(Received for publication, February 3, 1997, and in revised form, May 4, 1997)
From the Department of Molecular Biology and
Pharmacology, Washington University School of Medicine, St. Louis,
Missouri 63110 and the Division of Reproductive Biology, Department of
Gynecology/Obstetrics, Stanford University Medical Center,
Stanford, California 94305-5317
The common subunit of
glycoprotein hormones contains five disulfide bonds. Based on the
published crystal structure, the assignments are 7-31, 59-87, 10-60,
28-82, and 32-84; the last three comprise the cystine knot, a
structure also seen in a variety of growth factors. Previously, we
demonstrated that the efficiency of secretion and the ability to form
heterodimers by
subunits bearing single cysteine residue mutants in
the cystine knot were significantly reduced. These results suggested
that the cystine knot is critical for the intracellular integrity of
the subunit. To assess if the presence of the free thiol affected the
secretion kinetics, we constructed paired cysteine mutants of each
disulfide bond of the
subunit. The secretion rate for these
monomers was comparable with wild type except for the
-10-60
mutant, which was 40% lower. The recovery of the
7-31 and
59-87 mutants was greater than 95%, whereas for the cystine knot
mutants, it was 20-40%. Co-expression of the wild-type chorionic
gonadotropin
subunit with double cysteine mutants did not enhance
the recovery of
mutants in the media. Moreover, compared with
wild-type, the efficiency of heterodimer formation of the
10-60 or
32-84 mutants was less than 5%. Because subunit assembly is
required for biological activity, studies on the role of these
disulfide bonds in signal transduction were not possible. To bypass the assembly step, we exploited the single chain model, where the
and
subunits are genetically fused. The recovery of secreted tethered
gonadotropins bearing mutations in the cystine knot was increased
significantly. Although dimer-specific monoclonal antibodies discriminated the conformation of single chain
10-60 and
32-84 mutants from the native heterodimer, these mutants were nevertheless biologically active. Thus, individual bonds of cystine knot are important for secretion and heterodimer formation but not for in
vitro bioactivity. Moreover, the data suggest that the native heterodimer configuration is not a prerequisite for receptor binding or
signal transduction.
Human chorionic gonadotropin (CG),1
lutropin (LH), follitropin, and thyrotropin are members of the
glycoprotein hormone family that share a common subunit but differ
in their hormone-specific
subunits. The amino acid sequence of the
subunit is identical within a species (1), suggesting that the
tertiary structure of the
subunit requires a degree of flexibility
to adapt to a unique conformation relative to the
domain(s). The
human
subunit has 10 highly conserved cysteine residues, which form five disulfide bonds. Based on the crystal structure of the
subunit
in hCG, the proposed pairs are 10-60, 28-82, 32-84, 7-31, and
59-87 (Fig. 1) (2, 3). Bonds 28-32 and 32-84 comprise a ring structure penetrated by a bond bridging cysteine residues 10 and
60 resulting in a core that forms three hairpin loops. This structure
is a common feature to the large growth factor family including tumor
growth factor-
, activins, nerve growth factor, and platelet-derived
growth factor (4-8). Chemical reduction (1) or site-specific mutations
of individual cysteine residues in either the common
or CG
subunit, alters dimer assembly and secretion rate (9, 10). Mutants
lacking either the 7-31 or 59-87 bond were secreted (10), and
heterodimers containing the
7-31 mutation bound to the LH/hCG
receptor with an affinity comparable with the wild-type hCG, whereas
the 59-87 mutant interacted with a lower binding affinity. However,
mutants comprising the cystine knot (10-60, 28-82, and 32-84) were
not secreted in sufficient quantities to allow examination of the
biological activity. To address this issue, we used a single chain
gonadotropin model where the CG
subunit was genetically fused to the
subunit (11, 12) to promote more efficient secretion of proteins
bearing the
and
subunit domains in the same complex. Using this
approach, the rate-limiting step of subunit assembly could be
circumvented, and mutations that otherwise block dimer formation could
be evaluated. Here we show that single chains containing mutations in
the cystine knot (
10-60, 28-82, and 32-84) are secreted and
biologically active, implying that quaternary relationships between the
subunits in the heterodimer primarily influence the intracellular
behavior rather than receptor binding/signal transduction.
Enzymes for preparing vectors and for polymerase chain reaction were purchased from Promega (Madison, WI), Boehringer Mannheim, or Stratagene (La Jolla, CA). Oligonucleotides were prepared by the Washington University Sequencing Facility. Cell culture media and reagents were prepared by the Washington University Center for Basic Cancer Research (St. Louis, MO). Fetal bovine serum, dialyzed fetal bovine serum, and the neomycin analogue, G418, were purchased from Life Technologies, Inc. For metabolic labeling, [35S]cysteine-methionine (Promix) (Amersham Corp.) or [35S]cysteine (ICN, Irvine, CA) were used. Monoclonal antibodies A407 and B109 were gifts from Dr. Robert Canfield and Dr. Steve Birken (Columbia University, New York).
Construction ofPreviously, the
intracellular behavior of single cysteine mutants was assessed (10,
13). Double mutants were only constructed that deleted the 7-31 and
59-87 disulfide bridges. Here, we designed all of the analogs to
contain double substitutions for each disulfide bond that were based on
the assignments from the recently published crystal structure (2, 3).
Thus, both monomeric subunit and the single chain analogs
containing such modifications were constructed.
The cysteine mutants previously
constructed in vector pM2 were used (10). Exon III of the
subunit contains a unique XbaI restriction site within
the sequence encoding amino acid residues 34 and 35 (Fig.
2A). Because the vector also contains a single XbaI site, fragments bearing
10,
28, or
32
mutations were exchanged with XbaI-digested pM2
containing
60,
82, or
84 substitutions. This resulted in
double cysteine mutants
10-60,
28-82, and
32-84,
respectively. Construction of the pM2
7-31 and
pM2
59-87 were described previously (10).
Tethered Analogs
Single chain variants were constructed
with the carboxyl end of the subunit fused to the amino end of the
subunit using overlap polymerase chain reaction mutagenesis (11)
(Fig. 2B). All constructs generated by
polymerase chain reaction were sequenced to ensure that the final
products contained no misincorporated residues.
Transfection, selection of stable CHO clones, metabolic labeling, and immunoprecipitation with subunit-specific antiserum were previously described (10, 14, 15). Analysis by Western blotting was performed using conditioned media that were concentrated with Centriprep-10 columns (Amicon, Beverly, MA). Samples were loaded on 12.5% SDS-polyacrylamide gels and electroblotted to nitrocellulose (Hybond ECL, Amersham International, UK), and the proteins were detected by the Western Light kit (Tropix, Bedford, MA).
To determine the biological activity, the variants were quantitated
either by an hCG RIA (Diagnostic Products Corporation, Los Angeles, CA)
containing CG polyclonal antiserum or an hCG dimer-specific
enzyme-linked immunosorbent assay kit (Organon, Oss, The Netherlands).
Both assays gave comparable results. Conditioned media were incubated
with human fetal kidney 293 cells stably transfected with human LH/CG
receptor, and the binding affinities (10) and cAMP accumulation were
determined (16). Iodination of hCG (CR-127; 14,900 IU/mg) was performed
as described (17); the specific activity and maximum binding of
125I-labeled hCG, as determined by radioligand receptor
assay (18), were 53,000 cpm/ng and 40%, respectively. Nonspecific
binding was determined by adding a 1000-fold excess of unlabeled ligand (Pregnyl, Organon); specific binding routinely was 10-12% of total 125I-labeled hCG added. The cAMP levels were also
determined in 293 cells expressing human LH receptors by radioimmune
assay (16).
The proposed disulfide bonds in the subunit are at
positions 7-31, 10-60, 28-82, 32-84, and 59-87 (Fig. 1).
Previously, we demonstrated that
subunits containing single
mutations at Cys-10, -28, -60, -82, and -84 were not secreted and in
most cases were degraded intracellularly (10). These mutants also
failed to assemble with hCG
subunit, whereas mutants with
alterations at Cys-7, -31, -32, -59, or -87 were secreted and
assembly-competent. Because a free thiol group could alter
intracellular behavior (9, 10, 19, 20), double mutants in the
subunit were constructed using the assignments based on the
crystallographic analysis (2, 3). To assess secretion/stability of
these mutants, transfected cells were labeled with
[35S]cysteine and subjected to pulse-chase analysis. The
intracellular (lysate) and extracellular (media) forms were
immunoprecipitated and resolved on SDS gels (Fig. 3).
Except for the 10-60 variant, the secretion rate of the mutants and
wild-type subunit were comparable. As described previously, the
WT,
7-31, and 59-87 mutants were efficiently secreted because greater
than 95% of the subunits were recovered from the media (10). In
contrast, the recovery of the
10-60, 28-82, and
32-84 double
mutants was 21, 41, and 20%, respectively (Fig. 3; Table
I). The decreased recovery was not due to lower
synthesis but rather to enhanced degradation, since at zero time of
chase the intracellular accumulation of the mutants is comparable (Fig.
3). These data suggest the cystine knot structure is critical for
maximum
subunit secretion. Although the secreted non-combined
WT
is heterogeneous, the media form of
10-60 is more homogeneous,
suggesting that a post-translational modification is affected by this
mutation.
|
To evaluate the efficiency of dimer formation, the constructs were
co-transfected with the hCG
gene (21), and clones synthesizing
excess
subunit were selected to ensure that it would not limit
assembly. The cells were labeled, and equal aliquots of lysate/medium
were precipitated with subunit-specific polyclonal antisera (Fig.
4, Table I). Recovery of the 10-60 and 32-84 mutants in the secreted heterodimer was less than 5% although synthesis of all
the mutants was at significant levels. In the case of the 28-82 mutant
(50% recovery), the combination efficiency is relatively unaffected,
since the amount of
subunit precipitated by either
or
subunit antiserum was comparable. Pulse-chase experiments demonstrated
that the
7-31, 28-82, and 59-87 mutants combined efficiently with
the
subunit, and their recovery paralleled the uncombined
subunit synthesized in the absence of co-transfected
subunit (Table
I). The low amount of dimers containing the
10-60 or
32-84
mutant reflects a decrease in combination efficiency and/or an enhanced
intracellular degradation of the mutated
subunit. The data show
that the disulfide bonds 10-60 and 32-84 (and to a lesser extent
28-82) are critical for assembly.
Expression of Tethered Variants
Due to decreased dimer
accumulation in the media, studies on the biologic action of the
heterodimers containing the 10-60 or 32-84 mutation are difficult.
However, if the subunit is covalently linked to the
subunit,
the rate-limiting assembly step is bypassed, and sufficient material
could be obtained to examine signal transduction. The disulfide bond
mutants described above were linked to the wild-type
subunit to
form single chain analogs (see "Materials and Methods"), and the
intracellular behavior was examined by pulse-chase experiments. The
secretion kinetics of the single chain variants (Fig. 5,
Table II) paralleled the corresponding monomers. The
variants including the nonmutated tether (CG
) were secreted at a
comparable t1/2 (CG
t1/2 = 106 min), while the single chain
10-60 mutant is secreted much
slower (t1/2 = 244 min; see Table II). Thus, even in
the single chain construct, the 10-60 bond is critical for secretion.
However, compared with the corresponding heterodimers (compare Tables I
and II), recovery of the mutants 10-60 and 32-84 dramatically
increases when incorporated in a single chain.
|
Because little heterodimer is formed with mutants 10-60 and
32-84, we suspected that the conformation of the single chain bearing
these mutated
domains would be changed. To address this issue,
conditioned media were examined by Western blot and probed by two
dimer-specific monoclonal antibodies: A407 and B109 (22-26) (Fig.
6). An A407 epitope resides in amino acid residues 5 and 6, 11-13, and 81 of the
subunit (23). Under nonreducing
conditions, A407 recognizes 7-31, 28-82, 59-87, nonmutated CG
single chain, and CG heterodimer (Fig. 6A). However, 10-60
and 32-84 are barely detected (Fig. 6A). These low signals
are not due to a difference in the recovery of the blotted protein,
since reprobing with polyclonal
antiserum shows comparable signals
for all mutants, including 10-60 and 32-84 (Fig. 6B).
Moreover, it is apparent that the migration of the mutants under
nonreduced conditions is altered by the mutations, implying that the
conformation of the single chain mutants and CG
are not the
same.
Similar data were obtained with B109, which is specific for the dimer
form of the subunit (24-26) (Fig. 6C). Tethers
containing mutations 7-31, 28-82, and 59-87 and the nonmutated
are recognized as well as the heterodimer, but single chains
containing either the 10-60 or 32-84 mutation show very weak signal
compared with the same blotted membrane using CG
antiserum (Fig.
6D). Both A407 and B109 recognized higher molecular weight
proteins in CG
28-82 and 59-87. They are likely noncovalently
linked because when the samples are boiled under nonreducing
conditions, the "aggregates" disappeared (data not shown). (It is
not clear why the aggregates are detected by monoclonal antibodies but
not by polyclonal antiserum; we are currently purifying these proteins for extensive characterization.) These results suggest that the conformation of single chain mutants 10-60 and 32-84 differ
substantially from the nonmutated single chain and the heterodimeric
forms.
Because we could
now obtain sufficient quantities of single chains bearing the cystine
knot mutants, the influence of the disulfide bonds on receptor
binding/signal transduction was examined using human kidney 293 cells
expressing the LH/hCG receptor. For these bioassays, conditioned media
were quantitated with a -specific polyclonal based radioimmune assay
(see "Materials and Methods"). It is apparent that except for the
59-87 mutant, the binding affinity of all of the mutants was
comparable with the nonmutated tether (Fig.
7A, Table III). The binding
affinity of single chain 59-87 was reduced 10-fold, similar to that
for the heterodimer containing this mutation in the monomeric
subunit as previously reported (10).
|
Adenylate cyclase activation by the single chain was also determined (Fig. 7B). The data demonstrate that signal transduction parallels receptor binding affinity, suggesting that the determinant(s) necessary for bioactivity is conserved between CG heterodimer and single chain (Fig. 7B, Table III). Thus, despite differences in the intracellular behavior and immunoreactivity to conformationally sensitive monoclonal antibodies, the mutants are nevertheless biologically active in vitro, with coupling of receptor binding to signal transduction unaffected by these structural modifications.
Within a species, the amino acid sequence of the subunit is
identical (1), and it is the unique structure of the
subunit that
ensures binding of each dimer to its cognate receptor. The
subunit,
therefore, must be very flexible, since it combines with each of the
four glycoprotein
subunits and
subunits from different species
(27). Further evidence for this adaptability is the conformational
changes occurring during dimer formation (for review, see Refs. 1 and
28-34). Presumably, the configuration of the disulfide bonds of the
subunit is critical for its interaction with the
subunit.
Previous studies on these bonds in the monomeric
subunit
demonstrated that single mutations in the cystine knot blocked
secretion and/or inhibited heterodimer formation (10). Although the
results implicated a critical role for these bonds in the intracellular
behavior of the subunit, we could not exclude the possibility that the
free thiol, rather than the disrupted bond per se, led to
the observed changes (9, 10, 19, 20). To address this issue, double
cysteine mutations of all the proposed pairs were constructed in the
subunit. It was apparent that mutations in the knot altered the
recovery of the subunit from the media. Recovery of variants containing
mutations outside the cystine knot,
7-31 and
59-87, was
comparable with the wild type
subunit. These changes in the
secretion patterns presumably reflect alterations in the folding of the
subunit. Consistent with this hypothesis are the experiments showing
that, compared with the wild-type
subunit, the cystine knot
variants 10-60 and 32-84 combined much less efficiently to the
subunit, with less than 5% of these mutants recovered as dimers. Thus,
assessing the role of the
10-60 and
32-84 bonds in the
biological action of hCG was precluded. Because tethers containing
either mutation at 10-60 or 32-84 were recovered in the media, we
could determine their biologic activity. Unexpectedly, it was observed
that all of the cystine knot and the
7-31 mutants exhibited high
affinity binding for the receptor. When corrected for binding, signal
transduction was unchanged regardless of the mutation. These results
imply that disulfide bonds of the cystine knot are required for maximal secretion/assembly but that for receptor recognition, not all of the
bonds in the core are needed. The single chain bearing the
59-87
mutant exhibited a 10-fold decrease in receptor binding. Several
studies have demonstrated the importance of the C-terminal residues
88-92 in the
subunit for maximum binding affinity (1, 35-38). It
is likely that disrupting the 59-87 bond perturbed the configuration
of the carboxyl-terminal region, leading to the observed decrease in
receptor binding. That the altered intracellular behavior of these
tethers is not accompanied by a significant change in biologic
activity, especially the 10-60 and 32-84 mutants, suggests that the
receptor could recognize different conformations of the hormone. While
this conclusion is supported by the absence of recognition by the
dimer-specific monoclonal antibodies, we cannot exclude a greater
liability of the epitopes in the 10-60 and 32-84 mutants to the
SDS-polyacrylamide gel electrophoresis conditions used.
Given that perturbing the cystine knot affects the intracellular
behavior of both the monomeric subunit and the single chain molecule, the conformation of this region is likely to be relatively conserved regardless of the associated hormone-specific
subunit. This fixed core determinant may be necessary for the "escort" (42,
43) function of the
subunit, which is critical for rescuing the
pituitary
subunits from the endoplasmic reticulum because few, if
any, uncombined LH, follitropin, and thyrotropin
subunits are
secreted in the absence of co-expressed
subunit (34, 39, 40). Thus,
we would propose that the major conformational changes of the
subunit resulting from assembly with the
subunit occurs in regions
outside of the cystine knot, i.e. the three hairpin loops
(2, 3, 23, 41). Consistent with this interpretation, the recovery of
the variants with the 7-31 and 59-87 mutations, which lie outside the
knot, was comparable with the wild type.
That the intracellular fate of single chains with mutations of paired
cysteines and corresponding monomers parallel each other suggests that
a misfolded epitope is not recognized by an intracellular transport
component and/or that the mutants are trapped by a chaperone, which
leads to enhanced degradation. Although the domain increases the
recovery of the mutated single chains, e.g. in
CG
10-60 and CG
32-84, they are still less than that of the
wild type. Thus, even the presence of the
subunit domain is not
sufficient to completely override the
Cys mutations.
Other members of the cystine knot growth factor superfamily display
similar characteristics (6-8). Mutations of the paired cysteine
residues within the knot of tumor growth factor-1 or activin
dramatically reduced secretion (6, 7). Thus, together with our data
presented here, the cystine knot is apparently the basic frame of the
glycoprotein hormone subunits and several growth factors and represents
a critical determinant for secretion and assembly of functional ligand.
Receptor recognition is apparently less dependent on the configuration
of this structure. The data also show the importance of the single
chain approach to study structure-function of multisubunit proteins
where dependence on the assembly step is essential for biologic
action.
We thank Drs. David Ornitz, Mesut Muyan, and Edward Grotjan for critical review of the manuscript and Dr. Steven Birken for helpful discussions. Also, we are grateful to Susan Carnes for invaluable assistance in preparing this manuscript.