pH
7.9, 25% glycerol, 420 mM NaCl, 1.5 mM
MgCl2, 0.2 mM EDTA, 0.5 mM
dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1 µg/ml leupeptin, 1 µg/ml aprotinin), incubated for 30 min at
4 °C, and membranes were pelleted. The protein concentration of the
supernatant was determined by the Bio-Rad protein assay according to
the manufacturer's protocol. Double-stranded oligonucleotides containing the B site from the ICAM-1 promoter
(5
-agcttctTGGAAATTCCggagc-3
) were labeled with
[32P]dCTP using the Klenow fragment of DNA polymerase I. Whole cell extracts (5 µg) were incubated with 10,000 cpm of probe
(0.1-0.5 ng) and 1 µg of poly(dI-dC) for 30 min at room temperature
in a total reaction mixture of 20 µl containing 20 mM
Hepes, pH 7.9, 100 mM KCl, 0.2 mM EDTA, 20%
glycerol, 1 mM dithiothreitol, 1 µg/µl bovine serum
albumin. Samples were loaded on a 5% polyacrylamide (29:1) gel,
containing 0.25 × TBE (90 mM Tris borate, 2 mM EDTA) as running buffer.
COS-1 cells and 293 cells were transfected as described above. Subsequently cells were harvested directly in sample buffer. Samples were separated on SDS-polyacrylamide gels and transferred to Immobilon (Millipore). Blots were blocked with Blotto (phosphate-buffered saline containing 4% non-fat milk powder and 0.05% Tween 20) for 30 min. All subsequent steps were carried out in Blotto:phosphate-buffered saline (1:1). Blots were probed with the polyclonal antibody SC-109 or SC-372 against RelA. After washing, blots were incubated with peroxidase-conjugated antibodies (1:10,000, Amersham Corp.). Blots were washed again, and immunoreactive bands were visualized with ECL according to the manufacturer's instructions (Amersham Corp.).
To determine the domain(s) in RelA involved in
repression of GR activity, several deletion constructs, lacking (part
of) the TADs or (part of) the RHD were used. Furthermore, a mutant was used containing a point mutation in the RHD (Fig.
1A). Expression levels of
these RelA constructs in 293 cells were similar, as detected on a
Western blot (Fig. 1B). To detect the RelA proteins in
lanes 1-12, a polyclonal antibody against the N-terminal
region of RelA was used. A different antibody, directed against the
C-terminal region of RelA, was used to detect RelA
1-55122-248 protein (lane 13). This
antibody shows an aspecific signal as indicated in Fig. 1B.
Similar levels of expression were observed in COS-1 cells (data not
shown).
To study the ability of the different proteins to activate
transcription from an NF-B reporter, the constructs encoding the different RelA mutants were cotransfected with this reporter into 293 cells. As shown in Fig. 1C, full-length RelA (1-551)
strongly activated the reporter (~80-fold), while deletions in
transactivation domains 1 and 2 (TA1 and TA2)
drastically decreased the transactivation potential. As expected, no
activity was observed when both activation domains were deleted (RelA
1-431 and RelA 1-305). Also the constructs containing either
deletions or mutations in the RHD, which are therefore no longer able
to bind DNA (see Fig. 4B and data not shown), could not
activate the NF-
B reporter (RelA 1-551
22-248,
RelA 1-551
79-95, and RelAE39I). Similar
results were obtained in COS-1 cells (data not shown). For all
constructs that were able to transactivate the NF-
B reporter (Fig.
1C), repression of their transcriptional activity could be
observed upon cotransfection of expression vector encoding GR (1 µg)
in the presence of dexamethasone (results not shown). This indicates
that not a specific part of the transactivation domain of RelA is
involved in the repression of RelA activity by GR.
To examine whether the RelA mutants were able to repress GR activity,
293 cells and COS-1 cells were transfected with a reporter construct
containing two glucocorticoid response elements (GREs) in front of the
thymidine kinase promoter coupled to the luciferase gene.
Cotransfection with GR expression vector (100 ng) resulted in a
hormone-dependent induction of luciferase activity, which could be clearly repressed by the presence of full-length RelA (1-551)
(Fig. 2), as described previously (13).
Deletion of either TA1 (1-521), or the
TA1-like domain in TA2
(1-551443-476) (11), hardly affected the repressive
activity of RelA, while deletion of both the leucine zipper-like
structure and part of the TA1-like domain in
TA2 (RelA 1-551
431-470) did have a small
effect. In 293 cells, further deletion of the TADs and combinations of
deletions in TA1 and TA2 resulted in a decrease
in repressive activity and RelA 1-431, lacking both TADs was no longer
able to repress GR activity (Fig. 2A). However, in COS-1
cells, these deletions in the TADs had only minor effects, and RelA
1-431 was still able to repress GR activity to around 35% (Fig.
2B). RelA 1-305, containing only the RHD, no longer showed
repressive activity in both cell lines. These results indicate that, in
293 cells and in COS-1 cells, the TADs of RelA are necessary for
repression of GR activity. In addition, there is no strict correlation
between the transactivation function and the transrepression function
of the RelA mutants.
When (part of) the RHD of RelA was deleted (RelA
1-55122-248, RelA 1-551
79-95), the
resulting mutants were no longer able to bind DNA. Cotransfection of
these constructs did not result in repression of GR activity in 293 cells and hardly showed repressive activity in COS-1 cells, whereas
point mutant E39I, also defective in DNA binding (25) was able to
repress GR activity in COS-1 cells but not in 293 cells. The DNA
binding-defective mutants that still contained intact TADs, able to
interact with transcription intermediary factors, were not able to
repress GR activity. This indicates that negative cross-talk between GR
and RelA is not only the result of competition for common coactivators,
a process named squelching (26, 27). It is clear that, although there
are cell-type specific differences in repressive activity of RelA, both
the RHD and the TADs of RelA are required for repression of GR
activity.
To
investigate the possible function of the RHD and the TADs of RelA in
the interaction with GR, reciprocal binding assays were performed.
First, the cDNAs of RelA 1-551, RelA 1-431, lacking the TADs,
RelA 1-305, containing only the RHD, and RelA
1-55122-248, lacking the RHD, were fused in-frame to
the GST gene. GST-NF-
B1 and GST-RAR
were used as controls. GST
fusion proteins were expressed in bacteria, purified with
glutathione-coated agarose beads, and subsequently incubated with
equivalent amounts of [35S]methionine-labeled GR or
RXR
protein, synthesized by in vitro transcription-translation. As shown in Fig.
3A, GR could not be precipitated by GST alone and hardly precipitated by GST-NF-
B1 and
GST-RAR
, whereas GST-RelA 1-551, GST-RelA 1-431, and GST-RelA 1-305 efficiently bound GR protein. However, GST-RelA
1-551
22-248, lacking the RHD, was not able to
precipitate GR protein. Furthermore, the GST-RelA proteins were
incubated with [35S]methionine-labeled RXR
, and it was
found that only GST-RAR
was able to precipitate RXR
protein,
clearly demonstrating the specificity of the interaction between the
RelA proteins and GR.
Next, in an alternative approach to determine the domain in RelA
involved in interaction with GR, the cDNA encoding GRAB (amino
acids 420-779) was fused to the GST gene and tested for its ability to
bind in vitro synthesized,
[35S]methionine-labeled mutant RelA proteins. As shown in
Fig. 3B, the labeled RelA proteins could not be precipitated
by GST alone, whereas GST-NF
B1, known to associate with RelA (4),
and GST-GR
AB clearly precipitated RelA 1-551, RelA 1-431, and RelA
1-305. RelA 286-551 could not be precipitated by both GST-NF-
B1
and GST-GR
AB, confirming our previous results shown in Fig.
3A. In all cases, the additional presence of hormone had no
effect (data not shown). These data indicate that, although negative
cross-talk between GR and RelA in intact cells requires both the RHD
and the TADs, physical association between GR and RelA involves the RHD
only.
RelA
mutants in which the domain involved in repression is mutated, but
which are still able to dimerize, can associate with wild type RelA
protein. The potential effect of these complexes on the repressive
activity of RelA was tested in transfection experiments. Several RelA
mutants were analyzed by transfection of 293 cells with a GRE reporter
construct and expression constructs encoding GR (100 ng) and RelA
1-551 (100 ng). RelA can readily repress GR activity to ~55% when
cotransfected in amounts as low as 100 ng (Fig.
4A). Cotransfection of wild
type RelA (1 µg) resulted in an increased repression of GR activity,
whereas cotransfection of RelA 1-431, RelA 1-305, or
RelAE39I (1 µg), showed a decrease in the repressive
activity of RelA, and therefore these constructs could be considered to
act as transdominant negative mutants. A deletion construct that was no
longer able to dimerize with wild type RelA did not influence this
repressive activity (RelA 1-55122-248).
To verify that the transdominant mutants indeed dimerized with RelA,
electrophoretic mobility shift assays were performed. Therefore RelA
and the transdominant mutants were overexpressed in 293 cells, and
their ability to form heterodimers was examined. As shown in Fig.
4B, heterodimer formation could be observed between RelA
1-551 and the mutants, resulting in either the formation of a complex
with intermediate mobility in the case of RelA 1-431 (lane
4) and RelA 1-305 (lane 6), or resulting in a decrease
in DNA binding of wild type RelA when RelAE39I was
co-transfected (lane 8). Coexpression of RelA
1-55122-248 had no effect on the DNA binding of wild
type RelA protein (lane 10). These results indicate that
RelA deletion mutants, which were not able to repress GR activity (but
still able to dimerize), can act as transdominant negative inhibitors
of repressive activity of the wild type RelA protein.
As
we showed (Fig. 2), there appeared to be cell type-specific differences
in repressive activity of RelA in COS-1 cells and 293 cells, which were
most pronounced in the case of RelA 1-521443-476, RelA
1-431, and RelA 1-551E39I. However, the transactivation
potential of these RelA mutants was identical in both cell types (Fig.
1C). The difference in repressive activity between both cell
types could be due to the presence of cell type-specific cofactors.
Nuclear receptors have been described to inhibit or enhance
transcription by recruiting specific coactivator or corepressor
proteins to the transcription complex (20). These cofactors might also
interfere with the cross-talk between GR and RelA. An important
difference between COS-1 cells and 293 cells is that 293 cells contain
E1A protein (28). Furthermore, it has previously been described that 13 S E1A was able to associate with the C-terminal part of RelA and to
stimulate the transcriptional activity of RelA (29). Therefore, we
investigated the possible role of E1A in the transrepression potential
of RelA. To study this, COS-1 cells and 293 cells were transiently
transfected with a GRE reporter construct (2 µg), an expression
construct for GR (100 ng) and expression vectors encoding RelA (500 ng)
and 13 S E1A (1 µg) as indicated. Fig. 5 clearly demonstrates that
cotransfection of 13 S E1A in COS-1 cells results in a decrease in
repressive activity of RelA and RelA 1-521
443-476.
This resembles the lower transrepression potential of RelA in 293 cells, in which cotransfection of 13 S E1A had no effect. Similar
results were obtained when RelA 1-431 and RelA 1-551E39I
were used (not shown). No effect was observed on the transcriptional activity of GR itself or on the repressive activity of RelA 1-305 (results not shown). Of the two known splice variants of E1A, only the
13 S form, and not 12 S E1A, was able to influence the transrepressive
activity of RelA (data not shown). These results suggest that
cofactors, such as 13 S E1A might be able to modify the negative
cross-talk between GR and RelA.
In this report, we demonstrate that both the RHD and the TADs of RelA are required for repression of GR activity in intact cells. Furthermore, a physical interaction between GR and RelA requires the presence of the RHD of RelA, which has already been shown to be sufficient for binding to estrogen receptor (16) and to AP-1 (12). Since the DBD of GR is essential for repression of RelA activity (13, 15), this domain in GR may directly interact with the RHD of RelA. Both the DBD of GR and the RHD of RelA are essential for the mutual repression but not sufficient. In addition, the LBD of GR (13) as well as the TADs of RelA are required for the functional repression. Both AF-2 in the LBD of steroid receptors and the TADs of RelA are known to interact with cofactors, which inhibit or enhance their transcriptional activity (20, 29, 30). Together with the fact that 13 S E1A is able to decrease the repressive activity of RelA, these findings suggest that cofactors might be involved in cross-talk between GR and RelA.
Our data show a difference in the functional interaction between GR and
RelA, requiring both the RHD and the TADs, and the physical
interaction, requiring the RHD only. One possible explanation for this
difference is that the RHD may bind to GR first and in this way
facilitates subsequent binding of the TADs to GR. However, this
explanation seems unlikely because our results suggest that not a
specific part of the C terminus of RelA determines its ability to
repress, but rather the length of the C terminus or the presence of a
functional TAD. Therefore, an alternative model could be that the RHD
accounts for the interaction with GR, whereas the C-terminal part of
RelA interferes with transcriptional activation by GR. This could be
attained by either masking the domain(s) in GR necessary for
interaction with the basal transcription machinery and/or coactivators,
or through binding of the TADs of RelA to these cofactors themselves.
This latter model is in agreement with the findings of several groups
that show that NF-B1, lacking the C-terminal extension present in
RelA, can interact in vitro with GR, estrogen receptor, and
androgen receptor, but is, in contrast to RelA, not able to repress the
transcriptional activity of these receptors (15, 16, 18).
We found that mutations in RelA leading to a loss of repressive activity, but still allowing protein dimerization, resulted in transdominant inhibition of the repressive activity of wild type RelA. These mutants dimerized with RelA in mobility shift assays, suggesting that interaction between the RelA mutants and wild type RelA results in a heterodimer, unable to repress GR activity. It remains unclear whether dimerization of RelA is required for the transrepression. The dominant negative effect of the RelA mutants could also be due to competition with wild type RelA for the interaction domain in GR, since both RelA 1-431 and RelA 1-305 were found to bind to GR in vitro.
By using both 293 cells and COS-1 cells to study the repressive
activity of RelA, it became clear that there were cell type-specific differences. Although the same pattern of repressive activity of the
RelA mutants was observed in both cell types, generally the RelA
constructs were more active in repression in COS-1 cells. Furthermore,
some deletion constructs, which did not display any repressive activity
in 293 cells, were still able to repress GR activity in COS-1 cells.
This could be due to a differential regulation of IB
in these
cells. Besides complex formation between GR and RelA, resulting in the
mutual repression, a second mechanism has been proposed in which
glucocorticoids induce I
B
synthesis and thereby inhibit NF-
B
activity (31, 32). Up-regulation of I
B
by dexamethasone in 293 cells could therefore result in an inhibition of RelA translocating to
the nucleus and in this way decrease the ability of RelA to repress GR
activity. However, Western blot analysis showed no up-regulation of
I
B
protein after transfection with an expression vector for GR
and treatment with dexamethasone in both COS-1 cells and 293 cells.2 Therefore,
differential regulation of I
B
seems not to account for the
differences observed between the two cell lines.
Recently it has been reported that inhibition of AP-1 activity by GR in HeLa cells is mediated by competition for limiting amounts of CREB-binding protein or p300, which serve as coactivators for both GR and AP-1 transcriptional activity (33). The fact that the TADs are required for the repression function of RelA could also indicate a role for cell type-specific coactivators in the negative cross-talk between GR and RelA. However, we have been unable to observe an effect of cotransfection of CREB-binding protein or p300 on RelA transcriptional activity or on the cross-talk between GR and RelA in COS-1 cells and 293 cells.2
In addition, RelA has been shown to interact with components of the
basal transcription machinery, such as TFIIB and TATA-binding protein,
and some cofactors have been described to enhance the transcriptional
activity of RelA, such as PC1 (30) and E1A (29). E1A is expressed in
293 cells (28) and not in COS-1 cells, possibly explaining some of the
differences in repressive activity of RelA in these cells. Our data
show that cotransfection of 13 S E1A and not 12 S E1A in COS-1 cells
results in a decrease in repressive activity of RelA, thus resembling
the repression potential of RelA in 293 cells. Possibly, 13 S E1A
interferes with the transrepressive activity of RelA via binding to
RelA and thereby preventing interaction of RelA with GR or coactivators
for GR. It has been described previously that 13 S E1A but not 12 S E1A
activates NF-B and interacts with RelA at the C terminus (29). In
conclusion, it seems likely that some of the many coregulatory
proteins, which function between transcription factors and the basal
transcription machinery and among transcription factors of distinct
families, may influence the effectiveness of the interaction between GR and NF-
B. However, the fact that this response occurs in many different cell types (13-15, 34), despite differences in the magnitude
of the negative cross-talk between GR and RelA, supports a mechanism of
repression involving direct protein-protein interaction.
We thank Dr. R. H. Goodman for the CREB-binding protein cDNA and Dr. R. Eckner for the p300 cDNA. We thank G. Folkers and Drs. C. L. Mummery and C. Kuil for useful discussions and suggestions and critical reading of the manuscript. We also thank J. Heinen and F. Vervoordeldonk for photographic reproductions. The work was carried out in the Graduate School for Developmental Biology, Utrecht.