Research ArticleBiochemistry

Phosphorylation of the phosphatase PTPROt at Tyr399 is a molecular switch that controls osteoclast activity and bone mass in vivo

See allHide authors and affiliations

Science Signaling  08 Jan 2019:
Vol. 12, Issue 563, eaau0240
DOI: 10.1126/scisignal.aau0240

A phosphoswitch for a phosphatase

The receptor-type tyrosine phosphatase PTPROt either stimulates or inhibits the kinase Src in different contexts. In osteoclasts, PTPROt stimulates Src activity, which promotes bone resorption. Roth et al. described mice lacking PTPROt entirely and mice lacking a putative phosphorylation site at the C terminus of PTPROt, Tyr399. Phenotypic analyses of the mice, combined with experiments in osteoclasts derived from them and in cultured cells, demonstrated that Tyr399 is a phosphoswitch that controls PTPROt activity toward Src. When Tyr399 was not phosphorylated, PTPROt dephosphorylated Src at an activating site, thus inhibiting Src activity. When PTPROt Tyr399 was phosphorylated, PTPROt recruited Src through the adaptor protein Grb2 and dephosphorylated an inhibitory site in Src, thereby activating the kinase. These findings explain how PTPROt both positively and negatively influences Src activity and suggest a potential switch function for C-terminal tyrosine residues in other tyrosine phosphatases.

Abstract

Bone resorption by osteoclasts is essential for bone homeostasis. The kinase Src promotes osteoclast activity and is activated in osteoclasts by the receptor-type tyrosine phosphatase PTPROt. In other contexts, however, PTPROt can inhibit Src activity. Through in vivo and in vitro experiments, we show that PTPROt is bifunctional and can dephosphorylate Src both at its inhibitory residue Tyr527 and its activating residue Tyr416. Whereas wild-type and PTPROt knockout mice exhibited similar bone masses, mice in which a putative C-terminal phosphorylation site, Tyr399, in endogenous PTPROt was replaced with phenylalanine had increased bone mass and reduced osteoclast activity. Osteoclasts from the knock-in mice also showed reduced Src activity. Experiments in cultured cells and in osteoclasts derived from both mouse strains demonstrated that the absence of phosphorylation at Tyr399 caused PTPROt to dephosphorylate Src at the activating site pTyr416. In contrast, phosphorylation of PTPROt at Tyr399 enabled PTPROt to recruit Src through Grb2 and to dephosphorylate Src at the inhibitory site Tyr527, thus stimulating Src activity. We conclude that reversible phosphorylation of PTPROt at Tyr399 is a molecular switch that selects between its opposing activities toward Src and maintains a coherent signaling output, and that blocking this phosphorylation event can induce physiological effects in vivo. Because most receptor-type tyrosine phosphatases contain potential phosphorylation sites at their C termini, we propose that preventing phosphorylation at these sites or its consequences may offer an alternative to inhibiting their catalytic activity to achieve therapeutic benefit.

INTRODUCTION

Osteoclasts (OCLs) are large, multinucleated phagocytic cells that degrade bone as part of homeostatic regulation of this tissue. OCLs are formed by fusion of precursor cells of the monocyte lineage in a process that is driven by the cytokines macrophage colony-stimulating factor (M-CSF) and receptor activator of nuclear factor κB ligand (RANKL) (1, 2). The mature cells adhere to bone tightly and secrete onto the bone surface proteases and protons that degrade the protein and mineral components, respectively, of the bone matrix (3, 4). OCLs function in close spatial and temporal proximity with bone-producing osteoblasts, and the balance between their opposing activities is critical for maintaining the mass and physical properties of bone. Aberrant activation of OCLs disrupts this balance as exemplified by osteoporosis, in which reduced production of estrogen after menopause increases OCL-mediated bone loss, leading to fractures and increased morbidity and mortality (5). Inhibiting the production or function of OCLs, such as by bisphosphonates or antibodies directed against RANKL (6), is an established strategy for treating bone diseases that are associated with increased OCL activity.

The production and activity of OCLs are heavily dependent on phosphotyrosine-based cell signaling. A prominent example illustrating this are mice lacking the tyrosine kinase Src, which exhibit greatly increased bone mass due to reduced OCL activity (7, 8). Protein tyrosine phosphatases (PTPs), which antagonize kinase activity and fulfill critical roles in the regulation of physiological processes in vivo (911), are also central to OCL function and bone homeostasis. The cytosolic PTP SHP-1 inhibits OCL production and function (12, 13), whereas the PTPs PTPε (1416), MKP-1 (also known as DUSP1) (17), CD45 (18), SHP-2 (19), and PTP-PEST (20) exert the opposite effect.

Another PTP that has been studied in the context of OCLs is PTPROt [also known as PTP-phi and PTP-oc (2123)], an orphan receptor-type PTP that is present mainly in hematopoietic cells. PTPROt is one of several proteins generated from alternately spliced transcripts of the Ptpro gene and contains a short extracellular domain, a single transmembrane domain, and a single cytosolic PTP catalytic domain. The Ptpro gene also encodes a second protein isoform, GLEPP-1, which is identical to PTPROt but contains additional extracellular sequences (24). Transcripts for PTPROt and GLEPP-1 are produced from different promoters of the Ptpro gene, and their expression patterns are distinct (25). Studies in which PTPROt was either knocked down or overexpressed in cultured cells collectively indicate that this PTP supports OCL activity (2628). In agreement, transgenic mice that overexpress PTPROt in their OCLs exhibit a male-specific decrease in bone mass (29). Molecularly, PTPROt has been reported to activate Src in these systems by dephosphorylating the kinase at Tyr527 (by chicken Src numbering) (2729), thus stimulating downstream signaling events such as signaling by integrin β3, the kinase Syk, and Jun N-terminal kinase and nuclear factor κB (30, 31).

PTPROt has also been reported to activate the Src family kinase Lyn in chronic lymphocytic leukemia B cells (32), and Src activity is enhanced in PTPROt-overexpressing hepatocellular carcinoma cells (33). However, PTPROt attenuates Src activity in colon cancer cells by dephosphorylating the kinase at Tyr416 (34). PTPROt also inactivates Lyn by dephosphorylating the analogous activating site Tyr397 in B cell lines (35), and phosphorylation of Lyn at this site is reduced in B cells of transgenic mice that overexpress PTPROt (36). These mixed results suggest that PTPROt is capable of both activating and inactivating Src, and that the choice between these activities is context dependent. Thus, a reversible regulatory mechanism likely exists to balance these opposing functions and to ensure a coherent signaling output.

To understand better the effects of PTPROt on Src and what determines whether this phosphatase stimulates or represses Src activity, we examined how PTPROt affects Src activity and OCL function in vivo using mice with modified PTPROt expression. These mice either completely lacked PTPROt or expressed a modified form of PTPROt from the endogenous locus, in which a putative C-terminal phosphorylation site, Tyr399, was replaced with phenylalanine (Y399F PTPROt). We show that PTPROt targeted both the inhibitory phosphorylated Tyr527 (pTyr527) and activating pTyr416 sites in Src and that phosphorylation of PTPROt at Tyr399 was a reversible molecular switch that selected between these possibilities. Accordingly, Y399F PTPROt mice exhibited reduced Src activity in OCLs, reduced OCL activity, and increased bone mass, whereas mice that completely lack PTPROt did not exhibit OCL or bone phenotypes. From a broader perspective, we propose that interfering with C-terminal phosphorylation of PTPs without affecting their catalytic activities can generate specific and focused physiological effects in vivo and may offer an alternative approach to inhibiting the catalytic activity of PTPs for therapeutic purposes.

RESULTS

Tagged forms of wild-type and mutant PTPROt are expressed similarly in mice

To examine the possible role of PTPROt in vivo in OCLs, we examined mice in which the distal P2 promoter of the Ptpro gene, which drives PTPROt production (25), was disrupted [ROKO (PTPROt knockout) mice] (32). Both PTPROt mRNA expression and production of PTPROt protein are abrogated in these mice, whereas expression of Glepp-1, the other product of the Ptpro gene, and production of GLEPP-1 protein is not affected (32). PTPROt, but not GLEPP-1, was present in OCLs from wild-type mice, and neither isoform was present in osteoblasts [Fig. 1A (32)]. PTPROt has a short C-terminal tail sequence that includes one tyrosine residue, Tyr399 (Fig. 1B). The analogous tyrosine residue (Tyr1220) in GLEPP-1 can be phosphorylated by Fyn and is required for this isoform to separately bind Grb2 and Fyn in human embryonic kidney (HEK) 293 cells (37). Moreover, a similarly placed tyrosine residue is required for the unrelated cyt-PTPε to activate Src in OCLs by dephosphorylating the kinase at its Tyr527 (14, 15), leading us to hypothesize that Tyr399 in PTPROt might participate in regulating the activity of this phosphatase toward Src. To enable testing this in vivo, we generated a second mouse model, Y399F-HA-PTPROt mice, in which the endogenous Ptpro gene was mutated to produce the mutant Y399F PTPROt protein that also carries an HA tag at its C terminus (Y399F-HA-PTPROt). In parallel, we generated mice whose endogenous PTPROt includes a similarly placed HA tag but retains Tyr399, for control purposes (WT-HA-PTPROt mice; Fig. 1B and fig. S1). OCLs from Y399F-HA-PTPROt and WT-HA-PTPROt mice produce similar amounts of HA-tagged PTPROt proteins (Fig. 1C).

Fig. 1 ROKO and Y399F-HA-PTPROt mice.

(A) Western blot (WB) showing PTPROt in osteoclasts and osteoblasts (Ob) of homozygous PTPROt knockout (ROKO, labeled RO) and wild-type (WT) mice, as well as in RAW 264.7 (Rw) cells and in HEK293 cells expressing exogenous PTPROt (293). The asterisk marks likely PTPROt alternative splicing products. The antibody against PTPROt cross-reacts with GLEPP-1 (32), and the arrow marks the expected location of the GLEPP-1 band. Molecular size markers are indicated in kilobases. Blotting for tubulin is a loading control. Blot is representative of n = 2 independent experiments using cells from different mice. (B) Schematic diagrams of WT-HA-PTPROt (WT-HA) and Y399F-HA-PTPROt (YF-HA) proteins. PTP, PTP catalytic domain; HA, C-terminal HA epitope tag. The arrow marks the position of the Y399F mutation. The sequence of the C terminus of PTPROt is also shown below the cartoon, with Tyr399 and Phe399 in bold. (C) Western blot showing endogenous PTPROt in OCLs of WT, ROKO (RO), WT-HA, and YF-HA mice. The HA tag slightly reduces the mobility of tagged PTPROt proteins. Blot is representative of n = 4 independent experiments using cells from different mice.

Mice expressing Y399F PTPROt have increased bone mass

To determine whether complete loss of PTPROt or expression of the Y399F mutant form affected bone structure in vivo, we isolated tibiae from mice of all four genotypes (ROKO, the genetic background–matched wild-type control for the ROKO line, Y399F-HA-PTPROt, and the matched WT-HA-PTPROt control). Micro–computerized tomography (μCT) analyses revealed that the trabecular bone structure of ROKO mice in both sexes did not differ from their sex- and age-matched controls (Fig. 2A and Table 1). The thickness of trabecular bone was reduced in female ROKO mice, but this result is likely not biologically meaningful because the other bone parameters in these mice were unaffected. In contrast, in Y399F-HA-PTPROt mice the bone volume fraction, number of trabeculae, and trabecular thickness were increased, whereas trabecular separation was decreased (Fig. 2A and Table 2). Similar results were obtained also when a second independent line of Y399F-HA-PTPROt mice was analyzed (table S1). We conclude that the presence of the Y399F mutation in PTPROt, but not the complete absence of this PTP, significantly increases bone mass in vivo.

Fig. 2 Impaired activity of Y399F-HA-PTPROt OCLs.

(A) Virtual μCT longitudinal sections of tibias from WT, ROKO, WT-HA-PTPROt (WT-HA), and Y399F-HA-PTPROt (YF-HA) 7-week-old female mice. All images are presented at the same magnification. Images are representative of n = 3 mice from each genotype. (B) Bar graph quantifying the surface area of bone in contact with OCLs (Oc.S/BS) from WT, ROKO, WT-HA, and YF-HA mice. n = 5 to 7 2-month-old mice per bar. Data shown are means ± SE; *P < 0.05 by two-way analysis of variance (ANOVA) with Bonferroni’s post hoc test. a.u., arbitrary units. (C) Bar graph quantifying the concentration of collagen telopeptides in circulation in WT, ROKO, WT-HA, and YF-HA mice. n = 10 to 11 2-month-old mice per bar. Data shown are means ± SE and were analyzed as in (B). (D) OCLs from WT-HA and YF-HA mice stained to show tartarate-resistant acid phosphatase (TRAP) activity, alongside quantification of OCL numbers (means ± SE) per well of a 24-well plate of each genotype, n = 5 mice per bar. (E) Pits indicating resorption (dark areas) in bovine bone that were generated by similar numbers of OCLs from WT-HA and YF-HA mice, alongside quantification of pit resorption by OCLs of both genotypes. Resorption was measured as percentage of bone area covered by pits of the total bone fragment surface area, mean ± SE. n = 7 (YF-HA) or 8 (WT-HA) mice per bar. **P = 0.011 by two-tailed Student’s t test. All scale bars, 500 μm.

Table 1 μCT analysis of tibiae from WT and ROKO mice.

Parameters shown are BV/TV (trabecular bone volume as percentage of total volume), Tb.Th (trabecular thickness), Tb.Sp (trabecular separation), and Tb.N (trabecular number). n = 8 mice of each sex per genotype, aged 2 months. Data are means ± SD and were analyzed by unpaired, two-tailed Student’s t test. NS, not significant.

View this table:
Table 2 μCT analysis of tibiae from WT-HA-PTPROt (WT-HA) and Y399F-HA-PTPROt (Y399F-HA) mice.

Parameters shown are BV/TV (trabecular bone volume as percentage of total volume), Tb.Th (trabecular thickness), Tb.Sp (trabecular separation), Tb.N (trabecular number), and tibial length. n = 6 mice of each sex per genotype, aged 2 months. Data are means ± SD and were analyzed by unpaired, two-tailed Student’s t test. Y399F-HA-PTPROt data are from mouse line 8; WT-HA-PTPROt data are combined from mouse lines 2 and 3. See table S1 for a similar comparison to Y399F-HA-PTPROt mouse line 16.

View this table:

OCLs expressing Y399F PTPROt have reduced bone resorption activity

To determine the cellular basis for the increased bone mass of Y399F-HA-PTPROt mice, we examined bone synthesis and degradation. Histomorphometry studies revealed significantly increased trabecular bone surface in contact with OCLs in Y399F-HA-PTPROt tibiae compared to matched WT-HA-PTPROt controls. In contrast, this parameter was similar in ROKO mice versus wild-type controls (Fig. 2B). Collagen telopeptide concentrations in serum, a clinical marker that generally correlates with the extent of bone resorption in vivo, were similar in ROKO and in Y399F-HA-PTPROt mice, each relative to their respective controls (Fig. 2C). These results suggest that OCLs from Y399F-HA-PTPROt mice, whose total contact area with bone is increased, are less active on a per-cell basis, whereas ROKO OCLs resorb bone normally. To examine the resorptive activity of OCLs, we produced OCLs in vitro by culturing bone marrow cells in the presence of M-CSF and RANKL. Cells from Y399F-HA-PTPROt and WT-HA-PTPROt mice grew well and produced similar numbers of mature OCLs (Fig. 2D). However, when plated on slices of bovine bone, Y399F-HA-PTPROt OCLs consistently resorbed less bone than did WT-HA-PTPROt controls (Fig. 2E), confirming that they were less active than control cells. In contrast, OCLs from ROKO and wild-type control mice grew and resorbed bone similarly (fig. S2, A and B). Because OCLs are cultured from bone marrow precursor cells that are grown and differentiated in vitro over a period of several days, this result indicates that the defect present in Y399F-HA-PTPROt OCLs is cell autonomous.

Further support for this latter conclusion was obtained from bone marrow transplantation studies, in which WT-HA-PTPROt mice were lethally irradiated and then rescued by injection of bone marrow cells from either Y399F-HA-PTPROt or WT-HA-PTPROt mice. Mice transplanted with Y399F bone marrow cells exhibited significantly increased bone volume fraction, as well as a trend for increased trabecular thickness relative to mice transplanted with WT-HA-PTPROt bone marrow cells (fig. S3 and Table 3). The reduced bone mass of mice shown in this experiment relative to mice shown in Table 2 may have been caused by the whole-body lethal irradiation procedure that the mice underwent during the transplantation process. These results indicate that the increased bone mass phenotype of Y399F-HA-PTPROt mice is due to effects that originate in the hematopoietic system, from which OCLs are derived. Studies in which Y399F PTPROt is expressed only in OCLs are required to examine whether hematopoietic cells from other lineages may contribute to the OCL and bone phenotypes of Y399F-HA PTPROt mice.

Table 3 Transplanting bone marrow from Y399F-HA PTPROt mice into WT-HA PTPROt mice increases bone mass.

Bone marrow cells from female WT-HA-PTPROt mice (WT-HA, controls) or Y399F-HA PTPROt mice (Y399F-HA) aged 5 weeks were implanted in lethally irradiated male WT-HA-PTPROt mice of the same age, and tibiae were analyzed 1 month later by μCT. Parameters shown are BV/TV (trabecular bone volume as percentage of total volume), Tb.Th (trabecular thickness), Tb.Sp (trabecular separation), and Tb.N (trabecular number). n = 7 or 8 acceptor mice in each category. Data are means ± SD and were analyzed by unpaired, two-tailed Student’s t test. The success of the transplantation procedure was evaluated by polymerase chain reaction (PCR) of DNA of macrophages isolated from the mice at sacrifice for the presence of DNA markers specific for the X and Y chromosomes (fig. S3).

View this table:

Additional histomorphometry studies indicated that the numbers of osteoblasts and osteocytes, bone area in contact with osteoblasts, and bone formation rates were unchanged in ROKO and in Y399F-HA PTPROt mice compared to the respective controls (Table 4). Complete loss of PTPROt or expression of the Y399F mutant form therefore does not affect osteoblasts and bone formation, in agreement with the absence of PTPROt in these cells (Fig. 1A).

Table 4 Bone formation parameters in femurs from WT versus ROKO and from WT-HA-PTPROt versus Y399F-HA-PTPROt mice.

Parameters shown are BFR/BS (bone formation rate normalized to bone surface), N.Ob/B.Pm (number of osteoblasts normalized to bone perimeter), Ob.S/BS (osteoblast surface normalized to bone surface), and N.Ot/B.Ar (number of osteocytes normalized to bone area). n = 7 or 8 male mice per genotype, aged 2 months. Data are means ± SD and were analyzed by two-tailed Student’s t test. NS, not significant.

View this table:

Y399F PTPROt reduces Src activation in OCLs

PTPROt has been reported in some cases to activate Src by dephosphorylating the inhibitory C-terminal Tyr527, and in other cases to attenuate Src activity by dephosphorylating the activating site Tyr416. To determine whether PTPROt can target both residues, we expressed PTPROt and Src in SYF cells, which are mouse fibroblasts cells that lack the Src family kinases Src, Yes, and Fyn and do not produce endogenous PTPROt. The presence of PTPROt decreased phosphorylation of Src at both Tyr527 and Tyr416, whereas an inactive, substrate-trapping mutant form of PTPROt, C325S PTPROt [fig. S4 (38)], did not (Fig. 3, A and B). We next examined how the absence of PTPROt or expression of the Y399F mutant affected Src in OCLs. OCLs from ROKO mice exhibited increased phosphorylation of Src at Tyr416 compared to OCLs from wild-type control mice (Fig. 3C), indicating that PTPROt targets Tyr416 and that other PTPs cannot compensate for its absence. In contrast, phosphorylation of Src at Tyr527 was increased in OCLs from Y399F-HA-PTPROt mice compared to OCLs from WT-HA-PTPROt mice (Fig. 3D), indicating that phosphorylation at Tyr399 is required for PTPROt to target pTyr527 of Src. In agreement with the known strong inhibitory effect of phosphorylation at Tyr527 on Src activity, Src kinase activity was reduced by 27% in Y399F-HA-PTPROt OCLs compared to WT-HA-PTPROt OCLs (Fig. 3E). Increased Tyr416 phosphorylation in ROKO OCLs did not, however, translate into increased Src activity (Fig. 3E), most likely because Src activation requires additional events, such as dephosphorylation of pTyr527, which do not occur in this context. These results indicate that the Y399F mutation preferentially reduces the ability of PTPROt to dephosphorylate Src at Tyr527, thereby upsetting the balance between the effects of this phosphatase on pTyr416 and pTyr527 of Src, leading to reduced Src kinase activity.

Fig. 3 PTPROt dephosphorylates Src at Tyr416 and Tyr527.

(A) FLAG-tagged WT PTPROt, the inactive, substrate-trapping mutant C325S PTPROt (CS), and Src were expressed in SYF cells as indicated. Cell lysates were subjected to FLAG immunoprecipitation (IP) and blotted (WB) for tyrosine phosphorylation (pTyr), PTPROt. Cell lysates were also blotted for Src phosphorylated at Tyr416, Src phosphorylated at Tyr527, and total Src. Blot is representative of n = 3 independent experiments. (B) Quantification of phosphorylation of Src at Tyr527 (pTyr527) and Tyr416 (pTyr416) in SYF cells expressing WT or CS PTPROt relative to cells not expressing PTPROt. Data are means ± SD from n = 5 (Tyr527) or n = 3 (Tyr416) independent experiments. *P ≤ 0.014 by one-way ANOVA with Tukey’s multiple comparisons test. (C) Phosphorylation of Src at Tyr527 and Tyr416 in OCLs from WT and ROKO mice as determined by protein blotting with phospho-specific antibodies alongside a representative blot showing Src phosphorylation. (D) Phosphorylation of Src at Tyr527 and Tyr416 in OCLs from WT-HA-PTPROt (WT-HA) and Y399F-HA-PTPROt (YF-HA) mice, alongside a representative blot showing Src phosphorylation. (E) Src kinase activity measured in OCLs from WT, ROKO, WT-HA, and YF-HA mice as indicated. Data in (C) to (E) are means ± SD obtained from OCLs from n = 7 to 9 male mice, aged 6 to 8 weeks, per bar. *P < 0.05, **P < 0.01 by Student’s t test.

PTPROt can undergo phosphorylation at Tyr399

Although the catalytically inactive mutant C325S PTPROt was phosphorylated when coexpressed with Src (Fig. 3A) in SYF cells, we were unable to detect tyrosine phosphorylation of wild-type PTPROt in the presence of Src (Fig. 3A). This suggested that wild-type PTPROt autodephosphorylates at Tyr399 (see below) and that use of catalytically inactive PTPROt molecules should stabilize their phosphorylation sufficiently to enable detection. We therefore used R331M PTPROt or C325S PTPROt and their corresponding Y399F mutants to further study phosphorylation of PTPROt (Fig. 4A). Of note, the Y399F mutation itself did not affect catalytic activity of wild-type PTPROt toward para-nitrophenyl-phosphate (PNPP, fig. S4). R331M PTPROt is virtually inactive [fig. S4 (38)] but, in contrast to C325S PTPROt, it does not have substrate-trapping abilities and does not bind substrate phosphotyrosine residues (38). As expected, R331M PTPROt and C325S PTPROt were strongly phosphorylated in the presence of Src in SYF cells (Fig. 4B). Phosphorylation was abolished when the Y399F mutation was added to R331M PTPROt or C325S PTPROt (Fig. 4B), confirming that Src can induce phosphorylation of PTPROt at this residue. Src coimmunoprecipitated with both wild-type PTPROt and Y399F PTPROt, as well as with R331M PTPROt and C325S PTPROt and with their respective Y399F mutants (Fig. 4B). Tyr399 is the major phosphorylation site in PTPROt because the Y399F mutation prevented detection of PTPROt phosphorylation by general antibodies against phosphotyrosine also when cells were treated with sodium pervanadate, an irreversible inhibitor of PTPs that induces massive tyrosine phosphorylation of cellular proteins (Fig. 4C).

Fig. 4 Phosphorylation and dephosphorylation of PTPROt at Tyr399.

(A) Schematic representations of the indicated WT and mutant-tagged PTPROt molecules. Arrows indicate the Y399F mutation in the C terminus and the R331M and C325S mutations, both of which are located in the PTP catalytic domain. C-terminal FLAG (F) or HA (HA) tags are indicated. (B) Western blot showing tyrosine phosphorylation of PTPROt and its association with Src in SYF cells expressing the indicated combinations of Src and tagged WT, R331M, or C325S PTPROt proteins. The WT PTPROt blot contains a sample from cells treated with 0.5 mM sodium pervanadate as a technical positive control for phosphorylation. Y, Tyr399 is intact; YF, Tyr399 is mutated to phenylalanine. Cell lysates were subjected to FLAG immunoprecipitation (IP) and blotted for phosphorylated tyrosine (pTyr), PTPROt, and Src. Input blots were probed for FLAG and Src. Blots are representative of three independent experiments per PTPROt construct. (C) Western blot showing tyrosine phosphorylation of C325S PTPROt (CS) or (C325S, Y399F) PTPROt (CSYF) in SYF cells in response to a 10-min treatment with 0.5 mM sodium pervanadate (PV), as indicated. Blot is representative of three independent experiments. (D) Western blot showing trans-dephosphorylation of inactive, HA-tagged WT PTPROt at pTyr399 at the indicated time points after 30 and 60 min of exposure to active, FLAG-tagged PTPROt (WT). Negative controls include replacement of WT PTPROt with the inactive C325S mutant (CS), or the addition of 0.5 mM sodium pervanadate (WT + V). Blot is representative of three independent experiments. (E) Quantification of PTPROt trans-dephosphorylation at Tyr399. Data are means ± SD from n = 3 independent experiments per bar. *P < 0.05, **P < 0.01 versus WT at time 0 by one-way ANOVA with Dunnett’s multiple comparison test.

We next examined whether PTPROt does undergo autodephosphorylation at Tyr399. The difficulty in obtaining active PTPROt that is also phosphorylated at Tyr399 led us to examine whether active PTPROt molecules can trans-dephosphorylate inactive PTPROt molecules that are phosphorylated at Tyr399. We therefore purified inactive, HA-tagged, wild-type PTPROt (HA-PTPROt) from cells treated with the PTP inhibitor sodium pervanadate, which in this context resulted in increased phosphorylation at Tyr399 (Fig. 4C). We purified active, FLAG-tagged wild-type PTPROt from separate cells that were not treated with sodium pervanadate. Incubation of inactive, phosphorylated PTPROt with active PTPROt resulted in a time-dependent decrease in phosphorylation of the former (Fig. 4, D and E). Phosphorylation of HA-PTPROt was not reduced in control experiments to which sodium pervanadate was added or in which active FLAG-PTPROt was replaced by FLAG-tagged C325S PTPROt, confirming that dephosphorylation of inactive HA-PTPROt was performed in trans by active FLAG-PTPROt molecules. We hypothesize that individual PTPROt molecules can dephosphorylate their own pTyr399 residues in cis, although proof of this requires additional studies.

Phosphorylation of PTPROt at Tyr399 creates docking sites for Grb2 and other SH2 domain–containing proteins

How does the Y399F mutation affect the function of PTPROt? The presence of this mutation in PTPROt did not affect its catalytic activity toward PNPP (fig. S4), suggesting that the Y399F mutation did not affect the specific activity of PTPROt toward its substrates. Alternatively, phosphorylation at Tyr399 could affect the cellular role of PTPROt by creating docking sites for other molecules, in particular those containing SH2 domains. To examine this possibility, we used a synthetic peptide containing the 19 C-terminal amino acid residues of PTPROt, including pTyr399, to screen an array on which glutathione S-transferase (GST) fusion proteins of 89 distinct SH2 domains had been spotted (Fig. 5A and fig. S5, A to E)). In this qualitative assay SH2 domains of Abl1, Abl2, FES, GRAP, GADS, Grb2, TNS4, PIK3R1, PIK3R2, and PIK3R3 bound the phosphorylated peptide but not its nonphosphorylated form (Fig. 5A and fig. S5, A and B), indicating that binding depended on Tyr399 phosphorylation and was direct. Of note, the pTyr399 peptide did not bind the SH2 domains of several Src family kinases, including Src and Fyn, that were included in the array (Fig. 5A). We conclude that in our system, pTyr399 does not bind directly to the SH2 domains of the Src family kinases.

Fig. 5 Grb2 binds PTPROt at pTyr399.

(A) Hybridization of a PTPROt peptide containing pTyr399 with a panel of 89 SH2 domains, each fused to GST. The sequence of the peptide is shown with pTyr399 highlighted in bold and bracketed. The ellipses labeled A to J indicate pairs of dots, with each pair representing binding to a specific SH2 domain from the proteins indicated at the right of the panel. Arrowheads and asterisks mark the positions of the SH2 domains of Src and Fyn, respectively, which did not bind the peptide. The complete peptide array analysis is shown in fig. S5. (B) Western blot (WB) showing PTPROt pulled down from lysates of pervanadate-treated OCLs from WT-HA-PTPROt (WT-HA) or Y399F-HA-PTPROt (YF-HA) mice using GST or GST fused to the SH2 domains of Grb2 or Src. Blot is representative of three independent experiments. (C) Western blots showing PTPROt pulled down from OCLs from WT-HA and YF-HA mice using GST or GST fused to full-length Grb2 (Grb2) or to the SH2, N-terminal SH3 (N-SH3), or C-terminal SH3 (C-SH3) domains of Grb2 as indicated. Blot is representative of two independent experiments.

The importance of the cyt-PTPε–Grb2 association in OCLs (15) and the requirement for phosphorylation of Tyr1220 in GLEPP-1 for its association with Grb2 (35) led us to examine the possible interaction between Grb2 and PTPROt. In agreement with the array results, the Grb2 SH2 domain pulled down endogenous wild-type PTPROt from extracts of pervanadate-treated OCLs (Fig. 5B). The Grb2 SH2 domain did not pull down Y399F PTPROt from extracts of OCLs from Y399F-HA-PTPROt mice, confirming that this association depends on pTyr399. The SH2 domain of Src did not pull down either form of PTPROt (Fig. 5B), in agreement with its inability to bind the C-terminal peptide of PTPROt (Fig. 5A). Further studies revealed that pull-down is mediated by the Grb2 SH2 domain alone because full-length Grb2 or its isolated SH2 domain bound PTPROt in a Tyr399-dependent manner, whereas neither one of the isolated SH3 domains of Grb2 did so (Fig. 5C). We conclude that the PTPROt-Grb2 interaction is mediated exclusively by the SH2 domain of Grb2 binding to pTyr399 of PTPROt.

PTPROt pTyr399 promotes recruitment of Src through Grb2

To examine how the PTPROt-Grb2 association affects Src, we cotransfected HEK293 cells with constructs encoding Src and either wild-type or the inactive R331M form of FLAG-tagged PTPROt. Src and endogenous Grb2 coimmunoprecipitated with both forms of PTPROt in a specific manner, confirming that they interacted with PTPROt in intact cells (Fig. 6A). R331M PTPROt also coimmunoprecipitated with exogenously expressed, HA-tagged Grb2 (fig. S6). We then examined the role of Tyr399 in this association using FLAG-tagged forms of R331M PTPROt and the R331M,Y399F PTPROt double mutant. Src and Grb2 immunoprecipitated with R331M PTPROt (Fig. 6B, 1st IP); Grb2 did not coprecipitate with R331M,Y399F PTPROt, confirming that the PTPROt-Grb2 interaction required pTyr399 in intact cells as well as in binding assays in vitro. Src coimmunoprecipitated with both PTPROt mutants, indicating that both molecules can interact also in the absence of the Tyr399 phosphorylation site. However, as seen before (Fig. 4B), less Src coprecipitated with the R331M,Y399F mutant than with R331M PTPROt (Fig. 6, B and C), indicating that phosphorylation of PTPROt at Tyr399 promotes recruitment of additional Src molecules to PTPROt.

Fig. 6 Influence of pTyr399 on association of PTPROt with Grb2 and Src.

(A) Western blot showing coprecipitation of transgenically expressed Src and endogenous Grb2 with FLAG-tagged PTPROt from lysates of HEK293 cells expressing WT or R331M PTPROt. Blots are representative of three independent experiments. (B) Western blots showing coprecipitation of transgenically expressed Src and endogenous Grb2 with R331M PTPROt (RM) and (R331M Y399F) PTPROt (RMYF) in HEK293 cells (1st IP). Precipitated material was eluted from the beads and then precipitated with Src antibodies (2nd IP). Blots are representative of three independent experiments. (C) Bar graph showing the amount of Src that coprecipitates with RM PTPROt and RMYF PTPROt in the 1st IP of (B). Data are means ± SD from n = 3 independent experiments per bar, P value by paired, two-tailed t test. (D) PLA performed on OCLs from WT-HA-PTPROt (WT-HA) and Y399F-HA-PTPROt (YF-HA) mice. Red signals indicate interaction between antibodies against HA and Src. Staining of OCLs from WT mice, which only produce nontagged endogenous PTPROt, is a negative control. Cells are also stained for actin (green), which marks the podosomal array (the sealing zone–like structure) at the cell periphery, and for DNA [4′,6-diamidino-2-phenylindole (DAPI), blue] to mark nuclei. The boxed areas in the top row of images are magnified below. (E) Quantification of PLA signals in OCLs from WT-HA and YF-HA mice. The number of red PLA signals in individual OCLs was divided by the number of nuclei in each OCL, which is proportional to OCL size. Data (means ± SD) are shown normalized to the average values of WT-HA cells. Each bar represents data from n = 8 cells from nonoverlapping fields obtained in two independent experiments. Statistical analysis was performed by two-way ANOVA on original log-transformed data, accounting for batch and treatment effects. (F) Single-antibody PLA (negative control) performed on OCLs from WT-HA mice with primary antibodies against either HA or Src as indicated, and with both secondary antibodies. Actin and DNA are stained as in (D). All scale bars, 20 μm.

We then examined whether the interaction between endogenous PTPROt and Src could be detected in OCLs. Immunofluorescence staining indicated that both proteins were present throughout the cell periphery (fig. S7, A and B). Proximity ligation assay (PLA) studies indicated that wild-type PTPROt and Src interacted in OCLs produced from WT-HA-PTPROt mice, as visualized by the red signals generated by the proximity-derived probes (Fig. 6D). Significantly fewer red PLA signals were detected when Y399F-HA-PTPROt OCLs were examined (Fig. 6, D and E), in agreement with the contribution of Tyr399 to the PTPROt-Src interaction. The PLA signals were specific and were not detected in OCLs from wild-type mice, in which PTPROt is present but is not HA-tagged, or when one of the two primary antibodies was omitted (Fig. 6, D and F). The PLA assays indicate that endogenous wild-type PTPROt interacts with Src in primary OCLs and that pTyr399 promotes this interaction, as was observed also in transfected cells.

The presence of both Src and Grb2 in PTPROt immunoprecipitates is consistent with the existence of a PTPROt-Grb2-Src complex and supports a role for Grb2 in regulation of Src by PTPROt. Alternatively, Src and Grb2 could each interact with distinct subpopulations of PTPROt. To differentiate between these possibilities, we examined whether Grb2 was present in specific complexes that contained both PTPROt and Src. We eluted the immunoprecipitated PTPROt from the beads after the first PTPROt immunoprecipitation step and then precipitated the eluate a second time with antibodies directed against Src. The precipitated material from this step, which contained both PTPROt and Src, also contained Grb2, confirming the existence of a common PTPROt-Grb2-Src complex (Fig. 6B, 2nd IP). The large reduction in PTPROt amounts present in the second, Src precipitate indicates that most PTPROt molecules do not participate in the PTPROt-Src-Grb2 complex or bind Src. Collectively, these results indicate that PTPROt can bind Src independently of its phosphorylation at Tyr399 or of its ability to bind Grb2. However, phosphorylation at Tyr399 and the resulting association with Grb2 enables PTPROt to recruit additional Src molecules and to activate them.

DISCUSSION

PTPROt has been described variably as an activator or as an inhibitor of Src and Src-related kinases in various physiological systems. We show here that PTPROt can perform both activities because it can target both the inhibitory pTyr527 and the activating pTyr416 sites of Src. PTPROt thus joins a small group of PTPs, such as CD45 (39, 40), that can fulfill both positive and negative Src regulatory functions. We also show that phosphorylation of PTPROt at its C-terminal Tyr399 functions as a reversible molecular switch between these opposing functions.

In the context of Src, the major effect of the Y399F mutation in PTPROt in OCLs is to reduce the ability of the PTP to dephosphorylate Src at Tyr527. As a result, OCLs from Y399F-HA-PTPROt mice exhibit increased Src phosphorylation at Tyr527 and reduced Src kinase activity. In a manner consistent with the central role Src plays in OCLs (7, 41), these mice also exhibit reduced OCL-mediated bone resorption and consequently increased bone mass. In contrast, normal Src activity is maintained in OCLs from PTPROt knockout (ROKO) mice, indicating that contrary to the Y399F point mutation, complete loss of PTPROt does not upset the balance between activation versus inactivation of Src in these cells.

Phosphorylation of PTPROt at Tyr399 helps to activate Src, most likely by recruiting Grb2 (Fig. 7). Because this activates Src, we conclude that the Src molecules recruited through the phosphotyrosine-Grb2 mechanism are predominantly inactive. Because Src is among the kinases that can phosphorylate PTPROt at Tyr399 (Figs. 3 and 4), PTPROt may function in a positive regulatory loop in which Src molecules that were activated by other means induce the phosphatase to activate additional Src molecules. The potent autodephosphorylating activity of PTPROt at Tyr399, either in trans or in cis, offers an attractive mechanism for negatively regulating and limiting this loop. We note that Grb2 plays a major role in the generation of OCLs (14); at least part of this role is due to its ability to recruit Src for activation by several PTPs, thereby ensuring activation of this key kinase. Nonetheless, the phenotypes observed in Y399F-HA-PTPROt mice and OCLs indicate that PTPROt fulfills a nonredundant role in this context.

Fig. 7 Model depicting the dual roles of PTPROt toward Src.

Active Src, which is phosphorylated at pTyr416, physically associates with PTPROt, which dephosphorylates pTyr416. The mode of interaction between Src with PTPROt in this case is not known, but the end result is a reduction in Src activity. When PTPROt is phosphorylated at Tyr399 by active Src molecules (and possibly by other kinases), the SH2 domain of Grb2 binds PTPROt at pTyr399, and Grb2 recruits Src to the phosphatase. PTPROt then dephosphorylates Src at the inactivating site pTyr527, thus activating the kinase. PTPROt autodephosphorylation (dashed line) in cis or trans helps to terminate its Src-stimulating activity.

Although phosphorylation of PTPROt at Tyr399 enables it to recruit and activate some Src molecules, other Src molecules physically associate with PTPROt independently of pTyr399 or of Grb2. The importance of pTyr399 specifically for Src activation leads us to hypothesize that this second population of Src molecules is already active; PTPROt would then dephosphorylate these Src molecules at Tyr416 and attenuate their activity. This interpretation is consistent with the significant increase in pTyr416 Src observed in OCLs from ROKO mice but not in OCLs from Y399F-HA-PTPROt mice (Fig. 3C). We also note that the model presented here does not rule out the recruitment of active Src molecules by the pTyr399-Grb2 axis. However, the data indicate that this axis is required in the case of Src activation, suggesting a more complex function of Grb2 toward inactive Src molecules. Inactive, pTyr527 Src assumes a “closed” conformation, in which pTyr527 is bound by the Src SH2 domain and is not easily accessible (42, 43). It is possible that the sequences in Src that bind PTPROt directly are inaccessible when Src is in its closed conformation and that Grb2 is required in this case to physically recruit Src to PTPROt by binding to other Src sequences. Alternatively, binding of Grb2 to Src may relax its closed conformation sufficiently to provide PTPROt with access to pTyr527, thereby facilitating activation of Src. A mechanism in which pTyr399 displaces pTyr527 of Src from its association with the Src SH2 domain by directly binding this domain, similar to what has been suggested for the unrelated phosphatase PTPRA (44) and for overexpressed GLEPP-1 in HEK293 cells (37), most likely does not function here because pTyr399 does not bind the Src SH2 domain directly (Fig. 5A). Although this study focuses on Src, PTPROt may also regulate the activity of additional Src family members in OCLs in a similar manner.

From a broader perspective, the importance of pTyr399 in controlling the function of PTPROt toward Src highlights the regulatory roles of the C termini of receptor-type PTPs (45) and illustrates that targeting interactions mediated by these sequences can produce discrete physiological effects in vivo. Despite anecdotal success (46), most efforts to develop active site–specific PTP inhibitors for therapeutic use have not been successful because of their inadequate substrate specificity and their charged or polar nature, which reduces cell permeability. Current efforts attempt to bypass these difficulties (4749), although they still aim to completely inhibit PTP activity. Our results suggest a conceptually distinct approach, namely, to affect the physiological role of a PTP in vivo by targeting C terminus–mediated interactions while retaining catalytic activity. The example of PTPROt presented here in the context of live mice indicates that disrupting such interactions can produce discrete molecular effects and specific physiological outcomes. In the present case, these outcomes may have beneficial effects in the context of bone loss that is associated with, for example, osteoporosis or metastatic bone disease. Similar outcomes might not be obtained when broader interference is used, such as when the catalytic activity of a PTP is inhibited or, as shown here for PTPROt, when its expression is abolished. Disruption of C terminus–mediated interactions can also theoretically allow activation of a PTP (such as a tumor-suppressor PTP) by preventing an inhibitory interaction. C-terminal phosphorylation that leads to physiological outcomes in cultured cells has been described for a small number of PTPs, including PTPRA (5054), PTPRE (14, 15, 5558), and PTPRJ (59, 60). Two-thirds of receptor-type PTPs have C-terminal tyrosine residues (45), illustrating the potential of C-terminal phosphorylation as a more general regulatory mechanism for PTPs and suggesting that these residues may be targeted in vivo for therapeutic benefit.

MATERIALS AND METHODS

Antibodies and complementary DNA constructs

The rabbit polyclonal antibody that recognizes both PTPROt and GLEPP-1 has been described (32). Also used were mouse monoclonal antibodies against phosphotyrosine (PY99, Santa Cruz Biotechnology and PY20, BD Transduction Laboratories); HA [clone HA-7, coupled to beads (immunoprecipitation), Sigma-Aldrich and clone HA.11 (blotting and PLA), BioLegend]; actin (clone AC-40, Sigma-Aldrich); tubulin (clone DMIA, Sigma-Aldrich); and FLAG (clone M2, Sigma-Aldrich). Rabbit polyclonal antibodies were used against Src, pTyr416-Src, and pTyr527-Src (Cell Signaling Technology) and Grb2 (S-255, Santa Cruz Biotechnology). A rabbit monoclonal antibody that recognizes Src (Clone 32G6, Cell Signaling Technology) was used for PLA.

Wild-type PTPROt was cloned into the pcDNA3 expression vector (Invitrogen/Thermo Fisher Scientific). C-terminal HA or FLAG tags, as well as the Y399F, C325S, and/or R331M mutations, were introduced into the mouse Ptprot complementary DNA using PCR-directed mutagenesis followed by sequencing.

Genetically manipulated mice

PTPROt-deficient mice were generated as described (32) in the 129SvEv genetic background. Y399F-HA-PTPROt and WT-HA-PTPROt mice were generated by CRISPR technology in the C57BL/6 genetic background. The Y399F mutation and addition of the HA tag immediately upstream to the termination codon of PTPROt were inserted into exon 26 of the Ptpro gene. Candidate single guide RNA (sgRNA) sequences were selected from PTPROt genomic sequences using sgRNA Scorer [https://crispr.med.harvard.edu/sgRNAScorer/ (61)] and the CRISPR design tool of the Zhang laboratory, MIT (http://crispr.mit.edu/). The sequence 5′- GCAAATCCTAGTTCGGAATC-3′ (coding strand) was used. Single-stranded DNA oligomers used for mutagenesis were as follows: Y399F-HA: AGTGTGTGCAGCTGATGTGGCTGAGGAAGAAGCAACAGTTCTGCATCAGCGACGTCATCTTCGAGAACGTCAGCAAATCCTACCCATACGATGTTCCAGATTACGCTTAGTTCGGAATCTGGAGCGGTGAGTAGAGAGGATCTTAGCATAAGGGGACTTGTGTAGAC. WT-HA: CTGAGGAAGAAGCAACAGTTCTGCATCAGCGACGTCATCTACGAGAACGTCAGCAAATCCTACCCATACGATGTTCCAGATTACGCTTAGTTCGGAATCTGGAGCGGTGAGTAGAGAGGATCTTAGCATAAGGGGACTTGTGTAGAC. The A > T mutation that induces Y399F and the HA tag in both oligos are underlined; the termination codon is in bold. Single-stranded donor oligonucleotides were synthesized by Integrated DNA Technologies and were used without further purification.

sgRNA (50 μg/ml), CAS9 mRNA (100 μg/ml), and donor DNA oligonucleotides for homologous recombination (200 μg/ml) were injected into the cytoplasm of fertilized oocytes of C57BL/6JOlaHsd mice (Harlan-Envigo) or B6(Cg)-Tyrc-2J/J (albino C57BL/6 J; Jackson Laboratories) mice. Injected oocytes were implanted in pseudopregnant female hosts; pups were genotyped by tail biopsy PCR as detailed below. Three Y399F-HA-PTPROt founder mice and two WT-HA-PTPROt founders were mated separately with wild-type 129SvEv strain mice. F1 offspring were mated again to wild-type 129SvEv mice, and F2 mice from each line separately were intercrossed. Genomic DNA from homozygous mice was sequenced to verify presence of the HA tag and the Y399F mutation (fig. S1). Potential off-target sites for the sgRNA were selected using the CRISPR design tools of the Zhang laboratory, MIT (http://crispr.mit.edu/ or https://benchling.com/crispr). The five off-target sites that scored highest were selected [sequences: GAAAATCCTTGTTCTGAATC (Chr. 6), TTAAATTCTAGTTTGGAATCAGG (Chr. 1), TGAAATTCCAGTTCGGAATCTGG (Chr. 3), GATAATGCTTGTTCGGAATCCAG (Chr. 5), and GCCATTCACAGTTCGGAATCCAG (Chr. 3)]. DNA segments of 450 to 500 base pairs (bp) in length centered around these sequences were amplified by PCR and sequenced; no mutations were detected.

Studies were performed using mice from two independent Y399F-HA-PTPROt mouse lines (lines 8 and 16; referred to as Y399F-HA-PTPROt mice) and two independent WT-HA-PTPROt mouse lines (lines 2 and 3; referred to as WT-HA-PTPROt mice). The notations ROKO and wild-type (WT in figures) are used to denote PTPROt knockout mice and their matched wild-type controls, respectively. In all experiments, Y399F-HA-PTPROt mice were compared with mice from WT-HA-PTPROt mice of the same genetic background. Similarly, ROKO mice were compared with wild-type mice in the 129SvEv background. This rule was also followed in studies of primary cells from these mice. Mice were housed in a barrier facility kept at 22° ± 2°C on a light/dark cycle of 12 hours:12 hours, with food and water ad libitum. All mouse studies were approved by the Weizmann Institute Institutional Animal Care and Use Committee (IACUC) and were performed in accordance with Israeli law.

Genotyping of ROKO mice has been described (32). WT-HA-PTPROt and Y399F-HA-PTPROt mice were genotyped by PCR as follows: forward primer, CATCAGCGACGTCATCTACGA; reverse primer, TCCAGCTGGCTCACATACTTG. PCR amplification protocol was as follows: 94°C for 5 min, followed by 30 cycles of (94°C for 30 s, 60°C for 30 s, and 72°C for 30 s) and terminal extension at 72°C for 5 min. Product size is 175 bp (WT, no HA tag) or 201 bp (both WT-HA-PTPROt and Y399F-HA-PTPROt). The Y399F mutation introduces a silent diagnostic TaqI restriction enzyme site in the PCR product.

Micro–computerized tomography

Tibiae were collected from mice, fixed in 4% formalin for 48 hours at room temperature (RT), and stored at RT in 70% ethanol. μCT scans (300 projections over 180°) were performed using the Zeiss-Xradia μCT system [MicroXCT-400 (Zeiss x-ray microscopy)], using 40 kV and 200 μA and reaching a voxel size of 4 μm. After three-dimensional reconstruction, trabecular bone microstructure was quantitatively determined on a volume of 1.8 mm3 of cancellous bone starting 0.3 mm distal to the proximal tibial growth plate using AVIZO 9.0 3D visualization and analysis software (Thermo Fisher Scientific–FEI).

Histomorphometry

For osteoblast, osteocyte, and bone formation rate measurements: Mice were injected intraperitoneally with calcein (10 mg/ml in a solution of 0.15 M NaCl/2% NaHCO3; 30 mg/kg body weight) on two occasions 7 days apart and sacrificed 2 days after the second injection. Bones were dissected and fixed in 4% paraformaldehyde for 3 days and then stored in 70% ethanol. The bones were dehydrated in increasing concentrations of ethanol (70 to 100%), incubated in infiltration solution [methyl-methacrylate (Merck), supplemented with 0.66% benzoylperoxide (Merck) and 10% nonylphenyl-polyethyleneglycol acetate (Sigma-Aldrich)] for 14 days, and then embedded in infiltration solution supplemented with 0.5% N-N-dimethyl-d toluidine (Merck). Sections (5 μm) were cut and deplasticized in 2-methoxyethyl acetate (Merck). For osteoblast counting, sections were stained with toluidine blue and osteoblast and osteocyte parameters were measured in femoral trabecular bone starting 100 μm distal to the proximal growth plate. Bone formation rate was established by measuring the calcein labels in cortical bone, starting 100 μm distal to the growth plate. All measurements were performed with the OsteoMeasure v3.0.0.0. program (OsteoMetrics). For OCL surface measurements, tibiae were fixed in 3% paraformaldehyde in phosphate-buffered saline (PBS) for 24 hours. After decalcification, samples were embedded in paraffin, sectioned at 5 μm, and mounted on slides. Samples were stained for TRAP activity using a leukocyte acid phosphatase kit (Sigma-Aldrich) and counterstained with hematoxylin and eosin. The perimeter of trabecular bone that was in contact with OCLs, as well as the total perimeter of trabecular bone visible, was measured in four to five nonoverlapping fields of trabecular bone for each sample.

Collagen telopeptide determination

Serum was prepared from retro-orbital bleeds of mice aged 2 months. Care was taken to use mice of the same age consistently because at young ages telopeptide abundances in serum change significantly with age. Samples were analyzed by enzyme-linked immunosorbent assay using the RATLAPS kit (Immunodiagnostic Systems) according to the manufacturer’s instructions.

OCL preparation and cell culture

OCLs were differentiated from mouse primary tibial and femoral bone marrow cells as described (14). Cells were grown in OCL medium [α-minimum essential medium (Sigma-Aldrich) containing 10% fetal calf serum (Biological Industries), 2 mM glutamine, penicillin (50 U/ml), and streptomycin (50 μg/ml)] supplemented with M-CSF (20 ng/ml; Peprotech) and RANKL (20 ng/ml; R&D Systems). SYF (62) and HEK293 cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum (Invitrogen), 1 mM sodium pyruvate, and glutamine and antibiotics as indicated above for OCL medium. Cells were transfected [SYF cells: Jet-PEI (Polyplus Transfection) and HEK293 cells: calcium phosphate method (63)] and used 2 to 3 days afterward.

PLA and immunofluorescence staining

One million bone marrow cells were cultured per glass coverslip and induced to differentiate with M-CSF and RANKL. After differentiation, OCLs were fixed in 4% paraformaldehyde, blocked with 3% bovine serum albumin (BSA) supplemented with 0.05% Triton X-100. Cells were exposed to primary antibodies (see “Antibodies and complementary DNA constructs” section) against HA (1:500) and Src (1:100) for 2 hours and washed with PBS three times. For PLA, cells were processed with the Duolink in situ PLA detection kit (Sigma-Aldrich), according to the manufacturer’s instructions. After exposure to primary antibodies and washes, cells were hybridized with secondary antibodies against Rabbit PLUS (DUO92002) and against Mouse MINUS (DUO92004). Negative control experiments included both secondary antibodies and a single primary antibody. In situ ligation followed by amplification of tetramethylrhodamine-5-isothiocyanate probe were then performed. For non-PLA immunofluorescence staining, cells were incubated with primary antibodies as above, followed by 1-hour incubation at RT with Rhodamine Red-X (RRX)–labeled secondary antibodies against mouse or rabbit immunoglobulins (Jackson ImmunoResearch Laboratories). All cells were also stained for actin (fluorescein isothiocyanate–conjugated phalloidin) and DNA (DAPI). The red signals representing PLA interactions in representative OCLs were counted and normalized to the number of nuclei present in the same OCL, which is proportional to OCL size.

Bone marrow transplantation

Five-week-old male WT-HA-PTPROt host mice were lethally irradiated by two doses of 5.5 Gy each at an interval of 5 hours. The next day, 8 × 106 freshly isolated bone marrow cells from 5-week-old female donor mice, WT-HA-PTPROt or Y399F-HA-PTPROt, were injected in a volume of 0.25-ml PBS through the tail vein into each host mouse. Transplanted host mice were sacrificed 1 month after transplant and tibiae analyzed by μCT. The degree of post-transplant engraftment was determined by PCR as described (64). In brief, we cultured femoral bone marrow cells collected at sacrifice for 2 days in OCL medium with M-CSF, isolated genomic DNA, and amplified fragments of the X-linked Jarid1c and the Y-lined Jarid1d genes. In all cases, the intensity of the Y-specific band did not exceed 10% of the X-specific band, indicating presence of very low numbers of remaining male host cells (e.g., fig. S3). For PCR, the following oligomers were used: CTGAAGCTTTTGGCTTTGAG (forward) and CCACTGCCAAATTCTTTGG (reverse). Note that the Jarid1d product is 19 bp shorter than the Jarid1c product. Reactions were initiated by 5-min incubation at 94°C, followed by 35 cycles of (94°C for 20 s, 54°C for 1 min, and 72°C for 40 s), and a final elongation step of 72°C for 10 min.

Pit resorption assays

Bovine bone slices were sterilized with 70% ethanol and washed with PBS. Before use, slices were incubated with OCL medium (without cytokines) at 37°C for 1 hour. Bone marrow cells (1 × 106) from mice 4 to 6 weeks old were seeded in a well of a 24-well plate containing a bone fragment. The cells were grown for 4 to 5 days in OCL medium (with M-CSF and RANKL), with daily medium changes. Cells were removed from the bone slices using 0.25 M ammonium hydroxide; the bone slices were washed in water and stained with acid hematoxylin (Sigma-Aldrich). Pit area was quantified by measuring the percentage of stained area compared with the entire bone surface in 24 to 105 nonoverlapping fields from three separate bone fragments per mouse per experiment, using ImageJ software.

Cell lysate preparation, immunoprecipitation, GST pulldowns, SDS-PAGE, and Western blotting

Cells were lysed in buffer A [50 mM tris-Cl (pH 8), 150 mM NaCl, and 1% NP-40] and protease inhibitors (1 mM AEBSF, 40 μM bestatin, 15 μM E-64, 20 μM leupeptin, and 15 μM pepstatin; Sigma-Aldrich). Sodium pervanadate (0.5 mM) was included unless noted otherwise. Protein immunoprecipitation, SDS–polyacrylamide gel electrophoresis (SDS-PAGE), and protein blotting were performed as described (14). Enhanced chemiluminescence signals were visualized and quantified using an ImageQuant LAS 400 Mini instrument (GE Healthcare). GST fusion proteins were purified from bacteria as described (65). In brief, GST, GST-Grb2 (full-length Grb2 or the isolated N-SH3, SH2, or C-SH3 domains), and GST-Src SH2 domain were grown in Escherichia coli. Bacteria were lysed by sonication in NETN buffer [0.5% NP-40, 20 mM tris-Cl, 100 mM NaCl, and 1 mM EDTA (pH 8.0)], and the GST proteins were adsorbed onto Glutathione-Agarose beads (Sigma-Aldrich). After three washes in NETN buffer, beads were added to cell lysates that were prepared as described above, incubated with gentle mixing for 1 hour, and washed in NETN buffer. The beads were then boiled in SDS-PAGE sample buffer and analyzed by SDS-PAGE and protein blotting as described above.

PTPROt autodephosphorylation activity assay

For protein preparation, wild-type PTPROt or the C325S mutant, each carrying a C-terminal FLAG tag, was expressed in HEK293 cells and purified by immunoprecipitation. Inactive, phosphorylated, HA-tagged wild-type PTPROt was prepared similarly from transfected HEK293 cells that had been treated with 0.5 mM sodium pervanadate for 5 min immediately before lysis. Lysis and precipitation buffers did not contain sodium pervanadate. After immunoprecipitation, beads carrying the FLAG-tagged proteins were washed three times with HNTG buffer [20 mM Hepes (pH 7.3), 150 mM NaCl, 0.1% Triton X-100, and 10% glycerol] and twice with elution buffer [20 mM tris (pH 7.5), 100 mM NaCl, and 0.5 mM EDTA]. FLAG-tagged proteins were eluted from the beads by incubation for 10 min at 37°C in 3XFLAG peptide (equal volume of 1 mg/ml; APExBIO in elution buffer). Elution was repeated and both eluates were combined. Inactive, HA-tagged, phosphorylated PTPROt was not eluted. Protein purity and amount were assessed relative to a BSA standard by SDS-PAGE.

For the PTPROt trans-dephosphorylation activity assay, equal amounts of beads carrying inactive, HA-tagged, phosphorylated PTPROt, prepared as described above, were suspended in 100 μl of phosphatase assay buffer [50 mM MES, 0.5 mM dithiothreitol, and BSA (0.5 mg/ml)] along with 200 ng of FLAG-tagged wild-type or C325S PTPROt eluted protein, and incubated at 32°C for 30 or 60 min. Sodium pervanadate (5 mM) was added to some reactions. The beads carrying HA-PTPROt were then washed, subjected to SDS-PAGE, and analyzed by protein blotting for pTyr and HA.

To assay the activity of mutant PTPROt molecules, FLAG-tagged wild-type PTPROt and its mutant forms (C325S, R331M, or Y399F PTPROt) were separately expressed in HEK293 cells. Cells were lysed in buffer A, and lysates were verified by protein blotting to contain similar amounts of the PTPROt proteins. Tyrosine phosphatase activity present in the lysates was measured in 96-well plate format, each well containing 25 or 50 μg of lysate protein in 200 μl of phosphatase assay buffer [50 mM MES, BSA (0.5 mg/ml), 0.5 mM dithiothreitol, and 10 mM PNPP (pH 7.0)]. Some reactions contained 0.5 mM sodium pervanadate to identify PTP activity, which is inhibited by pervanadate. Activity was monitored as the increase in absorption at 405 nm at 30°C for 1 hour and was linear with respect to time and amount of lysate. Activity of endogenous tyrosine phosphatases was determined in lysates of mock-transfected cells and was subtracted from activity measured in PTPROt-expressing cells.

Protein array experiments

The cloning of the human SH2 domain library has been described previously (66). All SH2 domains were expressed as GST fusions in E. coli and purified on Glutathione-Sepharose beads. The recombinant SH2 domains were arrayed onto nitrocellulose-coated glass slides (Oncyte Avid slides, Grace Bio-Labs) using a pin arrayer (67). A list of the SH2 domains on this array is presented in fig. S5. Biotinylated peptides were purchased from CPC Scientific. The fluorescent labeling of the biotinylated peptide probe and slide binding were done as described (67). Fluorescence signals were detected using a GeneTACTM LSIV scanner (Genomic Solutions). The array experiments were performed by the Protein Array and Analysis Core at MD Anderson Cancer Center.

Src kinase activity

OCLs were lysed in buffer A. Lysates were incubated with antibodies directed against Src and Protein-AG beads (Santa Cruz Biotechnology) for 3 hours at 4°C with gentle shaking. Beads were washed in buffer A and then once with kinase buffer [20 mM MOPS and 5 mM MgCl2 (pH 7.0)]. Beads were then incubated in 25 μl of kinase buffer to which 1 μl (5 μCi) of [γ-32P] adenosine 5′-triphosphate (3000 Ci/mmol, 10 mCi/ml; Amersham Biosciences) and 5 μg of acid-denatured enolase (Sigma-Aldrich) were added. Tubes were incubated at 30°C for 30 min, during which Src activity was linear with respect to time. Reactions were stopped by adding SDS-PAGE sample buffer and boiling. Samples were electrophoresed and blotted, and radioactivity present in enolase was quantified using a phosphorimager (Typhoon FLA 7000, GE Healthcare). The same blots were then probed with Src antibodies for normalization of Src activity to the amount of the kinase present in the immunoprecipitates.

Statistical analysis

Statistical analyses were performed by Student’s t test or by ANOVA, as indicated. Statistical significance thresholds were set at P = 0.05.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/12/563/eaau0240/DC1

Fig. S1. Detection of the Y399F mutation engineered into the Ptpro gene of mice.

Fig. S2. OCLs from WT and ROKO mice.

Fig. S3. Verification of engraftment after bone marrow transplantation.

Fig. S4. Activity of WT PTPROt and several PTPROt mutants toward PNPP.

Fig. S5. PTPROt pTyr399 binds to specific SH2 domains.

Fig. S6. PTPROt coprecipitates with Grb2 in a phosphorylation-dependent manner.

Fig. S7. Staining of OCLs with antibodies against Src and HA.

Table S1. μCT analysis of tibiae from WT-HA-PTPROt versus Y399F-HA-PTPROt (line 16) mice.

REFERENCES AND NOTES

Acknowledgments: We thank T. Dahlem and the University of Utah Mutation Generation and Detection Core Facility for reagents, S. Li (University of Western Ontario, Canada) for many of the GST fusion constructs used in the array study, T. M. Roberts (Dana Farber Cancer Institute) for GST fusion constructs of full-length Grb2 and its isolated domains, and V. Fedyuk for assistance in the initial stages of this study. We also thank R. Haffner-Kraus, G. Damari, A. Berkovitz, and S. Peretz of the Weizmann Institute Transgenic and Knockout Core Facility for help in preparing the mouse models used in this study, O. Higfa and N. Sharabi for expert animal care, and R. Rotkopf of the Weizmann Institute’s Department of Life Sciences Core Facilities for assistance with statistical analyses. Funding: This study was supported by grants from the Israel Science Foundation (the Joint Canada-Israel Program, no. 2640/16) (to A.E.), the Kekst Family Center for Medical Genetics at the Weizmann Institute (to A.E.), and the Gutwirth Foundation (to A.E.) and by the Trilateral Program of the Deutsche Forschungsgemeinschaft [(Germany), no. Tu220/12] (to A.E. and J.T.). M.T.B. and the Protein Array and Analysis Core at MD Anderson Cancer Center are supported by funding from the Cancer Prevention and Research Institute of Texas (CPRIT)—RP130432 and RP180804. Author contributions: L.R., J.W., E.W., M. Shalev., E.A., M. Stein, V.B., and C.A.S. performed the research; J.W., L.R., M. Shalev., M.T.B., and A.E. developed the concept of the research; M.T.B., J.T., and A.E. supervised the research; A.E. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials. Genetically manipulated mice are obtainable through a Weizmann Institute of Science material transfer agreement.
View Abstract

Navigate This Article