Integrated Mechanism of Lysine 351,PARK2,and STUB1 in AβPP Ubiquitination

Abstract

Intracellular accumulation of aggregated amyloid-β, misfolded and non-functional proteinopathy, is the hallmark feature in Alzheimer’s disease (AD). There are several mechanisms to clear the amyloid burden in a cell, including transcytosis across the blood-brain barrier, immune mediated, lysosomal pathway associated autophagy, enzymatic degradation by insulin degrading enzyme/neprilysin, and the proteasomal pathway. Among them, the ubiquitin proteasome system (UPS) is playing a critical role to prevent the intracellular amyloid-β deposition and to clear off the cellular burden in association with ubiquitin E3 ligase enzymes in AD. For ubiquitination, lysine moiety in a protein acts like a docking site for the attachment of ubiquitin molecule and different lysine residues act differently in this reaction. Therefore, it is pertinent to understand and link the role of various lysine residues along with their effector molecules, for instance, E3 ligases PARK2 and STUB1 in the ubiquitination cascade. Herein, we 1) modeled the structure of AβPP and determined its topologies and studied the impact of lysine residues in AβPP stability, 2) reported K351 as the most promising target for AβPP ubiquitination, 3) investigated the plausible role of lysine residues in non-covalent interactions mediated ubiquitin positioning in the ubiquitination, 4) detected conserved amino acids that is crucial for AβPP ubiquitination, and 5) identified the key ubiquitination enzymes and their interaction network playing major role in the ubiquitination of AβPP.

Keywords

Amyloid-β amyloid-β protein precursor lysine neurodegeneration therapeutics ubiquitination

INTRODUCTION

Alzheimer’s disease (AD) is characterized by the aberrant protein accumulation and aggregates such as amyloid-β (Aβ), leading to the development of senile plaques in the brain [1]. In this regard, the amyloid cascade hypothesis affirms the Aβ deposition as an early pathogenic event in the progression of AD [2]. The intra-and extracellular Aβ_40/42 deposits perturb the synaptic transmission between neurons and trigger the memory and cognitive decline in AD patients [3]. Here, Aβ and its precursor, AβPP, are the central players in the pathology of AD whose levels are regulated by the protein quality control: ubiquitin proteasome system (UPS). There are other mechanisms that regulate Aβ clearance including enzymatic pathways utilizing insulin degrading enzyme and neprilysin [4], immune-mediated trans-cytosis across the blood-brain barrier [5], and the lysosomal pathway through autophagy [6]. However, the ubiquitination pathway is central to regulate the level of proteins involved in all these pathways; therefore, it has a great importance in the cellular biology for the clearance of Aβ. The ubiquitination process involves the selection of key lysine residues of target protein by E3 ligases for ubiquitin attachment. This ubiquitin ligation at key lysine and their type of poly-ubiquitin chain determines the cellular fate of the target protein [7]. For instance, studies have identified K6 poly-Ub-chain to trigger DNA repair responses and K33 poly-Ub-chain to initiate stress responses. Moreover, K63 poly-Ub-chain to govern DNA repair, endocytosis and inflammatory responses while K11, K27, K29, and K48 poly-Ub-chain to elicit proteasomal degradation [8, 9].

In this manner, UPS greatly influences Aβ production through AβPP ubiquitination and through proteasomal degradation of their regulatory enzymes (β- and γ-secretases). Moreover, there is increasing evidence of non-functional UPS, i.e., ubiquitinating and deubiquitinating enzymes and the proteasomal subunits that are responsible for the altered Aβ clearance in AD patients [10]. However, there are numerous reports depicting the ubiquitination of Aβ or AβPP in vitro, ex vivo, and in vivo but their exact site for ubiquitination remained unknown for most cases (Table 1). Moreover, the AβPP ubiquitination research is at the preliminary stage, where much needs to be investigated to unravel the mystery of amyloid burden mitigation. Therefore, identification of the key lysine residues and the ubiquitination components are crucial for determining the mechanism behind Aβ clearance. In this regard, we have identified the key lysine residues having great potential for AβPP ubiquitination. Moreover, the importance of these key lysine residues in AβPP processing, non-covalent interactions with ubiquitin and other functions have also studied. Further, we modeled the 3D structure of holo-AβPP to investigate the impact of key lysine residues on AβPP stability and their mutational disease susceptibility. Nevertheless, we have also identified the potential ubiquitination enzymes, E1s, E2s, E3s, and DUBs, and their complex interplay in the ubiquitination process of AβPP. In summary, this work demonstrates the AβPP ubiquitination mechanism and Aβ clearance to provide novel therapeutic targets against AD.

Table 1

Experimental studies reporting the ubiquitination of AβPP and Aβ to rescue the amyloid burden in AD

S. No.	Experimental Models	UPS Enzymes	Ubiquitination Sites	Ubiquitination Effects	Reported Studies
1	APP/PS1 Transgenic mice	Parkin	_	AβPP degradation	[11]
2	Human Neuroblastoma M17 Cells, Rat Brains		_	Aβ degradation	[12]
3	SHSY5Y Cells, Human and Transgenic Mouse (Parkin K/O) Brain Samples		_	Aβ/AβPP degradation	[13, 14]
4	APP23/PS45 Mice	UCHL1	_	Regulate AβPP degradation	[15]
5	Crbn-KO Mouse Brains	CRL4	K676(AβPP695)/K751(AβPP770)	AβPP interactions with other proteins	[16]
6	HeLa Cells, 2×Tg Mice (FBL2/AD1)/3×Tg Mice	FBL2	K649–651(AβPP695)/K724–726(AβPP770)	AβPP metabolism, Endocytic trafficking	[17, 18]
7	Hippocampal Neurons, Fbxo2 KO Mice	Fbxo2	_	AβPP processing and degradation	[19]
8	N2a Cells, P0 Mice Brain		_	K612(AβPP695)/K687(AβPP770)	Endosomal sorting of AβPP	[20]
			K624(AβPP695)/K699(AβPP770)
9	SHSY5Y Cells, ddY Mice	HRD1	_	AβPP degradation	[21]
10	CHO Cells	HRD1	_	AβPP degradation	[22]
11	SHSY5Y Cells, Human Brain Sample	CHIP	_	AβPP degradation	[23]
12	Primary Neuron Culture	_	_	Aβ degradation	[24]
13	Rat Cortical Neurons	_	_	Aβ/AβPP degradation	[25]
14	PC12 Cells	Ubiquilin-1	K688(AβPP695)/K763(AβPP770)	AβPP biosynthesis, trafficking, and degradation	[26]
15	SHSY5Y Cells	_	_	AβPP degradation	[27]

MATERIALS AND METHODS

AβPP structure modeling

The AβPP structure was modeled with a protein homology/analogy recognition tool Phyre². It included the assembly of homologous sequences by multiple sequence alignments of query sequence with their sequence homologs and their consequent secondary structure prediction by PSI-pred pooled query hidden Markov model. Further, fold library was scanned for crude backbone construction based on top scoring alignments. Further, loop modeling for correcting insertions and deletions in the models followed by side chain placement to backbone to obtain a final protein structure model [28]. Here, six templates (c3ktmB, c3dxeB, c2yszA, d1rw6a, c1amlA, c2lp1A) were employed to model the AβPP protein structure based on heuristics method to maximize the confidence, percentage identity, and alignment coverage to the queried sequence.

AβPP domain analysis

The protein domain analysis was done to identify the functional sites in the AβPP with MOTIF Search tool. It extracted domain results on the basis of matched bit score for AβPP sequence and each domain found in Pfam library with an E-value cut-off of 0.001 [29].

Structure refinement simulations

The predicted 3D-structure of AβPP was refined with an atomic-level, high-resolution protein structure refinement tool-ModRefiner [30]. We performed the conformational search for the backbone and side chain atoms to simulate the predicted model to their native state in terms of backbone topology, side chain positioning and hydrogen bonds directed by combined physics-and knowledge-based force field energy calculations.

AβPP structural validations through PROVE, ERRAT, VERIFY 3D, and RAMPAGE

The reliability of 3D-atomic models of predicted AβPP validated through the stereo chemical quality measurements of modeled proteins with different structural validation programs. PROVE analyzed the volume-based quality of protein crystal structure by computing the statistical Z-score deviations of the atomic volumes from their standard values [31]. The buried atoms less than 1% passed the structural quality test while greater than 5% implied the structural irregularities. ERRAT determined overall quality factor by identifying the statistical differences between the patterns of non-bonded atomic interactions (ordered versus randomized distribution) and the error functions of the predicted model with the statistics of highly refined structures [32]. VERIFY 3D compared 3D atomic model compatibility with its own primary sequence (1D) associated structural class assigned based on their environment and location [33]. RAMPAGE geometrically validated C-alpha neighboring residues of modeled AβPP [34], which allowed the detection of C-beta bond angle distortions.Moreover, it defined favored and allowed phi-psi regions for glycine, proline, and pre-proline residues to validate the accuracy of the model.

Identification of potential ubiquitination sites in AβPP

The important lysine residues for AβPP ubiquitination was identified with four different approaches, including 1) the sequence identity between AβPP and ubiquitin lysine sites, 2) sequence similarity between AβPP and ubiquitin lysine sites, 3) ubiquitin-ubiquitin and ubiquitin-AβPP lysine peptides docking, and 4) machine learning based ubiquitination site predictions.

Sequence identity analysis

Pairwise sequence alignment was performed using BioEdit Sequence Alignment Editor Software [35] using BLOSUM 62 substitution matrix. Further, it calculated the identity scores among the 21 window-size central lysine’s neighboring residues of ubiquitin and AβPP sequences.

Sequence similarity analysis

Conserved sequence analysis of amino acid residues neighboring lysine sites was performed by aligning the 21 window-size multiple lysine site sequences in AβPP by the BioEdit Sequence Alignment Editor Software [35] at different thresholds to identify the important residues common to both ubiquitin and AβPP critical for the ubiquitination.

Ubiquitin-AβPP lysine peptides docking

The protein-peptide interactions of ubiquitin protein and AβPP lysine peptides were modeled by flexible docking approach using CABS-dock web server. It executed simulation searching for the binding sites in the receptor protein allowing full flexibility to the peptides being docked [36]. For ubiquitin-AβPP peptide docking, 21 window-size peptide sequences were prepared by taking 10 amino acid residues on either side of the lysine sites from AβPP protein as test set and from ubiquitin protein as control set, respectively. Further, ubiquitin protein (PDB ID: 1UBQ) is flexibly docked with both the test set peptides, and the control set peptides to identify the potential lysine for ubiquitination.

Machine learning techniques: UbiSite, BDM-PUB, hCKSAAP_UBSITE, UbPred, UbiPred

The potential ubiquitination sites for AβPP was predicted by diverse machine learning tools, including UbiSite, BDM-PUB, CKSAAP, UbPred, and UbiPred on the basis of the experimentally verified ubiquitination site datasets. UbiSite employed maximal dependence decomposition method (specificity level High at 95%) and support vector machine based on the hybrid features for the large-scale ubiquitin conjugation sites data to predict ubiquitination sites [37]. BDM-PUB employed Bayesian Discriminant analysis (at balanced cut off) between high specificity and high sensitivity to apply a probabilistic approach for ubiquitination pattern recognition [38]. hCKSAAP_UBSITE utilized SVM classifiers based diverse amino acid pattern and propensities trained by logistic regression to predict the ubiquitination sites in AβPP [39]. UbPred relied upon a random forest-based prediction model trained on experimentally verified ubiquitination sites to predict the ubiquitination pattern in AβPP [40]. UbiPred employed SVM to predict the ubiquitination sites based on the principle of the informative physicochemical property mining algorithm [41].

Functional studies of lysine’s impact on AβPP ubiquitination

The refined AβPP structure was incorporated with both positive (Arginine (R), Histidine (H)) and negative (Aspartic acid (D), Glutamic acid (E)) charged mutations at their lysine sites by Pymol software [42]. The above mutations were studied for their impact on AβPP stability, ubiquitination, modifications, and disease susceptibility.

Investigation of AβPP stability upon lysine mutations

The refined AβPP model and mutated AβPP (Lysine (K)⟶Arginine(R)/Histidine(H)/Aspartic acid(D)/Glutamic acid(E)) analyzed for their total force field energies by Swiss PDB viewer 4.0.2 software [43]. The variations in the total force field energies were estimated to observe the impact of lysine mutations on the amyloid beta precursor protein’s stability.

Physico-chemical property analysis for AβPP ubiquitination

The physico-chemical properties of Aβ, AβPP, and ubiquitin proteins were computationally determined from their peptide sequences by ExPASy server tool ProtParam [44]. It computed various physico-chemical properties, including the atomic and amino acid compositions along with the count of positively and negatively charged residues, the instability index, aliphatic index, and the grand average of hydropathicity of the desired proteins. The physico-chemical properties of AβPP was compared with ubiquitin to investigate the crucial parameters important for the ubiquitination.

Examination of lysine sites crucial for AβPP modifications

The molecular mechanisms associated with the pathogenic amino acid substitutions in AβPP were identified by a machine-learning tool MutPred [45]. It quantified the pathogenicity of mutations based on the probabilistic modeling of a large repertoire of structural-functional alterations, including the disruptions in structure, stability, macromolecular binding, and post-translational modification sites in amino acid sequences.

Mutational analysis of lysine residues for disease susceptibility

The functional impacts of the introduced mutations were studied with different mutation analysis tools, including PANTHER [46], SNAP2 [47], Polyphen2 [48], Pmut [49], PhD-SNP [50], and SIFT [51]. The obtained results were transformed into numerical values to analyze them on the stacked bar graph. For instance, results with “Probably Benign”, “Neutral”, and “Tolerated” were assigned with ‘0’ numerical value and “Possibly Damaging” with ‘1’ numerical value. Similarly, results with “Probably Damaging”, “Effect”, and Not tolerated” were assigned with “2” numerical value and the threshold was taken ‘6’, i.e., more than half prediction tools to predict high confidence lysine having disease susceptibility.

Characterization of ubiquitination pattern of AβPP

The ubiquitination pattern of AβPP was investigated through the combination of multiple studies, including 1) the AβPP-ubiquitin interaction studies, and 2) the prediction of ubiquitination enzyme’s interactional network for AβPP or Aβ clearances in AD biology.

AβPP-ubiquitin interaction prediction and interface residue analysis

The interactions among the different domains of AβPP and ubiquitin proteins were predicted by the protein-protein interaction prediction server-PSOPIA [52]. It compared the sequence similarities to a known interacting protein pair, statistical propensities of the domain-domain interactions, and the sum of edge weights along the shortest path between the homologous proteins in a PPI network. Moreover, the potential ubiquitin interacting domains of AβPP (PDB IDs: 4JFN_A, 2FKL_A, 1AAP_A, 3UMK_A, 1IYT_A, 2LP1_A) were docked with ubiquitin (PDB ID: 1UBQ) by rigid body docking approach using Fast Fourier Transformation methodology of GRAMM-X software [53]. Furthermore, their docked interface residues having distance <4.5A were identified by Pymol software [42] to identify the key lysine residues important for ubiquitin and AβPP interactions.

Prediction of ubiquitination enzymes interactional network for AβPP

The potential ubiquitination enzymes regulating the AD biology of Aβ and AβPP proteasomal clearance were identified by determining the interaction among all the ubiquitin E1-activating enzymes, E2-conjugating enzymes, E3-ligating enzymes, and deubiquitinating enzymes with amyloid forming proteins: AβPP, β-secretases and γ-secretases. Further, the protein-protein interactional network among the identified proteins was designed by functional protein association network prediction STRING tool [54]. Lastly, the overall methodology along with the tools and techniques employed in this study has shown in Fig. 1.

Fig. 1

Flow chart depicting the tools and techniques employed for the study of AβPP ubiquitination.

RESULTS

3D-model of AβPP

The three-dimensional structure of AβPP is predicted to analyze the potential lysine residues important for their ubiquitination, since its full experimental structure was lacking. The structural topology of the modeled AβPP has identified an N-terminal signal peptide MLPGLALLLLAAWTARALEVPT (1–22), an extracellular region (1–698), a trans-membrane helix S1 (699–721), and a cytoplasmic region (722–770). Further, domain analysis of full length AβPP (770aa) revealed six functional domains, namely: 1) N-terminal domain, 2) Copper binding domain, 3) Protease inhibitor domain, 4) E2 domain, 5) Aβ peptide domain, and 6) C-terminal domain. The combined results suggest that the first four domains parses in the extracellular region and the Aβ peptide in the trans-membrane region while the C-terminal domain in the cytoplasmic region. Moreover, the modeled structure was spanned by 22 alpha-helices (37%), 7 beta-strands (4%), and one trans-membrane alpha-helix (3%). In addition, the tertiary structure of the AβPP model is enriched with two disulfide linkages one at cysteine144-cystein174 and another at cysteine158-cysteine186 that are very important for their stability. Further, the structural validation of the refined model by PROVE analysis passed their structural quality test with less than 1% buried outlier atoms. Similarly, the sequence information based tertiary structure, i.e., 3D–1D score ≥0.2 was found to be more than 50% for our refined model as per the verify-3D tool. Moreover, the geometrical conformation of the refined model analyzed by RAMPAGE server was found to be greater than 99.5% for the favored and allowed region residues except four amino acids, which included Alanine 35, Aspartic acid 360, Proline 365, and Arginine 653. The refined model of AβPP along with their molecular descriptors and structural refinement simulations is shown in Fig. 2. Further, the structure validation scores of modeled and refined models obtained by verify 3D and RAMPAGE can be accessed from Supplementary Table 1.

Fig. 2

AβPP Modeling and Structural Refinement. A) AβPP Model: The modeled AβPP (shown in cartoon and surface view) represented an N-terminal signal peptide (Cyan in color), Trans-membrane helix/domain (Hot pink in color), and a C-terminal region (Purple in color) along with the disulfide bonds (Green/Yellow in color). The domain analysis identified six domains including N-terminal, copper binding, protease inhibitor, E2, Aβ, and C-terminal domain spanned across the full length 770aa protein. The structural topology of the predicted model has identified a membrane spanning trans-membrane helix S1 at 699 to 721 amino acid residues with an extracellular N-terminal (1–699) carrying signal peptide (1–22) and a cytoplasmic C-terminal (721–770). The structural analysis of the predicted model revealed 22 alpha-helices, 7 Beta-sheets, and 1 Trans-membrane helix and 2 disulfide linkages (Cys144–Cys174 and Cys158–Cys186). Moreover, the solvent accessible surface area (30775.238 angstrom∧2) was less than the molecular surface area (80013.477 angstrom∧2) of the model due to inaccessibility of the solvent in trans-membrane region. B) Structural Refinement of AβPP: The modeled AβPP was refined with structural refinement simulations up to 72 cycles (designated P1,P2, … P72) and validated for the refinements by different structural validation programs-PROVE, ERRAT, Verify-3D, and RAMPAGE. C) AβPP Ramachandran Plot: Evaluated 99.5% residues in geometrical (Φ–ψ) favored (719 residues (93.6%)) + allowed residues (45 (5.9%)) regions while only four residues Ala35, Asp360, Pro365, and Arg653 ∼0.5% in outlier region; other structural valition programs including PROVE and ERRAT identified to PASS the modeled structure for their overall quality. D) Verify-3D: The 3D atomic model compatibility assessment revealed the refinement of modeled AβPP from zero cycle (P)– 33.12% to 72 cycle (P72)– 43.9% residues, with 50.13% residues of the best obtained model P30 at 30th cycle showing 2D–3D structural compatibility. E) RAMPAGE: The geometrical validation of C-alpha neighboring residues revealed the refinement from zero cycle (P)– 655 (favored + allowed residues) and 113 (outlier residues) to 72 cycle (P72)– 764 (favored + allowed residues) and 4 (outlier residues). The best geometrical configuration with minimum simulations was achieved at 30th cycle (P30) signifying it as a best-predicted model.

Fig. 3

Potential lysine prediction for ubiquitination in AβPP based on: A) Sequence Identity: The sequence identity scores among 21 window size polypeptides carrying central lysine (K6-Blue, K11-Pink, K27-Brown, K29-Green, K33-Yellow, K48-Red, K63-Purple) in ubiquitin and (K51, K60, K66, K99, K103, K106, K132, K134, K155, K161, K178, K224, K315, K351, K363, K377, K393, K395, K401, K421, K425, K428, K429, K438, K495, K496, K503, K510, K521, K522, K568, K601, K662, K670, K687, K699, K724, K725, K726, K751, K763) in AβPP are plotted on a bar graph. The identity scores greater than 0.2 represented the conservation of more than two amino acid residues in the aligned sequences signifying the presence of potential lysine for ubiquitination based on sequence identity (shown in Red)- K60, K66, K99, K106, K132, K134, K155, K161, K351, K363, K377, K393, K395, K401, K503, K510, K522, K568, K601, K662, K670, K687, K699, K724, K725, K726. B) Sequence Similarity: The lysine neighboring amino acid conservation analysis among 21 window size polypeptides carrying central lysine have identified the conservation of Asp(D), Glu(E), Ile(I), Leu(L), and Val(V) at 57% similarity threshold and Asp(D), Gln(Q), Glu(E), Ile(I), Leu(L), Met(M), Thr(T), and Val(V) at 42% similarity threshold in ubiquitin while the conservation of Gln(Q) and Glu(E) at 34% similarity threshold; Gln(Q), Glu(E), Ile(I), Leu(L), Met(M), and Val(V) at 31% similarity threshold and Asn(N), Gln(Q), Glu(E), Ile(I), Leu(L), Met(M), and Val(V) at 29% similarity threshold in AβPP. The amino acid residues conserved at the common sites of both ubiquitin and AβPP are enclosed by red rounded rectangles signifying the critical amino acids for ubiquitination including Glu (E), Gln (Q), Val (V), Leu (L), Met (M), and Ile (I). Further, on the basis of best conserved amino acids, i.e., glutamic acid and glutamine common to both ubiquitin and AβPP (marked by yellow arrow), the potential lysine for ubiquitination are identified and encircled in pink rectangles including K224, K351, K377, K393, K401, K503, K510, K522, K662, K724, K751, and K763.

K351 is the most promising target for AβPP ubiquitination

The potential lysine sites which are favorable for the ubiquitination of AβPP are identified by four different methodologies, including sequence identity, sequence similarity, protein-peptide docking, and machine learning techniques. Since, ubiquitin is the pre-eminent protein that is ubiquitinated the most in any cellular processes due to the poly-ubiquitination phenomenon; therefore, its lysine site features are of great importance to unravel the ubiquitination mystery. These informative sites of ubiquitin are utilized by above mentioned methodologies to deduce the potential lysine sites in AβPP.

Sequence identity based potential lysine in AβPP

The sequence identity between the 21 window-size peptide sequence with central lysine at seven lysine sites of ubiquitin and forty-one lysine sites of AβPP is determined to infer the promising ubiquitination sites in AβPP. The sequence-identity of at least three amino acids, i.e., identity score >0.2 was taken as the threshold, which identified 26 key lysine sites in AβPP including K60, K66, K99, K106, K132, K134, K155, K161, K351, K363, K377, K393, K395, K401, K503, K510, K522, K568, K601, K662, K670, K687, K699, K724, K725, and K726 as favorable sites for the ubiquitination (Fig. 3A). Among them these seven K60, K66, K351, K363, K601, K662, and K687 lysine sites displayed the higher identity scores. For detailed sequence identity scores obtained for each pairwise sequence alignment, Supplementary Table 2 can be explored.

Sequence similarity based potential lysine in AβPP

The sequence conservation analysis among lysine neighboring residues in ubiquitin revealed the conservation of hydrophobic leucine (L), isoleucine (I), valine (V) and polar negatively charged aspartic acid (D), glutamic acid (E) at 57% similarity threshold. Further, the conservation in ubiquitin at 42% similarity threshold provided the majority of informative amino acid residues in common with AβPP that are crucial for the ubiquitination process. These conserved residues included the hydrophobic methionine (M), leucine (L), isoleucine (I), and valine (V), polar negatively charged glutamic acid (E), and polar uncharged glutamine (Q). Moreover, in AβPP only glutamic acid (E) and glutamine (Q) is conserved at the higher threshold of 34% sequence similarity, which depicted K224, K351, K377, K393, K401, K503, K510, K522, K662, K724, K751, and K763 as potential lysine sites (Fig. 3B).

Ubiquitin-AβPP lysine site peptide docking based potential lysine in AβPP

The affinity of lysine specific ubiquitin-ubiquitin interaction in the poly-ubiquitination process is taken as the principal to investigate the potential lysine sites in AβPP important for ubiquitination. The flexible docking of ubiquitin with the best conformations of ubiquitin’s seven lysine peptides, K6, K11, K27, K29, K33, K48, and K63, resulted in the average root mean square deviation (Avg. RMSD) of 2.92, 5.31, 4.92, 3.88, 4.49, 1.96, and 4.63 respectively. Further, the obtained docking results of ubiquitin with ubiquitin and AβPP lysine peptides are summarized in Table 2. Interestingly, the Avg. RMSD of K48 was minimal, i.e., 1.96 showing the best affinity for K48 linked poly-ubiquitination than the K11 site, which had highest Avg. RMSD of 5.31 implying the least propensity of K11 linked poly-ubiquitination. Here, the minimal Avg. RMSD, i.e., best binding affinity than the K11 (5.31) was taken as a threshold to find the best propensity ubiquitination sites in AβPP. The 21 potential sites were identified with good propensities for ubiquitination, including K51, K60, K99, K132, K161, K178, K351, K393, K401, K425, K495, K496, K503, K510, K521, K522, K568, K687, K724, K725, and K726 (Fig. 4A). Among them, the lowest Avg. RMSD of 0.906538 was obtained for the K687 site in AβPP or corresponding K16 site in Aβ showing best potential for ubiquitination.

Table 2

Ubiquitin and AβPP lysine peptides docking scores with ubiquitin protein using CABS-dock

Ubiquitin
S.No.	Ubiquitin Lysine Site	21 residues key lysine sequence	Cluster Density	Average RMSD	Max RMSD	No of elements
1	K6	MQIFVKTLTGKTITLE	26.338	2.92353	19.4007	77
2	K11	MQIFVKTLTGKTITLEVEPSD	23.1489	5.31343	23.9976	123
3	K27	VEPSDTIENVKAKIQDKEGIP	27.6251	4.92306	25.651	136
4	K29	PSDTIENVKAKIQDKEGIPPD	38.8683	3.88491	23.829	151
5	K33	IENVKAKIQDKEGIPPDQQRL	27.5913	4.49417	17.6504	124
6	K48	PDQQRLIFAGKQLEDGRTLSD	24.0229	1.95647	7.09756	47
7	K63	GRTLSDYNIQKESTLHLVLRL	25.5963	4.64911	26.4507	119
AβPP
S.No.	AβPP Lysine Site	21 residues key lysine sequence	Cluster Density	Average RMSD	Max RMSD	No of elements
1	K51	LNMHMNVQNGKWDSDPSGTKT	21.5806	5.09718	27.0102	110
2	K60	GKWDSDPSGTKTCIDTKEGIL	26.0199	4.3044	24.3425	112
3	K66	PSGTKTCIDTKEGILQYCQEV	19.0914	9.00931	24.3529	172
4	K99	NQPVTIQNWCKRGRKQCKTHP	41.1493	3.57236	14.4164	147
5	K103	TIQNWCKRGRKQCKTHPHFVI	24.9614	6.08941	22.8975	152
6	K106	NWCKRGRKQCKTHPHFVIPYR	19.9487	9.17352	24.6746	183
7	K132	FVSDALLVPDKCKFLHQERMD	34.6382	3.0602	18.6577	106
8	K134	SDALLVPDKCKFLHQERMDVE	17.7294	7.38886	30.4831	131
9	K155	ETHLHWHTVAKETCSEKSTNL	18.8584	9.22667	27.4895	174
10	K161	HTVAKETCSEKSTNLHDYGML	25.5239	3.44775	14.9316	88
11	K178	YGMLLPCGIDKFRGVEFVCCP	31.9959	4.53183	19.5177	145
12	K224	DTDYADGSEDKVVEVAEEEEV	21.3651	7.58246	19.827	162
13	K315	SRWYFDVTEGKCAPFFYGGCG	28.418	5.38391	18.2749	153
14	K351	CGSAMSQSLLKTTQEPLARDP	26.6421	4.42907	21.7612	118
15	K363	CGSAMSQSLLKTTQEPLARDP	22.2974	9.59752	30.1062	214
16	K377	TAASTPDAVDKYLETPGDENE	30.3162	5.50861	30.4716	167
17	K393	GDENEHAHFQKAKERLEAKHR	34.6566	4.09735	12.2452	142
18	K395	ENEHAHFQKAKERLEAKHRER	26.0746	6.36634	25.9803	166
19	K401	FQKAKERLEAKHRERMSQVMR	23.396	4.2315	18.9197	99
20	K421	REWEEAERQAKNLPKADKKAV	21.1402	10.4067	31.0374	220
21	K425	EAERQAKNLPKADKKAVIQHF	32.8454	3.83616	27.7128	126
22	K428	RQAKNLPKADKKAVIQHFQEK	24.0608	5.94329	22.5957	143
23	K429	QAKNLPKADKKAVIQHFQEKV	24.0077	6.70617	28.9284	161
24	K438	KKAVIQHFQEKVESLEQEAAN	37.1462	6.38019	28.0285	237
25	K495	PRPRHVFNMLKKYVRAEQKDR	22.7752	4.1712	14.4324	95
26	K496	RPRHVFNMLKKYVRAEQKDRQ	39.035	4.73934	17.5878	185
27	K503	MLKKYVRAEQKDRQHTLKHFE	26.2182	4.61512	15.6449	121
28	K510	AEQKDRQHTLKHFEHVRMVDP	26.9482	5.19515	20.7974	140
29	K521	HFEHVRMVDPKKAAQIRSQVM	20.0593	5.23448	19.223	105
30	K522	FEHVRMVDPKKAAQIRSQVM	29.0115	4.10182	18.1209	119
31	K568	IQDEVDELLQKEQNYSDDVLA	36.6961	2.9431	28.6506	108
32	K601	DALMPSLTETKTTVELLPVNG	19.4168	11.5879	26.6436	225
33	K662	TRPGSGLTNIKTEEISEVKMD	22.9797	8.39872	29.6782	193
34	K670	NIKTEEISEVKMDAEFRHDSG	27.64	8.53834	32.0268	236
35	K687	HDSGYEVHHQKLVFFAEDVGS	40.8146	0.906538	1.81447	37
36	K699	VFFAEDVGSNKGAIIGLMVGG	19.2171	9.62683	22.5839	185
37	K724	TVIVITLVMLKKKQYTSIHHG	20.3169	5.1189	28.5382	104
38	K725	VIVITLVMLKKKQYTSIHHGV	29.4156	2.00574	19.6385	59
39	K726	IVITLVMLKKKQYTSIHHGVV	31.7439	4.75682	19.5129	151
40	K751	AVTPEERHLSKMQQNGYENPT	21.713	11.0533	30.6823	240
41	K763	QQNGYENPTYKFFEQMQN	23.4377	5.54662	28.2897	130

Machine learning ubiquitination tools based potential lysine in AβPP

The different machine learning approaches, including maximal dependence decomposition, Bayesian discriminant analysis, random forest models, and support vector machine classifier-based tools (UbiSite, BDM-PUB, CKSAAP, UbPred and UbiPred) have been employed to determine the potential lysine sites in AβPP for ubiquitination on the basis of experimentally verified ubiquitination site datasets. The predicted ubiquitination sites and scores obtained by UbiSite, BDM-PUB, CKSAAP, UbPred, and UbiPred tools are summarized in Table 3. Moreover, the stacked bar graph of the ubiquitination prediction scores versus AβPP lysine sites are plotted to identify the potential lysine in AβPP as shown in Fig. 4B. The best ubiquitination aggregate scores were obtained for K351 and K377, i.e., 3.19 and 2.47, respectively, with maximal predictions by four tools: UbiSite, BDM-PUB, UbPred, and UbiPred.

Table 3

Ubiquitination Prediction Scores of Ubiquitination-site Prediction Tools

Further, these potential lysine residues were classified into different confidence levels: very high, high, medium, low, and very low based on the evidence from ubiquitination prediction tools and their verification by other methods, including sequence identity, sequence similarity, and flexible protein-peptide docking. For instance, the potential ubiquitination sites predicted by at least three ubiquitination prediction tools and is verified by at least any two other methods are classified as “Very High” confidence, while those verified by any one of the other methods are classified as “High” confidence. In a similar way, the potential ubiquitination sites predicted by any two ubiquitination prediction tools and are verified by at least any one of the other methods are categorized into “Medium” confidence. Additionally, the potential ubiquitination sites predicted by any one of the ubiquitination prediction tools, which is verified by at least, any two of the other methods, are assigned as “Low” confidence whereas if verified by any one of the other methods then classified as “Very Low” confidence. The key lysine residues important for the ubiquitination of AβPP with predicted confidence are listed in Table 4. Moreover, the Venn diagrams comprehensively reviewed the predicted ubiquitination sites and illustrated the logical relations among the key ubiquitination sites in AβPP (Fig. 4C). The Venn diagram analysis clearly depicted K351 as the most promising ubiquitination site at high confidence; K393, K401, K510, and K522 at medium confidence; and K503 and K724 at low confidence.

Table 4

Predicted Confidence of the Key Ubiquitination sites in AβPP

Fig. 4

Potential lysine prediction in AβPP and their comparative analysis: A) Ubiquitin and K-AβPP site Peptide Docking: The average root mean square deviation (Avg. RMSD) graph of the docked 21 window size K-site AβPP peptide with ubiquitin is dot plotted and is compared with the Avg. RMSD of the docked 21 window size K-site Ubiquitin peptide with the ubiquitin protein. The Avg. RMSD of K-11 ubiquitin, i.e., 5.31, is taken as threshold to identify potential ubiquitination site in AβPP (K51, K60, K99, K132, K161, K178, K351, K393, K401, K425, K495, K496, K503, K510, K521, K522, K568, K687, K724, K725, and K726) with lower Avg. RMSD values. B) Ubiquitination Prediction Tools: The machine learning based ubiquitination prediction tools-UbiPred (Light Blue), UbPred (Purple), CKSAAP (Green), BDM-PUB (Dark Red), and UbiSite (Blue) has predicted potential ubiquitination sites in AβPP including K51, K60, K103, K134, K155, K161, K224, K351, K363, K377, K393, K395, K401, K421, K425, K428, K429, K438, K495, K496, K503, K510, K521, K522, K568, K601, K662, K670, K699, K724, K751, and K763. C) Comparative Analysis of Ubiquitination Sites: The comparison of the potential ubiquitination sites predicted by all the methods have revealed the most potential ubiquitination sites which is predicted by at least three ubiquitination prediction tools and verified by at least any two other methods are K351 and K377 (marked Red) with very high confidence, while verified by any one of the other methods are K224, K363, K601, K751, and K763 (marked Red) with high confidence. Similarly, the potential ubiquitination sites predicted by any two ubiquitination prediction tools and verified by at least any one of the other methods are K60, K161, K393, K401, K496, K510, K522, and K662 (marked Red) with medium confidence. Likewise, the potential ubiquitination sites predicted by any one of the ubiquitination prediction tool and is verified by at least any two of the other methods are K503, K568, and K724 (marked Red) with low confidence while verified by any one of the other methods are K51, K134, K155, K395, K425, K495, K521, K670, and K699 (marked Red) with very low confidence.

Lysine residues are crucial for AβPP stability, ubiquitination, and other functions

The mutational studies on AβPP revealed the importance of lysine residues in AβPP processing and ubiquitination. When we mutated the key lysine (Polar; positively charged; basic amino acid; pK = 10.5) with other polar; positively charged; basic amino acids-Arginine (pK = 12.5), Histidine (pK = 6.0) and polar; negatively charged; acidic amino acids-Aspartate (pK = 3.9) and Glutamate (pK = 4.2), we observed some interesting effects on the internal potential energy of mutated AβPP summarized in Table 5. We found that only arginine had imparted site-specific stability to even more than half of the predicted ubiquitination sites, including K51, K60, K134, K161, K224, K393, K401, K425, K496, K510, K521, K522, K699, and K724, and instability at the rest site of AβPP (Fig. 5A). Moreover, the lysine residues at high confidence ubiquitination sites were found intolerant towards all mutations against the stability of AβPP. Apart from arginine, histidine was observed to greatly impact the stability of AβPP followed by glutamate and aspartate. However, the effect of glutamate on AβPP stability was less than the arginine and histidine, but it affected nearly all the ubiquitination sites (Fig. 5A). Further, the detailed internal potential energies obtained for the modeled and mutated AβPP, and the mutational impact of lysine mutations on their total potential energies can be inferred from Supplementary Tables 3 and 4, respectively.

Table 5

Effect of Lysine Mutation on Total Potential Energy of AβPP

Confidence	Mutational	ΔE Total (KJ/mol) (E_Mutation– E_NoMutation)
Level	Site	Mutation Type
		Lys(K)– Arg(R)	Lys(K)– His(H)	Lys(K)– Asp(D)	Lys(K)– Glu(E)
Very High	K351	5876.014	222.42	4.633	61.496
	K377	1194.364	62.217	15.471	14.557
High	K224	– 238.418	35.037	17.395	20.996
	K363	1026.26	9837.761	33.565	13848.825
	K601	2083620.973	16503.902	11.123	8.77
	K751	2473.202	165.922	– 0.629	38950.161
	K763	4890663.473	6062.487	6.621	156472.582
Medium	K60	– 236.179	106.526	48.961	34.491
	K161	– 184.562	7.19899E+11	27.44	12.717
	K393	– 234.58	74.135	23.825	20.399
	K401	– 194.841	86.977	52.08	62.194
	K496	– 239.636	59.368	20.416	12.114
	K510	– 240.533	616957.348	21.922	17.953
	K522	– 231.222	90.121	34.071	29.496
	K662	765427662.5	68410	226.784	601.836
Low	K503	487.243	70.155	13.244	173.287
	K568	1.21228E + 11	51.993	23.744	23.338
	K724	– 245.779	23655.275	35.864	19.457
Very Low	K51	– 228.095	0	14.649	26.098
	K134	– 131.127	17.27	7.151	9.86
	K155	1569.943	31.192	10.395	143.977
	K395	17638.653	237.323	36.303	11.287
	K425	– 231.168	64.469	24.192	29.85
	K495	331718094.5	153.008	309.342	7639277.973
	K521	– 232.041	49.873	27.586	24.532
	K670	2799.446	139.305	39005702.47	1061.897
	K699	– 235.707	4256364558	23.08	24.27

Further, the disease susceptibility of lysine mutations were checked through the mutational analysis tools, PANTHER, SNAP2, Polyphen2, PMut, Phd-SNP, and SIFT. It revealed that mutations had an effect on all the sites, but they had very least effect on the high confidence ubiquitination sites in comparison with the medium and low confidence sites. The highly intolerant mutations that were most susceptible to the diseases are shown in Fig. 5B. In addition, the detailed results for the mutation associated disease susceptibility predicted by PANTHER, SNAP2, Polyphen2, PMut, PhD-SNP, and SIFT can be accessed through Supplementary Table 5. Further, we analyzed the physico-chemical and amino acid compositional analysis of ubiquitin, Aβ₄₂ peptide, and AβPP to understand the crucial factors that determine the ubiquitination of a protein. Interestingly we found that total number of positively and negatively charged residues were in same proportion in ubiquitin along with the aliphatic index-100 and GRAVY-– 0.489 which could be the decisive factor for ubiquitin to be the top ubiquitination protein. Moreover, some amino acid compositions, including arginine, asparagine, aspartate, glutamate, phenyl-alanine, proline, and serine were found in the comparable range±1% in ubiquitin, Aβ₄₂ peptide, and AβPP (Fig. 5C). Further, lysine sites in AβPP were investigated for their role in the other cellular processes and functions by MutPred, which has been outlined in Table 6. The collective results depicted that apart from ubiquitination; lysine residues are also crucial for protein stability, modifications, and other functions.

Table 6

Importance of Lysine residues in AβPP Ubiquitination, Processing and other Functions

AβPP Mutational Site		Affected molecular mechanisms		Affected	Pathogenic Score
(Ubiquitination Confidence)		(p-values≤0.05)		Motifs
K351 (Very High)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of O-linked glycosylation (T353)	Loss of Proteolytic cleavage (Q347)	GSK3, CK2, PIKK	0.647 0.658
			Gain of Proteolytic cleavage (Q347)	Phosphorylation site and FHA Phosphopeptide ligand	0.629
K377 (Very High)	Lys(K)– Asp(D)	Gain of Phosphorylation and Sulfation (Y378); Loss of Acetylation, SUMOylation, and Ubiquitylation (K377); Gain of Proteolytic cleavage (D376); Altered Coiled coil	_		0.517
K224 (High)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of SUMOylation and Ubiquitylation (K224); Loss of Proteolytic cleavage (D219); Altered Coiled coil and Trans membrane protein	_		0.7
				CK2 Phosphorylation site, SUMO Interaction site	0.755 0.722
K601 (High)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of SUMOylation (K601); Altered Trans membrane protein	NEK2 Phosphorylation site and FHA Phosphopeptide ligand	_	0.611
				SUMO Interaction site	0.601 0.567
K751 (High)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of Allosteric site (R747); Altered Metal binding; Altered Trans membrane protein	Altered Ordered and Disordered interface	_	0.737
				_	0.713
					0.705
K763 (High)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of N-terminal acetylation (M768); Altered Disordered interface	Tyrosine based sorting signal, Phosphotyrosine binding ligand, PKC Phosphorylation site, Amyloidogenic glycoprotein intracellular domain		0.601 0.829 0.839 0.832
		Loss of Methylation (K763); Gain of N-terminal acetylation (M768)	Altered Metal binding; Altered Ordered and Disordered interface Gain of Sulfation (Y762)
K60 (Medium)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of Relative solvent accessibility; Altered Trans membrane protein; Loss of Disulfide linkage (C62)	Gain of ADP-ribosylation (K60);	GSK3, CK1 Phosphorylation site, FHA Phosphopeptide ligand, N-myristoylation site	0.509 0.812 0.797 0.79
_
	Altered Metal binding; Loss of Loop	CK2 Phosphorylation site
Loss of Relative solvent accessibility; Altered Trans membrane protein; Loss of Loop
K161 (Medium)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Gain of Loop; Loss of Disulfide linkage (C158); Altered Trans membrane protein	Altered Metal binding; Loss of Relative solvent accessibility	GSK3, CK1, PKC Phosphorylation site, FHA Phosphopeptide ligand	0.619 0.635 0.629
_
K 393 (Medium)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Coiled coil; Loss of Acetylation (K393); Loss of Helix	_	_	0.783
					0.782
			Gain of SUMOylation (K395)		0.737
K401 (Medium)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of SUMOylation and Acetylation (K401); Altered Disordered interface	_	_	0.525
Loss of Helix	0.798
					0.774
					0.782
K496 (Medium)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of Acetylation (K496); Altered Disordered interface	_	0.613
0.544
0.516
K510 (Medium)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Metal binding; Altered DNA binding; Loss of Allosteric site (R505); Loss of Acetylation (K510); Altered Disordered interface	_	PKC Phosphorylation site	0.524
Loss of Helix	0.814
0.801
_	0.806
K522 (Medium)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of Helix; Altered Disordered interface; Altered Coiled coil	_		0.773
					0.745
					0.754
K662 (Medium)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Gain of ADP-ribosylation (K662)	Loss of SUMOylation, Acetylation, Ubiquitylation (K662); Altered Trans membrane protein	GSK3, CK1 Phosphorylation site, SUMOylation site, FHA Phosphopeptide ligand	0.541 0.808 0.813 0.803
		Loss of Phosphorylation (S667)
		Gain of Phosphorylation (S667)			CK2 Phosphorylation site
		Loss of Phosphorylation (S667)
K503 (Low)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Disordered interface; Loss of Acetylation (K503); Altered DNA binding; Loss of Allosteric site (R505)	_	PKB Phosphorylation site	0.571 0.805 0.815 0.786
Altered Metal binding; Loss of Helix	LATS Kinase Phosphorylation site
_
		Loss of Helix	TRAF2 Binding site
K568 (Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Coiled coil; Loss of Helix; Loss of Phosphorylation (Y572); Altered Trans membrane protein; Loss of N-linked glycosylation (N571)	Loss of Sulfation (Y572)	_	0.608
Gain of Sulfation (Y572)	0.629
0.621
K724 (Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Trans membrane protein; Altered DNA binding; Altered Metal binding; Gain of Pyrrolidone carboxylic acid (Q727)	Altered Disordered interface	LATS Kinase Phosphorylation site	0.641
_	Clathrin box, SUMO Interaction site	0.648
0.638
K51 (Very Low)	Lys(K)– Arg(R) Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Trans membrane protein; Loss of Relative solvent accessibility; Loss of Ubiquitylation (K51)	_	Integrin Binding site	0.511
			Altered Ordered interface	_	0.773
			Altered Ordered interface; Altered Metal binding		0.778 0.736
_
K134 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Metal binding; Altered Trans membrane protein	Loss of Relative solvent accessibility	_	0.744 0.769 0.726
_
			Gain of SUMOylation (K132)
K155 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Metal binding; Altered Trans membrane protein	Loss of Helix; Loss of Disulfide linkage (C158)	CK2 Phosphorylation site, FHA Phosphopeptide ligand	0.815 0.801 0.811
			Gain of Disulfide linkage (C158)
			Loss of Disulfide linkage (C158)
K395 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of Helix; Loss of Acetylation(K393); Altered Disordered interface; Loss of SUMOylation(K395)	_	0.751 0.709 0.73
		Altered Coiled coil; Loss of Helix; Loss of Acetylation and SUMOylation (K393); Altered Disordered interface
K425 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D)	Altered Coiled coil; Loss of Helix; Altered Disordered interface; Loss of Acetylation (K425); Gain of Ubiquitylation (K428)	_			0.506 0.527

Can lysine mediated non-covalent interactions prevail ubiquitin positioning and ubiquitin-AβPP conjugation during ubiquitination?

The non-covalent interactions of ubiquitin with different proteins, such as human DNA repair proteins, insulin-degrading enzyme, and SUMO protein, are well known for modulating their activity or correctly orienting ubiquitin for lysine specific conjugation [55 –57]. These interactions can also impart stability to the ubiquitin-protein conjugates in the ubiquitination process [58]. In this regard, we predicted the interaction among all the domains of AβPP and ubiquitin and examined them for the lysine mediated polar interactions among their interacting residues. The ubiquitin-AβPP interaction predictions reported that ubiquitin has nearly same interactive affinity toward all the domains of AβPP except Aβ peptide. The prediction of ubiquitin-AβPP interactions was highest on the basis of homologous protein interactional network (S_Net), rather than by the statistical domain-domain interactions (S_Dom) and sequence similarity based interacting protein pairs (S_Seq) shown in Fig. 6A. Further, lysine investigation in non-covalent electrostatic, polar interactions revealed the presence of lysine interactions with N-terminal domain (K66, K99), Copper binding domain (K155), E2 domain (K447), Aβ peptide (K28), and C-terminal domain of AβPP (K687). Apart from AβPP, ubiquitin’s lysine residues, including K6, K11, K33, K48, and K63 were also involved in the polar interactions (Fig. 6B). The presence of lysine as interacting residues provided a clue for their role in ubiquitin positioning or ubiquitin-AβPP conjugation but further researches are required for clear understanding. These informative residues are the foundations for future avenues of AβPP ubiquitination mechanistic research.

Interaction network of ubiquitination proteins for AβPP clearance

Fig. 5

Mutational Analysis of Lysine residues for AβPP Ubiquitination. A) Impact of Lysine Mutation on AβPP Stability: The total energy change in AβPP upon Lysine mutations have identified the prominent effect of Lysine-Arginine and Lysine-Histidine mutations as the most detrimental to AβPP stability in comparison with Glutamic and Aspartic acid. Here, nearly 14 Lysine-Arginine mutations including K51, K60, K134, K161, K224, K393, K401, K425, K496, K510, K521, K522, K699, and K724 are reported to increase the stability of AβPP by reducing their total energy in KJ/mol. B) Lysine Mutation and Disease Susceptibility: The investigation by various mutation analysis tools including PANTHER, SNAP2, Polyphen2, Pmut, PhD-SNP, and SIFT have identified the key lysine sites where mutations are susceptible to disease pathogenesis and are marked with red color taking the threshold of predictions by more than three tools for best accuracy. The highly disease susceptible mutations include K51H/D/E, K60H/D/E, K134H/D/E, K155H/D/E, K161D, K393H/D/E, K395R/H/D/E, K401H/D/E, K425H/D/E, K495H/D/E, K496H/D/E, K503R/H/D/E, K510D/E, K521R/H/D, K522R/H/D/E, K662H, K670D, K699H/D/E, K724H/D/E, K751H/D, and K763H/D/E. C) Physico-chemical Properties and Amino acid Composition: The physico-chemical properties of Ubiquitin showed equal no of positive and negative charged residues, with low hydrophobicity and high stability, while Aβ₄₂ showed high hydrophobicity and stability in comparison to AβPP low hydrophobicity and instability. Amino acid compositional analysis revealed the comparable percentage within 1% range of Ubiquitin with Aβ₄₂ and AβPP including Arginine, Asparagine, Aspartic acid, Glutamic acid, Phenyl alanine, Proline, and Serine.

The accumulation of well-known pathogenic Aβ can be regulated by the clearance of its precursor, AβPP, through ubiquitination. However, ubiquitination is a well-defined process, but the association of different ubiquitination E1s, E2s, E3s, and deubiquitinating enzymes (DUBs) always remained an interesting field of research for target specific protein clearance. Here, we investigated the interaction network of E1s, E2s, E3s, and DUBs enzymes for the clearance of AβPP in AD biology. The interaction network identified ubiquitin E3 ligases, Park2 and STUB1, to be associated with AβPP ubiquitination while deubiquitination was by USP25 and UCHL1 (Fig. 7A). The ubiquitination processing of other enzymes that govern the synthesis of Aβ peptide are also crucial to restrict them. Here, certain E3 ligases like FBXO2 is reported to regulate the ubiquitination of BACE1; TRIM13 to regulate PSENEN; CDH1 to regulate PSEN1 and NCSTN; while TRIM55 and SART1 to regulate the ubiquitination of NCSTN. Likewise, specific deubiquitinase enzymes, including USP25 and UCHL, are found to regulate AβPP; USP8 to BACE1; and USP39 to NCSTN, respectively (Fig. 7A). Further, examinations on the potential E1 activating and E2 conjugating enzymes were done to identify the pre-processors of AβPP ubiquitin ligation by Park2 and STUB1 E3 ligases. We reported an array of ubiquitin E1 activating enzymes UBA1, UBA6, UBA7, and ATG7 that can possibly activate specific ubiquitin E2 conjugating enzymes including UBE2A, UBE2B, UBE2G1, UBE2J1, UBE2K, UBE2L3, UBE2L6, UBE2N, UBE2Q1, UBE2S, UBE2U, and UBE2Z to conjugate lysine to Park2 (Fig. 7B); while additional UBE2I, UBE2T, and UBE2W enzymes to conjugate lysine to STUB1 (Fig. 7C). The PPI network provided us the substantial information about the ubiquitination enzymes associated with the clearance of AβPP in the AD biology that can be further explored for the therapeutic avenues.

DISCUSSION

The structural topology of the modeled AβPP has identified 1) an N-terminal signal peptide of 22AA residues (MLPGLALLLLAAWTARALEVPT) and a trans-membrane segment S1 (699–721). Here, positively charged arginine (Arg16) and the N-terminal hydrophobic amino acids are crucial for AβPP’s post-translational translocation across the ER membrane. Since, the removal of positively charged amino acid from signal peptide selectively impairs the translocation, while the presence of hydrophobic residues favors it through lipid bilayer of ER membrane toward the secretory pathway [59]. 2) The domain analysis revealed six domains that are contributing toward the functionality of AβPP. Among them four domains, including N-terminal, copper binding, protease inhibitor (Kunitz_BPTI), and E2 domain spans over the extracellular surface region and are responsible for copper binding, protease inhibition, platelet aggregation, and hemostasis. While the proteolytic product of cytoplasmic c-terminal domain acts as a transcriptional regulator in neurons. In addition, the cleavage of their trans-membrane domain is responsible for Aβ peptide formation [60]. 3) The secondary structure analysis identified a higher alpha-helical content ∼37% in AβPP indicating their potent ability to tolerate mutations, since helices can accrue more mutations than beta strands without structural distortion due to their higher numbers of inter-residue contacts [61]. 4) Their tertiary structure revealed two disulfide linkages (cysteine144-cystein174 and cysteine158-cysteine186) that imparted stability to the AβPP. Further, the potential ubiquitination sites in AβPP has identified by adopting the sequence identity, sequence similarity and protein-peptide docking approach using lysine site features of ubiquitin, for being the eminent protein for poly-ubiquitination [62, 63]. The comparative analysis of thus predicted ubiquitination sites with the potential sites identified by the machine learning approaches revealed K351 as most promising ubiquitination site at high confidence; K393, K401, K510, and K522 at medium confidence; and K503 and K724 at low confidence. Additionally, the site specific conserved amino acids, glutamic acid (E), glutamine (Q), valine (V), leucine (L), isoleucine (I), and methionine (M), were identified neighboring the lysine residue in AβPP. These conserved polar residues Glu(E) and Gln(Q) at ubiquitination site would display their propensities for salt-bridges while conservation of hydrophobic residues Val(V), Leu(L), Ile(I), and Met(M) indicate their propensities for buried contacts signifying their role in the ubiquitination reaction [64]. Interestingly, we can comprehend the presence of polar negatively charged Glu(E) amino acids neighboring lysine for their electrostatic interactions among each other in their native folded state [65] while ubiquitination directs lysine’s covalent attachment with c-terminal glycine of ubiquitin molecule that opened the avenues for further validations.

Table 8

K 495 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Disordered interface; Loss of Acetylation (K495); Altered Coiled coil		_	0.779 0.781 0.731
Altered Disordered interface; Gain of Acetylation (K496)	_
Altered Coiled coil
K521 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Loss of Helix; Altered Coiled coil	_	_	0.729
Altered Disordered interface	0.72
0.731
K670 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Trans membrane protein; Loss of Acetylation (K670)	Loss of Phosphorylation (S667)	_	0.655 0.642 0.634
Gain of Phosphorylation (S667)	CK2 Phosphorylation site
K699 (Very Low)	Lys(K)– His(H) Lys(K)– Asp(D) Lys(K)– Glu(E)	Altered Trans membrane protein; Loss of Ubiquitylation (K699);	Loss of GPI-anchor amidation (N698)	PKC Phosphorylation site, N-myristoylation site	0.621 0.661 0.641
Gain of GPI-anchor amidation (N698); Altered Metal binding
Gain of GPI-anchor amidation (N698)

GSK3, Glycogen synthase kinase-3; CK1, Casein kinase-1; CK2, Casein kinase-2; PIKK, Phosphatidylinositol 3-kinase-related kinase; FHA, Forkhead-associated domain; SUMO, Small Ubiquitin-like Modifier; NEK2, NIMA Related Kinase 2; PKB, Protein Kinase B; PKC, Protein Kinase C; LATS, Large tumor suppressor kinase 1; TRAF2, TNF receptor-associated factor 2.

Fig. 6

Lysine residues in Ubiquitin and AβPP Interactions. A) Prediction of AβPP– Ubiquitin Interactions: The strongest evidence for AβPP–Ubiquitin interactions was reported by S_Net, i.e., sum of edge weights along the shortest path between homologous proteins in a protein-protein interaction network followed by overall interactions (S_All), statistical propensities of domain-domain interactions (S_Dom), and sequence similarities to a known interacting protein pair (S_Seq). B) Important Lysine for AβPP– Ubiquitin Interactions: The AβPP-Ubiquitin interactions revealed the role of diverse lysine residues in the interaction of different AβPP domains and Ubiquitin including K6, K11, K33, K48, and K63 in ubiquitin and K66, K99, K155, K447, K687 (K16 in Aβ), and K699 (K28 in Aβ) in AβPP.

Fig. 7

Protein-Protein Interaction Network. A) Ubiquitination Enzyme Network for AβPP: The interaction network of AβPP and amyloid cascade proteins (BACE1, BACE2, PSEN1, PSEN2, PSENEN, NCSTN, APH1A) with the ubiquitin E3 ligases are shown in molecular action view. Ubiquitin E3 ligases, Park2 and STUB1 along with deubiquitinases, USP25 and UCHL1 interact with AβPP to regulate its ubiquitination process. Other E3 s including FBXO2, CDH1, TRIM13, TRIM55, and SART1 along with deubiquitinases USP8, USP25, USP39, and UCHL1 regulate the ubiquitination of amyloid cascade enzymes β-secretase (BACE1) and γ-subunit complex (PSEN1, PSNEN, NCSTN). The network’s whole genome statistical analysis identified the interacting proteins associated with the proteolysis process, endopeptidase activity, intracellular membrane-bound organelle cellular component and AD pathway. B) E1s and E2s Interaction with PARK2: The network identified different ubiquitin E2 conjugating enzymes, UBE2A, UBE2B, UBE2G1, UBE2J1, UBE2K, UBE2L3, UBE2L6, UBE2N, UBE2Q1, UBE2S, UBE2U, and UBE2Z; and ubiquitin E1 activating enzymes, UBA1, UBA6, UBA7, and ATG7 associated with the conjugation of lysine to Park2. C) E1s and E2s Interaction with STUB1: The network identified different ubiquitin E2 conjugating enzymes-UBE2A, UBE2B, UBE2G1, UBE2I, UBE2J1, UBE2K, UBE2L3, UBE2L6, UBE2N, UBE2Q1, UBE2S, UBE2T, UBE2U, UBE2W, UBE2Z; and ubiquitin E1 activating enzymes-UBA1, UBA6, UBA7, ATG7 associated with the conjugation of lysine to STUB1. STUB1-STIP1 homology and U-box containing protein 1; PSENEN-Presenilin enhancer 2 homolog; CDH1-Cadherin 1; UCHL1-Ubiquitin carboxyl-terminal esterase L1; APP-Amyloid beta (A4) precursor protein; USP25-Ubiquitin specific peptidase 25; NCSTN-Nicastrin; TRIM13-Tripartite motif containing 13; USP8-Ubiquitin specific peptidase 8; SART1-Squamous cell carcinoma antigen recognized by T cells; USP39-Ubiquitin specific peptidase 39; BACE1-Beta-site AβPP-cleaving enzyme 1; TRIM55-Tripartite motif containing 55; PSEN1-Presenilin 1; BACE2-Beta-site AβPP-cleaving enzyme 2; FBXO2-F-box protein 2; PSEN2-Presenilin 2; PARK2-Parkinson protein 2; APH1A-Anterior pharynx defective 1 homolog A.

Moreover, the mutational analysis of lysine’s impact on AβPP stability revealed only arginine for imparting site-specific stability to the AβPP while rest mutations were detrimental for the structural conformation of AβPP. The stability associated with arginine mutations can be attributed to its highly basic nature and resonating states than lysine that can stabilize the positive charges in protein. The arginine has also shown protein stability results with other proteins [66] thereby also indicating their role in structural stability of AβPP. Interestingly, lysine residues at high confidence ubiquitination sites were intolerant against all mutations thereby affecting AβPP stability and signifying their importance in protein’s structural and functional regulation. Further, the evaluation of lysine mutations for disease susceptibility indicated their least effect on the high confidence ubiquitination sites than the other sites. This situation can be extrapolated that ubiquitination may be more selective towards highly conserved or stable lysine sites than the least stable sites as observed by Kim and Hahn, who reported the gain of ubiquitination sites in the highly conserved region of proteins [67]. Further analysis by Mutpred identified lysine residues to play a role in AβPP acetylation, glycosylation, phosphorylation, and SUMOylation apart from the ubiquitination. Additionally, investigations on AβPP-ubiquitin non covalent electrostatic, polar interactions reported K66, K99, K155, K447, K687, and K699 lysine residues having their plausible role in ubiquitin positioning or ubiquitin-AβPP conjugation that are the foundations for further research. Last but not the least, we investigated the interaction network of E1s, E2s, E3s, and DUBs enzymes for the clearance of AβPP in AD biology. The network revealed Parkin and STUB1 to be the key ubiquitin E3 ligases and USP25 and UCHL1 to be the key deubiquitinases directly involved in the ubiquitination of AβPP along with more than a dozen of E2 conjugating and E1 activating enzymes. While other E3 ligases such as FBXO2, TRIM13, CDH1, TRIM55, and SART1 are reported to regulate the ubiquitination of BACE1, PSENEN, PSEN1, and NCSTN, respectively. Similarly, other deubiquitinases, like USP8 and USP39, regulate BACE1 and NCSTN deubiquitination, respectively. In summary, the indepth studies pertaining to the lysine potential in AβPP processing, stability, interaction, ubiquitination, and other functions are provided herein that needed further investigations in vivo at the molecular level to devise novel therapeutic modalities against AD.

Conclusions

The Aβ synthesis is the consequence of AβPP processing by the β- and γ-secretases in neurons through amyloidogenic pathway, which is tightly regulated by the ubiquitination process. Moreover, any disruption in the ubiquitination of AβPP and Aβ, selectively amplify Aβ level that subsequently triggers AD pathogenesis. The identification of key lysine governing ubiquitination is not only crucial for regulating Aβ level in neurons but also for determining the mechanism of AβPP’s subcellular trafficking and processing. Further identification of potential E3 ligases for directly (Parkin and STUB1) or indirectly (FBXO2, TRIM13, CDH1, TRIM55, and SART1) restraining Aβ production served as key therapeutic candidates for targeting neurodegenerative pathologies. Moreover, their identification is also crucial for developing chimera products like PROTACs for achieving selective protein degradation in the diseased state, which is an interesting area of research for the scientist [68]. Moreover, the finding of conserved residues near ubiquitination sites will aid in better understanding the mechanism behind lysine selection by E3 ligases and their interactions for ubiquitin positioning to govern lysine specific (K6, K11, K27, K29, K33, K48, K63) polyubiquitination. Additionally, we also reported that lysine residue of target protein may be involved in the ubiquitin positioning for determining the type of poly-ubiquitination chain during ubiquitin attachment. The present research has revealed the dynamics of Aβ or AβPP ubiquitination, which is essential to device, the strategies to regulate the AβPP metabolism. This would enable us to avoid the complications arisen from the elimination of secretases mediated biological functions of AβPP fragments: ectodomain-sAβPPα, sAβPPβ, N-terminal-APP-NTFs, and intracellular-AICD in regulating gene transcriptions [69]. Moreover, AβPP is proposed to be a cell-surface receptor [70]; therefore, AβPP ubiquitination may also serve as a signaling event for some unknown cellular processes apart from merely a signal for degradation. Further investigations are required to develop the prospective therapeutic agents that can address the clearance of such toxic proteins (Aβ) or their progenitors (AβPP) in a regulated way to ameliorate the neurodegenerative diseases like AD globally.

Footnotes

ACKNOWLEDGMENTS

We would like to thanks the senior management of Delhi Technological University for constant support and encouragement.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Singh

, Srivastav

, Yadav

, Srikrishna

, Perry

(2016) Overview of Alzheimer’s disease and some therapeutic approaches targeting Aß by using several synthetic and herbal compounds. Oxid Med Cell Longev 2016, 7361613.

Serrano-Pozo

, Frosch

, Masliah

, Hyman

(2011) Neuropathological alterations in Alzheimer disease. Cold Spring Harb Perspect Med 1, a006189.

Shankar

, Walsh

(2009) Alzheimer’s disease: Synaptic dysfunction and Abeta. Mol Neurodegener 4, 48.

Jha

, Jha

, Kumar

, Kejriwal

, Sharma

, Ambasta

, Kumar

(2015) Impact of insulin degrading enzyme and neprilysin in Alzheimer’s disease biology: Characterization of putative cognates for therapeutic applications. J Alzheimers Dis 48, 891–917.

Storck

, Meister

, Nahrath

, Meißner

, Schubert

, Di Spiezio

, Baches

, Vandenbroucke

, Bouter

, Prikulis

, Korth

, Weggen

, Heimann

, Schwaninger

, Bayer

, Pietrzik

(2016) Endothelial LRP1 transports amyloid-ß(1-42) across the blood-brain barrier. J Clin Invest 126, 123–136.

Baranello

, Bharani

, Padmaraju

, Chopra

, Lahiri

, Greig

, Pappolla

, Sambamurti

(2015) Amyloid-beta protein clearance and degradation (ABCD)pathways and their role in Alzheimer’s disease. Curr Alzheimer Res 12, 32–46.

Suryadinata

, Roesley

, Yang

, Sarcevic

(2014) Mechanisms of generating polyubiquitin chains of different topology. Cells 3, 674–689.

, Ye

(2008) Polyubiquitin chains: Functions, structures, and mechanisms. Cell Mol Life Sci 65, 2397–2406.

Dammer

, Na

, Xu

, Seyfried

, Duong

, Cheng

, Gearing

, Rees

, Lah

, Levey

, Rush

, Peng

(2011) Polyubiquitin linkage profiles in three models of proteolytic stress suggest the etiology of Alzheimer disease. JBiolChem 286, 10457–10465.

10.

Gong

, Radulovic

, Figueiredo-Pereira

, Cardozo

(2016) The ubiquitin-proteasome system: Potential therapeutic targets for Alzheimer’s disease and spinal cord injury. Front Mol Neurosci 9, 4.

11.

Hong

, Liu

, Zhu

, Zhuang

, Suo

, Wang

, Huang

, Xu

, Huang

, Yu

, Bian

, Sheng

, Fei

, Song

, Behnisch

, Huang

(2014) Parkin overexpression ameliorates hippocampal long-term potentiation and ß-amyloid load in an Alzheimer’s disease mouse model. Hum Mol Genet 23, 1056–1072.

12.

Burns

, Zhang

, Rebeck

, Querfurth

, Moussa

(2009) Parkin promotes intracellular Abeta1-42 clearance. Hum Mol Genet 18, 3206–3216.

13.

Rosen

, Moussa

, Lee

HK,

Kumar

, Kitada

, Qin

, Fu

, Querfurth

(2010) Parkin reverses intracellu-lar beta-amyloid accumulation and its negative effects on proteasome function. J Neurosci Res 88, 167–178.

14.

Kumar

, Pradhan

, Karunya

, Ambasta

, Querfurth

(2012) Cross-functional E3 ligases Parkin and C-terminus Hsp70-interacting protein in neurodegenerative disorders. J Neurochem 120, 350–370.

15.

Zhang

, Cai

, Zhang

, Song

(2014) Overexpression of ubiquitin carboxyl-terminal hydrolase L1 (UCHL1) delays Alzheimer’s progression in vivo. Sci Rep 4, 7298.

16.

Del Prete

, Rice

, Rajadhyaksha

, D'Adamio

(2016) 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. J Biol Chem 291, 17209–17227.

17.

Watanabe

, Hikichi

, Willuweit

, Shintani

, Horiguchi

(2012) FBL2 regulates amyloid precursor protein (APP) metabolism by promoting ubiquitination-dependent APP degradation and inhibition of APP endocytosis. J Neurosci 32, 3352–3365.

18.

Morel

, Chamoun

, Lasiecka

, Chan

, Williamson

, Vetanovetz

, Dall'Armi

, Simoes

, Point Du Jour

, McCabe

, Small

, Di Paolo

(2013) Phosphatidylinositol-3-phosphate regulates sorting and processing of amyloid precursor protein through the endosomal system. Nat Commun 4, 2250.

19.

Atkin

, Hunt

, Minakawa

, Sharkey

, Tipper

, Ten-nant

, Paulson

(2014) F-box only protein 2 (Fbxo2) regulates amyloid precursor protein levels and processing. JBiol Chem 289, 7038–7048.

20.

Williamson

, Laulagnier

, Miranda

, Fernandez

, Wolfe

, Sadoul

, Di Paolo

(2017) Disruption of amyloid precursor protein ubiquitination selectively increases amyloid ß (Aß) 40 levels via presenilin 2-mediated cleavage. J Biol Chem 292, 19873–19889.

21.

Kaneko

, Koike

, Saito

, Kitamura

, Okuma

, Nomura

(2010) Loss of HRD1-mediated proteindegradation causes amyloid precursor protein accumulation and amyloid-beta generation. J Neurosci 30, 3924–3932.

22.

Jung

, Hong

, Kim

, Mook-Jung

(2015) Acute ER stress regulates amyloid precursor protein processing through ubiquitin-dependent degradation. Sci Rep 5, 8805.

23.

Kumar

, Ambasta

, Veereshwarayya

, Rosen

, Kosik

, Band

, Mestril

, Patterson

, Querfurth

(2007) CHIP and HSPs interact with beta-APP in a proteasome-dependent manner and influence Abeta metabolism. Hum Mol Genet 16, 848–864.

24.

Lopez Salon

, Pasquini

, Besio Moreno

, Pasquini

, Soto

(2003) Relationship between beta-amyloid degradation and the 26S proteasome in neural cells. Exp Neurol 180, 131–143.

25.

Favit

, Grimaldi

, Alkon

(2000) Prevention of beta-amyloid neurotoxicity by blockade of the ubiquitin-proteasome proteolytic pathway. JNeurochem 75, 1258–1263.

26.

El Ayadi

, Stieren

, Barral

, Boehning

(2012) Ubiquilin-1 regulates amyloid precursor protein maturation and degradation by stimulating K63-linked polyubiquiti-nation of lysine 688. Proc Natl Acad Sci USA 109, 13416–13421.

27.

Scuderi

, Steardo

, Esposito

(2014) Cannabidiol promotes amyloid precursor protein ubiquitination and reduction of beta amyloid expression in SHSY5YAPP+cells through PPAR7 involvement. Phytother Res 28, 1007–1013.

28.

Kelley

, Mezulis

, Yates

, Wass

, Sternberg

(2015) The Phyre2 web portal for protein modeling, prediction and analysis. Nat Protoc 10, 845–858.

29.

Finn

, Coggill

, Eberhardt

, Eddy

, Mistry

, Mitchell

, Potter

, Punta

, Qureshi

, Sangrador-Vegas

, Salazar

, Tate

, Bateman

(2016) The Pfam protein families database: Towards a more sustainable future. Nucleic Acids Res 44(D1), D279–285.

30.

, Zhang

(2011) Improving the physical realism and structural accuracy of protein models by a two-step atomic-level energy minimization. Biophys J101, 2525–2534.

31.

Pontius

, Richelle

, Wodak

(1996) Deviations from standard atomic volumes as a quality measure for protein crystal structures. J Mol Biol 264, 121–136.

32.

Colovos

, Yeates

(1993) Verification of protein structures: Patterns of nonbonded atomic interactions. Protein Sci 2, 1511–1519.

33.

Luthy

, Bowie

, Eisenberg

(1992) Assessment of protein models with three-dimensional profiles. Nature 356, 83–85.

34.

Lovell

, Davis

, Arendall

3rd , de Bakker

, Word

, Prisant

, Richardson

(2003) Structure validation by Calpha geometry: Phi, psi and Cbeta deviation. Proteins 50, 437–450.

35.

Hall

(1999) BioEdit: A user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucl Acids Symp Ser 41, 95–98.

36.

Ciemny

, Kurcinski

, Kozak

, Kolinski

, Kmiecik

(2017) Highly flexible protein-peptide docking using CABS-Dock. Methods Mol Biol 1561, 69–94.

37.

Huang

, Su

, Kao

, Jhong

, Weng

, Lee

(2016) UbiSite: Incorporating two-layered machine learning method with substrate motifs to predict ubiquitin-conjugation site on lysines. BMC Syst Biol 10(Suppl 1), 6.

38.

Xue

, Li

, Wang

, Feng

, Yao

(2006) PPSP: Prediction of PK-specific phosphorylation site with Bayesian decision theory. BMC Bioinformatics 7, 163.

39.

Chen

, Zhou

, Song

, Zhang

(2013) hCK-SAAP_UbSite: Improved prediction of human ubiquitina-tion sites by exploiting amino acid pattern and properties. Biochim Biophys Acta 1834, 1461–1467.

40.

Radivojac

, Vacic

, Haynes

, Cocklin

, Mohan

, Heyen

, Goebl

, Iakoucheva

(2010) Identification, analysis, and prediction of protein ubiquitination sites. Proteins 78, 365–380.

41.

Tung

, Ho

(2008) Computational identification of ubiquitylation sites from protein sequences. BMC Bioinfor-matics 9, 310.

42.

DeLano

(2002) The PyMOL Molecular Graphics System, DeLano Scientific LLC, San Carlos, CA, 700.

43.

Guex

, Peitsch

, Schwede

(2009) Automated comparative protein structure modeling with SWISS-MODEL and Swiss-PdbViewer: A historical perspective. Elec-trophoresis 30 Suppl 1, S162–173.

44.

Gasteiger

, Hoogland

, Gattiker

, Duvaud

, Wilkins

, Appel

, Bairoch

(2005) Protein identification and analysis tools on the ExPASy Server. In The Proteomics Protocols Handbook, Walker JM, ed. Humana Press, pp. 571–607.

45.

Pejaver

, Urresti

, Lugo-Martinez

, Pagel

, Lin

, Nam

, Mort

, Cooper

, Sebat

, Iakoucheva

, Mooney

, Radivojac

(2017) MutPred2: Inferring the molecular and phenotypic impact of amino acid variants. bioRxiv 134981, doi: 10.1101/134981

46.

Tang

, Thomas

(2016) PANTHER-PSEP: Predicting disease-causing genetic variants using position-specific evolutionary preservation. Bioinformatics 32, 2230–2232.

47.

Hecht

, Bromberg

, Rost

(2015) Better prediction of functional effects for sequence variants. BMC Genomics 16(Suppl 8), S1.

48.

Adzhubei

, Schmidt

, Peshkin

, Ramensky

, Gerasi-mova

, Bork

, Kondrashov

, Sunyaev

(2010) A method and server for predicting damaging missense mutations. Nat Methods 7, 248–249.

49.

Ferrer-Costa

, Gelpi

, Zamakola

, Parraga

, de la Cruz

, Orozco

(2005) PMUT: A web-based tool for the annotation of pathological mutations on proteins. Bioinformatics 21, 3176–3178.

50.

Capriotti

, Calabrese

, Casadio

(2006) Predicting the insurgence of human genetic diseases associated to single point protein mutations with support vector machines and evolutionary information. Bioinformatics 22, 2729–2734.

51.

Sim

, Kumar

, Hu

, Henikoff

, Schneider

, Ng

(2012) SIFT web server: Predicting effects of amino acid substitutions on proteins. Nucleic Acids Res 40(Web Server issue), W452–457.

52.

Murakami

, Mizuguchi

(2014) Homology-based prediction of interactions between proteins using Averaged One-Dependence Estimators. BMC Bioinformatics 15, 213.

53.

Tovchigrechko

, Vakser

(2006) GRAMM-X public web server for protein-protein docking. Nucleic Acids Res 34(Web Server issue), W310–4.

54.

Szklarczyk

, Franceschini

, Wyder

, Forslund

, Heller

, Huerta-Cepas

, Simonovic

, Roth

, Santos

, Tsafou

, Kuhn

, Bork

, Jensen

, von Mering

(2015) STRING v10: Protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res 43(Database issue), D447–452.

55.

McKenna

, Spyracopoulos

, Moraes

, Pastushok

, Ptak

, Xiao

, Ellison

(2001) Noncovalent interaction between ubiquitin and the human DNA repair protein Mms2 is required for Ubc13-mediated polyubiquitination. J Biol Chem 276, 40120–40126.

56.

Saric

, Muller

, Seitz

, Pavelic

(2003) Non-covalent interaction of ubiquitin with insulin-degrading enzyme. Mol Cell Endocrinol 204, 11–20.

57.

Ouyang

, Garner

, Hallet

, Nguyen

, Rickman

, Gill

, Smogorzewska

, Zou

(2015) Noncovalent interactions with SUMO and ubiquitin orchestrate distinct functions of the SLX4 complex in genome maintenance. Mol Cell 57, 108–122.

58.

Nick Pace

, Scholtz

, Grimsley

(2014) Forces stabilizing proteins. FEBS Lett 588, 2177–2184.

59.

Guo

, Sun

, Li

, Xiong

, Wang

, Shu

, Zhu

, Liu

, Huang

, Madley

, Wang

, Cui

, Arvan

, Liu

(2018) Positive charge in the n-region of the signal peptide contributes to efficient post-translational translocation of small secretory preproteins. J Biol Chem 293, 1899–1907.

60.

Cappai

(2014) Making sense of the amyloid precursor protein: Its tail tells an interesting tale. J Neurochem 130, 325–327.

61.

Abrusan

, Marsh

(2016) Alpha helices are more robust to mutations than beta strands. PLoS Comput Biol 12, e1005242.

62.

Cai

, Jiang

(2016) Computational methods for ubiqui-tination site prediction using physicochemical properties of C protein sequences. BMC Bioinformatics 17, 116.

63.

Swatek

, Komander

(2016) Ubiquitin modifications. Cell Res 26, 399–422.

64.

Fornili

, Pandini

, Lu

, Fraternali

(2013) Specialized dynamical properties of promiscuous residues revealed by simulated conformational ensembles. J Chem Theory Comput 9, 5127–5147.

65.

Fallas

, Dong

, Tao

, Hartgerink

(2012) Structural insights into charge pair interactions in triple helical collagen-like proteins. J Biol Chem 287, 8039–8047.

66.

Sokalingam

, Raghunathan

, Soundrarajan

, Lee

(2012) A study on the effect of surface lysine to arginine mutagenesis on protein stability and structure using green fluorescent protein. PLoS One 7, e40410.

67.

Kim

, Hahn

(2012) Gains of ubiquitylation sites in highly conserved proteins in the human lineage. BMC Bioin-formatics 13, 306.

68.

Lai

, Crews

(2017) Induced protein degradation: An emerging drug discovery paradigm. Nat Rev Drug Discov 16, 101–114.

69.

Wang

, Saunders

(2014) The role of ubiquitin-proteasome in the metabolism of amyloid precursor protein (APP): Implications for novel therapeutic strategies for Alzheimer’s disease. Discov Med 18, 41–50.

70.

Thinakaran

, Koo

(2008) Amyloid precursor protein trafficking, processing, and function. J Biol Chem 283, 29615–29619.