Creating and Applying a Reference to Facilitate the Discussion and Classification of Proteins in a Diverse Group

Summary

The goal of this protocol is to develop a reference for divergent proteins in a group that lacks coherent criteria for nomenclature and classification. This reference will facilitate analyses and discussion of the group as a whole and can be used in addition to established names.

Abstract

Related proteins that have been studied in different labs using varying organisms may lack a uniform system of nomenclature and classification, making it difficult to discuss the group as a whole and to place new sequences into the appropriate context. Developing a reference that prioritizes important sequence features related to structure and/or activity can be used in addition to established names to add some coherency to a diverse group of proteins. This paper uses the cysteine-stabilized alpha-helix (CS-αβ) superfamily as an example to show how a reference generated in spreadsheet software can clarify relationships between existing proteins in the superfamily, as well as facilitate the addition of new sequences. It also shows how the reference can help to refine sequence alignments generated in commonly used software, which impacts the validity of phylogenetic analyses. The use of a reference will likely be most helpful for protein groups that include highly divergent sequences from a broad spectrum of taxa, with features that are not adequately captured by molecular analyses.

Introduction

A protein's name should reflect is characteristics and relationship to other proteins. Unfortunately, names are generally assigned at the time of discovery and, as research continues, the understanding of the larger context may change. This can lead to multiple names if a protein was independently identified by more than one lab, to changes in nomenclature or in the characteristics thought to be definitive when assigning the name, and to the name no longer sufficiently differentiating the protein from others.

Invertebrate defensins provide a good example of degeneration in nomenclature and classification. The first invertebrate defensins were reported from insects, and the name "insect defensin" was proposed based on the perceived homology to mammalian defensins^1,2. The term defensin is still used, even though it is now clear that invertebrate and mammalian defensins do not share a common ancestor^3,4. Depending on the species, an invertebrate "defensin" may have six or eight cysteines (that form three or four disulfide bonds) and a variety of antimicrobial activities. To complicate the situation, proteins with the same characteristics as defensins are not always called "defensins," such as the recently identified cremycins from Caenorhabditis remanei⁵. In addition, invertebrate big defensins are more likely to be evolutionarily related to vertebrate β-defensins than to other invertebrate defensins⁶. Despite this, researchers sometimes rely on the name "defensin" when determining which sequences should be included in analyses.

Structural studies revealed the similarity between insect defensins and scorpion toxins⁷, and the CS-αβ fold was subsequently established as the defining structural characteristic of insect defensins⁸. This fold defines the scorpion toxin-like (CS-αβ) superfamily in the Structural Classification of Proteins (SCOP) database⁹, which currently includes five families: insect defensins, short-chain scorpion toxins, long-chain scorpion toxins, MGD-1 (from a mollusk), and plant defensins. This superfamily is synonymous with the recently described cis-defensins⁴ and Superfamily 3.30.30.10 in the CATH/Gene 3D database^10,11. Studies from a variety of invertebrate taxa, plants, and fungi show that the names of proteins that contain this fold are not clearly related to cysteine number or bonding pattern, antimicrobial activity, or evolutionary history¹².

The lack of consistency and clear criteria make it challenging to name and classify newly-identified sequences in this superfamily. A major obstacle to comparing proteins in this superfamily is that cysteines are numbered with respect to each individual sequence (the first cysteine in each sequence is C1), with no way to account for the structural role. This means that only sequences with the same number of cysteines can be compared. There is little sequence conservation other than the cysteines forming the CS-αβ fold, which makes alignments and phylogenetic analyses difficult. By developing a numbering system that prioritizes structural features, superfamily sequences can be more easily compared and aligned. Conserved features, as well as those defining subgroups, can be visualized quickly, and new sequences can be more easily placed into the appropriate context.

This paper uses a spreadsheet software (e.g., Excel) to generate a reference numbering system for the CS-αβ superfamily. It shows how this clarifies comparisons between sequences and applies it to new CS-αβ sequences identified from tardigrades. Using the CS-αβ superfamily as an example, the protocol was written to provide guidance when using sequences of interest; however, it is not intended to be specific to this superfamily or to cysteine-rich sequences. This method will likely be most useful for groups of proteins that have been researched independently in divergent taxa and/or have little overall sequence homology, with discrete characteristics that may not be easily recognized by molecular analysis software. This method requires some a priori decisions regarding important features, so it will be of limited utility if no important features have been identified. The primary goal is to show how a simple visualization of the sequence relationships can be achieved. This can then be used to inform sequence alignment and analysis, but if alignment and analysis are the primary goals, a barcode method would be a suitable alternative that has more capacity for automation¹³. The current method displays the features of each peptide in a linear form, so it will not be helpful for the direct visualization of 3D structure.

Protocol

1. Determine the Defining Features of the Protein Group of Interest

1. Consult previous publications to determine if there is a consensus regarding the features that are necessary to be considered part of the group. Take note of any inconsistencies or differences in opinion between research groups, and include characteristics that may serve to differentiate one subgroup from another.
2. If previous literature does not address defining characteristics, use sequences that are considered representative of the group as a starting point to identify conserved features.

2. Collect Relevant Sequences

1. If reviews have been written that include analyses of sequences that are representing the group, include these sequences in the raw dataset. Retrieve sequences using accession numbers referenced in the literature and save in a standard sequence editing program (e.g., EditSeq in the Lasergene suite or one of many available for free online).
2. If the group in question has been defined in one of the structural databases, include the sequences the database lists as being part of the group. Retrieve sequences using accession numbers provided in the database and save in a standard sequence editing program, as above.
   NOTE: For example, the sequences categorized in the CS-αβ (scorpion toxin-like) superfamily in the SCOP database can be found here: http://scop.mrc-lmb.cam.ac.uk/scop/data/scop.b.h.c.h.html. 
3. Perform Basic Local Alignment Search Tool (BLAST)¹⁴ searches of public, online databases available through the National Center for Biotechnology Information (NCBI) to find sequences that may have not been included in the literature or structural databases. For the most complete results, use both the protein BLAST (blastp) and translated blast with protein query (tblastn) programs; these are both available at: https://blast.ncbi.nlm.nih.gov/Blast.cgi.
   1. Use sequences known to be part of the group of interest as query sequences. Copy and paste the sequence into the search box at the top, or provide a GenBank accession number or gi identifier, if available.
   2. Choose the database from the dropdown menu. Choose non-redundant protein sequences (nr) for blastp and expressed sequence tags for tblastn.
   3. Search for results in specific taxa in the organism setting by typing the organism or taxon name and choosing from the list that appears while typing. To add additional organisms or taxa to exclude, click the "+" button and another field will appear. Exclude any unwanted taxa in the organism box by typing the organism or taxon name, choosing from the list that appears while typing, and checking the "Exclude" box on the right.
   4. Access additional parameters by clicking on "Algorithm parameters" near the bottom of the page. Leave at default unless there is a rationale for changing a parameter.
   5. Click the "BLAST" button to run the analysis; it may take some time for the results to appear. In general, retrieve hits with an expect value (or e-value) of "-05" or better and save in a standard sequence editing program.
      1. If all hits are above this threshold, rerun the search with an increased number of target sequences (in the algorithm parameters section) to obtain all relevant sequences.
4. If necessary, trim the sequences to exclude irrelevant information (e.g., the CS-αβ fold only applies to the mature peptide). Identify signal peptides and pro-peptides for removal using ProP¹⁵ (available online), or SignalP for more sophisticated signal peptide prediction¹⁶ (available online).

3. Generate a Reference in a Spreadsheet Based on the Important Features That Were Identified

1. Identify the defining characteristics of the group of interest. For example, use the CS-αβ fold definitively established by the solution structure of insect defensin A from Phormia terraenovae (Figure 1)⁸.
   1. This fold includes a smaller motif called the cysteine-stabilized helix (CSH)¹⁷; identify this motif by a CXXXC (where X is any amino acid) upstream of a CXC that form two disulfide bonds (Figure 1, solid pink lines).
      NOTE: To complete the CS-αβ motif, a third disulfide bond is formed from additional cysteines placed before each half of the CSH motif (Figure 1, dotted pink lines).
2. Enter these defining features into a spreadsheet. See Figure 2.
   1. Use columns for the conserved features and to represent the spaces between these features. Keep the columns wide enough to fit numbers and ensure that they have a consistent width. Set the width using the "Format| Column Width" function (Figure 2, pink arrow).
   2. Use the rows for the sequence names.
   3. When a sequence has the feature, fill in the box using the fill function (Figure 2, pink square). For spacing between features, enter the number of amino acids in the box between and leave it unfilled. For example, using the insect defensin sequence gives a reference that includes six cysteines, with defined spacings between C2 and C3 and between C5 and C6.
3. Add representative sequences that have been previously established as members of the group based on the structural databases and literature.
   NOTE: For example, previous literature and the SCOP database identify several groups for inclusion: insect defensins, short-chain scorpion toxins, long-chain scorpion toxins, MGD-1, plant defensins, nematode ABFs, drosomycins from Drosophila, and macins. The literature also identifies a bacterial sequence with only four cysteines that might represent the ancestor of this superfamily¹⁸. Adding these sequences increases the number of cysteines in the reference from six to ten but maintains the alignment of the important structural features (Figure 3).
   1. To add a feature that is likely to define a subgroup of sequences (for example, an extra cysteine), use the "Insert" function (Figure 3, pink arrow).
   2. If there are features missing from a given sequence, leave the box unfilled and combine it with boxes representing intervening amino acids. If necessary, merge the cells using the merge and center feature (Figure 3, pink box).
4. Continue adding sequences to the groups to gain a better picture of the variation in each group of the larger superfamily. Summarize the group characteristics to facilitate comparisons (Figure 4).
   1. When the number of amino acids between major features varies, use a hyphen to indicate a range, such as 6 - 12 (6 to 12 amino acids), and a slash to indicate either/or, such as 7/10 (7 or 10 amino acids).
   2. Choose a way to annotate features of sequences that may be relevant but do not occur often enough to include in the reference. For example, since cysteines are important in this superfamily, label additional cysteines (Figure 4, pink boxes).
5. Add newly-identified sequences to the spreadsheet using the established sequences as a guide. For example, adding sequences from tardigrades (yellow) shows that the tardigrade sequences fall into several different groups of the superfamily (Figure 5 shows summaries instead of a row per sequence for space purposes).
6. Show variability within a taxonomic group by rearranging the rows (Figure 6).

4. Use the Reference to Refine Amino Acid Alignments

NOTE: There are many programs that can be used for multiple sequence alignments, but this demonstration will use Molecular Evolutionary Genetics Analysis (MEGA6)¹⁹ because it is available to download for free.

1. Download and install the software.
2. Begin a new alignment in MEGA by selecting "Edit/Build Alignment" under the Align tab. Select "Create a new alignment" in the box that appears and click "OK." Then select "Protein."
3. Select "Insert Sequence from File" in the "Edit" menu to import the sequences.
   NOTE: Sequences will need to be in FASTA format for import into MEGA. Background colors that reflect different amino acid types are used by default, but this option can be turned off under the "Display" menu.
4. Once all sequences are entered, click the flexing arm icon and then "Align Protein" to align the sequences using the MUSCLE algorithm²⁰.
   NOTE: ClustalW is also available.
   1. If a message saying that nothing has been selected pops up and asks to select all, click "OK."
   2. NOTE: This opens a window that allows one to change some parameters, but they should only be changed there is reason to do so. This analysis uses a subset of the sequences analyzed in a previous paper¹².
5. Check the alignment based on the important features; note that the top bar above the sequences will show any columns where the amino acid is completely conserved (*). See Figure 7. See that the initial alignment shows only three of the four conserved cysteines (Figure 7, pink boxes); looking down the column, the AlCRP sequence is clearly misaligned (Figure 7, pink arrow).
6. To get rid of the large gap between the I and the conserved C, highlight the dashes and press the "delete" key. Do not highlight any amino acids, or they will be deleted as well.
7. To move amino acids to the right, highlight and press the space bar.
   1. Note that the AlCRP now has the structural cysteines aligned and that the last C of the CXXXC motif is conserved throughout the alignment (Figure 8). Adjust the alignment as necessary to prioritize the most important features of the sequences.

5. Compare the Groups Identified Using the Reference with Results from Phylogenetic Analyses

1. From preliminary alignments, determine which sequences should be included in a phylogenetic analysis; for a small number of sequences, this step may be unnecessary.
   1. Keep an alignment file that includes all sequences, but for a phylogenetic analysis, remove redundant sequences (Figure 9, pink boxes show pairs of redundant sequences).
   2. If the data set includes a large number of sequences, run a preliminary analysis and select representatives from groups that always form a clade.
2. Determine the best amino acid substitution model.
   1. Export the alignment in MEGA format (under the data tab).
   2. Go to the Models menu and select "Find Best DNA/Protein Model." Choose the file just saved and open it; this will open a window that has some parameters that can be changed.
   3. Use the default parameters unless there is a reason to change them. Click "Compute" to begin the analysis.
3. Run a maximum likelihood (ML) analysis in MEGA.
   1. Choose "Construct/Test Maximum Likelihood Tree" from the Phylogeny menu.
   2. Choose the model determined to be the best fit for the data from step 5.2 (the output will give the substitution model as well as the best "rates among sites" parameter).
   3. Choose 1,000 bootstrap replicates to obtain the measures of support for the tree.
   4. Click "Compute" to run the analysis; MEGA has a "Tree Explorer" to visualize the tree.
4. Run a Bayesian analysis in MrBayes open-source software²¹.
   NOTE: A MrBayes manual is also available from this site. This is intended to provide basic steps and is not a comprehensive guide to conducting Bayesian phylogenetic analysis.
   1. Export the MEGA alignment in PAUP (Nexus) format to the same folder as the MrBayes program.
   2. Open MrBayes and type "exe Filename" (e.g., "exe Alignment.nex").
   3. Specify the model and analysis parameters. Choose either the model specified in step 5.2 or choose the "mixed" setting that will try various models and report the frequency of the model in the trees with the best posterior probabilities (prset aamodelpr=mixed). Type "showmodel" to report the current model settings and "help mcmc" to show current parameter settings, with a brief explanation of each.
   4. Set the number of generations using the "mcmcp ngen=" command (1 million is typical).
   5. Type "mcmc" to begin the analysis.
   6. When the number of generations has completed, the program will ask to add more generations. If the average standard deviation of split frequencies is less than 0.1, type no. If it is above 0.1, the analysis should be allowed to continue, or some parameters should be changed (see the manual).
   7. Use the "sumt" command to generate the tree files.
   8. After the analysis is complete and a consensus tree is generated, the tree can be viewed in FigTree (available online).
5. Compare the trees to see if the methods generate consistent results.
   NOTE: Some sequences do not provide a lot of information: the trees may not be well resolved and the branches may have minimal support (Figure 10).
6. Compare trees to the groups identified using the reference to see if the phylogenetic analyses support these groups.

Results

Groups of sequences in the CS-αβ superfamily reported in the literature are shown in Figure 4. The cysteine pairings based on the numbering for each sequence suggest five basic groups (Table 1, middle column). Group 1 has six cysteines that from three disulfide bonds and includes sequences from insects, arachnids, mollusks, nematodes, and fungi. Groups 2, 3, and 4 have 8 cysteines that form four disulfide bonds. Group 2 includes insect, arach...

Discussion

The criteria for naming a protein within a group should be clear, but this is not always the case. Sequences that have the CS-αβ fold have been studied in many labs using a variety of organisms, resulting in different systems of nomenclature, as well as varying levels of characterization. Attempting to impose a completely new nomenclature is not reasonable and would result in a great deal of confusion when consulting previous literature. A reference numbering system can be used in addition to the name of a prot...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
BLAST webpage			https://blast.ncbi.nlm.nih.gov/Blast.cgi
EditSeq (Lasergene suite)	DNASTAR		https://www.dnastar.com/t-allproducts.aspx
Excel 2013	Microsoft
FigTree			http://tree.bio.ed.ac.uk/software/figtree/
MEGA			www.megasoftware.net
MrBayes			http://mrbayes.sourceforge.net/
SCOP database			http://scop.mrc-lmb.cam.ac.uk/scop/