1. Neuroscience
Download icon

APP and APLP2 interact with the synaptic release machinery and facilitate transmitter release at hippocampal synapses

  1. Tomas Fanutza
  2. Dolores Del Prete
  3. Michael J Ford
  4. Pablo E Castillo  Is a corresponding author
  5. Luciano D’Adamio  Is a corresponding author
  1. Albert Einstein College of Medicine, United States
  2. MS Bioworks, LLC, United States
Research Article
  • Cited 17
  • Views 1,537
  • Annotations
Cite this article as: eLife 2015;4:e09743 doi: 10.7554/eLife.09743

Abstract

The amyloid precursor protein (APP), whose mutations cause familial Alzheimer’s disease, interacts with the synaptic release machinery, suggesting a role in neurotransmission. Here we mapped this interaction to the NH2-terminal region of the APP intracellular domain. A peptide encompassing this binding domain -named JCasp- is naturally produced by a γ-secretase/caspase double-cut of APP. JCasp interferes with the APP-presynaptic proteins interaction and, if linked to a cell-penetrating peptide, reduces glutamate release in acute hippocampal slices from wild-type but not APP deficient mice, indicating that JCasp inhibits APP function.The APP-like protein-2 (APLP2) also binds the synaptic release machinery. Deletion of APP and APLP2 produces synaptic deficits similar to those caused by JCasp. Our data support the notion that APP and APLP2 facilitate transmitter release, likely through the interaction with the neurotransmitter release machinery. Given the link of APP to Alzheimer’s disease, alterations of this synaptic role of APP could contribute to dementia.

https://doi.org/10.7554/eLife.09743.001

eLife digest

Alzheimer’s disease has been linked to mutations in a gene encoding the amyloid precursor protein (APP). Mutations in this gene cause early onset Alzheimer’s disease in some families. Studies in animals suggest this mutant form of the protein may interfere with the messages sent between brain cells. But it remains unclear how this protein works even when it isn’t mutated.

One reason it has been difficult to figure out how APP works is because it is a long protein and is often cut into smaller, but still functional proteins. It is unclear which parts of the protein are involved in sending messages between brain cells, but previous research has allowed scientists to zero in on one stretch of the protein containing just 50 amino acids.

Now, Fanutza, Del Prete et al. reveal that this short stretch of APP interacts with the proteins that control the release of glutamate, one of the brain’s most important excitatory neurotransmitter. This stretch is cut from the full protein and released into the cell as a protein fragment (or peptide). Further experiments showed that this naturally occurring peptide interferes with the activity of APP, and in doing so reduces the release of glutamate from brain cells in mice. However, there was no effect on glutamate release when this peptide was introduced inside brain cells from mutant mice that lacked APP.

Together, the results show that this 50-amino acid stretch of APP promotes glutamate release between brain cells, and that a peptide containing the same stretch of the protein interferes with this process. Altered glutamate neurotransmission could explain the memory loss and thinking problems seen in early Alzheimer’s disease. Therefore, drug treatments that aim to restore normal glutamate transmission may prove a beneficial therapeutic strategy in Alzheimer’s disease.

https://doi.org/10.7554/eLife.09743.002

Introduction

Human genetic evidence and studies in App-knock-out (KO) mice indicate that the amyloid precursor protein (APP) plays an important physiological role in the central nervous system. In humans, a polymorphism in APP that reduces APP processing protects from sporadic Alzheimer’s disease (AD) and normal aging-dependent cognitive decline (De Strooper and Voet, 2012, Jonsson et al., 2012). In contrast, mutations in APP and in genes that regulate APP processing – such as PSENs and BRI2/ITM2B – cause familial AD (FAD) and the AD-like familial British dementia and familial Danish dementia (De Strooper, 2007, De Strooper et al., 2010, De Strooper and Voet, 2012, Garringer et al., 2010, Matsuda et al., 2005, Giliberto et al., 2008, Matsuda et al., 2009, Tanzi, 2012, Vidal et al., 1999, 2000). Analysis of App-KO mice has also suggested a synaptic role for APP (Korte et al., 2012, Marcello et al., 2012, Octave et al., 2013, Randall et al., 2010, Turner et al., 2003, Venkitaramani et al., 2007). For example, hippocampal App-KO neurons in culture show enhanced α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) receptor-mediated synaptic transmission and increased size of the readily releasable synaptic vesicle pool (RRP). In addition, long-term potentiation (LTP), a form of synaptic plasticity that underlies memory formation, is impaired in App-KO mice; and these LTP deficits are clearly noticeable in aged but not in juvenile App-KO mice (Dawson et al., 1999, Korte et al., 2012, Priller et al., 2006, Venkitaramani et al., 2007).

Although the evidence indicating that APP plays a role in several processes controlling synaptic function is strong, the molecular mechanisms are poorly understood. A confounding factor when studying APP function is the complex APP metabolism. The best-characterized cascade involves a sequential β/γ-secretase cleavage of APP. In this pathway, APP is cleaved by β-secretase/BACE1 into a soluble ectodomain (sAPPβ) and a membrane-bound COOH-terminal fragment of 99 amino acids (C99 or β C-terminal fragment [CTF]). C99/β-CTF is cleaved by the γ-secretase to produce Aβ peptides, which are believed to be the main culprit of AD, and the APP intracellular domain (AID/AICD). Alternatively, α-secretase cleaves APP in the Aβ sequence into sAPPα and a membrane-bound COOH-terminal fragment of 83 amino acids (C83 or α-CTF). C83/α-CTF is cleaved by the γ-secretase to produce a short peptide comprising the COOH-terminus of Aβ, called P3, and AID/AICD (Annaert et al., 2000, De Strooper, 2003, Selkoe and Kopan, 2003, Sisodia and St George-Hyslop, 2002, Vassar et al., 1999). Another processing pathway, potentially important yet less studied, involves cleavage of APP in the cytoplasmic domain – between D664 and A665, following the numbering of the main brain APP isoform of 695 amino acids – by caspase-6, -3, and -8. Caspases can in theory process all APP metabolites comprising this intracellular motif, such as full-length APP, C99/β-CTF, C83/α-CTF, and AID/AICD. So far, cleavages of full-length APP and AID/AICD by caspases have been described (Bertrand et al., 2001, Gervais et al., 1999, Lu et al., 2000, Madeira et al., 2005, Pellegrini et al., 1999, Weidemann et al., 1999). Processing of full-length APP generates a membrane bound NH2-terminal fragment called APP-Ncas and a COOH-terminal intracellular peptide called APP-Ccas or C31 (Bertrand et al., 2001, Gervais et al., 1999, Lu et al., 2000, Madeira et al., 2005, Pellegrini et al., 1999, Weidemann et al., 1999). Caspase cleavage of AID/AICD splits the cytoplasmic domain of APP into two soluble cytosolic peptides: the aforementioned Ccas/C31 and JCasp. The latter comprises the NH2-terminal peptide segment of the APP intracellular domain (Bertrand et al., 2001, Madeira et al., 2005). Recently, a new secretase activity (called η-secretase) has been described (Willem et al., 2015). Cleavage by η-secretase occurs primarily at amino acids 504–505 of APP695 and releases a ~500 amino acid-long ecto-domain and a membrane-bound fragment called CTF-η. Interestingly, CTF-η is further processed by α- and β-secretases to release a long and a short Aη peptide (Aη-α and Aη-β, respectively).

Many of the APP metabolites produced by these cleavages are biologically active. AID/AICD modulates programmed cell death, gene transcription and Ca2+ homeostasis (Baek et al., 2002, Cao and Südhof, 2001, Checler et al., 2007, Cupers et al., 2001, Hamid et al., 2007, Kim et al., 2003, Leissring et al., 2002, Liu et al., 2007, Madeira et al., 2005, Pardossi-Piquard et al., 2005, Passer et al., 2000, von Rotz et al., 2004). The caspase-derived APP fragments Ccas/C31 and JCasp (Bertrand et al., 2001, Lu et al., 2000, Madeira et al., 2005) also possess toxic activities. Aβ both regulates normal synaptic transmission and prompts deficits in synaptic plasticity (Fogel et al., 2014, Mota et al., 2014, Puzzo and Arancio, 2013, Puzzo et al., 2008, Shankar et al., 2007, Sivanesan et al., 2013, Um et al., 2012). sAPPα has neurotrophic functions and is involved in synaptic plasticity (Hick et al., 2015, Milosch et al., 2014). sAPPβ promotes axon pruning and neurodegeneration (Nikolaev et al., 2009). Aη-α can reduce LTP (Willem et al., 2015). β-CTF mediates long-term synaptic plasticity deficits, behavioral deficits and neurodegeneration (Neve et al., 1996, Neve and Robakis, 1998, Oster-Granite et al., 1996, Tamayev et al., 2012b, 2012c).

The presence of the functionally redundant APP-like protein-1 and -2 (APLP1 and APLP2) is another confounding factor. APLP1 and APLP2 are processed like APP (Heber et al., 2000, Herms et al., 2004, Müller and Zheng, 2012, Scheinfeld et al., 2002, von Koch et al., 1997) and release intracellular peptides – called APLP2-intracellular domain (ALID) 1 and ALID2, respectively – that, like AID/AICD, can potentially regulate transcription (Scheinfeld et al., 2002, 2003). The evidence that Aplp2-KO and App-KO mice are viable, whereas combined App/Aplp2-dKO mice develop neuromuscular junction deficits and die shortly after birth (Wang et al., 2005) vividly illustrates the functional redundancy of APP and APLP2. The robust phenotype of App/Aplp2-dKO mice can be used to map the functional domains of APP by expressing App knock-in mutations on an Aplp2-KO genetic background. This in vivo functional mapping has identified an essential role for the highly conserved APP intracellular domain in the patterning of neuromuscular junction and survival (Barbagallo et al., 2011a, 2011b, Li et al., 2010), indicating that the intracellular region of APP is functionally important in vivo

The APP cytoplasmic domain is very short (~50 amino acids) and does not possess recognizable enzymatic domains, raising the possibility that APP functions could be dictated by molecular interaction of the APP cytoplasmic domain with intracellular proteins. Using an unbiased proteomic approach, we found that the brain APP intracellular domain-interactome includes synaptic vesicles trans-membrane proteins, proteins associated with synaptic vesicles and the presynaptic active zone (Del Prete et al., 2014). Many of these proteins, such as cis- and trans-SNAREs, complexin, and the Ca2+ sensors synaptotagmins, function as key regulators of synaptic vesicles exocytosis (McMahon and Boucrot, 2011, Rizzoli, 2014, Royle and Lagnado, 2010, Südhof, 2012, 2013, Südhof and Rizo, 2011, Xu et al., 2009). Remarkably a similar brain APP–presynaptic interactome was identified in transgenic mice overexpressing human APP (Kohli et al., 2012, Norstrom et al., 2010). These findings are consistent with the evidence that APP in the brain is found at presynaptic terminals and synaptic vesicles (Del Prete et al., 2014, Groemer et al., 2011, Laßek et al., 2014, Wilhelm et al., 2014, Yang et al., 2005).

The fact that APP is present at synaptic vesicles and the presynaptic active zone – specialized region of the presynaptic membrane where neurotransmitter release occurs – and that the APP intracellular domain interacts with the neurotransmitter release machinery suggests that APP regulates transmitter release. Here, we investigated this proposition. By recording neuronal activity in hippocampal brain slices while using and APP-derived peptide that interferes with the APP-neurotransmitter release machinery interaction and acts as a dominant negative inhibitor of APP, we implicate APP in the physiological regulation of the synaptic release of glutamate at hippocampal excitatory synapses. Moreover, we found that APLP2 can compensate for loss of APP function and that genetic ablation of both APP and APLP2 protein expression produces glutamatergic synaptic transmission deficits similar to those caused by JCasp, confirming that JCasp acts as a dominant negative inhibitor of some functions of APP in wild-type (WT) animals. Our results indicate that APP and APLP2 fine-tune synaptic vesicle release likely through the interaction with, and regulation of, the molecular machinery that controls synaptic vesicle exocytosis.

Results

APP interacts with presynaptic proteins via the NH2-terminal portion of its intracellular domain

Given that the intracellular domain of APP interacts with proteins that mediate synaptic vesicle exocytosis (Del Prete et al., 2014), APP could regulate transmitter release. To test this hypothesis, one should target those functions associated with these presynaptic APP interactions, possibly without directly interfering in the function of other APP domains/regions, such as Aβ or the long APP extracellular domain and derived products (sAPPα and sAPPβ). As an initial step toward the generation of biological tools to test the hypothesis, we mapped the APP intracellular domain that interacts with the exocytosis machinery, and used a proteomic approach to identify the brain proteins that interact with the N-terminus (JCasp) and C-terminus (Ccas) sub-domains of the APP intracellular region. JCasp and Ccas are two naturally occurring fragments that are generated by a double γ-secretase/caspases cleavage of APP (Figure 1A and B) (Bertrand et al., 2001, Gervais et al., 1999, Lu et al., 2000, Madeira et al., 2005, Pellegrini et al., 1999, Weidemann et al., 1999). It is therefore conceivable that this physiological proteolytic event has been selected to separate two functionally distinct intracellular regions of APP. In line with this thought, presynaptic interactors of APP bind to the NH2-terminal intracellular region JCasp. Other APP-binding proteins, such as the APP interactor Ddb1 (Watanabe et al., 1999) and the Ddb1-interacting proteins Cul4A and Cul4B (O’Connell and Harper, 2007), specifically bind the COOH-terminal APP intracellular region Ccas (Table 1, Source Data were uploaded to the PRIDE repository – European Bioinformatics Institute (EBI), Cambridge, UK – with Proteome Xchange identifier #48018), underscoring the reliability of the mapping method. In addition, JCasp, but not ScJCasp, which contains the same amino acids of JCasp but in a scrambled order, competes for the binding of Syp, Vamp2, and Syt2 to St-AID, supporting the notion that JCasp contains the binding domain of APP for these presynaptic proteins (Figure 1B and C).

The AID–presynaptic interactome binds to JCasp.

(A) The APP intracellular domain, composed by the two sub-regions JCasp and Ccas, is always cytosolic whether APP molecules are localized at synaptic vesicles or the active zone. (B) Sequences of AID, JCasp, and Ccas used as baits in the proteomic experiments. (C) Brain lysates were incubated for 30 min with 10 μM of either JCasp or ScJCasp. After incubation, the lysates were affinity purified on StrepTactin columns bound to the indicated St-peptides. Bound proteins were analyzed by Western blot. JCasp reduces binding of Syp, Vamp2, and Syt2, but not Ddb1, to St-AID. (D) Triplicate experiment showing that JCasp, but not ScJCasp, significantly reduces binding of Syp, Vamp2, and Syt2 to St-AID. (E) Quantification of the data shown in (D). (F) Hippocampal slices from 8-week-old WT or App-KO mice were stained with an αNeuN (red, left panels) and an αAPP (green middle panel) antibody. The images were merged in the right panels. APP is widely distributed in the hippocampus. The strongest staining is seen in the Hi, MF, CA3, and CA1. The staining for APP is specific since there is no signal in App-KO hippocampi. (G) Staining for APP (red, left panels) and Bassoon (green middle panel) show that APP is expressed in the Stratum radiatum where it partially co-localizes with Bassoon (arrows in the merged image, right panel). Again, the staining for APP and the co-localization spots are specific as shown by their absence in App-KO hippocampi. AID, APP intracellular domain; APP, amyloid precursor protein; WT, wild type. 

https://doi.org/10.7554/eLife.09743.003
Table 1

Proteins that bind JCasp are in redbold. Proteins that bind Ccas are in bluethe other entries. Source Data containing all search results, coverage maps, peptide lists and product ion data were uploaded to the PRIDE repository (European Bioinformatics Institute (EBI), Cambridge, UK) with Proteome Xchange identifier #48018. Details of protein identification data related to the proteins discussed in this manuscript can be found in Table1. The table file contains: the list of proteins identified (Column 1); the database accession numbers (Column 2); the spectral counts (SpC, Columns 3-7); Spectral Abundance Factor (SAF, Columns 8-12); subsequent Normalized Spectral Abundance Factor (NSAF, Columns 13-17). This was based on the equation: NSAF = (SpC/MW)/Σ(SpC/MW)N, where SpC = Spectral Counts, MW = Protein MW in kDa and N = Total Number of Proteins, NSAF contains NSAF values without alone.

https://doi.org/10.7554/eLife.09743.004
ST ST-AID ST ST-Ccas ST-JCasp ST ST-AID ST ST-Ccas ST-JCasp ST ST-AID ST ST-Ccas ST-JCasp
Prot.Acc. Num.SpC SpC SpC SpC SpC SAFSAFSAFSAFSAFNSAFNSAFNSAFNSAFNSAF
Ddb1 sp|Q3U1J4| 0 418 0 730 0   0.0000 3.2913 0.0000 5.7480 0.0000   0.0000 0.0139 0.0000 0.0366 0.0000
Syt1sp|P46096|088001130.00001.87230.00000.00002.40430.00000.00790.00000.00000.0129
Stxbp1sp|O08599|04408460.00000.64710.00000.11760.67650.00000.00270.00000.00070.0036
Cul4a sp|Q3TCH7| 0 40 0 36 0   0.0000 0.4545 0.0000 0.4091 0.0000   0.0000 0.0019 0.0000 0.0026 0.0000
Sv2bsp|Q8BG39|05500660.00000.71430.00000.00000.85710.00000.00300.00000.00000.0046
Syngr3sp|Q8R191|04006790.00001.60000.00000.24003.16000.00000.00680.00000.00150.0170
Cul4b sp|A2A432| 0 19 0 15 0   0.0000 0.1712 0.0000 0.1351 0.0000   0.0000 0.0007 0.0000 0.0009 0.0000
Stx1bsp|P61264|02300310.00000.69700.00000.00000.93940.00000.00290.00000.00000.0050
Vamp2sp|P63044|02300460.00001.76920.00000.00003.53850.00000.00750.00000.00000.0190
Syngr1sp|O55100|0800300.00000.30770.00000.00001.15380.00000.00130.00000.00000.0062
Sv2asp|Q9JIS5|01101230.00000.13250.00000.01200.27710.00000.00060.00000.00010.0015
Snap25sp|P60879|016012350.00000.69570.00000.52171.52170.00000.00290.00000.00330.0082
Stx1asp|O35526|0200210.00000.06060.00000.00000.63640.00000.00030.00000.00000.0034
Stx7sp|O70439|030040.00000.10000.00000.00000.13330.00000.00040.00000.00000.0007
Syt2sp|P46097|01500360.00000.31910.00000.00000.76600.00000.00130.00000.00000.0041
Syt12sp|Q920N7|0300100.00000.06380.00000.00000.21280.00000.00030.00000.00000.0011
Stx12sp|Q9ER00|060050.00000.19350.00000.00000.16130.00000.00080.00000.00000.0009

These interactions suggest a presynaptic function of APP in vivo. This hypothesis is substantiated by a number of studies showing localization of APP at presynaptic terminals and synaptic vesicles. Biochemical studies have shown that endogenous APP is present in fractions highly enriched in synaptic vesicles and the active zone (Del Prete et al., 2014, Groemer et al., 2011, Laßek et al., 2013, Lundgren et al., 2015). Immune-precipitation of APP from mouse brains showed interaction of APP with key regulators of synaptic vesicles exocytosis. Although the possibility that these interactions happen post-cellular lysis cannot be formally discarded, these data are consistent with a presynaptic localization of APP in vivo. Immuno-electron microscopy experiments (performed using the specific Y-188 anti-APP antibody, see below) have indicated that APP is localized at presynaptic terminals and in synaptic vesicles in the hippocampus (Del Prete et al., 2014). Finally, localization of APP at presynaptic terminals has been reported also in primary neuronal cultures where it can regulate synaptic transmission functioning as a presynaptic receptor for Aβ (Fogel et al., 2014).

The data discussed above support a presynaptic function of APP. Next, we analyzed the pattern of expression of APP in the hippocampus, a brain structure critically involved in memory formation. We revisited this question in view of a recent report showing that most of the antibodies that have been used to determine the localization of endogenous APP are non-specific as they label App-KO and WT cells/tissues in a similar fashion (Guo et al., 2012). The authors showed that Y188 was the only APP-specific antibody. The evidence that Y188 does not stain App-KO hippocampi confirmed the specificity of this antibody (Figure 1E and F). In addition, low magnification analysis shows that APP is expressed, albeit at different levels, in all hippocampal circuits, including the CA1 neurons and the stratum radiatum containing the CA3 projections (Schaffer collaterals, or SC) to CA1 neurons (Figure 1E). Higher magnification analysis confirms that APP staining is stronger in CA1 pyramidal neurons as compared to the stratum radiatum (Figure 1F). Several factors may explain these differences: 1) APP is synthesized and undergoes maturation in the endoplasmic reticulum (ER) and Golgi that are highly concentrated in cell bodies; 2) APP may be expressed at synapses at levels that are, albeit functionally important, too low to be clearly detected by Y188; 3) Y188 recognizes the C-terminal epitope of APP, and this epitope may be functionally engaged in multi-molecular interactions at presynaptic terminals and may therefore be scarcely accessible to Y188. Nevertheless, the staining shown in Figure 1F indicates that APP is indeed expressed in the stratum radiatum, where it partially co-localizes with the presynaptic protein Bassoon.

Intracellular delivery of JCasp decreases excitatory synaptic transmission

This presynaptic APP-interacting network can stem from APP molecules localized at synaptic vesicles, presynaptic membranes or both, and it is likely formed via the direct interaction of the intracellular domain of APP with some release machinery protein. The other components are most likely indirectly associated to APP (Figure 2A). Since JCasp interferes with these interactions in vitro, if delivered inside neurons, JCasp could hamper the function of this presynaptic APP–interacting complex in vivo (Figure 2B) and reveal the function of this intracellular APP domain.

Pen1-JCasp is delivered intra-neuronally.

Models of assembly of the amyloid precursor protein (APP) presynaptic interactome and mechanism of action of Pen1-JCasp. (A) SV = synaptic vesicles; AZ = active zone. SV proteins and proteins associated to SV and to the AZ bind via direct and indirect interactions with the JCasp region of APP molecules localized to either SV or the AZ. The interactions depicted here are hypothetical. (B) Pen1-JCasp is delivered intracellularly and can interfere with the interaction of APP with presynaptic proteins. (C) Immune-fluorescence experiment showing that Pen1 delivers the JCasp peptide in hippocampal neurons.

https://doi.org/10.7554/eLife.09743.005

Thus, we systematically characterized the effect of JCasp on excitatory synaptic transmission at SC–CA1 pyramidal cell synapses. To provide intracellular delivery, we synthesized JCasp and the negative control peptide ScJCasp linked to the cell-penetrating peptide (Davidson et al., 2004) Penetratin1 (Pen1) at the N-terminus (Figure 2B), a strategy that we have previously used to deliver peptides to murine hippocampal neurons in vivo (Tamayev et al., 2012c). First, we verified whether Pen1-JCasp efficiently delivers JCasp at hippocampal SC–CA1 synapses. To this end, we incubated hippocampal slices with 1 μM of either biotinylated JCasp or biotinylated Pen1-JCasp. Later, 5 hr after incubation, slices were stained for the peptide using Streptavidin. As shown in Figure 2C, Pen1-JCasp, but not JCasp, is detected in neurons of the CA1 pyramidal neurons and in the stratum radiatum, which contains the SCs. Thus, we determined whether Pen1-JCasp alters synaptic transmission at SC–CA1 synapses. The experiments were performed as follows: soon after preparation the brain slices were incubated in 200 ml of artificial cerebrospinal fluid (ACSF) containing the indicated concentrations of either Pen1-JCasp or the negative control peptide Pen1-ScJCasp. Recordings were performed between ~5 and 8 hr after preparation of slices and incubation with Pen1 peptides. We found that Pen1-JCasp impaired basal synaptic transmission, as indicated by a reduction in the input/output (I/O) function (Figure 3A). These effects were dose-dependent and specific since the negative control peptide Pen1-ScJCasp did not alter synaptic transmission.

Pen1-JCasp impairs synaptic plasticity.

(A) CA1 recordings of hippocampal slices incubated with either artificial cerebrospinal fluid (ACSF), 1 μM Pen1-ScJCasp or Pen1-JCasp at the indicated concentrations. The synaptic input/output (I/O) relationship was obtained by plotting the fiber volley amplitude against the initial slope of the evoked field excitatory postsynaptic potential (fEPSP). Representative traces are shown on top. The slope values of I/O recording for each group were compared for statistical assessments. It has been found that 1 μM Pen1-JCasp significantly reduces synaptic transmission as compared to ACSF (*), Pen1-ScJCasp (*), 10 nM (*), and 100 nM Pen1-JCasp (*p). Representative traces are shown. (B) Average paired-pulse facilitation (PPF) (2nd fEPSP/1st fEPSP) plotted as a function of the inter-stimulus interval. Representative traces of fEPSPs evoked at 40 ms inter-stimulus interval are shown. Pen1-JCasp increases PPF. (C) Synaptic facilitation elicited by stimulus trains at 1, 5, 10, and 20 Hz. fEPSP slopes are normalized to the slope of the first fEPSP of the stimulus train. Representative traces of fEPSPs evoked at 20 Hz are shown. Stimulus artifacts are removed for clarity. Pen1-JCasp increases frequency facilitation (FF) in a frequency and dose-dependent manner. Statistical assessments were performed by: one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test for slopes of I/O curves; two-way repeated measures (RM) ANOVA followed by Tukey’s multiple comparisons test for PPF and FF (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). In FF, 10 nM Pen1-JCasp increased facilitation in a statistically significant manner only at the 10th stimulation (indicated with (*) and (*)). The number of recordings and the number of mice analyzed for each group are shown in (C). All data represent means ± SEM. The complete statistical analyses are shown in the attached Excel file.

https://doi.org/10.7554/eLife.09743.006

To test whether the reduction in I/O function could be due to presynaptic impairment, we first estimated the probability of release (Pr) by analyzing two forms of short-term synaptic plasticity: paired-pulse facilitation (PPF) and synaptic frequency facilitation (FF) in response to a stimulus train (10 stimuli at either 1, 5, 10, or 20 Hz). These forms of short-term synaptic plasticity are determined, at least in part, by changes in Pr, such that a decrease in Pr leads to an increase in facilitation and vice versa (Zucker and Regehr, 2002). Pen1-JCasp but not Pen1-ScJCasp significantly increased both PPF (Figure 3B) and FF (Figure 3C) in a dose-dependent manner. In addition, Pen1-JCasp reduced miniature excitatory postsynaptic current (EPSC) (mEPSC) frequency (Figure 4A), but had no effect on mEPSC amplitude (Figure 4B) or waveform (Figure 4C), suggesting a presynaptic mechanism of action. We also delivered a high-frequency train of stimuli to deplete the RRP. The rate of RRP depletion, as measured by the suppression of synaptic transmission during repetitive stimulation, is proportional to the initial Pr (i.e., a reduction in Pr would decrease this rate and vice versa, an increase in Pr would accelerate it). Indeed, Pen1-JCasp slowed down the rate of synaptic suppression induced by a 500 stimuli, 28 Hz train (Figure 5A and B). After depletion, the recycling of synaptic vesicles replenishes the RRP. We found that the recovery phase after depletion – measured by stimulating at 2 Hz – was unaffected by Pen1-JCasp (Figure 5C and D), arguing against an alteration of synaptic vesicle recycling as a cause of the reduced synaptic strength. Taken together, our findings strongly suggest that Pen1-JCasp reduces basal synaptic transmission by interfering with glutamate release.

Pen1-JCasp reduces frequency of excitatory miniature currents.

(A) Cumulative probability of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) mediated miniature excitatory postsynaptic current (EPSC) (mEPSC) inter-event intervals. (B) Cumulative probability of AMPAR-mediated mEPSC amplitudes. Insets in cumulative probability graphs represent average mEPSC frequency (A) and amplitudes (B). mEPSC frequency was significantly reduced by Pen1-JCasp. (C) Average mEPSC of the two groups were not significantly different. (D) Representative recording traces of miniature EPSCs are shown. The number of recordings and the number of mice analyzed for each group are shown in (A). All data represent means ± SEM. Statistical assessments were performed by paired t test (**p < 0.01).

https://doi.org/10.7554/eLife.09743.013
Pen1-JCasp reduces the rate of depletion of glutamatergic vesicles.

(A) Artificial cerebrospinal fluid (ACSF) vs Pen1-JCasp, p < 0.05; Pen1-SCJCasp vs. Pen1-JCasp, p < 0.01; (B) ACSF vs. Pen1-JCasp, p< 0.01; Pen1-SCJCasp vs. Pen1-JCasp, p < 0.01. Recovery kinetics (C and D) are not affected. The experiments were performed in 3-month-old C57Bl/6j (A and C) and C57Bl/6j-129 hybrid mice (B and D). All experiments shown in this paper were performed in C57Bl/6j mice, except for the experiments shown in Figure 7; App-KO and wild-type (WT) littermate animals were on a C57Bl/6j-129 hybrid genetic background. Data represent means ± SEM. Statistical assessments were performed by nonlinear regression curve fit, one phase decay.

https://doi.org/10.7554/eLife.09743.015

Given that several APP metabolites have been involved in a wide range of neuronal functions, including synaptic plasticity, Pen1-JCasp could reduce synaptic transmission indirectly by altering APP processing. To address this possibility, we tested the effect of Pen1-JCasp on APP processing. Several soluble APP metabolites resulting from APP processing, such as APPβ, sAPPα, and Aβ, are diluted into the large volume (200 ml) of ACSF used for culturing hippocampal slices during recordings. Therefore, it is difficult to accurately measure the levels of these soluble APP metabolites. However, other APP metabolites, that is, β-CTF and α-CTF, remain membrane bound and can be measured by Western blot, together with full-length APP. Pen1-JCasp does not change levels of full-length APP, β-CTF, and α-CTF (Figure 6A and B), arguing against the possibility that Pen1-JCasp alters synaptic plasticity via modulation of APP processing.

Pen1-JCasp does not alter amyloid precursor protein (APP) processing.

(A) Western blot analysis of Actin, APP, β C-terminal fragment (CTF) and α-CTF in hippocampal slices incubated for 8 hr with either 1 μM Pen1-JCasp or 1 μM Pen1-ScJCasp. β-CTFp and α-CTFp are phosphorylated forms of β-CTF and α-CTF, respectively. (B) Quantification of the data: for each protein, the average signal obtained with 1 μM Pen1-ScJCasp is arbitrarily considered 1; the average signal obtained with 1 μM Pen1-JCasp is expressed as a percent of the control. The * indicates an aspecific band, which is also present in App-KO lysates (not shown).

https://doi.org/10.7554/eLife.09743.018

Based on our findings, we hypothesized that JCasp reduces transmitter release by interfering with the intracellular interactions of APP with key regulators of vesicle exocytosis. To test this hypothesis we compared the effects of Pen1-JCasp and JCasp side-by-side . Pen1-JCasp, but not JCasp, which lacks the cell penetrating properties of Pen1 and is not detected inside neurons (Figure 2C), reduced basal synaptic transmission (Figure 7A) and increased PPF (Figure 7B) and FF (Figure 7C). Thus, to suppress transmitter release, the JCasp peptide must access the intracellular compartment.

Pen1-JCasp impairs excitatory synapses via an intracellular mechanism.

Pen1-JCasp, but not JCasp, reduces basal synaptic transmission (A) paired-pulse facilitation (PPF) (B) and frequency facilitation (FF) at 20 Hz (C). Representative traces are shown. Stimulus artifacts are removed from the FF traces for clarity. The number of recordings and of mice analyzed are shown in (B). Statistical assessments were performed by: one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test for slopes of input/output (I/O) curves; two-way repeated measures (RM) ANOVA followed by Tukey’s multiple comparisons test for PPF and FF. *Pen1-JCasp 1 μM vs. artificial cerebrospinal fluid (ACSF), *Pen1-JCasp 1 μM vs. JCasp; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. The complete statistical analyses are shown in the attached Excel file. All data represent means ± SEM.

https://doi.org/10.7554/eLife.09743.019

A recent report shows that Aβ in cultured neurons triggers APP dimerization that, by interacting via the intracellular domain with Gαoβ1γ2 protein, increases Ca+2 influx through Cav2 channels, thereby enhancing Pr (Fogel et al., 2014). Thus, Pen1-JCasp could reduce Pr by inhibiting this pathway and reducing Ca+2 influx. However, we found that Pen1-JCasp was equally effective in the presence of 100 μM Cd+2, a manipulation that non-selectively blocks Ca+2 influx through voltage-gated Ca+2 channels (Figure 8).

Pen1-JCasp reduces the frequency of miniature excitatory postsynaptic currents (EPSCs) (mEPSCs) independently of alterations of Ca+2 influx.

(A) Cumulative probability of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) mediated mEPSC inter-event intervals. (B) Cumulative probability of AMPAR-mediated mEPSC amplitudes. Insets in cumulative probability graphs represent average mEPSC frequency (A) and amplitudes (B). mEPSC frequency was significantly reduced by Pen1-JCasp in the presence of Cd+2. (C) Average mEPSC of the two groups were not significantly different. (D) Representative recording traces of miniature EPSCs are shown. The number of recordings and the number of mice analyzed for each group are shown in (A). All data represent means ± SEM. Statistical assessments were performed by paired t test (*p < 0.05).

https://doi.org/10.7554/eLife.09743.023

JCasp inhibits the function of endogenous APP

Next, we determined whether JCasp has any effect on SC–CA1 synaptic transmission in App-KO and WT littermates. Interestingly, Pen1-JCasp reduced I/O function (Figure 9A) and increased PPF (Figure 9B) and FF (Figure 9C) in WT but not in App-KO animals, suggesting that Pen1-JCasp does not alter transmission at APP-lacking synapses.

Amyloid precursor protein (APP) is required for the inhibitory effect of Pen1-JCasp on synaptic transmission.

(A) Analysis of slopes of each input/output (I/O) recording by one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test showed no significant differences among the groups. However, uncorrected Fisher’s LSD test shows that Pen1-JCasp reduces basal synaptic transmission in wild-type (WT) hippocampal slices as compared to WT slices treated with Pen1-ScJCasp. Pen1-JCasp increases paired-pulse facilitation (PPF) (B) and frequency facilitation (FF) (C) in WT but not App-KO excitatory Schaffer collateral (SC) synapses. Representative traces are shown on the right of summary plots. The number of recordings and of mice analyzed for each group are shown in (A). All data represent means ± SEM. Statistical assessment by two-way repeated measures (RM) ANOVA shows significant differences in PPF and FF between WT mice treated with Pen1-JCasp and the other three experimental groups (*, vs. WT+Pen1-SCJCasp; *, vs. App-KO+Pen1-SCJCasp; *, vs. App-KO+Pen1-JCasp; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). The complete statistical analyses are shown in the attached Excel file.

https://doi.org/10.7554/eLife.09743.025

The fact that Pen1-JCasp-mediated suppression of synaptic transmission requires APP expression not only indicates that Pen1-JCasp acts as a dominant negative inhibitor of endogenous APP but also suggests that endogenous APP regulates glutamate release at excitatory synapses. However, we found no differences in synaptic facilitation between App-KO and WT hippocampi (Figure 9B and C), suggesting that compensatory mechanisms operate in App-KO mice to counterbalance the loss of APP function. Good candidates for these redundant functions are APLP1 and APLP2, the other two members of the APP protein family; particularly APLP2 since functional redundancy between APP and APLP2 has already been demonstrated (Barbagallo et al., 2011b, Dawson et al., 1999, von Koch et al., 1997, Wang et al., 2005). Of note, APLP1 and APLP2 have also been localized to presynaptic terminals (Laßek et al., 2013), further supporting a compensatory function at presynaptic terminals.

Loss of APP and APLP2 reduces excitatory synaptic transmission

As a first step to determine whether APLP2 shares this presynaptic function of APP, we characterized the brain interactome of the ALID2 using the same unbiased proteomic approach used for Ccas and JCasp. We synthesized two peptides: control Strep-tag only peptide and St-ALID2. Interestingly, the brain interactome of ALID2 comprises most of the key proteins of the presynaptic release machinery that bind to JCasp and AID (Table 2, Source Data were uploaded to the PRIDE repository – European Bioinformatics Institute (EBI), Cambridge, UK – with Proteome Xchange identifier #58725).

Table 2

Interaction of ALID2 with presynaptic protein. Source Data containing all search results, coverage maps, peptide lists and product ion data were uploaded to the PRIDE repository (European Bioinformatics Institute (EBI), Cambridge, UK) with Proteome Xchange identifier #58725. The table file contains: the list of proteins identified (Column 1); the database accession numbers (Column 2); the spectral counts (SpC, Columns 3-7); Spectral Abundance Factor (SAF, Columns 8-12); subsequent Normalized Spectral Abundance Factor (NSAF, Columns 13-17). This was based on the equation: NSAF = (SpC/MW)/Σ(SpC/MW)N, where SpC = Spectral Counts, MW = Protein MW in kDa and N = Total Number of Proteins, NSAF contains NSAF values without alone.

https://doi.org/10.7554/eLife.09743.029
ST ST-ALID2 ST ST-ALID2 ST ST-ALID2
Prot.Acc. Num.SpCSpCSAFSpCNSAFSpC
Syt1sp|P46096|0290.00000.32350.00000.0045
Stxbp1sp|O08599|0220.00000.64710.00000.0027
Sv2bsp|Q8BG39|0150.00000.19480.00000.0034
Syngr3sp|Q8R191|070.00000.280.00000.0067
Stx1bsp|P61264|060.00000.12120.00000.0029
Vamp2sp|P63044|0110.00000.84610.00000.0203
Syngr1sp|O55100|060.00000.88460.00000.0055
Sv2asp|Q9JIS5|0140.00000.40190.00000.0023
Syt2sp|P46097|0180.00000.38290.00000.0053
Syphsp|Q62277|080.00000.23530.00000.0049

The preceding data are consistent with the idea that APLP2, via its intracellular domain, can compensate the function of the cytoplasmic domain of APP in App-KO synapses. Yet, the evidence that Pen1-JCasp does not alter synaptic transmission in App-KO synapses (Figure 9) indicates that Pen1-JCasp acts as an inhibitor of the intracellular domain of APP but not APLP2 in vivo. Consistent with this possibility JCasp and the corresponding APLP2 region (RKRQYGTISHGIVEVD) are only 63% identical, and JCasp does not interfere with the interaction of ALID2 with Vamp2, Syt2, and Syp (Figure 10). Thus, Pen1-Jcasp is a tool to selectively study the function of APP.

JCasp does not compete for the interaction of amyloid precursor protein (APP) like protein-2 (APLP2) intracellular domain (ALID2) with Syt2, Syp, and Vamp2.

Brain lysates were incubated for 30 min with 10 μM of either JCasp or ScJCasp. After incubation, the lysates were affinity purified on StrepTactin columns bound to the indicated St-peptides. Bound proteins were analyzed by Western blot. JCasp does not reduce binding of Syp, Vamp2, and Syt2 to St-ALID2. The graphs at the bottom of the figure show quantitative analysis of the Western blot data.

https://doi.org/10.7554/eLife.09743.030

To directly test whether APLP2 can compensate for loss of APP function, we monitored the effect of constitutive APP, APLP2, and APP+APLP2 deletion on excitatory synaptic transmission at SC–CA1 synapses in acute brain slices. Because App/Aplp2-dKO mice die within the first 4 weeks of age, for comparison reasons all experiments were performed on 16- to 24-day-old mice of both sexes. Animals were on a mixed C57BL/6J and 129 genetic background. Aplp2-KOand App/Aplp2-dKO showed reduced basal synaptic transmission as compared to both WT and App-KO mice; however, these differences reached statistical significance only when compared to App-KO samples (Figure 11A).

Amyloid precursor protein (APP) like protein-2 (APLP2) compensates for loss of APP function.

(A) Input/output (I/O) recording showed significant differences between different genotypes (App-KO vs. Aplp2-KO, **p < 0.01; App-KO vs. App/Aplp2-dKO, *p < 0.05). PPR (B) and frequency facilitation (FF) (C) are significantly increased in App/Aplp2-dKO as compared to all other genotypes. Representative traces are shown on the right of summary plots. The number of recordings and of mice analyzed for each group are shown in (A). Statistical assessment by two-way repeated measures analysis of variance (RM ANOVA) followed by Tukey’s multiple comparisons tests. All data represent means ± SEM (*p < 0.05; **p < 0.01; ***p < 0.001).

https://doi.org/10.7554/eLife.09743.031

The App/Aplp2-dKO mice showed an increase in PPF (Figure 11B) and FF (Figure 11C), suggesting a reduction in Pr (Zucker and Regehr, 2002). The magnitude PPF and FF in App-KO and Aplp2-KO SC–CA1 synapses was not significantly different than in WT mice (Figure 11B and C).

Next we monitored mEPSCs. Interestingly, we found that mEPSC frequency was decreased by genetic ablation of APP and APLP2 expression, with APLP2 deletion having a greater effect as compared to APP deletion (Figure 12A, WT>App-KO>Aplp2-KO>App/Aplp2-dKO). However, these differences reached statistical significance only between WT Vs. Aplp2-KO and WT Vs. App/Aplp2-dKO samples. Thus, our results showing changes in short-term plasticity and mEPSC activity strongly suggest that while APP and APLP2 have redundant effects on glutamate release, APLP2 compensates for loss of APP function in App-KO mice more efficently than APP compensates loss of APLP2 function in Aplp2-KO hippocampi. Remarkably, the synaptic deficits observed in App/Aplp2-dKO SC–CA1 excitatory synapses are strikingly similar to those induced by JCasp in WT hippocampi, supporting the notion that a common mechanism underlies these deficits.

Reduced frequency of miniature excitatory postsynaptic currents (EPSCs) (mEPSCs) in Aplp2-KO and App/Aplp2-dKO CA1 pyramidal neurons.

(A) Cumulative probability of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-mediated mEPSC inter-event intervals. (B) Cumulative probability of AMPAR-mediated mEPSC amplitudes. Insets in cumulative probability graphs represent average mEPSC frequency (A) and amplitudes (B). mEPSC frequency was significantly reduced in Aplp2-KO and App/Aplp2-dKO CA1 pyramidal neurons as compared to wild-type (WT) but not App-KO littermates. (C) Average mEPSC of the four groups were not significantly different. (D) Cumulative probability of AMPAR-mediated mEPSC decay time with insets representing average mEPSC decay time. mEPSC decay time was significantly increased in App/Aplp2-dKO CA1 pyramidal neurons as compared to WT, App-KO, and Aplp2-KO littermates. (E) Representative recording traces of miniature EPSCs are shown. The number of recordings and the number of mice analyzed for each group are shown in (A). All data represent means ± SEM. Statistical assessment was performed using ordinary one-way ANOVA followed by uncorrected Fisher’s least significant difference (LSD) multiple comparisons test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

https://doi.org/10.7554/eLife.09743.035

We also found that the mEPSC decay time was increased in App/Aplp2-dKO as compared to WT, App-KO, and Aplp2-KO (Figure 12C, D), suggesting that APP and APLP2 may possess some redundant postsynaptic function, consistent with previous observations indicating a postsynaptic role for some APP metabolites, such as Aβ (Gu et al., 2009, Hsieh et al., 2006). Interestingly, Pen1-JCasp does not prompt this sharp decline in mEPSCs’ decay, suggesting that the JCasp domain of APP is primarily involved in the presynaptic functions of APP.

Discussion

Using an unbiased proteomic methodology and competition experiments, we found that the NH2-terminal activation-induced cytidine deaminase (AID) region, JCasp, is the APP domain that binds presynaptic proteins regulating synaptic transmission. We have used a peptide encompassing this JCasp region to test whether the interaction of APP with proteins that regulate synaptic vesicles exocytosis enables APP to tune synaptic vesicles release. Strikingly, intracellular delivery of JCasp caused a strong reduction in basal synaptic transmission. Consistent with a presynaptic action, this reduction was associated with an increase in PPF and FF, as well as in mEPCS frequency but not amplitude, rise time, and decay time. These effects are evident in WT synapses but absent at APP-lacking synapses. In addition, Pen1-JCasp reduced the rate of synaptic vesicles depletion without affecting recycling. Taken together, these results indicate that JCasp acts as a dominant negative inhibitor of APP, and that APP facilitates the release of excitatory synaptic vesicles, likely via the interaction with and functional regulation of the molecular machinery that controls synaptic vesicles exocytosis.

The evidence that glutamatergic synaptic vesicles release is normal in the hippocampus of young App-KO mice suggests that compensatory mechanisms counterbalance the loss of APP function. Given the functional redundancy of APP and APLP2 in neuromuscular junctions patterning and survival (Barbagallo et al., 2011b, Li et al., 2010, Wang et al., 2005), we thought that APLP2 could compensate for loss of APP function at hippocampal synapse. Consistent with this hypothesis, we found that the intracellular domain of APLP2, ALID2, interacts with key regulators of synaptic vesicle exocytosis. A systematic characterization of SC–CA1 synaptic transmission revealed that APP and APLP2 have redundant functions in the hippocampus as indicated by the following observations: 1) App-KO animals showed normal synaptic transmission at SC–CA1 excitatory synapses; 2) Aplp2-KO mice presented a significant reduction in mEPSCs frequency; 3) App/Aplp2-dKO mice showed increased PPF and FF, which was not observed in single App-KO and Aplp2-KO mice, as well as reduced mEPSCs frequency.

Overall, the synaptic phenotype observed in App/Aplp2-dKO mice is very similar to that induced by Pen1-JCasp in WT animals, suggesting a primary functional role for the intracellular domains of APP and APLP2 in tuning glutamate release. A notable exception is represented by an increase in mEPSC decay time – present in App/Aplp2-dKO mice but not in hippocampal slices incubated with Pen1-JCasp. This observation suggests that Pen1-JCasp can unveil the role of the JCasp domain of APP without directly affecting the function of other APP region, such as, for example, Aβ or the APP extracellular domain.

Our results indicate that JCasp is an inhibitor of APP but not APLP2. This specificity of JCasp for APP allows to study the function of the APP intracellular domain in WT adult animals, eliminating the confounding effects associated with both constitutive and conditional App-KO models. In fact, it permits acute inhibition of the functions of this APP domain in WT adult animals, circumventing developmental issues and compensatory mechanisms, such as those mediated by APLP2 and/or APLP1. The results presented here indicate that APP and APLP2 facilitate the release of excitatory synaptic vesicles, likely via the interaction with and functional regulation of the molecular machinery that controls synaptic vesicle exocytosis. How exactly these presynaptic APP interactions tune vesicles exocytosis and how this function is regulated remain to be elucidated.

While a recent study in primary neuronal cells has shown that Aβ can increase Pr by enhancing presynaptic Ca+2 influx (Fogel et al., 2014), our findings support a Ca+2 influx-independent mechanism. These two mechanisms of action of APP are not mutually exclusive and could act synergistically to control neurotransmitter release. Inherited or acquired factors leading to altered APP metabolism and Aβ levels could negatively impact this physiological function of APP and, eventually, contribute to the pathogenesis of AD.

It is worth noting that AID/AICD and JCasp are natural proteolytic products of APP. The mechanisms by which cleavage of APP can regulate synaptic transmission are twofold; 1) severing the APP–presynaptic multi-molecular complex from membranes; and 2) liberating cytosolic proteolysis fragments of APP, such as AID/AICD and JCasp, which could act, as we show for Pen1-JCasp, as dominant negative inhibitors of APP. AID/AICD and JCasp have very short half-lives and are hardly detectable in vivo (Cupers et al., 2001, Giliberto et al., 2010, 2008, Passer et al., 2000). It therefore appears that evolution has selected mechanisms, which are perhaps expensive in terms of genetic information and energy usage, to tightly regulate the levels of AID/JCasp, suggesting that AID/AICD and JCasp are biologically active and need to be strictly controlled.

APP plays a central role in the pathogenesis of both sporadic AD and FAD. Many previous studies investigating synaptic dysfunction induced by Aβ, which derives from APP processing and is considered the main mediator of AD pathogenesis, have suggested that early pathogenic changes in AD were driven by postsynaptic impairments (Cissé et al., 2011, Ferreira et al., 2012, Li et al., 2009, Mucke and Selkoe, 2012, Parameshwaran et al., 2008, Roberson et al., 2011, Shankar et al., 2007, Um et al., 2012). However, the presynaptic function of APP described here , together with evidence that the other familial Alzheimer’s proteins PS1 and PS2 regulate glutamate release in mature neurons by a presynaptic mechanism (Fogel et al., 2014, Wu et al., 2013, Xia et al., 2015, Zhang et al., 2009), support the possibility that deficits in neurotransmitter release represents a convergent mechanism involved in AD.

Materials and methods

Mice and ethics statement

Mice were handled according to the Ethical Guidelines for Treatment of Laboratory Animals of the NIH. The procedures were described and approved by the Institutional Animal Care and Use Committee (IACUC). The App-KO (Stock No: 004133) and Aplp2-KO (Stock No: 004142) mice were originally purchased from The Jackson Laboratory on a C57BL/6J background. Mice were crossed to obtain App+/-/Aplp2+/- animals. App+/-/Aplp2+/- mice were crossed to 129 mice to select App+/-/Aplp2+/- animals on a C57BL/6J/129 mixed background. App+/-/Aplp2+/- animals on a C57BL/6J/129 were crossed and WT, App-KO, Aplp2-KO, or App/Aplp2-dKO were used for experiments. C57BL/6J mice were generated at the Einstein animal facility.

Mouse brain and hippocampal preparation

The brain samples used for proteomic experiments were prepared as follows. Brains were homogenized in 5 mM (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)/NaOH pH 7.4, 1 mM ethylenediaminetetraacetic acid (EDTA), 1 mM ethylene glycol tetraacetic acid (EGTA), 0.25 M sucrose supplemented with protease and phosphatase inhibitors (Thermo Scientific). Brain homogenates were centrifuged at 800g for 10 min. Supernatant was collected and centrifuged at 9200g for 10 min to obtain the pellet (P2) and the supernatant (S2) fractions. The P2 fraction, which contains synaptosomes, was solubilized in 0.32 M sucrose, 20 mM Tris-base pH 7.4, 1 mM EDTA, incubated for 30 min at 4°C and centrifuged at 21000g for 15 min. The supernatant (LS1) was collected. The S2 and LS1 fractions were combined and used for the proteomic experiments. For Western blot analysis on hippocampal slices treated with either Pen1-JCasp or Pen1-ScJCasp (Figure 5), P2 fractions were prepared and solubilized in 1% TritonX100 RIPA buffer for 30 min at 4°C and centrifuged at 21000g. The triton soluble supernatant fractions were separated on a 16.5% Tris-Tricine sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), to obtain a better separation of APP-CTF species, and transferred onto nitrocellulose membranes. For Western blot analysis of pull-downs pre-incubated with JCasp or ScJCasp (Figure 1D), samples were separated on a 4–20% SDS-PAGE and transferred onto nitrocellulose membranes. The following antibodies were used in Western blots: anti-APP C-terminal AbD (Zymed), anti-APP N-terminal 22C11 (Millipore), anti-Actin (Cell Signaling), anti-Syp (Synaptic Systems, Goettingen, Germany), anti-Vamp2 (Synaptic Systems), anti-Syt2 (Synaptic Systems) and anti-Ddb11 (Cell Signaling).

Pull-down assays with StrepTag peptides and mass spectrometry

The StrepTag peptides were immobilized on StrepTactin column (IBA, St. Louis, MO), incubated with StrepTactin pre-cleared WT brain S2+LS1 fractions, and washed and eluted with desthiobiotin. The following analysis was performed by MSBioworks (Ann Arbor, Michigan). To prepare samples for mass spectrometry, the volume of each sample was reduced to 50 μL by vacuum centrifugation, 20 μL of each concentrated sample was processed by SDS-PAGE using a 10% Bis-Tris NuPAGE gel (Invitrogen) with the 2-(N-morpholino)ethanesulfonic acid (MES) buffer system, the gel was run approximately 2 cm. The mobility region was excised into 10 equal sized segments and in-gel digestion was performed on each using a robot (ProGest, DigiLab) with the following protocol: washed with 25 mM ammonium bicarbonate followed by acetonitrile. Reduced with 10 mM dithiothreitol at 60°C followed by alkylation with 50 mM iodoacetamide at RT. Digested with trypsin (Promega) at 37°C for 4 hr. Quenched with formic acid and the supernatant was analyzed directly without further processing. Each digest was analyzed by nano liquid chromatography–mass spectrometry (LC/MS/MS) with a Waters NanoAcquity High Performance Liquid Chromatography (HPLC) system interfaced to a Thermo Fisher Q Exactive mass spectrometer. Peptides were loaded on to a trapping column and eluted over a 75 μm analytical column at 350 nL/min; both columns were packed with Jupiter Proteo resin (Phenomenex). The mass spectrometer was operated in data-dependent mode, with MS and MS/MS performed in the Orbitrap at 70,000 and 17,500 full width at half maximum (FWHM) resolution, respectively. The 15 most abundant ions were selected for MS/MS.

Immunofluorescence

The mice were anesthetized by intraperitoneal injection (IP) of Avertin and perfused with ~20–30 ml of Soresan’s Buffer followed by 30–50 ml of 4% paraformaldehyde (PFA) fixative. After the perfusion brains were quickly removed and fixed in 4% PFA in Soresan’s Buffer overnight at 4°C. Brains were sectioned (50 μm thick) with a vibratome; slices were washed with phosphate buffer saline (PBS) and incubated for 90 min with pre-blocking agent containing 10% normal goat serum (NGS) in PBS, 0.02% saponin, and 0.1% Triton. Slides were washed in PBS and incubated overnight with primary antibodies Neun (1:300, Millipore), APP Y188 (1:200, Abcam), and Snap25 (1:300, Synaptic Systems) prepared in pre-blocking agent solution. After washes, sections were incubated with secondary Alexa 488 and 594 fluorescent antibodies (Molecular Probes) at room temperature for 2.5 hr. After washes, slices were fixed on glass slides and incubated with the antifade (Prolong Gold). Images were captured using SP5 confocal microscope (Leica), under 10X and 63X magnification.

Electrophysiology on brain slices

We used C57BL/6J WT mice aged 3–12 weeks of both sex (experiments shown in Figures 3, 4, 5, 7, 8), as well as 16–28-day-old C57BL/6J/129 hybrid mice, which were either WT, App-KO, Aplp2-KO, or App/Aplp2-dKO (experiments shown in Figures 9, 11, 12). The experimenter was blind to genotype and/or treatment. Coronal brain slices containing the hippocampal formation were prepared as previously described (Talani et al., 2011). In brief, animals were anesthetized with isoflurane, following the University of California San Diego Institutional Animal Care and Use Committee (IACUC) approved protocol. Brains were rapidly removed from the skull and placed in an ice-cold modified ACSF solution containing (in mM): 215 sucrose, 2.5 KCl, 1.6 NaH2PO4, 4 MgSO4, 1 CaCl2, 4 MgCl2, 20 glucose, 26 NaHCO3 (pH 7.4 equilibrated with 95% O2 and 5% CO2).Coronal brain slices (400 μm thick) were prepared with a Vibratome VT1200S (Leica Microsystems, Germany) and then incubated at room temperature in a physiologic ACSF, containing (in mM): 120 NaCl, 3.3 KCl, 1.2 Na2HPO4, 26 NaHCO3, 1.3 MgSO4, 1.8 CaCl2, 11 Glucose (pH 7.4 equilibrated with 95% O2 and 5% CO2). The hemi-slices were transferred to a recording chamber perfused with ACSF at a flow rate of ~2 ml/min using a peristaltic pump; experiments were performed at 28.0 ± 0.1°C. All recordings were performed using a multiclamp 700B amplifier (Molecular Devices). For extracellular field recordings (fEPSP recordings), a patch-type pipette was fabricated on a micropipette puller (Sutter Instruments, Novato, CA, USA), filled with 1M NaCl, and placed in the middle third of stratum radiatum in area CA1. Field excitatory postsynaptic potentials (fEPSPs) were evoked by activating SCs with a patch-type pipette (monopolar stimulation) filled with ACSF and placed in the middle third of stratum radiatum 150–200 μm away from the recording pipette and at approximately the same slice depth (150–200 μm). Square-wave current pulses (60 μs pulse width) were generated by a stimulus generator (Master 8, AMPI) and delivered through a stimulus isolator (Isoflex, AMPI). I/O function was generated by stimulating in 0.1 mA steps from 0.1 to 1.0/1.2 mA. Paired-pulse ratio (PPR) was measured by delivering two stimuli at 20, 40, 80, 200, 500, and 1000 ms inter-stimulus intervals. Synaptic facilitation was examined by repetitive stimulation (10 stimuli) at 1, 5, 10, and 20 Hz. Depletion of synaptic transmission was elicited by a long train (500 stimuli, 28 Hz) in 5 mM extracellular Ca2+ and in the presence of the NMDA receptor antagonist 2-amino-5-phosphopentanoic acid (DL-AP5) (100 µM). Recovery was assessed with nine stimuli at 2 Hz. Depletion and recovery plots were generated by normalizing responses to the peak amplitude of the fEPSP in the depletion train. The stimulus strength was adjusted so that the basal synaptic amplitude was similar (e.g., 0.5–1.0 mV) across experiments. To monitor miniature excitatory postsynaptic currents (mEPSCs), patch-clamp whole-cell recordings were performed from CA1 pyramidal cells, and we used an infrared differential interference contrast microscope (Nikon eclipse E600FN, Morrel Instrument Company Inc, Melville New York). The cells were voltage-clamped at -65 mV, and the patch pipette (resistance 3–6 MΩ) was filled with an internal solution containing (in mM): 131 CsCl, 8 NaCl, 1 CaCl2, 10 EGTA, 10 Glucose, 10 HEPES (pH = 7.3, 286 mmol/Kg). Series resistance (5 and 15 MΩ) was constantly monitored throughout the experiments by delivering a -5 mV, 80 ms voltage step, and cells with >10% change in series resistance were excluded from analysis. mEPSCs were recorded in the presence of 1 μM TTX and 15 μM 1(S), 9(R)(−)bicuculline methiodide, to block action potentials and GABA-A receptors, respectively. To calculate average mEPSCs recorded from at least three (Figures 4 and 8) and five (Figure 12) CA1 pyramidal neurons (between ~1000 and 600 events each) were normalized, aligned, and averaged.

Data acquisition and statistical analysis. Output signals were acquired at 5 kHz, filtered at 2.4 kHz, and stored online using pCLAMP 10.3 Electrophysiology Data Acquisition and Analysis Software (Molecular Devices). mEPSCs were analyzed using Mini Analysis software (Synaptosoft). In all cases the experimenter was blind to genotype and/or treatment. Data are presented as means ± SEM. Statistical comparisons of pooled data were performed by unpaired t-test or ANOVA (one- or two-way) using Prism software (GraphPad, San Diego,CA, USA). A p value of <0.05 was considered statistically significant.

Drugs were obtained from Sigma-Aldrich 1(S), 9(R)(−)bicuculline methiodide, Biotium (TTX) and Enzo Life Science (DL-AP5). Stock solutions were prepared in water and dimethyl sulfoxide (DMSO) and added to the ACSF as needed.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
    Mice with combined gene knock-outs reveal essential and partially redundant functions of amyloid precursor protein family members
    1. S Heber
    2. J Herms
    3. V Gajic
    4. J Hainfellner
    5. A Aguzzi
    6. T Rülicke
    7. H von Kretzschmar
    8. C von Koch
    9. S Sisodia
    10. P Tremml
    11. HP Lipp
    12. DP Wolfer
    13. U Müller
    (2000)
    The Journal of Neuroscience 20:7951–7963.
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
    ADAM10 and BACE1 are localized to synaptic vesicles
    1. JL Lundgren
    2. S Ahmed
    3. S Schedin-Weiss
    4. GK Gouras
    5. B Winblad
    6. LO Tjernberg
    7. S Frykman
    (2015)
    Journal of Neurochemistry, 135, 10.1111/jnc.13287.
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
    Age-dependent neuronal and synaptic degeneration in mice transgenic for the c terminus of the amyloid precursor protein
    1. ML Oster-Granite
    2. DL McPhie
    3. J Greenan
    4. RL Neve
    (1996)
    The Journal of Neuroscience 16:6732–6741.
  60. 60
  61. 61
  62. 62
    Generation of an apoptotic intracellular peptide by gamma-secretase cleavage of alzheimer's amyloid beta protein precursor
    1. B Passer
    2. L Pellegrini
    3. C Russo
    4. RM Siegel
    5. MJ Lenardo
    6. G Schettini
    7. M Bachmann
    8. M Tabaton
    9. L D'Adamio
    (2000)
    Journal of Alzheimer's Disease : JAD 2:289–301.
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
    Synaptic vesicle exocytosis
    1. TC Südhof
    2. J Rizo
    (2011)
    Cold Spring Harbor Perspectives in Biology, 3, 10.1101/cshperspect.a005637.
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
    The genetics of alzheimer disease
    1. RE Tanzi
    (2012)
    Cold Spring Harbor Perspectives in Medicine, 2, 10.1101/cshperspect.a006296.
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
  90. 90
  91. 91
  92. 92
  93. 93
  94. 94
  95. 95
  96. 96
  97. 97
  98. 98
  99. 99
  100. 100
  101. 101
  102. 102
  103. 103

Decision letter

  1. Bart De Strooper
    Reviewing Editor; VIB Center for the Biology of Disease, KU Leuven, Belgium

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for submitting your work entitled "APP interacts with the synaptic release machinery and facilitates transmitter release at hippocampal synapses" for peer review at eLife. Your submission has been favorably evaluated by a Senior editor and three reviewers, one of whom is a member of our Board of Reviewing Editors. One of the three peer reviewers, Jie Shen, has agreed to reveal his identity.

The reviewers have discussed the reviews with one another and the Reviewing editor has drafted this decision to help you prepare a revised submission.

Summary:

This paper by Fanutza et al. focuses on the functional significance of the N-terminal part of the intracellular APP domain (JCasp) in synaptic transmission and short-term plasticity in CA3-CA1 synapses. This work builds further on their previous study (Del Prete et al., 2014). They first show specific interaction of a series of synaptic vesicle associated proteins with the membrane proximal part of the APP intracellular domain (JCasp). They then show that this fragment, when modified with penetratin inhibits basal synaptic transmission and enhances short-term synaptic facilitation in CA3-CA1 synapses. Moreover, the authors found that JCasp imposes a tonic inhibition on spontaneous vesicle release, decreasing the mEPSC frequency. This study addresses an important question in neurobiology, i.e. the physiological impact of the intracellular APP domain in Schaffer collateral presynaptic function. The manuscript further explores the interaction between the intracellular domain of APP and proteins involved in neurotransmitter release. Specifically, the authors showed that JCasp, a fragment that is the N-terminus part of the APP intracellular domain and is cleaved by g-secretase and caspase, is involved in presynaptic short-term plasticity and neurotransmitter release. Overall, the paper provides much needed mechanistic insight into normal function of APP at presynaptic terminals where it is abundantly localized.

Essential revisions:

1) The quality of the immunofluorescent data needs to be upgraded. It is not possible to decide whether the APP staining is cellular or in synapses.

2) The whole story critically depends on one type of experiment: the effect of penetratin-coupled Pen1-JCasp on synaptic transmission. The authors provide a series of important controls on the experiment (penetratin deficient, scrambled peptide…), but no independent experimental approach to confirm their hypothesis that APP is involved in neurotransmission in the way they propose. The fact that the APP KO neurons do not show the same phenotype as the neurons treated with Pen1-JCasp is worrying. I can see that compensation by APLP2 could provide an explanation, but it would be great to actually demonstrate this and to show that APP, but not APP in which the JCasp part is deleted can rescue this in double knock out neurons. Can APPL2 overexpression rescue the phenotype induced by penetratin-JCasp? Alternatively does overexpression of the identified JCasp interacting proteins rescue the penetratin-JCasp mediated effect? Or have the authors other ideas to confirm their working hypothesis? Not all of these experiments have to be performed but some work along these lines could strengthen considerably the claims of the paper.

3) The authors should present the evidence for Pen1-JCasp penetration to the intracellular compartments of neurons in slices after 5 hr of pre-incubation. All the experiments were done following 5 hr pre-incubation of slices with the peptides. Was it 5 hr after the recovery of slices in large ACSF volume? Given that the experiments take at least 1 hr, all the measurements in the paper were done in late slices with modified properties. I wonder whether the authors find no difference in synaptic facilitation between WT and APP-KOs also in fresh slices (1-2 hr after recovery). I think it is important to repeat the major finding – increase in synaptic facilitation – after ICV injection of Pen1-JCasp and preparing the slices several hours after it.

4) Figure 7: mEPSC amplitude – something is wrong with the scale bar and the average amplitude of 3 pA. What is the noise level in the system? In Figure 3, the average amplitude is ~7 pA. The authors should show cumulative distributions for all the mEPSC measurements in the paper and explain in Methods clearly the detection threshold and which filters were applied for mEPSC recording/analysis. Why is the rise time ~40% larger in Figure 7 in comparison to Figure 3?

5) In the Figure 7, the authors show that the effect of JCasp on miniature spontaneous events persists in the presence of cadmium, a non-selective blocker of voltage-gated calcium channels. Based on these results, they concluded that JCasp generally reduces synaptic vesicle release in a Ca-independent manner. As spontaneous and evoked synaptic vesicle release may be differentially regulated, the authors should test the effect of JCasp on presynaptic Ca transients evoked by action potentials or, at least, discuss this caveat in the text and carefully interpret the mEPSC data. Discussion on the mechanism of Ca-independent, JCasp-mediated tonic inhibition would be helpful – could the authors suggest the specific stage in the exocytosis that is under a tonic block of JCasp.

6) The "Appresyome" is a vague description to indicate presynaptic proteins interacting with APP. I would not use this terminology, it is imprecise. What proteins are precisely responsible for the effects? The authors remain also very vague whether these interactions occur at the level of the synaptic vesicle or at the level of the presynaptic membrane. Is APP part of the snare complexes?

https://doi.org/10.7554/eLife.09743.039

Author response

Essential revisions:

1) The quality of the immunofluorescent data needs to be upgraded. It is not possible to decide whether the APP staining is cellular or in synapses.

We apologize for not having clearly stated the rationale for the IF experiments. The purpose of the experiment shown in Figure 1D of the original manuscript (which is now Figure 1E of the revised version) was to show the distribution of endogenous APP in the hippocampus. Albeit this may seem as a superfluous exercise, we believe it is important in view of the recent report form Hui Zheng’s laboratory (Guo, Li et al., 2012) showing that most of the antibodies that have been used in the past to determine the localization of APP were non-specific (i.e. they stained App-KO and WT material in a similar fashion). The authors established with clarity that, among those tested, the Y188 antibody was the only specific antibody. We are unaware of any publication showing the expression of endogenous APP in mouse hippocampus using the Y188 antibody or other antibodies with appropriate controls (i.e. App-KO tissues to determine specificity). All other information concerning APP distribution in the hippocampus is derived from analysis of mice overexpressing APP. It is worth noting that: a) the distribution of overexpressed APP does not necessarily reflect the “normal” distribution of endogenous APP; b) detecting endogenous APP is objectively a much more difficult task then detecting overexpressed APP.

The staining with Y188 (the specificity of Y188 was confirmed by the evidence that this antibody does not stain App-KO hippocampi) showed high level of APP staining in cell bodies of DG, CA3 and CA1 neurons. On the contrary, the staining in other regions, such as the stratum radiatum containing the CA3 projections to CA1 neurons, is more diffused and less intense. This is not surprising since cell bodies are where the ER and Golgi structure are highly concentrated: this is where APP, a type I membrane protein, is synthesized and undergoes maturation.

From the cell bodies, APP will be transported along dendrites at synapses. Thus, it is expected to find that APP in more concentrated in cellular bodies as compared to dendrites and synapses. We have now added a staining in Figure 1F showing that APP is found in the stratum radiatum, where it partially co-localizes with the specific presynaptic marker Bassoon, supporting a presynaptic function of APP in the SC pathway. This staining is specific since it is absent in App-KO hippocampi.

Technical issues could also explain the quantitative differences in Y188 staining between CA1/CA3 cell bodies and stratum radiatum. Y188 recognize the C-terminal epitope. This epitope may be functionally engaged in multi molecular interactions (this is actually our working model) in dendrites/synapses -perhaps immature APP that is in cell body is much less BUSY interacting- therefore it may be scarcely accessible to Y188.

Overall, our findings are consistent with previous observations showing a pre-synaptic localization of APP (Del Prete et al., 2014, Fogel, Frere et al., 2014, Gautam et al., 2015, Kohli, Pflieger et al., 2012, Lassek, Weingarten et al., 2014, Lassek et al., 2013, Lundgren et al., 2015, Norstrom et al., 2010).

2) The whole story critically depends on one type of experiment: the effect of penetratin-coupled Pen1-JCasp on synaptic transmission. The authors provide a series of important controls on the experiment (penetratin deficient, scrambled peptide...), but no independent experimental approach to confirm their hypothesis that APP is involved in neurotransmission in the way they propose. The fact that the APP KO neurons do not show the same phenotype as the neurons treated with Pen1-JCasp is worrying. I can see that compensation by APLP2 could provide an explanation, but it would be great to actually demonstrate this and to show that APP, but not APP in which the JCasp part is deleted can rescue this in double knock out neurons. Can APPL2 overexpression rescue the phenotype induced by penetratin-JCasp? Alternatively does overexpression of the identified JCasp interacting proteins rescue the penetratin-JCasp mediated effect? Or have the authors other ideas to confirm their working hypothesis? Not all of these experiments have to be performed but some work along these lines could strengthen considerably the claims of the paper.

The fact that App-KO hippocampi of young mice showed normal synaptic transmission is expected. Although App-KO mice have been available since 1992, deficits in PPF and FF in the SC pathway have never been reported in these animals. In addition, deficits in long-term plasticity (long-term potentiation, LTP) are obvious only in aged, but not in juvenile, App-KO mice [reviewed in (Korte, Herrmann et al., 2012)]. As well pointed out by the reviewers, APP and APLP2 have redundant functions; APLP2 compensates for APP in App-KO mice and, vice versa, APP compensates for APLP2 in Aplp2-KO mice. App/Aplp2-dKO mice have dramatic deficits in NMJ patterning; yet App-KO and Aplp2-KO mice have absolutely normal NMJ development. More vividly, App/Aplp2-dKO mice have a very short lifespan (~30 days); yet App-KO and Aplp2-KO mice have substantially normal lifespans. These observations demonstrate that App/Aplp2-dKO mice have extremely striking phenotypes, whereas single KO mice are essentially normal, clearly illustrating the power of the compensatory mechanisms between APP and APLP2.

We thought that the data obtained with JCasp are important and could stand alone mainly for two reasons: 1) the evidence that Pen1-JCasp can acutely inhibit the normal function of APP in young wild type mice can be considered a strength, since it is a great tool to unveil the functional role of APP (and more specifically of a specific APP domain, the intracellular domain) avoiding compensatory mechanisms and/or the ablation of functions associated with other domains of APP; 2) given that JCasp is a naturally occurring APP fragment, understanding its function is intrinsically important, just like studying the function of other APP metabolites including Aβ, AID/AICD, the recently characterized Aη-α, sAPPβ, sAPPα, β-CTF, α-CTF etc.

Having established that JCasp is functionally active, we went one step further and tested JCasp in App-KO hippocampi to better understand the mechanism of action of JCasp. We felt that experiments addressing compensation issues may have sidetracked the reader and diluted the main messages of the paper.

Yet, it is of value to determine whether APP and APLP2 share common synaptic functions and whether APLP2 can compensate the loss of APP function in App-KO hippocampi. In fact, we have been testing this hypothesis.

First, we have used a proteomic approach to characterize the brain interactome of the APLP2 intracellular domain. The rational is that if the interaction of APP with key regulators of exocytosis is important for the described synaptic function of APP, and if APLP2 compensates the loss of this function in App-KO, then the intracellular domain of APLP2 should also interact with key regulators of exocytosis. The data are now included in Table 2 and show that also the APLP2 intracellular domain interacts with pre-synaptic proteins, including of Vamp2, Syp and Syt2.

Second, we have tested whether JCasp interferes with these interactions, since the electrophysiological data suggest that Pen1-JCasp is a dominant negative of APP but not APLP2 (Pen1-JCasp does not alter synaptic transmission in App-KO mice, which still express APLP2). As shown in Figure 10, JCasp does not alter the interaction of Vamp2, Syp and Syt2 with ALID2.

Then, we have tested synaptic transmission in WT, App-KO, Aplp2-KO and App/Aplp2-dKO mice derived from AppKO/WT/Aplp2KO/WT breeding pairs. These experiments, that are shown in Figures 11 and 12 and were underway when the manuscript was originally submitted, indicate that: 1) APP and APLP2 facilitate glutamate release; 2) APLP2 compensates for loss of APP function in App-KO mice and, vice versa, that APP compensates for loss of APLP2 function in Aplp2-KO mice. These experiments are described in a new sub-chapter of the Results section entitled “Loss of APP and APLP2 reduces excitatory synaptic transmission”.

3) The authors should present the evidence for Pen1-JCasp penetration to the intracellular compartments of neurons in slices after 5 hr of pre-incubation.

To visualize intracellular peptides, we have synthesized JCasp and Pen1-JCasp linked to Biotin at the NH2-terminus. As we show in Figure 2C of this revised manuscript, Pen1-JCasp, but not JCasp, penetrates into the intracellular compartments of neurons. The results are discussed in the section entitled “Intracellular delivery of JCasp decreases excitatory synaptic transmission”, second paragraph.

All the experiments were done following 5 hr pre-incubation of slices with the peptides. Was it 5 hr after the recovery of slices in large ACSF volume? Given that the experiments take at least 1 hr, all the measurements in the paper were done in late slices with modified properties. I wonder whether the authors find no difference in synaptic facilitation between WT and APP-KOs also in fresh slices (1-2 hr after recovery). I think it is important to repeat the major finding – increase in synaptic facilitation – after ICV injection of Pen1-JCasp and preparing the slices several hours after it.

We apologize for not having explained more clearly the design of our experiments. In this revised version, we specify (in the section “Intracellular delivery of JCasp decreases excitatory synaptic transmission”): “The experiments were performed as follows: soon after preparation, the brain slices were incubated in 200 ml of ACSF containing the indicated concentrations of either Pen1-JCasp or the negative control peptide Pen1-ScJCasp. Recordings were performed between ~5 and 8 hr after preparation of slices and incubation with Pen1-peptides.”

As suggested by the reviewers, to further stress the point that the results were not artificially affected by the experimental design, we compared facilitation on fresh vs. old slices in wild type and App-KO mice. Normally, we allow the brain slices to recover for 3 hr after cutting. For this experiment, we have defined as “fresh slices” those analyzed between 3 and 5 hr post-cutting, and as “old slices” those analyzed between 5 and 8 hr post-cutting. As shown in Author response image 1, synaptic facilitation is not significantly affected within this time frame in neither genotype, and there is no significant difference between wild type and App-KO recordings both in “fresh” and in “old” slices. Moreover, a new set of experiments conducted on WT, App-KO, Aplp2-KO and App/Aplp2-dKO (and discussed above) confirms that, at least at 3-4 weeks of age, App-KO mice do not show deficits in facilitation.

Author response image 1
The “age” of the hippocampal slices does not affect facilitations in both WT and App-KO SC:CA1 synapses.

Frequency facilitation was tested at 20 Hz.

https://doi.org/10.7554/eLife.09743.038

In addition, if Pen1-JCasp were affecting synaptic transmission by altering neurons’ recovery and not by altering APP functions, Pen1-JCasp would probably have a similar effect in WT and App-KO hippocampal neurons. In view of these clarifications and new data, we hope the reviewer will agree in that testing synaptic facilitation in slices prepared from mice after ICV injection of peptides does not seem to be required. ICV injection experiments – especially when using peptides that have never been tested before – are not trivial to set up. The effects of Pen1-JCasp on synaptic facilitation are highly dependent on the concentration of the peptide (a clear example is illustrated in Figure 3 of the paper). Determining the delivery conditions (peptide concentration in the solution being injected/volume injected/frequency of injection/location of delivery etc.) needed to achieve working concentrations of Pen1-JCasp and Pen1-ScJCasp in the hippocampus of a live animal is not straightforward.

4) Figure 7: mEPSC amplitude – something is wrong with the scale bar and the average amplitude of 3 pA. What is the noise level in the system? In Figure 3, the average amplitude is ~7 pA. The authors should show cumulative distributions for all the mEPSC measurements in the paper and explain in Methods clearly the detection threshold and which filters were applied for mEPSC recording/analysis. Why is the rise time ~40% larger in Figure 7 in comparison to Figure 3?

We thank the reviewers for pointing out these errors. We confirm that there was a mistake not only in the calibration bars but a digital filter was mistakenly applied to the raw data. Moreover, while the experiments shown in the old Figure 3 were performed in 4 weeks old C57Bl6 mice, the experiments in Figure 8 were performed in p18/20 C57Bl6/129 mixed background mice. The recording conditions and the different genetic background may contribute to the differences between experiments that are noted by the reviewer and the low amplitude observed in the experiment reported in Figure 8. We have repeated these experiments in hippocampal slices prepared from 3/4 weeks old C57Bl6/J mice, and removed the digital filter. The results are now shown in the revised Figure 4 and 8. In response to the reviewers we have also included cumulative probability curves. These new experiments not only confirm that the Pen1-JCasp reduces the frequency of mEPSCs but also confirm that Cd+2 does not alter the effects of Pen1-JCasp.

5) In the Figure 7, the authors show that the effect of JCasp on miniature spontaneous events persists in the presence of cadmium, a non-selective blocker of voltage-gated calcium channels. Based on these results, they concluded that JCasp generally reduces synaptic vesicle release in a Ca-independent manner. As spontaneous and evoked synaptic vesicle release may be differentially regulated, the authors should test the effect of JCasp on presynaptic Ca transients evoked by action potentials or, at least, discuss this caveat in the text and carefully interpret the mEPSC data. Discussion on the mechanism of Ca-independent, JCasp-mediated tonic inhibition would be helpful – could the authors suggest the specific stage in the exocytosis that is under a tonic block of JCasp.

Our experiments using the voltage-gated calcium channel blocker cadmium show that these channels do not play a significant role in the JCasp-mediated suppression of synaptic transmission. We therefore conclude that JCasp reduces glutamate release in a “calcium influx-independent manner” (Discussion). The fact that JCasp was equally effective in the presence of cadmium, strongly suggests that JCasp mainly acts downstream calcium influx. We do not claim that JCasp acts in a calcium-independent manner, and we have no evidence that JCasp differentially targets evoked vs. spontaneous release.

6) The "Appresyome" is a vague description to indicate presynaptic proteins interacting with APP. I would not use this terminology, it is imprecise. What proteins are precisely responsible for the effects? The authors remain also very vague whether these interactions occur at the level of the synaptic vesicle or at the level of the presynaptic membrane. Is APP part of the snare complexes?

In response to the reviewer’s suggestion, we have removed the term Appresyome. Defining which interaction (or interactions) is (are) “directly” responsible for the effects, whether synaptic vesicle APP and/or presynaptic membrane APP are involved, and whether APP is an integral part of the SNARE complex are extremely important. However, addressing these questions will likely require multiple approaches and several years of experimentation. As we discuss in our manuscript (in the section “Intracellular delivery of JCasp decreases excitatory synaptic transmission”): “This presynaptic APP-interacting network can stem from APP molecules localized either at synaptic vesicles or at presynaptic membranes and is almost certainly formed via the direct interaction of the intracellular domain of APP with some release machinery protein”. We also acknowledge (in the Discussion): “How exactly these presynaptic APP interactions tune vesicles exocytosis and how this function is regulated remain to be elucidated.”

https://doi.org/10.7554/eLife.09743.040

Article and author information

Author details

  1. Tomas Fanutza

    Department of Microbiology and Immunology, Albert Einstein College of Medicine, New York, United States
    Contribution
    TF, Prepared samples for electrophysiology recordings. Performed the electrophysiology recordings. Analyzed and interpreted the electrophysiology recordings, Acquisition of data, Analysis and interpretation of data
    Contributed equally with
    Dolores Del Prete
    Competing interests
    The authors declare that no competing interests exist.
  2. Dolores Del Prete

    Department of Microbiology and Immunology, Albert Einstein College of Medicine, New York, United States
    Contribution
    DDP, Performed biochemical experiments. Generated the KO mice. Performed IF experiments. Prepared samples for electrophysiology recordings. Analyzed and interpreted the electrophysiology recordings, Acquisition of data, Analysis and interpretation of data
    Contributed equally with
    Tomas Fanutza
    Competing interests
    The authors declare that no competing interests exist.
  3. Michael J Ford

    MS Bioworks, LLC, Ann Arbor, United States
    Contribution
    MJF, Performed the proteomic analysis
    Competing interests
    The authors declare that no competing interests exist.
  4. Pablo E Castillo

    Dominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, New York, United States
    Contribution
    PEC, Contributed to the analysis, interpretation and planning of electrophysiology experiments. Contributed to writing the paper. Analysis and interpretation of data, Drafting or revising the article
    For correspondence
    pablo.castillo@einstein.yu.edu
    Competing interests
    The authors declare that no competing interests exist.
  5. Luciano D’Adamio

    Department of Microbiology and Immunology, Albert Einstein College of Medicine, New York, United States
    Contribution
    LD’A, Designed experiments. Performed biochemical experiments. Generated the KO mice. Analyzed and interpreted the data. Wrote the manuscript; Conception and design, Acquisition of data, Analysis and interpretation of data; Drafting or revising the article
    For correspondence
    luciano.dadamio@einstein.yu.edu
    Competing interests
    The authors declare that no competing interests exist.

Funding

Alzheimer's Association (Zen-11-201425)

  • Tomas Fanutza
  • Dolores Del Prete
  • Luciano D’Adamio

National Institute on Aging (1RO1AG041531, 1RO1AG033007, 5R21AG048971)

  • Tomas Fanutza
  • Dolores Del Prete
  • Luciano D’Adamio

BrightFocus Foundation (A2013046F)

  • Dolores Del Prete

National Institute on Drug Abuse (DA017392)

  • Pablo E Castillo

National Institute of Mental Health (MH081935)

  • Pablo E Castillo

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee(IACUC) at the Albert Einstein College of Medicine in animal protocol number 20130509. All surgery was performed under sodium pentobarbital anesthesia, and every effort was made to minimize suffering.

Reviewing Editor

  1. Bart De Strooper, VIB Center for the Biology of Disease, KU Leuven, Belgium

Publication history

  1. Received: June 29, 2015
  2. Accepted: November 8, 2015
  3. Accepted Manuscript published: November 9, 2015 (version 1)
  4. Accepted Manuscript updated: November 10, 2015 (version 2)
  5. Version of Record published: February 3, 2016 (version 3)

Copyright

© 2015, Fanutza et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,537
    Page views
  • 570
    Downloads
  • 17
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)