Tyrosine phosphorylation of the transmembrane protein SIRPα: Sensing synaptic activity and regulating ectodomain cleavage for synapse maturation

Synapse maturation is a neural activity–dependent process during brain development, in which active synapses preferentially undergo maturation to establish efficient neural circuits in the brain. Defects in this process are implicated in various neuropsychiatric disorders. We have previously reported that a postsynaptic transmembrane protein, signal regulatory protein-α (SIRPα), plays an important role in activity-dependently directing synapse maturation. In the presence of synaptic activity, the ectodomain of SIRPα is cleaved and released and then acts as a retrograde signal to induce presynaptic maturation. However, how SIRPα detects synaptic activity to promote its ectodomain cleavage and synapse maturation is unknown. Here, we show that activity-dependent tyrosine phosphorylation of SIRPα is critical for SIRPα cleavage and synapse maturation. We found that during synapse maturation and in response to neural activity, SIRPα is highly phosphorylated on its tyrosine residues in the hippocampus, a structure critical for learning and memory. Tyrosine phosphorylation of SIRPα was necessary for SIRPα cleavage and presynaptic maturation, as indicated by the fact that a phosphorylation-deficient SIRPα variant underwent much less cleavage and could not drive presynaptic maturation. However, SIRPα phosphorylation did not affect its synaptic localization. Finally, we show that inhibitors of the Src and JAK kinase family suppress neural activity–dependent SIRPα phosphorylation and cleavage. Together, our results indicate that SIRPα phosphorylation serves as a mechanism for detecting synaptic activity and linking it to the ectodomain cleavage of SIRPα, which in turn drives synapse maturation in an activity-dependent manner.

The precise development of synapses, the sites of information transfer between neurons in the brain, is critical for proper information processing and brain function. The impaired formation of synapses leads to numerous neurological and psychiatric disorders, such as autism and schizophrenia (1)(2)(3)(4)(5). Syn-apses develop through a series of highly regulated stages: initial synapse differentiation, synapse maturation, and synapse maintenance stages (6 -8). These stages of synapse development are orchestrated by a number of bidirectional, trans-synaptic signals that organize and coordinate the development of both preand postsynaptic terminals (6 -11). Whereas the initial synapse differentiation stage, during which immature synapses are formed, is considered to be synaptic activity-independent, the synapse maturation stage is regulated by synaptic activity (12)(13)(14)(15). During synapse maturation, only the active and functional synapses are strengthened and stabilized to form effective synaptic networks in the brain. However, the molecules and mechanisms by which synapses mature in an activity-dependent manner are largely unknown. Defects in synapse maturation have been implicated in various neurological and psychiatric disorders (1-5, 16, 17). Thus, understanding the molecular mechanisms underlying activity-dependent synapse maturation should further our understanding of and provide novel therapeutic strategies for neurological and psychiatric disorders.
To identify the molecular mechanisms by which synapses mature, we have purified synaptogenic molecules from developing mouse brains using a synapse formation assay in cultured neurons (18). One of the synaptogenic molecules we identified was signal regulatory protein-␣ (SIRP␣), 2 a member of the Ig superfamily transmembrane proteins. SIRP␣ is highly expressed in the brain and is concentrated at the postsynaptic terminals of excitatory synapses (19). Using in vitro culture systems and in vivo mouse mutants, we showed that in the hippocampus, the region central to learning, memory, and emotional processing, SIRP␣ serves as a retrograde, trans-synaptic signal that promotes presynaptic maturation (19). Importantly, synaptic activity is necessary for SIRP␣ to drive presynaptic maturation. When synapses are active, the ectodomain of SIRP␣ is cleaved and shed and drives maturation of the presynaptic terminal. Clearly, there are two important questions to address. 1) How does SIRP␣ detect synaptic activity? 2) How does synaptic activity regulate the cleavage of the SIRP␣ ectodomain? Addressing these questions will advance our understanding of the mechanisms underlying activity-dependent synapse maturation, a process fundamental to the proper wiring of the brain, and shed light on the pathophysiology of neuropsychiatric disorders caused by impaired synapse maturation.
In nonneural systems, an important way in which SIRP␣ signaling is transduced is via SIRP␣ tyrosine phosphorylation. The SIRP␣ intracellular domain contains immunoreceptor tyrosine-based inhibitory motifs. The tyrosine phosphorylation of these motifs in response to various stimuli, such as growth factors and ligand binding, recruits and subsequently activates the Src homology 2 domain-containing tyrosine phosphatases (SHPs). These phosphatases regulate intracellular signaling pathways, specifically the mitogen-activated protein kinase and the NF-B pathways, to moderate various cellular functions (20,21). In cultured cells, the tyrosine phosphorylation of SIRP␣ regulates cell migration and cell proliferation (22,23). The expression of WT SIRP␣, but not a phosphorylationdeficient SIRP␣ mutant, positively regulated Chinese hamster ovary cell migration in response to insulin (22) and breast cancer cell proliferation (23). In the immune system, the tyrosine phosphorylation-dependent signaling of SIRP␣ regulates inflammatory responses, including monocyte adhesion, macrophage migration, phagocytosis, and cytokine release (24 -28). SIRP␣ tyrosine phosphorylation has also been implicated in osteoblast differentiation (29). In the brain, SIRP␣ is known to be tyrosinephosphorylated in response to stress, hypothermia, or light exposure (30 -33). However, the role of SIRP␣ tyrosine phosphorylation in the brain is not known.
Here we identify a critical role for the tyrosine phosphorylation of SIRP␣ in activity-dependent synapse maturation. We show that 1) SIRP␣ is highly phosphorylated on its tyrosine residues during synapse maturation, 2) neural activity induces tyrosine phosphorylation of SIRP␣, 3) SIRP␣ phosphorylation is critical for its synaptogenic activity, 4) SIRP␣ phosphorylation is necessary for its ectodomain cleavage, and 5) inhibitors of Src and JAK family tyrosine kinases suppress the activity-dependent tyrosine phosphorylation and cleavage of SIRP␣. These results demonstrate a critical role of SIRP␣ tyrosine phosphorylation in sensing synaptic activity, regulating SIRP␣ ectodomain cleavage, and, ultimately, driving synapse maturation in an activity-dependent manner. Our results reveal a novel molecular mechanism by which synaptic activity is converted to synaptogenic activity to establish functional synaptic networks.

Characterization of SIRP␣ mutant constructs and antibodies
To determine the role of SIRP␣ tyrosine phosphorylation during synaptic development, we created a tyrosine phosphorylation-deficient SIRP␣ mutant (SIRP␣4YF). In this construct, all four intracellular tyrosine residues in SIRP␣ were mutated to phenylalanine residues, preventing tyrosine phosphorylation (Fig. 1A). We confirmed that SIRP␣4YF is indeed tyrosine phos-phorylation-deficient by examining its JAK2-dependent tyrosine phosphorylation (34,35). In transfected COS cells, SIRP␣WT was tyrosine-phosphorylated by an active form of JAK2 (JAK2VF) (36), whereas SIRP␣4YF was not (Fig. 1B, top  left). We further verified the phosphorylation deficiency of SIRP␣4YF by examining its phosphorylation-dependent SHP2 interaction; the adaptor protein SHP2 is known to bind SIRP␣ following the tyrosine phosphorylation of SIRP␣ (37)(38)(39). We co-transfected COS cells with SIRP␣4YF or SIRP␣WT, SHP2, with or without JAK2VF. We then immunoprecipitated SIRP␣ and determined its SHP2 interaction. We found that in the presence of JAK2VF, SHP2 was physically associated with SIRP␣WT but not with SIRP␣4YF (Fig. 1B, second panel from  top left). These results verify that the SIRP␣4YF mutant is tyrosine phosphorylation-deficient.
We then characterized the two SIRP␣ antibodies we used to examine the role of SIRP␣ tyrosine phosphorylation: an antibody recognizing the extracellular domain of SIRP␣ (SIRP␣ N-Term) and an antibody recognizing the intracellular domain of SIRP␣ (SIRP␣ C-Term) (Fig. 1C). We have verified that both antibodies 1) similarly recognize SIRP␣WT and SIRP␣4YF by Western blotting and 2) effectively immunoprecipitate both SIRP␣WT and SIRP␣4YF (83-96% efficiency; Fig. 1, D and E). The C-Term antibody recognized both the SIRP␣ full length and intracellular C-terminal fragment, which is produced after ectodomain cleavage, by Western blotting (Fig. 1E). Both antibodies specifically recognize SIRP␣ by immunostaining, as confirmed by lack of staining in SIRP␣ knockout brains (Fig. 1F). These results confirm that the antibodies are suitable for examining the role of SIRP␣ tyrosine phosphorylation by Western blotting, immunoprecipitation, and immunostaining.

SIRP␣ is highly phosphorylated on tyrosine residues during synapse maturation in the hippocampus in vivo
The synaptic Ig superfamily molecule, SIRP␣, is a critical activity-dependent regulator of excitatory synapse maturation in the hippocampus (19). Our aim here was to understand how SIRP␣ detects synaptic activity and how, in turn, this regulates SIRP␣'s synaptogenic activity. For this, we focused on tyrosine phosphorylation of SIRP␣. SIRP␣ has four tyrosine residues in its intracellular domain (37). Because tyrosine phosphorylation is a rapid and localized mechanism that moderates the function of a protein (40), we hypothesized that SIRP␣ is tyrosine-phosphorylated in response to synaptic activity during synapse maturation and that this phosphorylation is necessary for SIRP␣'s role in driving activity-dependent synapse maturation. To test this idea, we first examined whether SIRP␣ is tyrosine-phosphorylated during synapse maturation in the mouse hippocampus in vivo.
In the mouse hippocampus, initial synapse differentiation occurs during the first two postnatal weeks, between postnatal day 0 (P0) and P14 (19,41). This differentiation stage is followed by an activity-dependent synapse maturation stage that occurs between P15 and P30 (13,19). Afterward, synapses are maintained throughout life. Hence, we evaluated tyrosine phosphorylation of SIRP␣ at P6 (during synapse differentiation), P15 (the start of synapse maturation), and P20 (during synapse maturation) and in adults (during synapse maintenance). We immu-noprecipitated SIRP␣ from hippocampal lysates and assessed the amount of total and tyrosine-phosphorylated SIRP␣ present ( Fig. 2A). The total amount of SIRP␣ continuously increased from P6 to adulthood ( Fig. 2C; the same results were obtained by directly blotting the lysates for SIRP␣. Note that the levels of N-cadherin and ␣-tubulin were similar across all ages tested, suggesting that the lysis buffer is equally capable of solubilizing synaptic proteins). In contrast, the amount of tyrosine-phosphorylated SIRP␣ robustly increased from P6 to P15, when synapse maturation starts, and then decreased afterward (Fig. 2, A and B). The percentage of phosphorylated SIRP␣ to total SIRP␣ also peaked around P15 (Fig. 2D) Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis demonstrate that both the absolute amount of tyrosine-phosphorylated SIRP␣ and the relative percentage of SIRP␣ that is phosphorylated are highest at the beginning of the synapse maturation stage. This suggests that tyrosine phosphorylation of SIRP␣ may play important roles in regulating synapse maturation in the hippocampus.

Neural activity induces tyrosine phosphorylation of SIRP␣
The synapse maturation stage is an activity-dependent stage where active synapses are the ones that undergo maturation (12)(13)(14)(15)19). Because SIRP␣ phosphorylation peaks during the synapse maturation stage in the hippocampus (Fig. 2), we next asked whether SIRP␣ phosphorylation is regulated by neural . COS cells were transfected with a SIRP␣ construct (WT or 4YF) and SHP2, together with or without an active form of JAK2 (JAK2VF). Cells were immunoprecipitated for SIRP␣ protein (with SIRP␣ N-Term antibody; described below), and the immunoprecipitates were blotted for phosphotyrosine (pY), SHP2, and SIRP␣ (SIRP␣ N-Term) (left). SIRP␣WT is phosphorylated by JAK2VF and binds SHP2; however, SIRP␣4YF is not phosphorylated by JAK2VF and does not bind SHP2. Expression levels of transfected proteins in the lysates are shown in the right panel. C, illustration showing the recognition sites of the anti-SIRP␣ antibodies used; SIRP␣ (N-Term) recognizes the SIRP␣ ectodomain, and SIRP␣ (C-Term) recognizes the SIRP␣ intracellular domain. D, immunoprecipitation efficiency of the SIRP␣ (N-Term) antibody. The same amount of lysates was used for direct blotting (Lysates) and immunoprecipitation (IP) followed by blotting. IP efficiency (%) was calculated by quantifying the ratio of SIRP␣ (IP) to SIRP␣ (Lysates). The IP efficiency of the SIRP␣ (N-Term) antibody in detecting SIRP␣WT and SIRP␣4YF was 83.88 and 86.66%, respectively. E, immunoprecipitation efficiency of the SIRP␣ (C-Term) antibody. The C-Term antibody recognizes both the full-length and the intracellular/transmembrane domain (lacking the ectodomain) SIRP␣. IP efficiency was calculated as in D. The SIRP␣ (C-Term) effectively immunoprecipitated SIRP␣WT (full-length, 95.93%; intracellular fragment, 87.89%) and SIRP␣4YF (full-length, 94.58%; intracellular fragment, 93.78%). F, verification of the specificity of SIRP␣ antibodies by immunostaining. SIRP␣ KO brains were stained with either SIRP␣ (N-Term) or SIRP␣ (C-Term). SIRP␣ KO brains showed a lack of staining with both antibodies.  Figure 2. SIRP␣ tyrosine phosphorylation peaks during synapse maturation in the mouse hippocampus. A, SIRP␣ protein was immunoprecipitated (with SIRP␣ N-Term antibody) from the mouse hippocampal lysates prepared at P6, P15, P20, and adulthood. Immunoprecipitates were subjected to Western blotting using antibodies against phosphotyrosine (pY) and SIRP␣ (N-Term). Lysates were also directly blotted for SIRP␣ (N-Term), N-cadherin, and ␣-tubulin. The levels of N-cadherin and ␣-tubulin were similar across all ages tested, suggesting that the lysis buffer is equally capable of solubilizing synaptic proteins at different ages. B, quantification of the absolute amount of tyrosine-phosphorylated SIRP␣ (pY-SIRP␣) normalized to that at P6. C, quantification of total SIRP␣ (black dots, after IP; red dots, lysates) normalized to that at P6. D, quantification of the percentage of tyrosine-phosphorylated SIRP␣ in total SIRP␣ (pY-SIRP␣/ total SIRP␣; normalized to P6). Both the absolute amount of phosphorylated SIRP␣ and the percentage of phosphorylated SIRP␣ are highest at P15, which corresponds to the beginning of the synapse maturation period in the mouse hippocampus (13,19). Equal amounts of total protein were used for immunoprecipitation and blotting. n ϭ 4 mice for each age group. Error bars, S.E.

Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis
activity. For this, we cultured hippocampal neurons, and at DIV (days in vitro) 12, which is during synapse maturation in cultured hippocampal neurons (19), we treated cultures for 20 min either with a neurotransmitter receptor inhibitor mixture (50 M picrotoxin, 10 M CNQX, and 50 M AP5) to silence all synaptic activity or with 55 mM potassium chloride (KCl) to depolarize neurons. This treatment length was chosen because the KCl treatment for 20 min showed the most robust and consistent response with regard to SIRP␣ phosphorylation relative to that of 3 min and 1 h (data not shown). After treatment, we immunoprecipitated SIRP␣ and examined the amount of total and tyrosine-phosphorylated SIRP␣. We found that the activation of neurons by KCl treatment induced a more than 2-fold increase in SIRP␣ phosphorylation relative to untreated and inhibitor-treated cultures (Fig. 3, A and B), indicating that neural activity induces the tyrosine phosphorylation of SIRP␣. The total level of SIRP␣ was not significantly different in all conditions (Fig. 3, C and D). These results support the idea that tyrosine phosphorylation of SIRP␣ contributes to activity-depen-dent synapse maturation. The addition of inhibitors showed similar levels of SIRP␣ phosphorylation as untreated cultures. This may be due to the relatively low level of basal synaptic activity in neuronal cultures in the 20-min period (Fig. 3B).

Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis
with Cre (Fig. 4A). Both SIRP␣WT and SIRP␣4YF expressed equally in neurons (Fig. 4, B and C). Additionally, to label transfected neurons, we co-transfected StopYFP, which expresses YFP only when Cre is co-expressed (42). This system allows us to temporally control SIRP␣ inactivation and rescue during synapse development. We transfected neurons sparsely to evaluate the effects of SIRP␣ KO and rescues in a neuron-specific and cell autonomous manner (Fig. 4D).
To evaluate glutamatergic presynaptic maturation, we analyzed the clustering of glutamatergic synaptic vesicles by staining for vesicular glutamate transporter 1 (VGluT1) (19,41) (Fig.  4E). We quantified the number of VGluT1 puncta on the dendrites of transfected neurons (Fig. 4F). The inactivation of endogenous SIRP␣ (SIRP␣ KO) significantly decreased the density of VGluT1 puncta on transfected dendrites relative to control (GFP, instead of Cre, was transfected), consistent with our previous report showing that SIRP␣ is necessary for glutamatergic presynaptic maturation (19). This decrease was completely rescued by the expression of SIRP␣WT. In contrast, SIRP␣4YF failed to rescue the decrease in VGluT1 density. These results suggest that tyrosine phosphorylation of SIRP␣ is necessary for presynaptic maturation.
To confirm the necessity of SIRP␣ tyrosine-phosphorylation in synapse maturation, we performed electrophysiological experiments. We recorded miniature excitatory postsynaptic currents (mEPSCs) from transfected hippocampal neurons (Fig. 4G). In agreement with our histological data, we found a marked decrease in mEPSC frequency (reflecting the number of presynaptic inputs) in SIRP␣ KO neurons relative to control neurons. This decrease in mEPSC frequency was completely rescued by the expression of SIRP␣WT but not by SIRP␣4YF (Fig. 4H). Together, our data demonstrate that SIRP␣ tyrosine phosphorylation is critical for SIRP␣'s synaptogenic function in increasing the number of functional presynaptic inputs.
Interestingly, we noticed that the mEPSC amplitude (reflecting postsynaptic response) was increased in SIRP␣4YF-rescue neurons ( Fig. 4I), whereas there was no difference among control, SIRP␣ KO, and SIRP␣WT-rescue neurons. This suggests that SIRP␣ phosphorylation may also influence postsynaptic maturation. SIRP␣4YF may be interfering with signals that control postsynaptic development.
To rule out the possibility that the inability of SIRP␣4YF to drive presynaptic development is due to its inability to localize to synapses, we examined the synaptic localization of SIRP␣WT and SIRP␣4YF. In neurons, endogenous SIRP␣ localizes predominantly to excitatory postsynaptic terminals (19). To visualize excitatory postsynaptic terminals in cultured hippocampal neurons, we co-transfected EGFP-tagged postsynaptic density 95 (PSD95-EGFP), which accumulates at excitatory postsynaptic terminals and effectively labels them (43). In control neurons, SIRP␣ accumulated at PSD95-EGFPpositive synapses. This expression of SIRP␣ on PSD95-EGFPpositive puncta was diminished in SIRP␣ KO neurons (Fig. 4J). Importantly, both SIRP␣WT and SIRP␣4YF localized to PSD95-EGFP-positive puncta (Fig. 4, J and K). Additionally, the intensity of SIRP␣ staining at synapses was not significantly different between SIRP␣WT and SIRP␣4YF (Fig. 4L). These results indicate that the tyrosine phosphorylation of SIRP␣ does not regulate the synaptic targeting of SIRP␣.

Tyrosine phosphorylation of SIRP␣ regulates SIRP␣ ectodomain cleavage
SIRP␣ regulates synapse maturation via activity-dependent ectodomain cleavage (19). Because we found that tyrosine phosphorylation of SIRP␣ is critical for synapse maturation (Fig. 4), we hypothesized that SIRP␣ tyrosine phosphorylation regulates the cleavage of SIRP␣. To test this idea, we transfected COS cells with SIRP␣WT only, SIRP␣4YF only, SIRP␣WT ϩ JAK2VF (to drive SIRP␣ phosphorylation), or SIRP␣4YF ϩ JAK2VF. Two days after transfection, we collected the culture media and immunoprecipitated the secreted SIRP␣ ectodomain. Without tyrosine kinases, the levels of secreted SIRP␣ were similar between SIRP␣WT and SIRP␣4YF (Fig. 5A). In contrast, in the presence of JAK2VF, SIRP␣WT had significantly more ectodomain released than SIRP␣4YF (Fig. 5B). Similar results were obtained using sodium orthovanadate, which inhibits protein-tyrosine phosphatases and activates tyrosine kinases (44), to drive SIRP␣ phosphorylation (Fig. 5C). Our quantitative analysis (percentage of cell content released per hour) showed that significantly more ectodomain was released from SIRP␣WT than SIRP␣4YF following vanadate treatment. Vanadate treatment induced the cleavage of SIRP␣WT at the rate of 2.15% per hour (percentage of cell content). In contrast, vanadate induced the cleavage of SIRP␣4YF only at the rate of 0.609%. This difference in ectodomain cleavage is not due to lack of surface expression of SIRP␣4YF, because both forms of SIRP␣ were detected on the cell surface when cells were stained for SIRP␣ without permeabilizing the cell membrane (Fig. 5D). These results indicate that tyrosine phosphorylation of SIRP␣ plays a critical role in regulating SIRP␣ ectodomain cleavage in COS cells. Interestingly, after the cleavage of the SIRP␣ ectodomain, the C-terminal fragment of SIRP␣ was dephosphorylated (Fig. 5E).
To determine whether SIRP␣ tyrosine phosphorylation regulates ectodomain cleavage in neurons as well, we prepared hippocampal cultures from Sirpa fl/fl mice (19) and transfected the cultures to create the SIRP␣WT-rescue or SIRP␣4YF-rescue conditions as described above (Fig. 4A). To identify synapses, we co-transfected PSD95-EGFP. We then left cultures untreated, or we treated these cultures with KCl for 2 h at DIV12. After the treatment, we stained the cultures using the SIRP␣ N-terminal antibody without permeabilization to detect the surface SIRP␣ ectodomain, followed by the SIRP␣ C-terminal antibody with permeabilization to evaluate total SIRP␣. We imaged these cultures and calculated the ratio of the SIRP␣ ectodomain to total SIRP␣. At the basal level (no KCl conditions), the ectodomain/total SIRP␣ ratio was significantly smaller in SIRP␣WT-rescue neurons relative to SIRP␣4YFrescue neurons, suggesting that the basal level of neuronal activity induced more cleavage of SIRP␣WT than SIRP␣4YF. In SIRP␣WT-rescue neurons, KCl treatment significantly decreased the ectodomain/total SIRP␣ ratio relative to control, suggesting that the SIRP␣WT ectodomain is further cleaved upon neuronal activation. In contrast, in SIRP␣4YF neurons, KCl treatment did not significantly change the ectodomain/ total SIRP␣ ratio (Fig. 6, A and B). The total level of SIRP␣ is not

Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis
significantly different in all conditions (Fig. 6C). These results suggest that neural activity-dependent SIRP␣ cleavage requires SIRP␣ tyrosine phosphorylation.
To ascertain that the change in the ectodomain/total SIRP␣ ratio is indeed due to SIRP␣ ectodomain cleavage, we treated SIRP␣WT-rescue cultures with the proteinase inhibitors, TIMPs, which inhibit SIRP␣ ectodomain cleavage (19). The treatment of cultures with TIMPs blocked the KCl-driven decrease in the ectodomain/total SIRP␣ ratio, suggesting that the change in the ratio is due to SIRP␣ ectodomain cleavage (Fig. 6, D and E).

Inhibitors of Src and JAK family kinases suppress neural activity-driven SIRP␣ tyrosine phosphorylation
We next wanted to identify the signaling molecules that regulate the tyrosine phosphorylation of SIRP␣ in response to neural activity. For this, we focused on two families of tyrosine kinases, the Src and JAK families, because they are highly expressed in hippocampal neurons and can be activated by neural activity (45)(46)(47). Additionally, we showed that in COS cells, JAK2 can phosphorylate SIRP␣ (Fig. 1B) (34,35).
To test whether Src and JAK kinase activity contributes to activity-dependent tyrosine phosphorylation of SIRP␣ in neurons, we treated hippocampal cultures prepared from WT mice at DIV12 with either 10 M PP2 (an inhibitor of Src family kinases) (48) or 10 M AG490 (an inhibitor of JAK family kinases) (46, 49) for 1 h. Control cultures were treated with DMSO only (PP2 and AG490 were dissolved in DMSO). These cultures were then treated with or without 55 mM KCl for 20 min to activate neurons. We then subjected these cultures to immunoprecipitation for SIRP␣ and immunoblotted for phosphotyrosine and SIRP␣. We found that treatment of neurons with either PP2 or AG490 blocked the increase in SIRP␣ tyrosine phosphorylation in response to neuronal activation (Fig. 7,  A and B). These results indicate that the inhibitors, likely through the inhibition of either Src or JAK family kinases, blocked the activity-dependent tyrosine phosphorylation of SIRP␣. Interestingly, the treatment of cultures with PP2, but not AG490, diminished the basal level of SIRP␣ phosphorylation (Fig. 7, C and D), suggesting that Src family kinases, but not JAK kinases, also contribute to the basal level of SIRP␣ phosphorylation in neurons.

CaMK activity is not involved in SIRP␣ tyrosine phosphorylation
We have previously shown that the cleavage of SIRP␣ requires CaMK activity (19). The inhibition of CaMK activity resulted in a substantial decrease in the amount of cleaved SIRP␣. Because SIRP␣ cleavage is regulated by its tyrosine phosphorylation (Figs. 5 and 6), we asked whether CaMK signaling influences tyrosine phosphorylation of SIRP␣. To test this, we treated cultured hippocampal neurons at DIV12 with 5 M KN62 (a CaMK inhibitor) and activated neurons with KCl for 20 min. Treatment of KN62 effectively suppressed CaMK activation at synapses (Fig. 7E). Cultures treated with KN62 still showed a marked increase in the level of SIRP␣ tyrosine phosphorylation in response to neuronal activation (Fig. 7, F and G), suggesting that CaMK activity is not involved in activity-dependent tyrosine phosphorylation of SIRP␣. This implies that SIRP␣ tyrosine phosphorylation and CaMK signaling regulate SIRP␣ ectodomain cleavage via distinct pathways.

Inhibitors of Src and JAK family kinases suppress neural activity-driven SIRP␣ ectodomain cleavage
Based on our data that inhibitors of Src and JAK family kinases suppress SIRP␣ tyrosine phosphorylation in response to neural activity (Fig. 7), we next wanted to determine whether these kinase inhibitors also suppress activity-dependent SIRP␣ ectodomain cleavage in neurons. We first determined that KCldependent neuronal activation affects the SIRP␣ cleavage by sequential immunostaining (see Fig. 6). After KCl treatment, we stained the neurons with the SIRP␣ N-terminal antibody without permeabilization (to detect surface "ectodomain-containing SIRP␣") and then with the SIRP␣ C-terminal antibody (to detect "total SIRP␣") and VGluT1 antibody (to mark synapses) with permeabilization. We quantified the ectodomain/ total SIRP␣ ratio at synapses. We found that KCl treatment resulted in a decrease in this ratio: the intensity of surface ectodomain-containing SIRP␣ at the synapse significantly decreased following KCl treatment, whereas the level of total SIRP␣ stayed constant (Fig. 8, A-C). TIMPs, the proteinase inhibitors that inhibit SIRP␣ ectodomain cleavage, blocked the KCl-driven decrease in the amount of ectodomain SIRP␣ and the ectodomain/total SIRP␣ ratio (Fig. 8, A and B). These results support our conclusion that neuronal activation induces the cleavage of SIRP␣ ectodomain.

Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis
We next performed these experiments with the addition of the kinase inhibitors, PP2 or AG490, which block activity-dependent phosphorylation of SIRP␣ (Fig. 8D). We found that PP2 or AG490 treatment inhibited the KCl-driven decrease in the amount of ectodomain SIRP␣ and the ectodomain/total SIRP␣ ratio (Fig. 8, E-G). These results indicate that the inhibition of Src or JAK family kinases inhibits the activity-dependent ectodomain cleavage of SIRP␣.

Discussion
Selective maturation of active synapses is critical for the establishment of functional neural circuits (1)(2)(3)(4)(5)(6)(7)(8)(12)(13)(14)(15). Although a few molecules have been identified as regulators of synapse maturation (19, 50 -52), the mechanisms through which these molecules detect synaptic activity and signal synapse maturation selectively at active synapses remain unclear. Here, we identified a critical role for the tyrosine phosphorylation of SIRP␣ in activity-dependent synapse maturation. We showed that 1) tyrosine phosphorylation of SIRP␣ peaks in vivo during the synapse maturation period; 2) SIRP␣ phosphorylation is driven by neural activity; 3) in contrast to WT SIRP␣, the phosphorylation-deficient SIRP␣ mutant cannot drive presynaptic maturation; 4) tyrosine phosphorylation of SIRP␣ controls the ectodomain cleavage of SIRP␣, which in turn, signals presynaptic maturation; and 5) inhibitors of Src and JAK family kinases suppress activity-dependent tyrosine phosphorylation and ectodomain cleavage of SIRP␣ in neurons. Collectively, our results demonstrate that tyrosine phosphorylation of SIRP␣ serves as a molecular switch that detects neuronal activity and turns on the maturation of active synapses by promoting the cleavage of SIRP␣'s ectodomain.

Tyrosine phosphorylation of SIRP␣ as a synaptic activity sensor
SIRP␣ is an activity-dependent regulator of excitatory presynaptic maturation (19). The overexpression of WT SIRP␣ in neurons drives presynaptic maturation; however, this only occurs when neuronal activity is present. So, how does SIRP␣ detect synaptic activity and drive the maturation of active synapses only? Here, we showed that tyrosine phosphorylation of SIRP␣ is driven by neuronal activity (Fig. 3) and that this phosphorylation is necessary for SIRP␣'s ability to drive presynaptic maturation during development (Fig. 4). This makes SIRP␣ tyrosine phosphorylation a key mechanism through which SIRP␣ senses neuronal activity and converts it to synaptogenic activity. Tyrosine phosphorylation of SIRP␣ is a temporally and spatially regulated, activity-dependent mechanism that allows SIRP␣ to function specifically at active synapses during development.
SIRP␣ tyrosine phosphorylation may be a general mechanism for SIRP␣ to detect various stimulations, not only during development, but also in adulthood. Light stimulation, hypothermia, and stress can affect SIRP␣ phosphorylation in the adult brain (30 -33). The precise roles of these phosphorylation events are not known, but tyrosine phosphorylation of SIRP␣ may serve as a sensor of various brain activities even in adults.

Tyrosine phosphorylation of SIRP␣ as a "cleave me" signal
How does SIRP␣ tyrosine phosphorylation regulate synapse maturation? Our previous work demonstrated that the ectodomain of SIRP␣ is cleaved, and this cleavage is essential for presynaptic maturation (18,19). Here we showed that SIRP␣ tyrosine phosphorylation regulates this cleavage (Figs. 5 and 6). Thus, we propose that the activity-dependent tyrosine phosphorylation of SIRP␣ serves as a mechanism to tag SIRP␣ for cleavage.
How the tyrosine phosphorylation of SIRP␣ controls the cleavage of its ectodomain is an important next question. Possible mechanisms include the following. 1) SIRP␣ phosphorylation activates downstream signaling that regulates proteinase expression. One possible downstream signaling molecule is SHP2. Upon phosphorylation, SIRP␣ recruits SHP2 (Fig. 1B). SHP2 has been implicated in the regulation of MMP-2 and -9 expression (53). 2) SIRP␣ phosphorylation leads to the recruitment of proteinases. For example, ADAM10, which has an intracellular substrate-binding domain, interacts with the ephrin family of cell adhesion molecules (54,55). This interaction facilitates ephrin cleavage. Similar mechanisms might exist for SIRP␣.
The cleavage of SIRP␣ is suggested to be regulated by MMP/ ADAM10 proteinases (19,56). At the synapse, there are several other synaptic proteins, including neuroligins, N-cadherin, and nectin, that are known to undergo MMP/ADAM-mediated proteolytic cleavage (51,(57)(58)(59)(60). How MMPs and ADAMs recognize these synaptic proteins is unknown, but the phosphorylation of these synaptic proteins may also play a role in tagging them.

Multiple signaling pathways coordinate to regulate SIRP␣ tyrosine phosphorylation and cleavage
Src and JAK family kinases are two major tyrosine kinase families in the brain. Furthermore, many members of these kinase families are known to be activated by neuronal activity (45)(46)(47). Here we identified that the activity of both Src and JAK family kinases regulates tyrosine phosphorylation of SIRP␣ during synapse maturation. Interestingly, the inhibition of Src family kinases resulted in a decrease in both basal and activity- . Two days after transfection, culture medium and cell lysates were collected. Culture medium was subjected to immunoprecipitation for secreted SIRP␣ protein (with SIRP␣ N-term antibody) and blotted for SIRP␣ (N-Term). Cell lysates were subjected to Western blotting for SIRP␣ (N-Term) and JAK2. Graphs show the quantification of the amount of secreted SIRP␣ ectodomain. A, when SIRP␣ was not tyrosine-phosphorylated (without JAK2VF; see Fig. 1B), a similar amount of SIRP␣ ectodomain was released in the medium from SIRP␣WT and SIRP␣4YF. B, in the presence of tyrosine kinase (JAK2VF), significantly more SIRP␣ ectodomain was released in the medium from SIRP␣WT compared with SIRP␣4YF. C, COS cells were transfected with SIRP␣WT or SIRP␣4YF. Two days after transfection, cultures were treated with 1 mM sodium orthovanadate for 7-17 h to promote tyrosine phosphorylation of SIRP␣ (SIRP␣WT, but not SIRP␣4YF, is tyrosine-phosphorylated (pY-SIRP␣; bottom right panel)). Culture medium and cell lysates were collected and subjected to Western blotting; the membranes were first blotted with the SIRP␣ N-Term antibody, stripped, and then blotted with the SIRP␣ C-Term antibody. The following SIRP␣ band intensities were quantified: Media (IB: N-Term, Media), Full-N (N-Term, Lysates), Full-C (C-Term, Lysates, top band), and C-Fragment (C-Term, Lysates, bottom band). The rate of SIRP␣ ectodomain cleavage was calculated using the formula, Media/((Full-C ϩ C-Fragment) ϫ (Full-N/Full-C))/hours (see "Experimental procedures"). Experiments were done as a pair (WT and 4YF), and the pairs are connected with lines in the graph. Significantly more SIRP␣ ectodomain was released in the medium from SIRP␣WT than SIRP␣4YF following vanadate treatment. D, COS cells were transfected with mCherry only, mCherry and SIRP␣WT, or mCherry and SIRP␣4YF. Cells were fixed and stained for SIRP␣ (N-Term) without permeabilization to detect surface expression of SIRP␣. Both SIRP␣WT and SIRP␣4YF are expressed on the surface. E, tyrosine phosphorylation of the SIRP␣ full-length and C-terminal fragment (arrows) after vanadate treatment. n ϭ 5 (A and B) in seven independent experiments (C), reproduced four times (E). Data are mean Ϯ S.E. (error bars). n.s., not significant (p Ͼ 0.05); *, p Ͻ 0.05, by Student's t test (A and B) or two-way ANOVA (C). Scale bar, 20 m.

Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis
dependent SIRP␣ phosphorylation, whereas the inhibition of JAKfamilykinasesonlydeterredactivity-dependentSIRP␣phosphorylation (Fig. 7). These results hint at the possibility that these different kinase groups may coordinate to regulate SIRP␣ phosphorylation during synapse development, perhaps by phosphorylating different tyrosine residues to send different signals downstream. These data also suggest that Src family kinases may phosphorylate and prime SIRP␣ to then be phosphorylat-ed by JAK kinases upon stimulation by neuronal activity. Signals from both kinase families need to coincide to drive the activity-dependent tyrosine phosphorylation of SIRP␣.

Tyrosine phosphorylation of SIRP␣ regulates synaptogenesis
In summary, our work identified a novel molecular mechanism that governs the function of SIRP␣ in sensing neural activity and driving the maturation of active synapses. SIRP␣-driven hippocampal synaptic maturation is a highly regulated process that requires proper signaling of multiple pathways in response to synaptic activity. A comprehensive understanding of the signals that govern synapse development can further our understanding of and suggest novel therapeutic strategies to prevent and/or treat various neuropsychiatric disorders associated with abnormal synapse maturation.

Mouse strains
Sirpa fl/fl and Actin-CreER mutant mice were described previously (19). WT mice were ICR/CD-1 (Charles River) (Figs. 3, 7, and 8) or C57/B6 (Jackson) (Fig. 2). All animal care and use was in accordance with institutional guidelines and approved by the institutional animal care and use committee at Boston Children's Hospital.
For the treatment with inhibitors described in Figs. 3 and 7, neurons were cultured for 12 days, after which culture medium was replaced by new medium containing inhibitor mixture (50 M picrotoxin, 10 M CNQX, and 50 M AP5) and incubated for 7 h. Cultures were treated with a kinase inhibitor: 10 M PP2 or 10 M AG490 (for the last 1 h) or 5 M KN62 (for 7 h). Because inhibitors were dissolved in DMSO, an equal volume of DMSO was added to control cultures. Medium was then replaced with new culture medium containing inhibitor mixture or 55 mM KCl, with or without the kinase inhibitor for 20 min.
For the treatment with inhibitors described in Fig. 6 and 8, neurons were cultured for 12 days, after which culture medium was replaced by new medium. Cultures were then either untreated or treated with 55 mM KCl with or without inhibitors: 10 M PP2, 10 M AG490, or 0.5 g/ml each of TIMP1 and TIMP2 (for 2 h). Because inhibitors were dissolved in DMSO, an equal volume of DMSO was added to control cultures.

COS cell culture and transfection
COS7 cells (ATCC) were cultured in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% fetal bovine serum (Corning). Transfections were done via the Lipofectamine 3000 (Thermo Fisher) method or via the calcium phosphate method. Cells were cultured for 1-3 days following transfection and subjected to biochemical and histological experiments. In the sodium orthovanadate experiments, cultures were treated with 1 mM sodium orthovanadate for 7-17 h before biochemical experiments.

Immunoprecipitation and Western blotting
For cultured cells, media and cells were collected separately. Cells were lysed in lysis buffer containing 1% Triton X-100, 50 mM Tris-HCl (pH 8), 150 mM NaCl, 1 mM Na 3 VO 4 with a protease inhibitor mixture tablet (Roche Applied Science). Dissected hippocampi were lysed by homogenization in 10 volumes of lysis buffer/g of tissue. Protein concentrations were measured by protein assays (BCA, Pierce). Media or lysates were incubated with 1 g of anti-SIRP␣ (p84; BD Biosciences; BDB552371) or 3 g of anti-SIRP␣ C terminus (QED; 2428) antibodies for 2 h at 4°C. Equal amounts of lysates and equal volume of media from each experimental condition were used for immunoprecipitation. The immune complexes were precipitated with Protein L or Protein A (Pierce). Immunoprecipitates and lysates were subjected to SDS-PAGE. Proteins were transferred to a polyvinylidene difluoride membrane (Bio-Rad), blocked with 3% nonfat milk, and blotted using anti-SIRP␣ (p84; BD; 1:1000), anti-phosphotyrosine (4G10; Millipore; 05-321X; 1:1000), anti-SHP2 (Santa Cruz Biotechnology, Inc.; SC-7384; 1:1000), anti-␣-tubulin (DM1A; Invitrogen; 62204; 1:1000), anti-N-cadherin (BD Biosciences; 610921; 1:1000), and anti-JAK2 (CST; 3230P; 1:1000). The proteins were visualized by chemiluminescence (Bio-Rad) and imaged with ImageQuant LAS4000. Band intensities were quantified with Fiji software. For this, rectangles of equal length and width were drawn around each protein band within the blot and used as the region of interest. The width of the rectangle was drawn to fit the width of the band, whereas the length of the rectangle was drawn to include 3 times the length of the band above and below the protein band of interest. An intensity profile plot was created using ImageJ for each rectangle. The intensity of each peak was measured from this plot, which allows for the background to be subtracted using the intensity of the regions above and below the protein band for each lane. For the quantification of phosphorylated SIRP␣ to total SIRP␣, the blots were probed Figure 8. Inhibition of Src and JAK family kinases prevents activity-dependent SIRP␣ ectodomain cleavage in neurons. A-C, hippocampal neurons were cultured to DIV12 and left untreated or were treated with KCl or KCl ϩ TIMPs for 2 h. The amount of ectodomain and total SIRP␣ (endogenous) at excitatory synapses (identified by VGluT1 staining) was examined as described in the legend to Fig. 6. B, quantification of the ratio of ectodomain/total SIRP␣. KCl-driven neuronal activation decreased the ratio of ectodomain/total SIRP␣. This decrease was blocked by TIMPs. C, total levels of SIRP␣ are not significantly different between all groups. D-G, hippocampal neurons were cultured to DIV12 and treated with either 10 M PP2, 10 M AG490, or DMSO, with or without 55 mM KCl for 2 h. The amount of ectodomain and total SIRP␣ at excitatory synapses was examined as described above. E, quantification of the ratio of ectodomain/total SIRP␣. AG490 and PP2 blocked the KCl-induced decrease in the ratio of ectodomain/total SIRP␣.
For the calculation of SIRP␣ ectodomain secretion (Fig. 5C), media and lysates were first blotted with the SIRP␣ N-Term antibody, stripped, and then blotted with the SIRP␣ C-Term antibody. The following SIRP␣ band intensities were quantified (Fig. 5C): Media (IB: N-Term, Media), Full-N (N-Term, Lysates), Full-C (C-Term, Lysates, top band), and C-Fragment (C-Term, Lysates, bottom band). The rate of SIRP␣ ectodomain cleavage was calculated using the formula, Media/((Full-C ϩ C-Fragment) ϫ (Full-N/Full-C))/hours. (Full-C ϩ C-Fragment), detected by SIRP␣ C-Term antibody, implies total SIRP␣. Because SIRP␣ in media was detected with SIRP␣ N-Term antibody, the ratio of Full-N/Full-C was used to normalize between two blots blotted with N-Term and C-Term antibodies.

Immunocytochemistry
Cultured neurons were fixed with 3% paraformaldehyde for 10 min at 37°C and blocked in 2% BSA, 2% normal goat serum, and 0.1% Triton X-100 for 1 h at room temperature. This was followed by incubation with primary antibodies for 3 h at room temperature or overnight at 4°C. The cultures were then washed with PBS, and secondary antibodies were applied for 1 h at room temperature. After being washed again with PBS, samples were mounted in glycerol with n-propyl gallate (Sigma). The following antibodies and dilutions were used: anti-VGluT1 (Millipore; AB5905; 1:4000), anti-GFP (Aves Labs; GFP-1020; 1:5000), anti-SIRP␣ (clone p84; BD; 1:200), anti-CaMKII-pT286 (CST; D21E4; 1:200). For surface staining, a similar protocol as described was used; however, Triton X-100 was omitted to prevent membrane permeabilization.

Image acquisition and analysis
Imaging was done with a confocal microscope (Zeiss LSM 700) using ϫ40 and ϫ60 objectives at a resolution of 1024 ϫ 1024 pixels. Images were taken as 16-bit, z-stack images with a 0.5-m step size. Identical settings for laser power, master gain, digital gain, and offset were used for all acquired images within each experiment.
For VGluT1 density analysis, transfected pyramidal-like neurons were selected for imaging at random. For analysis, images were stacked and merged using Fiji software, and VGluT1 puncta on transfected dendrites after the primary branch point (identified by GFP or YFP expression) were analyzed using MetaMorph (Molecular Devices). The quantification was done blind. Puncta that were within a region 0.3 m away from the dendritic shaft and spines were included for analysis. The staining intensity of VGluT1 in the dendritic shaft was subtracted as background. This background intensity was similar between conditions. Puncta smaller than 4 pixels (ϳ0.04 m 2 ) were excluded from analysis. Dendritic lengths were measured by manual tracing of analyzed dendritic lengths in Fiji.
For ectodomain/total SIRP␣ analysis, transfected pyramidal-like neurons (Fig. 6) or fields of neuronal cultures (Fig. 8) were selected for imaging at random. For analysis, images were stacked using Fiji software, and regions of interest (ROI) were selected as follows. 1) For analysis in Fig. 6, PSD95-EGFP-positive puncta were selected as the ROI, and 2) for analysis in Fig. 8, SIRP␣ (C-Term)-positive puncta with Ͼ10% overlap with VGluT1 puncta were selected as the ROI. The average intensity of SIRP␣ (C-Term) and SIRP␣ (N-Term) staining within the ROI was measured. 5-10 ROIs were selected at random per transfected neuron (Fig. 6) or per field of image (Fig. 8).

Statistical analysis
Statistical analyses were performed using GraphPad Prism software. Statistical tests used were the Kolmogorov-Smirnov test, Student's t test, and one-way ANOVA or two-way ANOVA, as indicated in the figure legends. In the case of oneway ANOVA, post hoc analysis was done with Tukey's test. In the case of two-way ANOVA, post hoc analysis was done with Sidak's test. All data are mean Ϯ S.E. In all figures, significance is indicated as follows: n.s., p Ͼ 0.05; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.005; ****, p Ͻ 0.001. Sample sizes were similar to those reported in previous publications (19,(41)(42)(43)52). Post hoc statistical calculation of sample sizes for imaging and electrophysiology was also done to ensure that sample sizes had sufficient power for subsequent statistical analyses (at least 80% power at the 0.05 level of significance for each set of experiments).