 |
INTRODUCTION |
The hormone vasopressin is synthesized as part of a larger
precursor that contains the vasopressin sequence at its amino terminus, followed by that of the disulfide-rich protein neurophysin
(NP)1 and a carboxyl-terminal
glycopeptide known as copeptin (Ref. 1 and reviewed in Ref. 2).
Oxytocin biosynthesis is similar (3), with the exception that the
precursor lacks the copeptin segment, the function of which is unclear
(4). The NP components of the two precursors (VPNP) and
(OxyNP) are highly homologous; the two bovine neurophysins
have almost identical properties in vitro, including similar
affinities for each of the two hormones (e.g. Ref. 2).
Following processing, which cleaves the hormones from NP, the mature
hormones remain noncovalently bound to NP by forces basically analogous
to those pre-existing within the precursor, leading to analogous NP
self-association properties (5). The structure of NP, the nature of the
noncovalent interactions between hormones and NP, and the effects of
these interactions on NP properties have been extensively investigated in solution (e.g. Ref. 2) and by crystallographic analysis (6, 7). Moreover, the central role of NP in vasopressin elaboration has
become evident by the demonstration that familial neurogenic diabetes
insipidus (FNDI), an autosomal dominant disease characterized by
vasopressin deficiency, appears most frequently to be due to mutations
in NP (e.g. Refs. 8 and 9) accompanied by loss of the proper
targeting of the hormone to regulated neurosecretory granules
(e.g. Refs. 10 and 11); retention of the mutated precursor
in the endoplasmic reticulum is generally considered the cause of the
ultimate death of the affected neurons (10).
Many of the mutations involved in diabetes insipidus can be predicted,
on the basis of what is already known, to exert their effects by
directly or indirectly interfering with NP folding and/or dimerization
(e.g. Refs. 12 and 13). Protein targeting to regulated
neurosecretory granules is thought to depend at least in part on
self-association, although the nature of the requisite self-association
process is not well defined (reviewed in Ref. 14). In the case of
oxytocin and vasopressin precursors, the self-association arises from
the self-association properties of their NP segments. The central
element of this self-association is the dimer, the formation of which
occurs through specific 
-sheet interactions between folded NP
subunits (6, 7) and is strongly enhanced by the binding of hormone or
related peptides to the binding site (15). The efficiency of NP folding
is in turn critically linked to the correct but metastable (16) pairing
of its 7 disulfides and depends, with one known exception (17), on the
ability of the protein to bind hormone and thereby provide the
necessary thermodynamic stability to drive formation of the correct
disulfides to completion. Thus, mutations (e.g. Ref. 9) that
interfere with disulfide formation by deleting or adding half-Cys
residues, or that alter binding site residues, can be expected to
affect adversely both folding and targeting.
However, there are a number of FNDI mutations for which the molecular
mechanisms underlying their biological effects are ambiguous and
therefore have the potential to provide new information on the
relationship between structure and function in this system. These
include mutations in residues involved neither in binding nor
dimerization, such as those in external -turns or in the 55-60-loop
region that connects the two NP domains, or a premature STOP mutation
that deletes both copeptin and the disordered (6, 7) carboxyl-terminal
NP tail (9). They also include mutations affecting more clearly
essential residues. Accordingly, we have investigated the effects of
these mutations on NP physical-chemical properties. Mutations studied
that do not involve obviously essential residues are G57S and G57R in
the inter-domain loop (8, 9), G17V in the first -turn (18), and
87STOP near the carboxyl terminus (9). Additionally, we investigated
the C67STOP mutation, which deletes most of the second domain of the
protein, together with half of the subunit interface and one cysteine,
and the 
E47 mutation that deletes the critical binding site residue
Glu-47 (e.g. 9), located in the 3,10 helix of the protein.
The effects of these mutations were principally investigated by
introducing them into bovine OxyNP and in one case into the
oxytocin precursor, for both of which we have developed an expression
system. This takes advantage of the strong homology in structure and
function between OxyNP and VPNP within and
among different mammalian species (2, 19). Fig.
1 compares the amino acid sequences of
bovine OxyNP (BNP-I), bovine VPNP (BNP-II), and
human VPNP. With the exception of the carboxyl-terminal
tail, for which no function is known, there are only 4 amino acid
differences between bovine and human VPNP, and none involve
residues at the hormone-binding or dimerization sites. Also, with the
exception of the carboxyl-terminal tail, which is eliminated when the
protein is terminated with residue 86, all the mutated residues are
strictly conserved. Therefore, it seems reasonable to expect that the
similarity in behavior between bovine OxyNP and bovine
VPNP will be paralleled by a similarity between bovine
OxyNP and human VPNP; a single exception to
this may be evident in these studies. It is also relevant that the
present study emphasizes the effects of mutation on mature NP, as
opposed to the NP segment of the precursor. We selected this route
because of the inefficiency associated with the in vitro
folding of the wild type precursor (20), which contrasts with the
efficiency of folding the wild type processed protein using added
dipeptide to drive the folding process (16, 21). In the single case in
which the effects of mutation on both the processed protein and
precursor are compared, the same effect is seen. A few studies of the
effects of chemical or enzymatic modification of the two bovine NPs are
also presented. Fig. 2 is a diagram of NP
structure, including its disulfide pairs, that shows the positions of
the different mutations investigated.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 1.
Sequence of human VPNP compared
with that of bovine VPNP (BNP-II) and bovine
OxyNP (BNP-I). Data are from Ref. 19. Note that the
sequence of BNP-I and its mutants used in this study terminate at
Ser-92 (20).
|
|
View larger version (22K):
[in this window]
[in a new window]
|
Fig. 2.
Diagram of the structure of BNP-II showing
the position of the mutations introduced into BNP-I. The
dotted lines show the disulfide connectivities.
|
|
| |
MATERIALS AND METHODS |
Preparation of Wild Type and Recombinant Proteins--
BNP-I and
-II, isolated from bovine posterior pituitary lobes as described
previously (22, 23), and affinity-purified (24), were used for all
studies of the wild type (WT) mature proteins. The preparation of
recombinant WT mature BNP-I and oxytocin precursor and mutagenesis of
wild type cDNA have been described previously (17, 20). The
recombinant mature mutant proteins in the present study were expressed
in Escherichia coli, folded at pH 8.0 in the presence of
-mercaptoethanol and 10 mM Phe-Tyr-NH2, and
purified by affinity chromatography as described for the recombinant mature WT protein (17). A large fraction of the protein so prepared is
covalently damaged prior to folding (17, 20), and the remaining protein
folds with variable efficiency depending on its structure (see below).
The covalent structures of mutants were confirmed by DNA sequencing of
the plasmids from which they were prepared and by mass spectrometry.
With proteins that fold efficiently, the binding-competent and
binding-incompetent proteins are distinguished by clear differences in
mass, reflecting the covalent damage of the binding-incompetent protein
(20). This was not the case for several of the mutants, indicating the
inability of undamaged protein to fold efficiently ("Results").
None of the E47 mutant was retained by the affinity column after
folding (as expected because of its incomplete binding site), but
because of CD evidence of folded product within the binding-incompetent fractions (245 nm positive ellipticity), this mutant was further fractionated by reverse phase HPLC using a Vydac C18
column, 10× 250 mm, and a gradient from 100% solution A (25%
CH3CN, 40 mM ammonium acetate) to 100%
solution B (80% CH3CN, 40 mM ammonium acetate)
at 150 min. A sharp peak at 13 min gave a CD spectrum characteristic of
folded protein and was used for further analysis. The same HPLC
protocol was applied to the C67STOP mutant, but only a diffuse peak
eluting approximately between 14 and 24 min was obtained
("Results").
The G57S mutation was also introduced into the full-length oxytocin
precursor. The precursor was folded, and the folded precursor was
purified essentially as described previously for the WT precursor (20).
Folding Efficiencies of Recombinant Proteins--
Although
purified WT BNP-I refolds from the reduced state with ~90%
efficiency under our folding conditions (21), the yield from the
initial folding reaction of folded recombinant WT protein, or of its
stable mutants, is considerably lower than 90% and is somewhat
variable, reflecting the covalently damaged protein formed during
expression. Nonetheless, the folding efficiencies of the G57S, G57R,
and P87STOP mutants were within the experimental range of that of WT
and other stable proteins, whereas yields of the E47 mutant and turn
AG insertion mutant were reduced by factors of ~50 and 80%
respectively. The folding efficiency of the G17V mutant was
significantly lower than that of the turn AG mutant, but was too low
for accurate estimates. Note that the theoretical folding efficiency of
the WT protein under the conditions used here can be calculated to be
~95%, indicating an equilibrium heavily in favor of the folded form.
Therefore, only relatively large reductions in stability should be
manifest as a significant reduction in yield.
Chemical and Enzymatic Modification of Wild Type
Protein--
Succinylated BNP-I was prepared from native BNP-I as
described elsewhere (25). Following succinylation, protein retarded by
the affinity column is succinylated at its two Lys residues and
randomly succinylated at Ser residues exclusive of Ser-56, whereas
protein not retarded by the affinity column differs from the retarded
fraction by its succinylation at Ser-56 (25).
For preparation of BNP-I with a specific internal clip after Lys-18, 13 mg of native BNP-I was digested with 3 units of endoproteinase Lys-C
(Sigma) for 24 h at 37 °C in 0.5 ml of 0.1 M
NH4HCO3 buffer, pH 9. Residual native protein
was separated from the cleaved protein by affinity chromatography (15),
the cleaved protein not binding to the affinity column. On native
polyacrylamide gel electrophoresis at pH 9.5, the non-binding protein
gave two bands, a major product (
80% of the protein) migrating with
the expected additional negative charge relative to the native protein
and a minor band with the same mobility as the native protein.
Automated Edman sequencing confirmed cleavage after Lys-18 but gave no
evidence of any undigested protein. Two controls were run in order to
determine whether factors other than cleavage after Lys-18 could
account for the altered conformational properties (see "Results")
of the digested protein. No significant effect on conformation was
noted if the native protein was subjected to enzymatic digestion
conditions in the absence of enzyme or if protein succinylated at its
two Lys residues was subjected to enzymatic digestion conditions in the
presence of enzyme.
Des-1-6 BNP-II was prepared as described earlier (26). The
des-1-6,des-87-95 protein was prepared from the purified des-1-6 derivative by limited digestion with trypsin (1% by weight) at room
temperature in 0.16 mM KCl, 50 mM Tris buffer,
pH 8.2. Digestion was monitored by non-denaturing polyacrylamide gel
electrophoresis at pH 9.5 and was terminated by snap-freezing aliquots
at 
80 °C. The crude modified protein was purified by HPLC on a
Vydac C18 reverse phase column using a gradient from time 0 (25%
acetonitrile, 0.1% trifluoroacetic acid in water) to 40 min (35%
acetonitrile, 0.1% trifluoroacetic acid in water). The method
separates the des-1-6,des-87-95 protein (retention time = 20 min) from des-1-6,des-93-95 protein (retention time = 22 min)
and from most of the more completely digested protein; the latter is
completely removed by a subsequent affinity chromatography (24) step.
Binding fractions from affinity chromatography, representing the
des-1-6,des-87-95 protein, were dialyzed against water and
lyophilized. Purity and identity were confirmed by native
polyacrylamide gel electrophoresis and amino acid analysis. Relative to
the initial native BNP-II used, the yield of the final
des-1-6,des-87-95 protein was ~4%.
Determination of Peptide Binding Constants and Dimerization
Constants--
Peptide binding constants were determined by CD using
the mononitrated derivatives of the different mutants and binding
methodology described elsewhere (e.g. Ref. 27); the peptides
used were Phe-Tyr-NH2 or Abu-Tyr-NH2,
depending on availability. Dimerization constants were determined from
one-dimensional proton NMR spectra obtained in D2O as
described previously (17), measuring the intensity ratio at different
protein concentrations of signals at ~6.4 and 6.2 ppm representing
dimer and monomer, respectively; the validity of the use of these
signals in the mutants was verified by demonstrating that the
calculated dimerization constants were independent of protein concentration.
Stability Measurements--
Stabilities to denaturation by
guanidine HCl were determined by one or both of the two methods. In the
conventional method, applied earlier to this system, CD spectra were
obtained by a constant concentration of protein in different
concentrations of guanidine HCl (17, 20). In the second method (28),
carefully weighed quantities of solid guanidine HCl are added serially
to a sample of protein; the guanidine concentration is calculated from
the weight of guanidine added and known dilution factors. This method,
which was used for many preparations because of its smaller protein
requirement, is intrinsically less accurate, due in part to the
concentration dependence of the stability of WT mature NP and many of
its mutants (17). Nonetheless, good agreement between the two methods
was seen where the two methods were applied to the same sample (Table
I). Both methods assume a two-state denaturation system, and both
typically have a reproducibility of better than 5%, although
occasional exceptions are evident.
Other Methods--
Most other methods have been described
previously, i.e. determination of protein folding kinetics
(16), NMR, CD (e.g. Refs. 16 and 20), mass spectrometry
(20), amino acid analysis, and protein sequencing (e.g. Ref.
25). Protein modeling utilized the Swiss Protein Data Bank modeling
system and the crystal structures of des-1-6 bovine NP-II in its
unliganded state2 and as its
lysine vasopressin complex.2
| |
RESULTS |
Effects of the G57S and G57R Mutations--
Gly-57 does not
participate directly in either binding or dimerization (6, 7), and the
G57S mutant folded with an efficiency within the range of the WT
protein (see "Materials and Methods"). The near-ultraviolet CD
spectrum of the affinity-purified G57S mutant (Fig.
3), which represents the contribution of
NP disulfides, is identical to that of the WT protein (Fig. 3),
consistent with a similarity in conformation. Note that this spectrum
is very sensitive to NP folding; the 245-nm positive band is absent in misfolded protein, and the 280-nm negative band is reduced in intensity
and shifted to shorter wavelengths (e.g. Ref. 29). One-dimensional NMR spectra were also largely similar to spectra of
BNP-I, but significant differences were evident in the intensity ratio
of monomer and dimer signals at ~6.2 and ~6.4 ppm at the same
protein concentration (Fig. 4).
Calculation of this ratio as a function of concentration indicated a
dimerization constant 20-30% that of BNP-I (Table II). Surprisingly
(cf. Ref. 17), the stability to guanidine denaturation was
unaffected (Table I), suggesting that the
reduced dimerization might reflect increased monomer stability. The
peptide binding constant of the G57S mutant was also reduced by
20-30%, but as with the WT protein the binding affinity increased
with protein concentration, signifying stronger binding by dimer than
by monomer (Table III). NMR studies (not shown) confirmed the
peptide-induced conversion of monomer to dimer. The effects of
concentration on binding and of binding on dimerization are tentatively
consistent with estimates of an ~10-fold higher binding constant to a
dimer site than to a monomer site found for the WT protein (27),
resulting in a 100-fold increase in dimerization constant in the bound
state (30).
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 3.
Near-ultraviolet CD spectra at neutral pH of
the folded forms of WT BNP-I and of various mutants. The spectrum
of the C67STOP mutant is shown for comparison. Top spectra,
curve 1, WT BNP-I; curve 2, G57S;
curve 3, G57R; and curve 4, C67STOP.
Lower spectra, curve 1, WT BNP-I; curve
2, P87STOP; and curve 3, E47. Results are expressed
as molar (not residue) ellipticities. To show differences in 245/280 nm
ellipticity ratios, data for all mutants except C67STOP are normalized
to a value of 20,000 at 280 nm, the ellipticity of the WT protein,
although small differences among the mutants in absolute ellipticity
might be present.
|
|
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 4.
600 MHz proton NMR spectra of the correctly
folded forms of WT BNP-I and of selected mutants in D2O at
a protein concentration of 0.3 mM, pH 6.2. Arrows point to the  -proton signals at ~6.4 ppm (dimer)
and 6.2 ppm (monomer) used to calculate dimerization constants. Signals
upfield from 0.55 ppm and the small signal at ~1.0 ppm are also
assigned to dimer. The partial spectrum above the full spectrum of the
87STOP mutant represents the same sample in the presence of 2 mM Phe-PheNH2; the peptide has no signals in
this region. The increase in the 6.4 ppm signal and disappearance of
the 6.2 ppm signal show the ligand-induced increase in dimer
content.
|
|
View this table:
[in this window]
[in a new window]
|
Table I
Stabilities to guanidine HCl denaturation of wild type and mutant
proteins
Studies were conducted at pH 6.0 in 0.1 M ammonium acetate
unless otherwise noted. A and B represent the denaturation method used,
where A is that used in earlier NP studies (17, 20), and B involves
addition of solid guanidine HCl (see "Materials and Methods"). The
value m is the slope of the guanidine denaturation free
energy curve (40). Data are reported to two significant figures ± S.E. of readings at 247 and 250 nm unless otherwise indicated.
|
|
The reduction in binding constant by G57S is greater than can be
accounted for by the reduction in its dimerization constant, e.g. the binding constant of G57S can be shown to only half
that calculated for BNP-I when the two are compared at equivalent
weight fractions of dimer as described for other mutants (17). This is
particularly evident with the G57R mutant. The dimerization constant of
this mutant is the same or slightly higher than that of the G57S mutant
(Table II), but, at similar protein
concentrations, its binding affinity for Abu-Tyr-NH2 is
~1/4 that of the G57S mutant, and its binding affinity for
Phe-Tyr-NH2 is only 13% that of BNP-I (Table
III). Again, the link between binding and
dimerization is preserved in the G57R mutant as judged by the
dependence of its binding on concentration (Table III) and by NMR
studies (not shown). Factors additional to reduced dimerization
therefore operate to reduce the binding affinities of both mutants. The
CD spectrum of the G57R mutant (Fig. 3) is slightly altered relative to
that of the wild type protein in the unliganded state, raising the possibility that it has a slightly altered conformation. This is
consistent with modeling studies (see "Materials and Methods") that
suggest weak steric hindrance from Arg-57 in the unliganded state that
becomes more pronounced in the liganded state. Modeling studies
similarly suggest that the G57S mutation should introduce steric
problems in the liganded state but significantly less so than the
Arg-57 mutation; again, fewer problems are evident in the unbound
state. The modeling results therefore provide an explanation for the
binding affinity reductions in the two mutants that do not arise from
the reduction in dimerization.
View this table:
[in this window]
[in a new window]
|
Table II
Effects of selected mutations on neurophysin dimerization in
D2O at pH 6.2
The estimated error in dimerization constants is ±20%.
|
|
View this table:
[in this window]
[in a new window]
|
Table III
Effects of mutations and modifications on neurophysin binding
affinities
Studies were conducted at pH 6.2 and 25 °C in 0.1 M
ammonium acetate, containing 20 mM Mes buffer. Values in
parentheses are the protein concentrations at which binding constants
were determined. Relative binding constants (last column) are reported
to one significant figure and represent the ratio of the experimental
constant to that of BNP-I (or BNP-II where applicable) at the same
protein concentration; where two values are given, the first is for the
Abu peptide. Binding constants for the WT proteins, when not
experimentally determined at a given protein concentration, were
calculated from the known concentration dependence of binding (27).
|
|
Because the effects of Gly-57 on dimerization suggested a modulating
role in dimerization for the loop between the two domains, the effects
on dimerization of succinylation of Ser-56 were also investigated. This
modification was shown previously to lead to a large decrease in
binding constant (25). Table II demonstrates a reduction in
dimerization in the unliganded state by a factor of 2-3 when NP
succinylated at Ser-56 is compared with protein similarly modified with
the exception of Ser-56 (see "Materials and Methods").
Precursor stability depends strongly on the strength of interaction
between hormone and NP segments (20); so the effects of Gly-57 mutation
on binding predict a diminution in the stability of the oxytocin
precursor containing such mutations (17, 20). The G57S mutation was
accordingly incorporated into the precursor (see "Materials and
Methods"), and its effect on stability was evaluated. Previous
studies of the WT precursor had demonstrated a pH-dependent
dissociation of the internal interactions between the oxytocin and NP
segments of the wild type precursor with a CD midpoint near pH 10 (20).
The clearest CD indicator of this transition (least influenced by
tyrosine ionization (20)) was the change in the ratio of ellipticity at
291 nm to that at 280 nm, which occurs with a midpoint of 9.7 in the WT
precursor (20). The neutral pH spectrum of the G57S precursor (Fig.
5) was similar to that of the wild type
(cf. Ref. 20) and similarly exhibited a
pH-dependent transition characteristic of internal
hormone-protein dissociation. However, the CD midpoint was shifted
downward to pH 9.25 in the G57S precursor (Fig. 5), indicating (20) a
0.6 kcal/mol weaker free energy of internal bonding between hormone and
NP than in the WT precursor.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 5.
Relative stabilities of the WT and G57S
precursors as monitored by the effects of pH and guanidine HCl.
Left, near-ultraviolet CD spectrum of the G57S precursor at
neutral pH (bold line) and at pH 11. The change in this
region is assigned to the dissociation of internal NP-hormone
interactions (20). Inset, plot of the % dissociation of the
internal complex as a function of pH, as calculated from the 280/291-nm
ellipticity ratio for the WT protein ( ) and G57S mutant ( ).
Right, plot of % unfolding at pH 6 as a function of
guanidine HCl concentration ([GuHCl]) for the WT and G57S
precursors; symbols are the same as in the
inset.
|
|
The difference between the WT and G57S precursors in the strength of
interaction between hormone and NP was also manifest by a difference in
stability to guanidine denaturation (Fig. 5). Denaturation midpoints
were at 3.85 M guanidine and 4.25 M guanidine for the G57S and WT precursors, respectively, and did not significantly change with protein concentration over the range studied. Comparison of
the stabilities of the WT and mutant precursors (Table I) calculated
from their individual denaturation profiles indicated a lower stability
of the mutant by 0.3-0.4 kcal/mol. However, the intrinsic uncertainty
in the absolute value of G0 for each protein
(see Table I) leads to still greater uncertainty in the differences
calculated in this manner. Since the shapes of the denaturation
profiles of both proteins are essentially identical (Fig. 5), and the
derived values of m for the two proteins are within the
experimental error of these measurements (Table I), we also estimated
the stability difference between the two proteins from the difference
in their denaturation midpoints (0.4 M), assuming the same
value of m for each (1.7), and the relationship (Equation 1),
|
(Eq. 1)
|
By either consideration, the lower stability of the G57S precursor
calculated from the guanidine data is reasonably consistent with that
estimated from the pH data and therefore appears primarily to arise
from weaker internal NP-oxytocin interactions.
Effects of the G17V Mutation and Related Studies of the Role of the
14-18 Turn--
Gly-17 is part of the turn connecting the first two
strands of -sheet (Fig. 2). Mutation to Val yielded very low
quantities of binding-competent protein (see "Materials and
Methods"). Mass spectrometry indicated that both the
binding-competent protein and a large fraction of the non-binding
protein had the correct mass, suggesting that much of the non-binding
protein was not covalently damaged (see "Materials and Methods")
but reflected an unfavorable folding equilibrium. This was confirmed by
resubjecting the non-binding protein to folding conditions, followed by
the same workup as given to the original folding mixture. Upon affinity chromatography, a similar percentage of protein was found in the binding fraction as found following the initial folding reaction (data
not shown).
The binding and non-binding fractions were distinguishable
electrophoretically; the binding fraction moved ~5% more rapidly to
the anode than the non-binding fraction, suggesting tighter folding. It
is significant that no component with the same mobility as the binding
fraction was identifiable in the non-binding fraction, indicating that
the binding affinity of the folded protein, while possibly not
normal, was not reduced to an extent that prevented complete retention
by the column. Although only very small quantities of binding protein
could be isolated, the differences in folding between binding-competent
and binding-incompetent protein were discernible in NMR spectra (see
below) and in disulfide CD spectra, the CD spectrum of the non-binding
component, as other non-binding components, completely lacking the
245-nm positive CD band (data not shown). However, CD spectra of the
binding fraction were also clearly atypical (Fig.
6). The intensity of the 245-nm positive peak characteristic of the folded state relative to that of the 280-nm
negative band was markedly lower than in the WT protein, and this
difference was increased at low pH. Note that the intensity of this
peak in the WT protein is essentially independent of pH (e.g. Ref. 31). With the assumption that the disulfides are correctly paired in the binding fraction, the results indicate that,
even with the correct pairing, mutation of Gly-17 to Val alters the
folded state.
View larger version (28K):
[in this window]
[in a new window]
|
Fig. 6.
Near-ultraviolet CD spectra of
binding-competent Gly-17 turn mutants as a function
of pH. Inset, G17V at pH 2.1 (curve 1) and
pH 6.2 (curve 2). Main figure, turn AG mutant at
pH 2.5 (curve 1), pH 3 (curve 2), and pH 6.2 (curve 3). CD data are expressed in millidegrees
ellipticity.
|
|
Fig. 7 compares regions of the proton NMR
spectra of the non-binding form of the G17V mutant with that of the
folded WT protein at pH 6.2. The region downfield of 4.7 ppm contains
aromatic protons and -protons that are downfield-shifted by
-sheet formation; the peak at 6.7-6.8 ppm represents the 3,5 ring protons of Tyr-49 and can be used as an intensity reference (26).
The results show the presence in the non-binding form of the G17V
mutant of a diminished number of downfield-shifted -protons relative
to WT, suggesting a significant reduction in -structure,
e.g. note the complete absence of a 6.2- or 6.4-ppm signal.
Quantities of the binding form of G17V were inadequate for good
spectra, but signals throughout the spectrum show its greater
similarity to WT than to the G17V non-binding species; particularly
note the 1.7-ppm region (Fig. 7).
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 7.
Proton NMR spectra of selected
binding-competent and binding-incompetent species in D2O at
pH 6.2. From top to bottom: C67STOP, 0.4 mM; G17V, binding-competent, concentration unknown; G17V,
binding-incompetent, 0.3 mM; BNP-I, Lys-C protease-treated,
0.8 mM; BNP-I (native), 0.3 mM. See Fig. 4 for
other details.
|
|
Two other studies of the 14-18 turn were carried out. We particularly
wanted to learn whether -sheet formation forced formation of the
turn, which was sterically hindered by a Val in position 17, or whether
the turn actively promoted the docking of -strands. In one study,
native BNP-I was treated with Lys-C protease to generate an internal
clip after Lys-18 (see "Materials and Methods"). Previous studies
(26) had demonstrated that insertion of an internal clip after Glu-47
in native NP did not significantly alter conformation. If the turn
played only a passive role, its cleavage after the correct disulfides
were formed should also not significantly alter conformation. In the
case of cleavage after Lys-18, however, all the protein became
binding-incompetent, and the 245-nm positive CD band was completely
lost (data not shown). Significantly, the NMR spectrum of this
derivative (Fig. 7) was almost identical to that of the non-binding
component of the G17V mutant, consistent with a similar effect on
-structure.
We also investigated the effect of expanding the size of the 14-18
turn by interposing the sequence AG between Gly-17 and Lys-18. With the
exception of a greater (but still low) folding efficiency (see
"Materials and Methods"), the folding properties of the mutant
(turn AG) were analogous to those of the G17V mutant. As with the G17V
mutant, additional folded protein could be formed from the
binding-incompetent fraction by resubjecting it to folding conditions.
In this case, we additionally demonstrated that this was not the case
if -mercaptoethanol was omitted during refolding, showing that
disulfide rearrangement was critical to the refolding process.
Analysis of the turn AG binding fraction by CD indicated spectral
properties similar to those of G17V (Fig. 6). The 245-nm positive band
was barely evident at pH 2.5 but increased in intensity with an
increase in pH to 6.2, although it did not achieve the intensity of the
WT protein under these conditions. Preliminary CD studies in the far UV
suggested a lower content of -sheet in that mutant than in the WT
protein. Native BNP-I and -II have atypical CD spectra in this region
for proteins rich in -sheet (e.g. Refs. 22 and 29), with
a single negative band at 208-209 nm that shifts to 203-204 nm on
denaturation (e.g. Ref. 29). The turn AG mutant exhibited a
negative band centered at ~204 nm and did not achieve the positive
ellipticity in the 190-200-nm region seen for the folded WT protein.
We assume that the binding-competent G17V and turn AG mutants have the
correct disulfide pairs. Nonetheless, the optical activity of these
mutants indicates that their conformations differ from that of the
folded WT protein.
Effects of Deletion of Glu-47--
Glu-47 plays a central role in
binding the hormone -amino group, the latter thus far critical to
recognition by NP (2). Its deletion also alters the 3,10 helix.
Accordingly, no binding-competent protein was isolated from the E47
mutant, and the presence of peptide during folding had no clear
influence on the final product. However, fractionation of the
binding-incompetent protein by HPLC (see "Materials and Methods")
led to isolation of a discrete component with a strong positive CD band
at 245 nm (Fig. 3). The 245/280 nm ratio for this peak was 60% that of
the WT protein and has so far not been improved by refractionation. The
yield of this component was approximately half that of the WT protein
(see "Materials and Methods").
Folding of the HPLC-purified protein was confirmed by NMR. The spectrum
of this component (Fig. 4) suggested -sheet formation similar, but
not necessarily identical, to that of WT protein, as evidenced
particularly by the chemical shifts and content of downfield
-protons. Signals at ~6.2 and ~6.4 ppm associated with monomer
and dimer, respectively, in the native protein are slightly shifted in
the mutant but were demonstrated to be similarly affected by changes in
dimerization induced by changes in concentration or pH. Assuming that
the fractionated protein is homogeneous, the dimerization
constant calculated from the relative intensity of these signals is
essentially identical to that of the WT protein in D2O
alone (Table II). However, we cannot exclude the possibility that the
low 245/280 nm ratio reflects the presence of unfolded protein that was
not separated from folded protein by HPLC. Since such protein can be
maximally estimated as ~50% in content (this assumes a normal
245/280 nm ratio for the folded Glu-47 deletion mutant and typical
spectral properties (29) for the unfolded protein) and should neither
dimerize nor contribute to signals at 6.2 and 6.4 ppm, the actual
dimerization constant might be twice as high as that shown. The
stability of the mutant was within experimental error of that of the WT
protein (Table III).
Although the hormone (or peptide) -amino group has always appeared
essential to binding to NP, it binds to groups additional to Glu-47 (6,
7). Accordingly, we looked for peptide binding by the G47 mutant
that would be too weak to allow its interactions with the affinity
support. CD binding studies using the mononitrated form of the protein
(see "Materials and Methods") indicated no perturbation of the
protein nitrotyrosine at concentrations of Phe-Tyr-NH2 as
high as 28 mM, but the nitrotyrosine appeared optically inactive in this mutant. We then investigated the effect of peptide (Phe-PheNH2) on the NMR spectrum of the (non-nitrated)
mutant. Small changes were reproducibly seen in the intensity ratio of the dimer and monomer signals at ~6.4 and 6.2 ppm, which were not
assignable to signals from the free peptide and which suggested an
increase in dimerization constant in the presence of peptide (data not
shown). However, these changes were unaccompanied by other changes
normally associated with increased dimerization and require additional investigation.
Effects of Termination of NP at Residue 66--
Essentially all of
the peptide-binding site on NP resides within the first 60 residues (6,
7). Nonetheless, no demonstrable binding-competent protein was obtained
from folding mixtures of the C67STOP mutant (see "Materials and
Methods"). The binding-incompetent protein gave a near-UV circular
dichroism spectrum (Fig. 3) similar in shape, but weaker in 280-nm
negative molar intensity, to that typical of binding-incompetent
protein. Because the incomplete set of disulfides present might itself
lead to an altered spectrum, an NMR spectrum was also obtained (Fig.
7). Several relatively weak downfield-shifted -protons were seen,
suggestive of -sheet. The intensity of these signals relative to the
2-proton signal from Tyr-49 at ~6.8 ppm allows the presence of
~30% of a species with partial -structure content, if all belong
to the same component. However, in contrast to the E47 mutant,
reverse phase HPLC analysis failed to resolve discrete components of
the C67STOP mutant, the protein migrating as a single very diffuse band
(see "Materials and Methods"). Far ultraviolet CD analysis of
different sections of this band revealed only subtle differences that
were not amenable to interpretation, and all the spectra generally
resembled that of unfolded NP.
Effects of Termination of NP at Residue 86--
The yield of
folded protein obtained from the P87STOP mutant (see "Materials and
Methods") and its near UV circular dichroism spectrum (Fig. 3) and
stability (Table I) were essentially the same as those of the WT
protein. Apart from the loss of signals from the deleted
carboxyl-terminal residues, proton NMR spectra (Fig. 4) were also
normal for folded NP. However, the peptide binding constant of the
87STOP mutant was twice that of the WT protein at the same protein
concentration (Table III), whereas the dimerization constant was only
half that of the WT protein (Table II). The results signify a higher
intrinsic binding constant to the monomeric and/or dimeric mutant
protein than to the WT protein; as with the WT protein, NMR spectra
indicated that addition of peptide to the P87STOP mutant converted the
equilibrium mixture of monomer and dimer to all dimer, the data
indicating a minimum 100-fold increase in dimerization constant in the
presence of ligand (Fig. 4).
Because these results did not provide clear evidence of a defect that
might contribute to diabetes insipidus (see "Discussion"), and
because the conditions used to fold the protein during preparation are
not sensitive to minor folding defects (see "Materials and Methods"), we also compared the rate of folding of the completely reduced protein with that of the WT protein under identical conditions, using procedures described earlier for studies of folding kinetics (16). The rate of folding of the P87STOP mutant was within experimental error of that of the WT protein (Fig. 8),
arguing that deletion of the carboxyl terminus does not negatively
impact on folding mechanism. It is further relevant that the first
order folding rate constant of both proteins (7 ± 1 × 10
3/min) was almost three times the value of 2.4 × 103/min obtained for WT bovine NP-II under identical
conditions (16); this difference paralleled the greater stability of
the folded state of BNP-I than of BNP-II relative to their
guanidine-denatured states (Table I) or to their disulfide-mispaired
states (20).3 The similarity
of folding rates and of stabilities to guanidine denaturation (Table I)
of BNP-I and the P87STOP mutant therefore also strongly suggest similar
stabilities relative to their disulfide-mispaired states and the lack
of importance of the 87-93 region to this property.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 8.
First order refolding plots of WT BNP-I ()
and the P87 STOP mutant ( ) from the reduced state. Rates were
monitored as described previously (16) at pH 7.3 in a redox buffer
containing 2 mM GSH, 2 mM GSSG, and 10 mM Phe-Tyr-NH2. The calculated first order rate
constants are 6.8 × 103 and 6.6 × 103/min for the P87STOP and WT proteins,
respectively.
|
|
We also enzymatically prepared a derivative of BNP-II lacking residues
1-6 and 87-95 (see "Materials and Methods"), and we compared its
properties with that of native BNP-II and with des-1-6 BNP-II. In this
case, the peptide binding constant of the des-1-6,des-87-95 mutant
was ~4× that of the WT protein at the same protein concentration (Table III). The binding constant of its parent des-1-6 protein could
not be determined under the same conditions because of the insolubility
of its complexes, but was only twice that of the WT protein at lower
protein concentrations (Table III); thus half of the increase in
binding by the des-1-6,des-87-95 protein appears to derive from
deletion of residues 87-95. The dimerization constant of the
des-1-6,des-87-95 protein was trivially higher than that of native
BNP-II but was only 1/4 that of its parent des-1-6 protein (Table II).
As with the P87STOP mutation, deletion of the carboxyl terminus
therefore negatively impacted on dimerization but increased binding
affinity. No clearly significant differences in stability between
native BNP-II and either the des-1-6 protein or the
des-1-6,des-87-95 protein were observed (Table I).
| |
DISCUSSION |
Implications for the Relationship between Neurophysin Structure and
Properties--
The sensitivity of folding to alterations in the
Gly-17 turn and the loss of structure upon cleavage of this turn in the
native folded protein are particularly striking. Modeling suggests that the destabilization associated with the G17V mutation arises
principally from the energy barrier involved in fitting Val-17 
and
angles (32) to the Gly-17 values (approximately +60 and +26 in the unbound state, respectively), with potential additional contributions from Val hydrophobicity. The instability arising from expansion of the
turn by the AG insert probably in part involves other mechanisms, since
such expansion should reduce the and angle constraints needed
to correctly position the loop ends and because Ala hydrophobicity is a
lesser issue; note that the side chain of Lys-18, which immediately follows the insert, is not involved in any critical interactions in the
folded structure. The Lys-C protease results are relevant here.
Assuming that cleavage of the turn by Lys-C is not accompanied by
disulfide rearrangement (see "Materials and Methods"), the conformational rearrangement associated with turn cleavage signifies that tethering of the ends of the -strands is critical to -strand docking in this case, perhaps because of the short length (3 residues) of one of the connected strands. It follows that the length and composition of the turn will play a central role in sheet alignment, in
this case the increased flexibility associated with the AG insert
reducing tethering efficiency.
The critical requirements for -sheet docking in this system contrast
with the lack of sensitivity of folding to changes in the 3,10 helix. Previous studies (26) demonstrated that this helix could be
cleaved in folded NP without significant effect on either conformation
or dimerization. The present results indicate that altering the helix
by deletion of Glu-47 does not interfere with folding from the reduced
state, except to the extent that the additional stabilization of the
folded state induced by peptide binding and required for efficient
folding is lost. The folding efficiency of the E47 mutant,
approximately half that of BNP-I in the presence of peptide, is
consistent with what would be expected for BNP-I in the absence of
peptide, given that the folded and disulfide-mispaired states of BNP-I
are of approximately equal stability (cf. Refs. 16 and 20).
Additionally, the structure of the folded state indicates a high
content of -sheet and at most only minor differences from unliganded
BNP-I in dimerization; the results in general indicate a high degree of
independence of global protein conformation from that of the helix.
We did not evaluate dimerization in the more severely
folding-compromised mutants, but clear reductions in dimerization were evident in the G57S, G57R, and 87STOP mutants. The reductions in
dimerization in the unliganded G57S and G57R mutants implicate a role
for loop structure in the relative orientation of the two NP domains,
both of which participate in dimerization (6, 7). This is similarly
suggested by the effects on dimerization of succinylation of Ser-56.
However, mutation of Gly-57 has so far not reduced ligand-facilitated
dimerization, whereas succinylation of Ser-56 appears to block this
process (25), suggesting that the Ser-56 modification has a greater
effect than the Gly-57 mutations on dimerization in the liganded state.
In the case of the 87STOP mutant, the small decrease in dimerization
constant is accompanied by an increased binding constant, and these
effects are also seen upon deletion of residues 87-95 in BNP-II. The
deleted region has not been seen in any BNP-II crystal structure (6,
7), but the structure of the unliganded state2 places
residue 86 within 8 Å of interface residues of the neighboring subunit, allowing the possibility that residues carboxyl-terminal to
residue 86 might be in still closer proximity. Thus, the decrease in
dimerization constant upon deletion of the carboxyl terminus, which
represents ~0.4 kcal/mol in BNP-I and 0.8 kcal/mol in the case of the
des-1-6 BNP-II, might reflect ancillary stabilizing interactions by
carboxyl-terminal residues across the subunit interface. It is relevant
that solution studies (e.g. Refs. 27 and 33) and crystal
structure comparisons of liganded and unliganded states2
indicate small ligand-induced changes in the subunit interface. Since
the increase in binding constant upon deletion of the carboxyl terminus
is similar energetically to the decrease in dimerization constant, the
binding constant increase may reflect a decreased uphill energy
requirement for ligand-induced interface alteration when
carboxyl-terminal interactions are lost.
The mutation that had the most negative impact on folding was, not
surprisingly, that of terminating the protein at residue 66. This
mutation deletes almost all of the carboxyl-terminal domain and
necessarily leaves one Cys residue without a partner. We attribute the
lack of folding both to the odd number of cysteines, which increases
the opportunity for disulfide mispairing in this metastable
disulfide-rich protein, and to the almost complete loss of the carboxyl
domain, the noncovalent interactions of which with the amino domain
appear to be necessary for stability. Thus, in preliminary studies, we
have found that mutation of Cys-85 to Phe, which retains the carboxyl
domain but leaves an unpaired Cys residue, and mutation of the codon
for Cys-61 to a STOP, which completely deletes the carboxyl domain
without leaving an unpaired Cys residue, are each associated with
complete loss of folding. The C61STOP mutation is also associated with
diabetes insipidus (9), providing further evidence that the amino
domain is too unstable to fold efficiently without stabilization by the
carboxyl domain.
Implications for the Etiology of Diabetes Insipidus--
The
intracellular accumulation of mutant vasopressin precursor in FNDI
might in principle result from a failure to fold and consequently a
failure to target to regulated neurosecretory granules, or a failure of
properly folded protein to target because of a loss of some sequence or
other attribute (such as the necessary self-association properties)
required for targeting. In the present studies, all mutants except
P87STOP exhibited defects that would, to varying extents, reduce
folding efficiency. Additionally, the greater intrinsic stability of
the folded state of bovine OxyNP than of bovine
VPNP and the stronger homology of human VPNP to
bovine VPNP than to bovine OxyNP (Fig. 1)
suggest that these mutations have the potential to impact folding of
the human vasopressin precursor more than seen here for bovine
OxyNP and its precursor.
The relative impact of the G17V, E47, G57S, and C67STOP mutations on
the trafficking and processing of the human vasopressin precursor in
cell culture generally parallels the relative magnitude of their
effects here on the properties of mature NP and their projected impact
on the folding of the corresponding precursors. In order of their
inhibitory effects on folding, our data indicate C67STOP > G17V > E47 > G57S. This agrees with studies by Olias et al. (34) and Nijenhuis et al. (11) which
indicated that the G17V mutant does not exit the ER (34) but that small
quantities of G57S and, to a lesser extent, E47 do exit the ER. It
also agrees with studies by Ito et al. (35) who reported
adverse effects of mutations on the secretion of the corresponding
precursors in the order C67STOP > E47 > G57S (35).
Although G57S appears to adversely affect cell viability more than
E47 (10), the agreement between the biological and physical-chemical
data is generally consistent with a model in which only the correctly folded human vasopressin precursor, with the ability to form normal dimers, has a significant ability to exit the ER. The finding that a
fraction of the E47 mutant exits the ER additionally allows the
possibility that the increase in dimerization constant associated with
ligand binding may not be required for efficient trafficking of the
fraction that folds, although further studies of potential hormone-NP
interactions within this precursor, and their impact on both folding
and dimerization, are clearly needed. Note that some mutants have been
shown to associate via disulfide formation, as opposed to the normal
non-covalent mechanism, yielding dimers and higher oligomers that do
not exit the ER (36). Such covalent self-association may account for
the apparent ability of the C67STOP mutant to dimerize, suggested by
cross-linking (35), since both our failure to find significant folding
of this mutant and the absence of half of its interface argue that it
is unlikely to dimerize by normal mechanisms.
Beyond this qualitative agreement between folding thermodynamics and
pathogenicity for the above mutations, however, unanswered questions
remain. Consider the G57S mutation. We find that this destabilizes the
bovine oxytocin precursor by only ~0.6 kcal/mol, while cell culture
studies indicate that secretion of arginine vasopressin
immunoreactivity into cell culture media from precursors with this
mutation is impaired by ~90% (11, 35). By using methods previously
described to estimate precursor folding efficiency (20), a 0.6 kcal/mol
destabilization is predicted to only reduce the theoretical folding
efficiency of the bovine oxytocin precursor from 98 to 94% under ER
conditions. Even allowing for a lower stability of the WT human
vasopressin precursor than of the bovine oxytocin precursor, there is
no mechanism based on folding thermodynamics alone by which a
mutation-induced 0.6 kcal/mol further decrease in stability would
reduce folding by 90%. One potential explanation lies in the response
of the ER to the presence of misfolded protein (37), which may amplify
the effect of a relatively small degree of misfolding, e.g.
the presence of misfolded mutant precursor in the ER has been shown to
impair the secretion of folded protein (35). Alternatively, the
mutation may have a greater energetic impact on the human vasopressin
precursor than on the bovine oxytocin precursor. We particularly note
the possibility that direct or indirect interactions of the G57S
mutation with copeptin, or with other elements of sequence of the human
vasopressin precursor not found in the bovine oxytocin precursor, might
be more destabilizing to the human protein.
The principal unanswered questions concern the 87STOP mutation. This is
the sole mutation for which we find no clear parallel in a change in
the physical properties of bovine NP. Recent cell culture studies (38)
indicate that incorporation of this mutation into the human vasopressin
precursor leads to a large fraction of it accumulating in the ER,
suggesting a folding defect; the remainder appears to be normally
handled. However, we find no evidence of a folding defect in the mature
mutant, and no change is seen in other physical properties that might
significantly impact on precursor stability. The small reduction in
dimerization constant in the unliganded state that accompanies loss of
the carboxyl terminus is accompanied by a compensating increase in binding affinity and the normal ligand-facilitated increase in dimerization. The stability of the mutant precursor should therefore be
similar to that of the WT precursor unless factors unique to the human
protein are involved. This possibility merits exploration. For example,
modeling of the sequence differences between human and bovine
VPNPs into the structure of BNP-II provocatively suggests
very close proximity between the Phe substitution at position 70 of
human NP (Fig. 1) and Arg-86.
Is it possible that the defect in the 87STOP mutant lies in a targeting
defect that does not involve a folding defect? We tentatively discount
decreased dimerization as causative in this case, since the high
dimerization constant of the liganded state might provide essentially
complete dimerization at ER protein concentrations. Similarly, deletion
of the carboxyl-terminal glycopeptide has been reported to facilitate,
not hinder, targeting (4). However, the potential role in targeting of
human VPNP residues 87-93 (Fig. 1) remains to be
evaluated. Additionally, in view of evidence for a role of protein
insolubility in targeting to regulated secretory granules (14, 39) and
the increased solubility of complexes of des-1-6 BNP-II upon cleavage
of its carboxyl terminus (see "Results"), a role for the carboxyl
terminus in higher order aggregation processes that might contribute to targeting is not implausible. Physical-chemical studies of the human
vasopressin precursor and of its mutants are clearly needed to resolve
these and other ambiguities noted here and, in doing so, have the
potential to increase understanding of the complex factors (14)
involved in targeting to the regulated secretory pathway.
,
TOP
ABSTRACT
INTRODUCTION
