Structural Study of the HD-PTP Bro1 Domain in a Complex with the Core Region of STAM2, a Subunit of ESCRT-0

EGFR is a key player in cell proliferation and survival signaling, and its sorting into MVBs for eventual lysosomal degradation is controlled by the coordination of multiple ESCRT complexes on the endosomal membrane. HD-PTP is a cytosolic protein tyrosine phosphatase, and is associated with EGFR trafficking by interacting with the ESCRT-0 protein STAM2 and the ESCRT-III protein CHMP4B via its N-terminal Bro1 domain. Intriguingly, the homologous domain of two other human Bro1 domain-containing proteins, Alix and Brox, binds CHMP4B but not STAM2, despite their high structural similarity. To elucidate this binding specificity, we determined the complex structure of the HD-PTP Bro1 domain bound to the STAM2 core region. STAM2 binds to the hydrophobic concave pocket of the HD-PTP Bro1 domain, as CHMP4B does to the pocket of Alix, Brox, or HD-PTP but in the opposite direction. Critically, Thr145 of HD-PTP, corresponding to Lys151 of Alix and Arg145 of Brox, is revealed to be a determinant residue enabling this protein to bind STAM2, as the Alix- or Brox-mimicking mutations of this residue blocks the intermolecular interaction. This work therefore provides the structural basis for how HD-PTP recognizes the ESCRT-0 component to control EGFR sorting.


Introduction
Epidermal growth factor receptor (EGFR) is a well-known cell-surface receptor tyrosine kinase, and one of the key regulators of cell survival and growth. Its aberrant expression or uncontrolled activity is directly implicated in a variety of tumors [1]. Binding of ligands, such as epidermal growth factor and transforming growth factor α, to the extracellular domain of EGFR triggers the homodimerization and autophosphorylation of the intracellular domain. A number of downstream signal transduction cascades subsequently initiate, eventually leading to cellular proliferation and differentiation and to the blockade of apoptosis [2,3]. EGFR activity is HD-PTP binds to STAM2(311-370) but not STAM2(260-310) (Fig 1B, middle and bottom panels). Using isothermal titration calorimetry (ITC), we next examined and quantified the interaction of HD-PTP(1-361) with three STAM2 peptides containing the residues 310-330, 330-350 and 350-370 of STAM2, respectively ( Fig 1C). The residues 350-370 of STAM2, referred to as STAM2(350-370), were identified to be necessary and sufficient for binding HD-PTP(1-361), with a resulting K D value of 6.06 μM (Fig 1C, third panel). We also confirmed using ITC that the STAM2 peptide compromising residues 371-416 is unable to bind HD-PTP(1-361) (Fig 1C, fourth panel).

Structure Determination of the HD-PTP Bro1 Domain Bound to the STAM2 Fragment
Based on this result, we subsequently attempted to determine the complex structure of HD-PTP  bound to the STAM2(350-370) peptide. Our initial trial was unsuccessful, because even though the protein and the peptide were mixed and incubated at a 1:5 molar ratio before crystallization, the resulting crystals (space group P1) contained four HD-PTP(1-361) molecules in the asymmetric unit but did not contain STAM2(350-370), which was revealed after the structure determination (S2A and S2B Fig; PDB code 5CRU). All four HD-PTP(1-361) monomers are very well matched with the previously determined crystal structure of the HD-PTP Bro1 domain (PDB code 3RAU) [20] with a root-mean-square deviation (RMSD) value in the range of 0.48-0.77 Å. It was previously indicated that the binding of the HD-PTP Bro1 domain to STAM2 can be compromised via an aspartate substitution of Leu202 and Ile206, implying that these residues play a role in the binding interaction [11]. We noticed that in our HD-PTP(1-361) structure, these residues (together with neighboring residues Arg205, Ala336, and Leu338) form hydrophilic and hydrophobic contacts with Asn33 and Tyr34 from an adjacent HD-PTP monomer (S2C Fig), suggesting that the crystal packing interaction between HD-PTP molecules will prevent STAM2(350-370) from being accommodated in HD-PTP(1-361). Therefore, with the expectation of altering crystal packing and promoting crystallization in a STAM2-bound form, we prepared a mutant HD-PTP(1-361) protein in which Asn33 and Tyr34 are substituted for alanine (referred to as HD-PTP(1-361;NAYA)). This mutant protein binds STAM2(350-370) as potently as the wild-type protein; this was confirmed using ITC (S3A Fig). After incubation at a molar ratio of 1:5 between HD-PTP(1-361;NAYA) and the STAM2(350-370) peptide overnight, we crystallized the protein sample and obtained novel crystals with the space group P2 1 containing two HD-PTP molecules in the asymmetric unit. Using these crystals, we finally determined the crystal structure of HD-PTP(1-361;NAYA) in a complex with STAM2(350-370) to 2.0 Å ( Table 1.

Analysis of the Interaction between HD-PTP and STAM2
Alix, Brox and HD-PTP are three human proteins containing a Bro1 domain that binds the ESCRT-III component CHMP4B, as is well established by means of structural determination and complex modeling studies [20,23,24]. Similar to that of Alix or Brox, the Bro1 domain of HD-PTP adopts a boomerang-like fold containing a concave binding pocket, where the C-terminal tail of CHMP4 may be accommodated [20]. In the HD-PTP(1-361;NAYA)−STAM2 (350-370) complex structure, the STAM2 peptide forms an amphipathic α-helix that binds to HD-PTP, mainly through hydrophobic interaction (Fig 2A and 2B and S6 Fig). In detail, the intermolecular hydrophobic interactions involve Val354, Leu358, Tyr361, Leu364 and Val365 of STAM2 and Leu189, Leu202, Ile206, Ala336, Leu338 and the hydrocarbon portions of Lys141, Lys192, and Arg198 of HD-PTP. Specifically, Tyr361 of STAM2 is located at the center of the hydrophobic cluster ( Fig 2B and S6 Fig), suggesting that this residue plays a key role in the complex formation. Alanine substitution of this residue completely abrogated the binding interaction between the two proteins, as confirmed by ITC measurements (S3B Fig). The involvement of Leu202 and Ile206 of HD-PTP in the hydrophobic interaction is in good agreement with the previous finding from a mutational study which showed that aspartate substitutions of these residues (the L/I-D/D mutation) impaired the complex formation [11]. Along with the hydrophobic interactions, we note that hydrogen bonds mediated by three water molecules also reinforce the intermolecular binding between HD-PTP and STAM2 (Fig 2C).

Thr145 Is a Key Determinant of HD-PTP in Binding STAM2
At a glance, CHMP4B(207-224) and STAM2(350-370) bind to the concave pocket of the Bro1 domain in a similar manner, mainly through their hydrophobic residues. We therefore superposed our HD-PTP(1-361;NAYA)−STAM2(350-370) complex structure onto the structure of the Bro1 domain of Alix bound to the CHMP4B peptide (residues 207-224; referred to as CHMP4B(207-224)). Indeed, STAM2(350-370) and CHMP4B(207-224) overlap when bound to the Bro1 domains, and the hydrophobic residues of the two proteins involved in the intermolecular interaction can be matched one by one ( Fig 3A). Thus, the ESCRT-III component CHMP4B should compete with and displace the ESCRT-0 component STAM2 from the Bro1 domain of HD-PTP, as previously suggested [11]. One notable difference between STAM2 (350-370) and CHMP4B(207-224) is that they associate with the Bro1 domains in opposite orientations (Fig 3A, bottom). Moreover, while the residues 354-367 of STAM2 are able to correspond to the residues 211-224 of CHMP4B, the preceding residues 351-353 of STAM2 do not, as the CHMP4B polypeptide terminates with Met224 matched to Val354 of STAM2 ( Fig 3A).
Despite the overall structural similarity among the Bro1 domains, a recent report by Ali et al. indicated that the Bro1 domain of Alix was not coprecipitated with STAM2 unlike that of HD-PTP [11]. We thus verified the interaction between the Bro1 domains and the core region of STAM2 using ITC. Indeed, unlike HD-PTP(1-361), neither the Bro1 domains of Alix (residues 1-359) nor Brox (residues 1-374) interact with STAM2(350-370) (Fig 3B), despite the fact that the key residues of HD-PTP in the binding to STAM2 are mostly conserved in Alix and Brox (S7 Fig). We therefore looked into the superposed complex structures. Intriguingly, we found that the STAM2 residues 351-353 play a key role in determining the binding partner of the protein; in the superposed models, this region brings about steric hindrance with the side chains of Lys151 of Alix and Arg145 of Brox, but not with that of the corresponding residue Thr145 of HD-PTP (Fig 3C; left and middle panels). This threonine residue at first did not appear as a key residue in the intermolecular interaction of HD-PTP with STAM2 in the complex formation. Indeed, the alanine substitution of Thr145 of HD-PTP(1-361) did not abolish its binding to STAM2(350-370) (S3C Fig). Nevertheless, structural superposition analysis revealed a possibility that the presence of threonine at that position in HD-PTP, instead of lysine as in Alix and arginine as in Brox, would facilitate the complex formation by "avoiding" or "not making" steric hindrance with the STAM2 helix. To corroborate this hypothesis, we prepared two mutant HD-PTP(1-361) proteins: Alix-mimicking HD-PTP(1-361;T145K) and Brox-mimicking HD-PTP(1-361;T145R). In the ITC experiments, neither HD-PTP(1-361; T145K) nor HD-PTP(1-361;T145R) interacted with the STAM2(350-370) peptide (Fig 3D), demonstrating the significance of the steric clash we found in the interaction between STAM2 and the Bro1 domains. Otherwise, CHMP4B does not contain residues corresponding to the STAM2 residues 351-353 causing steric hindrance (Fig 3C; right panel). We thus predicted that the mutation of Thr145 would not affect the interaction between HD-PTP and CHMP4B. This was also confirmed using ITC; CHMP4B(207-224) bound well to wild type and the HD-PTP mutant proteins (K D values of less than 8 μM; Fig 3E and S3D Fig) as anticipated.
Collectively, these results demonstrate that Thr145 is a unique functional determinant residue of HD-PTP, which enables the protein to bind to the core region of STAM2 without steric hindrance. Next, the interactions in human cells between STAM2, CHMP4B, and the Bro1 domain of HD-PTP were confirmed by pull-down assays. In order to concentrate on the Bro1 domainmediated intermolecular association and to exclude the effect of the additional binding between the SH3 domain of STAM2 and the central proline-rich domain of HD-PTP [11], Flag-tagged HD-PTP(1-712) variants together with HA-tagged CHMP4B and STAM2 proteins were transiently expressed in human embryonic kidney 293 (HEK293) cells, and immunoprecipitation assays were performed. Due to low expression level of full-length STAM2 ( Fig  4A), STAM2(1-370) and STAM2(1-370;Y361A) were subjected to immunoprecipitation. The  (Fig 4C), which is consistent with our structural analysis and binding measurements. We also confirmed that the STAM2 binding to HD-PTP is abrogated by the alanine substitution of Tyr361 of STAM2 ( Fig 4B; third and sixth columns), the key residue of the intermolecular hydrophobic interaction between the two proteins (see Fig 2B and S6 Fig).

Discussion
As previous studies have indicated, the boomerang-shaped Bro1 domain of HD-PTP shares considerable sequence and structural similarity with that of Alix or Brox (S7 Fig, bottom table). Together with their structural similarity, all the three Bro1 domain-containing proteins are known to interact with the ESCRT-III component CHMP4B through their concave pocket. Although the role of Brox in protein sorting is not yet well defined, both HD-PTP and Alix were determined to be involved in the EGFR sorting to the MVB [8,11]. Nevertheless, despite such similarities, Alix, Brox and HD-PTP also have their own unique structural features that enable the three proteins to function differently and specifically. For instance, only the Bro1 domain of Alix functions during the release of human immunodeficiency virus-1 (HIV-1), despite the fact that all three are able to interact with the nucleocapsid domain of the Gag protein of HIV-1 [25,26]. Structural studies revealed that this is due to a distinguishing Phe105 loop only present in the Bro1 domain of Alix, which is essential for the HIV-1 release [20,27]. On the other hand, the C-terminal tail of CHMP5 binds to the Bro1 domain of Brox but not to that of Alix or HD-PTP, in which nonconserved Tyr348 of Brox plays a critical structural role in constituting a unique binding pocket for the β-hairpin structure of CHMP5 [24]. Likewise, our structural and biochemical studies provide a rational explanation of the binding selectivity of STAM2 to the Bro1 domain of HD-PTP over that of Alix or Brox; the avoidance of steric hindrance due to the presence of a threonine residue instead of lysine or arginine is the key feature of the HD-PTP Bro1 domain, enabling this protein to accommodate STAM2 in its binding pocket.
In HD-PTP and Alix, but not in Brox, the Bro1 domain is followed by a V domain (residues 362-699 of HD-PTP). The two V domains share 17% sequence identity and 45% similarity with each other, and both were reported to interact with the Lys63-linked polyubiquitin chain [28][29][30]. The V domain of Alix is also known to provide a binding module for the YPX 3 L motif of protease-activated receptor 1 [31], the YPX n L late-domain motif of the p6 domain of the Gag protein of HIV and other viruses that require it for viral budding [32,33]. The residues in the V domain of Alix that associate with those motifs are mostly conserved in that of HD-PTP as well, including the key phenylalanine residue (Phe676 of Alix; Phe678 of HD-PTP) [34]. Interestingly, a study by Stefani et al. addressed that ubiquitin-associated protein 1 (UBAP1), an ESCRT-I component involved in EGFR sorting to the MVB, binds the V domain of HD-PTP but not the corresponding domain of Alix [35], providing another case of an ESCRT protein that selectively binds HD-PTP over Alix. Aspartate substitution of Phe678 of HD-PTP abolished its interaction with UBAP1, suggesting that the UBAP1-binding region might overlap with the presumed YPX n L motif-binding region in HD-PTP. We thus consider that, as in case of the HD-PTP Bro1 domain and the ESCRT-0 component STAM2, structural study of the V domain of HD-PTP would be necessary to elucidate the basis of its selective binding to UBAP1, which might also contribute to the understanding of the precise role and function of HD-PTP in EGFR sorting In this work, we discovered that the residues 350-370 of STAM2 constitute the HD-PTPbinding region, and presented the crystal structure of the Bro1 domain of HD-PTP in a complex with the core fragment of STAM2. We further delineated the structural feature of the Bro1 domain of HD-PTP, discriminating it from that of Alix or Brox, as certified by structural analyses, ITC binding measurements, and coimmunoprecipitation assays. We believe that this structural information will be a rational basis for future investigations to unravel the overall working mechanism of EGFR trafficking, which has received great interest due to its direct association with cell survival and cancer signaling.

Materials and Methods
Preparation, Crystallization, and Structure Determination of HD-PTP(1-361) and HD-PTP(1-361;NAYA) Bound to STAM2(350-370) The DNA fragment coding for the Bro1 domain of human HD-PTP (residues 1-361) was amplified by a polymerase chain reaction and cloned into the pET28a plasmid (Novagen), which was used as the template for the preparation of the mutant form of HD-PTP containing N33A and Y34A substitutions. Wild type or mutant HD-PTP protein was produced in the E. coli BL21(DE3) RIL strain (Novagen) at 18°C and initially purified using a Ni-NTA column (QIAGEN). After the removal of the N-terminal (His) 6  (v/v) acetonitrile. Before data collection, the complex crystals were immersed briefly in a cryoprotectant solution, which was the reservoir solution plus 5% glycerol. Diffraction data were collected on the beamline 5C at the Pohang Accelerator Laboratory, Korea, and processed using the program HKL 2000 [36]. The structures were determined by the molecular replacement method with the program Phaser [37] using the structure of the Bro1 domain of HD-PTP [20] as a search model. The programs Coot [38] and PHENIX [39] were used for model building and refinement, respectively. Crystallographic data statistics are summarized in Table 1.

Isothermal Titration Calorimetry
All measurements were carried out at 2°C on an iTC200 microcalorimetry system (GE Healthcare). Protein samples were dialyzed against the solution containing 20 mM Tris-HCl (pH 7.5) and 100 mM NaCl. The samples were centrifuged to remove any residuals prior to the measurements. Dilution enthalpies were measured in separate experiments (titrant into buffer) and subtracted from the enthalpies of the binding between the protein and the titrant. Data were analyzed using the Origin software (OriginLab Corp.).

Accession Numbers
The coordinates of HD-PTP(1-361) and HD-PTP(1-361;NAYA) in a complex with STAM2 (350-370) together with the structure factors have been deposited in the Protein Data Bank with the accession codes of 5CRU and 5CRV, respectively.