Although BMPs are characterized by versatility in receptor binding, referred to as promiscuity, tremendous differences underlie in vivo signaling by either BMP-2⁄ 4 or BMP-6⁄ 7: whereas
Trang 1is dependent on N-glycosylation of the ligand
Stefan Saremba1,2,*, Joachim Nickel1,*, Axel Seher1, Alexander Kotzsch1,2, Walter Sebald1
and Thomas D Mueller1,2
1 Lehrstuhl fu¨r Physiologische Chemie II, Biozentrum der Universita¨t Wu¨rzburg, Germany
2 Lehrstuhl fu¨r Molekulare Pflanzenphysiologie und Biophysik, Julius-von-Sachs Institut der Universita¨t Wu¨rzburg, Germany
Bone morphogenetic protein (BMP)-6, BMP-5, BMP-7
and BMP-8 constitute a subgroup of the transforming
growth factor (TGF)-b superfamily proteins Besides
the ability of BMP-6 to induce bone formation at
ecto-pic and orthotoecto-pic sites, BMP-6 transcripts have been
localized in numerous studies to developing organs
and tissues, such as the heart, the brain, and
hyper-trophic cartilage, throughout the developing skeletal system, and also to adult tissues, such as brain and uterus [1–5] BMP-6 and its closest relative, BMP-7, show overlapping expression patterns as well as over-lapping functions For example, in the developing heart, BMP-6 and BMP-7 are required for cushion for-mation and septation [5] In the brain, BMP-6 and
Keywords
crystal structure; ligand–receptor specificity;
protein–protein interaction; recognition;
transforming growth factor-b superfamily
Correspondence
T D Mueller, Lehrstuhl fu¨r Molekulare
Pflanzenphysiologie und Biophysik,
Julius-von-Sachs Institut der Universita¨t Wu¨rzburg,
Julius-von-Sachs Platz 2,
D-97082 Wu¨rzburg, Germany
Fax: +49 931 888 6158
Tel: +49 931 888 6146
E-mail: mueller@biozentrum.
uni-wuerzburg.de
Website: http://www.bot1.biozentrum.
uni-wuerzburg.de
*These authors contributed equally
Database
The coordinates and structure factors for
the structures of wild-type BMP-6 and
B2-BMP-6 have been deposited with the
Pro-tein Data Bank, entry codes 2R52 and 2R53
(Received 3 September 2007, revised 6
November 2007, accepted 12 November
2007)
doi:10.1111/j.1742-4658.2007.06187.x
Bone morphogenetic proteins (BMPs), together with transforming growth factor (TGF)-b and activins⁄ inhibins, constitute the TGF-b superfamily of ligands This superfamily is formed by more than 30 structurally related secreted proteins The crystal structure of human BMP-6 was determined
to a resolution of 2.1 A˚; the overall structure is similar to that of other TGF-b superfamily ligands, e.g BMP-7 The asymmetric unit contains the full dimeric BMP-6, indicating possible asymmetry between the two mono-meric subunits Indeed, the conformation of several loops differs between both monomers In particular, the prehelix loop, which plays a crucial role
in the type I receptor interactions of BMP-2, adopts two rather different conformations in BMP-6, indicating possible dynamic flexibility of the pre-helix loop in its unbound conformation Flexibility of this loop segment has been discussed as an important feature required for promiscuous bind-ing of different type I receptors to BMPs Further studies investigatbind-ing the interaction of BMP-6 with different ectodomains of type I receptors revealed that N-glycosylation at Asn73 of BMP-6 in the wrist epitope is crucial for recognition by the activin receptor type I In the absence of the carbohydrate moiety, activin receptor type I-mediated signaling of BMP-6
is totally diminished Thus, flexibility within the binding epitope of BMP-6 and an unusual recognition motif, i.e an N-glycosylation motif, possibly play an important role in type I receptor specificity of BMP-6
Abbreviations
ActR, activin receptor; ALP, alkaline phosphatase; BMP, bone morphogenetic protein; BMPR, bone morphogenetic protein receptor; GlcNAc, N-acetylglucosamine; h, human; MPD, 2-methyl-2,4-pentanediol; TGF, transforming growth factor.
Trang 2BMP-7 have important effects at early and late stages
of nervous system development (e.g specification of
nervous system patterning [6] and decision of neuronal
fate [7]) BMP-6-deficient mice are viable and fertile
without displaying overt effects in tissues known to
express BMP-6 mRNA [3], implying also a functional
redundancy among the factors of this subgroup
BMPs exert their biological effects by inducing the
formation of a heteromeric receptor complex from
type II [activin receptor (ActR)-II, ActR-IIB, or BMP
receptor (BMPR)-II] and type I chains (BMPR-IA,
BMPR-IB, and ActR-I) [8–10] The constitutively
active type II kinase then phosphorylates and thereby
activates the type I chain [11], which subsequently
propagates the signal downstream by acting on
BMPR-regulated Smads-1⁄ 5 ⁄ 8 [12] Although BMPs
are characterized by versatility in receptor binding,
referred to as promiscuity, tremendous differences
underlie in vivo signaling by either BMP-2⁄ 4 or
BMP-6⁄ 7: whereas BMP-2 ⁄ 4 exert their function by initial
binding to the high-affinity receptors BMPR-IA or
BMPR-IB, it has been shown that ActR-I is the
pre-dominant type I receptor used by BMP-7 in a variety
of cell lines [13]
The overall fold and dimer architecture seem to be
highly conserved for the ligands of the TGF-b
super-family, when the receptor unbound conformation is
considered Several structures have been determined,
i.e TGF-b1, TGF-b2, TGF-b3, BMP-2, BMP-7, and
growth and differentiation factor (GDF)-5, all showing
the canonical fold Here, we present the crystal
struc-ture of BMP-6 and compare the strucstruc-ture with that of
other BMPs Differences resulting in possible
altera-tions of the type I receptor-binding profile were
inves-tigated Comparison of the receptor binding of
Escherichia coli and CHO-cell derived BMP-6 reveals
the importance of N-glycosylation for ActR-I binding
and activation For proteins other than TGF-b ligands
(e.g hormones), it has been shown that glycosylation
can ameliorate the receptor binding [14]; however, this
is the first report in which glycosylation of a TGF-b
superfamily ligand is essential for the versatility in
receptor binding
Results and Discussion
Structure of human BMP-6
Structures of BMP-6 and B2-BMP-6 were determined
by molecular replacement using the coordinates of
human (h)BMP-7 (Protein Data Bank entry 1BMP
[15]) as a start model Analysis of the unit cell content
suggested the presence of a complete BMP-6 dimer in
the asymmetric unit Assuming the presence of the BMP-6 dimer in the asymmetric unit, the Matthews coefficient VMis 3.94 A˚3ÆDa)1, corresponding to a sol-vent content of 70% With only one monomer pres-ent in the asymmetric unit, the solvpres-ent contpres-ent would exceed 86%, making this possibility unlikely Calcula-tion of a self-rotaCalcula-tion funcCalcula-tion confirmed the presence
of a two-fold noncrystallographic symmetry, which is distinct from all other structures of members of the TGF-b superfamily that have so far been determined (e.g BMP-2 [16], BMP-7 [15], GDF-5 [17,18], TGF-b2 [19,20], and TGF-b3 [21]) As a result, the biological dimer is not formed by a crystallographic dyad run-ning through the intermolecular disulfide bond, as is observed in all the other structures of TGF-b mem-bers, but by a noncrystallographic two-fold axis, indicating possible asymmetry in the homodimeric structure (Fig 1A,B)
Indeed, our structure data imply that loop regions, e.g the prehelix loop, in BMP-6 can adopt two differ-ent conformations (Fig 1C) Superimposing the struc-ture of free BMP-2 [16] onto the two different monomer conformers of BMP-6 yields rmsd values of 2.5 and 1.4 A˚ for the Ca positions, clearly indicating that one of the two BMP-6 conformers adopts a loop conformation similar to that of BMP-2 (Fig 1D) If only the Ca atoms of the b-sheet core of the BMP-2 dimer are considered (without helix, fingertip and pre-helix loops), an rmsd of 1 A˚ (0.8 and 0.6 A˚ for the individual monomer subunits) is observed, showing that the core b-sheet and the dimer architecture are almost identical between BMP-6 and BMP-2 (Fig 1D) The region exhibiting the largest difference between the two monomers is the prehelix loop comprising resi-dues Phe66 to Met72 of BMP-6 (Fig 1C) In con-former 1, the loop strongly deviates from the canonical backbone conformation observed in all other BMP (see Fig 1E,F for BMP-2 as an example) members [22] As compared with BMP-2, distances between the
Ca atoms of individual residues of up to more than
8 A˚, e.g between His71 (BMP-6) and His54 (BMP-2), are found The smallest distance (1.8 A˚) between two
Ca atoms of BMP-6 and BMP-2 within this loop seg-ment is observed for Leu68 (BMP-6) and Leu51 (BMP-2) (Fig 1E,F)
The prehelix loop, however, was shown to contain the main binding and specificity determinants for type I receptor recognition in BMP-2 [23] and GDF-5 [17] Structure analysis of receptor–ligand complexes of BMP-2 [23,24] and BMP-7 [25] suggested that receptor binding and recognition is accompanied by an induced
fit mechanism affecting the side chain and backbone conformation in this loop region Thus, the fact that
Trang 3BMP-6 shows two possible, largely different
conforma-tions for this loop segment suggests that recognition
and binding of type I receptors might be influenced by
this unique feature (Fig 1C) Superimposing the
struc-ture of the binary complex of BMP-2 bound to its
high-affinity receptor BMPR-IA indeed shows that the
noncanonical conformation of the prehelix loop of
BMP-6 would prevent binding of BMPR-IA, due to
steric hindrance, whereas the canonical loop
conforma-tion (BMP-2-like conformaconforma-tion) could form similar
noncovalent interactions with the type I receptor
(Fig 2) Although the noncanonical loop
conforma-tion of BMP-6 seems not to be able to form a stable
ligand–receptor interface using the type I receptor
structures known so far, the two different loop
confor-mations clearly show that the prehelix loop seems to
be dynamically disordered in the unbound ligand
Together with the fact that the a-helix of the type I
receptor BMPR-IA, which carries the main binding determinants for BMP-2 interaction (Phe85 and Gln86
of BMPR-IA), also seems not to be folded in the free receptor [26], a large portion of the core interface seems to be flexible and undergoes a disorder-to-order transition upon complex formation This induced-fit mechanism might explain the high degree of promiscu-ity in the BMP ligand–receptor interaction, as it allows the ligand as well as the receptor surfaces to adapt to the binding partner
Receptor binding and activity of BMP-6 depend
on the nature of the expression system Signaling of BMP-6 and BMP-7 has been shown to be mediated mainly via the ActR-I receptor in many cell types [9,13,27], although this receptor binds both BMPs only with weak affinities [25] In contrast,
C
D
Fig 1 The prehelix loop of BMP-6 adopts two conformations (A) Ribbon representa-tion of BMP-6 viewed from the top and from the side (B) The central intermolecular disulfide bond is indicated by ball-and-stick, secondary structure elements, and struc-tural features are marked The prehelix loop adopts two vastly different conformations, with the largest distances between the Ca atoms of the same amino acid residue in both segments (C) Stereoview of a super-position of the prehelix loop of conform-ers A and B of BMP-6 Residues occupying similar positions are indicated in black; resi-dues having different orientations in the two conformers are marked A and B according
to the conformer (D) Stereoview of a super-position of BMP-6 and BMP-2 (Protein Data Bank entry 3BMP), showing the differences
in the loop conformations of the fingertip loops as well as the prehelix loop (E) Super-position of the prehelix loops of BMP-2 (red carbon atoms) and BMP-6 (cyan carbon atoms) in its canonical loop conformation, which is very similar to that of BMP-2 (F) Same as in (E) except for the BMP-6 prehe-lix loop of conformer 1 (green carbon atoms), which adopts a noncanonical confor-mation, and is therefore different from BMP-2.
Trang 4BMP-2 uses the type I receptors BMPR-IA and
BMPR-IB, both of which are bound with high
affini-ties To further elucidate the molecular basis for the
different type I receptor specificity profiles of BMP-2,
BMP-6, and BMP-7, we used in vitro interaction
analy-sis (Table 1) Ligand proteins of BMP-2 (E coli),
BMP-6 (CHO cells) and BMP-7 (NS0 cells) were
immobilized onto a biosensor chip, and interaction
with the receptor ectodomain proteins BMPR-IA,
BMPR-IB and ActR-I was measured using surface
plasmon resonance spectroscopy (BIAcore technique)
As expected, immobilized BMP-2 showed high binding
affinities for BMPR-IA and BMPR-IB (KD= 10 and
95 nm, respectively), whereas binding to ActR-I was
below the detection level (KD> 400 lm) In contrast,
BMP-7 bound to BMPR-IA with a much lower affinity
of 10 lm and to BMPR-IB with a slightly higher
affinity of about 1 lm Binding of BMP-7 to ActR-I
yielded affinities (KD 50 lm) similar to those
described by Greenwald et al [25], and CHO
cell-derived BMP-6 showed a receptor binding profile
simi-lar to that of BMP-7 But, to our surprise, BMP-6
expressed from E coli did not bind to ActR-I
(Table 2), whereas basically identical binding
parame-ters for the type I receptors BMPR-IA and BMPR-IB
and for the type II receptor ActR-II were observed
(Table 2) Therefore, it can be ruled out that the
differ-ence in binding is caused by misfolding or unfolding of
the E coli-derived BMP-6 We thus investigated
whether E coli-derived BMP-6, which does not bind
ActR-I, is inactive in cell-based assays, as would be expected if ActR-I were the main signaling receptor for BMP-6 and BMP-7 Indeed, glycosylated BMP-6 and BMP-7 induced alkaline phosphatase (ALP)
A
Fig 2 Analysis of a BMP-6–BMPR-IA
com-plex model (A) A putative model of BMP-6
bound to BMPR-IA was built by docking the
BMPR-IA molecules of the BMP-2–BMPR-IA
complex (Protein Data Bank entry 1REW) to
BMP-6 (B) The noncanonical prehelix loop
is incompatible with complex formation, due
to several steric clashes between residues
of the BMP-6 prehelix loop and the
b 5 a 1 -loop of BMPR-IA (C) In its canonical
(or BMP-2 like) form, the prehelix loop
adopts a conformation that is very similar to
that of BMP-2 in the BMP-2–BMPR-IA
complex No severe steric clashes are
found, suggesting that this loop
conforma-tion might be adopted in a BMP-6–BMPR-IA
interaction.
Table 1 Receptor binding profile of BMP-6, BMP-7, and BMP-2 (BIAcore analysis) Biosensor analysis using surface plasmon reso-nance was performed to determine binding affinities of the BMP ligand–receptor interaction Ligands were immobilized onto the sur-face of a CM5 sensor chip, and receptor ectodomain proteins were used as analyte Thus, interaction analysis yields the 1 : 1 interac-tion of BMPs and their receptor ectodomain proteins NB, no bind-ing within detection limit (upper limit: KD> 400 l M ).
Receptor proteins
Ligands [affinity (l M )] a
Type I
Type II
a KD(eq) as deduced from the dose dependency of equilibrium bind-ing b KD(kin) as deduced from the association and dissociation rates of the interaction; analysis of dose dependency of equilibrium binding yields higher (three-fold) values for KD(eq), as real equilib-rium binding cannot be achieved, due to the slow association rates (k on 3 · 10 4
Æ M )1Æs)1) of the BMP-2–BMPR-IA and BMPR-IB 1 : 1
interaction In contrast, association, and especially dissociation, for interaction of BMPs with ActR-I and ActR-II are faster (k on > 10 5 Æ M )1Æs)1, k
off > 10)1Æs)1), impeding the analysis of the dis-sociation rate and thus requiring analysis of the equilibrium binding.
Trang 5expression in ATDC-5 cells in a dose-dependent
man-ner, with EC50 values of 9 and 57 nm (Fig 3A) In
contrast, BMP-6 derived from E coli was practically
inactive at ALP induction, even at the high
concentra-tions tested (Fig 3A) It is also interesting to note that
BMPR-IA seems to be not able to rescue activity of
BMP-6 in ATDC5 cells, despite the fact that in vitro
interaction analysis shows that BMP-6 (and BMP-7)
can bind to BMPR-IA and BMPR-IB with higher
affinity than ActR-I That signaling of BMP-6 and
BMP-7 is mediated via ActR-I can be seen from the inhibition of proliferation in the human myeloma cell line INA6, which expresses ActR-I but not BMPR-IA
or BMPR-IB [28] Both BMP-6 and BMP-7 showed high activity in this cell line, whereas BMP-2, which signals via BMPR-IA and BMPR-IB, did not (Fig 3B) This is the first time that such a large differ-ence in binding and activity has been observed for members of the TGF-b superfamily For example, BMP-2 derived from either prokaryotic or eukaryotic expression systems has similar biological activities when tested in various cells, e.g ALP induction in C2C12 or ATDC5 cells (W Sebald, unpublished results) Receptor binding is only marginally influ-enced, with a slightly decreased affinity of CHO cell-derived BMP-2 for the type II receptor ActR-IIB [29]
Recognition of BMP-6 by ActR-I depends on N-glycosylation
To determine the molecular basis for these differences between BMP-6 derived from E coli or CHO cells, we investigated whether post-translational modifications might play a role in receptor binding and activity The crystal structure analysis of recombinant BMP-7 expressed in CHO cells (Protein Data Bank entry 1LXI [25]) or complexes of BMP-7 (Protein Data Bank entries 1LX5 [25] and 1M4U [30]) showed that the N-glycosylation sequence Asn-X-Ser⁄ Thr in the cystine-knot motif, which is conserved among BMP ligands of the BMP-2⁄ 4 and the BMP-5 ⁄ 6 ⁄ 7 family, does indeed
Table 2 N-glycosylation of BMP-6 is required for ActR-I binding
(BIAcore analysis) Biosensor analysis using surface plasmon
reso-nance was performed to determine binding affinities of the BMP-6
ligand–receptor interaction BMP-6 proteins were immobilized onto
the surface of a CM5 sensor chip via amino-coupling, and receptor
ectodomain proteins were used as analyte Thus, interaction
analy-sis yields the 1 : 1 interaction of BMPs and their receptor
ectodo-main proteins NB, no binding within detection limit (upper limit:
KD> 400 l M ).
Receptor
proteins
Ligands [affinity (l M ) a ]
BMP-6
(CHO)
BMP-6 (CHO) PNGase F
BMP-6 (CHO) PNGase F3 ⁄ H
BMP-6 (E coli) Type I
Type II
a KD(eq) as deduced from the dose dependency of equilibrium
bind-ing.
0 2000 4000 6000 8000
10 000
12 000
14 000
BMP-2 (n.d.) BMP-6 (12.4 n M +/–0.8) BMP-7 (24.0 n M +/–2.2)
BMP-6 E.coli (n.d.)
control
0.0
0.5
1.0
1.5
2.0
2.5
BMP- 2 ( 5.3 n M +/–0.3)
BMP- 6 ( 9.3 n M +/–0.5)
BMP- 7 ( 56.7 n M +/–3.4)
BMP- 6 E.coli
A B
Fig 3 Biological activities of BMP-6, BMP-7, and BMP-2 (A) Induction of ALP expression in ATDC5 cells is stimulated by BMP-6, BMP-7, and BMP-2 BMP-6 derived from E coli expression is inactive in these cells, probably due to its lack of binding to ActR-I (B) Signaling of BMP-6 and BMP-7 via ActR-I is shown in the inhibition of proliferation in the myeloma cell line INA6, which lacks the type I receptors BMPR-IA and BMPR-IB Whereas BMP-6 and BMP-7 show high activity in this cell line, BMP-2 is almost completely inactive, due to its requirement for BMPR-IA or BMPR-IB E coli-derived BMP-6 is also inactive, due to its inability to bind ActR-I; the green dashed line indi-cates maximal proliferation in the absence of any BMP ligand.
Trang 6carry carbohydrate moieties The two putative
N-gly-cosylation sites in the N-terminus of the mature part
of BMP-7, which are also present in BMP-6, have been
shown not to be glycosylated [31,32] The binding site
of the type I receptors is located in the so-called wrist
epitope of BMPs, which comprises a part of both
fin-gers and the prehelix loop, suggesting that the
carbo-hydrate moieties linked to Asn73 of BMP-6 (Asn80 on
BMP-7) could contact the type I receptors and thus
modulate receptor binding To confirm this hypothesis,
we performed deglycosylation of CHO cell-expressed
BMP-6 and determined its receptor-binding properties
by BIAcore interaction analysis First, we removed all
N-linked carbohydrate by N-endoglycosidase F
treat-ment under nondenaturing conditions to ensure that
the folding of BMP-6 was not altered by the enzymatic
reaction Endoglycosidase F hydrolyzes the
N-glyco-sidic bond between the asparagine and the first
N-acet-ylglucosamine (GlcNAc) residue, resulting in a
nonglycosylated protein (Fig 4A) The completeness
of the deglycosylation was checked by SDS⁄ PAGE
(Fig 4B) and MS analysis; the BMP-6 was then
immo-bilized onto a biosensor, and the properties of binding
to BMP type I and type II receptors were determined
by BIAcore analysis Whereas binding to the type I
receptors BMPR-IA and BMPR-IB, as well as to the
type II receptor ActR-II, was essentially identical to
binding of fully glycosylated BMP-6, no binding to
ActR-I could be determined (Table 2) This clearly shows that binding of BMP-6 to ActR-I requires car-bohydrate moieties attached to Asn73 as binding deter-minants, whereas the other type I receptors do not As the parameters for binding to BMPR-IA, BMPR-IB and ActR-II are not influenced by the removal of the N-glycosylation, large, and even small, local structural changes can be excluded We examined how many car-bohydrate residues might be involved in the binding of ActR-I by using a mixture of N-endoglycosidase H and N-endoglycosidase F3 The latter cleaves the b1–4 glycosidic bond between the first and the second Glc-NAc residue, leaving the first carbohydrate (GlcGlc-NAc) attached to the protein (Fig 4A,B) Measurement of the binding affinities of this partially glycosylated BMP-6 for BMPR-IA and BMPR-IB confirms that binding to these two type I receptors is not altered by different N-glycosylation levels However, binding affin-ity for ActR-I is now very close (less than a factor of 2)
to that of CHO cell-derived BMP-6 with full N-glycosyl-ation (Table 2), showing that the first carbohydrate moiety at Asn73 is a main binding determinant for ActR-I interaction, whereas further carbohydrate resi-dues in the carbohydrate chain are not required The aspartyl side chain generated from Asn by deglycosyla-tion using endoglycosidase F cannot be responsible for this lack of activity, as the unglycosylated E coli BMP-6 containing an Asn at position 73 is also inactive
R 1
R 2
GlcN
β1-4 α1-4 GlcN GlcN
GlcN
GlcN
M M
M
β1-4
β1-2
β1-2
α1-3 α1-6
R 1
R 2
Asp
GlcN
β1-4
GlcN
R
PNGase F
PNGase F3/H
45 35
25
18.4 14.4
45 35
25
18.4 14.4
PNGases
Asp
Fig 4 Deglycosylation of BMP-6 expressed from CHO cells (A) Scheme to illustrate the restriction sites for the endoglycosidases used PNGase F hydrolyzes the N-glycosidic bond immediately after the asparagine residue, leaving a fully deglycosylated protein A mixture of PNGase F3 and PNGase H is used to trim complex carbohydrate structures to a single GlcNAc moiety attached to the asparagine residue (B) SDS ⁄ PAGE analysis under reducing conditions of the deglycosylation reactions of BMP-6, showing the completeness of the enzymatic reactions M: molecular weight marker Lane 1: BMP-6 derived from CHO cells Lane 2: BMP-6 after PNGase F treatment Lane 3: BMP-6 before PNGase F3 ⁄ H treatment Lane 4: BMP-6 after PNGase F3 ⁄ H treatment BMP-6 runs as two bands, i.e dimer and monomer.
Trang 7Modeling the interaction of BMP-6 and ActR-I
To gain insights into how ActR-I might interact with
the N-glycosylation site at Asn73 of BMP-6, we
con-structed a model of the binary complex of BMP-6
bound to ActR-I on the basis of our BMP-6 structure
and the structures of the BMP-7–ActR-II (Protein
Data Bank entry 1LX5 [25]) and BMP-2–BMPR-IA
(Protein Data Bank entry 1REW [23]) complexes A
putative carbohydrate chain was added to BMP-6 by
using the complex glycosylation structure present on
BMP-7 as identified in BMP-7–ActR-II and which also
presents a typical N-glycosylation from expression
in mammalian cells The model of the extracellular
domain of ActR-I is based on the structure of
BMPR-IA in its bound conformation to BMP-2 (Protein Data
Bank entry 1REW); insertions and deletions were built
manually using quanta2006 software The putative
complex model of BMP-6(glycosylated)–ActR-I was
then formed by superimposing BMP-6(glycosylated)
and ActR-I with the ligand and receptor structures in
BMP-2–BMPR-IA (Protein Data Bank entry 1REW)
The model of BMP-6–ActR-I shows that several
resi-dues in the N-terminus, the b1b2-loop and the short
loop before the a-helix of ActR-I are in close
proxim-ity to the carbohydrate chain of the BMP-6(glycosylat-ed) model (Fig 5A,B) Residues of ActR-I in these regions, namely Lys11 and Tyr54, can possibly form several hydrogen bonds with the first carbohydrate moiety (Fig 5C,D), showing how the first carbohy-drate plays an important role in recognition and the generation of binding affinity for the BMP-6–ACTR-I interaction
The model also gives some hints as to why the bind-ing affinity of BMP-6 for the type I receptors
BMPR-IA and BMPR-IB is not dependent on the presence of the carbohydrate structure The b1b2-loop of ActR-I is shortened by three residues in comparison to
BMPR-IA and BMPR-IB, possibly resulting in a less flexible loop in ActR-I Interactions between residues within this loop and the carbohydrate chain might thus con-tribute significantly to the binding free energy, whereas
in the more flexible b1b2-loop of BMPR-IA⁄ IB it does not
In summary, our analysis shows the first structure of
a BMP ligand member, which exhibits two vastly dif-ferent conformations for the prehelix loop, which has been shown to be important for BMP type I receptor interaction Although, due to the lack of other BMP ligand–receptor complex structures with type I receptors
Fig 5 Model of the binary complex of N-glycosylated BMP-6 bound to ActR-I Ribbon representation of the binary complex of BMP-6 bound
to its type I receptor ActR-I The carbohydrate chain [GlcNAcb1–4GlcNAcb1–4Man(a1,3Man)(a1,5Man)(b1,4Man)] – shown as thick lines – was added from a crystal structure analysis of BMP-7 (expressed in CHO cells) bound to the type II receptor ActR-II Several residues in the N-terminus, b1b 2 -loop or the loop in front of the a-helix of ActR-I are in close contact with the carbohydrate, namely the first two to three carbohydrate moieties, as would be predicted from our deglycosylation studies (A) Viewed from the top (B) Viewed from the side (C) BMP-6 (cyan ⁄ green) and ActR-I (magenta) are shown as surface representations to visualize the close packing of the carbohydrate in between the ligand–receptor interface; putative contact residues are indicated (D) Putative hydrogen bond interactions between ActR-I and the first GlcNAc residue of BMP-6 glycosylated at Asn73 Hydrogen bonds between the first carbohydrate residue (GlcNAc) and residues Lys11 and Tyr54 of ActR-I are shown.
Trang 8different from BMPR-IA, we cannot say whether the
second noncanonical loop conformation plays a direct
role in complex formation, however the second loop
conformation confirms that this loop is highly mobile
in BMP-6 and possibly also in other BMPs An NMR
relaxation study on TGF-b3has shown that this region
is dynamically disordered in solution [33] As the
bind-ing epitope of BMPR-IA also seems to be not fully
folded in solution, formation of complexes of BMP
ligands and receptor seem to involve a large induced
fit mechanism This conformational rearrangement
upon binding might explain the promiscuous binding
of BMPs, as it allows the binding epitopes to adapt to
various different binding partners Furthermore, so
far, functional analysis has highlighted the importance
of the central hydrogen bond pair in
BMP-2–BMPR-IA for BMP type I receptor recognition [23] However,
in the binding of BMP-6 to ActR-I, a new, so far
unknown, main binding determinant has been
discov-ered This new hot spot of binding involves an
N-gly-cosylation motif conserved between BMP-2, BMP-7,
and BMP-6, which is specifically required for binding
of BMP-6 (and possibly BMP-7) to ActR-I but does
not play a role in binding to the other type I receptors
BMPR-IA and BMPR-IB This finding suggests that,
in addition to the above-mentioned flexible binding
epitope, usage of different main binding determinants
might also add to the broad binding specificity
observed in the BMP family
Experimental procedures
Expression and purification of recombinant
proteins
The mature part of hBMP-6, comprising amino acids 375–
513 plus an N-terminal extension MAPT (single-letter
amino acid code) [34], was expressed in E coli
Alterna-tively, a BMP-6 variant with residues 375–410 replaced with
the sequence MAQAKHKQRKRLK was used
(B2-BMP-6) The protein was expressed in insoluble form in inclusion
bodies BMP-6 isolated from these inclusion bodies was
refolded and purified as previously described [35]
Recombi-nant hBMP-6 obtained by eukaryotic expression, i.e CHO
cells, was purchased from R&D Systems (Minneapolis,
MN, USA) The extracellular domains of the receptors
BMPR-IA and BMPR-IB were expressed as
thioredoxin-fusion proteins in E coli and purified as previously
described [36] The extracellular domains of hActR-I
(resi-dues 21–123 [37]) and hActR-II (resi(resi-dues 18–135 [38]) were
expressed in baculoviral-infected Sf9 insect cells as
previ-ously described [20] The receptor proteins hActR-I and
hActR-II were purified by metal affinity chromatography
using Ni–nitrilotriacetic acid–agarose (Qiagen, Hilden, Germany); the eluate was dialyzed against HBS buffer (10 mm Hepes, pH 7.4, 3.4 mm EDTA, 20 mm NaCl), and subjected to anion exchange chromatography The flow-through of the latter step contains the monomeric, biologi-cally active receptor protein, which was then finally purified
by RP-HPLC
Interaction analysis by surface plasmon resonance
A BIAcore2000 system (BIAcore Life Science; GE Health-care, Freiburg, Germany) was used for all biosensor experi-ments Ligand proteins were directly immobilized onto a CM5 biosensor chip at a density of about 800 resonance units (1 RU = 1 pgÆmm)2), using the amine coupling kit (BIAcore Life Science; GE Healthcare) according to the manufacturer’s protocol Sensor chips were first activated
by perfusing an ethyl-N-(3-diethylaminopropyl)carbodi-imide (EDC)⁄ N-hydroxysuccinimide (NHS) mixture for
7 min; ligands were dissolved in 10 mm sodium acetate (pH 4.5) at a concentration of 1 lgÆmL)1and perfused over the activated chip surface until the required surface density was achieved Sensor chips were subsequently deactivated with 1 m ethanolamine (pH 8.0) for 7 min
All interaction experiments were carried out using HBS500 buffer (10 mm Hepes, pH 7.4, 500 mm NaCl, 3.4 mm EDTA, 0.005% surfactant P20) Sensorgrams of receptor–ligand interaction were recorded at a flow rate of
10 lLÆmin)1at 25C The association and dissociation time was set to 5 min After each cycle, 4 m MgCl2was perfused for 2 min of regeneration
Evaluation of recorded sensorgrams
Apparent binding affinities were calculated using biaevalu-ation software 2.2.4 Bulk face effects, i.e unspecific bind-ing to the biosensor or buffer exchange, were removed by subtracting a reference flow cell (FC1) from all sensor-grams Briefly, equilibrium binding constants of the interac-tion of BMP-2 with the type I receptors BMPR-IA and BMPR-IB were calculated by fitting the kinetic data to a
1 : 1 Langmuir binding model [KD(kin)], and those of the interaction of BMPs with hActR-I and hActR-II as well as
of BMP-6 and BMP-7 with BMPR-IA and BMPR-IB were determined from the dose dependency of the equilibrium binding [KD(eq)] The relative standard deviations for mean
KD(eq) values were below 25%; those of mean KD(kin) val-ues were below 50%
Deglycosylation of BMP-6
Recombinant hBMP-6 obtained from a eukaryotic expres-sion system (R&D Systems, Minneapolis, MN, USA) was
Trang 9fully or partially deglycosylated using either the
endogly-cosidase PNGase F (New England Biolabs, Frankfurt,
Ger-many), or a mixture of the endoglycosidases PNGase F3
and PNGase H (New England Biolabs) For complete
deglycosylation under nondenaturing conditions, 1 U of
PNGase F was used per microgram of BMP-6, with a
reac-tion time of 8 h at 37C For glycosylation trimming, a
mixture of 1 mU of PNGase F3 and 1 mU of PNGase H
was used per microgram of hBMP-6, with a reaction time
of 24 h at 37C The completeness of the carbohydrate
removal was analyzed by SDS⁄ PAGE and MS
Crystallization and structure analysis of BMP-6
Lyophilized E coli-derived wild-type BMP-6 and the
vari-ant B2-BMP-6 were dissolved in water at a concentration
of 5–10 mgÆmL)1 and submitted to crystallization trials
using Hampton Crystal Screens I and II (Hampton
Research, Aliso Viejo, USA) Crystals were obtained using
several sets of conditions and organic solvents, i.e
2-propa-nol, 2-methyl-2,4-pentanediol (MPD) or dioxane, or
poly-ethylene glycols (polypoly-ethylene glycol 4000 to polypoly-ethylene glycol 6000) Wild-type BMP-6 and the B2-BMP-6 crystal-lized under identical conditions, however, due to the increased solubility of B2-BMP-6, reproducibly yielded lar-ger crystals Suitable crystals of B2-BMP-6 grew from 25% MPD and 0.1 m sodium citrate (pH 4.0), and for wild-type BMP-6, the largest crystals were obtained from 20% 2-pro-panol and 0.1 m sodium citrate (pH 4.0) Diffraction data for B2-BMP-6 were collected from a single crystal at 100 K
at the beamline XS06SA at the Swiss Light Source (SLS; Paul Scherrer Institute, Switzerland), and data for wild-type BMP-6 were acquired at 100 K using a home source (Rigaku RU300, MarResearch Imageplate 345, Osmic Con-focalBlue) Data were processed using xds software [39] or HKL2000⁄ Scalepack [40]; a summary of the processing statistics is given in Table 3
Structure analysis was performed by applying molecular replacement using cns software [41] and the structure of BMP-7 (Protein Data Bank entry 1BMP [15]) as a search model The initial models were refined by iterative manual model building using quanta2006 software (Accelrys Inc.,
Table 3 Processing and refinement statistics Statistical analyses for the highest-resolution shell are shown in parentheses.
c = 86.8 A ˚
a = b = 90, c = 120
a = b = 97.7 A ˚ ,
c = 85.1 A ˚
a = b = 90, c = 120
Refinement statistics
rmsd
Procheck analysis c
a
Wild-type BMP-6 was analyzed for comparison The structures of BMP-6T2 and wild-type BMP-6 are identical within the accuracy of the resolution (rmsd of 0.7 for all Ca atoms); residues in the segment differing between BMP-6T2 and wild-type BMP-6 show no electron den-sity, indicating a high degree of flexibility b Cut-off for reflections F > 0r c Number of residues is shown in parentheses.
Trang 10San Diego, CA, USA), and either Refmac5 [42] or CNS
[41] was used for subsequent refinement Progress of
refine-ment was monitored using the R-factors Rcryst and Rfree;
the latter was calculated from a test dataset comprising 5%
of randomly selected reflections In the final rounds of
refinement, electron density difference maps Fobs– Fcalc
were used to identify 78 water molecules (for wild-type
BMP-6, 53 water molecules) and four MPD molecules (for
wild-type BMP-6, 10 2-propanol molecules could be
identi-fied) The final minimization cycle yielded R-factors of 25.9
for Rcryst and 27.9 for Rfree for B2-BMP-6 (for wild-type
BMP-6, Rcrystis 23.5 and Rfreeis 27.0)
ALP induction
The teratocarcinoma AT508-derived cell line ATDC5
(RIKEN, Ibaraki, Japan, No RCB0565) was cultured in
DMEM⁄ F12 (1 : 1) medium containing 5% fetal bovine
serum, and antibiotics (100 UÆmL)1 penicillin G and
100 lgÆmL)1streptomycin) For ALP assays, the cells were
serum starved (2% fetal bovine serum) and exposed to
ligands for 72 h in 96-well microplates After cell lyses,
ALP activity was measured by p-nitrophenylphosphate
con-version using an ELISA reader at 405 nm
BMP-induced inhibition of INA6 cell proliferation
Cells of the human myeloma cell line INA6 were seeded in
DMEM in 96-well plates at densities of 5· 103
cells per well The dose-dependent inhibition of proliferation was
measured by adding increasing concentrations of BMP-6,
BMP-7 (R&D Systems, Minneapolis, MN, USA) or
BMP-2 After 72 h, 10 lL of [3H]thymidine (0.25 lCi;
GE Healthcare⁄ Amersham, Munich, Germany) was added
to each well The cells were immobilized after 24 h on fiber
mats (Skatron Instruments A⁄ S, Lier, Norway), and the
thymidine incorporation was determined using a RITA
counter (Raytest, Straubenhardt, Germany) All assays
were performed in duplicate, and the experiments were
repeated twice
Acknowledgements
The authors thank Maike Gottermeier and Christian
So¨der for excellent assistance We wish to acknowledge
access to the X-ray facility at TU Munich (A Skerra)
and the beamline XS06SA at the Swiss Light Source,
and thank C Schulze-Briese for local support
References
1 Lyons KM, Pelton RW & Hogan BL (1989) Patterns of
expression of murine Vgr-1 and BMP-2a RNA suggest
that transforming growth factor-beta-like genes
coordi-nately regulate aspects of embryonic development Genes Dev 3, 1657–1668
2 Dudley AT & Robertson EJ (1997) Overlapping expres-sion domains of bone morphogenetic protein family members potentially account for limited tissue defects in BMP7 deficient embryos Dev Dyn 208, 349–362
3 Solloway MJ, Dudley AT, Bikoff EK, Lyons KM, Hogan BL & Robertson EJ (1998) Mice lacking Bmp6 function Dev Genet 22, 321–339
4 Jones CM, Lyons KM & Hogan BL (1991) Involvement
of bone morphogenetic protein-4 (BMP-4) and Vgr-1 in morphogenesis and neurogenesis in the mouse Develop-ment 111, 531–542
5 Kim RY, Robertson EJ & Solloway MJ (2001) Bmp6 and Bmp7 are required for cushion formation and sep-tation in the developing mouse heart Dev Biol 235, 449–466
6 Dale JK, Vesque C, Lints TJ, Sampath TK, Furley A, Dodd J & Placzek M (1997) Cooperation of BMP7 and SHH in the induction of forebrain ventral midline cells
by prechordal mesoderm Cell 90, 257–269
7 Tomizawa K, Matsui H, Kondo E, Miyamoto K, Tokuda M, Itano T, Nagahata S, Akagi T & Hatase O (1995) Developmental alteration and neuron-specific expression of bone morphogenetic protein-6 (BMP-6) mRNA in rodent brain Brain Res Mol Brain Res 28, 122–128
8 Rosenzweig BL, Imamura T, Okadome T, Cox GN, Yamashita H, ten Dijke P, Heldin CH & Miyazono K (1995) Cloning and characterization of a human type II receptor for bone morphogenetic proteins Proc Natl Acad Sci USA 92, 7632–7636
9 ten Dijke P, Yamashita H, Sampath TK, Reddi AH, Estevez M, Riddle DL, Ichijo H, Heldin CH &
Miyazono K (1994) Identification of type I receptors for osteogenic protein-1 and bone morphogenetic protein-4 J Biol Chem 269, 16985–16988
10 Liu F, Ventura F, Doody J & Massague J (1995) Human type II receptor for bone morphogenic proteins (BMPs): extension of the two-kinase receptor model to the BMPs Mol Cell Biol 15, 3479–3486
11 Wrana JL, Attisano L, Wieser R, Ventura F &
Massague J (1994) Mechanism of activation of the TGF-beta receptor Nature 370, 341–347
12 Massague J & Wotton D (2000) Transcriptional control
by the TGF-beta⁄ Smad signaling system EMBO J 19, 1745–1754
13 Macias-Silva M, Hoodless PA, Tang SJ, Buchwald M
& Wrana JL (1998) Specific activation of Smad1 signal-ing pathways by the BMP7 type I receptor, ALK2
J Biol Chem 273, 25628–25636
14 Sairam MR, Bhargavi GN & Yarney TA (1990) Hormone glycosylation required for lutropin receptor recognition in sheep testis FEBS Lett 276, 143–146