Block-based characterization of protease specificity from substrate sequence profile
- 855 Downloads
The mechanism of action of proteases has been widely studied based on substrate specificity. Prior research has been focused on the amino acids at a single amino acid site, but rarely on combinations of amino acids around the cleavage bond.
We propose a novel block-based approach to reveal the potential combinations of amino acids which may regulate the action of proteases. Using the entropies of eight blocks centered at a cleavage bond, we created a distance matrix for 61 proteases to compare their specificities. After quantitative analysis, we discovered a number of prominent blocks, each of which consists of successive amino acids near a cleavage bond, intuitively characterizing the site cooperation of the substrate sequences.
This approach will help in the discovery of specific substrate sequences which may bridge between proteases and cleavage substrate as more substrate information becomes available.
KeywordsProtease Block Entropy Site cooperation
Proteases are a category of enzymes capable of hydrolyzing peptide bonds and irreversibly modifying functions of substrate proteins. These hydrolyzations and modifications are essential for cell growth and differentiation [1, 2]. Recognition of the target substrate of a protease depends partly on the complementation between the protease active site and the sequence surrounding the scissile bond in the substrate. Proteases have pockets that accommodate substrate residues. Substrate sequences that bind the pockets are indexed by P4, P3, P2, P1, P1’, P2’, P3’, P4’ in order from N-terminal to C-terminal following the convention of Schechter and Berger .
Some proteases show strict specificities on the cleavage sequences of the substrates. For example, trypsin 1 requires Lys and Arg at the P1 site , and granzyme B shows strict specificity for Asp at the P1 site . The specificity of protease has been widely used not only in identifying the biologically relevant substrates, but also in applying protease to site-specific proteolysis [6, 7]. Proteases participate in various disease processes, exhibiting a potentially huge future application in the design of new drug targets for enzyme [8, 9] and protease inhibitors . Although all the proteases function in hydrolyzing peptide bonds, almost all are linked to a particular cleavage pattern .
The MEROPS database is a manually curated information resource for peptidases . According to MEROPS, more than 10,000 known substrates are profiled for some proteases , so it is necessary to develop an approach to map the abundant substrate-sequence information to specificities of proteases to highlight the enzymatic preferences, especially for specific catalytic types . Integrating features of substrate sequences characteristics, PoPS  and PROSPER  are proposed to predict protease substrate cleavage sites. A well-designed approach of identifying the specificity of the protease will contribute to a better method of predicting the substrate cleavage site.
Previous analyses of protease cleavage data, such as visualized sequence logos , iceLogo , heat maps  and several techniques [20, 21, 22], have been focused on qualitative interpretation. Using LC-MS/MS sequencing , a simple and rapid multiplex substrate profiling method was presented to demonstrate the substrate specificity. Further measures include using fluorogenic substrates , specific labeling techniques of N-terminal [24, 25], and proteome-derived peptide libraries [26, 27, 28, 29, 30]. Fuchs  developed a method to quantify protease specificity and rank proteases with the cleavage entropy of a single position. Several quantitative measures were developed [32, 33, 34, 35], in which the specificities of proteases were shown by the occurrences of amino acids at the binding sites. As mentioned by Schilling and Overall  in profiling the specificity of the MMP2, the preferred amino acid residues at different site may cooperate in the hydrolysis process, therefore, it is critically important to elucidate the hydrolyzation process by the closely cooperative relationship of successive positions on the substrate sequences.
In this study, we designed a novel approach to present the protease specificity based on blocks which are composed by successive amino acids from the substrate sequence. The essential difference between our approach and previous ones lies in that we characterize the specificity of proteases based on successive amino acids rather than a single binding site. This new approach could more reliably identify protease specificity by considering cooperation among the successive sites of the substrate peptides during the hydrolyzation process.
The dataset is composed of 61 proteases for analysis as described by Fuchs . The cleavage information from all experimental sources is obtained from the MEROPS database  and is updated according to MEROPS 10.0. This study focuses on the protease specificity directly on the active sites, ignoring differences in allosteric sites and exosite interactions. Among the data, signal peptidase complex (XS26.001) has been deleted from the dataset since the complex contains two peptidases, and it is not possible to assign a particular cleavage to one activity due to not a single component.
Greedy algorithm for filtering the data
First, the substrate sequence with less than two amino acids is filtered out. Then all of substrate sequences left primarily are aligned pairwise. Starting from sequences with the maximum number of similar amino acids, we remove redundant sequences by greedy algorithm  to make sure that there is no pair of sequences whose similarity is greater than or equal to 0.875. Therefore, there are at least two different amino acid residues between any two remaining substrate sequences.
Construction of blocks
Calculation of entropy
Calculation of distance matrix
Where E i (P) and E i ’(P) are the entropies of blocks B i and B i ′ of protease P respectively. This yields a symmetric distance matrix. The elements on the diagonal are 0, which is the distance of identical proteases.
Principal components analysis
All the eight blocks for each of 61 proteases are used for principal components analysis (PCA). Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy is computed as 0.733 which indicates that the sample size is sufficient for the application of PCA. The PCA is performed in SPSS 19.0 (SPSS Inc., Chicago, IL, USA) with the correlation method and Varimax with Kaiser Normalization as the rotation method.
Fisher’s exact test
2 × 2 contingency table for Fisher’s exact test
a + b
c + d
a + c
b + d
a + b + c + d
Creation of sequence profile
To depict the substrate preferences at different sites, the data of substrate sequences after removing the redundancy is submitted to Weblogo [17, 40] to generate sequence profiles of substrate cleavage site.
Distance character of 61 proteases
The entropies of eight blocks B4, B3, B2, B1, B1’, B2’, B3’, B4’ are calculated and denoted as E4, E3, E2, E1, E1’, E2’, E3’, E4’ correspondingly (Additional file 2: Table S1). There are three blocks with entropy 0, including, caspase 6 with E1 = 0 implying the unique amino acid Asp at site P1; peptidyl-Lys metallopeptidase with E1’ = 0 implying the unique amino acid Lys at site P1’; lysyl peptidase with E1 = 0 implying the unique amino acid Lys at site P1.
Principal components analysis
The entropies of eight blocks reveal the complexity of different combination types. In order to mine the blocks which play the crucial role in the specificity recognition of substrate sequence, we used principal components analysis.
The distribution of eight different eigenvalues is shown in a scree plot (Additional file 2: Figure S1.). Three principal components (PC1: the first principal component; PC2: the second principal component; PC3: the third principal component) are obtained according to the principle of eigenvalues more than 1. Among the three principal components, PC1, PC2 and PC3 contribute 57.938%, 23.284% and 15.960% to the total variance respectively, and the cumulative contribution of three principal components is 97.183% (Additional file 2: Table S3). Thus, the three principal components may represent the main features in the recognition of substrate specificities of different proteases.
Block-based sequence profile
There are a few prominent blocks from prime side. For instance, except for strict specificity for Lys at the P1’ site, peptidyl-Lys metallopeptidase has block B2’ with LysGlu = 179 from 1869 substrates, and signal peptidase 1 has block B3’ with AlaGluAla = 19 from 297 substrates (The number behind the equal sign represents the amount of combination of amino acids in the corresponding block).
Meanwhile, a few blocks from non-prime side show the specificity. For example, kexin has block B2 LysArg = 147 from 171 substrates. With caspase 3 having 571 substrates, besides the prominent block B3 GluValAsp = 43, we still find the prominent block B4 AspGluValAsp = 19.
The top prominent B2 blocks of proteases listed in Fig. 6
(a) The top prominent B2 blocks of proteases listed in Fig. 6a
(b) The top prominent B2 blocks of proteases listed in Fig. 6b
Val, Glu, Ile
Ala, Ser, Gly
Ala, Gly, Asn
Ala, Gly, Pro
Gly, Ala, Pro
Some blocks B2’ show the similar combination property as in blocks B2. For example, the top three amino acids of HIV-1 retropepsin at sites P1’ and P2’ are Leu, Val, Phe and Glu, Val, Ala, respectively, yet the prominent block B2’ with the highest number of combination are LeuAla = 33. For LAST_MAM peptidase, amino acids on the top at sites P1’ and P2’ are Asp, Ala, Glu and Pro, Ala, Glu respectively, yet the top one Asp at the site P1’ shows no preference of the top amino acid Pro at the site P2’, and the prominent block B2’ with the highest number of combination is AlaPro = 44 from 429 substrates. For the proteases which cleavage sites possess two or more preferred residues, the prominent combinations in the blocks reflect the cooperation of the residues in one position with other positions, characterizing the specificity of proteases detailedly.
Some specificities of certain proteases have been determined, such as trypsin 1 , caspase 3 , kexin , furin  and so on. However, by focusing on single positions and not taking into consideration the interaction of adjacent amino acids, the study of substrate specificity is too limited.
Taking the cooperation of amino acids into account, we propose a quantitative method to characterize substrates specificity of different proteases. By calculating entropies of different blocks, some distinctions of substrates between different proteases can be conceived (Fig. 2a). The principal component analysis gives evidence on the existence of blocks which play the crucial role in the specificity recognition of substrate sequence, and most of them are block B2, B1 and B1’. This is confirmed by the statistical analysis showing the ratios at B2, B1, and B1’ are higher than those in other blocks (Fig. 5). With Fisher’s exact test, a number of prominent blocks of different proteases have been discovered. For example, blocks B2 in kexin and furin are consistent with the previous discovery that both of the proteases cleave after dibasic residues . Other block B2, e.g. GluLeu in HIV-1 retropepsin, AlaAla in MMP2 and ProGly in MMP 9, are more likely to reflect the preferences and the cooperation of the successive amino acids in the substrate sequences which could not be found previously.
Cathepsin B is an endopeptidase and as an exopeptidase acts as a peptidyl-dipeptidase, releasing a dipeptide from the C-terminus of a protein or peptide. As no distinction is made in MEROPS between cleavages resulting from either activity, a view of the endopeptidase activity would be clear if the substrates of the exopeptidase activity were filtered out.
From the specificity matrix in MEROPS and the heat map , the preference of the protease is shown by the amino acids at one single binding site. However, it will not show the combinations of amino acids if proteases show multiple preferences at each binding site. Our method indicates interactions of different compositions of successive amino acids which can’t be obtained previously. For example, MMP9 has preferences for Ala, Gly and Pro at the site P2, Gly, Ala and Pro at the site P1 from the specificity matrix, yet the combination is clear using our method, such as ProGly, AlaAla in block B2. Whether a prominent combination exists in a block is obviously presented in the heat map of prominent combinations in each block (Fig. 4). These findings of specific blocks will shed light on future experiments and further investigation of proteolytic specificity.
Although in this study we only focused on the specificity of selected proteases, the method would be applicable to other proteases for mining the specificity pattern of substrates. In conclusion, we can obtain more substrate specificity patterns by site cooperation as more and more substrates data becomes available. Further investigations of the substrate specificity will be important to reveal the hydrolyzation mechanism of proteases.
Generally, the design of experiments and the description of the specificity of the protease are based on the assumption that the process of binding amino acid residue to the corresponding subsite is independent. However, it is not exactly true and the binding of amino acid residues at one site can more or less influence the binding at other subsites. It is essential to take the site cooperation into consideration for understanding fully the active site.
Our approach provides a new framework for dealing with the specificity pattern of substrates of the proteases. The combinations of site cooperation in the substrates offer a new sight in mining the specificity of the protease. We successfully find the significant blocks B2 in kexin and furin which are consistent with the previous discovery that both of the proteases cleave after dibasic residues. Other significant combinations found by the new approach could be more reliable to capture the activity of the active site. In principle, this method is useful for the further research relying on the substrate dataset, such as the identification of the novel substrate and the design of the inhibitor for the protease.
The authors thank the editorial staff for their help in editing this manuscript and thank the anonymous reviewers for their suggestions and comments to improve the manuscript.
This work was supported by National Science Foundation (NSF Grant No.1553680), and National Science Foundation of China (NSFC Grant No. 61432010, 61,272,016 and 31,571,354).
Availability of data and materials
The datasets used in this research are available at http://www.merops.ac.uk.
E.Q and G.L conceived and designed the approach. E.Q, B.G and Y.L implemented the software. D.W and Y.L performed the data analysis. E.Q wrote the manuscript. G.L contributed to revise the manuscript. All author approved the final version of this manuscript.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 11.Hedstrom L.Introduction: proteases. 2002;102(12):4429.Google Scholar
- 23.O'Donoghue AJ, Eroy-Reveles AA, Knudsen GM, Ingram J, Zhou M, Statnekov JB, Greninger AL, Hostetter DR, Qu G, Maltby DA, Anderson MO, Derisi JL, McKerrow JH, Burlingame AL, Craik CS. Nat Methods 2012;9(11):1095–100.Google Scholar
- 42.Demon D, Van Damme P, Vanden Berghe T, Deceuninck A, Van Durme J, Verspurten J, Helsens K, Impens F, Wejda M, Schymkowitz J, Rousseau F, Madder A, Vandekerckhove J, Declercq W, Gevaert K, Vandenabeele P. Proteome-wide substrate analysis indicates substrate exclusion as a mechanism to generate caspase-7 versus caspase-3 specificity. Mol Cell Proteomics. 2009;8(12):2700–14.CrossRefPubMedPubMedCentralGoogle Scholar
- 44.Remacle AG, Shiryaev SA, ES O, Cieplak P, Srinivasan A, Wei G, Liddington RC, Ratnikov BI, Parent A, Desjardins R, Day R, Smith JW, Lebl M, Strongin AY. Substrate cleavage analysis of furin and related proprotein convertases, A comparative study. J Biol Chem. 2008;283(30):20897–906.CrossRefPubMedPubMedCentralGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.