2017 Newsletter Article 2, Zika virus and drug discovery: a classroom-based bioinformatics research project and general method for repurposing existing drugs

Introduction

I teach an on-line bioinformatics course through Austin Community College that focuses on common applications for bioinformatics tools in biotechnology.  Students learn how to search databases, use molecular models, analyze molecular sequences, and characterize genetic variation.  Although bioinformatics provides us with interesting real-life examples to study, our abilities are limited when it comes to finding student projects that include opportunities to make new discoveries. 

Last spring, I developed a research project designed to engage students by having them apply their skills and knowledge to the problem of finding drugs to treat Zika virus.  It seemed that the devastating effects of Zika infection combined with the possibility (albeit low) that students taking our class might be impacted, would have a positive effect on student engagement. 

Students begin the project by looking at Health Map to see where Zika virus has been reported and read about the outcomes of Zika virus infections.  Through their reading, they learn that the NCBI structure database contains structures of proteins bound to inhibitors and drugs and that it might be possible to repurpose an existing drug to treat Zika virus. Students read about the Zika virus genome and life cycle and decide whether to look for a drug that inhibits the RNA polymerase or the protease.  They find the protein sequence for one of these proteins and use it to search the structure database to find similar proteins that are bound to drugs / inhibitors.  After that, they use the Molecule World™ iPad app (1) to identify amino acids that contact the drug and determine if those same residues are present in the Zika virus protein.  

In developing an example to show students how this process would work, I determined that the Zika virus polymerase would be likely to bind to Gilead’s anti-Hepatitis C drug Sovaldi^® (Sofosbuvir).  This prediction was borne out by two recently published studies showing that Sofosbuvir does inhibit Zika virus replication (2, 3).

This general process for repurposing existing drugs is summarized below.  Using this process in a class-based research project provides a new avenue for students to apply bioinformatics tools to real-life problems and potentially, make discoveries.

The method for identifying potential drugs

1.  Decide on a drug target.

2.  Find the amino acid sequence for the protein you wish to target.  

3.  Use protein blast to query the Protein Database Proteins (pdb) database at the NCBI.

4.  Scan the results to find proteins bound to drugs (or inhibitors).

5.  Use Molecule World™ to identify amino acids in the drug binding site.

6.  Determine if amino acids in the drug binding site are present in both proteins.

How do we apply this method? 

Students in our bioinformatics course have used this process for the past three semesters.  In this section, I describe the process in more detail and present an example that I developed with sofosbuvir (Sovaldi^®).

1.  Decide on a drug target.

The first step is to decide which protein to target.  Although many different proteins can be used as drug targets, our students choose between two kinds of proteins:  the viral RNA polymerase and viral protease.  Polymerases play an important role in the life cycle of a virus by making copies of the viral genome.  These copies get packaged into new viral particles that can go on and infect new cells.  Proteases play an important role in viruses that make polyproteins.  These polyproteins must be cut into smaller parts by a protease in order to function. Both polymerase and protease inhibitors are used to treat infections with HIV.  Sofosbuvir, from Gilead, is an example of an FDA-approved drug that inhibits the RNA polymerase from Hepatitis C virus.  

2.  Find the amino acid sequence for the protein you wish to target.  

The NCBI (National Center for Biotechnology Information) was established in 1988 as a division of the National Library of Medicine, at the National Institutes of Health (NIH).  Their overall mission is to create systems for storing and analyzing knowledge about molecular biology, biochemistry, and genetics and to facilitate the use of such databases and software by the research and medical community  (http://ncbi.nlm.nih.gov).  As part of this mission, the NCBI has set up resource pages for specific viral genomes, in particular pathogenic viruses such as Zika virus, MERs, and Ebola (Fig1 https://www.ncbi.nlm.nih.gov/genome/viruses/)

Figure 1.  The Viral Genome portal at the NCBI.

Students navigate to the viral genomes page and choose Zika virus to access a collection of Zika virus resources (Fig. 2  https://www.ncbi.nlm.nih.gov/genome/viruses/variation/Zika/).  

Figure 2.  Zika Virus Resources at the NCBI.

The Zika Virus Resource page has links to Health Map, the CDC (Centers for Disease Control), WHO (the World Health Organization), and publications. From here, we use the “NCBI Zika virus reference genome” link followed by the GenBank record link to locate the accession number for protein sequences from either the protease or the Zika virus RNA polymerase (Fig 3).

Figure 3.  Accession numbers for the Zika virus protease and Zika virus RNA polymerase.

3.  Use protein blast to query the Protein Database Proteins (pdb) database at the NCBI.

The accession number for the RNA polymerase is YP_009227205.1.  Accession numbers act like catalog numbers.  They allow us to find specific sequences in a database.  Clicking the linked accession number from the GenBank record takes us to the protein record (Fig 4).  

Figure 4.  The protein record for the Zika virus RNA polymerase with a link to run BLAST on the right.

On the right hand side of the protein record (Fig. 4) is a link to Run BLAST.  Students select this link to access the protein blast search algorithm.  Protein blast is an algorithm that searches databases of protein sequences to find similar proteins.

From the BLAST search form, we choose the Protein Data Bank database (Fig. 5).  All the proteins in this database can be found in files of molecular structures.  

Figure 5.  The BLAST form.

Next, we enter Zika in the organism field and click Exclude so that we can avoid getting results from Zika virus proteins.  It’s not necessary to know the taxid for this organism.  This information is entered by autocomplete after we type “Zika” in the organism field (Fig. 6). 

Figure 6.  Excluding Zika virus from the BLAST search.

Now, we click BLAST to start the search.  

Our BLAST results appear shortly.  The top graph shows the polymerase domain in tan on the right side (Fig. 7).  The graph below shows where related proteins align to the Zika protein sequence that we used as a query.

Figure 7.  BLAST results showing conserved domains in the Zika virus protein sequence

4.  Scan the results to find proteins bound to drugs (or inhibitors).

Now that we have a set of similar proteins, the next step is find proteins that are bound to potential drugs.  We scroll down the page and read the titles to identify structures where a protein is bound to some kind of inhibitor or potential drug.  

In Figure 8, we can see one structure where the RNA polymerase from Dengue virus is bound to compound 29 and another where it’s bound to compound 27.  Both compounds 27 and 29 might be drug candidates.  These are shown in the same alignment because the protein sequences are identical.  

Figure 8. Protein sequences from molecular structure files showing that the protein may be bound to an inhibitor.

We can confirm that these compounds might be drug candidates or inhibitors by following the links to the structure record and skimming the title of the paper or abstract if there’s one accompanying the structure.

Both of these compounds are described in a paper discussing potent inhibitors of Dengue virus.  The title and the abstract tell us the authors were looking for potential drugs.

As mentioned earlier, I wanted to provide students with an example to show how this process might work.  Since Sofosbuvir has been approved by the FDA to treat Hepatitis C infections and the Hepatitis C virus belongs to the same class of viruses as Zika virus, I decided to compare the Zika virus RNA polymerase to the Hepatitis C RNA polymerase to see if the Zika virus protein might contain a likely binding site for Sofosbuvir.  

I used blastp to compare these two proteins and found a region with a significant E value (Fig. 9).  A an E value of 0.001 shows there is a 1 in 1000 chance that two proteins would match to this extent.  Of the amino acids, 25% are identical and 43% are either identical or conserved.

Figure 9.  A statistically significant alignment between a region of the Zika virus RNA polymerase query sequence and the RNA polymerase from Hepatitis C virus.

5.  Use Molecule World™ to identify amino acids in the drug-binding site.

To look at the binding site, I downloaded 4WTG in Molecule World from the MMDB (Molecular Modeling Database).  This structure contains the RNA polymerase, RNA in the process of being copied, and Sofosbuvir.  Sofosbuvir is a nucleotide analogue that blocks RNA synthesis by preventing RNA polymerase from adding new residues to the chain after the drug has been incorporated.  Looking at the structure shows me that Sofosbuvir has been incorporated into the RNA chain (Fig. 10).      

To identify the amino acids in the binding site, I touch the Show sequence button to open the sequence viewer and touch the name of the drug to select it.  Two manganese atoms appeared to be bound to the drug, so I selected those as well.

Figure 10.  Hepatitis C virus RNA polymerase bound to RNA and Sofosbuvir from structure 4WTG in Molecule World.  

The next step is to touch the Selection button and choose Select nearby.  This highlights everything within 5-6 angstroms of the drug.  In general, anything within 5 angstroms of a substrate would be considered to be located in the binding site.

Next, I open the Show/Hide menu and choose “All atoms in residue” to see the amino acid side chains.  Viewing the side chains shows me if any amino acid residues form bonds to the drug.  I can also hide everything that’s not selected to make the interaction easier to see.  Since some of the highlighted objects in my structure are RNA, I deselect those and hide them as well.  Figure 11 shows some of the amino acids in the binding site for sofosbuvir with the amino acid sequence below.

Figure 11.   Amino acids in the Sofosbuvir binding site are colored by residue in both the structure and the sequence.

6.  Determine if amino acids in the drug-binding site are present in both proteins.

Once we know which amino acids bind to the drug and which are located in the binding site, we can add these data to the sequence alignment to see if these same amino acids are present in both proteins.

A common way to show the similarity between two sequences is to align them with one sequence on top and the related sequence below.  Figure 12 shows this alignment, with the RNA polymerase sequence from Zika virus on top (the Query) and the Hepatitis C virus RNA polymerase is shown on the bottom.  When the two sequences match, the amino acid is shown in the middle.  If the amino acids are similar, but not identical, a + sign is shown.  This helps us to visualize the similarity between two protein sequences.

We copy the alignment from the BLAST results and paste it into a Word file to make it easy to annotate.  This often requires a bit of formatting work as well.

I also find it helpful to touch the Molecule icon and select the option to color the amino acids by residue.  Then, I scroll through the sequence to find highlighted amino acids.  When I find them, I highlight the corresponding amino acids in the aligned sequences.

In Figure 11, I can see that arginine 158 may form a salt bridge with oxygens in the nearby phosphate group.   This amino acid is in both proteins (Fig. 12).  Aspartic acid 225 is likely to form a hydrogen bond to the drug and Aspartic acid 220 forms a metal bond to one of the manganese atoms.

Figure 12.  Aligned sequences from the Zika virus RNA polymerase (Query) and the Hepatitis C virus RNA polymerase.  Amino acids in the Sofosbuvir binding site that are found in both proteins are highlighted. 

Based on the number of shared amino acids in the binding site and the types of interactions between shared amino acids, and the drug, I predicted that sofosbuvir would inhibit the Zika virus RNA polymerase.  This prediction was recently confirmed by in vivo experiments in cell culture and in mice (2, 3).

Conclusion

The method described in this article has been successfully used with students in three semesters.  Students have enjoyed it and commented that they like doing things that relate to real life.  This technique is suitable for student projects for a number of reasons.  Students can can apply this technique to multiple kinds of drug targets and carry out the steps involved using publicly available data and low-cost, user-friendly software.  In addition, this gives students an opportunity to learn and practice skills that are used in the biotech industry.  Furthermore, this same method was used in the two papers discussed earlier and predictions made through this method were shown to be correct when experiments were performed with cell cultures and mice.

References:

1.  Molecule World™ is a user-friendly iPad app for working with molecular structures and is available in the iTunes app store:  https://itunes.apple.com/us/app/molecule-world/id863565223?mt=8   For more information, see DigitalWorldBiology.com

2.  Sacramento, C. et. al.  The clinically approved antiviral drug sofosbuvir inhibits Zika virus replication.  Sci Rep. 2017 Jan 18;7:40920.

3.  Bullard-Feibelman, K.M., Govero, J., Zhu, Z., Salazar, V., Veselinovic, M., Diamond, M.S., Geiss, B.J., The FDA-approved drug sofosbuvir inhibits Zika virus infection, Antiviral Research  (2016), doi: 10.1016/j.antiviral.2016.11.023