Though γ-secretase appears primarily in endosomes, scientists previously detected the protease at the plasma membrane (see Lah et al., 1997), and at synapses (see Georgakopoulos et al., 1999). More recent work showed that presenilin-1 (PS1), the catalytic center of γ-secretase, interacts with δ-catenin (Kosik et al., 2005), a cell adhesion regulator that senior author Edward Ziff and others have also reported at synapses (Silverman et al., 2007). This hinted to Restituito that PS1 does something important in those tiny spaces between neurons.

By immunofluorescence, her team found PS1 colocalizing with the post-synaptic marker PSD-95 in embryonic rat hippocampal neurons. PSD-95 also shows up with the PS1 substrates N-cadherin and EphRB2, as well as with three metalloproteases (aka “α-secretases”)—ADAM10, ADAM17, and MT5-MMP. α- and γ-secretase sequentially cleave a variety of transmembrane proteins including N-cadherin and EphRB2. In addition, the researchers detected γ-secretase components and substrates on Western blots of rat brain synaptosomes, and showed that the protease is enzymatically active there, cutting N-cadherin at pre- and post-synaptic areas.

Harking back to the lab’s earlier report of PS1 associating with δ-catenin, Restituito and colleagues found that the cell adhesion factor is, in fact, required to keep PS1 at the synapse. In brain extracts from δ-catenin knockout mice, PS1 was down, relative to wild-type levels, in post-synaptic sites, but not in other subcellular fractions, the researchers found.

If α- and γ-secretase reside at the synapse, could neuronal activity determine whether they snip their synaptic substrates? The literature to date is murky, suggesting synaptic stimulation can prevent (see Tanaka et al., 2000; Tai et al., 2007) or promote N-cadherin cleavage (see Marambaud et al., 2003; Uemura et al., 2006). The experiments in the current study add support for the latter—activation of cultured rat neurons with glutamate or NMDA drove up production of the C-terminal fragment (CTF1) formed when metalloproteases cut N-cadherin. When neuronal firing was dampened using the sodium channel blocker tetrodotoxin, levels of the N-cadherin fragment dropped. Src or JNK kinase inhibitors also blocked formation of this fragment, suggesting those signaling pathways operate in the NMDAR-mediated proteolysis. However, synaptic activity had little bearing on γ-secretase cleavage of CTF1, indicating that NMDA stimulation enhances N-cadherin processing by promoting the initial cut only. Just a small fraction of full-length N-cadherin gets cleaved, and what those C-terminal fragments do is not entirely clear, though accumulating evidence points toward a synaptic function. Of the little CTF1 that neurons do produce, much of it shows up in synapses. γ-secretase cleavage of CTF1 produces a smaller CTF2 fragment that represses transcription (Marambaud et al., 2003).

If synaptic activity promotes N-cadherin proteolysis, what happens if that cleavage is blocked? The researchers showed that γ-secretase or metalloprotease inhibitors enhanced synaptic transmission (measured by mini-excitatory post-synaptic currents) and increased levels of synaptic proteins needed for neurotransmission (judged by immunofluorescent staining of PSD-95, synaptotagmin, and the AMPA receptor subunit GluA2). Taken together, the data describe a “feedback loop” where synaptic activity activates proteolysis, which in turn modulates synaptic activity, Restituito told ARF.

Showing that γ-secretase and metalloproteases are “not just physically present at the synapse, but also influence other proteins known to be important for synaptic transmission—that is a key finding,” said Jane Sullivan of the University of Washington, Seattle. In a Nature Neuroscience paper published a few weeks ago, she and colleagues reported that PS1 regulates homeostatic scaling, a type of synaptic plasticity that operates in neuronal networks (ARF related news story on Pratt et al., 2011).

Kenneth Kosik of the University of California, Santa Barbara, finds the current paper novel in that it “gets at a global mechanism for the regulation of synaptic proteins. It shows that concentrations of these proteins may be linked to synaptic transmission through proteolytic activity [of γ-secretase and MMPs],” he told ARF.

Though the present study focused on N-cadherin, Restituito’s team has also looked at amyloid precursor protein (APP), the MMP/γ-secretase substrate that grabs much of the attention in AD. Preliminary data suggest that some APP may also be processed at synapses and regulated by synaptic activity, Restituito told ARF.

The relationship between Aβ and synaptic activity was examined in a study published earlier this month by John Cirrito of Washington University School of Medicine, St. Louis, Missouri (see ARF related news story on Verges et al., 2011). That work and the current paper show that NMDAR stimulation activates α-secretase. (ADAM10, ADAM17, and MT50MMP—the metalloproteases Restituito’s team detected at synapses—are α-secretases.) Cirrito found that α-secretase activation decreases Aβ production, whereas Restituito discovered it leads to N-cadherin-dependent inhibition of synaptic activity. “It would appear NMDARs may activate one pathway which then has multiple consequences,” Cirrito suggested in an e-mail to ARF (see full comment below).

Given prior reports that some PS1 FAD mutations cause loss of γ-secretase function (Marambaud et al., 2003; Georgakopoulos et al., 2006), the new data may link dysfunction in the γ-secretase/MMP proteolytic system with the synaptic abnormalities of familial AD, Anastasios Georgakopoulos, Mount Sinai School of Medicine, New York, suggested in an e-mail to ARF (see full comment below).

Ultimately, “how NMDARs, secretases, MMPs, cadherins, and Aβ function in various combinations to regulate synaptic activity will be important to parse apart, though the relationships are likely to be very complex,” Cirrito noted.—Esther Landhuis.

Reference:
Restituito S, Khatri L, Ninan I, Mathews PM, Liu X, Weinberg RJ, Ziff EB. Synaptic autoregulation by metalloproteases and gamma-secretase. J Neurosci. 24 Aug 2011;31(34):12083-12093. Abstract