WFIKKN1 and WFIKKN2 Bind Growth Factors TGF1, BMP2 and BMP4 But

WFIKKN1 and WFIKKN2 bind growth factors TGFb1, BMP2 and BMP4 but do not inhibit their signalling activity Gyo¨ rgy Szla´ ma, Katalin Konda´ s, Ma´ ria Trexler and La´ szlo´ Patthy

Institute of Enzymology, Budapest, Hungary

Keywords WFIKKN1 and WFIKKN2 are large extracellular multidomain proteins BMP; GDF11; GDF8; TGFb; WFIKKN consisting of a WAP domain, a follistatin domain, an immunoglobulin domain, two Kunitz-type protease inhibitor domains and an NTR domain. Correspondence Recent experiments have shown that both proteins have high affinity for L. Patthy, Institute of Enzymology, Budapest, Karolina ut 29, Hungary growth and differentiation factor (GDF)8 and GDF11. Here we study the Fax: +361 466 5465 interaction of WFIKKN proteins with several additional representatives of Tel: + 361 209 3537 the transforming growth factor (TGF)b family using SPR measurements. E-mail: [email protected] Analyses of SPR sensorgrams suggested that, in addition to GDF8 and GDF11, both WFIKKN proteins bind TGFb1, bone morphogenetic pro- (Received 1 June 2010, revised 5 October )6 tein (BMP)2 and BMP4 with relatively high affinity (Kd 10 M). To 2010, accepted 8 October 2010) assess the biological significance of these interactions we studied the effect doi:10.1111/j.1742-4658.2010.07909.x of WFIKKN proteins on the activity of GDF8, GDF11, TGFb1, BMP2 and BMP4 using reporter assays. These studies revealed that WFIKKN1 and WFIKKN2 inhibited the biological activity of GDF8 and GDF11 in the nanomolar range, whereas they did not inhibit the activities of TGFb1, BMP2 and BMP4 even in the micromolar range. Our data indicate that WFIKKN proteins are antagonists of GDF8 and GDF11, but in the case of TGFb1, BMP2 and BMP4 they function as growth factor binding proteins. It is suggested that the physical association of WFIKKN proteins with these growth factors may localize their action and thus help to establish growth factor gradients in the extracellular space.

Structured digital abstract l A list of the large number of protein-protein interactions described in this article is available via the MINT article MINT-8044119

family are usually assigned to three main subfamilies: Introduction activins, TGFbs and bone morphogenic proteins Growth factors of the transforming growth factor b (BMPs) ⁄ growth and differentiation factors (GDFs). (TGFb) family regulate many cellular processes, includ- TGFb family members are secreted as large precursor ing cell proliferation, differentiation and lineage deter- proteins and the mature growth factors are released from mination. In humans, more than 30 structurally related these precursors through cleavage by furin-type prote- proteins belong to this family [1]. Members of this protein ases. In several cases, the prodomain and the mature

Abbreviations ACRIIB, activin receptor IIB; BMP2, bone morphogenetic protein 2; BMP3, bone morphogenetic protein 3; BMP4, bone morphogenetic protein 4; BMP8b, bone morphogenetic protein 8b; BMP11, bone morphogenetic protein 11 or growth and differentiation factor 11; BMPRIA, bone morphogenetic protein receptor IA; ECD, extracellular domain; GDF11, growth and differentiation factor 11 or bone morphogenetic protein 11; GDF8, growth and differentiation factor 8 or myostatin; TGFb1, transforming growth factor b1; TGF-bsRII, recombinant protein corresponding to the extracellular domain of TGFb1 receptor TGF-bbRII; WFIKKN1 and WFIKKN2 – WAP, follistatin, immunoglobulin, kunitz and netrin domain containing protein 1 and 2.

5040 FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS G. Szla´ma et al. WFIKKNs bind several members of the TGFb family disulfide-bonded homodimer growth factor remain asso- in the micromolar range. Our data suggest that ciated after proteolytic cleavage [2–4]. The prodo- WFIKKN proteins may function not only as antago- main ⁄ growth factor complexes confer latency on the nists of GDF8 and GDF11, but also as proteins that growth factors and the active homodimeric growth fac- localize the action of growth factors. tors may be liberated from the latent complexes through degradation of the propeptides by proteases [4–6]. Results TGFb family proteins signal through type I and type II serine–threonine kinase receptors; in verte- Characterization of the interaction of WFIKKN1 brates, seven type I receptors and five type II receptors and WFIKKN2 with BMP2, BMP3, BMP4, BMP8b have been identified [1]. Homodimeric growth factors and TGFb1 by SPR bind to two type I and two type II receptors to form a hexameric signalling complex. In these complexes, SPR analyses suggested that both WFIKKN proteins type II receptors phosphorylate a short segment of may bind BMP2, BMP3, BMP4, BMP8b and TGFb1 type I receptors, which in turn phosphorylate down- (Fig. 1), although the affinities of WFIKKN1 and stream targets [7,8]. WFIKKN2 for these growth factors are significantly The number of growth factors available for signal- lower than those determined for GDF8 and GDF11 ling is tightly regulated by several, structurally differ- (Table 1). No interaction was detected with activin A, ent antagonists that, by interacting with the growth even when high (up to 4 lm) concentrations of factors, alter or diminish their binding to the receptors. WFIKKNs were injected on the surface of immobi- Similar to the prodomain in latent complexes, inhibi- lized activin A. tory proteins, like chordin, noggin, follistatin, follista- The Kd values calculated for the interactions of ) tin-related protein and gremlin bind various members WFIKKN1 with BMP2 (7.2 · 10 7 m), BMP3 (3.3 · ) ) ) of the TGFb family with high affinity and block their 10 6 m), BMP4 (8.2 · 10 7 m), TGFb1 (4.5 · 10 7 m) interaction with their receptors [9]. or for the interactions of WFIKKN2 with BMP2 ) ) ) Recent studies have expanded the list of TGFb (4.3 · 10 8 m), BMP3 (1.8 · 10 7 m), BMP4 (6.5 · 10 8 ) antagonists to include WFIKKN1 and WFIKKN2 m), TGFb1 (2.8 · 10 8 m) were suggestive of relatively proteins: these proteins bind GDF8 (myostatin) and high affinities, raising the possibility that these interac- GDF11 (BMP11) with high affinity [10]. WFIKKN tions may have biological importance. proteins are large extracellular multidomain proteins It should be noted, however, that there was a major that contain a WAP domain, a Follistatin ⁄ Kazal additional difference between sensorgrams obtained domain, an immunoglobulin domain, two Kunitz-type with GDF8 ⁄ GDF11 and TGFb1 ⁄ BMP2 ⁄ BMP3⁄ BMP4⁄ protease inhibitor domains and an NTR domain BMP8b: the former gave good fits with the simple [11,12]. The fact that, in luciferase reporter assays, model of a 1 : 1 Langmuir interaction [10], whereas the WFIKKN2 inhibited the activity of myostatin and association and dissociation curves of the interaction of GDF11 [13] suggests that WFIKKNs may play crucial WFIKKN proteins with the various BMPs gave accept- roles in the regulation of processes (muscle growth, able fits only with the model of ‘two state reaction with anterior ⁄ posterior patterning of the axial skeleton, etc.) conformational change’. The association and dissocia- that are under the control of these growth factors. tion curves of the interaction of WFIKKN proteins WFIKKN proteins are, however, expressed in with TGFb1 could be fitted to the model of ‘heteroge- numerous tissues other than those controlled by GDF8 neous ligand parallel reaction’ (see Experimental proce- or GDF11. For example, the WFIKKN1 gene is dures). Because it has been pointed out recently in a expressed in pancreas, thymus, liver, kidney, lung, tes- critical review of the biosensor literature that parame- tis and inner ear, and the WFIKKN2 gene is expressed ters calculated with the 1 : 1 interaction model are most in ovary, testis, pancreas, brain and lung [11,12,14], likely to give reliable estimates of binding constants raising the possibility that the proteins may have addi- [15], it may be doubtful whether the Kd values calcu- tional functions. To investigate this possibility, we used lated for the interaction of WFIKKN proteins with SPR and luciferase reporter assays to study the inter- TGFb1, BMP2, BMP3, BMP4, BMP8b are valid, and action of WFIKKN1 and WFIKKN2 proteins with whether WFIKKNs are efficient inhibitors of the bind- several representatives of the TGFb family. ing of these growth factors to their cognate receptors. Analyses of SPR sensorgrams have shown that both To answer these questions, we studied the ability of WFIKKN proteins bind TGFb1, BMP2 and BMP4 WFIKKN proteins to block the binding of growth fac- )6 with relatively high affinity (Kd 10 m), but in tors to their receptors using SPR in a solution-competi- reporter assays they do not inhibit their activities, even tion format, as well as luciferase reporter assays.

FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS 5041 WFIKKNs bind several members of the TGFb family G. Szla´ma et al.

WFIKKN1-BMP4 WFIKKN2-BMP4 A B 380 100 220 RU 20 60

0 200 400 600 800 0 200 400 600 800 Time (s) Time (s)

WFIKKN1-TGFβ1WFIKKN2-TGF1 C 100 D 220 60 160

RU 100 20 40

0 200 400 600 800 0 200 400 600 800 Time (s) Time (s)

E WFIKKN1-BMP2F WFIKKN2-BMP2 700 380 460 220 RU 220 60 0 200 400 600 800 0 200 400 600 800 Time (s) Time (s)

G WFIKKN1-BMP3H WFIKKN2-BMP3 300 950

140 RU550 RU RU RU RU

150

0 200 400 600 800 0 200 400 600 800 Time (s) Time (s )

WFIKKN1-BMP8b WFIKKN2-BMP8b I J 620 80

RU 300 40 RU

0 200 400 600 800 0 200 400 600 800 Time (s) Time (s)

Fig. 1. Characterization of the interaction of WFIKKN1 and WFIKKN2 with members of the TGFb family using SPR assays. Sensorgrams of the interactions of: (A) WFIKKN1 (500 nM, 1, 1.25, 1.5 and 2 lM) with BMP4; (B) WFIKKN2 (25, 50, 100, 250, 500 and 1000 nM) with BMP4; (C) WFIKKN1 (1, 2, 3, 4 and 6 lM) with TGFb1; (D) WFIKKN2 (250 AND 500 nM, and 1, 2 and 4 lM) with TGFb1; (E) WFIKKN1 (400 and 750 nM, 2 and 3 lM) with BMP2; (F) WFIKKN2 (50, 100, 250 and 500 nM) with BMP2; (G) WFIKKN1 (500 and 750 nM, 1 and 1.5 lM) with BMP3; (H) WFIKKN2 (100, 250 and 500 nM, 1 and 2 lM) with BMP3; (I) WFIKKN1 (250, 500 and 750 nM, 1, 1.25, 1.5 and 2 lM) with BMP8b; (J) WFIKKN2 (250, 500 and 750 nM, 1 and 1.5 lM) with BMP8b. Various concentrations of WFIKKN1 or WFIKKN2 in 20 mM Hepes buffer, pH 7.5, containing 150 mM NaCl, 5 mM EDTA, 0.005% Tween 20 were injected over CM5 sensorchips containing the immobilized growth factors. For each type of experiment, one set of representative data from three parallel experiments are shown. For the sake of clarity, the concentrations of WFIKKN proteins are not indicated in the panels; in each case, the SPR response increased parallel to the increase in WFIKKN concentration.

5042 FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS G. Szla´ma et al. WFIKKNs bind several members of the TGFb family

Table 1. Kinetic parameters of the interaction of BMP2, BMP3 inhibition even at the highest (4 lm) concentration of BMP4, BMP8b, TGFb1, GDF8 and GDF11 with WFIKKN1 and WFIKKN1 (Fig. 3C,D). WFIKKN2 proved to be a WFIKKN2. The equilibrium dissociation constants of the interactions slightly more efficient inhibitor of both BMP2 and were determined from SPR measurements with BIAEVALUATION soft- BMP4 than WFIKKN1. As shown in Fig. 3E,F, ware 4.0. The Kd values of the interaction of GDF8 and GDF11 with WFIKKN proteins are taken from Konda´s et al. [10]. WFIKKN2 decreased both the recorded SPR response and the association rate; WFIKKN2 caused a 50% Interacting proteins Kd (M) Interacting proteins Kd (M) decrease in the rate of association of BMP2 and ) ) m BMP2a–WFIKKN1 7.2 · 10 7 BMP2a–WFIKKN2 4.3 · 10 8 BMP4 to the ECD of BMPRIA at 2 and 3 l , BMP3a–WFIKKN1 3.3 · 10)6 BMP3a–WFIKKN2 1.8 · 10)7 respectively. ) ) BMP4a–WFIKKN1 8.2 · 10 7 BMP4a–WFIKKN2 6.5 · 10 8 TGFb1 had high affinity for the immobilized ECD BMP8ba–WFIKKN1 3.0 · 10)5 BMP8ba–WFIKKN2 5.3 · 10)5 of TGFbRII; based on analyses of sensorgrams, the a )7 a )8 TGFb1 –WFIKKN1 4.5 · 10 TGFb1 –WFIKKN2 2.8 · 10 interaction is characterized by a K value of )5 )5 d 8.9 · 10 3.3 · 10 )11 ) ) 5 · 10 m (Table 2). Preincubation of TGFb1 with GDF8a–WFIKKN1 3.3 · 10 8 GDF8a–WFIKKN2 2.8 · 10 10 ) ) increasing concentrations of WFIKKN2 caused a 50% GDF11a–WFIKKN1 2.2 · 10 9 GDF11a–WFIKKN2 1.6 · 10 10 decrease in the rate of association to its receptor at a These proteins were immobilized on the sensorchips. 1 lm. In the case of WFIKKN1, even the highest concentration used (2 lm) caused only a 20% decrease in the rate of association of the growth factor to its Effect of WFIKKN1 and WFIKKN2 on binding of receptor (data not shown). GDF8, GDF11, BMP2, BMP4 and TGFb1 to the extracellular domains of their cognate receptors Effect of WFIKKN1 and WFIKKN2 on growth As shown in Fig. 2A, GDF8 binds tightly to the extra- factor activity of GDF8, GDF11, BMP2, BMP4 cellular domain (ECD) of its receptor, activin recep- )9 and TGFb1 tor IIB (ACRIIB), with a Kd value of 1.6 · 10 m, comparable with that determined for intact receptors As shown in Fig. 4A, both WFIKKN1 and WFIKKN2 present on cells (Table 2). Preincubation of GDF8 efficiently inhibited the activity of GDF8 in luciferase with increasing concentrations of WFIKKN1 (Fig. 2B) reporter assays, half-maximal inhibition being achieved or WFIKKN2 (Fig. 2C) efficiently decreased the by 6nm WFIKKN1 and by 3nm WFIKKN2. recorded SPR response and association rate, indicating Similarly, WFIKKN1 and WFIKKN2 inhibited the that GDF8–WFIKKN1 and GDF8–WFIKKN2 com- activity of GDF11; 50 nm WFIKKN1 and WFIKKN2 plexes formed are unable to bind to the ECD of the caused 80% and 90% inhibition, respectively (data not receptor protein. Analysis of the data revealed that shown). By contrast with GDF8 and GDF11, in the WFIKKN1 and WFIKKN2 caused a 50% decrease in case of TGFb1, BMP2 and BMP4, neither WFIKKN1 the rate of association of GDF8 to the ECD of ACRI- nor WFIKKN2 was able to cause inhibition even at IB at 20 and 12 nm, respectively. GDF11 also the highest concentrations (1 lm) used (Fig. 4B–D). binds tightly to the ECD of its receptor (Fig. 2D), with ) a K value of 4.8 · 10 10 m (Table 2). Preincubation of d Discussion GDF11 with increasing concentrations of WFIKKN1 (Fig. 2E) or WFIKKN2 (Fig. 2F) efficiently decreased SPR studies on the interaction of WFIKKN1 and the recorded SPR response and association rate: WFIKKN2 proteins have raised the possibility that WFIKKN1 and WFIKKN2 caused a 50% decrease both proteins may bind several members of the TGFb in the rate of association of GDF11 to the ECD of family: their affinities for ligands were found to decrease ACRIIB at 40 and 5nm, respectively. in the order GDF8 ⁄ GDF11 >> TGFb1 ⁄ BMP2 ⁄ In similar experiments, we investigated the effect of BMP4 > BMP3 > BMP8b, with no detectable affinity WFIKKN1 and WFIKKN2 on the binding of BMP2 for activin A (Fig. 1 and Table 1). The lack of affinity and BMP4 to the ECD of their high-affinity receptor of WFIKKNs for activin A is in harmony with the bone morphogenetic protein receptor IA (BMPRIA). observation that, in luciferase reporter assays, As shown in Fig. 3, both BMP2 (Fig. 3A) and BMP4 WFIKKN2 had no effect on the activity of activin [13]. (Fig. 3B) had high affinity for the ECD of BMPRIA, Our observation, however, that both WFIKKN1 and with Kd values comparable with those determined by WFIKKN2 appeared to bind TGFb1, BMP2 and others (Table 2). BMP4 with relatively high affinity in SPR experiments Preincubation of BMP2 or BMP4 with increasing does not necessarily mean that they are efficient inhibi- concentrations of WFIKKN1 resulted only in weak tors of these growth factors. The use of SPR to

FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS 5043 WFIKKNs bind several members of the TGFb family G. Szla´ma et al.

A GDF8 D GDF11

15 15 10 10

RU 5 RU 5

0 0

50 150 250 350 450 550 650 750 50 150 250 350 450 550 650 Time (s) Time (s)

B GDF8-WFIKKN1 E GDF11-WFIKKN1 15 15

5 10

RU 0 RU 5 0 –10

50 150 250 350 450 550 650 750 50 150 250 350 450 550 650 Time (s) Time (s)

C GDF8-WFIKKN2 F GDF11-WFIKKN2

15 15 10 5

RU 0 RU 5 0 –10 50 150 250 350 450 550 650 50 150 250 350 450 550 650 750 850 Time (s) Time (s)

Fig. 2. Effect of WFIKKN1 and WFIKKN2 on the binding of GDF8 and GDF11 with the ECD of ACRIIB monitored using SPR. Sensorgrams of the interactions of immobilized ECD of ACRIIB with: (A) GDF8, 5, 10, 20 and 50 nM; (B) 50 nM GDF8 preincubated with 0, 25, 50, 100 and 250 nM WFIKKN1; (C) 50 nM GDF8 preincubated with 0, 12.5, 25, 30 and 50 nM WFIKKN2; (D) GDF11, 2, 2.5, 5, 7.5 and 10 nM; (E) 10 nM GDF11 preincubated with 0, 5, 10, 50, 100 and 200 nM WFIKKN1; (F) 10 nM GDF11 preincubated with 0, 0.5, 1, 2, 4, 6 and 8 nM WFIKKN2. Various concentrations of WFIKKN1 or WFIKKN2 and 50 nM GDF8 or 10 nM GDF11 were preincubated in 20 mM Hepes buffer, pH 7.5, containing 150 mM NaCl, 5 mM EDTA, 0.005% Tween 20 for 30 min at room temperature and were injected over CM5 sensorchips containing immobilized ECD of ACRIIB. For the sake of clarity, the concentrations of the proteins injected over the sensorchips are not indicated; in (A) and (D) the SPR response increased parallel to the increase in GDF8 and GDF11 concentration, respectively. In (B), (C), (E) and (F) the SPR response decreased parallel to the increase in WFIKKN concentration.

determine dissociation constants of protein–protein factors to their receptors using SPR solution-competi- interactions has numerous pitfalls, including problems tion assay formats as well as reporter assays. associated with the immobilization of one of the inter- These studies have shown that the conclusions acting partners. As pointed out by Rich and Myszka drawn from the different types of assays are in agree- [15], parameters calculated with the 1 : 1 Langmuir ment in the case of GDF8 and GDF11. For example, interaction model are most likely to give reliable esti- the high afﬁnity of WFIKKN1 for GDF8 determined mates of binding constants, therefore that the interac- with SPR (Kd 33 nm) is in agreement with its tion of WFIKKN proteins with TGFb1, BMP2, BMP3, efﬁciency in inhibiting the binding of GDF8 to its

BMP4, BMP8b gave acceptable ﬁts only with alterna- receptor in vitro (IC50 20 nm) and its ability to block tive models raised doubts about whether the dissocia- the activity of GDF8 in reporter assays (IC50 6nm). tion constants are valid and, consequently, whether the Similarly, the efﬁciency of WFIKKN2 to inhibit interactions detected by SPR have physiological rele- the binding of GDF8 to its receptor in vitro (IC50 vance. To overcome this problem, we studied the ability 12 nm) is in agreement with its ability to block the of WFIKKN proteins to block the binding of growth activity of GDF8 in reporter assays (IC50 3nm).

5044 FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS G. Szla´ma et al. WFIKKNs bind several members of the TGFb family

Table 2. Interaction of GDF8, BMP2 and BMP4 with immobilized proteins may control the action of growth factors not extracellular domains of their receptors. The equilibrium dissocia- only by inhibiting their action, but also, by serving as tion constants of the interactions were determined from SPR mea- a reservoir for growth factors, may localize their action surements with BIAEVALUATION software 4.0. For comparison K d in the vicinity of the binding proteins and thus help to values determined by others for the interaction of the growth factors with ECDs of receptors or intact receptors present on cell establish growth factor gradients in the extracellular surfaces are also shown. space through physical association. As a result, the same growth factor binding protein Interacting proteins Kd (M) References may serve as either an agonist or antagonist of a given GDF8–cell surface 1.4 · 10)10 3 growth factor in a context-dependent manner [16–20]. ) GDF8–ECD ACRIIB 1.6 · 10 9 It may be relevant in this respect that WFIKKN1 was GDF11–ECD ACRIIB 4.8 · 10)10 shown to be preferentially expressed in the developing )12 TGFb1–ECD TbRII 6 · 10 26 inner ear, mainly in the BMP4-positive presumptive TGFb1–cell surface 25 · 10)12 27 ) cristae, and it was suggested that WFIKKN1 may be TGFb1–ECD TbRII 5 · 10 11 ) involved in the early development of the inner ear BMP4–ECD BMPRIA 9.6 · 10 9 28 ) BMP4–cell surface 2.5 · 10 10 29 sensory organ by controlling the action of BMP4 [14]. BMP4–ECD BMPRIA 1.2 · 10)9 Because BMP4 not only specifies the sensory epithe- ) BMP2–ECD BMPRIA 2.6 · 10 9 30 lium of the inner ear, but also regulates its structural BMP2–ECD BMPRIA 3.3 · 10)10 development [21,22] we suggest that WFIKKN1 may influence this process by acting as a short-range diffus- ible protein that binds BMP4. The differences in the parameters in the different assays are likely to reflect true variations in the assay formats: in interactions with immobilized partner Experimental procedures versus competition in the solution phase, immobilized ECD may not properly represent the structure of the Reagents, enzymes, PCR primers, proteins, intact receptor complex present on the cell-surface. The bacterial strains, cell lines and media three types of experiments suggest that WFIKKN1 and Restriction enzymes, T4 DNA Ligase and Klenow polymer- WFIKKN2 are potent inhibitors of GDF8 and GDF11 ase were from New England Biolabs (Beverly, MA, USA). and may play a significant role in the regulation of PCR primers were obtained from Integrated DNA Technol- biological processes controlled by these growth factors. ogies (Coralville, IA, USA). For amplification reactions, we In the case of BMP2 and BMP4, solution-competi- used Taq DNA polymerase from Fermentas (Vilnius, Lith- tion assays revealed that WFIKKNs inhibit the bind- uania) or the proofreading thermostable polymerase Accu- ing of these growth factors to immobilized ECDs of zyme (Bioline, London, UK). DNA purification was their cognate receptors at higher concentrations (IC50 performed with Nucleospin Extract PCR purification kit values of 2–3 lm) than expected from the affinities of (Macherey-Nagel, Duren, Germany). Escherichia coli WFIKKNs for immobilized BMP2 and BMP4 JM109 bacterial strain was used for DNA propagation (Table 1). In reporter assays, WFIKKNs failed to inhi- during DNA manipulation steps. Mature mouse GDF8 bit the activity of BMP2 and BMP4. (O08689, GDF8_MOUSE), human GDF11 (O95390, In the case of TGFb1, solution-competition assays GDF11_HUMAN), activin A (A4D1W7, A4D1W7_HU- also indicated that WFIKKNs inhibit the binding of MAN), BMP2 (P12643, BMP2_HUMAN), BMP3 (P12645, BMP3_HUMAN), BMP4 (P12644, BMP4_HUMAN), this growth factor to its receptor at higher concentra- BMP8b (P34820, BMP8B_HUMAN), TGFb1 (P01137, tions than expected from the affinities of WFIKKNs TGFB1_HUMAN) and TGF-bsRII (P37173, TGFR2_HU- for immobilized TGFb1 (Table 1). In reporter gene MAN) (corresponding to the ECD of TGF-bRII) were assays, neither WFIKKN1 nor WFIKKN2 inhibited purchased from R&D Systems (Wiesbaden, Germany). the biological activity of this growth factor, even at CM5 sensorchips and the reagents for protein coupling to m 1 l (Fig. 4B). This observation is in agreement with the chips were from Biacore AB (Uppsala, Sweden). the finding of Hill et al. [13] that WFIKKN2 did not Recombinant WFIKKN1 (Q96NZ8, WFKN1_HUMAN) inhibit the activity of TGFb1. and WFIKKN2 (Q8TEU8, WFKN2_HUMAN) were Although WFIKKN proteins do not inhibit the sig- produced as described previously [10]. The Cignal SMAD nalling activites of BMP2, BMP4 and TGFb1, this Reporter Kit was purchased from SaBiosciences (Frederick, does not necessarily mean that the interactions of MD, USA), the firefly and Renilla luciferase kits were from WFIKKNs with BMP2, BMP4 and TGFb1 do not Biotium (Hayward, CA, USA). Rhabdomyosarcoma A204 have physiological relevance. Growth factor binding cells were from the German Collection of Microorganisms

FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS 5045 WFIKKNs bind several members of the TGFb family G. Szla´ma et al.

A BMP2B BMP4 120 250 100 200 80 150 60

RU 100 RU 40 50 20

0 0

0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800 Time (s) Time (s)

C BMP2-WFIKKN1D BMP4-WFIKKN1 200 120 100 150 80 100 60

RU RU 40 50 20

0 0

0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800 Time (s) Time (s)

E BMP2-WFIKKN2F BMP4-WFIKKN2 200 120

100 150 80

100 60 RU RU 50 40 20 0 0

0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800 Time (s) Time (s)

Fig. 3. Effect of WFIKKN1 and WFIKKN2 on the binding of BMP2 and BMP4 to the ECD of BMPRIA monitored using SPR. Sensorgrams of the interactions of immobilized ECD of BMPRIA with: (A) BMP2, 5, 10, 15 and 25 nM; (B) BMP4, 10, 15, 20 and 25 nM; (C) 15 nM BMP2 preincubated with 0, 750, 2250 and 3900 nM WFIKKN1; (D) 25 nM BMP4 preincubated with 0, 500, 2500 and 4000 nM WFIKKN1; (E) 15 nM BMP2 preincubated with 0, 750, 2250 and 3900 nM WFIKKN2; and (F) 25 nM BMP4 preincubated with 0, 500, 2500 and 4000 nM WFIKKN2. Various concentrations of WFIKKN1 or WFIKKN2 and 50 nM GDF8 were preincubated in 20 mM Hepes buffer, pH 7.5, containing 150 mM NaCl, 5 mM EDTA, 0.005% Tween 20 for 30 min at room temperature and were injected over CM5 sensorchips containing immobilized ECD of BMPRIA. For the sake of clarity, the concentrations of the proteins injected over the sensorchips are not indicated; in (A) and (B) the SPR response increased parallel to the increase in BMP concentration, in (C), (D), (E) and (F) the SPR response decreased parallel to the increase in WFIKKN concentration. and Cell Cultures (DSMZ, Braunschweig, Germany). Mink Expression of the ECD of human BMPR1a and lung epithelial cells stably transfected with a truncated PAI- AVRIIB in Pichia pastoris 1 promoter ⁄ firefly luciferase construct (MLEC-clone32) [23] and HepG2-BRA cells stably transfected with the BRE–luc The cDNA fragment coding for the extracellular domain of reporter construct [24] were generously provided by Profes- BMPR1A (P36894, BMR1A_HUMAN) was amplified from sor Daniel Rifkin (New York University). Culture media a human prostate first-strand cDNA library (Clontech, Dulbecco’s modified Eagle’s medium, McCoy’s 5A and Mountain View, CA,USA) using 5¢-GAGGAATTCCAG heat-inactivated fetal bovine serum were obtained from AATCTGGATAGTATGCTT-3¢ sense and 5¢-GAGGTCG Sigma-Aldrich (St Louis, MO, USA). ACTCGAATGCTGCCATCAAAAAACGG-3¢ antisense

5046 FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS G. Szla´ma et al. WFIKKNs bind several members of the TGFb family

A B 160

100 140 WFIKKN1 120 80 WFIKKN2 100 60 80

RLU% 40 RLU% 60 WFIKKN1 20 40 WFIKKN2 0 20

0 1020304050 0 200 400 600 800 1000 WFIKKN (nM) WFIKKN (nM)

C D 120 140 100 120

100 80

80 60 RLU% 60 RLU% WFIKKN1 40 WFIKKN1 40 WFIKKN2 WFIKKN2 20 20

0 0 0 200 400 600 800 1000 0 200 400 600 800 1000 WFIKKN (nM) WFIKKN (nM)

Fig. 4. Effect of WFIKKN1 and WFIKKN2 on the growth factor activities. (A) Rhabdomyosarcoma A204 cells transiently transfected with Cig- nal SMAD Luciferase Reporter vector and a Renilla luciferase vector were incubated for 16 h with 0.8 nM GDF8 preincubated with different concentrations of WFIKKN1 ( ) and WFIKKN2 ( ). Firefly luciferase units were normalized to Renilla luciferase units and background values obtained from control cells were subtracted. (B) Mink lung epithelial cells stably transfected with a truncated PAI-1 promoter ⁄ firefly luciferase construct (MLEC-clone32) were incubated for 15 h with 8 pM TGFb1 preincubated with different concentrations of WFIKKN1 ( ) and WFIKKN2 ( ). (C) HepG2 cells stably transfected with the BRE–luc reporter construct were incubated for 15 h with 250 pM BMP2 preincubated with different concentrations of WFIKKN1 ( ) and WFIKKN2 ( ). (D) HepG2 cells stably transfected with the BRE–luc reporter construct were incubated for 15 h with 250 pM BMP4 preincubated with different concentrations of WFIKKN1 ( ) and WFIKKN2 ( ). In the case of (B), (C) and (D), the luciferase activities were normalized to the protein content of the wells and background values obtained from control cells were subtracted. The figure shows the mean values of three parallel experiments. Error bars represent the SEM. primers. The cDNA of the ECD of human AVRIIB Pichia pastoris GS115 cells and the expression of the recom- (Q13705, AVR2B_HUMAN) was amplified with 5¢-GAGG binant protein was performed according to the protocol AATTCTCTGGGCGTGGGGAGGCTGAG-3¢ sense and described for WFIKKN proteins [10]. 5¢-GAGGTCGACCGTGAGCAGGGTGGGGGCTGT-3¢ The calculated molecular mass of the ECDs of AVRIIB antisense primers from a skeletal muscle cDNA library. Both and BMPRIA are 14 971 and 15 499 Da, respectively. reactions were performed with Accuzyme proofreading Taq SDS ⁄ PAGE analysis of the proteins purified from the DNA polymerase over 35 reaction cycles. The annealing induction media showed diffuse bands with molecular temperature was 60 C for amplification of the ECD of masses higher than the expected, suggesting that the recom- BMPRIA and 63.5 C in the case of the ECD of AVRIIB. binant proteins may be glycosylated. Deglycosylation of the In both cases, amplified DNAs were digested with EcoRI proteins by EndoH digestion decreased the molecular and SalI restriction enzymes and ligated into pPiczalphaA mass of the proteins to 15–16 kDa. The N-terminal Pichia pastoris expression vector digested with the same sequence of the recombinant ECD of BMPRIA was enzymes. The introduction of the linearized pPICZal- EFQNLDSMLHGT and that of the recombinant ECD of phaA_BMPR1A and pPICZalphaA_AVRIIB plasmids into AVRIIB was EFSGRGEAETRE (residues in bold corre-

FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS 5047 WFIKKNs bind several members of the TGFb family G. Szla´ma et al. spond to residues of the ECDs, the N-terminal residues EF 150 mm NaCl, 5 mm EDTA, 0.005% Tween 20 pH 7.5 originate from the expression constructs). buffer. After each cycle the chips were regenerated with The structural integrity and stability of recombinant 20 mm Hepes, 150 mm NaCl, 5 mm EDTA, 0.005% ECDs was checked by CD spectroscopy. CD spectra were Tween 20, pH 7.5 buffer containing 8 m urea. measured over the range 195–250 nm by using a JASCO In solution-competition assays, constant concentrations J-720 spectropolarimeter thermostatted with a Neslab RT-111 of growth factors were incubated with increasing concentra- water bath. The measurements were carried out in 1 mm tions of WFIKKN1 or WFIKKN2 in 20 mm Hepes, pathlength cells and protein solutions of 0.1 mgÆmL)1 in 150 mm NaCl, 5 mm EDTA, 0.005% Tween 20 pH 7.5 buf- 10 mm Tris ⁄ HCl buffer, pH 8.0. Spectra were measured at fer for 30 min at room temperature prior to injection on ) 25 C with a 16 s time constant and 20 nmÆmin 1 scan rate. chips with immobilized ECDs of growth factor receptors. The spectral slit width was 1.0 nm. The spectra of the Control flow cells were prepared by executing the cou- recombinant proteins were also recorded at different temper- pling reaction in the presence of coupling buffer alone. atures from 25 to 95 Cat10C intervals. The thermal Control flow cells were used to obtain control sensorgrams unfolding of the recombinant ECD of BMPRIA and the showing nonspecific binding to the surface as well as refrac- ECD of AVRIIB were monitored at 213 and 230 nm, tive index changes resulting from changes in the bulk prop- respectively, where the difference of the CD spectra recorded erties of the solution. Control sensorgrams were subtracted at different temperatures was the largest. The heating rate from sensorgrams obtained with immobilized ligand. To ) was 60 CÆh 1. The native fold of both recombinant ECDs correct for differences between the reaction and reference collapsed with single melting temperatures: Tm =72C for surfaces, we also subtracted the average of sensorgrams the ECD of BMPRIA and 67 C for the ECD of AVRIIB. obtained with blank running buffer injections. The kinetic parameters for each interaction were determined by globally fitting the experimental data with bia- Protein analyses evaluation software 4.0 and the closeness of the fit was The composition of protein samples was analysed by characterized by the chi-square values. Fits were accepted 2 SDS ⁄ PAGE under both reducing and nonreducing condi- only if the v values were < 5% of Rmax [25]. The associa- tions. The gels were stained with Coomassie Brilliant Blue tion and dissociation curves of the interaction of GDF8, G-250. The concentration of the recombinant proteins was GDF11, BMP2 and BMP4 with the ECDs of their recep- determined using the following extinction coefficients: tors gave good fits with the model of 1 : 1 Langmuir inter- WFIKKN1, 64 440 m)1Æcm)1; WFIKKN2, 57 470 m)1Æcm)1; action. The sensorgrams of the interaction of WFIKKN ECD of BMPRIA, 5095 m)1Æcm)1; ECD of AVRIIB, proteins with BMPs, however, gave acceptable fits only 26 065 m)1Æcm)1. The extinction coefficients were calculated with the model of ‘two state reaction with conformational with the online protein analysis tool protparam. change’. In the case of the interaction of WFIKKN proteins with TGFb1, the data gave acceptable fits with the model ‘of heterogeneous ligand parallel reaction’. SPR analysis

SPR measurements were performed on a BIACORE X (GE Cell culture Healthcare, Stockholm, Sweden) instrument. Proteins to be immobilized were dissolved in 50 mm sodium acetate, pH Rhabdomyosarcoma A204 cells were cultured in McCoy’s 4.5, and 100 lL of 0.7 lm activin A or 50 lL of 0.8 lm 5A medium supplemented with 10% fetal bovine serum, )1 )1 BMP2 or 100 lL of 0.83 lm BMP3 or 75 lL of 0.7 lm penicillin (100 UÆmL ) and streptomycin (100 lgÆmL )at BMP8b or 50 lL of 0.8 lm BMP4 solutions were injected 37 C, 5% CO2. Mink lung epithelial cells and HepG2- with a 5 lLÆmin)1 flow rate on a CM5 sensor chip activated BRA cells were cultured in Dulbecco’s modified Eagle’s by the amine coupling method, according to the manufac- medium supplemented with 10% fetal bovine serum, peni- )1 )1 turer’s instructions. TGFb1 was dissolved in 50 mm sodium cillin (100 UÆmL ) streptomycin (100 lgÆmL ) and geneti- )1 acetate, pH 4.1, and 42 lL of a 0.8 lm solution was cin at a concentration of 200 lgÆmL (MLEC-clone32) or )1 )1 injected with a 3 lLÆmin flow rate. The ECDs of ACRIIB 700 lgÆmL (HEPG2-BRA) at 37 C, 5%CO2. and BMPRIA were dissolved in 50 mm sodium acetate, pH 4.0 or 4.2 respectively, and 100 lL of 17.5 lm solutions Reporter assays were injected with 5 lLÆmin)1 flow rate. TGFbsRII was dissolved in 50 mm sodium acetate, pH 4.2, and 100 lLof TGFb1 activity was measured with MLEC-clone32 cells, 6.6 lm solution was injected with 5 lLÆmin)1 flow rate. whereas the activities of BMP2 and BMP4 were monitored For interaction measurements, 80-lL aliquots of pro- with HEPG2-BRA cells, using 96-well tissue culture dishes. tein solutions were injected over the sensor chips with a In these reporter assays MLEC-clone32 cells (1.6 · 104 ) ) ) 20 lLÆmin 1 flow rate, followed by 10 min flow of buffer. cellsÆwell 1) or HEPG2-BRA cells (5 · 103 cellsÆwell 1) were Binding and washes were performed in 20 mm Hepes, allowed to attach for 3 or 24 h respectively, then the

5048 FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS G. Szla´ma et al. WFIKKNs bind several members of the TGFb family medium was changed to Dulbecco’s modified Eagle’s med- References ium supplemented with 0.1% BSA, penicillin (100 UÆmL)1) streptomycin (100 lgÆmL)1) containing 8 pm TGFb1, 1 de Caestecker M (2004) The transforming growth fac- 250 pm BMP2 or 250 pm BMP4, preincubated for 30 min tor-beta superfamily of receptors. Cytokine & Growth with different concentrations of WFIKKN1 or WFIKKN2. Factor Rev 15, 1–11. Control experiments were performed similarly, except that 2 Miyazono K & Heldin CH (1991) Latent forms of no growth factor was added. TGF-beta: molecular structure and mechanisms of activation. Ciba Found Symp 157, 81–89. After incubation for 15 h at 37 C, 5% CO2, cells were lysed in 100 lL lysis buffer and the luciferase activity of the 3 Thies RS, Chen T, Davies MV, Tomkinson KN, Pear- samples was determined using the firefly luciferase assay kit son AA, Shakey QA & Wolfman NM (2001) GDF-8 of Biotium on an Appliskan luminometer (Thermo Electron propeptide binds to GDF-8 and antagonizes biological Corp., Beverly, MA, USA). The protein content of the activity by inhibiting GDF-8 receptor binding. Growth samples was determined with the Bio-Rad protein assay Factors 18, 251–259. (Bio-Rad, Hercules, CA, USA) and the luciferase activity 4 Ge G, Hopkins DR, Ho WB & Greenspan DS (2005) was normalized to the protein content of the wells. GDF11 forms a bone morphogenetic protein 1-acti- The activities of GDF8 and GDF11 were studied on vated latent complex that can modulate nerve growth Rhabdomyosarcoma A204 cells. 2.5–3 · 104 cells were pla- factor-induced differentiation of PC12 cells. Mol Cell ted in wells of a 96-well plate and were allowed to attach Biol 25, 5846–5858. for 24 h, then transiently transfected with 150 ng Cignal 5 Jenkins G (2008) The role of proteases in transforming SMAD Luciferase Reporter vector mixture with 0.4 lL growth factor-beta activation. Int J Biochem Cell Biol Lipofectamin 2000 reagent per well, according to the manu- 40, 1068–1078. facturer’s instructions. 6 Ge G & Greenspan DS (2006) Developmental roles of Transfections were performed in serum-free McCoy’s 5A the BMP1 ⁄ TLD metalloproteinases. Birth Defects Res medium containing 1 mgÆmL)1 BSA without antibiotics. C Embryo Today 78, 47–68. The plasmid preparation used for transfections was a 40 : 1 7 Wrana JL (1998) TGF-beta receptors and signalling ratio mixture of an inducible, TGFb-responsive firefly lucif- mechanisms. Miner Electrolyte Metab 24, 120–130. erase construct and a constitutively expressing Renilla lucif- 8 ten Dijke P, Miyazono K & Heldin CH (1996) Signal- erase construct. Sixteen hours later, the transfection ing via hetero-oligomeric complexes of type I and medium was changed to McCoy’s 5A containing type II serine ⁄ threonine kinase receptors. Curr Opin 1mgÆmL)1 BSA and conditioned for 5 h. WFIKKN1 and Cell Biol 8, 139–145. WFIKKN2 were incubated with 0.8 nm growth factors for 9 Rosen V (2006) BMP and BMP inhibitors in bone. 30 min at 37 C and were added to the cells. After 16 h, Ann NY Acad Sci 1068, 19–25. 10 Konda´s K, Szla´ma G, Trexler M & Patthy L (2008) the cells were washed with NaCl ⁄ Pi and lysed using 50 lL passive lysis buffer from the Biotium luciferase kit, and fire- Both WFIKKN1 and WFIKKN2 have high affinity for fly and Renilla luciferase activities were measured. growth and differentiation factors 8 and 11. J Biol The firefly luciferase units obtained were normalized to Chem 283, 23677–23684. the Renilla luciferase units and background values obtained 11 Trexler M, Bańyai L & Patthy L (2001) A human pro- from cells grown in McCoy’s 5A medium–1 mgÆmL)1 BSA tein containing multiple types of protease-inhibitory alone were subtracted to generate relative luciferase units. modules. Proc Natl Acad Sci USA 98, 3705–3709. Three parallel experiments were performed in all cases 12 Trexler M, Bańyai L & Patthy L (2002) Distinct expres- and were repeated at least twice. Control experiments were sion pattern of two related human proteins containing also performed to check whether WFIKKNs have any multiple types of protease-inhibitory modules. Biol influence on luciferase activity in the absence of added Chem 383, 223–228. growth factor. 13 Hill JJ, Qiu Y, Hewick RM & Wolfman NM (2003) Regulation of myostatin in vivo by growth and differentiation factor-associated serum protein-1: a novel pro- Acknowledgements tein with protease inhibitor and follistatin domains. Mol Endocrinol 17, 1144–1154. This work was supported by grant 72125 of the 14 Nishida AT, Kobuke K, Kojima K, Ito J, Honjo T & National Scientific Research Fund of Hungary Tashiro K (2004) OC29 is preferentially expressed in (OTKA) and by grant RET14 ⁄ 2005 of the National the presumptive sensory organ region of the otocyst. Office for Research and Technology of Hungary Dev Dynam 231, 766–774. (NKTH). The authors thank Professor Daniel Rifkin 15 Rich RL & Myszka DG (2010) Grading the commercial (New York University) for the generous gift of optical biosensor literature – class of 2008: ‘the mighty HepG2-BRA and MLEC-clone32 cells. binders’. J Mol Recognit 23, 1–64.

FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS 5049 WFIKKNs bind several members of the TGFb family G. Szla´ma et al.

16 Ashe HL & Levine M (1999) Local inhibition and long- 24 Zilberberg L, ten Dijke P, Sakai LY & Rifkin DB range enhancement of Dpp signal transduction by Sog. (2007) A rapid and sensitive bioassay to measure Nature 398, 427–431. bone morphogenetic protein activity. BMC Cell Biol 8, 17 Decotto E & Ferguson EL (2001) A positive role for 41. Short gastrulation in modulating BMP signaling during 25 Lin X, Takahashi K, Campion SL, Liu Y, Gustavsen dorsoventral patterning in the Drosophila embryo. GG, Penã LA & Zamora PO (2006) Synthetic peptide Development 128, 3831–3841. F2A4-K-NS mimics fibroblast growth factor-2 in vitro 18 Shimmi O, Ralston A, Blair SS & O’Connor MB (2005) and is angiogenic in vivo. Int J Mol Med 17, 833–839. The crossveinless gene encodes a new member of the 26 De Crescenzo G, Pham PL, Durocher Y & O’Connor- Twisted gastrulation family of BMP-binding proteins McCourt MD (2003) Transforming growth factor-beta which, with Short gastrulation, promotes BMP signal- (TGF-b) binding to the extracellular domain of the ing in the crossveins of the Drosophila wing. Dev Biol type II TGF-b receptor: receptor capture on a biosensor 282, 70–83. surface using a new coiled-coil capture system demon- 19 Rentzsch F, Zhang J, Kramer C, Sebald W & strates that avidity contributes significantly to high Hammerschmidt M (2006) Crossveinless 2 is an affinity binding. J Mol Biol 328, 1173–1183. essential positive feedback regulator of Bmp signaling 27 Lin HY, Moustakas A, Knaus P, Wells RG, Henis YI during zebrafish gastrulation. Development 133, & Lodish HF (1995) The soluble exoplasmic domain of 801–811. the type II transforming growth factor (TGF)-beta 20 Fung WY, Fat KF, Eng CK & Lau CK (2007) crm-1 receptor. A heterogeneously glycosylated protein with facilitates BMP signaling to control body size in high affinity and selectivity for TGF-beta ligands. J Biol Caenorhabditis elegans. Dev Biol 311, 95–105. Chem 270, 2747–2754. 21 Chang W, Nunes FD, De Jesus-Escobar JM, Harland 28 Natsume T, Tomita S, Iemura S, Kinto N, Yamaguchi R & Wu DK (1999) Ectopic noggin blocks sensory and A & Ueno N (1997) Interaction between soluble type I nonsensory organ morphogenesis in the chicken inner receptor for bone morphogenetic protein and bone mor- ear. Dev Biol 216, 369–381. phogenetic protein-4. J Biol Chem 272, 11535–11540. 22 Gerlach LM, Hutson MR, Germiller JA, Nguyen-Luu 29 Koenig BB, Cook JS, Wolsing DH, Ting J, Tiesman JP, D, Victor JC & Barald KF (2000) Addition of the Correa PE, Olson CA, Pecquet AL, Ventura F, Grant BMP4 antagonist, noggin, disrupts avian inner ear RA et al. (1994) Characterization and cloning of a development. Development 127, 45–54. receptor for BMP-2 and BMP-4 from NIH 3T3 cells. 23 Abe M, Harpel JG, Metz CN, Nunes I, Loskutoff DJ Mol Cell Biol 14, 5961–5974. & Rifkin DB (1994) An assay for transforming growth 30 Allendorph GP, Isaacs MJ, Kawakami Y, Belmonte JC factor-beta using cells transfected with a plasminogen & Choe S (2007) BMP-3 and BMP-6 structures illumi- activator inhibitor-1 promoter–luciferase construct. Anal nate the nature of binding specificity with receptors. Biochem 216, 276–284. Biochemistry 46, 12238–12247.

5050 FEBS Journal 277 (2010) 5040–5050 ª 2010 The Authors Journal compilation ª 2010 FEBS