Comparison of tyrosine kinase domain properties for the neurotrophin receptors TrkA and TrkB.

The tropomyosin-related kinase (Trk) family consists of three receptor tyrosine kinases (RTKs) called TrkA, TrkB, and TrkC.  These RTKs are regulated by the neurotrophins, a class of secreted growth factors responsible for the development and function of neurons.  The Trks share a high degree of homology and utilize overlapping signaling pathways, yet their signaling is associated with starkly different outcomes in certain cancers.  For example, in neuroblastoma, TrkA expression and signaling correlates with a favorable prognosis, whereas TrkB is associated with poor prognoses.  To begin to understand how activation of the different Trks can lead to such distinct cellular outcomes, we investigated differences in kinase activity and duration of autophosphorylation for the TrkA and TrkB tyrosine kinase domains (TKDs).  We find that the TrkA TKD has a catalytic efficiency that is ~2-fold higher than that of TrkB, and becomes autophosphorylated in vitro more rapidly than the TrkB TKD.  Studies with mutated TKD variants suggest that a crystallographic dimer seen in many TrkA (but not TrkB) TKD crystal structures, which involves the kinase-insert domain, may contribute to this enhanced TrkA autophosphorylation.  Consistent with previous studies showing that cellular context determines whether TrkB signaling is sustained (promoting differentiation) or transient (promoting proliferation), we also find that TrkB signaling can be made more transient in PC12 cells by suppressing levels of p75NTR.  Our findings shed new light on potential differences between TrkA and TrkB signaling, and suggest that subtle differences in signaling dynamics can lead to substantial shifts in cellular outcome.


INTRODUCTION
The tropomyosin receptor kinase (Trk) family [1] comprises three receptor tyrosine kinases (RTKs) called TrkA, TrkB, and TrkC that are activated by neurotrophins. TrkA functions as the primary receptor for nerve growth factor (NGF), TrkB is the primary receptor for brainderived neurotrophic factor (BDNF) and neurotrophin-4 (NT-4), and TrkC is the primary receptor for neurotrophin-3 (NT-3). The Trks primarily regulate neuronal survival and differentiation through a highly complex set of tightly controlled cell-specific and receptorspecific interactions. In addition, they are associated with certain cancers, notably neuroblastoma, where the different receptors play starkly different roles [2,3]. Activating TrkA and TrkB in the same cell type can yield very different cellular outcomes. For example, activating TrkA signaling in neuroblastoma cells inhibits cell growth, whereas activation of TrkB does not [4]. Indeed, whereas TrkA expression and activation promotes cell differentiation and a favorable prognosis in neuroblastoma, TrkB expression and activation conversely causes cell proliferation and is associated with a poor prognosis [2,3,5,6]. TrkA and TrkB are 49% identical overall, with >76% sequence identity in the intracellular region. How can two homologous RTKs, both thought to engage similar downstream signaling pathways, produce opposite cellular responses?
Despite their association with favorable and unfavorable outcome respectively in neuroblastoma, studies of global gene regulation following ligand stimulation in engineered neuroblastoma cells have suggested a surprising degree of similarity, apart from differential expression of a few target genes [7]. This raises the possibility that differences in kinetics or time course of signaling -rather than pathways -may be the primary determinant of cellular outcome. Now-classic studies in rat pheochromocytoma (PC12) cells [8,9] showed that duration of extracellular regulated kinase (ERK) signaling downstream of RTKs defines the nature of the response. NGF activates TrkA to promote sustained signaling and differentiation, whereas activation of EGFR with EGF leads to transient signaling and proliferation. These different cell fates arise in part from phosphorylation-dependent regulation of immediate-early gene product stability [10]. Recent studies argue that the dynamics of ERK signaling downstream of an RTK are substantially defined at the level of receptor activation kinetics [11][12][13][14]. Moreover, we and others have shown that the same intracellular tyrosine kinase domain (from EGFR) can signal with different kinetics and cellular outcome depending on how it is activated [14][15][16][17][18].
The different prognoses associated with TrkA and TrkB signaling in neuroblastoma [2,3] and certain other cancers [19,20] suggest the possibility of intrinsic signaling differences between the two receptors. So do differences in response to optogenetically activated TrkA and TrkB kinase domains that have been reported in Xenopus embryos [21]. Motivated by these observations, we initiated an investigation of possible differences in kinase activity between the TrkA and TrkB tyrosine kinase domains (TKDs). Through in vitro kinase and autophosphorylation assays, we characterize the effects of autophosphorylation on TKD activation for TrkA and TrkB. We find that they are quite similar kinetically, but TrkA does have a slightly -but significantly -higher catalytic efficiency than TrkB when fully active. Moreover. TrkA undergoes more efficient autophosphorylation than TrkB in vitro, which may be enhanced by crystallographically observed TrkA-specific interactions involving its unique kinase insert domain (KID). Finally, motivated by studies of the effects of the low affinity neurotrophin receptor p75 NTR on neuroblastoma [22,23], we found that siRNA knockdown of this receptor makes TrkB signaling more transient when exogenously expressed in PC12 cells. These data, and others [24] suggest that the kinetic characteristics of TrkB signaling, and its association with proliferation or differentiation depend on the expression of co-receptors and other molecules that influence its signaling networks. Our results have implications for Trk receptor targeting in neuroblastoma and other settings, focusing attention on the kinetics of their signaling.

Plasmid Construction
Sequence and structure alignments facilitated the design of analogous TrkA TKD and TrkB TKD constructs used for protein expression. DNA encoding intracellular domain residues 498-796 of human TrkA (NCBI reference sequence NM_002529.3) and residues 526-822 of human TrkB (NCBI reference sequence NM_001018064.3) were amplified by PCR, also adding a sequence encoding an N-terminal hexahistidine tag and unique EcoRI and XhoI restriction sites for subcloning. The TrkA B-KID TKD variant, with its kinase insert domain (KID) engineered to resemble that of TrkB was generated by introducing the following mutations into TrkA TKD: Glu 615 Asn, Asp 616 _Val 617 del, Ala 618 Pro. EcoRI/XhoI digested PCR fragments were subcloned into pFastBac1 (Invitrogen) cut with the same restriction enzymes. The Bac-to-Bac expression system (Invitrogen) was then used for generation of recombinant baculoviruses and for protein expression in Spodoptera frugiperda Sf9 cells.

Production and Purification of Proteins
TrkA, TrkB, or mutated TrkA B-KID TKDs were produced as previously described for the TrkA TKD [25]. Sf9 cell lysate was mixed with Ni-NTA beads (Qiagen) for 1 h at 4°C. Beads were washed in 50 column-volumes of lysis buffer: 50 mM KH 2 PO 4 /Na 2 HPO 4 , pH 8.0, containing 300 mM NaCl, 5% w/v glycerol, 10 mM imidazole, 10 mM 2mercaptoethanol, 0.5 mM PMSF, and Roche protease inhibitor cocktail. The bound TKD was eluted with increasing concentrations of imidazole in the same buffer, and was further purified using a Fractogel SO 3 − cation exchange column (EMD) in 25 mM MES, pH 6, containing 5% glycerol, 2 mM DTT (eluting with a gradient from 10 mM to 1 M NaCl). The TKD proteins were concentrated to 5-6 ml by ultrafiltration, and were then treated with 2 μM YopH phosphatase for 1 h at 30°C in the presence of 1mM EDTA. YopH was purified away by repeating the cation exchange step. Trk TKDs were then concentrated again to 5-6 ml and treated with 2 μM λ phosphatase in the presence of 5 mM MnCl 2 for 1 h at 30°C, with subsequent removal of λ phosphatase using a cation exchange step. The TKDs were next applied to a HiTrap Butyl Sepharose HP column (GE Healthcare) in 25 mM MES, pH 6.0, containing 150 mM NaCl and 2 mM DTT, and were eluted with a gradient from 0. λ phosphatase was expressed in E. coli and purified essentially as described [26]. Cell lysate was loaded onto a phenylsepharose column and peak fractions were pooled, concentrated and loaded onto a 10/300 Superdex 200 size exclusion column (GE Healthcare) equilibrated in 25 mM Tris, pH 7.5, containing 150 mM NaCl. Purified λ phosphatase was stored at −80°C at 20 mg/ml in 25 mM Tris, pH 7.5, 150 mM NaCl, plus 10% w/v glycerol.

Autophosphorylation Assays
All autophosphorylation reactions were conducted in the presence of 1 mM ATP (Sigma-Aldrich) and 10 mM MgCl 2 in a buffer containing 100 mM HEPES pH 7.4, 150 mM NaCl, 2 mM DTT and 1x Halt phosphatase inhibitor cocktail (Thermo Scientific #1862495), which contains sodium fluoride, sodium orthovanadate, sodium pyrophosphate, and βglycerophosphate. Figure legends indicate the concentrations of TrkA and TrkB TKD proteins used for autophosphorylation reactions, ranging from 1 to 10 μM, and also indicate the temperature at which reactions were incubated (ranging from 15°C-37°C). Autophosphorylation reactions were analyzed by different gel-based methods including SDS-PAGE autoradiography, SDS-PAGE Western blotting, and Phos-Tag SDS-PAGE. For SDS-PAGE autoradiography, gels were dried and imaged using a PhoshorImager, and normalized using the near infrared fluorescence signal from protein-bound Coomassie blue stain from the same gel using ImageStudio software (LI-COR Biosciences).

Generation and purification of phospho-Trk TKD
Autophosphorylation reactions were conducted to generate phospho-Trk (pTrk) TKDs for kinetic studies. Western blotting of Phos-Tag gels with phosphospecific antibodies was used to monitor the reactions and ensure phosphorylation was as close to stoichiometric as possible, with examples shown in Fig. S3. TrkA TKD at 10 μM was incubated at room temperature with 1 mM ATP, 10  Reaction progress was monitored essentially as described [27]. Samples were taken at each time point from the reactions listed above, spotted onto squares of P81 phosphocellulose paper (Millipore # 20-134), and immediately quenched with 0.5% phosphoric acid. The phosphocellulose squares were then washed extensively with 0.5% phosphoric acid and dried with acetone. Incorporated radioactivity was then measured for each phosphocellulose square by scintillation counting using a Tri-Carb 2900TR liquid scintillation analyzer (PerkinElmer). Initial rates were determined under conditions of <10% product formationensuring linearity. These rates were converted to specific activity, normalized for enzyme concentration, and data were fit to the Michaelis-Menten equation using GraphPad Prism version 8.4.2. Unpaired two-tailed t-tests were used to determine statistical significance of differences in enzymatic parameters between TrkA TKD and TrkB TKD in Table 1.

Radioactive autophosphorylation assays
Parallel radioactive autophosphorylation assays were performed at RT with 4 μM of TrkA TKD, TrkA B-KID , or TrkB TKD protein in the presence of 2 mM ATP, 10 mM MgCl 2 100 mM HEPES pH 7.4, 150 mM NaCl, 2 mM DTT, 1x Halt phosphatase inhibitor cocktail and trace amounts of γ− 32 P ATP (40 μCi), but no peptide substrate. Samples taken at each time point were processed as above for the quantitative kinase assay samples. Counts from the scintillation analyzer were converted to moles of phosphate incorporated using measured γ − 32 P-ATP specific activity and then normalized to enzyme concentration. Each reaction data set was converted to percent maximum moles of phosphate incorporated normalized to enzyme concentration.

Western Blotting
For SDS PAGE Western blotting, samples were separated on a 12.5% Laemmli gel, transferred to nitrocellulose and then blocked with blocking buffer (LI-COR Biosciences). Membranes incubated with primary antibody diluted in blocking buffer overnight at 4°C, washed with PBS containing 1% Tween-20, and then incubated with secondary antibody diluted in blocking buffer. Membranes were washed with PBS containing 1% Tween-20 and then subjected to a final wash with just PBS. Membranes were imaged using the Odyssey® FC (LI-COR Biosciences) imaging station.
Phos-Tag gels were made using the Phos-Tag acrylamide reagent (Wako Pure Chemical Industries, Ltd., catalog # AAL-107) along with ZnCl 2 as previously described [28]. The composition of reagents for optimal separation of phosphorylated Trk TKD species was 25 μM Phos-Tag acrylamide reagent and a 10% acrylamide gel using acrylamide:bis solution (29:1). Gels were run at a constant 70V for 3 hours. Phos-Tag gels were stained at room temperature for 15-20 min with Coomassie brilliant blue R-250. For Western blotting, prior to transferring to nitrocellulose, Phos-Tag gels were soaked in transfer buffer containing 1 mM EDTA for 15 min and then washed for 15 min in transfer buffer without EDTA. The rest of the Western blot protocol was performed as described above.
For the SDS PAGE Western blots, each phosphospecific signal was normalized for total Trk. Normalized signal from a phosphospecific antibody was then expressed as a fraction of the maximum signal seen with that antibody for TrkA or TrkB TKD autophosphorylation respectively. Using these data, the normalized relative signals for each phosphospecific antibody could be compared directly between sites and TKDs.

Structure visualization and alignment
Structure figures and alignments were generated using PyMOL (The PyMOL Molecular Graphics System, Version 2.0.7, Schrödinger, LLC). All structures were aligned to our previously determined inactive TrkA structure (PDB code 4GT5).

Studies of Trk signaling in PC12 cells
PC12 cells and/or PC12 cells stably expressing TrkB (TrkB-PC12) were cultured at 37°C, humidified 5% CO 2 in complete RPMI-1640 medium containing L-glutamine and HEPES (ThermoFisher Sci #22400089), supplemented with 1 mM sodium pyruvate (ThermoFisher Sci #11360070), 5% fetal bovine serum (FBS) (Atlanta Biologicals #S11550), 10% heat inactivated horse serum (ThermoFisher Sci #26050088) as well as 100 U/ml penicillin and 100 μg/ml streptomycin (ThermoFisher Sci #15140122). To subculture, harvested cells were centrifuged at 200 × g for 10 min, supernatant removed and the cell pellet was resuspended in fresh medium. Cell culture was aspirated several times with a syringe outfitted with a 22gauge needle to break up cell clusters. Aliquots of cell suspension then were added to new T-75 flasks containing 10-15 ml fresh medium and cells were cultured up to cell density 2-4 × 10 6 cells/ml.

Generation of TrkB-PC12 stable cell line:
A cDNA construct directing expression of full length human TrkB was transfected into PC12 cells using electroporation with Nucleofector 2b device as described below for siRNA procedure. Transfected cells were selected for 2 weeks in medium containing 0.5 mg/ml G418. Stable clones were maintained in the presence of 250 mg/ml G418. TrkB expression levels were confirmed by Western blotting using anti-TrkB antibody (CST #4603).
Transient cell transfection with siRNA: TrkB-PC12 or parental PC12 cells were harvested and aspirated several times with a syringe outfitted with a 22-gauge needle to break up cell clusters. 2 × 10 6 cells per sample were aliquoted into Eppendorf tubes and centrifuged at 100 × g for 10 min. The cell pellet was resuspended in 100 μl of Ingenio transfection solution (Mirus #MIR50111) containing siRNA. 100 nM of rat specific SMARTpool siGENOME p75 NTR siRNA (Dharmacon # M-080041-01-0005) was used to compare cell responses to those transfected with 100 nM AllStars Negative Control siRNA (Qiagen #1027280). Cell suspensions containing siRNA were electroporated using the U-029 program on Nucleofector 2b device (Lonza, Basel, Switzerland). The pre-equilibrated antibiotic-free complete medium was added to the cuvette immediately after electroporation and transferred into 60 × 15 mm Corning BioCoat collagen I-coated dishes (Fisher Sci #356401) to a final volume of 2 ml. Penicillin-streptomycin solution was added to the cells after 6 h. On the next day, cell culture medium was replaced with 5 ml of fresh complete medium. Cell signaling experiments were performed 72 h post-transfection.
Cell stimulation: parental PC12 cells and/or TrkB-PC12 were harvested from T-75 flasks, centrifuged and subjected to the needle aspiration procedure described above. Cells were seeded on 100 × 20 mm Corning BioCoat collagen I-coated dishes (Fisher Sci #356450) in complete growth medium at a density of 1 × 10 6 cells/ml (10 ml per dish) and grown for one day at 37°C in 5% CO 2  Haven, CT). Signal intensities for a given protein were normalized by signal intensities of the loading control (Grb2 protein from the same gel) at each time point as described [14,15]. Time courses of protein phosphorylation from different experiments were also compared by normalizing peak values at 5 min stimulation and were expressed in percent maximum.

Quantitation of TrkA and TrkB kinase activity and effects of autophosphorylation
To explore possible differences between TrkA and TrkB, we first compared the activities of their intracellular tyrosine kinase domains (TKDs) in both the dephosphorylated (unactivated) and fully phosphorylated states. TrkA and TrkB contain five primary autophosphorylation sites [31][32][33]. Three are in the activation loop as part of the YxxxYY motif. Another lies in the juxtamembrane (JM) region (Tyr 496 in TrkA, Tyr 516 in TrkB) and was removed in our constructs to optimize protein production levels. The fifth lies in the Cterminal tail (Tyr 791 in TrkA, Tyr 817 in TrkB), which is intact in the proteins used here. We fully dephosphorylated the purified TKDs by serial treatment with the YopH and λ phosphatases as described in Materials and Methods. To generate maximally phosphorylated proteins, autophosphorylation was allowed to proceed as described below (see also Materials and Methods), and the reaction was quenched at a time determined to correspond to its peak. We then monitored activity of the dephosphorylated and phosphorylated TKDs by following incorporation of 32 P from 32 P-γ-ATP into peptides that mimic the activation loop of either TrkA or TrkB (see Materials and Methods). Fig. 1 shows representative data used to obtain the kinetic parameters, and Table 1 lists kinetic parameters measured (and errors) for TrkA and TrkB TKDs in these assays.
The kinase activities of the TrkA and TrkB TKDs were very similar. Values for k cat were slightly (10-30%) higher for TrkA than for TrkB in experiments with both phosphorylated and dephosphorylated protein, but these differences were not statistically significant across multiple repeats when subjected to Mann-Whitney-Wilcoxon tests. Similarly, K m, ATP values were slightly lower for TrkA than TrkB in both sets of experiments (Table 1), but this comparison alone did not reach statistical significance. The results did suggest a small but statistically significant difference in catalytic efficiency (k cat /K m,ATP ) for the phosphorylated Trk TKDs, however, with the value for pTrkA being approximately 1.7-fold greater than that for pTrkB (P = 0.036 from a Mann-Whitney-Wilcoxon test, or 0.0005 from a Student's ttest).
As expected, autophosphorylation led to a dramatic increase in kinase activity for both TrkA TKD and TrkB TKD, with 70-fold and 85-fold increases in k cat respectively. Autophosphorylation also reduced K m values for ATP by ~10-fold, and for peptide by 3-5 fold Table 1). All phosphorylation-dependent changes in Table 1 were statistically significant. These changes upon phosphorylation are typical for tyrosine kinases in the insulin receptor family. Indeed, TKDs from the insulin receptor (IRK), IGF1 receptor, and MuSK have been reported to show phosphorylation-dependent increases in k cat of between 20 and 230-fold, with associated 6-10-fold reductions in K m for ATP and peptide [34][35][36][37].
The IRK, IGF1-R and MuSK TKDs are the closest homologues of the Trk TKDs. All share >40% sequence identity and adopt very closely related conformations in their inactive states [25]. Like TrkA and TrkB, the IGF1-R and MuSK TKDs share a characteristic mode of peptide binding site occlusion by the second tyrosine of the YxxxYY motif first seen in IRK [38], and of ATP-binding site occlusion by the remainder of the activation loop. Interestingly, TKDs that contain an activation loop YxxxYY motif but do not adopt the IRKlike autoinhibitory conformation, namely c-Met and ALK [25], show smaller changes in K m upon autophosphorylation. In the case of c-Met, K m for peptide is reduced by only ~1.5-fold by autophosphorylation, and K m,ATP by ~5-fold [39]. More notably, autophosphorylation of ALK changes K m for peptide by less than 2-fold, and actually increases (rather than decreases) K m,ATP [27]. These differences in the effects of activation loop autophosphorylation provide evidence for the altered modes of (or lack of) of ATP-binding site occlusion in the unique autoinhibitory conformations seen for c-Met and ALK (and Ron) in this tyrosine kinase family.

Autophosphorylation and dephosphorylation of TrkA and TrkB TKDs
In the course of optimizing autophosphorylation of the TrkA and TrkB TKDs for these experiments, we found by monitoring 32 P incorporation that TrkA-TKD becomes autophosphorylated more rapidly than TrkB-TKD when analyzed at the same concentration (10 μM) under the same conditions (Fig. S4). This is a protein concentration that mimics a receptor density of ~10 5 receptors per cell [40,41], so is likely to be relevant for events at the cell surface. TrkA-TKD autophosphorylation peaked at approximately 1 min at 37°C (Fig. S4A,B), whereas TrkB-TKD autophosphorylation peaked at ~2.5 min, effectively displaying a lag phase at the beginning of the reaction -as reported in earlier studies of the TrkB intracellular region [42]. This difference was maintained, but spread out in time, in studies at room temperature (Fig. S4C,D), with TrkA autophosphorylation peaking at around 5 min and TrkB at 10 min, with a longer lag phase than seen at 37°C. Interestingly, once autophosphorylation had reached its peak, the 32 P signal was progressively lost over time, indicating dephosphorylation of the purified TKDs. This is particularly noticeable in the autoradiograms shown for TrkA and TrkB TKDs in Fig. S4A. Similar dephosphorylation has been reported for IRK [43][44][45][46], where it was shown to be ADP dependent, to coincide with transfer of 32 P-phosphate from IRK into ATP [45], and to be inhibited by EDTA or staurosporine [43,46]. Similar phenomena have also been observed for the IGF1-R TKD [37] and the fibroblast growth factor (FGF) receptor-1 [47]. Moreover, the vascular endothelial growth factor (VEGF) receptor-2 was shown to be capable of catalyzing transfer of phosphate from a peptide substrate to Mg 2+ -ADP [48], albeit with a k cat of just 0.1 s −1 .
We tried to identify contaminating protein tyrosine phosphatases in our Trk TKD preparations using mass spectrometry -but found none. As in the earlier IRK studies, we also could not arrest Trk TKD dephosphorylation with phosphatase inhibitors (data not shown). Thus, the Trk TKDs appear to share this feature with other IRK-family kinases. Interestingly, the auto-dephosphorylation seen here was not observed in our extensive studies of the ALK TKD [27], and was not evident in detailed studies of the c-Met TKD [39]. Although the physiological importance of such apparent autodephosphorylation remains unclear, it appears to be primarily a characteristic of TKDs that closely resemble IRK in terms of the structural details of their autoinhibition and phosphorylation-dependent activation [25,38,49].

TrkA TKD autophosphorylates more rapidly than TrkB TKD, with a similar order of phosphorylation site usage that resembles IRK
To analyze Trk TKD autophosphorylation in more detail, we followed the reaction by Western blotting with phospho-specific antibodies -with a view to comparing phosphorylation site usage in addition. Three antibodies were used. One recognizes phosphorylation of the first tyrosine in the pTyrxxxTyr-Tyr motif (pTyr 676 in TrkA, pTyr 702 in TrkB). Another recognizes the second and third phosphotyrosines in the TyrxxxpTyr-pTyr motif (pTyr/pTyr 680/681 in TrkA, pTyr/pTyr 706/707 ). The third recognizes the C-terminal tail phosphotyrosine (pTyr 791 in TrkA, pTyr 817 in TrkB).
Autophosphorylation studies using 5 μM TKD, at 15°C (Fig. 2), showed a statistically significant difference in phosphorylation of TrkA and TrkB with each of the phosphospecific antibodies at 10 min. In the 2-color Western blots shown in Fig. 2A, the 10 min time-point shows a clear signal with all three phospho-specific antibodies for TrkA but not TrkBillustrating a further extension of the lag phase for TrkB autophosphorylation as temperature is reduced. In the plots shown in Fig. 2B, P values for the TrkA/TrkB difference (adjusted using the Bonferroni correction for multiple comparisons) ranged from 0.006 (pTyr 791/817 ) to 0.01 (pTyr 676/702 ) at 10 min, with no significant differences at other time points. This trend was also maintained at other temperatures (Fig. S4).
The order of autophosphorylation site usage appears to be the same for the TrkA and TrkB TKDs in these experiments, with the activation loop tyrosines being phosphorylated first, and phosphorylation of the C-terminal tail tyrosine (Tyr 791 in TrkA, Tyr 817 in TrkB) following (Fig. 2). We used Phos-tag gels [28] to further assess this, as shown in Fig. 3, which show that essentially all of the TKD is phosphorylated within 10 min. We cannot distinguish between TrkA TKD that is singly and triply phosphorylated in the YxxxYY motif using the pTyr 676 and pTyr/pTyr 680/681 antibodies. The signals for both of these antibodies increase with the same kinetics, and they recognize bands with the same mobilities in Phos-tag gels -suggesting that the three tyrosines are phosphorylated with very similar kinetics. It is clear from Fig. 3C, however, that Tyr 791 is phosphorylated later than the YxxxYY motif, as also suggested by Fig. 2B -resulting in the 4p TrkA TKD species in Fig. 3D. TrkB TKD is broadly similar, but the pTyr/pTyr 706/707 antibody appears to recognize a more mobile (presumed 2p) species at 10 min (Fig. 3B) that is followed at 20 min by a 3p species recognized by the pTyr 702 antibody (Fig. 3A). Thus, in TrkB TKD, the first Tyr of the YxxxYY motif appears to be phosphorylated after the 706/707 pair. Again, the C-tail tyrosine (Tyr 817 ) is phosphorylated last (Fig. 3C), to yield the 4p species that appears at 30 min (Fig. 3D). With the caveat that we cannot distinguish between species phosphorylated at one or both of the 'YY' tyrosines in the YxxxYY motif, these characteristics appear similar to IRK [50], IGF1-R [37], and MuSK [36]. The TrkA and TrkB TKDs both appears to be phosphorylated first at one of the tyrosines in the YY pair, and then at the first tyrosine of the YxxxYY motif. Tris phosphorylation of this motif occurs late in the reaction for IRK [50], and would not change recognition by the pTyr/pTyr 680/681 antibody in the case of the Trks. We therefore conclude that phosphorylation site usage of TrkA and TrkB resembles IRK, IGF1-R, and MuSK -with no substantial differences between TrkA and TrkB. Again, the TrkA/B TKDs resemble their closest structural relatives in terms of phosphorylation site usage rather than ALK or c-Met, where the order of YxxxYY phosphorylation appears to be distinct [39,[51][52][53].

TrkA TKD crystallographic dimer not observed in TrkB TKD crystal structures
Since autophosphorylation of IRK-related tyrosine kinases appears to be intermolecular [37,50], we hypothesized that the more rapid autophosphorylation seen for the TrkA TKD might reflect a higher preference for self-association than for TrkB TKD. In studies using sedimentation equilibrium analytical ultracentrifugation (SE-AUC) we could see no difference in TKD oligomerization at >10 μM, with both proteins sedimenting as ideal monomers of 34.8 and 35.1 kDa, respectively (Fig. S2). However, in our pilot experiments for kinase assays with dephosphorylated Trk TKDs, we observed a significant concentration dependence in our initial velocity values for TrkA but not TrkB (Fig. S5), and only when the TrkA TKD concentration exceeded approximately 60nM (Fig. S5A). This finding prompted us to perform the subsequent assays discussed above at concentrations that we established do not influence activity (30 nM for TrkA, 400 nM for TrkB, see Materials and Methods). However, we hypothesized that this phenomenon might be related to the higher autophosphorylation rate seen for TrkA TKD than for TrkB TKD in Figs. 2 and 3, and might reflect a degree of self-association that was not evident in SE-AUC studies. Indeed in our initial crystal structure of the (inactive) TrkA TKD -PDB ID 4GT5 [25] we observed two crystallographic dimers (Fig. 4A,B) that searches using the PISA server [54] identified in 23 additional TrkA TKD crystal structures that have since been deposited to the Protein Data Bank (PDB). It should be noted that there are an additional 24 PDB entries for TrkA TKD that show different crystal packing -so the TrkA TKD does not always crystallize in the dimers shown in Fig. 4A and B. However, the 24 cases in which it is seen represent at least 3 different crystal forms/environments and bury a total surface area of 1,290 Å 2 (Fig. 4A) or 1,250 Å 2 in PDB ID 4GT5 (Fig. 4B). Similar packing is not seen in the 4 TrkB TKD structure entries in the PDB. Importantly, the two TrkA TKD dimers involve residues that differ between TrkA and TrkB (Fig. 4C). For the 'face-to-face' dimer in Fig. 4A, side chains of residues between αEF and αF and in αG that differ between TrkA and TrkB contribute to the dimer, notably the side-chains of Thr 734 (Asn in TrkB), Asp 738 (Glu in TrkB) and Leu 700 (Met in TrkB), marked in both Fig, 4A and (with arrows) in Fig. 4C. The other dimer is largely mediated by the kinase insert domain (KID), which is the region of greatest sequence divergence between TrkA and TrkB (which are 76% identical) as shown in Fig. 4C.
Stabilization of the face-to-face dimer for TrkA as a result of the sequence differences between the two TKDs in its interface could promote its trans-autophosphorylation compared with that of TrkB, and this might explain our findings. It seems more plausible, however, that the KID-mediated dimer is the key. TrkA-specific residues play a key role in interactions between the two TKDs in this dimer, with Glu 615 playing a key role, interacting across the interface with Arg 593 in the hinge region of the dimer partner (Fig. 5). Asp 616 in the KID also plays a key role in setting up this interaction, forming predicted hydrogen bonds with Arg 593 and His 594 in its own hinge region. The fact that KID residues that seem to mediate the TrkA TKD 'KID' dimer shown in Fig. 4B also interact with residues in the hinge region of the kinase domain further suggests that dimerization might modulate conformational dynamics in the hinge region to alter TKD activity -as has been reported for other kinases [55,56]. Moreover, as pointed out by Bertrand et al. [57], there are no such interactions in TrkB, and no similar dimer has been seen in TrkB TKD crystals.

KID mutations in TrkA mutant slow progression of TKD autophosphorylation
The results outlined above prompted us to hypothesize that swapping the TrkA KID for the TrkB KID would yield a TKD that resembles TrkB in autophosphorylation assays. We therefore mutated the TrkA KID to resemble that seen in TrkB as shown in Fig. 5A, to yield TrkA B-KID . The TrkA B-KID TKD has a shorter KID (by 2 residues) and lacks Glu 615 and Asp 616 , which are central in intermolecular and intramolecular interactions respectively in the TrkA TKD KID/hinge dimer as mentioned above (see Fig, 5B). To examine the functional consequences of this change we used a radioactive autophosphorylation assay to compare autophosphorylation profiles for TrkA, TrkA B-KID , and TrkB TKDs. As shown in Fig. 5C, TrkA (at 4 μM) became autophosphorylated significantly more rapidly than TrkB TKD as expected, with significant differences at 1.5, 2, 3, and 4 min (after which TrkB catches up). The TrkA B-KID variant showed a slightly slower rate of autophosphorylation than wild-type TrkA TKD, but was not reduced to the rate seen for TrkB TKD. Unpaired Student's t-tests suggested significant differences (P < 0.05) at the 2, 3, and 4 min time points (Fig. 5C), although significance was lost following correction for multiple comparisons using the Bonferroni method. Taken together, these data are consistent with the crystallographically observed TrkA TKD dimer interface shown in Fig. 4B playing some role in the accelerated autophosphorylation of TrkA TKD. However, the effect is clearly only partial, possibly because we did not fully abolish dimerization with the KID mutations made, and possibly because both dimers seen across multiple TrkA TKD crystal structures play a role.

Knockdown of p75 converts TrkB (but not TrkA) signaling from sustained to transient
As mentioned in the Introduction, the initial motivation for this project was to understand why upregulated TrkB signaling causes cell proliferation and a poor prognosis in neuroblastoma, whereas TrkA conversely promotes cell differentiation and a favorable prognosis [2,3,5,6]. The TrkA TKD autophosphorylates more rapidly than TrkB TKD in vitro, which might suggest a difference in signaling kinetics for these two neurotrophin receptors. Moreover, subtle differences in signaling outcome have been reported when optogenetic activation of the membrane-associated intracellular regions of TrkA and TrkB was compared in Xenopus embryos [21]. It is well established, with TrkA as a key subject of the early studies in PC12 cells, that signaling kinetics can define cellular outcome. Sustained ERK signaling correlates with cell differentiation and transient ERK activation with cell proliferation [8,9]. NGF stimulates neurite outgrowth in parental PC12 cells, and BDNF also stimulates neurite outgrowth in PC12 cells engineered to express TrkB [58] -which they do not normally express [59]. As shown in Fig. 6A, this is associated with sustained activation of TrkB and ERK by BDNF. A similar signaling profile can also be recapitulated by sustained optogenetically-induced TrkB dimerization [60]. However, there have been several reports that TrkB signaling kinetics can be altered, in particular by neuronal activity [24,61], to make it more transient. We also recently showed for the EGF receptor that the same RTK can signal either in a transient manner (leading to proliferation) or a sustained manner (to promote differentiation), depending on how it is activated [14,15]. Such a difference -dependent on cellular context -may explain why TrkB promotes proliferation in neuroblastoma, but differentiation in PC12 cells.
One other aspect of neurotrophin signaling that correlates with an improved prognosis in neuroblastoma is expression of the low affinity p75 neurotrophin receptor, p75 NTR [23]. p75 NTR is well known to promote NGF-induced TrkA signaling and support differentiation of sympathetic neurons [62], and also facilitates the ability of TrkB to promote neuronal survival [63]. Upregulation of p75 NTR appears to reduce the tumorigenicity of neuroblastoma cells in vivo [22]. Since p75 NTR interacts with TrkB and facilitates its ability to promote neuronal survival [63], we wondered if knocking down p75 NTR in TrkB-PC12 cells would make BDNF-induced TrkB signaling more transient. As shown in Fig. 6B [64], which we confirmed (Fig. 6D). We therefore suggest that interaction with p75 NTR has different effects on the kinetics of ligand-induced signaling by TrkA and TrkB, and may be required for TrkB (but not TrkA) signaling to be fully sustained. We speculate that this difference could be related to the ability of the TrkA TKD to self-associate and trans-autophosphorylate more efficiently than TrkB TKD without facilitation of dimerization by p75 NTR .

CONCLUSIONS
The TrkA and TrkB neurotrophin receptors are homologous RTKs that initiate broadly similar signaling pathways upon ligand-induced activation of their TKDs [65]. However, several reports have demonstrated that TrkA and TrkB expression and activation can result in distinct cellular outcomes in the same cell type [4,7,58,66]. This is particularly notable in neuroblastoma, where TrkB signaling correlates with a poor prognosis, and TrkA signaling with a favorable one [2,67]. Our biochemical analysis of the TrkA and TrkB TKDs uncovered no substantial differences that can satisfyingly explain the distinct signaling of the two receptors on their own, but did find that TrkA TKD autophosphorylates faster than TrkB TKD in vitro. The initial lag phase seen for TrkB, and its slower autophosphorylation, might allow protein tyrosine phosphatases to 'compete' more effectively with kinase activity to make TrkB activation more transient than that of TrkA. NGF-induced TrkA signaling has been well studied, and its sustained nature has underpinned studies that link signaling kinetics to cellular outcome -with sustained responses promoting differentiation and transient responses promoting proliferation. If TrkB signaling is intrinsically more transient, this could explain its association with proliferation and poor prognosis in neuroblastoma, and the contrast with TrkA's association with favorable outcome.
Interestingly the differences between TrkA and TrkB, which seem to depend in part on deviations in sequence of the kinase insert domain, resemble distinctions reported between the non-receptor tyrosine kinases Itk and Btk [68]. By making Itk/Btk chimeras, Joseph et al.
reported that an increase in Itk's catalytic efficiency (k cat /K m , ATP ) of just ~3.5-fold was sufficient to make its signaling more Btk-like, with enhanced, more prolonged signaling of the receptor that controls it [68]. This also correlated with enhanced activation loop phosphorylation. These findings resemble the situation with TrkA, which has a ~2-fold higher catalytic efficiency than TrkB (Table 1), and is more readily phosphorylatedpossibly to allow a more readily activatable and sustained response. Along similar lines, studies of autophosphorylation kinetics of the fibroblast growth factor receptor (FGFR)-1 and −2 [69] have shown that FGFR2's TKD is less stringently autoinhibited than that of FGFR1, leading to faster in vitro autophosphorylation kinetics for FGFR2. This may explain some signaling differences between these receptors, where signaling duration determines phenotype [16,17,70] -as we are suggesting for TrkA and TrkB. The differences might arise through kinetic proofreading-type effects [71] or through differences in trafficking that have been described for TrkA and TrkB [72,73], and might control access to intracellular tyrosine phosphatases as described for EGFR [74].
Several recent studies have indicated that the phenotypic outcome of cell signaling by RTKs depends on the kinetics of receptor activation itself [12][13][14][15]. The studies described here for TrkA and TrkB suggest that even subtle differences can be important -consistent with previous results for FGFR isoforms as well as Itk and Btk [16,17,68,75]. The link between poor prognosis in neuroblastoma and TrkB signaling may reflect its more stringent autoinhibition and ability to signal transiently in certain cellular contexts such as those in neuroblastoma, whereas TrkA signaling remains sustained -promoting differentiation and more favorable outcome. With selective inhibition of one Trk isoform over another remaining a major therapeutic challenge [76], these data suggest that exploiting differential kinetics at intermediate inhibitor doses may have some value in certain settings.  Relative phosphotyrosine signals in Western blots (green) were determined by first normalizing the maximum signal for each phosphospecific antibody to the total Trk signal (red), and plotting (in panel B) the signal at each time as a proportion of the maximum for that site. At least three biological repeats were performed for each experiment. Errors bars are standard deviations. Statistical significance of differences between TrkA and TrkB for each pTyr-specific antibody was determined using the Bonferroni-Dunn method (α = 0.05) to correct for multiple comparisons. Significant differences were seen only at the 10 min time point. Adjusted P values (after Bonferroni correction) were **P < 0.01.  The face-to-face TrkA TKD dimer mentioned in the text, seen in 50% of available TrkA TKD crystal structures, shown specifically for PDB ID 4GT5 [25]. Tyrosine autophosphorylation sites are marked where visible in the orientation shown. So are the Clobe residues that interact across the interface (Leu 700 between αEF and αF and Thr 734 / Asp 738 in αG), which are all different in TrkB.

Ni-NTA
(B) The KID-mediated TrkA TKD dimer described in the text, also seen in 50% of available TrkA TKD crystal structures, shown specifically for PDB ID 4GT5 [25]. Again, tyrosine autophosphorylation sites are marked where visible in the orientation shown. So are the kinase insert domain (KID) residues that interact across the interface: Lys 609 (Val in TrkB), Glu 615 (Asn in TrkB), Asp 616 (deleted in TrkB), Val 617 (deleted in TrkB), Ala 618 (Pro in TrkB), and Pro 621 (Pro in TrkB). The KID is also marked, as are N-and C-termini where visible.