Structural basis of the transmembrane domain dimerization and rotation in the activation mechanism of the TRKA receptor by nerve growth factor

Tropomyosin-receptor kinases (TRKs) are essential for the development of the nervous system. The molecular mechanism of TRKA activation by its ligand nerve growth factor (NGF) is still unsolved. Recent results indicate that at endogenous levels most of TRKA is in a monomer–dimer equilibrium and that the binding of NGF induces an increase of the dimeric and oligomeric forms of this receptor. An unsolved issue is the role of the TRKA transmembrane domain (TMD) in the dimerization of TRKA and the structural details of the TMD in the active dimer receptor. Here, we found that the TRKA–TMD can form dimers, identified the structural determinants of the dimer interface in the active receptor, and validated this interface through site-directed mutagenesis together with functional and cell differentiation studies. Using in vivo cross-linking, we found that the extracellular juxtamembrane region is reordered after ligand binding. Replacement of some residues in the juxtamembrane region with cysteine resulted in ligand-independent active dimers and revealed the preferred dimer interface. Moreover, insertion of leucine residues into the TMD helix induced a ligand-independent TRKA activation, suggesting that a rotation of the TMD dimers underlies NGF-induced TRKA activation. Altogether, our findings indicate that the transmembrane and juxtamembrane regions of TRKA play key roles in its dimerization and activation by NGF.

Tropomyosin-receptor kinases (TRKs) are essential for the development of the nervous system. The molecular mechanism of TRKA activation by its ligand nerve growth factor (NGF) is still unsolved. Recent results indicate that at endogenous levels most of TRKA is in a monomer-dimer equilibrium and that the binding of NGF induces an increase of the dimeric and oligomeric forms of this receptor. An unsolved issue is the role of the TRKA transmembrane domain (TMD) in the dimerization of TRKA and the structural details of the TMD in the active dimer receptor. Here, we found that the TRKA-TMD can form dimers, identified the structural determinants of the dimer interface in the active receptor, and validated this interface through site-directed mutagenesis together with functional and cell differentiation studies. Using in vivo cross-linking, we found that the extracellular juxtamembrane region is reordered after ligand binding. Replacement of some residues in the juxtamembrane region with cysteine resulted in ligand-independent active dimers and revealed the preferred dimer interface. Moreover, insertion of leucine residues into the TMD helix induced a ligand-independent TRKA activation, suggesting that a rotation of the TMD dimers underlies NGF-induced TRKA activation. Altogether, our findings indicate that the transmembrane and juxtamembrane regions of TRKA play key roles in its dimerization and activation by NGF.
Nerve growth factor (NGF) 4 is a member of the mammalian neurotrophin protein family implicated in the maintenance and survival of the peripheral and central nervous systems (1)(2)(3). NGF is a dimer that interacts with two distinct receptors: TRKA, a cognate member of the Trk receptor tyrosine kinase family, and the p75 neurotrophin receptor, which belongs to the tumor necrosis factor receptor superfamily of death receptors (4 -6). TRKA signaling is essential for sensory and sympathetic neuron survival during development (7). Genetic mutations in the gene that encodes TRKA, NTRK1, cause congenital insensitivity to pain with anhidrosis (8), and somatic mutations and chromosomal rearrangements generate aberrant protein fusions with constitutive kinase activation causing several types of cancer (9,10).
Despite all these important roles, the molecular mechanisms of TRKA activation have been poorly studied compared with those of other receptor-tyrosine kinase (RTK) family members (11,12). The first three extracellular domains of TRKA consist of a leucine-rich region (Trk-d1) that is flanked by two cysteinerich domains (Trk-d2 and Trk-d3). The fourth and fifth domains (Trk-d4 and Trk-d5) are Ig-like domains, and they are followed by a 30-residue-long linker that connects the extracellular portion of the receptor to the single transmembrane domain and a juxtamembrane intracellular region that is connected to the kinase domain. TRKA is activated by NGF a member of the neurotrophin family (3). The NGF-binding domain is located in the Trk-d5(Ig2) domain (13), although other domains also participate in the activation by neurotrophins through an unknown mechanism (14,15).
Two models for TRKA activation are postulated; a ligandinduced dimerization of TRKA monomers and a ligand-induced conformational activation of preformed inactive dimers. The first model, which is based on the crystal structure of NGF with the ligand-binding domain of TRKA (13,16), assumed that the dimerization of TRKA is solely ligand-mediated and that receptor-receptor interactions are not present in the absence of its ligand. In the second model, TRKA exists as a preformed inactive dimer, suggesting receptor-receptor contacts in the absence of NGF (17,18). The most recent data using singleparticle tracking (19) and FRET studies (20) suggest that TRKA is, at endogenous levels, predominantly monomer (80%), and NGF binding induces an increase and a stabilization of the TRKA dimers and the formation of oligomers, together with a conformational change leading to kinase activation. This mechanism of activation has been called the "transition model" (20) and postulates a dynamic transition from a monomer to an inactive dimer to a ligand-bound active dimer, suggesting that Trk receptors are activated through a combination of the two mentioned models.
Whatever the model, it is clear that dimerization of TRKA is required for its activation. Deletion constructs suggested that dimerization of TRKA is mediated by the transmembrane (TM) and by the intracellular domains (ICDs) (20). In the case of the ICDs, this is supported by the crystal structure of the kinase domain of TRKA that showed the presence of dimers in the crystallographic unit (21,22). However, the structural determinants of the TMD dimerization are not known, and in this regard, it is important to understand the conformation of the TRKA-TMD dimer and identify the active dimer interface that may represent the functional state of the full-length receptor. In addition biochemical data supporting a conformational activation of TRKA are lacking. In the present work, we investigated the structural basis of TRKA-TMD dimerization in the activation of TRKA by NGF, using complementary structural and biochemical approaches.

Structure of TRKA transmembrane domain dimers
It has been shown that the isolated TMDs of all human RTKs form dimers in bacterial membranes (23). In addition functional studies indicate that TMDs play an important role as a modulator of RTK homodimerization and kinase activation (reviewed in Refs. [23][24][25]. Switching between two dimerization modes of the transmembrane helix has recently been described as part of the activation mechanism of the epidermal growth factor, vascular endothelial growth factor, and fibroblast growth factor receptors (26 -29). However, to date the role of TRKA-TMD dimerization in TRKA receptor activation has not been studied in detail.
To obtain a structural insight into TRKA-TMD dimerization, we solved the structure of human TRKA-TMD dimers in detergent micelles using NMR (Fig. 1). For this study, the human TRKA-TMD was produced in a cell-free system (see "Experimental procedures") as previously described (30). When the peptide is solubilized in dodecylphosphocholine (DPC) micelles at a detergent-to-protein molar ratio (DPR) of 50:1, the TRKA-TMD is in equilibrium between monomeric, dimeric, and other oligomeric states. The ratio of these states varies as the DPR value is altered (Fig. S1A). We then titrated TRKA-  15 N HSQC spectrum of the human TRKA-TMD-wt. The human TRKA-TMD was solubilized in an aqueous suspension of DPC micelles at a DPR of 50:1 and at 45°C and pH 5.9. 1 H-15 N backbone and side-chain resonance assignments are shown. B, schematic representation of the NMR spatial structure of the human TRKA-TMD dimer in detergent micelles, from different angles. The surface of one TMD helix is represented with the NMR dimer interface in red. The TMD of the other monomer within the dimer is represented as ribbons, and side chains are indicated as sticks. The location of the NMR dimer interface LXXFAXXF on the NMR structure is colored red, and other mutated residues are colored orange. The putative dimerization motif SXXXG is shown in blue. The Protein Data Bank accession code is 2n90. C, alignment of the amino acid sequences of the transmembrane helical domain of TRKA from different species (h, Homo sapiens; r, Rattus norvegicus; m, Mus musculus; c, Gallus gallus; z, Danio rerio; x, Xenopus laevis). The residues participating in the NMR dimer interface are highlighted in red.

TRKA transmembrane domain dimerization
TMD in DPC micelles to measure the standard free energy of dimerization (⌬G 0 ) using standard methods (31) (see details under "Experimental procedures" and Fig. S1B). The ⌬G 0 value obtained (Ϫ1.9 Ϯ 0.2 kcal/mol) suggested that the TRKA-TMD dimer is quite stable compared with the TMD dimers of other RTKs. Thus, although its dimerization energy is weaker than that of the vascular endothelial growth factor receptor 2 dimer (⌬G 0 ϭ Ϫ2.5 kcal/mol in DPC) (32), it is stronger than those of fibroblast growth factor receptor 3 (⌬G 0 ϭ Ϫ1.4 kcal/ mol in DPC/SDS 9:1 mixture) and ErbB4 (⌬G 0 ϭ Ϫ1.4 kcal/mol in DMPC/DHPC 1:4 bicelles) (33). The 15 N HSQC spectrum of 15 N-labeled TRKA-TMD (Fig. 1A) contained the expected number of cross-peaks, and the good quality of the spectra allowed solving of the structure of the dimer in DPC micelles (Figs. 1B, Figs. S2-S4, and Table S1). The ␣-helical region of the TRKA-TM dimer starts at Gly 417 , ends at Asn 440 , and is ϳ38 Å in length (Fig. S2). The crossing angle of the TRKA-TM helices is 40°, and the minimal distance between two monomers is 8.8 Å ( Fig. 1B and Table S1). The hydrophobicity plot and contact surface area of the dimer is shown in Fig. S4. The dimerization interface lies along the sequence motif 424 LXXFAXXF 431 ( Fig.  1B and Fig. S4) that is conserved in the TRKA-TMD of several species and also in TrkC but not in TrkB (Fig. 1C). Although the TRKA-TMD sequence contains a putative dimerization motif of the form SXXXG (shown in blue in Fig. 1B), this motif resides in an opposite helix interface.
These analyses of TRKA-TMD dimerization suggested the existence of the dimerization motif 424 LXXFAXXF 431 (Fig. 1B). The biological relevance of this motif can be questioned, because the presence of large extracellular and intracellular globular domains or the lipid environment of the plasma membrane may favor or hinder a specific interaction interface (34). We therefore used different functional assays to verify the found dimer interface in the context of the fulllength receptor.

Functional identification of the dimer interface upon NGF stimulation
The state of full-length TRKA was followed by assay of three different aspects of its activity: dimerization of the receptor, phosphorylation of intracellular tyrosine residues, and neurite differentiation of PC12nnr5 cells. To investigate the dimerization of TRKA, we individually mutated most of the N-terminal residues of the rat TRKA-TMD to cysteine ( Fig. 2A), expressed these constructs in HeLa cells, and then measured the amount of cross-linked species. To facilitate cross-linking via these cysteine residues in the transmembrane domains, we used oxidation with molecular iodine (I 2 ) as previously described (35). Such oxidation allows the formation of a disulfide bond between two close cysteine residues inside the lipid bilayer. Plasma membrane fractions from cells expressing different single-cysteine mutants were incubated in the absence or presence of NGF, together with molecular I 2 and were then analyzed by nonreducing SDS-PAGE and Western immunoblotting. Upon transfection in HeLa cells, the mutants G417C and V418C formed covalent dimers in the absence of NGF (Fig. 2, B and C). NGF stimulation increased the amount of G417C and V418C dimers, and low quantities of dimers in the V420C, A421C, and V422C mutants were observed (Fig. 2, B and C). Overexpression of TRKA induces a ligand-independent activation. To see whether the constitutive dimerization of some of these mutants induces the activation of TRKA, we transfected the cysteine mutants in PC12nnr5 cells and studied the differentiation in NGF-independent manner. As we can see in Fig. 2D, only the mutant V418C is able to induce the differentiation in the absence of NGF, suggesting that V418C is part of the active dimer interface.
To further study the significance of the found interfaces, we mutated the small residues Ala, Gly, and Ser within this region to the bulky Ile residue, and we then assayed TRKA activation upon NGF stimulation (Fig. 2, E and F). The rationale behind this approach was that the mutation of a small residue to a bulky one on the relevant interface would prevent the formation of the active dimeric state by inducing steric clashes and would therefore reduce TRKA activation. To perform this assay, we transfected HeLa cells that do not express endogenous TRKA with these mutants, stimulated these cells with nonsaturating concentrations (10 ng/ml) of NGF, and then assayed TRKA activation by analysis of TRKA autophosphorylation using Western blotting. Upon transfection, two TRKA electrophoretic bands are present in the TRKA immunoblots of HeLa cells: a lower band (ϳ110 kDa) of intracellular immature TRKA that has not completed Golgi-mediated processing of highmannose N-glycans (36) and an upper band (ϳ140 kDa) with mature sugars that is expressed in the plasma membrane. Exposure to NGF substantially increased the phosphorylation of the upper TRKA band as assessed by blotting with a phospho-specific antibody against the phosphotyrosine residues of the activation loop, P-Tyr 674 and P-Tyr 675 . This autophosphorylation was quantified to follow TRKA activation. Constitutive (t ϭ 0, no NGF added) and ligand-dependent phosphorylation of plasma membrane-localized TRKA after 5 and 15 min were measured. Because overexpression of TRKA induces ligandindependent autophosphorylation, we first transfected the HeLa cells with increasing concentrations of TRKA to determine a TRKA level that could still be detected but that displayed no autophosphorylation in the upper band in the absence of NGF (Fig. S5). It is noteworthy that all mutants are expressed at the plasma membrane as evidenced by immunofluorescence localization in the absence of Triton X-100 using an antibody against an epitope in the TRKA N terminus (Fig.  S6) and by flow cytometry (Fig. 2F). Of the seven single-point mutants tested, only the A428I substitution demonstrated a pronounced inhibitory effect on receptor autophosphorylation (Fig. 2E). Ala 428 is the only small-chain residue that is found deep and in the closest position in the dimerization interface of the TRKA TMD structure determined using NMR, which supports the relevance of the obtained NMR structure. The inhibitory effect of A428 substitution on receptor activity was further enhanced when all three Ala residues that are at least somehow involved in the TMD dimerization in the NMR-based structure: Ala 421 , Ala 425 , and Ala 428 , were simultaneously substituted (TRKA-3A/3I).
Lastly, we studied the effect of the same mutations on the NGF-induced differentiation of transfected PC12nnr5 cells TRKA transmembrane domain dimerization ( Fig. 2H). Again, the A428I mutant displayed substantial inhibition of this TRKA activity. Unexpectedly the mutation G423I did have an impact on cell differentiation (Fig. 2F).
The residues Gly 417 , Val 420 , and Ala 421 share the same helix interface as the LXXFAXXL motif found in the NMR structure (green and yellow, respectively, in Fig. 2G). However, the resi- An arrow shows the putative rotation from the two interfaces. Bars represent the standard error of at least three independent experiments. Statistical analysis was performed using ordinary one-way ANOVA, using Bonferroni's multicomparison test compared with wt. The p values of significant differences are shown. ns, not significant.

TRKA transmembrane domain dimerization
dues Val 418 and Val 422 are in a different interface. Interestingly the mutation V418C induces a ligand-independent differentiation of PC12nnr5 cells, indicating that this residue belongs to the active dimer interface (Fig. 2G).
The combined results of the functional assays suggest a transition from an inactive to an active dimer interface and support the importance of the NMR-derived TMD structure for TRKA activation. The residue Ala 428 plays a critical role in TRKA activation by NGF. Its location in the closest dimer interface suggests a pivoting role in the transition from the ligand-free to the ligand-bound dimer interface (Fig. 2G). Because activation of TRKA is a consequence of this change in the dimer interface in the next sections, we study the nature of this conformational change induced by NGF binding.

NGF induces a rearrangement of the extracellular juxtamembrane region of TRKA
Stimulation of HeLa cells transfected with TRKA-wt with NGF induces the formation of TRKA dimers that are crosslinked with BS3 (Fig. 3). Although it has been described that TRKA dimers are formed in the absence of NGF, we were not able to detect cross-linking without the ligand, even at overexpression levels, supporting that NGF binding is not only promoting a TMD dimerization but is accompanied by changes in the conformation of the extracellular part of the protein.
Because BS3 reacts only with free amines (the side chains of Lys residues or a free N terminus), we searched for possible sites in TRKA that might have caused the observed cross-linking. According to the crystal structure of the TRKA-NGF complex (15) (Fig. 3A), there are no lysine residues in the TRKA-ECD that are located in a position where cross-linking of the side chains of Lys residues of two monomers could occur. Because BS3 does not cross-the plasma membrane and because we used the full-length TRKA receptor in our assays, we wondered whether the observed BS3 cross-linking was mediated via cross-linking of Lys 410 and Lys 411 in the extracellular juxtamembrane region (eJTM) of TRKA (Fig. 3A) because this region is not observed in the crystal structure (16). To verify this hypothesis, we mutated both Lys 410 and Lys 411 to Arg and repeated the initial experiment using HEK293 cells transfected with this TRKA-KK/RR construct (Fig. 3B). No BS3-induced TRKA cross-linking was observed in the TRKA-KK/RR-transfected cells, suggesting that NGF binding brings this region of the eJTM into close proximity.
We considered that if NGF indeed induces contacts between these eJTM regions, then we should be able to mimic this activity of NGF by forcing the dimerization of eJTMs in the absence of NGF. For this purpose, we individually mutated most of the residues in the eJTM of TRKA to cysteine and subsequently analyzed the dimerization of these transfected single point mutants (Fig. 4A). After transfection of HeLa cells, disulfide dimers were spontaneously formed in all constructs but the amount of dimers differed between the various mutants (Fig.  4B). The amount of dimer is significantly higher in the positions D412C and K411C. As a functional assay, we then transfected these mutants into HeLa cells, which do not express endogenous TRKA and quantified the phosphorylation of the tyrosines from the kinase activation loop (Tyr 674/675 ) in the absence and presence of NGF (Fig. 4C). This analysis showed the presence of active dimers (D406C, K410C, and K411C) that are activated constitutively in the absence of NGF and dimers that are not active in the absence of NGF (V408C, D412C, E413C, and T414C). In the presence of NGF, these mutants showed no further activation by NGF (Fig. 4C), suggesting they are fully active and the dimer interface adopted by the cysteine dimers is similar or identical to the one obtained with NGF binding. As a whole, there is a poor correlation (R 2 ϭ 0.07; Fig. 4D, red line) between the amount of dimer formation and constitutive activation, suggesting that dimerization by itself is not enough for TRKA activation. However. the mutants with higher constitutive activation (D406C, K410C, and K411C) showed a good correlation between dimer formation and activation (R 2 ϭ 0.92; Fig. 4D, green line). We then transfected some of the active mutants in PC12nnr5 cells. In the absence of NGF, the R405C, K410C, and K411C mutants induced the formation of neurites in PC12nnr5 cells (Fig. 4, F and G), supporting the constitutive activation of these mutants and suggested that disulfide bond formation through this interface mimics the binding of NGF. If we assume that the TMD ␣-helix continues in the juxtamembrane region, the residues Asp 406 , Lys 410 , and Lys 411 are in one face of the helix (in green in the Fig. 4H). By contrast the residues whose mutation to cysteine do not activate constitutively the receptor are located in another face (in red in Fig. 4H). All the mutants are correctly expressed at the plasma membrane as found by flow cytometry (Fig. 4E). In summary, our results support the notion of the insufficiency of TRKA dimerization alone for higher receptor activation and support the existence of a preferred active dimer interface that is formed upon ligand binding.

Insertion of leucine residues into the TMD constitutively activates TRKA
Overexpression of TRKA is able to activate the receptor in the absence of ligand, and as we show above, the TMD dimer interface is quite similar, although not identical, to the one

TRKA transmembrane domain dimerization
obtained with NGF stimulation, suggesting that overexpression may induce a basal activation of TRKA. The Western blotting shown in Fig. 5A shows that although the overexpression of TRKA induces a ligand-independent activation, the presence of NGF is required for a higher and complete activation of the receptor. This could be the result of the conformational change induced by NGF binding in the JTM region concomitantly with the rotation of the dimer interface of the TMD dimer. This complete sequence of the events caused by the ligand is of great importance for receptor activation. To test this mechanism, we introduced a different number of leucine residues into the TRKA-TMD and analyzed the resulting TRKA activation in the absence of NGF (Fig. 5B). The insertion of each Leu should rotate the intracellular region around an angle of ϳ100°. Thus, Leu insertion allows evaluation of whether a change in the rotation angle of the intracellular domain plays any role in TRKA activation (Fig. 5C). We transfected the constructs TRKA-ins1L, TRKA-ins2L, TRKA-ins3L, and TRKA-ins4L that included 1-4 inserted leucines, respectively. The insertion of one Leu, TRKA-ins1Leu, significantly increased both the constitutive activation of TRKA in transfected HeLa cells compared with that of transfected TRKA-wt (Fig. 5, D and E) and the differentiation of PC12nnr5 cells compared with wttransfected cells (Fig. 5F) in a ligand-independent manner. Equal levels of all constructs were expressed at the plasma membrane as determined using flow cytometry (Fig. S7). These data indicate that NGF binding is accompanied by a

TRKA transmembrane domain dimerization
conformational rearrangement in the JTM that is transmitted, as a rotation of the TMD, to the intracellular region for TRKA activation.

Discussion
TRKA belongs to a subfamily of RTKs that includes the other family members TrkB and TrkC. These RTKs are essential for the formation of the nervous system and mediate a variety of cellular responses in normal biological processes and in pathological states (37). An understanding of their mechanism of action is necessary to facilitate the design of new pharmacological agents targeted to the processes in which they play a role.
In this regard, in the present work we posed two major questions: 1) what is the dimer interface of TRKA TMD in the active/inactive states? and 2) how is the coupling between the ligand binding and receptor activation? To answer the first question, we employed structural characterization using NMR spectroscopy together with mutagenesis studies, disulfide cross-linking, and functional assays. We described the highresolution NMR structure of the TRKA transmembrane domain dimer, which is the first such description of a neurotrophin receptor of the Trk family. The obtained spatial structure was verified using functional assays and cross-linking, which confirmed both the relevance of this structure for TRKA activity and assignation of the found dimer conformation to the receptor active state. This result combined with the crystal structure of the extracellular domain of the TRKA complex with NGF (16, 38), the crystal structure of the TRKA inactive kinase domain (21,22), and the recently reported structures of the entire TRKA intracellular region (39,40) provides an almost complete picture of the full-length Trk receptor family, lacking only the structure of the small JTM regions. Our results highlight the key role of Ala 428 in the activation of TRKA by NGF. The importance of this residue may reside in its pivotal position modulating the transition between the inactive to the active dimer interface. This transition could be hampered by the introduction of a bulky residue like Ile. This is also reflected by its high conservation in the TRKA and TrkC protein sequences from several species. The protein sequence divergence of TrkB-TMD is notable (Fig. 1) and may reflect a totally different mechanism of activation as has been recently proposed (41). The finding that the mutation of G423I reduces significantly the differentiation of PC12 cells with NGF suggests that small rearrangements in this TMD region could play an important role in the activation of the downstream signaling, leading to cell differentiation.
To answer the second question posed above, we investigated the role of the eJTM regions of the TRKA receptor upon NGF binding in receptor activation and dimerization. We showed that full-length TRKA receptors could be activated by specific single-point cysteine mutations in the eJTM and in the TMD, in which the position of the mutation relative to the TMD was more important for activation than the dimerization propensity of the mutant. In addition the insertion of Leu residues upstream of the TMD dimerization motif activated TRKA in

TRKA transmembrane domain dimerization
the absence of ligand. This suggests that rotation of the downstream domain may be behind the activation of the kinase domain. Other authors have suggested the rotational mechanism of RTK activation (42,43). In this model the ligand would induce a rotation of the TMD dimer interface that will reorient the kinase domains to facilitate the transphosphorylation. Although our results support this model of activation, other alternative possibilities may exist. For instance the insertion of extra residues increases the TMD length and could induce a piston-like mechanism of activation. However, an increase in the length by two, three, or four residues should also activate the kinase, and this was not the case, because only the Ins1L mutant showed activation. Also, because the insertion of the residues are located into the TMD, it may alter the dimer interface of the TMD dimer, leading to the formation of another dimer interface compatible with a higher activation of the kinase domains. Although we cannot discard this possibility, in the constructs we made, the Leu residues were inserted upstream of the active dimer interface (Fig. 5B) to not alter the dimer interface found by NMR studies.
Bearing all of these findings in mind, we propose a mechanism of receptor activation that suggested a ligand-induced dimerization (or stabilization of preformed dimers) accompanied by a conformational change in the JTM that is transmitted to the intracellular regions by a rotation of the TMD (Fig. 6). This model is supported by our data showing that cysteine mutants in the eJTM and in the TMD in some specific positions can activate TRKA without ligand and that NGF binding induce the formation of new cysteine dimers in the TMD and the ligand-independent activation of TRKA by an induced rotation of its TMD. In addition our model may allow the existence of preformed inactive dimers of TRKA and a conformational activation by NGF as others have suggested (18,44). Future structural and functional studies of the TRKA should address how the kinase domain is activated if they are connected to the TMD by a flexible intracellular juxtamembrane region. In summary, we have provided functional and structural evidence of the roles played by the JTM and the TMD in TRKA dimerization and activation by NGF.

DNA constructs
A plasmid encoding rat TRKA with an N-terminal hemagglutinin tag was kindly provided by Dr. Y. Barde. All TRKA mutants and constructs were derived from this plasmid. Mutagenesis was done using the site-directed mutagenesis kit (Agilent) according to the manufacturer's protocol. The oligonucleotide sequences of all of the constructs are available upon request. All DNA constructs were sequenced using local facilities.

Cell culture and transfection
HeLa cells, which do not express endogenous TRKA, were cultured in Dulbecco's modified Eagle's medium (Fisher) supplemented with 10% fetal bovine serum (Fisher) at 37°C in a humidified atmosphere with 5% CO 2 . PC12 and PC12nnr5 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 5% horse serum (Fisher). Transfection of HeLa cells was performed using polyethylenimine (PEI; Sigma) at a concentration of 1-2 g/l. The use of PEI as the transfection reagent for HeLa cells resulted in suboptimal transfection (10 -15% of cells transfected) and in the expression of only a small amount of TRKA in the cells. In contrast, when the same PEI/DNA ratio was used for transfection of HEK293 cells, TRKA was expressed in higher amounts, and ligand-independent activation of TRKA was seen. A concentration of 500 -1000 ng of DNA per 10-cm cell plate was used for the TRKA activation experiments. Twenty-four hours after transfection, the cells were lifted and replated into 12-well plates at a density of 100,000 cells/well. By using this procedure, the percentage of cells transfected was identical in all the wells. Forty-eight hours after transfection, the cells were starved in serum-free medium for 2 h and were then stimulated with NGF (Alomone) at the indicated concentrations and time intervals. The cells were lysed with TNE buffer (Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA) supplemented with 1% Triton X-100 (Sigma), protease inhibitors (Roche), 1 mM phenylmethylsulfonyl fluoride (Sigma), 1 mM sodium orthovanadate (Sigma), and 1 mM sodium fluoride (Sigma). In the experiments involving the TRKA cysteine mutants, 10 mM iodoacetamide (Sigma) was added to the lysis buffer. Lysates were kept on ice for 10 min and centrifuged at 12,000 ϫ g for 15 min in a tabletop centrifuge. The protein level of the lysates was quantified using a Bradford kit (Pierce), and lysates were analyzed by SDS-PAGE

Western blotting analysis
Cellular debris was removed by centrifugation at 12,000 ϫ g for 15 min, and the protein level of cell lysates was quantified using the Bradford assay (Pierce). The proteins were resolved in SDS-PAGE gels and transferred to nitrocellulose membranes that were incubated overnight at 4°C with one of the following antibodies: mouse monoclonal anti-hemagglutinin (1:2000, Sigma); rabbit polyclonal MBP-probe (1:1000, Santa Cruz); rabbit anti-P-Tyr 674/675 (1:1000, Cell Signaling); and rabbit anti-TRKA (1:1000, Millipore). Following incubation with the appropriate secondary antibody, the membranes were imaged, and the bands were quantified using enhanced chemiluminescence and autoradiography.

TRKA-TMD constructs for cell-free expression
The gene encoding the transmembrane domain of the human TRKA-TM ( 409 MKKDETPFGVSVAVGLAVFACLFL-

TRKA transmembrane domain dimerization
STLLLVLNKAGRRNK 447 ) was amplified by PCR from six chemically synthesized oligonucleotide templates (Evrogen) whose sequences partially overlapped along its sequence. The PCR products were cloned into a pGEMEX-1 vector by threecomponent ligation using the NdeI, AatII, and BamHI restriction sites.

Cell-free gene expression
A bacterial S30 cell-free extract was prepared from a 10-liter culture of the Escherichia coli Rosetta(DE3)pLysS strain according to a previously described protocol. The S30 cell extract was stored in 500-l aliquots at Ϫ80°C. The continuous exchange mode of preparation using a 12.5-kDa membrane was used in this study. Preparative-scale reactions (2-3 ml of reaction mixture) were carried out in 50-ml tubes. Optimal reaction conditions such as Mg 2ϩ and K ϩ concentrations, the ratio of the reaction mixture to the feeding mixture or the DNA concentration were established using homemade reactors based on the Mini-CECF reactor previously described (45,46). The final reaction mixture was a standard feeding mixture:reaction mixture ratio of 8:1 and a cell-free reaction mixture containing 100 mM HEPES, 0.83 mM EDTA with KOH added to achieve a pH of 8.0, 0.1 mg/ml folinic acid, 20 mM acetyl phosphate, 1.2 mM ATP, 0.8 mM each of G/C/UTP, 2 mM 1,4-DTT, 0.05% sodium azide, 2% PEG 8000, 20 mM magnesium acetate, 270 mM potassium acetate, 60 mM creatine phosphate, 1 mM each of 20 amino acids or 0.25% of a 20-amino acid mix (Cambridge Isotope Laboratories, USA), 1 tablet/50 ml of complete protease inhibitor (Roche, Switzerland), 0.5 mg/ml E. coli tRNA (Roche, Switzerland), 0.25 mg/ml creatine kinase from rabbit muscle (Roche, Switzerland), 0.05 mg/ml T7 RNA polymerase prepared using a previously described protocol (47), 0.1 unit/l Ribolock (Fermentas), 0.02 g/l plasmid DNA, and 30% S30 cell-free extract. All reagents were provided by Sigma unless otherwise specified. Plasmid DNA was purified using a Promega MaxiPrep kit. The reactions were conducted overnight at 34°C and in an Innova 44R shaker (New Brunswick) set at 150 rpm.

Protein purification
The cell-free reaction mixture was diluted three-times with buffer A (50 mM Tris, pH 8.0, and 200 mM NaCl). After 10 min of incubation, the mixture was centrifuged for 10 min at 18,000 ϫ g at room temperature. The precipitate was washed consecutively with buffer A containing 30 g/ml RNase A (Fermentas) and buffer B (50 mM Tris, pH 8.0, and 100 mM NaCl). The target protein was solubilized with 200 l of buffer B containing 1% lauryl sarcosine. After each step, the protein was centrifuged for 10 min at 18,000 ϫ g at room temperature, and aliquots of the supernatant were analyzed using 12.5% Tricine SDS-PAGE (53). The clarified protein solution was applied onto a 10/300 Tricorn column prepacked with Superdex 200 (GE Healthcare) and pre-equilibrated with buffer B containing 0.2% lauryl sarcosine. Protein-containing fractions were combined and precipitated using the TCA/acetone procedure (48).

Preparation of NMR samples in a membrane mimetic medium
The so-called "isotopic heterodimer" (1:1 mixture of unlabeled and 15 N/ 13 C-labeled peptides) samples were prepared corresponding to the TRKA-TMD construct to solve its structure. The powder containing the peptides of both samples was first dissolved in a 1:1 trifluoroethanol-H 2 O mixture with the addition of deuterated DPC (d 38 , 98%, Cambridge Isotope Laboratories, USA) and phosphate buffer and was then kept for several minutes in an ultrasound bath and lyophilized. Subsequently, the dried samples were dissolved in 350 l of a 9:1 H 2 O:D 2 O mixture. To attain a uniform micelle size and uniform distribution of the peptide throughout the micelles, the samples were sonicated in an ultrasound bath for several minutes until the solution was completely transparent. The TRKA-TMD concentration in the isotopic-heterodimer sample was 1.9 mM, and other conditions were as follows: DPR 50:1, pH 5.9, and 20 mM phosphate buffer. The samples were placed in Shigemi NMR tubes with a glass plunger. Selective-residue labeling was implemented to avoid peaks overlapping while processing the NMR spectra.

NMR spectroscopy and spatial structure calculation
NMR spectra were acquired at 45°C using 600 and 800 MHz AVANCE III spectrometers (Bruker BioSpin, Germany) equipped with pulsed-field gradient triple-resonance cryoprobes. 1 H, 13 1 H/ 13 C/ 15 N HNCO, 3D-HccH total correlation spectroscopy, and 13 C-and 15 N-edited NOESY HSQC (recorded on 600 and 800 MHz spectrometers, respectively). Dimeric spatial structures were calculated with the CYANA program (51) based on torsion angle restraints estimated from the chemical shift values obtained with the standard protocol of the TALOS-N program (52) and with intra-and intermonomeric NOE distance restraints derived through analysis of the three-dimensional 15 N-and 13 C-edited NOESY and 15 N, 13 C F1-filtered/F3-edit-edNOESY spectra (50) acquired for isotopic heterodimer samples. MOLMOL software was used to calculate the contact areas between the dimer subunits and to visualize the structures (53). Hydrophobic properties of the ␣-helices in the TRKA-TMD dimers were calculated using the molecular hydrophobicity potential approach implemented in the PRED-DIMER program (54).

Free energy measurements
To measure the free energy we used an NMR-based approach (31). 1 mM 15 N-labeled TRKA-TMD sample was prepared in DPC micelles at DPR 50 in NMR buffer and then bxgradually diluted by 10% DPC solution. At each point, the band-selective excitation short transient-transverse relaxation optimised spectroscopy spectrum with a 0.8-s relaxation delay was acquired to ensure the equilibrium longitudinal magnetization of amide protons (55). The populations of monomeric, dimeric, and oligomeric states were measured from the integrals of separate cross-peaks in these spectra, applying the correction to take into account the coherence losses caused by the transverse relaxation (31). Obtained populations were then converted to monomer and dimer concentrations and approx-

TRKA transmembrane domain dimerization
imated by the micelle-based model of TMD dimerization (31): , where M and D stand for the concentrations of monomer and dimer, whereas [Emic] is the concentration of "empty" micelles. An additional constraint equation was applied: N e [Emic] ϩ N m M ϩ N d D ϭ [Dpc], where N e , N m , and N d stand for the number in detergent molecules in empty, monomer-bearing, and dimer-bearing micelles, respectively. N e was fixed to 55 (56), whereas K D , N m , and N d were the parameters of approximation. The resulting free energy was additionally corrected by RTln(N e ), to change to the generally accepted standard conditions of 1 M detergent (the original value corresponded to the standard conditions of 1 M of empty micelles) (57).

Isolation of membrane fractions
TRKA mutants were overexpressed in HeLa cells by transfection using PEI (1 mg/ml; Sigma). 48 h after transfection, the cells were resuspended in 1 ml of ice-cold homogenization buffer (250 mM sucrose, 1 mM EDTA, 10 mM Tris buffer, pH 7.1, plus protease inhibitors) and broken by sonication in two time intervals (30 and 30 s) with 50 W and frequency at 30 MHz on ice. The broken cell homogenate was centrifuged for 10 min at 500 ϫ g at 4°C to remove whole cells and nuclei. To collect the membrane fraction the cleared supernatant was centrifuged at 100,000 ϫ g at 4°C for 1 h in a Beckman Optima MAX ultracentrifuge with a TLA110 rotor using polycarbonate thick-wall centrifuge tubes (Beckman Coulter). The supernatant that contains soluble proteins was removed, and the pellet-containing membranes were resuspended in 1 ml of ice-cold homogenization buffer by sonication and recentrifuged at 100,000 ϫ g. The final pellet contains the membrane fraction used to the iodine oxidation protocol.

Iodine oxidation for cross-linking of TRKA mutants
Because cysteine residues are buried deep in the phospholipid bilayer, the membranes were isolated, and I 2 was used as the oxidation agent as described in Ref. 58. The membrane fractions were diluted in the homogenization buffer (250 mM sucrose, 1 mM EDTA, 10 mM Tris buffer, pH 7.1 plus protease inhibitors). Protein content was quantified using a Bradford kit (Pierce), and equal amounts of isolated membranes fractions were incubated for 10 min with or without NGF (10 ng/ml). A solution of 2.5 mM I 2 in absolute ethanol was freshly prepared immediately before the cross-linking of cysteine residues and was added to the incubated membrane fractions with NGF (250 M final iodine concentration) for 30 s at room temperature. The reaction was stop adding 1/10 volume of a freshly made solution of sodium thiosulfate (60 M final concentration). Nonreducing SDS-PAGE sample buffer was added, and the samples were boiled for 5 min before analyzed by nonreducing SDS-PAGE.

Differentiation of PC12nnr5 cells
Transfection in PC12nnr5 cells was performed using Lipofectamine 2000 as per the manufacturer's instructions. The mutant or the WT TRKA-transfected cells and mock-transfected cells, as well as nontransfected cells, were treated under the same conditions in a 6-well tissue culture plate. The cells were washed three times with serum-free medium and incubated for 48 h in a medium containing 1% fetal bovine serum and 50 ng/ml of NGF (Alomone). At time 0, 24 h, and 48 h, the cells were washed with cold PBS and fixed with 4% paraformaldehyde for 15 min at room temperature. The cells were imaged using a Leica SP8 spectral confocal microscope. The percentage of cells with a neurite twice as long as the cell body was counted as differentiated.

Accession codes
The atomic coordinates and experimental restraints were deposited in the Protein Data Bank under accession code 2n90 for TRKA-TM-wt.