Structural Insights into Pseudokinase Domains of Receptor Tyrosine Kinases

1 Structural insights into pseudokinase domains of receptor tyrosine kinases Joshua B. Sheetz1,2,9, Sebastian Mathea3,4,5,9, Hanna Karvonen6,9, Ketan Malhotra1,2,10, Deep Chatterjee3,4, Wilhelmiina Niininen6, Robert Perttilä6, Franziska Preuss3,4, Krishna Suresh7, Steven E. Stayrook1,2, Yuko Tsutsui1,2, Ravi Radhakrishnan7, Daniela Ungureanu6,8,*, Stefan Knapp3,4,5,*, and Mark A. Lemmon1,2,11,* 1Department of Pharmacology, Yale University School of Medicine, New Haven, CT 06520, USA. 2Yale Cancer Biology Institute, Yale University West Campus, West Haven, CT 06516, USA. 3Institute for Pharmaceutical Chemistry, Johann Wolfgang Goethe-University, D-60438 Frankfurt am Main, Germany 4Buchmann Institute for Molecular Life Sciences, Structural Genomics Consortium, D-60438 Frankfurt am Main, Germany 5German Cancer Consortium DKTK Frankfurt/Mainz, D-60438 Frankfurt, Germany 6Cancer Signaling, Faculty of Medicine and Health Technology and BioMediTech, Tampere University, Tampere, 33014, Finland. 7Department of Bioengineering, University of Pennsylvania, Philadelphia, PA 19104, USA. 8Genome-Scale Biology, Research Programs Unit, University of Helsinki, FI-00014, Helsinki, Finland 9These authors contributed equally 10Present address: MRC Laboratory of Molecular Biology, Cambridge Biomedical Campus, Cambridge CB2 0QH, UK. 11Lead contact: mark.lemm[email protected] *Correspondence: [email protected] (D.U.), [email protected]-frankfurt.de (S.K.), [email protected] (M.A.L.) Running Title: RTK pseudokinase structures This is the accepted manuscript of the article, which has been published in Molecular cell, 2020, 79:3, 390-405. https://doi.org/10.1016/j.molcel.2020.06.018 2 SUMMARY Despite their apparent lack of catalytic activity, pseudokinases are essential signaling molecules. Here, we describe the structural and dynamic properties of pseudokinase domains from the Wntbinding receptor tyrosine kinases (PTK7, ROR1, ROR2, RYK), which play important roles in development. We determined structures of all pseudokinase domains in this family, and found that they share a conserved inactive conformation in their activation loop that resembles the autoinhibited insulin receptor kinase (IRK). They also have inaccessible ATP binding pockets, occluded by aromatic residues that mimic a cofactor-bound state. Structural comparisons revealed significant domain plasticity, and alternative interactions that substitute for absent conserved motifs. The pseudokinases also showed strikingly similar dynamic properties to IRK. Despite the inaccessible ATP site, screening identified ATP competitive type-II inhibitors for ROR1. Our results set the stage for an emerging therapeutic modality of “conformational disruptors” to inhibit or modulate non-catalytic functions of pseudokinases deregulated in disease. KEYWORDS Receptor tyrosine kinases, pseudokinases, Wnt signaling, growth factor signaling, protein conformation, targeted therapies, ponatinib, GZD824, cancer 3 INTRODUCTION As receptor tyrosine kinase (RTK) signaling mechanisms have become increasingly well understood (Lemmon and Schlessinger, 2010), those mediated by the subset of RTKs thought to lack kinase activity remain largely enigmatic (Manning et al., 2002; Mendrola et al., 2013). Approximately 10% of the >500 predicted human protein kinases are classed as pseudokinases, assumed to lack phosphotransferase activity because of substitutions of conserved and essential catalytic residues (Boudeau et al., 2006; Manning et al., 2002). Five of the 58 human RTKs fall into this category: ErbB3/HER3, protein tyrosine kinase 7 (PTK7)/colon carcinoma kinase 4 (CCK4), EphB6, EphA10, and SuRTK106 (which lacks an extracellular region and was not included in our study). Three additional RTKs (ROR1, ROR2 and RYK) have substitutions in conserved motifs and lack both kinase activity (Debebe and Rathmell, 2015; Gentile et al., 2011; Katso et al., 1999) and ATP binding (Murphy et al., 2014). These ‘RTK pseudokinases’ are depicted in Fig. 1A, and the pseudokinase domain sequences are compared in Fig. S1. Beyond the basic question of how RTKs can signal without kinase activity, strong links of RTK pseudokinases to disease have prompted significant interest in developing therapeutic approaches to inhibit their signaling functions (Bailey et al., 2015; Kung and Jura, 2019). ErbB3 is associated with lung, colon, gastric, and other cancers (Amin et al., 2010; Jaiswal et al., 2013), especially in resistance to inhibitors of other EGFR family members. PTK7, ROR1/2, and RYK are all involved in Wnt signaling (Green et al., 2014; Karvonen et al., 2018; Peradziryi et al., 2012), and elevation of their expression is associated with cancer development and metastasis in several hematological and solid tumors (Debebe and Rathmell, 2015; Karvonen et al., 2018). EphB6 and EphA10 are also reported to play important roles in tumorigenesis (Liang et al., 2019; Nagano et al., 2014). Antibodies targeting the ErbB3 extracellular region are in clinical development for several solid tumors (Jacob et al., 2018). ROR1 antibodies are being tested clinically in chronic lymphocytic leukemia (CLL) and other cancers (Choi et al., 2018; 4 Choi et al., 2015), as is a small molecule claimed to target the ROR1 extracellular region (Liu et al., 2019). PTK7 is also the target of an antibody-drug conjugate in clinical development (Damelin et al., 2017; Katoh, 2017). No validated pharmacological agents that target intracellular regions of RTK pseudokinases have been described, however – largely because their mechanistic role in signaling is poorly understood. The nucleotide binding properties of RTK pseudokinase domains vary. Those from PTK7, ROR1/2, and RYK all fail to bind ATP (Murphy et al., 2014), whereas those from ErbB3 and EphB6 bind ATP quite strongly (Jura et al., 2009; Murphy et al., 2014; Shi et al., 2010). ErbB3 also displays low level kinase activity (Shi et al., 2010; Steinkamp et al., 2014), inhibition of which does not block signaling (Novotny et al., 2016; Shi, 2012). Without phosphotransferase activity, pseudokinases are thought to signal by allosterically regulating binding partners (Zeqiraj and van Aalten, 2010). To help understand this role, we combined crystallographic, biophysical, and chemical biology studies. We show that RTK pseudokinase domains adopt structures closely related to the inactive state of their closest kinase-active homologue, and with comparable dynamics – suggesting a similar propensity for conformational changes. We also demonstrate that a pseudokinase domain that fails to bind ATP can nonetheless be targeted with small molecules that bind its vestigial ATP-binding site. The insight provided here into the structure and dynamics of pseudokinase domains suggests approaches for defining the mechanisms of their function and for targeting them pharmacologically. RESULTS AND DISCUSSION As a first step, we aimed to establish a comprehensive structural description of RTK pseudokinases – adding to structures of the ErbB3 (Jura et al., 2009; Shi et al., 2010) and ROR2 (Artim et al., 2012) pseudokinase domains. We determined crystal structures of pseudokinases from the remainder of the Wnt-binding RTK (PTK7/ROR/RYK) family, and also modeled the Eph RTK pseudokinases. 5 The activation loops of PTK7, ROR, and RYK pseudokinase domains mimic autoinhibitory interactions seen in inactive IRK We determined crystal structures of the PTK7 and RYK pseudokinase domains (to 1.95 Å and 2.38 Å respectively) plus that of ROR1 bound to a small molecule (described in a later section), as summarized in Table 1. These pseudokinase domains all contain a YxxxYY motif (or its vestiges) in the activation loop, similar to that in the insulin receptor kinase domain (IRK). The activation loop conformations also closely resemble that of inactive, autoinhibited, IRK (Fig. 1B) – immediately raising the question as to what activity or property is ‘autoinhibited’ in the pseudokinases. Inactive IRK is autoinhibited by simultaneous projection of a key tyrosine side chain (Y1162) into the protein substrate-binding site and occlusion of the ATP-binding site by the IRK DFG motif (Hubbard et al., 1994). The PTK7, RYK, and ROR pseudokinases all similarly place a tyrosine in precisely the same location as Y1162 of IRK (Fig. 1B), and the YxxxYY motif configuration (FxxxYH in RYK) is highly conserved. By contrast, the ErbB3 activation loop adopts a different conformation – instead resembling the inactive EGFR kinase domain (Jura et al., 2009; Shi et al., 2010). ErbB3 does place its single activation loop tyrosine (Y848; not conserved in EGFR) in a very similar location to Y1162 in IRK, however (Littlefield et al., 2014), suggesting that projecting an activation loop tyrosine into the (protein) substrate binding site may be common across RTK pseudokinases (Fig. 1C). The ATP-binding site of inactive IRK is occluded in two ways (Hubbard et al., 1994). First, the side-chain of the DFG phenylalanine (F1151) projects into the adenine ring binding site (Fig. 1D: IRK inactive). Second, the main chain following the DFG motif occludes the binding site for the ATP βand γ-phosphate groups (Fig. 1D). This is the ‘DFG-out’ conformation (Huse and Kuriyan, 2002; Modi and Dunbrack, 2019), frequently seen in inactive kinases – in which the aspartate side-chain (D1150) is also oriented away from the active site so that is cannot coordinate Mg2+. In active kinases, the DFG motif instead adopts an ‘in’ conformation, 6 allowing Mg2+ coordination and repositioning the phenylalanine side-chain to allow ATP binding and to complete the so-called regulatory spine (Fig. S2) or ‘R-spine’ (Kornev and Taylor, 2010). The Wnt-binding RTK pseudokinases all have an altered DFG motif (‘ALG’ in PTK7, ‘DLG’ in the RORs and ‘DNA’ in RYK; Fig. 1D). The conformation of this altered motif is always closer to ‘DFG-out’ than ‘DFG-in’ (Fig. 1D). As a result, each R-spine is broken (Fig. S2) – although in a different way for RYK than for PTK7 or ROR2 (which more closely resemble inactive IRK). Moreover, the leucine side-chains in the PTK7 ‘ALG’ motif (L949) and the ROR2 ‘DLG’ motif (L634) occlude the ATP-binding site (Fig. 1D). RYK is an exception, as its ‘DNA’ motif lacks a corresponding hydrophobic side-chain – although the main chain after this motif passes through (and occludes) the β/γ phosphate group-binding region (Fig. 1D). In contrast with the Wnt pseudokinase RTKs, the ErbB3 DFG motif adopts an ‘in’ conformation that completes the Rspine and leaves the ATP-binding site open to allow nucleotide binding (Figs. 1D and S2). Aromatic side-chains from the hinge region occlude the PTK7, ROR, and RYK ATPbinding sites ATP binding to PTK7 and ROR2 is prevented by projection of a tyrosine side-chain from the β5/αD hinge region (Y877 and Y555 respectively) into the adenine ring binding site (Fig. 1D). This position is two residues C-terminal to the gatekeeper, and is a tyrosine in ~50% of tyrosine kinases – although replaced by leucine (L1078) in IRK. The side-chain at this position typically projects away from the ATP binding site (as for L1078 in IRK and Y770 in ErbB3; Fig. 1D). The alternative side-chain orientation in PTK7 and ROR2 results from a rotation of their N-lobes about the β5/αD hinge (Fig. S3A,B). RYK does not undergo this N-lobe rotation (Fig. S3C), allowing its corresponding tyrosine (Y413) to adopt the canonical orientation (Fig. 1D). The RYK adenine ring binding site is instead occluded by the side-chain of a phenylalanine (F362) in the 7 altered β3 VAIK motif (AFVK in RYK). A leucine at this position in PTK7 (L828 in its VLVK variant of the VAIK motif) also contributes to ATP-binding site occlusion. Whereas the ATP binding sites in RTK pseudokinases are occluded by side-chains from multiple motifs, the IRK ATP-binding site is only obstructed by its DFG motif (Figs. 1D and S4A), suggesting that simply flipping the DFG motif from ‘out’ to ‘in’ might allow ATP binding. Indeed, IRK can bind Mg2+-ATP without activation loop phosphorylation (Till et al., 2001) as can be seen using differential scanning fluorimetry (DSF). DSF monitors the influence of ligand binding on the melting temperature (TM) of a protein, as reported by increased fluorescence of SYPRO Orange as it binds to thermally unfolded protein (Fedorov et al., 2012). Binding of Mg2+-ATP to unphosphorylated IRK destabilized the kinase domain by ~3.5˚C (Fig. S4B) – rather than stabilizing – which may reflect disruption of the autoinhibitory activation loop interactions described above. In contrast to IRK, the PTK7, ROR1, ROR2, and RYK pseudokinase domains showed no significant stability change upon adding 5 mM Mg2+-ATP (Fig. S4B), suggesting no binding, in agreement with previous work (Murphy et al., 2014). Moreover, a D644A mutation in ROR2 (equivalent to D1161A in IRK, which promotes ATP binding) failed to restore ATP binding. Another important difference between IRK and the Wnt-binding RTK pseudokinases is revealed by comparing their catalytic spines (‘C-spines’). The ‘C-spine’ is a non-contiguous hydrophobic motif defined by Kornev and Taylor (Kornev and Taylor, 2010). In ATP-bound active kinases, it is completed by the adenine ring of bound ATP, as seen in PKA and active IRK in Fig. S2. In an inactive kinase, by contrast, the C-spine is broken (see inactive IRK in Fig. S2) and the side-chain of the DFG-motif phenylalanine (from the R-spine) projects into the gap. Interestingly, the C-spine is fully contiguous in PTK7, ROR2 and RYK despite no direct contribution from R-spine (DFG motif) residues or ATP (Fig. S2) – although the side-chain of the ALG/DLG motif leucine does abut the intact C-spine in PTK7 and ROR2. 8 PTK7, ROR, and RYK pseudokinases adopt different ‘hybrid’ conformations stabilized by unique sets of interactions Although the PTK7, ROR, and RYK pseudokinase domains all closely resemble inactive IRK in their activation loop configurations (Fig. 1B), they vary substantially in αC helix position (Fig. 2A) and in other structural features, details of which can be described as follows. PTK7: Although PTK7 appears autoinhibited by its activation loop (and hinge tyrosine), it has an active-like ‘in’ αC disposition (Fig. 2A). As a result, the distance between the side-chains of the conserved αC glutamate (E846) and the VAIK motif β3 lysine (K830) is just 2.7 Å. This allows the αC/β3 salt bridge – a hallmark of active kinases – to form (Fig. 2B), and places K830 so that it could support ATP binding if the ATP-binding site were not occluded. PTK7 thus adopts an unusual ‘pseudo’ active-like kinase domain conformation. ROR2: The ROR2 αC helix adopts the ‘out’ position as in inactive IRK (Fig. 2A). The sidechains of the αC glutamate (E524) and β3 lysine (K507) are therefore too far apart (~5 Å) for the hallmark αC/β3 salt bridge of active kinases to form (Fig. 2C) – and the β3 lysine (K507) is >7 Å away from where the ATP phosphates would reside in a putative nucleotide-bound state (Artim et al., 2012). A hydrogen bond between D633 in the degenerated ROR2 DFG motif (DLG) and R528 in αC also connects these two motifs in an unusual way (Fig. 2C). RYK: The RYK pseudokinase domain has several unique structural features. First, as in PTK7, its αC helix adopts an active-like ‘in’ position (Fig. 2A), apparently ‘pushed’ into this position by a 2.5-turn α-helix (αB) at the domain’s amino-terminus (Fig. 2D). This arrangement is reminiscent of the EphB2 kinase domain (Wybenga-Groot et al., 2001) and involves intimate clustering of hydrophobic side-chains at the αB/αC interface (Fig. 2D, left). Surprisingly, although the conserved β3 lysine (K364) is retained in RYK’s degenerated VAIK motif (AFVK), and αC is ‘in’, the αC/β3 salt bridge that is a hallmark of active kinases does not form in RYK (Fig. 2D, right). Instead, K364 forms a predicted salt-bridge (2.6 Å) with D483 in RYK’s 9 degenerated DFG (DNA) motif, and the αC glutamate (E381) is close (3.0 Å) to R488 in the activation loop (three residues C-terminal to the degenerated DFG motif). The absence of a large hydrophobic side-chain in RYK’s degenerated DFG motif also appears to be compensated for by the positioning of F491 in the small helix C-terminal to this motif (Fig. S4A). In addition, the presence of a glutamine (Q337) in place of the first glycine of the glycine rich loop in RYK contributes to steric exclusion of RYK’s ATP binding site (Fig. S4A). Differences in Eph pseudokinase domain autoinhibition To extend these comparisons across the RTK pseudokinases, we modeled the EphA10 and EphB6 pseudokinase domains, so far of unknown structure. Apart from EphA10 and EphB6, the remaining 12 human ephrin receptors (Ephs) are catalytically active (Liang et al., 2019). They are regulated by phosphorylation of key juxtamembrane (JM) tyrosines, which relieves intramolecular autoinhibition to allow trans-autophosphorylation of a conserved activation loop tyrosine (Hornbeck et al., 2015) that corresponds to Y1162 in IRK. EphA10 retains this conserved activation loop tyrosine, but not the regulatory JM tyrosines. Conversely, EphB6 retains the regulatory JM tyrosines but not the activation loop tyrosine (and has an unusually short activation loop). In the few published Eph kinase structures in inactive conformations (Dong et al., 2014; Kung et al., 2016; Overman et al., 2014), the activation loop tyrosine is ordered and projects into the substrate binding site like Y1162 of inactive IRK. Similarly, our EphA10 pseudokinase model, (based on EphB3; see STAR Methods) places this tyrosine (Y801) in the autoinhibitory location (Fig. 3A,B). Y801 simultaneously contacts the arginine close to the end of the catalytic loop (R774 in EphA10) and an adjacent EphA10-specific histidine (H775) – substituting for interaction with the (absent) HRD aspartate (Fig. 3B). This IRK-like docking of Y801 constrains the EphA10 activation loop so that the backbone of the DFG motif (GFG in EphA10) passes through and occludes the ATP-binding site as in IRK (Fig. 16 the static crystallographic view, but our HDX data (Fig. 7A,B) argue that binding of ponatinib or GZD824 significantly increases its solvent accessibility and dynamics. Moreover, the C-terminal part of the activation loop – immediately following the YxxxYY motif – was disordered in the ROR1/ponatinib complex structure, suggesting that accommodating ponatinib partly disrupts autoinhibitory docking of the activation loop in this region. Another ponatinib-induced conformational change is seen in the DFG motif (DLG in ROR1/2). Ponatinib restrains this motif in ROR1 so that the leucine side-chain is disordered and the DLG aspartate (D633) side-chain is displaced (by ponatinib’s trifluoromethylphenyl group) towards H613 of the HRD motif (see Fig. 7C inset). This forces ROR1’s DLG motif into the same conformation seen for the DFG motif in ponatinib-bound ABL (O'Hare et al., 2009) – typical of that in kinases bound to type-II inhibitors (Modi and Dunbrack, 2019). The largest-scale difference between the structures of the ROR1/ponatinib complex and ROR2 (which does not bind ponatinib) is a rotation (~20˚) of the entire N-lobe about an axis (yellow rod in Fig. 7D) nearly coincident with strand β4. Ponatinib appears to simultaneously ‘push’ on the left-hand side of the N-lobe (near β1 and the β5/αD hinge) and ‘pull’ on its righthand side (near αC) to achieve this rotation. Ponatinib displaces the side-chain of the β5/αD hinge tyrosine (Y554 in ROR1, Y555 in ROR2) that occludes ROR2’s ATP-binding site, as indicated by the red arrow on the left of Fig. 7D – inserting its imidazo(1,2-b)pyridazine moiety into that pocket. At the same time, ponatinib (docked onto the C-lobe) associates intimately with helix αC of ROR1, through an interaction between the nitrogen of its amide linker and E523 in αC plus van der Waal’s contacts with the (hydrophobic) side-chains of L526, M527, and L530 in αC. All the corresponding residues are solvent-exposed in ROR2 (E524, L527, R528, and L531). Importantly, an arginine at position 528 of ROR2 (replaced by M527 in ROR1 – which contacts ponatinib) likely explains why ponatinib binds only ROR1, and not ROR2. An arginine in this position would interfere with ponatinib/αC interactions. 17 Together with our HDX-MS studies, this comparison of the ponatinib-bound ROR1 and unliganded ROR2 pseudokinase domain structures reveals one way in which a kinase inhibitorlike small molecule binder can modulate the conformation of a pseudokinase domain. Many effectors and activators bind to the N-lobe region of other kinases that is the focus of the conformational changes described here (Jura et al., 2011). Any similar mode of interaction that ROR1 participates in will thus be significantly affected by ponatinib or GZD824 binding. Ponatinib and GZD824 may therefore serve as pharmacologically privileged potential starting points for developing selective small molecule ROR1 interactors that could induce sufficiently extensive conformational changes in the pseudokinase domain to impair scaffolding functions of ROR1 in disease. Although efforts to investigate the effects of these inhibitors on the ROR1mediated signaling seen in Fig. 5C are currently confounded by the promiscuity of ponatinib and GZD824 as kinase inhibitors, both compounds clearly inhibit ROR1-dependent activation of ERK and AKT phosphorylation by Wnt5a (Fig. S7D) at concentrations that we know maximally stabilize the receptor in CETSA experiments (Fig. S6). Moreover, it is interesting to note that Bicocca et al. found that ponatinib robustly inhibits AKT phosphorylation in acute lymphoblastic leukemia cells (Bicocca et al., 2012), in which ROR1 also appears to signal to AKT and ERK. CONCLUSIONS Our structural and modeling studies argue that the inactive kinase domains of RTK pseudokinases nonetheless retain the autoinhibitory interactions seen in their kinase-active relatives – the Wnt-binding RTKs sharing that autoinhibitory YxxxYY activation loop motif conformation seen in inactive IRK. RTK pseudokinases all seem either to bind nucleotide (as in ErbB3 and EphB6) or to mimic an ATP-bound form (as in PTK7, ROR1/2 and RYK) – as reported for others (Hammaren et al., 2015; Scheeff et al., 2009). Which of these two classes they fall into appears to correlate with whether or not the ATP-binding site is occluded in the inactive conformation of their closest kinase-active homologues. 18 One important question that arises from our results is why the pseudokinases retain the autoinhibitory interactions seen in their kinase-active relatives. One hypothesis is that pseudokinase domains function as rigid scaffolds, and the autoinhibitory interactions could promote rigidity (Kornev and Taylor, 2009; Murphy et al., 2014; Patel et al., 2017; Scheeff et al., 2009). On the contrary, our HDX-MS data argue that the PTK7, ROR, and RYK pseudokinase domains all display similar conformational dynamics to IRK, despite mimicking an ATP-bound state. It is well known that canonical protein kinases such as IRK undergo transitions between active and (family-specific) inactive conformations (Huse and Kuriyan, 2002). The RTK pseudokinases are likely to have similar capabilities, providing a possible mechanism for controlling their ability to interact with binding partners and thus regulate signaling (Jacobsen and Murphy, 2017; Kung and Jura, 2019). In IRK, the inactive-to-active conformational transition is promoted by tyrosine phosphorylation of the activation loop YxxxYY motif (Hubbard, 2013). YxxxYY motif tyrosine phosphorylation has been reported for both PTK7 and ROR2 (Hornbeck et al., 2015), and could promote structural changes similar to those seen in IRK activation. Indeed, phosphorylation of the ROR1 and ROR2 pseudokinase domains has been suggested to be mediated by SRC recruitment (Akbarzadeh et al., 2008; Gentile et al., 2014). A related phosphorylation-dependent mechanism has also been described for the pseudokinase MLKL (mixed lineage kinase domain-like protein), where phosphorylation of the pseudokinase domain induces conformational changes, which control intraand intermolecular interactions that regulate necroptosis (Petrie et al., 2019). Thus, pseudokinases may not only be scaffolds that allosterically regulate active kinases, but may in fact be regulated as conformational switches. As with catalytically competent kinases, the dynamic structural changes involved in pseudokinase function might also be modulated by small molecules that bind the ATP site and modulate interand/or intra-molecular interactions. It might be possible to exploit this therapeutically. Several small molecules that bind the ATP-binding site of pseudokinases have been reported (Dhawan et al., 2016; Kung and Jura, 2019). For example, BMS-986165, an 19 ATP-competitive inhibitor that binds the pseudokinase domain of TYK2 (a JAK family member) is effective in several murine models of autoimmune disease and is now in clinical testing (Wrobleski et al., 2019). JAK pseudokinase domains play important roles in regulating activity of their adjacent catalytically-competent (JH1) kinase domains, highlighted by frequent patientderived JAK-activating mutations in the pseudokinase domain (Shan et al., 2014). In the case of TYK2, the unusual ATP site of its pseudokinase domain allowed selective inhibitor design that resulted in outstanding JAK isoform and kinome selectivity for BMS-986165 (Moslin et al., 2019; Wrobleski et al., 2019). Our findings support the value of similar concerted efforts to find small molecule modulators of the pseudokinase RTKs. Even if modulation of the properties of the pseudokinase domain by small molecules proves insufficient to block pseudokinase function, the ligands developed for the ATP-binding site of these proteins could instead be used for developing proteolysis targeting chimeras (PROTACs), as reported for both kinase-active RTKs (Burslem et al., 2018) and the RTK pseudokinase ErbB3 (Xie et al., 2014). In conclusion, the results presented here demonstrate the feasibility of targeting the apparently inaccessible ATP site of ROR1 and related RTK pseudokinases. Given the important role of ROR1 and the other Wnt-binding RTKs in mediating growth signals, their selective targeting in this way represents a potentially valuable strategy for the development of future therapeutic agents. 20 STARMETHODS Detailed methods are provided in the online version of this paper and include the following: • KEY RESOURCES TABLE • RESOURCE AVAILABILITY o Lead contact for reagent and resource sharing o Materials availability o Data and code availability • EXPERIMENTAL MODEL DETAILS o Cell culture • METHOD DETAILS o Plasmid construction for recombinant protein expression o Protein production and purification o Crystallization o Structure determination o Hydrogen-deuterium exchange-mass spectrometry (HDX-MS) o Differential scanning fluorimetry (DSF) o Small molecule screening o Homology models of EphA10 and EphB6 o BaF3 cell transfections o Cell viability and cell proliferation o Western blotting o Cellular thermal shift assays (CETSA) o in vitro kinase assays • QUANTIFICATION AND STATISTICAL ANALYSIS o Structure determination and analysis 21 o Analysis of HDX dynamics o Western blot image processing o Cell proliferation assays o Cellular thermal shift assays (CETSA) • DATA AND SOFTWARE AVAILABILITY SUPPLEMENTAL INFORMATION Supplemental information includes seven figures and two tables, and can be found in the online version of this article. ACKNOWLEDGMENTS We thank Kate Ferguson and members of the Lemmon, Knapp, Radhakrishnan, and Ungureanu laboratories for critical evaluation and discussion of the manuscript and the Tampere Flow Cytometry facility for their service. This material is based in part upon work supported in part by a National Science Foundation Graduate Research Fellowship (DGE1122492 to J.B.S.), the National Institutes of Health (R35-GM122485 to M.A.L. and U54CA193417 to R.R. and M.A.L.), the Academy of Finland (Grants 275525 and 284663 to D.U.), the Cancer Society of Finland (D.U.), and the Sigrid Jusélius Foundation (D.U.). The group of S.K. is grateful for support by the German cancer network DKTK and the SGC, a registered charity (no. 1097737) that receives funds from AbbVie, Bayer Pharma AG, Boehringer Ingelheim, Canada Foundation for Innovation, Eshelman Institute for Innovation, Genome Canada (through Ontario Genomics Institute), Innovative Medicines Initiative (EU/EFPIA), Janssen, Merck KGaA, MSD, Novartis Pharma AG, Ontario Ministry of Research, Innovation 22 and Science (MRIS), Pfizer, São Paulo Research Foundation-FAPESP, Takeda, and the Wellcome Trust. AUTHOR CONTRIBUTIONS J.B.S., D.U., S.M., K.M., S.K. and M.A.L designed the overall project and wrote the manuscript. J.B.S., S.M., K.M., S.K., and M.A.L. analyzed structures and generated structure figures. K.M., S.M., and S.E.S. carried out crystallographic structural studies, and J.B.S. and Y.T. performed HDX-MS. D.C., F.P., and J.B.S. performed DSF and other biochemical studies, H.K., R.P., W.N. and D.U. performed all cellular and CETSA studies, and K.S and R.R. carried out molecular modeling studies. All authors contributed to analysis of results and editing of the manuscript. DECLARATION OF INTERESTS The authors declare no competing interests. 23 LEGENDS TO FIGURES FIGURE 1 Pseudokinases retain autoinhibitory interactions of the insulin receptor (A) Domain composition of human RTK pseudokinases, with the extracellular region above, and intracellular region below the membrane. Domains are listed in the legend: pseudokinase (red), leucine-rich (L), cysteine-rich (Cys Rich), immunoglobulin (Ig), Frizzled cysteine-rich domain (Fz CRD), Kringle domain (Kr), Sterile Alpha Motif (SAM), and Wnt Inhibitory Factor (WIF) domain. (B) Activation loops of pseudokinase domains from PTK7 (slate blue), ROR2 (magenta), and RYK (green) are superimposed on that of IRK (black) in the context of the IRK surface (from PDB: 1IRK). YxxxYY tyrosines in IRK are labeled. The ROR2 structure is chain B of PDB: 3ZZW. (C) ErbB3 activation loop structure (Littlefield et al., 2014) from PDB: 4RIW, superimposed on the 1IRK surface as in (B), with ErbB3 colored orange. Y849 is labelled. (D) Close-up of residues surrounding the ATP binding site for active and inactive IRK (black), PTK7, ROR2, RYK, and ErbB3, colored as above. AMP-PNP is solid when seen in the relevant crystal structure and transparent when not. Residues from the DFG motif (and 4 residues beyond), the β5/αD hinge, and second residue of the VAIK motif are shown (VAVK in IRK, VLVK in PTK7, VAIK in ROR2, AFVK in RYK, VCIK in ErbB3). See also Figure S1 and Table 1. FIGURE 2 Overall structure of pseudokinase domains (A) Cartoon representing pseudokinase αC helix positions. Two orthogonal views of inactive IRK (1IRK) are shown in grey. Helix αC of IRK, which adopts the ‘out’ position is colored black. 24 Helix αC from ROR2 (3ZZW; magenta) and ErbB3 (4RIW; orange) is also ‘out’. By contrast, αC is ‘in’ for PTK7 (slate blue), RYK (green), and active IRK (1IR3). (B) PTK7 pseudokinase domain. The insert shows the predicted salt bridge between the αC glutamate (E846) and β3 lysine (K830). (C) ROR2 pseudokinase domain (chain B from 3ZZW). Insert shows absence of salt bridge between the αC glutamate (E524) and β3 lysine (K507), and alternate contact between D633 (in the DLG motif) and R528 (in αC). (D) RYK pseudokinase domain. Inserts show hydrophobic side-chains involved in packing between helices αB and αC (left) and connections between the DFG (DNA) motif region and helix αC (right). Interactions between D483 (in DNA motif) and the β3 lysine (K364) and between R488 (close to DNA motif) and αC glutamate (E381) are shown. Vestigial ATP-binding sites are labelled in inserts of (B-D) using a transparent AMP-PNP molecule positioned as in active IRK. See also Figure S2. FIGURE 3 Insights into EphA10 and EphB6 pseudokinases from modeling (A) Activation loops in models of the EphA10 (deep red) and EphB6 (olive) pseudokinases, superimposed on that of IRK (black) in the context of the IRK surface (from PDB: 1IRK). Y1162 in IRK and its EphA10 equivalent (Y801) are labeled. (B) Close-up of interactions involving Y801 for EphA10 (left) and Y1162 for IRK (right). Where Y1162 of IRK interacts with D1132 in the HRD motif and R1136 from the end of the catalytic loop, in EphA10 Y801 retains the arginine interaction (with R774), but interacts with a histidine at the very end of the catalytic loop (H775) making up for the lack of an HRD aspartate (replaced by glycine in EphA10). 25 (C) Close-up of residues around the ATP binding sites of EphA10 (left), and EphB6 (right), colored as above. An AMP-PNP molecule is shown – solid when predicted to bind and transparent when not. Residues are shown from the DFG motif (GFG in EphA10, and RLG in EphB6) – and 4 residues beyond – as well as the β5/αD hinge. See also Figure S3. FIGURE 4 Comparison of pseudokinase domain dynamics by HDX (A) HDX data at 1 minute for unphosphorylated IRK across 73 peptides (represented as short horizontal lines). Lines are colored according to percent exchange at 1 min (using the scale at right). Secondary structure of IRK is shown at top. Data for longer timepoints are plotted in the lower part of the figure, with the x-axis representing the median residue number of the peptide. Locations of the β5/αD hinge, HRD motif, activation loop (A-loop) and YxxxYY motif are noted. Errors represent SD from three independent labeling experiments. (B) Data (1 min) from the upper part of (A) plotted on the (inactive) IRK structure using the same color scheme as in (A). Blue represents less, and red more, exchange. (C) Comparison of HDX data (mean ± SD) for PTK7 (slate blue), ROR2 (magenta), RYK (green) and ROR1 (cyan) pseudokinase domains with data for IRK (grey) at 10 s, 1 min, 10 min, and 2 h. X-axis represents IRK-equivalent median residue number. IRK data are depicted as the range for each point. Errors represent SD from three independent labeling experiments. (D) Variance of ‘exchangeability index’ values – determined as in STAR Methods – for pseudokinases versus IRK (expressed as how many times SD/σ was the IRK value away from the pseudokinase mean), plotted on the inactive IRK structure. A location was assigned no color if within 2σ, but colored black or orange as marked if beyond 2σ (95% confidence). See also Figure S4 and Table S1. 32 Crystallization Crystals of PTK7774-1069 were obtained using the hanging-drop vapor diffusion method, by mixing equal volumes of protein and reservoir solutions and equilibrating over the reservoir solution at 21˚C. The final protein concentration used was 10.5 mg/ml in 20 mM HEPES (pH 7.0), 150 mM NaCl, and 2 mM DTT. Crystals were obtained with a reservoir solution of 30% (w/v) PEG monomethyl ether 5,000, 100 mM ADA (pH 6.5), and 100 mM (NH4)2SO4 (Jena Bioscience JBKinase Screen 3; condition D1). Crystals of the RYK pseudokinase domain were also obtained using the sitting-drop vapor diffusion method, by transferring 130 nl drops of the RYK protein solutions at 11 mg/ml to a 3well crystallization plate (Swissci) mixed with 70 nl precipitant solution (consisting of 100 mM HEPES (pH 7.5), 1.5 M Li2SO4) and incubating at 20˚C. Crystals of the ROR1-ponatinib complex were obtained using the same method and volumes, but using 13 mg/ml ROR1 protein, 0.5 mM ponatinib, and a precipitant solution of 100 mM MES (pH 7.2), 12% (w/v) PEG 20K – and with incubation at 4˚C. Before flash-freezing in liquid nitrogen, PTK7 crystals were cryoprotected in reservoir solution containing 10% (w/v) glycerol, and RYK and ROR1-ponatinib crystals were cryoprotected in reservoir solution containing 25% ethylene glycol. Structure determination For PTK7, diffraction data were collected at beamline 23ID-C of GM/CA at Advanced Photon Source (APS) and were processed using HKL2000. PTK7 crystallized in space group C2221 with three molecules in the asymmetric unit. Structures were solved by molecular replacement with Phaser, using co-ordinates for the MuSK TKD (PDB: 1LUF) as a search model (Till et al., 2002). Cycles of manual building/rebuilding using Coot (Emsley and Cowtan, 2004) were alternated with rounds of refinement using REFMAC (CCP4, 1994), plus composite omit maps calculated with CNS. PROCHECK identified no residues in the disallowed region of the Ramachandran plot. For RYK and ROR1/ponatinib, data were collected at Swiss Light Source (SLS) and analyzed, scaled and merged with Xia2 (Winter et al., 2013). Structures were solved by molecular replacement with Phaser using a ROR2 model (PDB: 4GT4) as a template for ROR1 (Artim et al., 2012) and an AXL model (PDB: 5TC0) as a template for RYK (Keung et al., 2017). The resulting models were refined with REFMAC5 (CCP4, 1994) and validated using MolProbity (Chen et al., 2010). Data collection and refinement statistics are listed in Table 1. Hydrogen-Deuterium Exchange Mass Spectrometry (HDX-MS) To obtain mass spectra of undeuterated and of D2O-labelled peptides, purified protein (IRK978–1283, PTK7774-1069, ROR1457-752, ROR2452–753, or RYK318-607) at 0.4 mg/ml in H2O-based buffer (20 mM HEPES pH 7.4, 200 mM NaCl, 100 µM TCEP) was diluted twenty-fold with either H2O-containing (undeuterated standards) or D2O-containing (labeling conditions) buffer (20 mM HEPES pD 7.4, 200 mM NaCl, 100 µM TCEP using deuterium as solvent) to label amide hydrogen atoms (with final 95% D2O concentration). Labeling was performed at 25˚C for a range of times (10 s, 1 min, 10 min, 1 h, or 2 h), and the reaction was quenched by adding cold (4˚C) 100 mM sodium phosphate buffer at pH 2.4, containing 2 M guanidine hydrochloride and 1% formic acid. For investigating the effect of GZD824 or ponatinib binding on ROR1, ROR1457-752 (10 µM) was incubated with freshly prepared GZD824 or ponatinib at a final inhibitor concentration of 50 μM (and final DMSO concentration of 5% v/v) for 30 min at 25˚C prior to deuterium labeling. The D2O buffer contained 20 mM HEPES (pD 7.4), 200 mM NaCl, 50 μM GZD824 or ponatinib (and a final 5% DMSO), and the labeled sample was quenched by adding cold (4˚C) 200 mM sodium phosphate buffer (pH 2.4) containing 2 M guanidine hydrochloride and 1% formic acid. Quenched samples were immediately injected onto a Waters HDX nanoAcquity UPLC (Waters, Milford, MA) with in-line digestion using an Enzymate BEH pepsin column (Waters, Milford, MA). Peptic fragments were trapped on an 33 Acquity UPLC BEH C18 peptide trap and separated on an Acquity UPLC BEH C18 column. For IRK samples, a CSH C18 peptide trap and column were instead used for a more optimal peptide separation. A 7 min, 5-35% acetonitrile (0.1% Formic acid) gradient was used to elute peptides directly into a Waters Synapt G2-Si mass spectrometer in TOF mode – except for one biological replicate each of ROR1 unliganded, 5% DMSO, and 50 μM GZD824, which were collected using a Waters Xevo G2-XS QTOF mass spectrometer. MSE data were acquired with a 15 to 30 V ramp CID for high energy acquisition of product ions as well as continuous lock mass (Leu-Enk) for mass accuracy correction. Peptide identification and protein sequence coverage maps were obtained from the undeuterated controls. Peptides were identified using the ProteinLynx Global SERVER 3.0.3 (PLGS) from Waters. Fully deuterated controls were performed by incubating in the presence of 2 M Guanidine-DCl for two hours prior to quenching (PTK7) or by incubating in D2O buffer for 2-7 days (ROR1, ROR2, RYK, IRK). All deuterium labeling time points and controls were repeated three times. Data and statistical analyses were carried out as described previously (Houde et al., 2011). Briefly, the deuterium uptake by the identified peptic fragments through increasing deuteration time and for the fully deuterated control was determined using DynamX 3.0 (Waters). The normalized percentage of deuterium uptake at incubation time t (%Dt) at an incubation time t for a given peptide was calculated as follows: %𝐷𝐷𝑡𝑡=100 ∗(𝑚𝑚𝑡𝑡− 𝑚𝑚0) (𝑚𝑚𝑓𝑓− 𝑚𝑚0) With mt being the centroid mass at incubation time t, m0 the centroid mass of the undeuterated control and mf the centroid mass of the fully deuterated control. For studies of changes in ROR1 dynamics upon inhibitor binding, the percent deuteration difference plots Δ%Dexchange (bound-unbound) presented in Fig. 7 were generated by subtracting the corresponding percent deuteration at incubation time t calculated for the DMSO control (unbound) from that calculated with ponatinib or GZD824 (bound) thus: 𝛥𝛥%𝐷𝐷𝑒𝑒𝑒𝑒𝑒𝑒ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑒𝑒 = %𝐷𝐷𝑒𝑒,𝑡𝑡(𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏)−%𝐷𝐷𝑒𝑒,𝑡𝑡(𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏) where x is each peptide, and t is the D2O incubation time. A summary of the HDX-MS data collected and the HDX uptake data are included in Table S1 following consensus guidelines (Masson et al., 2019). Differential scanning fluorimetry (DSF) For thermal shift assays to study ATP binding in Fig. S4B, proteins were diluted in 20 mM HEPES (pH 7.5), 150 mM NaCl, and 100 μM TCEP to a final protein concentration of 5 μM. Nucleotide (5 mM) and/or MgCl2 (10 mM) was added. SYPRO Orange (ThermoFisher Scientific) was then added (diluting 2500-fold into the sample), and 25 µl of the reaction mixtures were transferred to Concord 96-well polycarbonate PCR plates (Bio-Rad) with three technical replicates per condition. A Bio-Rad CFX96 Touch Real Time PCR machine was then used to monitor fluorescence at 530 nm while the temperature was raised by 1˚C per minute from 25˚C to 95˚C, with fluorescence measured at each increment. Fluorescence values were then plotted as a function of temperature after normalizing to the maximum fluorescence signal. Melting temperatures (TM) were determined as the temperature at halfmaximum fluorescence. Plots were generated using GraphPad Prism and represent the mean from three technical replicates. For thermal shift assays used for small molecule screening (Fig. 6) and analysis of ponatinib binding (Fig. 6C), a 2 µM solution of the relevant purified pseudokinase domain in 34 assay buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 0.5 mM TCEP, 5% glycerol) was mixed 1:1000 with SYPRO Orange (Millipore Sigma). Test compounds were added to a final concentration of 10 µM (containing a final 2% DMSO). 20 µl of each sample were then placed in a 96-well plate, and the temperature increased stepwise from 25˚C to 95˚C. Fluorescence was monitored using a Mx3005P real-time PCR instrument (Stratagene) with excitation and emission filters set to 465 and 590 nm, respectively. Data were analyzed using MxPro software. Compound screening was performed as a single shot experiment. Potential pseudokinase binders were confirmed by repeating the experiment with an inhibitor concentration series (final concentrations up to 25 μM) – as seen for GZD824 in Fig. 6C. Plots were generated using GraphPad Prism and represent the mean from three technical replicates. Small molecule screening Small molecule screening was performed using the DSF assay described above. Our screening collection consisted of the kinase inhibitor library (#L1200 Selleckchem) and FDAapproved drug library (#L1300 Selleckchem), which includes a total of 1486 compounds. Failure to get hits with other pseudokinases does not simply reflect protein stability; the most stable Wnt-signaling pseudokinase was PTK7 (TM ~ 59˚C) followed by RYK (54˚C), ROR1 (53˚C) and ROR2 (51˚C). Homology models of EphA10 and EphB6 The crystal structure of EphB3 kinase (PDB: 3ZFY) in the inactive conformation (Overman et al., 2014) was selected as the template for modeling the inactive conformation of the EphA10 and EphB6 pseudokinase domains. After running BLAST for sequence alignment, the top crystal structure matches with EphA10 and EphB6 were found to be EphA4 (PDB: 2YGM) and EphB1 (PDB: 3ZFX) respectively, which were then used as templates for modeling active conformations of EphA10 and EphB6 pseudokinase domains. Coordinates of the active and inactive template kinases were downloaded from the Protein Data Bank, and missing residues were added by generating a homology model of the kinase using its own partial structure as the template. Five candidate models were generated from each of the templates by satisfying a set of static and dynamic spatial restraints in MODELLER. These restraints are expressed in terms of molecular probability density function, or objective function, which is optimized and applied in the ranking of the set of models constructed in MODELLER. The stereochemical quality of each model was further evaluated using the Discrete Optimized Protein Energy (DOPE) method – an atomic distance-dependent statistical potential optimized for model assessment in MODELLER, with the lowest DOPE score defining the preferred model. We structurally aligned all of the homology models for EphA10 and EphB6 to their template structures using Visual Molecular Dynamics (VMD), and generated Root Mean Square Deviation (RMSD) plots to find regions of variability. Most variation was seen in the first 50 residues for the homology models of the inactive structures of EphA10 and EphB6, so a hybrid approach was used to improve models for the inactive EphA10 and EphB6 conformations, where the coordinates for the first 50 residues came from the homology model for the active conformation of the kinase, and the rest of the coordinates come from the homology model for the inactive conformation. VMD was used to display structural models, and MODELLER was used for homology modeling of the pseudokinase domains for both EphA10 and EphB6. BaF3 cell transfections For expression in BaF3 cells, human ROR1, ROR2, PTK7 and RYK (or their truncated variants) were cloned into the pEFIRES-P vector (Hobbs et al., 1998) and electroporated into BaF3 cells using the 4D-Nucleofector X Kit L (Lonza) and a 4D Nucleofector X (Lonza), using program 35 code DS-137. After 72 h, transfected cells were selected with increasing amounts of puromycin (0.5-1 µg/ml) for 2 weeks and tested for protein expression by Western blot (Figs. 5 and S5). Cell viability and cell proliferation Cell counts were determined by Trypan-blue exclusion using a Countess II automated cell counter (Life Technologies). Cell viability was measured using the CellTiter-Glow 2.0 Assay (CTG, Promega, Madison, USA) according to the manufacturer’s instructions. Western blotting Whole cell lysates were prepared by lysing cells in ice cold NP-40 lysis buffer (50 mM Tris-HCl pH 7.4, 10% glycerol, 50 mM NaCl, 0.5% sodium deoxycholate, 1% NP-40, 20 mM NaF), supplemented with protease and phosphatase inhibitor cocktails (Bimake, Houston, TX, USA). Cell lysates were incubated 15 min on ice, clarified by centrifugation (4˚C, 20 min, 20,000 x g), resuspended in 2 x SDS sample buffer and boiled at 95˚C for 5 min. Samples were separated by SDS-PAGE and transferred to nitrocellulose membrane. Blots were blocked (4% BSA in 0.05% Tween 20 in 1xTBS) at RT for 1 h and incubated with primary antibody at 4˚C overnight. Blots were then washed 3 times with TBS/0.1% Tween 20 buffer, and subjected to secondary antibodies for 1 h at RT. Blots were scanned with an Odyssey® CLx Imaging System (LI-COR) and images analyzed using Image Studio Lite (LI-COR). For Western blotting, the following antibodies were used: pAKT (S473, #6942), AKT (#9272), pERK1/2 (#9101), ERK1/2 (#4696), PTK7 (#25618), and SRC (#2109) from Cell Signaling Technology (Danvers, MA, USA); ROR1 4A5 (#564464) and ROR2 (#565550) from BD Biosciences (San Jose, CA, USA); anti-pTYR 4G10 (#05-321) from MerckMillipore; β-tubulin (#sc-166729) from Santa Cruz Biotechnology (Dallas, TX, USA); HA (#901513) from BioLegend (San Diego, CA). As secondary antibodies, IRDye® 800CW Donkey anti-Mouse IgG or IRDye® 680RD Donkey anti-Rabbit IgG (LI-COR, Lincoln, NE, USA) were used at 1:10.000 dilution. Cellular thermal shift assay (CETSA) CETSA was carried out according to the protocol described (Martinez Molina et al., 2013). Briefly, BaF3-ROR1 cells were treated with control (DMSO) or drug (10 µM) and incubated at 37˚C for 2 h. Cells were washed once and divided into aliquots with 0.5 x 106 cells each. Samples were heated (Bio-Rad T100TM Thermal Cycler) pairwise (control and drug-treated samples) for 3 min at 42-60˚C with 2˚C increments between pairs. After heating, cells were kept at room temperature for 3 min before transferring to ice. Cells were collected and lysed in Triton X100-containing lysis buffer (50 mM Tris-HCl, pH 7.5, 10% glycerol, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 50 mM NaF) supplemented with protease and phosphatase inhibitor cocktails (Bimake, Houston, TX, USA), and were subsequently analyzed by SDS-PAGE and Western blotting using anti-HA (ROR1) and β-tubulin antibodies. Protein levels were quantified with Image Studio Lite (Li-COR) and normalized to 42˚C samples for both treatments. in vitro kinase assays Kinase assays were performed using either ATP hydrolysis as a readout (ADP-Glo) or fluorescence detection of phosphorylation of a peptide. For the ADP-Glo assay (Promega), a kit was used as recommended by the manufacturer. Briefly, reactions with 1 µM purified ROR1 pseudokinase domain or kinase controls (at 1 µM) were set up in kinase assay buffer (20 mM HEPES, pH 7.4, containing 10 mM MgCl2, 0.02 mg/ml BSA, 0.1 mM Na3VO4, 2 mM DTT) with addition of 1 mM ATP and 0.1 mg/ml poly(Glu, Tyr) peptide substrate (Sigma-Aldrich). Reactions were incubated for 1 h at ambient temperature, and were then quenched by addition of an equal volume of ADP-Glo reagent – unreacted ATP was depleted over the course of 40 minutes at room temperature. An equal volume of detection reagent was then added to convert 36 ADP in the reaction mixture into newly synthesized ATP, which serves as a substrate for the luciferase reaction. Samples were incubated for 1 h at room temperature before luminescence measurements were taken in a 384-well plate using a BioTek Synergy 2 Multi-Detection Microplate Reader. Experiments were performed in technical triplicate. For fluorescent peptide assays, materials from Assay Quant Technologies were used as recommended by the manufacturer. Briefly, reactions with 1 µM purified ROR1 pseudokinase domain or kinase controls (at 1 µM) were set up in wells of a 384-well plate with the final assay conditions: 50 mM HEPES, pH 7.5, containing 10 mM MgCl2, 0.01% Brij-35, 0.1 mg/ml BSA, 1 mM ATP, and 20 µM peptide substrate sensor (AQT0001) that contains an incorporated Sox chromophore (Assay Quant). Reactions were incubated at 30°C for 2 h, and fluorescence intensities were collected every 2 min using a BioTek Synergy 2 Multi-Detection Microplate Reader. Experiments were performed in technical triplicate. QUANTIFICATION AND STATISTICAL ANALYSIS Structure determination and analysis The statistical analysis of the structural models is provided in Table 1. Analysis of the molecular contacts and RMSD values were calculated using the CCP4 software package (CCP4, 1994). Analysis of HDX dynamics Raw MSE mass spectra of undeuterated controls were used for peptide identification using ProteinLynx Global Server 3.0.3 (PLGS) from Waters. MS data from all controls and labeling conditions were then processed using DynamX to identify peptic peptides for each condition. All raw spectra for each peptide, labeling condition, drug condition, and charge state were then manually assessed for quality and for accurate peak assignment, at which point poor quality or incorrectly assigned peaks were unassigned. Average mass shifts of centroids and their standard deviations were then used to calculate percent uptake for each time point relative to a fully deuterated standard as described in Method Details. Uptake plots in Figs. 4A,C represent standard deviations from three independent labeling experiments for all conditions shown. Uptake plots in Fig. 7B represent mean ± SD from three independent protein preparations, for each of which three independent labelling experiments were performed in all conditions. For ‘exchangeability indices’ (Fig. 4D), an R script was written to first calculate an average percent exchange for each residue in each pseudokinase (and IRK) at each time point. The value assigned to each residue was the mean (per residue) percent exchange seen for all peptides containing that residue. In parallel, each residue in each pseudokinase was assigned a corresponding homologous residue in IRK (based on sequence alignments). For the position corresponding to each residue in IRK, the mean (± SD) percent exchange was calculated across the 4 pseudokinases (ROR1, ROR2, RYK, PTK7) for each time point. This pseudokinase ‘exchangeability index’ value was then compared with the exchange value assigned to the corresponding residue in IRK – with deviations reported in terms of number of standard deviations (σ) away from the average pseudokinase uptake. Differences between pseudokinase exchangeability index and IRK exchange for each residue were averaged across all time points to depict areas that show more (orange) or less (black) HDX in pseudokinases than in IRK, as shown in Fig. 4D. Western blot image processing Raw images from the LI-COR Odyssey were imported in Adobe Photoshop, and linear contrast stretching was manually applied using the ‘Levels’ function – so that the darkest 37 points of the images are black, and background is brought into the visible grey scale so that all features are registered. Cell proliferation assays All cell proliferation assays experiments were analyzed using Prism v8.0 software (GraphPad, San Diego, CA, USA). Cell proliferation data are represented as mean ± SD from four independent experiments each performed in technical triplicate. Statistical significance was determined using two-tailed t-tests. Two-sided p values were used to determine statistical significance, set as follows: *p ≤ 0.05, **p ≤ 0.01 and ***p ≤ 0.001. Cellular thermal shift assays All cellular thermal shift experiments were analyzed using Prism v8.0 software (GraphPad, San Diego, CA, USA). Individual bands were quantified using Image Studio Lite software and all values (band intensities) were normalized to unheated control sample (no inhibitor treatment) that denotes relative band intensity value of one. Standard deviation (SD) was calculated using three independent experiments. Curve fitting was done in Prism using Boltzmann sigmoidal algorithm. For concentration-dependent CETSA in Fig. S6, a constant heating temperature of 48˚C was chosen based on the results from Fig. 6E. All values (band intensities) were normalized to heated samples without inhibitor. Curve fitting was done in Prism using one site (specific binding) algorithm. 38 SUPPLEMENTAL EXCEL TABLE TITLES TABLE S1 – Related to Figures 4 and 7. HDX-MS summary and uptake data. Experimental details are listed for all HDX-MS experiments performed in this study, and the uptake data for each peptide reported are included. TABLE S2 – Related to Figure 6. Compound list and results for screening to identify pseudokinase-binding molecules. A library of 1486 small molecules consisting largely of protein kinase inhibitors or FDA approved drugs and their respective effects on the differential scanning fluorometry-based Tm shift values of PTK7, ROR1, ROR2, and RYK. 39 REFERENCES Ablooglu, A.J., Frankel, M., Rusinova, E., Ross, J.B., and Kohanski, R.A. (2001). Multiple activation loop conformations and their regulatory properties in the insulin receptor's kinase domain. J. Biol. Chem. 276, 46933-46940. Adams, P.D., Afonine, P.V., Bunkoczi, G., Chen, V.B., Davis, I.W., Echols, N., Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). PHENIX: a comprehensive Pythonbased system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213221. Akbarzadeh, S., Wheldon, L.M., Sweet, S.M., Talma, S., Mardakheh, F.K., and Heath, J.K. (2008). The deleted in brachydactyly B domain of ROR2 is required for receptor activation by recruitment of Src. PLoS One 3, e1873. Amin, D.N., Campbell, M.R., and Moasser, M.M. (2010). The role of HER3, the unpretentious member of the HER family, in cancer biology and cancer therapeutics. Semin. Cell Dev. Biol. 21, 944-950. Artim, S.C., Mendrola, J.M., and Lemmon, M.A. (2012). Assessing the range of kinase autoinhibition mechanisms in the insulin receptor family. Biochem. J. 448, 213-220. Bailey, F.P., Byrne, D.P., McSkimming, D., Kannan, N., and Eyers, P.A. (2015). Going for broke: targeting the human cancer pseudokinome. Biochem. J. 465, 195-211. Becher, I., Savitski, M.M., Savitski, M.F., Hopf, C., Bantscheff, M., and Drewes, G. (2013). Affinity profiling of the cellular kinome for the nucleotide cofactors ATP, ADP, and GTP. ACS Chem. Biol. 8, 599-607. Bicocca, V.T., Chang, B.H., Masouleh, B.K., Muschen, M., Loriaux, M.M., Druker, B.J., and Tyner, J.W. (2012). Crosstalk between ROR1 and the Pre-B cell receptor promotes survival of t(1;19) acute lymphoblastic leukemia. Cancer Cell 22, 656-667. Boudeau, J., Miranda-Saavedra, D., Barton, G.J., and Alessi, D.R. (2006). Emerging roles of pseudokinases. Trends Cell Biol. 16, 443-452. Bunney, T.D., Inglis, A.J., Sanfelice, D., Farrell, B., Kerr, C.J., Thompson, G.S., Masson, G.R., Thiyagarajan, N., Svergun, D.I., Williams, R.L., et al. (2018). Disease variants of FGFR3 reveal molecular basis for the recognition and additional roles for Cdc37 in Hsp90 chaperone system. Structure 26, 446-458. Burslem, G.M., Smith, B.E., Lai, A.C., Jaime-Figueroa, S., McQuaid, D.C., Bondeson, D.P., Toure, M., Dong, H., Qian, Y., Wang, J., et al. (2018). The advantages of targeted protein degradation over inhibition: An RTK case study. Cell Chem. Biol. 25, 67-77. CCP4 (1994). The CCP4 suite: Programs for protein crystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763. Chen, V.B., Arendall, W.B., 3rd, Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010). MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12-21. Choi, M.Y., Widhopf, G.F., 2nd,, Ghia, E.M., Kidwell, R.L., Hasan, M.K., Yu, J., Rassenti, L.Z., Chen, L., Chen, Y., Pittman, E., et al. (2018). Phase I trial: Cirmtuzumab inhibits ROR1 signaling and stemness signatures in patients with chronic lymphocytic leukemia. Cell Stem Cell 22, 951959. Choi, M.Y., Widhopf, G.F., 2nd,, Wu, C.C., Cui, B., Lao, F., Sadarangani, A., Cavagnaro, J., Prussak, C., Carson, D.A., Jamieson, C., et al. (2015). Pre-clinical specificity and safety of UC- 40 961, a first-in-class monoclonal antibody targeting ROR1. Clin. Lymph. Myeloma Leuk. 15, S167S169. Collins, M.K., Downward, J., Miyajima, A., Maruyama, K., Arai, K., and Mulligan, R.C. (1988). Transfer of functional EGF receptors to an IL3-dependent cell line. J. Cell. Physiol. 137, 293-298. Damelin, M., Bankovich, A., Bernstein, J., Lucas, J., Chen, L., Williams, S., Park, A., Aguilar, J., Ernstoff, E., Charati, M., et al. (2017). A PTK7-targeted antibody-drug conjugate reduces tumorinitiating cells and induces sustained tumor regressions c. Sci. Transl. Med. 9, eaag2611. Daneshmanesh, A.H., Porwit, A., Hojjat-Farsangi, M., Jeddi-Tehrani, M., Tamm, K.P., Grandér, D., Lehmann, S., Norin, S., Shokri, F., Rabbani, H., et al. (2013). Orphan receptor tyrosine kinases ROR1 and ROR2 in hematological malignancies. Leuk. Lymphoma 54, 843-850. Debebe, Z., and Rathmell, W.K. (2015). Ror2 as a therapeutic target in cancer. Pharmacol. Ther. 150, 143-148. Dhawan, N.S., Scopton, A.P., and Dar, A.C. (2016). Small molecule stabilization of the KSR inactive state antagonizes oncogenic Ras signalling. Nature 537, 112-116. Dong, J., Zhao, H., Zhou, T., Spiliotopoulos, D., Rajendran, C., Li, X.D., Huang, D., and Caflisch, A. (2014). Structural analysis of the binding of type I, I1/2, and II inhibitors to Eph tyrosine kinases. ACS Med. Chem. Lett. 6, 79-83. Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126-2132. Fedorov, O., Niesen, F.H., and Knapp, S. (2012). Kinase inhibitor selectivity profiling using differential scanning fluorimetry. Methods Mol. Biol. 795, 109-118. Gentile, A., Lazzari, L., Benvenuti, S., Trusolino, L., and Comoglio, P.M. (2011). Ror1 is a pseudokinase that is crucial for Met-driven tumorigenesis. Cancer Res. 71, 3132-3141. Gentile, A., Lazzari, L., Benvenuti, S., Trusolino, L., and Comoglio, P.M. (2014). The ROR1 pseudokinase diversifies signaling outputs in MET-addicted cancer cells. Int. J. Cancer 135, 23052316. Girdlestone, C., and Hayward, S. (2016). The DynDom3D Webserver for the analysis of domain movements in multimeric proteins. J. Comput. Biol. 23, 21-26. Green, J., Nusse, R., and van Amerongen, R. (2014). The role of Ryk and Ror receptor tyrosine kinases in Wnt signal transduction. Cold Spring Harb. Perspect. Biol. 6, a009175. Gustafson, W.C., Meyerowitz, J.G., Nekritz, E.A., Chen, J., Benes, C., Charron, E., Simonds, E.F., Seeger, R., Matthay, K.K., Hertz, N.T., et al. (2014). Drugging MYCN through an allosteric transition in Aurora kinase A. Cancer Cell 26, 414-427. Hammaren, H.M., Ungureanu, D., Grisouard, J., Skoda, R.C., Hubbard, S.R., and Silvennoinen, O. (2015). ATP binding to the pseudokinase domain of JAK2 is critical for pathogenic activation. Proc. Natl. Acad. Sci. U. S. A. 112, 4642-4647. Hayes, M., Naito, M., Daulat, A., Angers, S., and Ciruna, B. (2013). Ptk7 promotes non-canonical Wnt/PCP-mediated morphogenesis and inhibits Wnt/β-catenin-dependent cell fate decisions during vertebrate development. Development 140, 1807-1818. Hobbs, S., Jitrapakdee, S., and Wallace, J.C. (1998). Development of a bicistronic vector driven by the human polypeptide chain elongation factor 1alpha promoter for creation of stable mammalian cell lines that express very high levels of recombinant proteins. Biochem. Biophys. Res. Commun. 252, 368-372. 41 Hornbeck, P.V., Zhang, B., Murray, B., Kornhauser, J.M., Latham, V., and Skrzypek, E. (2015). PhosphoSitePlus, 2014: mutations, PTMs and recalibrations. Nucleic Acids Res. 43(Database issue), D512-D520. Houde, D., Berkowitz, S.A., and Engen, J.R. (2011). The utility of hydrogen/deuterium exchange mass spectrometry in biopharmaceutical comparability studies. J. Pharm. Sci. 100, 2071-2086. Hsu, Y.H., Johnson, D.A., and Traugh, J.A. (2008). Analysis of conformational changes during activation of protein kinase Pak2 by amide hydrogen/deuterium exchange. J. Biol. Chem. 283, 36397-36405. Hubbard, S.R. (2013). The insulin receptor: both a prototypical and atypical receptor tyrosine kinase. Cold Spring Harb. Perspect. Biol. 5, a008946. Hubbard, S.R., Wei, L., Ellis, L., and Hendrickson, W.A. (1994). Crystal structure of the tyrosine kinase domain of the human insulin receptor. Nature 372, 746-754. Hunter, J.D. (2007). Matplotlib: A 2D graphics environment. Computing in Science and Engineering 9, 90-95. Huse, M., and Kuriyan, J. (2002). The conformational plasticity of protein kinases. Cell 109, 275282. Iacob, R.E., Pene-Dumitrescu, T., Zhang, J., Gray, N.S., Smithgall, T.E., and Engen, J.R. (2009). Conformational disturbance in Abl kinase upon mutation and deregulation. Proc. Natl. Acad. Sci. U. S. A. 106, 1386-1391. Inoue, T., Oz, H.S., Wiland, D., Gharib, S., Deshpande, R., Hill, R.J., Katz, W.S., and Sternberg, P.W. (2004). C. elegans LIN-18 is a Ryk ortholog and functions in parallel to LIN-17/Frizzled in Wnt signaling. Cell 118, 795-806. Jacob, W., James, I., Hasmann, M., and Weisser, M. (2018). Clinical development of HER3targeting monoclonal antibodies: Perils and progress. Cancer Treat. Rev. 68, 111-123. Jacobsen, A.V., and Murphy, J.M. (2017). The secret life of kinases: insights into non-catalytic signalling functions from pseudokinases. Biochem. Soc. Trans. 45, 665-681. Jaiswal, B.S., Kljavin, N.M., Stawiski, E.W., Chan, E., Parikh, C., Durinck, S., Chaudhuri, S., Pujara, K., Guillory, J., Edgar, K.A., et al. (2013). Oncogenic ERBB3 mutations in human cancers. Cancer Cell 23, 603-617. Jura, N., Shan, Y., Cao, X., Shaw, D.E., and Kuriyan, J. (2009). Structural analysis of the catalytically inactive kinase domain of the human EGF receptor 3. Proc. Natl. Acad. Sci. U. S. A. 106, 21608-21613. Jura, N., Zhang, X., Endres, N.F., Seeliger, M.A., Schindler, T., and Kuriyan, J. (2011). Catalytic control in the EGF receptor and its connection to general kinase regulatory mechanisms. Mol. Cell 42, 9-22. Karvonen, H., Barker, H., Kaleva, L., Niininen, W., and Ungureanu, D. (2019). Molecular mechanisms associated with ROR1-mediated drug resistance: Crosstalk with Hippo-YAP/TAZ and BMI-1 pathways. Cells 8, E812. Karvonen, H., Perttilä, R., Niininen, W., Barker, H., and Ungureanu, D. (2018). Targeting Wnt signaling pseudokinases in hematological cancers. Eur. J. Haematol. 101, 457-465. Katoh, M. (2017). Canonical and non-canonical WNT signaling in cancer stem cells and their niches: Cellular heterogeneity, omics reprogramming, targeted therapy and tumor plasticity Int. J. Oncol. 51, 1357-1369. EphA10 Y801 R774 H775 H-R-G B C EphA10 EphB6IRK Y801 Y1162 A Y725 F789 G788 G790 Y723 GFG-‘out’ R813 L814 G815 F750 T748 RLG-‘in’ EphA10 EphB6 IRK Y1162 H-R-D D1132 N1137 R1136 980 1280 αHαGαFαEFαEα D αCαIβ7β5β4β3β2β1 IRK# β5/αD Hinge β8β10β11 HRD A-loop Peptides Recovered 100 80 60 20 0 40 10 sec 1 min 10 min 1 hr 2 hrs ROR1 ROR2 PTK7 RYK 0 50 100 % Exchange 0 50 100 % Exchange 0 50 100 % Exchange 980 1080 1180 1280 0 50 100 Residue Number % Exchange 980 1080 1180 1280 0 50 100 Residue Number % Exchange Exchange time (IRK) IRK A C 10 sec 1 min 10 min 2 hr B -6σ6σ more exchange in pseudokinase -2σ2σ less exchange in pseudokinase YxxxYY % Exchange at 1 min D β1 β2 β3 β4 β5 αD αC αE αF αG αH αI β1 β2 β3 β4 β5 αD αC αE αF αG αH αI β5/αD Hinge HRD A-loop A C ROR1 BaF3-ROR1 β-tubulin BaF3 ROR1ΔICR ROR1ΔICR BaF3 pAKT pERK Wnt5a (hrs) BaF3-ROR1 - AKT ERK β-tubulin BaF3-ROR1 ΔICR ROR1 ROR1ΔICR SRC pSRC 0.50.5 0.52 2 2 44 4 - - Relative proliferation BaF3 BaF3-ROR1 BaF3-ROR1ΔICR *** *** D Parental ROR1 FxxxFF K506A 0 5 10 15 20 Relative proliferation HA BaF3 ROR1 FxxxFF BaF3 ROR1 K506A BaF3 ROR1 01234 0 5 10 15 Time (days) *** *** β-tubulin E B BA C Ponatinib Ror1 ΔTM = 4.1˚C GZD824 Ror1 ΔTM = 6.0˚C N N HN O HN F F F N N O HN F F F N N N NN ROR2PTK7 6 6 6 6 4 4 4 4 0 2 2 2 2 0 0 0 -2 -2 -2-2 TM shift (˚C) TM shift (˚C)TM shift (˚C) TM shift (˚C) 1500 compounds1500 compounds 1500 compounds1500 compounds ROR1 RYK GZD824 Ponatinib HA-ROR1 β-tubulin Inhibitor Control Ponatinib (10μM) GZD824 (10μM) 42 44 46 48 50 52 54 56 58 60˚C - + - + - + - + - + - + - + - + - + - + 42 44 46 48 50 52 54 56 58 60˚C - + - + - + - + - + - + - + - + - + - + D Temp HA-ROR1 β-tubulin Inhibitor Temp GZD824 Ponatinib E 10-8 0 2 4 6 [GZD824] (M) ROR1 TM shift (˚C) 10-7 10-6 10-5 10-4 40 42 44 46 48 50 52 54 56 58 60 0.0 0.2 0.4 0.6 0.8 1.0 Temperature (˚C) Relative band intensity A -10 -5 +5 +10 Δ Exchange at 1 min(%) (GZD824 - DMSO) Protection upon binding Increased HX upon binding β5/αD Hinge HRD A-loop 450 750 αHαGαFαEFαEα D αCαIβ7β5β4β3β2β1 ROR1# β8β10β11 B i ii iii iv 476-481 MEELGE 521-526 QQEASL 552-560 FEYINQGDL 640-645 IYSADY iii iii iv Unliganded GZD824 Ponatinib αCHinge β1/β2 A-loop C D 14° Ponatinib E523 L526 F611 L530 K506 Y554 L622 S632 H613 L606 V536 I487 F552 D633 M527 β1 β2β3β4 β5 αC αC β1 αD αE αF αG αH αI β7 β8 αE β7β8 β5 β3 Ponatinib L526 E523 L530 M527 Y554 Y555 C-lobe ROR2 ROR1 ~20˚ αC αC DLG N-lobe: 1234 0 50 100 Log Time (sec) % Exchange 1234 0 50 100 Log Time (sec) % Exchange 1234 0 50 100 Log Time (sec) % Exchange 1234 0 50 100 Log Time (sec) % Exchange ROR1 Y554 displacement 1 Figure S1 – Related to Figure 1. Amino acid sequence alignment of RTK pseudokinase domains alongside IRK, EGFR, and protein kinase A (PKA) Highly conserved residues across all kinases/pseudokinases are boxed in grey; sequence variations in pseudokinases for key conserved residues are in bold red text. Secondary structure elements are noted, as are the catalytic and activation loop regions as well as the G-rich loop, HRD, and DFG motifs. Putative regulatory activation-loop tyrosine residues (in the YxxxYY motif) are in bold blue text, and the gatekeeper residue is in orange text. IRK ITLLRELGQGSFGMVYEGNA-RDIIK-GEAETRVAVKTVNESAS--LRERIEFLNEASVMKGFTCH-HVVR 1061 PTK7 LQPITTLGKSEFGEVFLAKA-QGL-EEGVAETLVLVKSLQSKDE---QQQLDFRRELEMFGKL-NHANVVR 860 ROR1 VRFMEELGECAFGKIYKGHL-YLPG-MDHA-QLVAIKTLKDYNN--PQQWTEFQQEASLMAEL-HHPNIVC 537 ROR2 VRFMEELGEDRFGKVYKGHL-FGPAPGEQT-QAVAIKTLKDKAE--GPLREEFRHEAMLRARL-QHPNVVC 538 RYK ITLKDVLQEGTFGRIFHGIL-IDEKDPNK-EKQAFVKTVKDQAS--EIQVTMMLTESCKLRGL-HHRNLLP 395 EphA10 VTLERSLGGGRFGELCCGCL-QLP-GRQEL--LVAVHMLRDSAS--DSQRLGFLAEALTLGQF-DHSHIVR 708 EphB6 IKIEEVIGTGSFGEVRQGRL-QPR-GRREQ--TVAIQALWAGGA--ESLQMTFLGRAAVLGQF-QHPNILR 733 ErbB3 LRKLKVLGSGVFGTVHKGVW-IPEGESIKI--PVCIKVIEDKSG--RQSFQAVTDHMLAIGSL-DHAHIVR 754 EGFR FKKIKVLGSGAFGTVYKGLW-IPEGEKVKI--PVAIKELREATS--PKANKEILDEAYVMASV-DNPHVCR 752 PKA FDRIKTLGTGSFGRVMLVKH-KESG------NHYAMKILDKQKVVKLKQIEHTLNEKRILQAV-NFPFLVK 105 IRK LLGV-VSKGQPTLVVMELMAHGDLKSYLRSLRP------EAENN-PGRPPPTLQEMIQMAAEIADGMAYLN 1124 PTK7 LLGL-CREAEPHYMVLEYVDLGDLKQFLRISK-------SKDEK-LKSQPLSTKQKVALCTQVALGMEHLS 922 ROR1 LLGA-VTQEQPVCMLFEYINQGDLHEFLIMRSPHSDVGCSSDEDGTVKSSLDHGDFLHIAIQIAAGMEYLS 607 ROR2 LLGV-VTKDQPLSMIFSYCSHGDLHEFLVMRSPHSDVG-STDDDRTVKSALEPPDFVHLVAQIAAGMEYLS 607 RYK ITHVCIEEGEKPMVILPYMNWGNLKLFLRQCK-----LVEANNP----QAISQQDLVHMAIQIACGMSYLA 457 EphA10 LEGV-VTRGSTLMIVTEYMSHGALDGFLRRHEG----------------QLVAGQLMGLLPGLASAMKYLS 762 EphB6 LEGV-VTKSRPLMVLTEFMELGPLDSFLRQREG----------------QFSSLQLVAMQRGVAAAMQYLS 787 ErbB3 LLG--LCPGSSLQLVTQYLPLGSLLDHVRQHRG----------------ALGPQLLLNWGVQIAKGMYYLE 807 EGFR LLGICLT--STVQLITQLMPFGCLLDYVREHKD----------------NIGSQYLLNWCVQIAKGMNYLE 805 PKA LEFS-FKDNSNLYMVMEYVAGGEMFSHLRRIG-----------------RFSEPHARFYAAQIVLTFEYLH 158 IRK AKKFVHRDLAARNCMVAHDFTVKIG GDF MTRDIYETDYYRKGGKGLLPVRWMAPESLKDGVFTTSSDMWSF 1195 PTK7 NNRFVHKDLAARNCLVSAQRQVKVSALGLSKDVYNSEYYHFRQ-AWVPLRWMSPEAILEGDFSTKSDVWAF 992 ROR1 SHFFVHKDLAARNILIGEQLHVKISDLGLSREIYSADYYRVQSKSLLPIRWMPPEAIMYGKFSSDSDIWSF 678 ROR2 SHHVVHKDLATRNVLVYDKLNVKISDLGLFREVYAADYYKLLGNSLLPIRWMAPEAIMYGKFSIDSDIWSY 678 RYK RREVIHKDLAARNCVIDDTLQVKITDNALSRDLFPMDYHCLGDNENRPVRWMALESLVNNEFSSASDVWAF 528 EphA10 EMGYVHRGLAARHVLVSSDLVCKISGFGRGPRDRSEAVYTTMS-GRSPALWAAPETLQFGHFSSASDVWSF 832 EphB6 SFAFVHRSLSAHSVLVNSHLVCKVARLGHSPQGPSC-----------LLRWAAPEVIAHGKHTTSSDVWSF 847 ErbB3 EHGMVHRNLAARNVLLKSPSQVQVADFGVADLLPPDDKQLLYSEAKTPIKWMALESIHFGKYTHQSDVWSY 878 EGFR DRRLVHRDLAARNVLVKTPQHVKITDFGLAKLLGAEEKEYHAEGGKVPIKWMALESILHRIYTHQSDVWSY 876 PKA SLDLIYRDLKPENLLIDQQGYIQVTDFGFAKRVKGRTWTLCGTP-----EYLAPEIILSKGYNKAVDWWAL 224 IRK GVVLWEITSLAEQPYQGLSNEQVLKFVMDGG-YLDQPDNCPERVTDLMRMCWQFNPKMRPTFLEIVNLL 1263 PTK7 GVLMWEVFTHGEMPHGGQADDEVLADLQAGKARLPQPEGCPSKLYRLMQRCWALSPKDRPSFSEIASAL 1061 ROR1 GVVLWEIFSFGLQPYYGFSNQEVIEMVRKRQ-LLPCSEDCPPRMYSLMTECWNEIPSRRPRFKDIHVRL 746 ROR2 GVVLWEVFSYGLQPYCGYSNQDVVEMIRNRQ-VLPCPDDCPAWVYALMIECWNEFPSRRPRFKDIHSRL 746 RYK GVTLWELMTLGQTPYVDIDPFEMAAYLKDGY-RIAQPINCPDELFAVMACCWALDPEERPKFQQLVQCL 596 EphA10 GIIMWEVMAFGERPYWDMSGQDVIKAVEDGF-RLPPPRNCPNLLHRLMLDCWQKDPGERPRFSQIHSIL 900 EphB6 GILMWEVMSYGERPYWDMSEQEVLNAIEQEF-RLPPPPGCPPGLHLLMLDTWQKDRARRPHFDQLVAAF 915 ErbB3 GVTVWELMTFGAEPYAGLRLAEVPDLLEKGE-RLAQPQICTIDVYMVMVKCWMIDENIRPTFKELANEF 946 EGFR GVTVWELMTFGSKPYDGIPASEISSILEKGE-RLPQPPICTIDVYMIMVKCWMIDADSRPKFRELIIEF 944 PKA GVLIYE-MAAGYPPFFADQPIQIYEKIVSGK--VRFPSHFSSDLKDLLRNLLQVDLTKRFGNLKNGVND 290 β1 β2 β3 αB αC β4 β5 αD αE β6 β7 β8 β9 β10 αEF αFβ11 αGαF αH αI YxxxYY motif G-rich loop gate keeper HRD motif DFG motif Catalytic loop Activation loop Figure S1 2 Figure S2 – Related to Figure 2. Comparison of hydrophobic spines in the RTK pseudokinases Top row: Cartoon representation of the catalytically active kinases PKA (white; PDB: 1ATP) and IRK (black) in its active (PDB: 1IR3) and inactive (PDB: 1IRK) conformations. Residues comprising the catalytic spine (C-spine) and the regulatory spine (R-spine) for each domain are shown in surface representation and colored yellow or red, respectively. Both are complete in the active kinases – with the C-spine completed by the adenine ring of ATP. In inactive IRK, F1151 of the DFG motif instead occupies the adenine ring position in the C-spine. Upon activation loop phosphorylation and ligation of the ATP-binding site, F1151 flips from this position towards the aC helix to complete the R-spine. The ‘gatekeeper’ residue (homologous to threonine 315 in ABL) is shown for each domain in sphere representation, and ATP or AMP-PNP is also shown where bound. 3 Middle row: Pseudokinase domains from PTK7 (slate blue), ROR2 (magenta; PDB: 4GT4), and RYK (green). The C-spine is completed in each case by aromatic side-chains that take the place of an ATP ligand. In PTK7 and ROR2, the leucine that replaces the DFG phenylalanine (L949 in PTK7’s ‘ALG’ motif, and L634 in ROR2’s ‘DLG’ motif) abuts the C-spine, but does not contribute to its completion like F1151 does in inactive IRK. Although these residues do not contribute to C-spine completion, they also do not contribute to the R-spine – which is broken in both cases. In RYK, the C-spine is complete, and the R-spine is broken because it lacks an equivalent hydrophobic residue to the DFG phenylalanine (replaced by N of DNA, which is in displaced from the spine. Bottom: The Cand R-spines of ErbB3 (orange; PDB: 4RIW) are shown, both complete, with an arrangement that resembles active kinases. 4 Figure S3 – Related to Figure 3. Rotation of the N-lobe and position of the a C helix in the RTK pseudokinases The N-lobes of the pseudokinase domains from (A): PTK7 (slate blue), (B): ROR2 (magenta), (C): RYK (green), and (D): active IRK (black; PBD: 1IR3) were individually overlaid on inactive IRK (white; PDB: 1IRK) based on superimpositions of the kinase C-lobes – to inspect differences in N-lobe orientation as a result of rotation about the b5/aD hinge. The C-lobe of only inactive IRK is represented (for clarity). Helix aC is colored red in each domain. In (A), the aC helix is clearly ‘out’ for inactive IRK and ‘in’ for PTK7. In addition, the entire N-lobe has undergone a rotation about the b5/aD hinge (marked in red) that also causes Y877 (marked) to project into the ATP-binding site (which it occludes) – whereas the corresponding IRK residue (L1078) projects away from the ATP-binding site. In (B), aC is ‘out’ for both ROR2 and inactive IRK, but ROR2 also displays the N-lobe rotation about the b5/aD hinge seen for PTK7 – in this case causing the Y555 side-chain of ROR2 to occlude the ATP-binding site. In (C), RYK displays an aC ‘in’ conformation (like PTK7), but its N-lobe has not undergone rotation about the b5/aD hinge, so the Y413 side-chain aligns with IRK L1078 and projects away from the ATP-binding site. In (D), active and inactive IRK are overlaid – with no N-lobe rotation, but a clear aC displacement. 5 Figure S4 – Related to Figure 4. Occlusion of the nucleotide-binding site prevents ATP-binding in Wnt-binding pseudokinases (A) Detailed structural view of the occluded ATP-binding sites of PTK7 (slate blue), ROR2 (magenta), RYK (green), and inactive IRK (black; PBD: 1IRK). Residues contributing to steric occlusion of the nucleotide-binding site are labelled and shown in both stick representation with a transparent surface and mesh for emphasis, with AMP-PNP placed based on superimposition with active IRK (PDB: 1IR3) – to highlight the degree of occlusion. (B) DSF analysis of the melting temperatures of IRK and the PTK7, ROR2, RYK, and ROR1 pseudokinases in the presence and absence of 5 mM ATP and/or MgCl2. For IRK, the presence of ATP and Mg2+ results in a leftward shift (ΔTM ~ 3.5°C) in the melting curve, indicating decreased IRK stability upon Mg2+-ATP binding. By contrast, the melting curves of PTK7, ROR2, RYK, and ROR1 were unchanged. Mutating the activation loop aspartate in ROR2 to alanine (D644A) – analogous to the IRK D1161A mutation that increases ATP-binding affinity ~20-fold even without phosphorylation (Ablooglu et al., 2001) – had no effect on the ATP-binding properties of ROR2. Curves represent means of three technical replicates.