In and out: Traffic and dynamics of thrombopoietin receptor

Abstract Thrombopoiesis had long been a challenging area of study due to the rarity of megakaryocyte precursors in the bone marrow and the incomplete understanding of its regulatory cytokines. A breakthrough was achieved in the early 1990s with the discovery of the thrombopoietin receptor (TpoR) and its ligand thrombopoietin (TPO). This accelerated research in thrombopoiesis, including the uncovering of the molecular basis of myeloproliferative neoplasms (MPN) and the advent of drugs to treat thrombocytopenic purpura. TpoR mutations affecting its membrane dynamics or transport were increasingly associated with pathologies such as MPN and thrombocytosis. It also became apparent that TpoR affected hematopoietic stem cell (HSC) quiescence while priming hematopoietic stem cells (HSCs) towards the megakaryocyte lineage. Thorough knowledge of TpoR surface localization, dimerization, dynamics and stability is therefore crucial to understanding thrombopoiesis and related pathologies. In this review, we will discuss the mechanisms of TpoR traffic. We will focus on the recent progress in TpoR membrane dynamics and highlight the areas that remain unexplored.

EpoR and GHR. This is due to the presence of two cytokine receptor modules (CRM-1 and CRM-2). The extracellular domain essentially constitutes a sensor. Each of the CRMs is composed of a pair of fibronectin-III (FNIII)-like domains (D1 and D2 in CRM-1 and D3 and D4 in CRM-2) and two pairs of cysteines. A conserved WSXWS box characteristic of type I receptors is present at the membraneproximal end of CRM-2. 21 The fibronectin-III-like domain present in each CRM is composed of 7 antiparallel β strands and is interconnected by the hinge region. 21,22 D261 and L265 residues 23 within CRM-1 (D2) are the primary site for TPO binding, 24 and D128 and P136 residues between D1 and D2 maintain the conformation of ligand-binding elbow. 25 Hence, deletion of CRM-1 results in loss of TPO binding. 24 Deletion of the WSXWS conserved motif of CRM-1 does not affect TPO-binding capacity. 23 Instead, the WSXWS motif present at the base of the extracellular domain is known to stabilize the ligand-binding conformation of type I receptors. 26,27 Analysis of selected mutations from congenital amegakaryocytic thrombocytopenia (CAMT) patients and structure-guided mutagenesis revealed that F45, L103, R102 and F104 are potential ligand-binding sites. 23,28 Among these, TpoR R102P found in CAMT patients is restricted in the endoplasmic reticulum (ER). Interestingly, TpoR R102P is rescued for traffic to the cell surface and subsequent activation by TpoR agonist eltrombopag 29 when coexpressed with CALR exon 9 mutants (found in myeloproliferative neoplasm). Recently, a novel TpoR mutation (TpoR R464G) has been detected in patients diagnosed with CAMT. 30 TpoR R464G could not be detected on the surface of platelets in these patients although low levels of surface expression coupled with limited activation by TPO were observed when expressed in Ba/F3 and UT-7 cell lines. However, unlike TpoR R102P, co-expression of CALR mutant could not rescue the traffic defect of TpoR R464G. The exact mechanism of action of CALR mutants on these traffic-deficient TpoR, however, remains elusive.
The transmembrane and juxtamembrane domain (22aa) folds into an α-helix and acts as the control centre for dimerization and activation of the receptor. This region may also exist in different dimeric-conformational orientations. 31 Experiments by fusing Put 3 coiled-coil domains to the transmembrane region of the TpoR and engineering the junction of dimeric coiled-coil and TpoR showed that it could signal from 6 different orientations. The extent of signalling may differ, as evidenced by the differences in the F I G U R E 1 Schematic TpoR structure depicting the position of the important residues. The conserved WSSWS motifs (shown in red) in the extracellular domain of receptor and Box 1 (green) and Box 2 (blue) part of the cytosolic domain are indicated. Four N-glycosylation N117, N178, N298 and N358 (in green) and residues mutated in hereditary thrombocytosis R102, F104 and P106 (in violet) are shown. The hydrophobic patch in the TpoR extracellular domain is indicated in yellow. Eltrombopag-binding site at residue H499 of human TpoR is indicated in magenta. Human H499 and its equivalent murine L492 are shown along with the position S505 (human) and S498 (murine) for comparison proliferation of Ba/F3 and UT-7 cell lines expressing these constructs. 31 Although the transmembrane region is composed of a relatively short stretch of amino acids when compared to the full receptor, this region plays a crucial role in halting ligand-independent activation of TpoR. Specifically, H499 and RW 515 QFP are two essential motifs for the prevention of ligand-independent activation of TpoR. 32,33 Incidentally, eltrombopag, an agonist of TpoR which binds at position H499, is used to treat thrombocytopenia. 32,34,35 W515, part of the RW 515 QFP motif located at the juxtamembrane region, is responsible for maintaining TpoR in an inactive state in the absence of its ligand. Hence, mutation at W515 to any residue except proline and cysteine (W515P and W515C) results in constitutive activation of TpoR. 36 In fact, myeloproliferative neoplasm (MPN) patients with essential thrombocythemia (ET) and primary myelofibrosis (PMF) were found to harbour activating transmembrane domain mutations W515L/K/R/A, S505N and V501A. 37 Of these, S505 and V501 appear at the dimeric interphase and maintain the inactive conformation of the receptor. Recently, saturation mutagenesis of the transmembrane domain revealed that second site mutations in the same domain modulated the effects of these driver mutations.
For example, R514K enhanced the ligand-independent activation of S505N. 38 Moreover, aberrant activation of the TpoR extracellular domain by oncogenic mutations S505N and W515A/L/K was found to depend upon W491 residue of the extracellular domain. 39 Taken together, amino acids at dimeric interface (V501, S505) and juxtamembrane region (W515, L498) prevent ligand-independent activation of the receptor. 38,39 Of note, the transmembrane domain of murine TpoR has higher propensity for dimerization in comparison with human TpoR. This has been attributed to the absence of the H499 residue in murine TpoR that results in an uninterrupted helical conformation of the transmembrane domain favouring the activation of murine TpoR. 32 Indeed, such a scenario is observed in v-MPL, an oncoprotein present in murine myeloproliferative leukaemia virus (MPLV). 3

| RECEP TOR DIMERIZ ATION
Previously, it was thought that TpoR exists on the surface as preformed dimers or in a monomer-dimer equilibrium. Biophysical studies using the transmembrane and juxtamembrane domains indicated that unstimulated human TpoR might be monomeric, while the murine TpoR might exist at least partially as preformed dimers. 32 These differences between the murine and human TpoR were attributed to the H499 residue in the transmembrane domain that is unique to the human receptor. 32,51 Indeed, H499 interrupts the alpha helix that might be required for preformed dimerization in an inactive orientation. 32 Dimerization of TpoR has also been detected in cells using various methods such as cysteine cross-linking assay, TOXCAT or FRET-based protein-protein interaction assays. 31,52 However, these approaches may give rise to a false interpretation of preformed dimers as these techniques can detect weak interactions between the receptor monomers. 52 Besides, elevated cell surface expression of TpoR and ensuing crowding and weak interactions of monomers cannot be ruled out. 42 Recently, utilizing physiological expression and single-molecule co-locomotion imaging of post-translationally labelled full-length surface monomers, it has been shown that class I cytokine receptors TpoR, EpoR and GHR predominantly exist on the surface as monomers which form stable signal-transducing dimers only when bound to their ligands. 42 It is important to note that using a similar technique, low-affinity IL6-R ligand (C7 and A1 IL6 engineered variants) displayed low to no dimerization but carried out strong STAT5 activation indicating that signalling and dimerization may be uncoupled at least for low-affinity IL6-R ligands. 53 When bound by its ligand, stable dimerization of TpoR occurs aided by the dimerization of the JAK2 pseudokinase (PK) domain.
This paves the way for the activation of the C-terminal tyrosine kinase domain of JAK2. Deletion of the PK domain leads to reduced stability of active TpoR dimers, while the absence of JAK2 causes lower levels of intrinsic dimerization of TpoR. 42 A stabilizing interaction between JAK2 PK domains has been hypothesized by Wilmes et al. to be important for the dimerization of cytokine receptors.
Along these lines, the FERM domain mutation of JAK2 (JAK2 L224E), which inhibits the PK-PK interaction, significantly inhibits TpoR dimerization even in the context of the activating PK domain mutation JAK2 V617F (driver mutation in MPN). Therefore, the extent of TpoR dimerization and signalling is determined by FERM and PK domains of JAK2 as shown by L224E and V617F mutations, respectively. 42 Furthermore, experiments with TpoR W515L and JAK2 V617F have provided insight into factors affecting ligand-independent dimerization of TpoR. W515L has been shown to induce strong dimerization in the presence of JAK2, only transient weak dimerization in the absence of JAK2, while minor dimerization has been observed with TYK2. JAK2 V617F drives effective ligand-independent dimerization of TpoR and EpoR, 54 while substantially less dimerization was observed for GHR. Similar to TpoR, and for all the three IL-4 receptor complexes, IL-4:IL-4Rα/IL-13Rα1 (type 2 complex), IL-13:IL-4Rα/ IL-13Rα1 (type 2 complex) and IL-4:IL-4Rα/IL-2Rγ (type 1 complex); fluorescence cross-correlation spectroscopy, as well as intramembrane dissociation constants, indicated the formation of short-lived transient dimers which are incapable of triggering signalling in the absence of the ligand. 55 Taken together, TpoR appears to exist as signalling competent dimers only in the presence of its ligand. Ligandindependent TpoR dimers are stabilized by specific mutations in JAK2 (JAK2 V617F) that enhances PK-PK interactions of adjacent JAKs or TpoR juxtamembrane mutations (TpoR W515) that relieve conformational inhibition.

| REG UL ATION OF TP OR S IG NALLING
Ligand stimulation leads to stable dimerization of TpoR. Dimerization triggers downstream signalling molecules such as STATs causing TpoR-specific transcription that maintains HSC population 11,12,15,16 as well as platelet production. 13,17 Having such an important role in haematopoiesis, TpoR signalling must be tightly regulated through multiple mechanisms. TpoR signalling is known to be regulated by two means. The first involves activation of the negative regulators of TpoR signalling cascade such as suppressor of cytokine signalling (SOCS), and 56,57 PIAS 58,59 -PIAS inhibits STAT DNA-binding activity, LYN 60 and LNK. 61 Further details may be found in the reviews. [62][63][64] The second method is by internalization of the active receptor complexes followed by degradation or recycling of the receptor to the surface resulting in attenuation of downstream signalling. 65 TpoR internalization has been followed by co-staining the cells for transferrin endocytosis. 66 Such experiments showed that endocytosed TpoR accumulates in early-endosomal vesicles. TpoR endocytosis is regulated by dynamin as chemical inhibition using Dynasore-blocked TpoR internalization. 29 Importantly, Dnm2 knock-out mice showed pronounced macrothrombocytopenia with impaired TpoR endocytosis. 67 Impaired endocytosis was accompanied by sustained TpoR signalling and JAK2 phosphorylation. Therefore, internalization of active TpoR complexes is essential to attenuate TpoR signalling. Additionally, the phenotype of macrothrombocytopenia appears to be paradoxical, suggesting that excessive signalling by TpoR may have negative consequences. 66,68 TpoR cytoplasmic domain associates with AP2 to induce TPO-stimulated and clathrin-mediated endocytosis of the receptor. 65 Internalization of TpoR is controlled by two intracellular motifs Y626 (fourth cytoplasmic tyrosine residue) and box 2 region (L 567 L 568 E 569 I 570 L 571 ) containing dileucine motifs (L 567 L 568 and I 570 L 571 ). However, box 2 exhibits internalization independent of JAK2 activation. 69 This is consistent with the study done by Royer et al. concluding that only box 1 and sequence between box 1 and box 2 (Q 532 Y 533 L 534 in murine homologue) are required by FERM domain of JAK2 to enhance cell surface expression as well as internalization of TpoR. 70 When stimulated with TPO, TpoR is ubiquitinated at K 553 and K 573 resulting in degradation of the receptor. 71 This is mediated by E3 ubiquitin ligases such as CBL. Crucially, CBL mutations are frequently detected in MDS/MPN. 72 It may be noted that G-CSFR continues signalling even after internalization in early endosomes. 73 As observed in the case of K5R mutant of G-CSFR where all the cytoplasmic lysines are mutated to arginine, STAT5 and ERK activation increases after internalization to early endosomes without being localized into late endosomes and lysosomes. Similarly, engineered high-affinity IL6 ligands (HyIL6) co-localized in early-endosomal compartment while low-affinity ligands displayed little or no colocalization in the same compartment. This correlated with robust STAT1/3 activation by HyIL6 as compared to low-affinity ligands. 53 Recent data have revealed that MPN-associated CALR mutants induced early-endosomal localization of TpoR. 29 Whether early-endosomal localization contributes towards increased signal amplitude remains unexplored for TpoR-mutant CALR complexes.
Platelet surface TpoR acts as a rheostat by regulating the availability of free circulating TPO in the blood. Similar to megakaryocytes, platelet TpoR binds to serum TPO resulting in the endocytosis of the complex. Endocytosis is mediated by Dynamins as evidenced by impaired TPO-induced TpoR endocytosis resulting in increased serum TPO levels in Dnm2 −/− mice. 67 Additionally, mislocalized early-endosomal markers (EEA1) and abnormal clustering of clathrin away from the plasma membrane were observed in Dnm2 −/− megakaryocytes. Together, these data indicate a clathrin and dynamindependent endocytosis of platelet surface TpoR. Platelets also show the presence of recycling endosomes 74 along with a characteristic TpoR surface recovery kinetics following TPO stimulation. 67 However, very little information is available regarding the other molecular components involved in endocytosis of the TPO-TpoR complex and recycling/degradation of TpoR, which may be different in HSCs and early MK progenitors. Importantly, TpoR expression correlates with the number of hematopoietic stem cells, 12 megakaryocyte progenitors, megakaryocytes and platelets. 75,76 Mice lacking TPO or TpoR are severely thrombocytopenic and deficient in megakaryocytes and their progenitors. 75 When TpoR is expressed in progenitors but not in megakaryocytes and platelets, platelet numbers increase due to lack of internalization and clearance of TPO from circulation. 76 Additionally, TPO has been shown to prime HSCs towards the megakaryocyte lineage. 77 This explains the paradoxical thrombocytosis observed in mpl −/− mice engineered to express low levels of TpoR wherein excess serum TPO enhances megakaryopoiesis. 78 Similarly, two partially traffic-deficient TpoR, viz. TpoR K39N and P106L mutants, result in hereditary thrombocytosis due to the presence of excess TPO in circulation, which stimulates megakaryocyte progenitor proliferation. 79 79 Low TpoR P106L activity is correlated with low surface expression and an internalization defect. Of note, the region between R102 and P106 is important for cell surface expression of TpoR as well as its ligand-binding activity. While R102P is blocked in the ER and is unresponsive to TPO, P106L shows the partial response to the ligand. 79

| TP OR AC TIVATI ON BY AG ONIS TS
TpoR agonists have been designed to tune TpoR signalling by either decreasing the distance between the monomers or changing the dimeric-conformational interface or dimeric topology, making the receptor active to various extents. These include eltrombopag, romiplostim and diabodies. Eltrombopag was first identified in a high-throughput screen of small molecule compounds capable of activating STAT in Ba/F3 cells expressing TpoR. 80 Crucially, eltrombopag was found not to compete with Tpo for binding to TpoR. Instead and as previously described, eltrombopag was observed to bind to H499 residue of human TpoR resulting in effective dimerization and activation of the receptor. 32 The dependency on H499 residue also marks a crucial difference between murine and human TpoR whereby only human TpoR containing H499 is activated by eltrombopag. Moreover, asparagine-scanning mutagenesis of the transmembrane domain of murine TpoR revealed that mutation at several residues was capable of activating the receptor. However, only S505N was found to activate the human TpoR. This differ- R464G. The TpoR R464G was observed to be unresponsive to TPO or eltrombopag. 30 However, upon co-expression of CALR del52, TpoR R464G showed selective activation with eltrombopag alone. It is possible that R464G mutation locks TpoR in an inactive topological orientation which is relieved upon binding to CALR del52 making it accessible to eltrombopag.

| TP OR SURFACE LO C ALIZ ATI ON IS MED IATED BY JAK 2 AND T YK 2
TpoR has four sites for N-glycosylation (N 117 , N 178 , N 298 and N 358 ).
While core glycosylation at the four Asn residues occurs in the ER, the addition of mature glycans requires passage through the Golgi.  86 In the absence of the ligand, TYK2 prevents the interaction between the internalization motif (Y466) on IFNAR1 and AP50 (subunit of AP2). These observations show that the masking effect of TYK2 reduces the basal internalization rate, thereby increasing the half-life of IFNAR1. 86

| GOLG I -INDEPENDENT TR AFFIC OF TP OR
TpoR utilizes both Golgi-dependent and Golgi-independent routes for traffic to the cell surface ( Figure 2). The Golgi-dependent route of TpoR traffic marks exits through the ER-Golgi to the cell membrane (anterograde secretion pathway) and is used by complex glycosylated TpoR. The Golgi-independent pathway is utilized by TpoR containing immature glycosylation and is believed to be processed through autophagosomes to the cell surface. 87 55 Association of TpoR with JAK2 is essential for the presentation of mature TpoR to the surface through an anterograde secretion pathway. 87 It is known that JAK2 and TYK2 increase the half-life of the mature form of TpoR. 70 However, mutant JAK2 V617F coexpressed with TpoR in Ba/F3 cell line has been shown to increase the half-life of the immature form of TpoR rather than mature TpoR. 66 Although surface expression of immature TpoR appears to be correlated with Golgi-independent traffic, it remains to be verified in the case of TpoR associated with CALR exon 9 mutants and JAK2 V617F.
Further details of surface expression, N-glycosylation status and traffic routes for the various mutants are provided in Table 1.

| C ALRE TI CULIN MUTATI ON S IN MPN
CALR is an ER-resident chaperone with three distinct functional domains. 88 The N-terminal domain contains glycan-dependent and F I G U R E 2 TpoR traffic routes. The classical Golgi-dependent (solid lines), autophagosome-lysosome dependent (dashed lines) traffic routes and the endocytic pathway (in green arrow) are depicted. Glycosylation status of TpoR-immature (in red) and mature (in blue) are indicated throughout the TpoR traffic routes glycan-independent polypeptide-binding sites essential for its chaperone activity. The high-affinity Ca 2+ -binding site containing prolinerich P-domain interacts with the thiol oxido-reductase Erp57 and is involved in glycan-independent chaperone activity. The acidic Cterminal domain contains multiple high capacity, low-affinity Ca 2+ binding sites that regulate ER Ca 2+ buffering and homeostasis. The  93,94 Interestingly, CALR mutants selectively activated TpoR and to a weaker extent G-CSFR. 95 Furthermore, shRNA-mediated TA B L E 1 Effects of the expression of WT and mutant TpoR, JAK2 and CALR on the cell surface expression of TpoR

| PER S PEC TIVE S
Defects in TpoR traffic are associated with multiple pathological conditions. Yet, the exact mechanisms of TpoR traffic and sorting post-receptor endocytosis remain unknown. Recent reports on fibrinogen endocytosis in platelets have implicated Arf6 (small Raslike GTP-binding protein) and VAMP-3 (v-SNARE) proteins in the process. 74,105 It remains to be seen whether these effectors also modulate endocytosis of TpoR. Although the conventional anterograde transport of TpoR has been widely studied, we are just beginning to appreciate the unconventional autophagosome-lysosomal route. It is not clear how much each of these routes contributes towards surface TpoR expression in HSCs, progenitors and platelets.
Moreover, how these routes affect the pathophysiology of TpoR in relation to TpoR/JAK2/CALR mutations needs to be explored.
For example, the unconventional autophagosome-lysosomal traffic of TpoR could be detected in cells with JAK2 V617F or TpoR P106L. 66,79 However, it remains to be established whether blocking the unconventional traffic of TpoR affects the disease pathology. Unconventional traffic that bypasses Golgi would necessitate changes in TpoR N-glycosylation. Of note, mutations targeting individual Asn residues responsible for N-glycosylation of TpoR showed little effect on its surface expression and response to TPO. 106 However, combinatorial mutations did indeed decrease surface expression and TpoR signalling. Nevertheless, we do not understand whether the N-glycosylation status affects receptor internalization and membrane dynamics. Answers to these questions will serve to better understand paradoxical thrombocytosis and aid in the development of effective TpoR agonists/antagonists.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.