Disrupting the Transmembrane Domain Oligomerization of Protein Tyrosine Phosphatase Receptor J Promotes its Activity in Cancer Cells

Despite the critical regulatory roles that receptor protein tyrosine phosphatases (RPTP) play in mammalian signal transduction, the detailed structural basis for the regulation of their catalytic activity is not fully understood, nor are they generally therapeutically targetable. It is due, in part, to the lack of known natural ligands or selective agonists. In contrast to conventional structure-function relationship for receptor tyrosine kinases (RTKs), the activity of RPTPs has been reported to be suppressed by dimerization, which may prevent their access to their RTK substrates. We report here that: (i) homodimerization of PTPRJ (also known as DEP-1) is regulated by specific transmembrane (TM) residues, and (ii) disrupting these interactions can destabilize full-length PTPRJ homodimerization in cells, reduce the phosphorylation of EGFR (a known substrate) and downstream signaling effectors, antagonize EGFR-driven cell phenotypes, and promote substrate access. We demonstrate these points in human cancer cells using both mutational studies and through the identification of a peptide designed to bind to the PTPRJ TM domain. This peptide is the first example of such allosteric agonist of RPTPs. This study, therefore, provides not only fundamental structure-function insights on how PTPRJ activity is tuned by TM interactions in cells but also opportunities to develop a unique class of agents that could be used as tools to probe RPTPs signaling regulating mechanisms or for therapeutic purposes in cancers driven by RTK signaling.


Introduction
Normally, cell signaling by receptor tyrosine kinases (RTKs) is tightly controlled by the counterbalancing actions of protein tyrosine phosphatases (PTPs), which dephosphorylate regulatory tyrosines on RTKs to attenuate their signal-initiating potency. Twentytwo PTPs are classified as receptor-like PTPs (RPTPs), which all share the same architecture: an extracellular region (often including fibronectin type III repeats and immunoglobin-like domains), a singlepass transmembrane segment, and either one or two cytosolic highly conserved PTP domains (1,2). These integral membrane PTPs are some of the most important regulators of RTKs, and their normal function is critical for homeostatic control (2,3). Interestingly, while RPTPs can act as suppressors of several tumors, including colon, lung, breast, and thyroid cancers, they can also act as oncogenes when deregulated either through reduced expression, often due to loss of heterozygosity or loss-of-function mutations (4)(5)(6)(7)(8). Notably, fewer than one-third of reported point mutations occur within the intracellular phosphatase domains indicating inactivation occurs without structural changes to the catalytic domain (9). RPTPs have therefore long been viewed as potential therapeutic targets.
Of the RPTPs characterized, PTPRJ (also known as DEP-1) is particularly intriguing as it has been identified as a regulator of C-terminal tyrosine phosphorylation in the epidermal growth factor receptor (EGFR) (10)(11)(12), a frequent driver of oncogenic signaling across numerous cancers. Deregulation of mechanisms controlling EGFR activation and expression results in abnormal cell growth and proliferation. PTPRJ has a direct role in regulating EGFR activity, implicating it as a relevant tumor suppressor. Furthermore, PTPRJ is often down-regulated in numerous cancers, and its restoration was shown to inhibit malignant phenotypes both in vitro and in vivo (13).
While small molecule allosteric inhibition of some non-receptor PTPs is now possible (14), methods to specifically target RPTPs are missing, due, in part, to the limited understanding of their mechanism of action and the lack of known natural ligands. Indeed, compared to the structure-function relationships for RTKs, relatively little has been elucidated for RPTPs. For instance, the reported ability of homodimerization to antagonize RPTP catalytic activity appears to be a general feature of the entire family, but there is no proposed common model. The "head-to-toe" dimerization model for the PTP D1 and D2 intracellular domains, proposed originally by Barr et al. for PTPRG, is generally accepted for RPTPs with tandemly repeated intracellular domains (1). However, since PTPRJ and other members of the R3 sub-family have only one catalytic intracellular PTP domain, the "head-to-toe" model and the "inhibitory wedge" model proposed for PTPRA (15,16) seem to be incompatible. Nevertheless, the reported ability of homodimerization to antagonize PTPRJ catalytic activity and substrate access presents potential opportunities to develop strategies to promote RPTP activity against their oncogenic RTK substrate (17). While the transmembrane (TM) domain of several RPTPs has been proposed to be involved in their homodimerization (17)(18)(19)(20), and therefore represents an attractive target, there is no clear structure-based proposal for how this occurs. Therefore, elucidating the contribution of the TM domain in RPTPs, and particularly in PTPRJ regulation, can provide significant insight into how these receptors function, interact, and eventually be modulated, leading to new methods to target signaling of oncogenic RTKs that may be less susceptible to common mechanisms of resistance.
Here, we employed mutational studies to show that PTPRJ homodimerization is regulated through specific TM residue interactions. Furthermore, disrupting these interactions antagonizes PTPRJ homodimerization, thereby promoting its phosphatase activity and ultimately inhibiting EGFRdriven cell phenotypes. We used these new findings to identify and characterize a synthetic peptide that interacts with and disrupts PTPRJ homodimers through specific TM interactions. We show the delivery of this peptide selectively modulates the dimerization state and activity of PTPRJ in EGFRdriven cancer cells. The present study represents a structure-function determination of the TM domain of PTPRJ, and a new way to selectively modulate the activity of this important class of phosphatases in cancer cells.

PTPRJ self-association is mediated by specific TM residues
To assess whether the transmembrane (TM) domain and specific amino acid residues play a role in the self-association of PTPRJ, we first used the dominant-negative AraC-based transcriptional reporter assay (DN-AraTM) (21,22). This assay reports on the propensity of TM domains to selfassociate and heterodimerize in cell membranes. Briefly, it relies on a protein chimera containing the receptor domain of interest fused to either the transcription factor AraC (which is active at the araBAD promoter as a homodimer), or to an AraC mutant unable to activate transcription (AraC*). Both chimeras include an N-terminal maltose-binding protein (MBP) fusion that directs chimera insertion in the inner membrane of AraC-deficient E. coli (SB1676). Homodimerization of AraC (a result from receptor domain self-association) induces the transcription of the gene coding for the green fluorescent protein (GFP). Thus, GFP fluorescence intensity is a measure of receptor domain dimerization.
When the TM domain of wild-type PTPRJ (WT) ( Figure 1A) was expressed as a fusion to AraC, a strong GFP signal was observed, indicating that the TM domain has a high propensity to self-associate ( Figure 1B; black bar). On the other hand, when PTPRJ WT-AraC* was co-expressed as a competitor to PTPRJ WT-AraC, a significant decrease in GFP signal was observed ( Figure 1B, hashed bar), consistent with specific TM-TM interactions. Having confirmed that PTPRJ self-associates, in part through the TM domain, we sought to identify the amino acids required for this interaction. To alter interfacial packing and introduce steric clashes, amino acids in the TM region with small side chains (Ala, Cys, Gly,) were mutated to leucine, while those with larger side chains (Ile, Leu, Phe, Val) were mutated to alanine (23). The propensity of each construct to selfassociate was then analyzed using the DN-AraTM assay. Among all the tested TM residues, three mutations (G979L, G983L, and G983L) caused the greatest decrease in GFP signal compared to WT ( Figure 1C), suggesting that these three residues are involved in stabilizing PTPRJ self-association. Interestingly, the arrangement of these three glycines places them on the same helix face ( Figure 1D) and is consistent with a double GxxxG zipper (18), a common motif for TM packing interfaces (24)(25)(26)(27)(28), and further pointing to specific homodimeric interactions.
Immunoblotting and maltose complementation assays showed that all constructs were expressed at similar levels ( Figure 1C, inset) and properly inserted in the membrane ( Figure S1), confirming that variation in the fluorescence signal was solely due to changes in association state of the constructs.

PTPRJ point mutants decrease EGFR phosphorylation, downstream signaling, and inhibit EGFR-driven cell migration
Next, we determined the functional significance of receptor point mutations in cells. We hypothesized that disrupting PTPRJ self-association through point mutations at G979, G983, and G983 would lead to a decrease in phosphorylation of EGFR and downstream signaling intermediates, and ultimately inhibit EGFR-driven cancer cell migration. To test our hypothesis, full-length WT, G979L, G983L and G987L PTPRJ were ectopically and stably expressed in UMSCC2 cells, an EGFR-driven human squamous cell carcinoma cell line that exhibits very little PTPRJ expression (Figure 2A; EV). All four cell lines expressed similar levels of PTPRJ and EGFR ( Figure S2). After serum starvation and treatment with EGF, the level of phosphorylation of EGFR (at Y1068 and Y1173, two representatives of EGFR phosphorylation across cytoplasmic tyrosines) (28,29), and of ERK1/2 was measured by immunoblotting. Consistent with our hypothesis, a significant decrease in EGFR and ERK1/2 phosphorylation was observed with the glycine to leucine point mutants when compared to wild-type PTPRJ ( Figure 2; + 10 ng/mL EGF). Moreover, not only does expression of WT PTPRJ significantly inhibit EGFR-driven cell migration when compared to empty vector ( Figure 3), but G983L and G987L receptor mutants also inhibit cell migration to an even greater extent than WT PTPRJ ( Figure 3). Remarkably, these results are consistent with the immunoblotting data, where G983L and G987L receptor mutants lowered EGFR phosphorylation the most (Figure 2).

PTPRJ point mutant disrupts full-length PTPRJ oligomerization and promotes its access to EGFR
To determine whether changes in phosphorylation and cell migration were direct functional consequences of the disruption of PTPRJ homodimers, we assessed the oligomerization of full-length PTPRJ WT and G983L (i.e., the mutant disrupting self-association, kinase phosphorylation, and cell migration the most) expressed in UMSCC2 cells by in situ proximity ligation assay (29)(30)(31). Figure 4 shows that strong PLA signal was observed with WT PTPRJ, indicating that PTPRJ exists as oligomers. Strikingly, a significant decrease in PLA signal was observed in cells expressing the receptor with the G983L point mutant compared to those with WT, confirming that the G983L point mutation disrupts PTPRJ oligomerization, and significantly reduces the number of PTPRJ oligomers.
Moreover, since physical interaction of PTPRJ with RTKs, such as EGFR, appears integral to its regulatory role of RTK signaling (10, 29), we investigated whether destabilizing full-length PTPRJ oligomerization promotes its interaction with EGFR. Figure 5 shows that cells expressing the G983L PTPRJ point mutant exhibited a stronger PLA signal between PTPRJ and EGFR than cells expressing wildtype PTPRJ. This suggests that, as hypothesized, disrupting PTPRJ self-association favors its access to EGFR.
Altogether, our results indicate that TM domain association stabilizes PTPRJ oligomerization through a specific contact interface consisting of G979, G983 and G987 (i.e., GxxxGxxxG motif) and that disrupting PTPRJ self-association through point mutations promotes PTPRJ phosphatase activity, likely by favoring its access to EGFR. Our results also suggest that this structure-function relationship may provide a new opportunity to augment PTPRJ activity in cancer cells driven by EGFR signaling.

Design and identification of a TM peptide sequence capable of antagonizing PTPRJ self-association
To leverage the identified regulatory TM domain interactions of PTPRJ, we set out to develop exogenous peptides that can modulate PTPRJ enzymatic activity by disrupting receptor oligomerization. To do so, we constructed a library of peptides designed to selectively bind to the TM domain of PTPRJ (see Experimental procedures). Similar peptide library design approaches have proven successful at identifying and selecting peptides to inhibit viral TM domain dimerization as well as transmembrane receptor dimerization (32,33). The degenerative oligonucleotide sequence corresponding to the designed TM binder consensus peptide sequence ( Figure 6A) was subcloned in fusion with AraC* to be tested against WT PTPRJ-AraC, using the DN-AraTM assay (34). After cotransforming these constructs into E. coli, individual colonies were tested for GFP emission. Of the 299 individual colonies that were screened ( Figure 6B), 28 clones ( Figure 6B, green dots) showed greater disruption of PTPRJ self-association than the wildtype PTPRJ construct ( Figure 6B, red line). After further DN-Ara-TM validation and DNA sequencing, the sequence of the clone that consistently showed the most GFP signal disruption was selected as the lead peptide sequence ( Figure 6C, lead). Importantly, this lead peptide sequence showed greater disruption of PTPRJ self-association than the WT sequence as designed ( Figure 6D). Next, we set out to identify residues important for the interaction of this lead peptide binder sequence with the TM domain of PTPRJ. Guided by preliminary molecular simulations indicating that Ser7, Ala11 and V15 ( Figure 6C) may be part of the contact interface with PTPRJ (data not shown), we mutated these 3 residues to leucines with the goal to disrupt helix packing, and tested their effect on PTPRJ self-association using the DN-AraTM assay. Figure 6E shows that all three pointmutations significantly hinder the ability of the lead peptide sequence to disrupt PTPRJ self-association, with A11L having the most effect.

Interaction of RJbinder with Membrane Mimics and Cell Membranes
Peptides corresponding to the lead and the control A11L sequences (referred to subsequently as RJbinder and RJcontrol, respectively) were synthesized and purified ( Figure S4). Due to the relatively high hydrophobicity of the peptide sequences, two lysine residues were appended to the N-and C-termini to enhance water solubility and facilitate purification ( Figure 6C) (35). However, despite these modifications, solubilization of such peptides in detergent is required to prevent aggregation in solution and allow partitioning into cell membranes (36)(37)(38)(39). Based on the design approach and the hydrophobicity of these peptides, one would expect that they insert and fold as TM α-helices in both detergent and lipid membrane mimics. Circular dichroism spectroscopy shows that both RJbinder and RJcontrol fold into α-helices in the presence of noctylglucoside micelles (39) and POPC large unilamellar vesicles ( Figure S4 A and B). Proper membrane insertion in both membrane mimetic environments was also confirmed by fluorescence emission measurements: A significant blue shift in tryptophan fluorescence emission (> 10 nm) was observed when peptides were incubated with either noctylglucoside micelles or POPC large unilamellar vesicles ( Figure 6C and D), blue line) when compared to fluorescence emission in buffer. Taken together, these results indicate that both RJbinder and RJcontrol adopt a α-helical transmembrane conformation in membrane mimetic environments. We also explored whether RJbinder can incorporate into UMSCC2 cell membranes. To this end, we studied the cell membrane partition of a fluorescently labeled version of RJbinder initially incorporated into n-octylglucoside micelles. To allow peptide partition into cell membranes, peptide/octylglucoside solutions were added to cells to obtain a final detergent concentration (3 mM) well below its critical micellar concentration (~ 15 mM as measured in cell media). Confocal microscopy showed the fluorescently-labeled RJbinder peptide to be localized on the cell surface ( Figure S5).

RJbinder disrupts PTPRJ self-association and inhibits EGFR signaling
Next, we investigated the effect of exogenously adding RJbinder to UMSCC2 cells on PTPRJ dimerization and cell phenotypes. We hypothesized that, similarly to dimerization-disrupting point mutants (e.g., G983L), treating cells with RJbinder (initially incorporated into n-octylglucoside micelles) would inhibit not only receptor dimerization, but also EGFR signaling. Figure 7 shows that a significant decrease in PLA fluorescence average signal was observed following treatment of cells expressing WT PTPRJ with RJbinder, indicating the ability of the peptide to disrupt PTPRJ self-association. Moreover, significant inhibition of EGFR phosphorylation (at both Y1068 and Y1173) (Figure 8), and cell migration was observed in cells treated with RJbinder ( Figure 9). We also hypothesized that RJbinder should have little effect (if any) on cells expressing either minimal PTPRJ (EV) or the homodimer-disrupting mutant G983L (as the PTPRJ homodimer is already destabilized). Figure S6 shows that RJbinder indeed had no effect on cell migration when added to these two cell lines. Crucially, cell treatment with RJcontrol did not affect receptor self-association (Figure 7), EGFR phosphorylation ( Figure 8) or wound closure ( Figure  9). Altogether, these results imply that the specific inhibitory effects observed in cells were due to a direct interaction of RJbinder with the cell endogenous PTPRJ receptor.

Discussion
The main goal of this work was to extend our structure-function understanding of the role of TMmediated interactions in modulating the activity of PTPRJ. We showed that PTPRJ dimerization is mediated by a specific single-contact TM interface constituted by two overlapping GxxxG motifs, a common motif for TM-TM homodimerization. Interestingly, when the corresponding full-length PTPRJ glycine point mutants (G979L, G983L and G987L) were individually expressed in UMSCC2 cells, we observed a decrease in EGFR and ERK phosphorylation, and inhibition of EGFR-driven wound closure. We also showed that expression of the most disruptive PTPRJ mutant (G983L) led to a decrease in full-length PTPRJ homodimer formation in cells when compared to the wild type receptor, validating the importance of this particular glycine residue in mediating receptor dimerization. As anticipated, the suppression of phosphorylation, wound closure and dimerization, while significant, is not complete. There are several potential reasons: interfering with a single amino acid is unlikely to completely disrupt full-length receptor selfassociation in cell membranes; TM-TM interaction may not be the only mediator of PTPRJ dimerization (1,40); and EGFR is known to be regulated by a number of other PTPs, such as PTPRS, PTPRK, LAR, SHP-1 and PTP1B (41).
Interestingly, we also observed an increase in PLA signal between PTPRJ and EGFR in cells expressing the G983L PTPRJ point mutant, suggesting that destabilizing PTPRJ self-association promotes its access to EGFR. While proximity detected by in situ PLA does not provide evidence of direct EGFR-PTPRJ interaction, the observed enhanced association, in addition to the decreased EGFR phosphorylation, strongly argues for a specific and functional interaction. It is then compatible with our hypothesis that PTPRJ TM-mediated dimerization prevents access to its substrates. Furthermore, these results and interpretations are consistent with the physical interaction at the cell surface between EGFR and PTPRJ identified by Yarden and coworkers (10). The same study reported that, while EGFR undergoes endocytosis after interacting with PTPRJ at the plasma membrane, PTPRJ remains confined to the cell surface. It is then reasonable to hypothesize that EGFR endocytosis would be forestalled by enhanced dephosphorylation by and/or association with G983L PTPRJ. These effects on EGFR internalization are the focus of on-going investigations in our laboratories.
In the present study, with the overarching goal to develop novel PTPRJ agonist peptides, we also rationally designed and selected a peptide (RJbinder) capable of binding to the TM domain of PTPRJ and antagonizing its self-association. Remarkably, treating UMSCC2 cells expressing WT PTPRJ with RJbinder led to inhibition of phosphorylation, wound closure and PTPRJ self-association. Similar to the effect of TM point mutations, RJbinder 10 µM treatments did not completely suppress EGFR phosphorylation. It could be due to the same reasons discussed above for the effect of single-point mutations and for the fact that interfering with TM-TM interaction may not be sufficient to completely overcome receptor self-association (39). Because peptides are reconstituted in detergent micelles prior to incubation with cells, another explanation is that the population of inserted peptides is a mixture of both possible orientation (i.e., N-terminus or Cterminus out), which would contribute to a lack of efficacy against PTPRJ. It is also conceivable that the activity of RJbinder may be highly cell-context dependent as possible RTK substrates and PTPRJ itself are likely to be differentially expressed across cancer cell type. Its activity could also be affected by the high level of PTPRJ expression observed in the transfected UMSCC2 cells, against which the peptide has to compete. It is therefore informative to compare the activity of the RJbinder with gefitinib, a potent inhibitor of EGFR activity used in cancer treatment. Indeed, cell treatment with 15 µM of gefitinib (IC50 for UMSCC2 cells expressing WT PTPRJ; Figure  S7), suppresses wound closure by 'only' ~60% (Figure 9), showing the importance of the targeted cells. Moreover, RJbinder showed no activity against cells expressing either minimal level of PTPRJ or the homodimer-disrupting mutant G983L, and the control peptide with hindered interaction with PTPRJ (i.e., RJcontrol) showed no effect on receptor selfassociation, phosphorylation or wound closure. Altogether, these results indicate that RJbinder interacts specifically with the TM domain of PTPRJ.
It is also important to point out that RJbinder represents a proof-of-concept for a new allosteric and possibly orthogonal way to target the activity of RPTPs and receptor kinase phosphorylation. Indeed, while other methods have been devised to modulate RPTPs (e.g., small molecules targeting the phosphatase domain, wedge peptide mimetic and decoy protein), most, if not all, act by inhibiting their activity and none of them target the TM domains (14). Moreover, the mechanism of activation of RPTPs remain unclear and no common model has been proposed. As mentioned in the introduction, it is due, in part, to the lack of known natural ligands that are able to trigger RPTPs intracellular pathways by binding to the extracellular domain in the same manner than with RTKs, for example. For instance, only two biological ligands have been identified for PTPRJ: heparan sulfate proteoglycan syndecan-2 (42) and thrombospondin-1 (43), both of which are able to bind to the extracellular domain and stimulate PTPRJ activity. However, there is no structural model for their mechanism of action. Therefore, being able to activate PTPRJ allosterically using TM peptide binders can provide significant insight into how this receptor functions, interacts, and eventually be modulated.
Because of its tumor-suppressor role, there are clear advantages in promoting PTPRJ activity towards not only EGFR, but also other deregulated RTKs in cancer cells. While fully suppressing EGFR phosphorylation through disruption of PTPRJ TM-TM interaction remains to be explored, PTPRJ is known to suppress other growth factor receptors and their signaling proteins, including VEGF receptor 2 (44,45), hepatocyte growth factor receptor (46)(47)(48), FGF receptor (46), and PGDF receptor (PDGFR) (45), and the RAS-and FLT3-mediated signaling pathways by dephosphorylating ERK1/2 (29, 49). Therefore, disrupting PTPRJ self-association may inhibit other RTKs known to be involved in cancer cell progression, which may potentiate the effect of a peptide such as RJbinder.
It is also true that activating PTPRJ may have potential unintended consequences on cell survival. For instance, in addition to these antiproliferative and tumor suppressor functions, PTPRJ activity may be driving oncogenic signaling through positive regulation of VEGF-induced Src and Akt activation, as well as of endothelial cell survival (45). However, one would expect that the role of PTPRJ, the consequences of inhibiting/activating it, and the magnitude of these effects to be cell type-dependent. Improving the efficacy of RJbinder and quantitatively determining the cancer cell context-dependent outcomes of interfering with PTPRJ activity through TM mutations and TM-directed peptide binders is the focus of on-going efforts in our laboratories. Moreover, since RJbinder appears to act allosterically and by activating PTPRJ, it is tempting to speculate that such peptides could potentiate the activity of conventional small molecule inhibitors that often also have incomplete suppression of downstream signaling pathways or overcome acquired drug resistance when secondary mutations alter small molecule binding. Finally, we expect that the basic framework developed here can be extended to other RPTPs because modulation of RPTP activity through dimerization of TM domains appears to be a general theme throughout the family, and TM domains may provide orthogonality not possible by targeting either the extracellular or PTP domains. Therefore, this approach may afford a new modality to study RPTPs activity and represent a novel class of therapeutics against RTK-driven cancers.

Methods
Subcloning. Unless otherwise stated, standard molecular biology techniques were used, and all constructs used were verified by DNA sequencing (Genewiz, Inc.). The DNA sequence coding for the TM domains of interest were cloned into either pAraTMwt (coding for AraC) or pAraTMDN (coding for the inactive form of AraC, AraC*) plasmids. The protein sequence of human PTPRJ used in AraTM assays contains five extracellular residues, the TM domain, and twenty cytoplasmic juxtamembrane residues:

QDPGVICGAVFGCIFGALVIVTVGGFIFWRKKRKDAKNNEVSFSQIKP
Transmembrane cytoplasmic The sequence coding for full-length human PTPRJ (Entrez Gene ID 5795; PlasmID Repository) was subcloned into the pBABE-puro retroviral vector as an EcoRI/SalI fragment. All site-directed mutagenesis was performed using the QuikChange II Site-Directed Mutagenesis Kit (Agilent). Peptide/octylglucoside solutions were added to serum-starved cells to obtain a final detergent concentration (3 mM) lower than its critical micellar concentration to allow peptide partition into cell membrane (39). Following a 3 h treatment at 37 ºC, treatment media was removed and replaced with fresh serum-free media containing EGF (10 ng/mL), and the cells were incubated for another 10 min at 37 ºC. Cells were solubilized by the addition of cell extraction buffer supplemented with broad-spectrum phosphatase and protease inhibitors (Pierce #88667), then removed and analyzed by immunoblot. Scratch Assay. UMSCC2 cells were seeded in 6well plates at a cell density necessary to reach confluency after 24 h. Cells were incubated with serum-free media for 2 h, after which a scratch was made in the confluent cell monolayer with a 200 µL pipette tip to create a thin gap. Cells were immediately washed once with serum-free media and imaged using phase contrast microscopy (t = 0 h). The cells were incubated with serum-free media containing EGF (50 ng/mL) to stimulate scratch closure or medium lacking EGF (control) at 37 ºC for 6 h at which point phase contrast images were taken (t = 6 h) with a 10x objective using an Eclipse Ti-S microscope. For peptide treatments, RJbinder and RJcontrol were prepared as previously described for immunoblot analyses and incubated at 37 ºC for 1 h before the initial scratch. The media was then replaced with complete media containing EGF (50 ng/mL) at 37 ºC for 12 h. Scratch areas were quantified with ImageJ using the MRI Wound Healing Tool, and the closure percent was found by calculating the percent change in area between the initial and final time points. Percent scratch closure was normalized to either empty pBabe-puro vector (EV) or wild-type cells treated with media containing EGF (50 ng/mL).

DN-AraTM Dimerization
Proximity Ligation Assay. UMSCC2 cells were seeded on 22 mm x 22 mm coverslips in 6 well plates at 300,000 cells/well to be ~60% confluent after 16 h. For peptide treatments, RJbinder and RJcontrol were prepared as previously described for immunoblot analyses and incubated with the cells in complete media at 37 ºC for 1 h. Following 3-5 washes with PBS, the cells were fixed and permeabilized with ice cold methanol for 10 min at RT. The coverslips were blocked in 1% BSA in PBS at RT for 30 m and washed 3 times with PBS. The coverslips were incubated overnight in a humidified chamber at 4 ºC with primary antibodies (anti-PTPRJ mouse and anti-PTPRJ rabbit at 1:50 dilutions). Following the manufacturer protocol (Sigma-Aldrich; Duolink® #DUO92101), the coverslips were then washed with PBS. The PLUS and MINUS PLA probes were diluted in the Duolink® Antibody Diluent (1:5), and the coverslips were incubated in a humidity chamber for 1 h at 37 ºC. After washing with the supplied Buffer A, the coverslips were incubated with the provided ligase in a humidity chamber for 30 min at 37 ºC. After a second wash with Buffer A, the cells were incubated with the provided polymerase in amplification buffer in a humidity chamber for 100 min at 37 ºC. The coverslips were then washed stained with Nucspot® Live 488 (Biotium #40081) and mounted on slides with Fluoromount (Sigma-Aldrich #F4680) before being imaged by spinning disc confocal microscopy. The images were analyzed and quantified using BlobFinder (Center for Image Analysis).

Peptide TM Binder Library Construction.
To design a library of peptide sequences to be screened for their ability to compete for PTPRJ self-association using DN-AraTM, we first generated sequence alignments based on the wild-type PTPRJ TM sequence. From this alignment, variability at each aligned position within the TM sequence was determined; positions where more than different 4 amino acids occurred within the alignment were chosen for the library, excluding positions where conservative substitutions only occurred. We also retained key amino acids important for dimer formation, particularly the small-x3-small motif present in PTPRJ and focused on varying positions flanking these key residues; previous approaches have shown that residues flanking core small-x3small motif residues can modulate specificity of dimerization. Based on these results, a consensus peptide sequence was generated ( Figure 6) and a corresponding degenerate codon sequence was chosen to capture as much of the sequence diversity as possible while minimizing the number of different degenerate codons required. Similar degenerate codon sequences and approaches have been used previously in library design and selection of TM peptides (32,33). where, θobs is the observed ellipticity in millidegrees, l is the optical path length in centimeters, c is the final molar concentration of the peptides, and n is the number of amino acid residues. Raw data was acquired from 260 to 200 nm at 1 nm intervals with a 100 nm/min scan rate, and at least five scans were averaged for each sample. The spectrum of octylglucoside micelles or POPC liposomes was subtracted out from all construct samples.

Tryptophan Fluorescence Spectroscopy.
Fluorescence emission spectra were acquired with a Fluorolog-3 Spectrofluorometer (HORIBA). The excitation wavelength was set at 280 nm and the emission spectrum was measured from 300 to 450 nm. The excitation and emission slits were both set to 5 nm.
Peptide Incorporation into Cell Membrane by Fluorescence Microscopy. RJbinder fluorescently labeled with 5(6)-carboxyfluorescein was prepared in n-octylglucoside as previously described for immunoblot analyses. Briefly, UMSCC2 cells were seeded on 22 mm x 22 mm chambered coverslips pretreated with poly-l-lysine at 300,000 cells/well to be ~60% confluent after 16 h. The cells were then treated with 10 µM peptide in complete media at 37 ºC for 1 h. The cells were washed with PBS before being imaged by spinning disc confocal microscopy.   The relative (ratio of phosphorylated to total protein) intensities are shown as mean ± SEM (n = 3-5). Statistical significance between WT and mutants (for each phosphorylation site) was assessed using one-way ANOVA corrected for multiple comparison (Dunnet test) at 95% confidence intervals: ***p ≤0.001; **p ≤ 0.01; and *p ≤ 0.05. were serum-starved, scratched (0 h) and incubated with serum-free media with or without EGF (50 ng/mL) for 6 h. Relative closure was quantified by calculating the percent change in area between 0 and 6 h using ImageJ, and then normalized to EV (+ EGF). (B) Results are shown as mean ± SEM (n = 6-9). Statistical significance between WT and mutants was assessed using one-way ANOVA corrected for multiple comparison (Dunnet test) at 95% confidence intervals): ***p ≤ 0.001; and *p ≤ 0.05. Oligomerization of full-length PTPRJ WT and G983L expressed in UMSCC2 cells was measured by spinning disk confocal microscopy (green: nucleus; red: PLA for PTPRJ self-association). The average PLA signal intensity per cell was determined using BlobFinder. (B) Results are shown as mean ± SEM (n > 100). Statistical significance was assessed using unpaired t test (at 95% confidence intervals): **p ≤ 0.01.

Figure 5. Full-length PTPRJ point mutant promotes interaction with EGFR in cells.
Representative confocal microscopy images (A) and quantification (B) of the effect of transmembrane point mutation on PTPRJ interaction with EGFR using in situ PLA. Complex formation with EGFR was measured by spinning disk confocal microscopy in UMSCC2 cells expressing WT or G983L PTPRJ (green: nucleus; red: PLA between PTPRJ and EGFR). The average PLA signal intensity per cells was determined using BlobFinder. (B) Results are shown as mean ± SEM (n > 100). Statistical significance was assessed using unpaired t test (at 95% confidence intervals): ***p ≤ 0.001.  Oligomerization of full-length PTPRJ WT in UMSCC2 cells was measured following treatment with RJbinder or RJcontrol for 1 h using spinning disk confocal microscopy (green: nucleus; red: PLA). The average PLA signal intensity per cells was determined using BlobFinder. (B) Results are shown as mean ± SEM (n > 100). Statistical significance was assessed using one-way ANOVA corrected for multiple comparison (Dunnet test) at 95% confidence intervals): **p ≤ 0.01 and *p ≤ 0.05.

Figure 8. RJbinder inhibits the phosphorylation of EGFR. (A)
Serum-starved UMSCC2 cells were treated with RJbinder or RJcontrol for 3 h prior to stimulation with EGF. Cell lysates were probed for EGFR, phospho-EGFR (Y1068 and Y1173), and PTPRJ by immunoblot. (B) The relative (ratio of phosphorylated to total protein) intensities are shown as mean ± SEM (n = 3-5). Statistical significance was assessed using unpaired t test (at 95% confidence intervals): **p ≤ 0.01 and *p ≤ 0.05. Figure 9. RJbinder inhibits EGFR-driven cell migration. Representative phase contrast images with tracings to identify open scratch areas (A) and quantification (B) of the effect of RJbinder, RJcontrol, and gefitinib on wound closure. Serum-starved UMSCC2 cells were treated with 3 mM n-octylglucoside, 10 µM RJbinder , 10 µM RJcontrol, or 15 µM gefitinib (a known inhibitor of EGFR activity) for 1 h, scratched (0 h), and incubated in media with EGF (50 ng/mL) for 12 h. Relative closure was quantified by calculating the percent change in area between 0 and 12 h using ImageJ, and then normalized to media containing EGF. (B) Results are shown as mean ± SD (n = 6-9). Statistical significance between WT and mutants was assessed using one-way ANOVA corrected for multiple comparison (Dunnet test) at 95% confidence intervals): ***p ≤ 0.001; and *p ≤ 0.05.