Solution structure of human insulin-like growth factor 11 - NCBI

The EMBO Journal vol.13 no.23 pp.5590-5597, 1994

Solution structure of human insulin-like growth factor 11; recognition sites for receptors and binding proteins Hiroaki Terasawal 2, Daisuke Kohdal, Hideki Hatanakal, Koji Nagatal, Nobuyuki Higashihashi2, Hiroyuki Fujiwara2, Katsu-ichi Sakano2 and Fuyuhiko Inagakil 3 IDepartment of Molecular Physiology, Tokyo Metropolitan Institute of Medical Science, 3-18-22, Honkomagome, Bunkyo-ku, Tokyo 113 and 2Molecular Biology Research Laboratory, Daiichi Pharmaceutical Co., Ltd, 16-13, Kitakasai 1-chome, Edogawa-ku, Tokyo 134, Japan 3Corresponding author Communicated by I.D.Campbell

The three-dimensional structure of human insulin-like growth factor II was determined at high resolution in aqueous solution by NMR and simulated annealing based calculations. The structure is quite similar to those of insulin and insulin-like growth factor I, which consists of an a-helix followed by a turn and a strand in the B-region and two antiparallel a-helices in the A-region. However, the regions of Alal-Glu6, Pro3l-Arg4O and Thr62-Glu67 are not well-defined for lack of distance constraints, possibly due to motional flexibility. Based on the resultant structure and the results of structure-activity relationships, we propose the interaction sites of insulin-like growth factor II with the type 2 insulin-like growth factor receptor and the insulin-like growth factor binding proteins. These sites partially overlap with each other at the opposite side of the putative binding surface to the insulin receptor and the type 1 insulin-like growth factor receptor. We also discuss the interaction modes of insulin-like growth factor II with the insulin receptor and the type 1 insulin-like growth factor receptor. Key words: insulin-like growth factor II/nuclear magnetic resonance/receptor binding/simulated annealing calcula-

tion/three-dimensional protein structure

Introduction Insulin-like growth factors II (IGF-II) and I (IGF-I) are single-chain polypeptides of 67 and 70 amino acid residues, respectively, that share a high degree of amino acid sequence homology with insulin (Rinderknecht and Humbel, 1978a,b). They consist of four regions designated B, C, A and D beginning from the N-terminus (Figure 1). In the A- and B-regions, -70% of the residues in IGF-II and IGF-I are identical and -50% in IGFs are identical with those of the A- and B-chains of insulin. The Cregion of IGF which corresponds to the C peptide of proinsulin has no sequence homology either with each other or with proinsulin. The D-region at the C-terminus of IGFs is not found in insulin. IGF-I and IGF-II possesses mitogenic activities as well

as insulin-like metabolic activities in vitro (Froesch et al., 1985). Furthermore, IGF-I mediates the growth promoting actions of growth hormone in vivo, which was well established (Froesch et al., 1985; Humbel, 1990). Whereas IGF-II seems to act as a fetal growth factor in some species (Moses et al., 1980; Daughaday and Rotwein, 1989; De Chiara et al., 1990), the biological role of IGFII in humans is not clear. The expression of IGF-II is found in a variety of tumor types in vitro (Schofield, 1992) and appears to be part of a rate-limiting step in multistage oncogenesis (Christofori et al., 1994). IGF-II, IGF-I and insulin bind to their respective receptors with high affinities. These receptors are the type 2 insulin-like growth factor receptor (type 2 IGFR), the type 1 insulin-like growth factor receptor (type 1 IGFR) and the insulin receptor (IR). Type 1 IGFR and IR are structurally related molecules which form a sub-group within the wider family of growth factor receptors with ligand-stimulated intrinsic tyrosine-specific protein kinase activity (Jacobs et al., 1983; Ebina et al., 1985; Sasaki et al., 1985; Ullrich et al., 1985, 1986; Ullrich and Schlessinger, 1990). On the other hand, type 2 IGFR, which is identical with the cation independent mannose6-phosphate receptor (Morgan et al., 1987), is not related to type 1 IGFR or IR and its function has not yet been elucidated. IGF-II also binds to type 1 IGFR and IR with moderate affinities, whereas IGF-I and insulin do not bind (or bind poorly) to each other's receptors (Casella et al., 1986; Bayne et al., 1990; Sakano et al., 1991). It is now accepted that most of the biological effects of IGF-II are not mediated via type 2 IGFR but rather via type 1 IGFR and/or IR (Czech, 1989; Roth and Kiess, 1994). In addition, IGFs also bind to a family of specific insulin-like growth factor binding proteins with high affinities designated IGFBP- 1 to IGFBP-6 that are thought to modulate IGF activities in blood and in tissues (Baxter and Martin, 1989; Rechler and Nissley, 1990; Clemmons, 1992; Rechler, 1993). Thus, IGF-II has a unique property in that it binds to several proteins; IR, type 1 IGFR and IGFBPs, as well as type 2 IGFR. Attempts to elucidate the residues required for interaction with these structurally and functionally distinct biological partners have been made using modified IGF-II. These attempts have been limited by the lack of detailed information concerning the tertiary structure of IGF-II, although a model for IGF-II was proposed based on the crystal structure of insulin (Blundell et al., 1978, 1983). In the present study, we report the three-dimensional structure of human IGF-II, determined by NMR and simulated annealing calculations, and compare it with those of the IGF-II model (Blundell et al., 1983), insulin (Dodson et al., 1979; Hua et al., 1991) and IGF-I (Cooke et al., 1991; Sato et al., 1993). We propose the interaction sites of IGF-II with type 2 IGFR and IGFBPs. We also

5 0 Oxford University Press 5590

Structure of human IGF 11 B-Region Human IGT-Il Human 1GW-I

30 1 20 10 AYRPSETLCGGELVDTLQFVCGDRGFYFSRPA

1

a-B

20

. 8.2 8Ie.0 .

7.6

7.6

7.8

10

GPETLCGAELVDALQFVCGDRGFYFNKPT 0 330 310 B1 FVNQHLCGSHLVEALYLVCGE RGFFYTPKA

Porcine Insulin C-Region Human IGIr-I1

10

8.I 8.4

(a)

40

CD

S- -RVSRRSR 40 GYGS SSRRAPQT

30

Humn IGF-I

A-Region Human XGJ-Il H-man

IGW-I

s0

GIVEECCFRSCDLALLETYCA a-A2 60 a-Al SO GIVDECCFRSCDLRRLEMYCA alO

Al

a20

Porcine Insulin GIVEQCCTSICSLYQLENYCN D-Region Human IGF-Il Human IGW-I

FD a

67

T

- -

PAKSE 70

P LKPAKSA

OD

.j

Fig. 1. Amino acid sequences of IGF-II, IGF-I and insulin. The sequences are aligned based on the IGF-II sequence with numbering of each peptide. Residues located at a-helix regions are underlined; residues contained in the hydrophobic core are in bold (Blundell et al., 1978; Cooke et al., 1991 ).

discuss the interaction modes of IGF-II with IR and type IGFR.

a.

10

1

(b)

30

20

AYRIS T LCG3LVDTLQFVCGDRGWYWSRPASR

dNM

t

-?

dNNi

-

t-

-??-?-

-t-?

_

Results Sequential assignments and secondary structure IGF-II was insoluble in H20 at millimolar concentration between pH 4 and so that initial efforts were devoted to find a solution condition suitable for NMR measurements. At acidic pH, IGF-II was apparently dissolved at 2 mM concentration in H20 but was still aggregated, as evidenced by the line broadening of 'H resonances. However, the aggregation was significantly removed by addition of acetic acid, as was observed in the case of insulin and IGF-I (Kline and Justice, 1990; Hua and Weiss, 1991; Hua et al., 1991; Sato et al., 1993). An increase in temperature further reduced the line width of the resonances without denaturation. Hence, NMR experiments were made in mixed solvent of 5:85:10 (v/v/v) CD3CO2D: H20:D20 or 5:95 (v/v) CD3CO2D:D2O at pH 3.2 (direct pH meter reading) and at 50°C with a peptide concentration of 2.4 mM. The resonances were assigned to individual protons in a sequence specific manner using a sequential assignment method (WUithrich, 1986). First, NMR resonances were assigned to the spin systems of specific amino acid types using double-quantum filtered correlation spectroscopy (DQF-COSY) and total correlation spectroscopy (TOCSY). Second, the spin systems were aligned according to the sequence of IGF-II using sequential connectivities obtained by NOE spectroscopy (NOESY) experiments. Figure 2a shows a NOESY spectrum in the amide proton region, where dNN(i, i + 1) NOE connectivities are manifested, suggesting that IGF-II is rich in a-helix. As is often the case with a-helix rich proteins, significant overlap of the resonances was observed. We therefore applied three-dimensional total correlation 15N-IH heteronuclear multiple quantum coherence (TOCSY-HMQC) and IH NOE 15N-'H HMQC (NOESY-HMQC) to resolve the overlapping resonances. Successive strong dNN(i, i + 1), daN(i, i + 3) and dp(i, i + 3) NOE connectivities, characteristic of an a-helix, were observed (Figure 2b), which revealed that IGF-II

dco(.S) 40

so

60

?

? ? ??_

67

VaRRSRGXVZZCC FRSCDLALLBTYCATPAKS3

do" -? ? dmNN-? dpZtt-

?

??

-?-?-~? = -?

-

d'MM4I.

? ?

-

.l

Fig. 2. (a) Down-field region of the NOESY spectrum of IGF-II in H20 solution measured at 50°C and with a mixing time of 200 ms. The sequential cross peaks linking residues of the B-region are labeled. (b) Summary of the sequential NOE connectivities involving the NH, C'H and CPH protons measured at 50°C and with 200 ms mixing time. NOEs are classified into five classes; very strong, strong, medium, weak and very weak, which are indicated by the height of the filled bar underneath the sequence. X indicates an undefined NOE connectivity. ? indicates an NOE connectivity which is observed but not clearly classified due to overlap with other NOE peaks.

comprised three a-helices: in the A-region.

one

in the B-region and two

Determination of the three-dimensional structure of IGF-II A total of 633 NOE distance constraints which included 244 intra-residue, 169 sequential (li-jl = 1), 129 short 5) and 91 long range (li-ji > 5) range (1 < Ii-Il constraints were derived from assigned NOE cross peaks measured with mixing times of 75 ms and 200 ms. Eight dihedral angle constraints of Xi were estimated from primitive exclusive COSY (PE-COSY) spectrum. The -

three-dimensional structures were calculated with XPLOR (Brunger, 1990) using the simulated annealing protocol (YASAP) on the basis of 641 above mentioned experimental constraints and three distance constraints 5591

H.Terasawa et aL Table I. Structural statisticsa of IGF-II

R.m.s. deviations from experimental distance constraints Number of experimental distance constraints violated by R.m.s. deviations from experimental dihedral constraints Number of experimental dihedral constraints violated by

(A) (636)b >0.4 A (0) (8)b >20

FNOE (kcal.mol -)C Ftor (kcal.mol-I)C F,rpe, (kcal.mol- I)C

EL-J (kcal.mol -)d

RMS deviations from idealized geometry bonds (A) (1032) angles (0) (1850) impropers (0) (452)



(SA)r

0.082 ± 0.002 (max. 0.55) 3.8 1.32 + 2.08 (max. 14.0) 0.5 211.5 ± 9.3 0.69 ± 1.35 142.3 ± 9.4 -190.9 ± 11.6

0.076 (max. 0.45) 4 0.76 (max. 2.1) 1 183.3 0.07 121.7 -180.0

0.007 ± 0.0000 2.154 ± 0.018 1.143 ± 0.017

0.006 2.133 1.096

a refers to the final set of dynamic simulated annealing structures; (SA)r refers to the structure obtained by restrained minimization of the averaged coordinate of the individual 10 structures. bThe total numbers of experimental distance constraints and dihedral constraints of Xi are 636 (including three from disulfide bridges) and eight, respectively. cThe values of the force constants used for the calculation of the square-well potentials are the original values of 50 kcal.mol-1 A-2. The van der Waals repulsion term is calculated with the original force constant 4 kcal.mol- A-4 with the van der Waals radii scaled by a factor 0.8 of the standard value used in the CHARMm empirical energy function (Brooks et al., 1983). dEL-J is the Lennard-Jones van der Waals energy calculated with the CHARMm empirical energy function (Brooks et al., 1983).

of the disulfide bonds (Cys9-Cys47, Cys2l -Cys6O, Cys46-Cys51). A total of 100 calculations were carried out, and a final set of 10 structures was selected on the critera of the smallest residual energy values of distance constraints, dihedral angle constraints and van der Waals repulsion (Kohda and Inagaki, 1992). There were no distance constraints which were violated by more than 0.5 A and no dihedral angle constraints violated by more than 200 in any structures. After fitting to the best converged structure, the 10 structures were averaged and restrainminimized to give a mean structure. Structural statistics for the converged and the mean structures are given in Table I. Small root-mean-square deviations (r.m.s.d.) of the bonds from idealized geometry (typically