Amyloid Precursor Protein (APP) May Act as a Substrate and a Recognition Unit for CRL4CRBN and Stub1 E3 Ligases Facilitating Ubiquitination of Proteins Involved in Presynaptic Functions and Neurodegeneration*

The amyloid precursor protein (APP), whose mutations cause Alzheimer disease, plays an important in vivo role and facilitates transmitter release. Because the APP cytosolic region (ACR) is essential for these functions, we have characterized its brain interactome. We found that the ACR interacts with proteins that regulate the ubiquitin-proteasome system, predominantly with the E3 ubiquitin-protein ligases Stub1, which binds the NH2 terminus of the ACR, and CRL4CRBN, which is formed by Cul4a/b, Ddb1, and Crbn, and interacts with the COOH terminus of the ACR via Crbn. APP shares essential functions with APP-like protein-2 (APLP2) but not APP-like protein-1 (APLP1). Noteworthy, APLP2, but not APLP1, interacts with Stub1 and CRL4CRBN, pointing to a functional pathway shared only by APP and APLP2. In vitro ubiquitination/ubiquitome analysis indicates that these E3 ligases are enzymatically active and ubiquitinate the ACR residues Lys649/650/651/676/688. Deletion of Crbn reduces ubiquitination of Lys676 suggesting that Lys676 is physiologically ubiquitinated by CRL4CRBN. The ACR facilitated in vitro ubiquitination of presynaptic proteins that regulate exocytosis, suggesting a mechanism by which APP tunes transmitter release. Other dementia-related proteins, namely Tau and apoE, interact with and are ubiquitinated via the ACR in vitro. This, and the evidence that CRBN and CUL4B are linked to intellectual disability, prompts us to hypothesize a pathogenic mechanism, in which APP acts as a modulator of E3 ubiquitin-protein ligase(s), shared by distinct neuronal disorders. The well described accumulation of ubiquitinated protein inclusions in neurodegenerative diseases and the link between the ubiquitin-proteasome system and neurodegeneration make this concept plausible.

In vivo studies have identified an essential role for the ACR in the patterning of neuromuscular junction and survival and in synaptic transmission (25)(26)(27)(28)(29). Other studies have suggested that release of AID modulates apoptosis, gene transcription, and Ca 2ϩ homeostasis (23, 30 -40). The caspase-derived APP fragments Ccas and JCasp also possess toxic activities (22)(23)(24). Overall, these data indicate that the ACR is functionally important in vivo.
APP belongs to a protein family that includes APLP1 and APLP2. APLP1 and APLP2 are processed like APP (41)(42)(43)(44)(45) and release intracellular peptides, called ALID1 and ALID2, respectively, that, like AID, can potentially regulate transcription (42,46). The evidence that Aplp2-KO and App-KO mice are viable and show normal synaptic vesicle release, whereas the combined App/Aplp2-dKO mice develop neuromuscular junction deficits, die shortly after birth, and have altered synaptic vesicle exocytosis (28,47), illustrates the functional redundancy of APP and APLP2.
The ACR is short and does not possess enzymatic activity, suggesting that it may function by modulating the activity of interacting proteins. As discussed above, several APP metabo-lites contain the ACR. These include the membrane-bound full-length APP, ␤-CTF, and ␣-CTF and the soluble AID peptide. Thus, complexes formed by ACR-interacting proteins may have distinct functional properties depending on the ACR-containing APP metabolite with which they interact.
Using an unbiased proteomic approach, we have characterized the ACR brain interactome (28,48). Here, we show that the ACR interacts with several proteins that regulate the UPS. The E3 ubiquitin-protein ligase Stub1 and the E3 ubiquitin-protein ligase complex CRL4 CRBN , which is formed by Cul4a/b, Ddb1, and Crbn (49), are the most abundant UPS-related proteins interacting with the ACR. By performing ubiquitome analysis on the ACR brain interactome and total mouse brain, we generated evidence implicating APP in the ubiquitination of ACR-interacting proteins, and E3 ubiquitin-protein ligases, including CRL4 CRBN , in the ubiquitination of APP. Moreover, we found that APLP2, but not APLP1, can potentially exert a similar role. Finally, we observed that several of the UPS-related ACR interactors and proteins ubiquitinated in vitro in an ACR-dependent manner are genetically linked to neurodegeneration.
Ubiquitination can either modulate protein function or promote protein degradation by the proteasomal and the autophagic/lysosomal pathway. Integrity of these two pathways is important for normal aging and to ensure efficient turnover of both functional and defective proteins. The finding that APP may play a role in the ubiquitination of proteins linked to neurodegenerative diseases suggests that dysregulation of a functional network in which APP functions as a modulator of E3 ubiquitin-protein ligase(s) could be a pathogenic mechanism shared by numerous neuronal disorders.

Results
APP Interacts, via Its ACR, with Proteins That Regulate the UPS-In vivo observations underscore a key physiological and pathological role of the ACR (25, 27-29, 48, 50, 51, 53, 54). This short sequence lacks enzymatic activity and may function as a docking domain for cytosolic as well as membrane-bound proteins (46,(55)(56)(57)(58)(59)(60)(61)(62)(63)(64)(65)(66)(67)(68)(69)(70)(71)(72). To explore the potential role of the ACR, we used a proteomic approach (28,48). Five synthetic peptides, i.e. control St peptide, St-ACR, St-ACR with on Tyr 682 (St-ACR Tyr(P) ), St-ACR with phosphorylation on Thr 668 (St-ACR Thr(P) ), and St-ACR with phosphorylation on both Thr 668 and Tyr 682 (St-ACR Thr(P)Tyr(P) ), were immobilized on Strep-Tactin resin. The numbering of phosphorylated residues is based on the APP isoform of 695 amino acids. Mouse brain fractions were first passed twice through StrepTactin resin columns to remove proteins that bind to the StrepTactin resin. Then, they were applied in parallel on separate columns packed with either StrepTactin-St-, StrepTactin-St-ACR-, StrepTactin-St-ACR Thr(P) -, StrepTactin-St-ACR Tyr(P) -, or StrepTactin-St-ACR Thr(P)Tyr(P) -coated resin. After extensive washing, the St, St-ACR, St-ACR Thr(P) , St-ACR Tyr(P) , and St-ACR Thr(P)Tyr(P) peptides were eluted, together with proteins specifically bound to them, with desthiobiotin. Eluted proteins were digested with trypsin and identified by nano-LC/MS/MS. We used St-ACR, St-ACR Thr(P) , St-ACR Tyr(P) , and St-ACR Thr(P)Tyr(P) peptides for two reasons. First is to identify interactions that are regulated by phosphorylation on these two residues. This goal is justified by previous reports showing that phosphorylation of either Thr 668 or Tyr 682 can modulate interaction of APP with some binding partners (73). Second is that the use of four independent ACR baits can help to pinpoint interactions that are possibly biologically relevant.
The E3 ubiquitin-protein ligase activity of CRL complexes is dependent on the neddylation of the cullin subunit and is inhibited by the association of the deneddylated cullin subunit with the Cullin-associated NEDD8-dissociated protein 1 (TIP120A/ CAND1) (78). CRL deneddylation is performed by the COP9 signalosome complex, a protein complex with isopeptidase activity that catalyzes the hydrolysis of NEDD8 protein from the cullins (79). Cand1 and all eight subunits composing the COP9 signalosomes (Cops1/2/3/4/5/6/7a/7b/8) were isolated in our ST-ACR pulldown experiments.
Ubiquilins contain an NH 2 -terminal ubiquitin-like domain and a COOH-terminal ubiquitin-associated domain. They physically associate with both proteasomes and ubiquitin ligases and thus are thought to functionally link the ubiquitination machinery to the proteasome (82). Three ubiquilins were isolated in our screening as follows: ubiquilin-1 (Ubqln1), -2 (Ubqln2), and -4 (Ubqln4).
CRL4 CRBN E3 Ligase Complex and the E3 Ligase Stub1 Bind the COOH-and NH 2 -terminal Regions of the ACR, Respectively-The UPS/ACR-interacting network is presumably formed via the direct interaction of the ACR with one or more UPS-linked proteins, and the other proteins are most likely indirectly associated with the ACR, for example via secondary interactions (i.e. mediated by binding to a direct ACR interactor) or tertiary interactions (i.e. mediated by binding to a protein that binds a direct ACR interactor) and so forth. It is reasonable to suppose that proteins binding to the ACR directly will be enriched more efficiently than proteins that bind indirectly. Normalized spectral abundance factor (NSAF) analysis indicates that Stub1, Ddb1, Crbn, Cul4a, and Cul4b are the five most abundant UPS-related proteins present in the St-ACR, St-ACR Thr(P) , St-ACR Tyr(P) , and St-ACR Thr(P)Tyr(P) pulldowns (Table 1; NSAF values for St-negative control, St-ACR, St-ACR Thr(P) , St-ACR Tyr(P) , and St-ACR Thr(P)Tyr(P) pulldowns were as follows: for Ddb1, 0.0019, 0.0141, 0.0228, 0.0114, and 0.0149; for Cul4a, 0, 0.0015, 0.003, 0.0009, and 0.0019; for Cul4b, 0, 0.0003, 0.0009, 0.0004, and 0.0007; for Crbn, 0, 0.0061, 0.0129, 0.0053, and 0.0061; and for Stub1, 0, 0.0091, 0.0072, 0.0055, and 0.0046). This quantitative analysis suggests that Stub1, Ddb1, Crbn, Cul4a, and Cul4b are the most likely UPSrelated proteins to directly interact with APP and that phosphorylation of the ACR on either Thr 668 or Tyr 682 does not appreciably alter their binding to the ACR.
As a further step toward discriminating biologically relevant interactions from background noise, we used the above proteomic approach to identify the brain proteins interacting with the NH 2 terminus (JCasp) and COOH terminus (Ccas) subdomains of the ACR (Fig. 1A, schematic). In this experiment as well we used five baits as follows: the negative control St peptide, St-Ccas, St-Ccas with phosphorylation on Tyr 682 (St-Ccas Tyr(P) ), St-Ccas with phosphorylation on Thr 668 (St-Ccas Thr(P) ), and St-JCasp.
To confirm the accuracy of the proteins detected in the proteomic analysis, we performed Western blotting analysis with pulldown lysates. As shown in Fig. 1B (55,59,83), and Pin1, which interacts with APP in a Thr 668 phosphorylation-dependent manner (84), were only detected in St-Ccas Tyr(P) and St-Ccas Thr(P) pulldowns, respectively, further validating the proteomic strategy.
Overall, the data suggest that of all the UPS-linked proteins pulled down with St-ACR, the CRL4 CRBN E3 ligase complex components and E3 ligase Stub1 are the most likely to interact directly with APP in vivo. Incidentally, an interaction between Stub1 and APP in vivo was previously described (85). CRL4 CRBN E3 ligase complex subunits and Stub1 bind to two distinct and non-overlapping domains of the APP intracellular domain: the NH 2 -terminal JCasp for the E3 ligase Stub1 and the COOH-terminal Ccas regions for the components of the CRL4 CRBN E3 ligase complex. Thus, it is conceivable that these two E3 ligases may interact with APP simultaneously.
Crbn Mediates Binding of the CRL4 CRBN E3 Ligase Complex to the ACR-Next, we investigated whether Ddb1, Cul4, and Crbn interact with the ACR as a complex. Ddb1 and Cul4a bind several substrate-recognition subunits, called Dcaf. Indeed, Crbn is essentially a Dcaf. If the ACR interacted with the CRL4 CRBN E3 ligase complex either via Cul4a, Cul4b, or Ddb1, several Dcafs should have been isolated by St-ACR. However, Crbn was the only Dcaf abundantly present in the pulldown with the four St-ACR baits (St-ACR, St-ACR Thr(P) , St-ACR Tyr(P) , and St-ACR Thr(P)Tyr(P) ) and the three St-Ccas baits (St-Ccas, St-Ccas Thr(P) , and St-Ccas Tyr(P) ); Dcaf5 and Dcaf8 were detected but in very low amounts and only in St-ACR and St-ACR Tyr(P) pulldowns (NSAF for Dcaf5-derived peptides was 0.0004 and 0.0001 in St-ACR and St-ACR Tyr(P) pulldowns, respectively; NSAF for Dcaf8-derived peptides was 0.0002 and 0.0003 in St-ACR and St-ACR Tyr(P) pulldowns, respectively) and not in St-ACR Thr(P) , St-ACR Thr(P)Tyr(P) , St-Ccas, Ccas Thr(P) , and St-Ccas Tyr(P) pulldowns (NSAF were 0 for both proteins in all five samples). Altogether, these data suggest that Dcaf5 and Dcaf8 may either have been non-specifically isolated in the St-ACR and St-ACR Tyr(P) pulldown or that Dcaf5 and Dcaf8 can weakly and indirectly interact with the ACR.
To directly test whether Crbn mediates the interaction of CRL4 CRBN with APP, we performed St-ACR Thr(P)Tyr(P) pulldowns using brain lysates isolated either from wild type (WT) or Crbn-KO mice (86). Both proteomic (Table 4) and Western blotting analysis (Fig. 2) of these pulldowns show St-ACR binds Ddb1 and Cul4 when challenged with brain lysates isolated from WT mice but not when the brain lysates were derived from Crbn-KO (86) animals, albeit Ddb1 and Cul4 were equally  (Fig. 2). These data strongly suggest that Cul4, Ddb1, and Crbn bind to the ACR as a complex and that the APP cytoplasmic tail binds CRL4 CRBN via Crbn. On the contrary and as predictable, Stub1 (Table 4), Grb2, and Pin1 (Table 4 and Fig. 2) bind to St-ACR Thr(P)Tyr(P) independently of Crbn. COOH Terminus of APP Binds the Substrate Recognition Pocket of Crbn-Although Thr 668 and Tyr 682 are in the Ccas ACR domain, phosphorylation of these residues does not modulate binding of CRL4 CRBN suggesting that they are not involved in the CRL4 CRBN -APP interaction. To further confirm this, we performed pulldowns with NH 2 -terminal Ccas deletions and found that the COOH-terminal 12 amino acids of APP (St-C-12, which do non include Thr 668 and Tyr 682 ) retain full binding capabilities for CRL4 CRBN , whereas the COOHterminal eight amino acids (St-C-8) do not (Fig. 3A). Next, we performed pulldowns using a series of COOH-terminal deletions of St-C-12. Removal of the last one and two amino acids of APP reduces the binding of Crbn and Ddb1 by ϳ50 and ϳ100%, respectively. Essentially, when the NH 2 -terminal amino acids of the 12-mer are deleted, binding was greatly reduced, as in the case with deleting the COOH-terminal two amino acids. It is therefore possible that reducing the ACR peptide below the length of 12 amino acids disrupts the peptide's secondary structure, thereby reducing Crbn and Ddb1 binding. Overall, the data indicate that the COOH-terminal 12 amino acids of APP contain the docking site for CRL4 CRBN .
Crbn is the substrate recognition subunit of CRL4 CRBN (49). To test whether APP binds the substrate-recognition pocket of Crbn, we pre-incubated brain lysates with either 10 or 100 M concentrations of either thalidomide (Thal) or lenalidomide (Len), two compounds that bind the substrate-recognition pocket of Crbn (87)(88)(89). After 1 h, lysates were used for pulldown experiments with St-Ccas Tyr(P) . Both thalidomide and lenalidomide reduced binding of Crbn and Ddb1 to St-Ccas Tyr(P) in a dose-dependent manner (Fig. 3C). Lenalidomide, which binds Crbn more efficiently than thalidomide, is more effective. As expected, neither thalidomide nor lenalidomide interfered with binding of Grb2 to St-Ccas Tyr(P) (Fig. 3C). In summary, the extreme COOH terminus of the ACR interacts with brain-derived CRL4 CRBN via the substrate-binding pocket of Crbn (model shown in Fig. 3D), suggesting that APP may be a substrate of the CRL4 CRBN E3 ligase.
APLP2, but not APLP1, Bind Stub1 and CRL4 CRBN -The APP protein family includes two other members, APLP1 and APLP2. These three proteins are functionally redundant, particularly APP and APLP2 (25,45,47,90). APLP1 and APLP2 cytoplasmic regions (named AL1CR and AL2CR, respectively) are similar to that of APP. The ACR and the AL2CR share 66% identity and 16% similarity; the ACR and AL1CR share 53% identity and 21% similarity; and the AL2CR and the AL1CR share 62% identity and 18% similarity (Fig. 4A). As a first step toward determining whether APLP1 and/or APLP2 may functionally interface with UPS-related proteins, we characterized the brain interactome of the AL1CR and the AL2CR. We synthesized the control St-only peptides, St-AL1CR and St-AL2CR. As shown in Table 5, St-AL2CR interacts with Cul4a, Ddb1, and Crbn (see also Fig. 4B); Stub1 and three other E3 ubiquitin-protein ligases found in the ACR pulldown were Nedd4, Arih1, and Ubr4. As was the case for St-ACR pulldowns, Ddb1, Crbn, Stub1 and Cul4a were the four most abundant UPS-related proteins found in the St-AL2CR pulldown. In contrast, UPS-related proteins were scarcely represented in AL1CR pulldowns.
The COOH terminus of the ACR is essential for binding CRL4 CRBN ; the AL2CR binds CRL4 CRBN , but the AL1CR does not; the COOH-terminal four amino acids of the ACR share 75 and 0% identity with the corresponding region of the AL2CR and AL1CR, respectively (Fig. 4A). These observations suggest that, like for the ACR, the COOH-terminal region of the AL2CR binds CRL4 CRBN via the substrate-binding pocket of Crbn. To test for this, we determined whether Crbn is required for binding of Ddb1 to the AL2CR. As shown in Fig. 4C, St-AL2CR binds Ddb1 when challenged with brain lysates isolated from WT animals but not when the input material is isolated from Crbn-KO mice. Next, we tested whether thalidomide and lenalidomide interfere with the AL2CR-Crbn interaction; indeed, both thalidomide and lenalidomide reduced binding of Crbn to St-AL2CR (Fig. 4D). Thus, like for the ACR, the AL2CR binds CRL4 CRBN via the substrate-binding pocket of Crbn. Overall, these data suggest that APP and APLP2, but not APLP1, may functionally interface with CRL4 CRBN and Stub1.    ACR and the AL2CR Pull Down Active E3 Ubiquitin-Protein Ligase(s)-Ubiquitination of proteins is a multistep reaction. First, an E1 ubiquitin-activating enzyme activates ubiquitin. Next, ubiquitin is transferred on one of many ubiquitin-conjugating enzyme (E2s). The E2 is loaded on an E3 ubiquitin-protein ligase, which directs the mono-and poly-ubiquitination of substrate proteins. The presence of many of these components in St-ACR and St-AL2CR pulldowns prompted us to test whether ubiquitinating activity could be detected. To this end, we performed an in vitro ubiquitination assay using FLAGubiquitin as an external source of ubiquitin. The reactions were analyzed by Western blotting using an ␣-FLAG antibody to determine whether the FLAG-ubiquitin (ϳ9 kDa) was incorporated into larger complexes. A smear of FLAG-ubiquitinated proteins was clearly detected in the St-ACR and St-AL2CR pulldowns after supplemental addition of recombinant E2 UbcH5a/UBE2D1 (Fig. 5A). Overimposed on the smear was a ladder of bands starting at ϳ25 kDa, which are ϳ9 kDa apart (Fig. 5A). The ϳ25-kDa band is compatible with mono-ubiquitinated recombinant E2; the higher bands likely represent recombinant E2 linked to a progressively increasing number of ubiquitin molecules. Addition of recombinant E1 Ube1, albeit not necessary, further potentiated the ubiquitination reactions (Fig. 5A). The observation that only recombinant E2 is required to initiate ubiquitination activity (Fig. 5A) indicates that St-ACR and St-AL2CR bring down from the mouse brain functional E1 and E3 ubiquitin-protein ligase(s). Of note, neither the smear nor the ladder was visible in the in vitro ubiquitination assay of St and St-AL1CR eluates (Fig. 5A), further stressing that APP and APLP2, but not APLP1, may functionally interface with the ubiquitination system.
Because active E3 ubiquitin-protein ligases undergo autoubiquitination, we tested whether Crbn was ubiquitinated in vitro. The St-ACR in vitro ubiquitination assay was analyzed by Western blotting with an ␣-Crbn antibody; in addition to fulllength Crbn, a ladder of signals compatible with Crbn molecules linked to an increasing number of ubiquitins was detected (Fig. 5B). This evidence suggests that St-ACR interacts with an active brain-derived CRL4 CRBN complex.
ACR Facilitates Ubiquitination of ACR-interacting Proteins in Vitro-The smear detected by the ␣-FLAG antibody hints at ubiquitination of numerous proteins. To test this hypothesis, we performed another in vitro ubiquitination assay on St-ACR Thr(P)Tyr(P) pulldown because it includes proteins that interact with APP in a Thr(P) 668 -and Tyr(P) 682 -dependent manner. The reaction was immunoprecipitated with the ␣-FLAG M2-agarose beads, and proteins were eluted from the M2-agarose beads using a competing 3ϫFLAG peptide. To identify the lysine residues ubiquitinated (K-ub) in vitro, the eluted material was subjected to the ubiquitin remnant motif (K-⑀-GG) assay (UbiScan). In parallel, we performed UbiScan experiments on total mouse brain lysates to determine whether K-ubs found in the in vitro assay are also ubiquitinated in vivo.
In addition, we used as further reference the Cell Signaling Data Bank collection of UbiScan results from rodent, human, and chicken tissue, which can be found on line.
As reported in Table 6, several brain-derived proteins associated with the ACR bait were ubiquitinated. These proteins can be separated into four groups as follows: (a) group 1, proteins of the ubiquitin-conjugating system; (b) group 2, presynaptic proteins; (c) group 3, other proteins implicated in AD; and (d) group 4, phosphorylation-dependent ACR interactors.
Group 1: consistent with the Western blot showing ubiquitination of Crbn (Fig. 5B), we found six K-ub from Crbn, four of which have been found in vivo. Two other components of   37 and Lys 615 , new; Lys 491 , found in our mouse brain UbiScan; Lys 516 , previously found in vivo), Ube3a (Lys 28 , new). Finally, brain-derived E1 Ube1 and E2s Ube2n and Ube2o (which is actually an E2/E3 hybrid ubiquitin-protein ligase) were also ubiquitinated: all these ubiquitinations have been found in vivo. Group 2: in this group are included the following single-pass and multi-pass synaptic vesicle membrane proteins (Fig. 6): the vesicular SNARE Vamp2; the Ca 2ϩ -sensors synaptotagmin-1 and -2 (Syt1 and Syt2); the secretory carrier-associated membrane protein 1 Scamp1; the synaptic vesicle glycoproteins Sv2a and Sv2b; the vesicular glutamate transporter Vglu1, and Atp6v0a1 that is required for assembly and activity of the vacuolar ATPase. All these K-ub are found in vivo. Moreover, these ubiquitinated lysine residues are all localized in the cytosolic domains (Fig. 6); this is relevant because in physiological conditions only lysine residues exposed to the vesicular lumen are not available to the ubiquitination system, suggesting that the in vivo ubiquitination of these two lysines is plausible. Other proteins associated with synaptic vesicles, such as the trans-SNARE Snap25, Synapsin2 and Synapsin1 (Syn1/2), and the G proteins Rab3a and Rab14, were ubiquitinated in vitro on lysine residues that, with the exception of Lys 280 of Syn2, are also ubiquitinated in vivo.
Group 3: this group includes the tubulin-binding protein Tau (93) and the cholesterol transporter apoE (94). Both proteins were detected in St-ACR pulldowns ( Table 7). The two Tau K-ubs found in vitro (Lys 455 and Lys 613 ) were also identified in our mouse brain UbiScan. It is worth noting that Stub1 can mediate Tau ubiquitination (95,96) suggesting that the ACR may facilitate Tau ubiquitination via interaction with Stub1 on one side and Tau on the other side. Alternatively, the ACR may interact with Tau indirectly, via Stub1. It is worth noting that a previous study has described a physical interaction between APP and Tau in vivo (97). ApoE was ubiquitinated on Lys 105 . ApoE is a secreted protein and its ubiquitination may be an in vitro artifact. However, ubiquitination of Lys 105 of apoE has been reported in vivo, suggesting that apoE is, at least in a small fraction or in certain conditions, resident in the cytosol. Indeed, previous work has shown that apoE is cleaved by a neuro-specific chymotrypsin-like serine protease that generates bioactive

In vitro ubiquitination of proteins present in the St-ACR Tyr(P)Thr(P) pulldowns
1st column lists some of the proteins ubiquitinated in vitro in an ACR Tyr(P)Thr(P) -dependent manner. 2nd column lists all the lysine residues found ubiquitinated in vitro. The K-ub found in our UbiScans from mouse brains are indicated with (m*) and in previous UbiScan experiments from mouse (m), human (h), rat (r) and chicken (c) tissues, which are reported online. 3rd column reports database accession numbers. intraneuronal truncated forms of apoE (98,99) and that apoE is ubiquitinated in cell lines (100). Group 4: we also found two Grb2 K-ubs and five Pin1 K-ubs. All of these K-ubs are found in vivo. This evidence suggests that phosphorylation of APP on Thr 668 and Tyr 682 could potentially alter the function of proteins that bind APP in a phosphorylation-dependent manner by regulating their ubiquitination.

Ubiquitin conjugating system
Although it is doubtful that all the ACR-dependent ubiquitinations detected in this in vitro assay are physiologically relevant, the data suggest that APP may work as a substrate recognition subunit of one or more E3 ubiquitin-protein ligases (possible candidates are CRL4 CRBN and Stub1), and APP may regulate ubiquitination of some of the proteins described here.

In Vitro Ubiquitination of the ACR-interacting Proteins Occurs on a Subset of the Lysine Residues Ubiquitinated in
Vivo-Next, we compared the in vitro UbiScan with the UbiScan performed in parallel on mouse brain lysates. We found that only few of the lysine residues ubiquitinated in the brain were also ubiquitinated in vitro. A few examples are shown in Table 8. Only one lysine residue of Scamp1, Sv2a, SNAP-25, and apoE was ubiquitinated in vitro. In contrast, additional K-ubs (six for Scamp1 and -3 for Sv2a and four for apoE) were detected in mouse brains. Syt1 and Tau were ubiquitinated on 19 and 16 lysine residues in mouse brains, respectively; of these, only five and two lysine residues were ubiquitinated in vitro, respectively. This evidence suggests the following. 1) APP may facilitate ubiquitination of a subset of "ubiquitinable" lysine residues on potential substrates and as a result APP could fine-tune the function of substrates with high accuracy by targeting lysine residues residing in specific functional domains. 2) Additional lysine residues of putative APP substrates are presumably ubiquitinated by distinct E3 ligases, possibly with distinct functional outcomes.
3) The stability of potential APP substrates can be regulated by APP-dependent and APP-independent mechanisms.
ACR Is Ubiquitinated in Vitro, Evidence for a Role of CRL4 CRBN in Ubiquitination of Lys 676 of the ACR-As expected, recombinant E1 and E2 were ubiquitinated in vitro (data not shown). Interestingly, we also found ubiquitination of synthetic ACR on Lys 649 , Lys 650 , Lys 651 , Lys 676 , and Lys 688 (see Table 9). Ubiquitination of Lys 651 , Lys 676 , and Lys 688 was also found in normal tissues, including our mouse brain UbiScan, whereas ubiquitination of Lys 649 and Lys 650 has been reported in cell lines (101). These data suggest that E3 ubiquitin-protein ligases present in the pulldown may be responsible for APP ubiquitination in vivo.
To test the role of CRL4 CRBN in ubiquitination of the ACR, we performed an in vitro ubiquitination assay of the ACR pulldowns from either WT or Crbn-KO mouse brains followed by UbiScan analysis. As shown in Table 9, ubiquitination of Lys 676 was significantly reduced in pulldowns from Crbn-KO brains; in contrast, ubiquitination of Lys 649 , Lys 650 , Lys 651 , and Lys 688 was similar in both WT-and Crbn-KO-derived samples. Together these data suggest the following. 1) CRL4 CRBN E3 ubiquitin-protein ligase, which is absent in pulldown from Crbn-KO brains, may have a primary, perhaps not exclusive, role in ubiquitination of APP on Lys 676 . 2) Other E3 ubiquitinprotein ligases participate in ubiquitination of residues Lys 649 , Lys 650 , Lys 651 , and Lys 688 .

Discussion
In this study, we provide in vitro evidence suggesting that APP may function as a substrate recognition unit for one or more E3 ubiquitin-protein ligases. This hypothesis is supported by the following findings. 1) The cluster of brain proteins that may interact with APP via the ACR is rich in proteins that regulate the UPS (Table 1). 2) The E3 ubiquitin-protein ligases CRL4 CRBN and Stub1 are the most abundant UPS-related proteins interacting with the ACR (Tables 1 and 3 and Fig. 1B). 3) The E3 ubiquitin-protein ligase interacting with the ACR is enzymatically active and ubiquitinates the five lysine residues Notably, all the lysine residues ubiquitinated in vitro are in segments that are exposed to the cytosol.

TABLE 7
Proteins ubiquitinated in vitro (see Table 6) interact with the ACR Table lists the proteins identified (1st column); the molecular mass in kDa (2nd column); NSAF (4th to 8th columns). Here we show the proteomic data documenting the interaction of Vglu1, Atp6v0a1, Scamp1, Rab14, Rab3a, Syn2, Syn1, ApoE, and Tau with the ACR. present in the APP cytoplasmic region ( Fig. 5A and Table 5). 4) The APP cytosolic domain facilitates ubiquitination of interacting proteins in vitro ( Table 6). The five cytoplasmic APP residues (Lys 649 , Lys 650 , Lys 651 , Lys 676 , and Lys 688 ), which are ubiquitinated in vivo, are also ubiquitinated in vitro in ACR pulldowns (Table 9) suggesting that the E3 ubiquitin-protein ligase(s) present in the pulldown interaction may be responsible for ubiquitination of APP in vivo. The evidence that ubiquitination of Lys 676 , but not that of the other four ACR lysine residues, is significantly reduced in the absence of CRL4 CRBN (Table 10) suggests the following: 1) Lys 676 of APP may be physiologically ubiquitinated mainly, but not exclusively, by the CRL4 CRBN E3 ubiquitin-protein ligase; 2) Lys 649 , Lys 650 , Lys 651 , and Lys 688 are probably targets of other E3 ubiquitin-protein ligases, although a role for CRL4 CRBN cannot be formally excluded, like Stub1 that is very abundant in the ACR pulldown. Because the subunits of CRL4 CRBN undergo auto-ubiquitination (Table 6 and Fig. 5B), the possibility that APP may at the same time act as a substrate recognition unit and a substrate of a CRL4 CRBN/APP E3 ubiquitin-protein ligase is not far-fetched.

Proteins kDa St ACR ACR Tyr(P)Thr(P) ACR Tyr(P) ACR Thr(P)
APP belongs to a protein family that includes APLP1 and APLP2. Analysis of single and double knock-out (KO and dKO) mice has shown that App-KO, Aplp1-KO, Aplp2-KO, and App/ Aplp1-dKO have minor deficits. In contrast, App/Aplp2-dKO mice have severe neuromuscular junctions deficits, are significantly smaller than App-KO and Aplp2-KO mice, and die within the first 28 days of life (25,43,102,103). These data indicate that APP and APLP2 share some essential function that cannot be compensated for by APLP1. However, the molecular mechanisms mediating this essential function (or functions) of APP and APLP2 are unclear. Here, we show that the brain interactomes of the ACR and the AL2CR, but not of the AL1CR, share many UPS-related proteins, including Stub1 and CRL4 CRBN (Table 5 and Fig. 4B). Moreover, the AL2CR interacts, like the ACR, with the substrate recognition pocket of Crbn (Fig. 4, C and D). Finally, the ACR and the AL2CR brain interactomes possess E3 ubiquitin-protein ligase activity, whereas the brain interactome of the AL1CR does not (Fig. 5A). Altogether, these data suggest that both APP and APLP2 may possess E3 ubiquitin-protein ligase substrate recognition activity, which is not compensated for by APLP1. Thus, the loss of this activity may mechanistically cause the severe phenotype of App/Aplp2-dKO mice.
APP facilitates glutamatergic transmitter release, likely through the interaction with the neurotransmitter release machinery (28). In addition, the APP intracellular domain has been linked to many other pathological and functional pathways, including caspase activation, transcription, Ca 2ϩ flux, and neurodegeneration (28, 31, 32, 34, 40, 46, 64, 104, 106 -108). It is tempting to speculate that APP may affect all these seemingly unrelated pathways via a single mechanism of action. Indeed, by acting as a substrate recognition unit for E3 ubiquitin-protein ligase(s), APP could modulate ubiquitination of proteins involved in these processes (model in Fig. 7). Thus, using a single modus operandi, APP may regulate disparate and apparently unrelated signaling pathways. As we show here, many of the proteins that control synaptic vesicle functions and interact with APP are ubiquitinated in vitro in an ACR-dependent manner (Table 6). Consequently, APP may aid synaptic vesicle activity by modulating ubiquitination of these proteins, thereby altering their stability and/or their function.
Eleven UPS-linked proteins isolated by St-ACR and five brain-derived proteins that are ubiquitinated in vitro in an ACR-dependent manner are associated with genetic diseases of the central nervous system and neuromuscular system. These include HUWE1, CRBN, CUL4B, USP9X, PARK7, STUB1, TABLE 8 In

vitro ubiquitination of ACR-interacting proteins occurs on a subset of the lysine residues ubiquitinated in vivo
The 1st column lists some of the proteins ubiquitinated in vitro in an ACR-dependent manner. The 2nd column reports all the lysine residues found ubiquitinated in the mouse brain lysate UbiScan that we performed (in vivo ubiquitinations). The 3rd column reports the lysine residues found ubiquitinated in our in vitro ubiquitination/ UbiScan experiment.  The 1st column shows the site of modification for the peptide assignment. The 2nd column reports the amino acid sequence for the peptide assignment with the ubiquitinated lysine (*). The 3rd column reports the relative fold-change between the integrated peak area of the experimental (numerator, Crbn-KO) and control (denominator, WT) conditions. A negative value indicates the peptide is more abundant in the control condition.   (109 -139). These observations suggest a molecular and functional connection between APP and other proteins genetically involved in Alzheimer and other neurodegenerative diseases. Tau is the main component of neuronal tangles that characterize AD, and TAU mutations are linked to genetic forms of frontotemporal dementia. Previous studies have shown that APP and Tau interact in vivo (97). Interestingly, Stub1 mediates ubiquitination of Tau, and Tau is polyubiquitinated at several sites in AD patients (140 -142). Thus, a role of APP in Tau ubiquitination would be both biologically and pathologically significant. APOE is the main genetic risk factor for sporadic AD, with the APOE4 allele increasing the risk of developing late onset AD (137). Our in vitro evidence suggesting that APP may both interact with and regulate ubiquitination of apoE is inciting and needs to be developed further. Four of the proteins discussed in this study (HUWE1, CRBN, CUL4B, and USP9X) are genetically linked to intellectual disability (previously known as mental retardation). Remarkably, this list includes two CRL4 CRBN E3 ubiquitin-protein ligase components, CRBN and CUL4A. Overall, these data suggest a functional network that comprises many proteins that, when functionally altered by genetic mutations, trigger neurodegenerative disorders. Hence, it is reasonable to speculate that dysregulation of this functional network, in which APP as a modulator of E3 ubiquitin-protein ligase(s) may play a crucial role, could be a pathogenic mechanism shared by numerous neuronal disorders. This concept is not outlandish because accumulation of ubiquitinated protein inclusions in neurodegenerative diseases and the link between mutations in proteins involved in the UPS and neurodegenerative disorders are both established truths (52,105,143).
Stub1 and CRL4 CRBN could contemporarily bind to the ACR because they interact with the NH 2 and COOH termini of this APP region, respectively. Thus, cleavage of APP by caspases at Asp 664 could functionally separate the activities linked to the APP-Stub1 and APP-CRL4 CRBN complexes.
In future studies, it will be important to test these hypotheses, to determine whether pathogenic APP mutation regulate this function of APP and whether this deregulation has any pathogenic role in neurodegeneration.  Pulldown Assays with St-peptides-The St-peptides were immobilized on StrepTactin column (catalog no. 2-1209-550, IBA-GmbH, Goettingen, Germany). S2 plus LS1 brain fractions were pre-cleared on StrepTactin columns containing no Stpeptides. Pre-cleared). Pre-cleared S2 plus LS1 brain fractions were next passed through the StrepTactin column loaded with St-peptides. The columns were then washed, and St-peptides, together with brain proteins specifically bound to the St-peptide, were eluted with desthiobiotin following the manufacturer's recommendations. In some pulldowns, brain lysates were incubated for 1 h at 4°C with the indicated concentrations of either lenalidomide (catalog no. T2800, lot 2570277, LKT Laboratories, Inc., St. Paul, MN) or thalidomide (catalog no. 0652, batch 11A/141284, Tocris Bioscience, Bristol, UK), prior to pulldown with St-peptides.
Immunoprecipitation of the in Vitro Ubiquitination Assays-To isolate proteins ubiquitinated in vitro, the in vitro ubiquitination assay performed on St-ACR pulldown was incubated with FLAG-M2 affinity gel (catalog no. A2220, lot SLBF8148, Sigma), under constant rotation for 3 h at 4°C. The agarose beads were collected by centrifugation and washed five times in phosphate-buffered saline plus 0.05% Tween. After washing, the proteins bound to FLAG-M2 affinity gel were eluted by incubation with 100 M concentration of the 3ϫFLAG competing peptide (catalog no. F4799, lot SLBG0131V, Sigma).
Mouse Brain Preparation for UbiScan-To process the brain tissue for UbiScan analysis, brains from at least eight animals of the same genotype for each experiment were cut into small pieces and lysed in freshly prepared urea lysis buffer (20 mM Hepes, 9 M urea, 1 mM sodium orthovanadate, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerol phosphate), using 4 ml of buffer for 100 mg of tissue. Samples were homogenized twice using a Polytron set to maximum speed, using 20-s-long pulses. Between pulses, samples were chilled on ice for 1 min. Successively, samples were sonicated using a microtip set to 15-watt output using three bursts of 30 s each. Between bursts, samples were chilled on ice for 1 min. Finally, the lysates were cleared from debris by centrifugation at 20,000 ϫ g for 15 min at 4°C.
UbiScan Analysis-This analysis was performed by the PTM-Scan Facility at Cell Signaling Technology. Briefly, samples were digested with trypsin; after digestion, peptides were loaded directly onto a 10 cm ϫ 75 m PicoFrit capillary column packed with Magic C18 AQ reversed-phase resin. The column was developed with a 90-min linear gradient of acetonitrile in 0.125% formic acid delivered at 280 nl/min. repeat duration 35.0; exclusion list size 500; exclusion duration 40.0; exclusion mass width relative to mass; exclusion mass width 10 ppm). MS/MS spectra were evaluated using SEQUEST 3G and the CORE platform from Harvard University. Searches were performed against the most recent update of the NCBI Mus musculus database with mass accuracy of Ϯ50 ppm for precursor ions and 1 Da for product ions. Results were filtered with mass accuracy of Ϯ5 ppm on precursor ions and presence of the intended motif (K-⑀-GG).
Mass Spectrometry-MS Bioworks, LLC, Ann Arbor, MI, performed this analysis. The volume of each pulldown sample was reduced to 50 l by vacuum centrifugation; 20 l of each concentrated sample was processed by SDS-PAGE using a 10% BisTris NuPAGE gel (Invitrogen); with the MES buffer system the gel was run ϳ2 cm. The mobility region was excised into 10 equally sized segments, and in-gel digestion was performed on each using a robot (ProGest, DigiLab) with the following protocol: washing with 25 mM ammonium bicarbonate followed by acetonitrile; reduced with 10 mM dithiothreitol at 60°C followed by alkylation with 50 mM iodoacetamide at room temperature; digested with trypsin (Promega) at 37°C for 4 h; quenched with formic acid; and the supernatant was analyzed directly without further processing. Each digest was analyzed by nano-LC/MS/MS with a Waters NanoAcquity HPLC system interfaced to a ThermoFisher Q Exactive mass spectrometer. Peptides were loaded on a trapping column and eluted over a 75-m 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 resolution, respectively. The 15 most abundant ions were selected for MS/MS.
Author Contributions-L. D. conceived and coordinated the study and wrote the paper. D. D. P. and L. D. designed, performed, and analyzed all the experiments. D. D. P. contributed to writing the paper. A. M. R. provided reagents/knock-out mice and contributed to writing the paper. R. C. R. generated knock-out mice. L. D. and D. D. P. prepared the figures. L. D. prepared the tables. All authors reviewed the results and approved the final version of the manuscript.