(Received for publication, December 11, 1996)
From the Department of Developmental Neuroscience, Biomedical Centre, Uppsala University, Uppsala S-751 23, Sweden, and the § Montreal Neurological Institute, McGill University, Montreal PQ H3A 2B4, Canada
Nerve growth factor (NGF) and neurotrophin-3
(NT-3) mediate activities such as survival, differentiation, and
proliferation in various subsets of neurons. In this report, we define
precisely the residues in human NGF responsible for NGF biological
activity and binding specificity to the neurotrophin receptor TrkA. In earlier studies we defined five amino acid residues of NGF which confer
NGF-like activity to NT-3 when replacing corresponding residues in the
120-amino acid long NT-3 molecule. Using this gain-of-function strategy
we report the further dissection of this functional epitope. We also
define another motif separated topographically in the NGF dimer and
determined to be independently responsible for NGF specificity. The
first of the two motifs determined to elicit NGF specificity is defined
by the residues Val-48, Pro-49, and Gln-96, which are situated in the
two top -loops of NGF. The second motif is represented by residues
Pro-5 and Phe-7 situated in the proximal part of the
NH2 terminus. Both motifs contain structurally
important residues revealing a novel principle, where specificity for
neurotrophin ligand-receptor interactions could be determined by
variable residues modifying the conformation of the neurotrophin
backbone. These findings will enhance further the possibility of
mimicking NGF with low molecular weight compounds.
Nerve growth factor (NGF),1
brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3), and
neurotrophin-4 (NT-4; NT-5 is the mammalian homolog to NT-4) are
members of a family comprising structurally related proteins termed
neurotrophins. These proteins promote growth and survival of neurons in
the nervous system (for review, see Ref. 1). Each of these proteins
binds to at least two membrane receptors. The p75 neurotrophin
receptor, which binds all members of the neurotrophin family, enhances
the NGF binding affinity and ligand binding specificity together with
the TrkA receptor. Moreover, p75 neurotrophin receptor has the ability to signal through the ceramide pathway in response to NGF (2), activate
nuclear factor-B, and mediate apoptosis in neural cells (3, 4). The
other known receptors for neurotrophins are tyrosine kinases that show
different specificities for binding to different neurotrophins; the
TrkA receptor preferentially binds NGF, the related TrkB receptor binds
BDNF and NT-4, and TrkC binds NT-3. Although data on the cross-binding
of NT-3 to all Trk receptors has been published, these studies have
been made using cell lines expressing ectopic receptors and with
neurotrophin concentrations above those expected to be present in
vivo. At physiological concentrations of NT-3, the preferred
receptor is TrkC (5).
The roles of NGF and NT-3 and their receptors in neural development have been studied extensively by gene targeting in mice (6-10). Data from these and many other types of studies have established that these neurotrophins are crucial for the development of the peripheral nervous system; their roles in the central nervous system are less well defined. In the peripheral nervous system NGF is a survival factor for sympathetic neurons and small sensory neurons associated with nociception. In the central nervous system, NGF mediates the differentiation and survival of basal forebrain cholinergic neurons, a population that deteriorates in Alzheimer's disease (9, 11, 12). NT-3 has been shown to be a survival factor for the proprioceptive neurons in the peripheral nervous system (7, 13), facial motor neurons, and noradrenergic neurons of the locus coeruleus (14). Thus, NGF and NT-3 have major roles in the development of the peripheral nervous system as well as several defined effects on populations of neurons in the central nervous system.
This family of closely related polypeptides has many appealing features
that make them suitable for mutational analysis. The neurotrophins are
small (usually around 120 amino acids) structurally related molecules
with clusters of conserved or variable residues. The three-dimensional
structure has been determined for NGF (15, 16) and a BDNF·NT-3
heterodimer (17), which revealed a flat molecule organized as a
symmetrical dimer with protruding -loops. The examination of
structural determinants for specificity among the neurotrophins can
provide insight into how ligand-receptor interactions have evolved in
the complex nervous system. In addition, precise knowledge of the
residues in the neurotrophins which bind their receptors will provide
clues for identification of low molecular neurotrophic antagonists and
agonists that may have therapeutic potentials for the treatment of
neurodegenerative diseases such as Alzheimer's and Parkinson's or in
stimulating or blocking neurotrophin actions in peripheral neuropathies
and pain conditions.
Several approaches have been used to elucidate which sequences of the
neurotrophins are required to specifically bind and activate their
cognate receptors (18). A first approach used chemical modification of
NGF where residues receptive for different agents could be blocked
and the effect analyzed (for review, see Ref. 19). Enzymatic cleavage
of NGF has identified the NH2 terminus to be critical for
binding to TrkA (20-22). Establishment of heterodimers between
neurotrophins has brought further insight regarding specificities, where a BDNF·NT-3 heterodimer was 10-fold less biologically active than a mixture of BDNF and NT-3 homodimers (23). Deletions of the NGF
coding sequence by site-directed mutagenesis have identified the
NH2 and COOH termini as necessary for binding to TrkA (24, 25). Furthermore, by exchanging blocks of amino acid residues among the
neurotrophins, it has been established that the NH2 terminus and the variable -loops are involved in neurotrophin specificity (26-28). When blocks of residues from BDNF variable regions I, III, and V were substituted into the corresponding segments
of NGF, there was no loss of NGF activity, implying that the remaining
-loops II and V or the NH2 and COOH termini, are involved in NGF specificity (26). Thus, the understanding of neurotrophin specificity and neurotrophin binding is emerging from
several studies, although thus far most studies have focused upon NGF.
Using a gain-of-function strategy, we have shown previously that
exchanges of NGF amino acids in NT-3 allowed NT-3 to act with a
NGF-like activity (28), thus defining a motif involved in NGF
specificity situated in the two top
-loops. Here we report the
further identification of the amino acids necessary for determining NGF
specificity in the two top
-loops as well as in the NH2
terminus, as assessed by biological effects on NGF-responsive neurons
and activation of TrkA.
Human NGF and NT-3 were
cloned by polymerase chain reaction using a cDNA library
constructed from primary astrocyte-like cells derived from a 9-week-old
human fetal brain (clinical abortion). For a thorough description of
the principle for this strategy of mutagenesis, see Ref. 28. Briefly,
single mutants were produced first, and a secondary polymerase chain
reaction using the single mutants as template produced the
combinatorial mutants shown in Fig. 1A. In
this work mutants were given their names according to the NGF amino
acids exchanged into the NT-3 backbone. VF-QAA and parts thereof were
amino acids in NGF (see Fig. 1A) residue positions 48, 49, and 96-98. N-term and parts thereof designate the amino terminus of
NGF corresponding to the first seven amino acids in the mature protein
SSSHPIF. For clarity, residues or position numbers in Roman type refer
to human mature NGF, and residues or position numbers in
italics refer to human mature NT-3. As seen in Fig. 1,
residue position numbers in NGF and NT-3 do not match because of gaps
inserted in the alignment for best fit of the protein sequences. The
fragments were inserted by direct cloning into the pBSKS+ cloning
vector (Stratagene), cleaved with EcoRV, and subsequently
incubated with Taq polymerase and 2 mM dTTP at
72 °C for 2 h. After sequencing the fragments were subcloned to
the expression vector pXM (Genetics Institute), and 30 mg was electroporated (Porator, InVitrogen) into 1 × 106 COS
cells in 0.6 ml of phosphate-buffered saline at 400 V and 250 microfarads at 22 °C. After 1 day of incubation in DMEM with 10%
fetal calf serum (FCS) the medium was changed to DMEM + 0.5% FCS for
harvest after another 3 days. Parallel porations were metabolically
labeled by incubation for 5 h in 200 ml of cysteine/methionine free DMEM supplemented with [35S]cysteine/methionine. 50 ml of this labeled medium was boiled in SDS sample buffer and run on a
15% SDS-polyacrylamide gel. To equalize the concentrations of the
produced proteins in the conditioned media, the gels were dried down
and exposed using a PhosphorImager, and the resulting bands were
quantified. The medium giving rise to the NGF band was measured in an
enzyme immunoassay for NGF (29) to obtain an absolute concentration
value corresponding to the density of the NGF band.
Fiber Outgrowth and Survival Assay
Sympathetic and Remak's ganglia were dissected from E9 chicken and embedded in a gel of native collagen fibrils (30). The cultures were incubated for 2 days under standard tissue culture conditions (Eagle's basal medium, 1% FCS) together with neurotrophins and then examined in dark-field illumination through an inverted microscope. The fiber outgrowth activity was determined on a scale from 0 to 1 (increments of 0.1 unit), where 1 is a full outgrowth with a dense halo of fibers originating from the ganglia (30). To check for survival effects, sympathetic ganglia dissected from E9 chicken were dissociated in 0.25% trypsin and plated in a thin collagen gel for incubation for 2 days. Neurotrophins or mutants were added immediately at different concentrations ranging from 0.5 to 100 ng/ml, and surviving neurons were counted in a phase-contrast microscope.
Binding StudiesRat pheochromocytoma PC12 cells were grown
in DMEM supplemented with 10% horse serum and 5% FCS and resuspended
in DMEM + 0.5% FCS to a concentration of 1 × 107
cells/ml. 3T3-TrkA cells were harvested with 1% EDTA in
phosphate-buffered saline and also used at a concentration of 1 × 107 cells/ml. Incubation was performed for 1 h with
agitation at 4 °C with 100 ml of cell suspension mixed with 100 ml
of medium containing different concentrations of purified mutant or
wild type protein. After blocking, iodinated mouse 2.5 S NGF (Amersham Corp., 1,500 Ci/mmol) was added to a final concentration of 4 × 1011 M for PC12 cells and 5 × 10
10 M for 3T3-TrkA cells and incubated for
another h. To separate unbound NGF from bound NGF the cells were
centrifuged through a sucrose gradient consisting of a 0.3 M sucrose bottom layer with 0.15 M sucrose
overlay in a microcentrifuge at 13,000 rpm for 10 min (31). The tubes
were frozen in liquid nitrogen, and the bottoms with the cell pellet
were cut off and counted in a LKB 1275 mini-
counter. To measure the
unspecific binding a 100-fold excess of cold mouse
-NGF was added
and the obtained value (never more than 10% of signal) subtracted from
the measurements. Experiments were repeated three or four times, and
the ratio of binding was calculated as cpm bound/(cpm bound + cpm
free).
3T3-TrkA expressing cells at a density of 2 × 107 cells/ml were stimulated with conditioned media or purified protein (100 ng/ml) for 5 min at +37 °C, washed with TBS buffer (20 mM Tris, pH 7.5, 137 mM NaCl), and then lysed in Nonidet P-40 buffer (20 mM Tris, pH 7.5, 137 mM NaCl, 500 mM Na3VO4, 10% glycerol, 1% Nonidet P-40, 1 µg/ml leupeptin, 10 µg/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride). The cells were centrifuged to pellet cell debris and incubated with an anti-Trk (C-14, Santa Cruz Biotechnology) antibody at a concentration of 1 µg/ml for 1 h. Protein A-Sepharose beads (Pharmacia Biotech Inc.) were added and the mixture incubated for an additional 30 min. All procedures were done at 4 °C. The beads were centrifuged and washed three times in Nonidet P-40 buffer and once in distilled water. The samples were subsequently boiled in SDS-gel loading buffer, run on a 7.5% acrylamide gel, transferred to nitrocellulose. and detected with 4G10 anti-phosphotyrosine antibody (Upstate Biotechnology). A secondary antibody conjugated with horseradish peroxidase was used together with chemiluminescence to detect phosphorylated tyrosine residues.
PC12 Cell Differentiation AssayPC12 cells or PC12 6-24 cells were grown in DMEM supplemented with 5% FCS and 10% horse serum. Neurotrophins or mutants were added to the cultures, and after 6 days (PC12) or 2 days (PC12 6-24) the ratio of cells with fibers longer than two cell diameters was counted. Three different plates were counted for each concentration, and at least 200 cells were analyzed in each plate.
Purification of Mutated NeurotrophinsConstructs with NT-3,
the P-F mutant, or the VF-QAA mutant were electroporated into
dihydrofolate reductase negative (dhfr) Chinese hamster
ovary cells. After selection in growth medium without nucleoside
analogs supplemented with 10% dialyzed FCS, positive clones were
tested for their ability to produce neurotrophin mutants with neurite
outgrowth activity on Remak's ganglia. Such clones were amplified for
higher expression levels of desired protein by using successively
higher concentrations of methotrexate in the growth media (32). Chinese
hamster ovary cells that after amplification produced ~1,000 ng/ml of
desired neurotrophin were grown in triple layered culture flasks, and 2 liters of medium was collected for further purification. Serum proteins
in the medium were precipitated away using 15% saturation of ammonium sulfate. Remaining proteins in the supernatant were then precipitated by increasing the ammonium sulfate concentration to 80% of saturation. The protein pellet was resuspended and dialyzed against 50 mM Tris-HCl, pH 9.0. The sample was loaded on an
equilibrated (50 mM Tris-Cl, pH 9.0) CM-Sepharose ion
exchange column (Pharmacia). The column was washed with 3 column
volumes of equilibration buffer before being eluted with 0.4 M NaCl in 50 mM Tris-HCl, pH 9.0. Fractions of
5 ml were collected of the eluate, tested for biological activity in
the ganglion assay, and analyzed on a 10-20% SDS-polyacrylamide gradient gel. Typically this process yielded about 1.5 mg of active neurotrophin.
By using polymerase chain reaction-based in vitro mutagenesis we have produced various NT-3 mutants containing different exchanges of NGF residues (Fig. 1). Mutants are produced in COS cells and metabolically labeled conditioned media are used to verify that the mutants are expressed at similar levels, ranging from 100 to 300 ng/ml. All mutants presented here (Table I) were found to be expressed within this range, and they were biologically active and capable of phosphorylating the TrkC receptor. As shown, the specific NGF-like activity varied considerably depending on which residues were introduced into the NT-3 backbone.
|
To analyze the function of the NH2 terminus of NGF, we
produced a NT-3 mutant with the six NH2-terminal residues
replaced with the corresponding seven NGF NH2-terminal
residues (protein sequences are shown in Fig. 1A).
Confirming previous reports (20-22, 25, 33), the
NH2-terminal mutant showed robust NGF-like responses (Fig.
1, B and C), as judged by both tyrosine
phosphorylation of the TrkA receptor in 3T3 cells (Fig. 1B)
and fiber outgrowth from chick E9 sympathetic ganglia (Fig.
1C). This prompted us to dissect the NH2
terminus further to establish the individual residues responsible for
this effect. Two mutants were produced in which the NH2
terminus of NT-3 was replaced with either the three most distal
NH2-terminal residues (SSS replacing YAE;
residues or position numbers in Roman type refer to amino acids in
human mature NGF, and residues or position numbers in
italics refer to amino acids in human mature NT-3) or the
three proximal residues following the shared His-4 (PIF replacing
KS, Fig. 1, B and C). The fiber
outgrowth response and tyrosine phosphorylation of the TrkA receptor
(Fig. 1, B and C) indicated that NGF-like
activity was achieved only when the PIF motif in NGF replaced the
KS motif in NT-3. Further analysis of the PIF motif, where
Pro-5 replaced Lys-5 in one mutant and Ile-6 and Phe-7
replaced Ser-6 in another mutant, revealed that these
replacements did not allow NT-3 to achieve NGF-like activity (Fig. 1,
B and C). However, a NT-3 mutant where Pro-5
replaced Lys-5 and Phe-7 was inserted after
Ser-6, was produced, and this mutant displayed substantial
NGF-like activity (Fig. 1, B and C). This mutant
also promoted survival in dissociated sympathetic neurons as seen in
Fig. 2. Thus, Pro-5 and Phe-7 are two residues important
for NGF specificity.
The motif VF-QAA (Fig. 1) was established earlier as a potent activator
of NGF-like activity when these residues were inserted in the
corresponding positions in the NT-3 (28). To establish further which of
these five amino acids are involved in the NGF-like response, a set of
five mutants was produced: V-QAA, F-QAA, VF-Q, VF-A(97), and VF-A(98).
Our earlier results showed that the VF or QAA motif by itself did not
achieve NGF-like responses (Fig. 1, D and E). Two
NT-3 mutants were generated to evaluate the VF motif, V(48)-QAA and
F(49)-QAA. These mutants did not exhibit NGF-like activity assessed by
tyrosine phosphorylation of TrkA (Fig. 1D) and the lack of
induced fiber outgrowth from chick E9 sympathetic ganglia (Fig.
1C). This suggests that both Val-48 and Phe-49 in the VF-QAA
motif are residues necessary to achieve NGF-like activity. Phe-49 is
placed like a shovel in the groove formed by the top -loop (II) and
the proximal
-loop from the other protomer (V) and could thereby be
available for a hydrophobic interaction with the TrkA receptor (see
Fig. 6, A-D). In contrast, three mutants in the QAA motif,
VF-Q(96), VF-A(97), and VF-A(98), all yielded NGF-like activity (Fig.
1, D and E), suggesting that the residues Gln-96,
Ala-97, and Ala-98 each independently contributes to the NGF-like
activity seen in the NT-3 mutants when combined with the VF motif. To
establish further whether the interactions from these residues in the
two top
-loops determine specificity by attraction or repulsion,
alanines were instead exchanged at the corresponding place in the NT-3
backbone. A mutant with the NGF residues VF replacing PV
(residues 48 and 49) in combination with alanines exchanging
LVG (residues 96, 97, and 98) was analyzed as well as a
mutant where alanines replaced KTG-PV-LVG (residues 42-44,
47-48, 96-98). Interestingly, none of these mutants showed any
significant NGF-like activity, although they all displayed normal NT-3
like activity (Table I). These results suggest that Gln-96 is actively
contributing to NGF specificity, and its presence overrides
Val-97 and Gly-98. In turn, Val-97 and
Gly-98 may suppress NGF-like activity of NT-3 by weak
repulsive actions directed toward TrkA. This is shown by the results
where replacing one of these residues with alanine results in an
NGF-like activity (Fig. 1, B and D). Also, we
have used a BDNF triple mutant, B3 (NGF residues NIN 43-45, VF 48-49,
and QAA 96-98 replacing corresponding BDNF residues) to analyze the
VF-QAA motif when put in a neurotrophin backbone other than NT-3. The
B3 mutant is somewhat more efficient in promoting survival of
sympathetic neurons than wild type BDNF (Fig. 2) but not as efficient
as comparable mutants using the NT-3 backbone as a framework (Table I),
indicating differences among the neurotrophins as to their specific
receptor determinants.
To analyze further the motifs identified above to be important in
determining NGF specificity we purified the VF-QAA mutant, the P-F
mutant, and wild type NT-3. Transfected Chinese hamster ovary cells
that expressed the desired mutants were selected, and the expression
was amplified to a yield a production of the neurotrophin mutants of
~1,000 ng/ml. The conditioned media were subjected to fractionated
precipitation with ammonium sulfate and further fractionated by ion
exchange chromatography (Fig. 3A). Following
this, the neurotrophin mutants were obtained at a concentration of 300 µg/ml and appeared as a single band with an apparent molecular mass
of 14 kDa on a 10-20% gradient gel (Fig. 3B). The purified
mutants were tested again for biological activity to confirm data from
transient transfections. The analysis showed that the
NH2-terminal mutant had a higher NGF-like activity than the
VF-QAA mutant and that the purified mutants displayed the expected
biological activities (Table I and Figs. 4 and
5). Examples of survival and fiber outgrowth assays
using purified proteins are shown in Fig. 4, A and
B. In survival assays using sympathetic neurons the effects
from three different protein concentrations (20, 5, and 2 ng/ml) were
analyzed and compared with the effects of equivalent concentrations of
NGF (Fig. 4). At 20 ng/ml there were no differences between the mutants
and NGF (Fig. 4C). Also, when testing lower concentrations,
the potency of the mutants compared with NGF was approximately
60-80%, with the P-F mutant being somewhat more potent than the
VF-QAA mutant. Similar results were obtained using a PC12
differentiation assay (Fig. 5, A and B), with the
neurotrophins assayed at a concentration of 20 ng/ml.
Finally, we examined the abilities of the neurotrophin mutants to induce differentiation in PC12 cells or PC12 6-24 cells that overexpress human TrkA 20-fold compared with endogenous rat TrkA (34). As shown in Fig. 5, A and B, the relative potencies of the mutants are comparable in both cell types, although the responses to the mutants were approximately 15% higher in the 6-24 cells. Furthermore, we confirmed that the purified mutants activated the TrkA receptor, as assessed in 3T3-TrkA cells (Fig. 5C). Competitive binding assays revealed differences between the two mutants when comparing displacement of iodinated NGF from either PC12 or 3T3-TrkA cells (Fig. 5, C and D). The P-F mutant was 10 times less efficient in displacing 50% of labeled NGF than NGF itself, whereas the VF-QAA mutant was 30 times less efficient. However, on TrkA receptors in 3T3 cells (not expressing p75), the P-F mutant was only five times less efficient and the VF-QAA mutant 10 times less efficient. These differences result from the presence of p75 neurotrophin receptor on PC12 cells and not on 3T3-TrkA cells. In summary, the P-F mutant is somewhat more NGF-like than the VF-QAA mutant when comparing biological activity, TrkA receptor activation, and NGF receptor binding.
In this study, NT-3 served as a framework in which potential residues defining NGF specificity replaced the corresponding NT-3 residues (Fig. 1A). The assay for neurotrophin specificity is based on two experimental systems, one biological and one biochemical. First, we use a robust biological assay measuring fiber outgrowth from explanted embryonic chicken ganglia (30). For the analysis of NGF fiber outgrowth activity we use the sympathetic ganglia that express TrkA mRNA (35) and only weakly respond to NT-3 (36, 37). To measure NT-3 fiber outgrowth activity we use Remak's ganglia, which do not respond to NGF and only express TrkC mRNA (38). Second, we examined ligand-induced receptor tyrosine phosphorylation using 3T3-TrkA fibroblasts expressing TrkA but not TrkC and 3T3-TrkC fibroblasts that express TrkC but not TrkA. Hence the two systems for analyzing mutants are dual, i.e. they analyze in comparative fashion the NGF and NT-3 activity in primary neurons and in cell lines. The ratios are less dependent on the concentration used and therefore a more reliable way to estimate the potential NGF-like activity of the mutants.
We establish that replacement of the NH2 terminus of NT-3
with corresponding residues of NGF confers NGF-like activity
independent of our previously defined VF-QAA motif in variable regions
II and V of neurotrophins (28). A further dissection of the
NH2 terminus showed that the inner part of this region
determines NGF specificity. More precisely, when Lys-5 was
exchanged for Pro-5 together with an insertion of Phe-7 in the
corresponding place in NT-3 (between Ser-6 and
His-7, see Fig. 1A) this NT-3 molecule was nearly
as potent as NGF (Figs. 1, B and C, 4, and 5).
When NGF NH2-terminal residues 3-9 were replaced by the
corresponding BDNF residues, different results were obtained: one group
reported only 5-fold decreased binding on TrkA receptors (33), whereas another group reported a 2-4-fold effect on bioactivity but no decrease in binding to PC12 cells (39). These results are interesting since the BDNF residues include a proline and also lack a residue where
NGF has Phe-7. It is thus tempting to argue that these amino acids
account for the differences seen when using BDNF and NGF rather than
NGF and NT-3 for the homolog scanning. The NH2 and COOH
termini are not resolved in the available crystal structures (Fig.
6) of the neurotrophins (15, 17). Recently, the
bioactive conformation of the NGF amino and carboxyl termini has been
predicted by molecular modeling (41). These studies demonstrate that
the conserved residues His-4, Pro-5, His-8, Glu-11, and Arg-118 cause a split of the amino-carboxyl termini complex into two distinct moieties: a ring-like rigid region (residues 9-11 of one protomer, and
residues 112
-118
of the second protomer), and a flexible loop
(residues 1-8). Although only the rigid region seems to be necessary
for TrkA-mediated activity, the flexible loop might be directly
involved in those NGF-receptor interactions that reinforce geometric
selectivity, primarily because of the exposed
-sheet motif provided
Phe-7 through Arg-9.
Apart from the NH2 terminus, another surface introducing NGF activity to NT-3 is comprised of amino acid residues VF-QAA of variable regions II and V (26, 28). These loop regions are structurally ideal candidates for specific interactions with receptor molecules since they can accommodate changes without affecting the overall structure (42). The combinatorial effect of the amino acid exchanges into NT-3 in the VF-QAA motif is crucial to achieve an NGF-like effect. The exchange of VF alone into the NT-3 framework did not yield NGF-like activity, but when exchanged in combination with the QAA motif this NT-3 mutant was nearly as potent as NGF (Fig. 1, D and E). When considering the three-dimensional model of NGF, the importance of this combination can be interpreted as an interaction between the two protomers. We proposed previously (28) that the VF motif from one protomer forms a patch with the QAA motif from the other protomer rather than with the QAA motif from the same protomer (Fig. 6, A-C). Similar conclusions regarding cooperation between neurotrophin protomers to achieve their specialized function have been reported by investigators using different approaches (17, 43). When replacements of alanines in the LVG (96-98) motif were carried out one at the time in combination with the VF exchange, we saw surprisingly that each of these mutants gave robust NGF-like activity (Fig. 1, D and E). In contrast, a mutant replacing the LVG residues with alanines, again in combination with VF, failed to result in NGF-like activity (Table I). This mutant only differed by Gln-96 compared with the VF-QAA mutant, and this suggests that the Gln-96 is necessary for NGF conversion. For Val-97 and Gly-98 in the QAA motif it must be considered that the effect from the alanine exchanges is of a negative nature because alanines are not likely to have a role in selection toward a receptor. More plausible is that the effect of taking away Val-97 and Gly-98 from NT-3 allows for interaction to the NGF receptor. Interestingly, when region V (containing the QAA motif in NGF) from BDNF was inserted into a NT-3 chimera (43) the interaction with TrkA was reduced, further emphasizing that this loop region contains determinants for specific TrkA binding.
Both motifs found here to determine specificity include residues known
to be critical for conformation of a peptide backbone. When
Pro-47 from the NT-3 wild type protein is replaced by Val-48 the peptide backbone is allowed to move more freely. If simultaneously exchanging Val-48 for Phe-49, a sterical effect to
reposition the loop is likely. Also in -loop V, a residue known to
be a structural determinant (Gly-98) is involved in the
specificity, but this effect is not as radical as in the two other
motifs since the change of the glycine is not necessary for the
conversion to NGF-like activity in the QAA motif (Fig. 1, D
and E). The flexibility of
-loops has been apparent when
analyzing several different crystals of NGF, and the differences in the
positioning of
-loops extended in some cases to 6 Å (16).
Furthermore, in the BDNF·NT-3 crystal,
-loop II resolved with
higher temperature, indicating flexibility, and the superpositioning of
the NT-3 structure over the NGF structure reveals conformational
differences in this
-loop region as well as in
-loop V (17). In
the NH2 terminus a proline and a phenylalanine are
involved in the conversion of specificity. In NT-3 Lys-5,
which is a long positively charged residue, is replaced by Pro-5, but
an insertion of Phe-7 is also necessary to achieve the NGF-like
activation (Figs. 2, 4, and 5). When modeling the amino terminus as a
peptide, the effect on the backbone is a profound bend determined by
the inserted Pro-5, and it is likely that this effect also perturbs the
solvent accessible NH2 terminus of the NT-3 mutant (Fig. 6,
E and F). Considering that the NH2 terminus and the three
-loops are highly flexible in the
three-dimensional structure (16, 17) it is plausible that the
positioning of the NH2 terminus and the
-loops
contributes to the differences in specificity of NGF and NT-3.
Reports regarding the binding areas for neurotrophins on the Trk
receptors have revealed two receptor surfaces as potential interaction
sites. The IgC2-like domains, especially the second IgC2-like domain
closest to the membrane, have been shown to be critical for binding of
neurotrophins (44, 45). In addition, the hinge region between the
IgC2-like domain and the transmembrane domain has been shown to be well
conserved among TrkA for different species but markedly different
compared with TrkB or TrkC (35). Deletions in this region abolish TrkA
binding, indicating a possible role in neurotrophin specificity (46).
Another extracellular TrkA domain, the second leucine repeat motif, has
also been identified to be a major binding domain for neurotrophins
(47, 48). However, these authors did not find evidence for the
IgC2-like domain to be responsible for neurotrophin interaction.
Interestingly, the exposed -sheet motif (Phe-7 through Arg-9) of the
flexible loop of the predicted bioactive conformation of the NGF amino
and carboxyl termini (41) has been suggested to be involved in NGF-TrkA
interactions.2 Phe-7 has thus been
predicted to show a stereochemical fit to a domain of a leucine-rich
motif within TrkA. Together with the findings presented here, clearly
indicating two independent areas conveying NGF specificity, an
intriguing possibility of specificity being determined by two sites on
both the receptor and ligand opens up. Moreover, a third site
representing binding to p75 must be taken into account in such a model
(49-51).
Even though we have defined two sites in purified mutants which are
found to confer NGF-like activity to NT-3 we have not seen the reverse
effect, i.e. that NT-3 loses its activity due to the
mutational alterations. This is not because of the dual independent
actions of the two motifs studied here since we have also analyzed a
mutant with both motifs replaced (Table I). This remarkable effect has
been reported in earlier studies where NGF and BDNF (26) or NT-3 (52)
activities were analyzed in chimeric molecules and indicates that the
different neurotrophins may each depend on different variable residues
for its specificity. This is also compatible with the findings (27)
showing that variable -loop II can determine NGF specificity in the
NT-3 backbone but not in the NT-4 backbone. Also in line with these
results are studies on NT-3 where mutants with alanine exchanges of
Arg-31 and His-33 were shown to determine the
binding site of NT-3 to its nonpreferred receptors TrkA and TrkB. These
NT-3 mutants also remained active on the TrkC receptor (53) as is the
case with the mutants presented in this study.
This study aimed at providing an answer to the question of what
residues determine NGF specificity. It is required to define the areas
responsible for specific receptor interaction when modeling how the
neurotrophins work and to be able to design agonists or antagonists as
potential drugs. The results presented in this work evidently point out
that small areas are enough to determine specificity within the
neurotrophin family. Similar findings have been gained regarding the
growth hormone-receptor complex (54). From the three-dimensional models
(Fig. 6) it can be concluded that the areas defined here are altered in
a manner accounting for the effects seen in this study. Conclusively,
adding to the information regarding specificity of NGF toward its
cognate receptor we have determined Gln-96 in cooperation with
positioning of -loop II through Val-48 and Phe-49 to distinguish NGF
from NT-3. Separated from this site but of equal, if not greater,
importance are residues Pro-5 and Phe-7 situated in the NH2
terminus.
Comparisons of the studies recently published on the erythropoietin
ligand-receptor complex (55, 56) with data currently available for
neurotrophin structure/function opens up exciting possibilities
synthesizing agonists or antagonists for a protein-protein interaction.
A challenging prospect for future studies is to compile available data
for rational drug design, an approach already tried in the neurotrophin
field (57, 58) but with the aid of phage expression cloning. One
problem experienced in these studies was the conformation of -loops;
and strikingly, in the studies of erythropoietin, a peptide sequence
with no resemblance to the original ligand forms a loop region that
interacts with the erythropoietin receptor.
We thank Annika Kylberg and Helena Vretman for excellent technical assistance and Rick Riopelle and Carlos Ibáñez for discussions and critical reading of the manuscript. We are grateful to Kristina Luthman for helping to generate computer models of NGF and NT-3. The Genetics Institute (Boston) provided the pXM expression vector.