Exercise:Malaria Vaccine

Exercise written by: Thomas Salhøj Rask and Henrik Nielsen — translated, revised and updated to BepiPred 2.0 by Rasmus Wernersson and Henrik Nielsen.

The aim of this exercise is to apply the bioinformatics methods and knowledge you have learned so far to a real-world biological challenge: the development of a malaria vaccine. You will explore the first steps of vaccine design by identifying protein regions from the malaria parasite that may be recognized by the human immune system and could therefore serve as potential vaccine candidates.Along the way, you will build on your existing knowledge and be introduced to new methods, particularly the prediction of B-cell epitopes (regions of proteins that may be recognized by antibodies).

The exercise is organized into the following steps:

1. Finding the infectious agent causing malaria.
2. Identification of membrane bound proteins (potential vaccine targets)
3. Analysis of membrane protein domain structure
4. Prediction of B-cell epitopes from membrane proteins
5. Modelling / visualization of predicted epitopes in the 3D structure of a protein domain.

 

Finding the infectious agent causing malaria

Malaria is a disease, but which type of organism/s causes malaria? Bacteria, protozoa (single cell eukaryote), worm or virus?

Investigate this by looking up malaria in the taxonomy database. NCBI Taxonomy: https://www.ncbi.nlm.nih.gov/datasets/taxonomy/ You will see that the menu displays several species options. Click on some of them to understand what is their relationship to the malaria disease. You can also get help from your AI sparring partner to understand how the different organisms play a role in the disease and identify what is the infectious agent.

Question 1a)

Identify the following taxonomical levels for the malaria-causing organism:

• Genus
• Phylum
• Domain

Question 1b) How "close" in taxonomy space are the malaria-causing organism to the following organisms. Find the common ancestor, upper level taxonomical group, that ties them together. Hint: as an alternative to manually comparing the taxonomy-strings (the "lineage"), you can use the NCBI Taxonomy Common Tree tool to automate the comparison.

• Homo sapiens
• Babesia microti    (It is a parasite, transmitted to humans by ticks (danish: "Skovflåt") that can lead to the disease babesiosis, where the red blood cells (erythrocytes) are invaded as in malaria, and which will lead to anemia ("blood loss", in this case lack of oxygen carrying capacity in the blood).

Finally, read more about malaria and the complicated life cycle of the malaria parasite here: CDC - DPDx Malaria.

Question 1c) Report the names of the five species of parasites causing malaria in humans, and use the NCBI Genome (https://www.ncbi.nlm.nih.gov/datasets/genome/) database to investigate which of them (if any) have had their genomes sequenced.

 

Identification of membrane proteins (potential vaccine targets)

Malaria caused by Plasmodium falciparum (Pf) is by far the most lethal malaria variant. This parasite species is responsible for 80%-90% of the ~1 million annual deaths due to malaria. It will therefore be a natural starting point to develop a vaccine against this type of malaria.

When the Pf genome was initially sequenced in the 1990s, it was based on Pf cells isolated from the blood of a Dutch malaria patient, who picked up the disease while traveling. Unfortunately, it was not recorded exactly where the patient had been. This isolate is named 3D7 and is the most studied malaria strain to this day (even though it's not known from where in the world it originates).

Task: Locate the entry for Pf 3D7 in NCBIs taxonomy browser. If you scroll down the page, you can see a section with the heading "Gene" including a pie chart.

Question 2a) How many protein-coding genes does Pf 3D7 have?

Malaria takes place in different stages within the human host (see figure below), and this is important to take into account when designing a malaria vaccine. The disease development can be divided into two phases: 1) The liver-stage and 2) the blood-stage. The liver-stage is defined by sporozoites injected by the malaria mosquito, which travel to the liver and invade hepatocytes (liver cells). The blood-stage is the second stage and is reached when merozoites developed within the hepatocytes are released into the blood stream, where they invade erythrocytes (red blood cells). In both stages the malaria parasite hides from the human immune system by staying inside native human cells.

Much of the effort towards developing malaria vaccines so far has been focused on surface exposed (cell-membrane) proteins from the sporozoites and merozoites as well as non-human proteins on the surface of infected hepatocytes and erythrocytes.

Searching UniProt

We will now explore how protein annotations in UniProtKB can help us identify potential malaria vaccine targets.

When designing a vaccine, it is important to choose proteins that can be recognized by the immune system. For antibodies to recognize a protein directly, at least part of that protein must be accessible on the surface of the parasite or an infected host cell.

Building on what we learned in the previous section, we are therefore interested in proteins that:

- Are produced by the malaria parasite (Plasmodium falciparum).

- Are present on the surface of "sporozoites", "merozoites", or "infected host cells".

To identify potential vaccine targets, we will explore several ways of searching UniProtKB:

Searching by organism and strain: Identify proteins from Plasmodium falciparum and compare the information available for different strains.

Searching by subcellular location: Use protein annotations to find proteins associated with the host cell membrane, where they may be accessible to antibodies.

Searching by protein name: Look for proteins whose names suggest that they are associated with erythrocytes (red blood cells) or their membranes, even if their subcellular location is not annotated.

Searching for known 3D structures: Check whether any of the identified proteins have experimentally determined structures available in the Protein Data Bank (PDB).

These approaches can help us narrow down the list of proteins worth investigating further. However, a protein's name or annotation does not necessarily confirm that it is exposed on the cell surface or suitable as a vaccine target.

For this exercise, we will use the same UniProt search interface introduced earlier in the course. We recommend keeping the original UniProt exercise open in another browser tab in case you need to review how to construct searches. UniProt Exercise manual

Note: Important: For each question, you must report both the number of hits and the search query you used.

Searching for proteins from Plasmodium falciparum

2b) Go to UniProt. Investigate how many Plasmodium falciparum (Pf) proteins there are in total in UniProtKB (i.e. proteins from all Pf strains, not only from 3D7). How many of these are from Swiss-Prot and how many from TrEMBL?

2c) Now try to see how many of the hits from the previous question are from the strain (isolate) 3D7. Is the number approximately equal to the number you got in question 2a)? How many of these are from Swiss-Prot and how many from TrEMBL?

Searching by subcellular location

Now, we shall investigate whether we can use the annotations of subcellular location in UniProt. Note: We go back to working with all strains of Pf, not exclusively 3D7.

2d) First, check how many Pf proteins have a "Subcellular location [CC]" comment at all (Tip: choose Subcellular location > Subcellular location [CC] > Subcellular location term in the menu and enter a * in the field). How many from each part of the database? (Note that the ratio between TrEMBL and Swiss-Prot numbers changes considerably relative to question 2b) — Swiss-Prot entries on average contain many more annotations than TrEMBL entries).

Now, let us focus on the life stage of the parasite where it is located inside an erythrocyte (a red blood cell), and thereby focus on the vaccine targets that are in the plasma membrane of the host cell.

2e) How many of the hits have the location "host cell membrane"?

These proteins could be very interesting as vaccine targets. However, the experimental researchers from your organization report that these have already been tried and do not work in practice, so they ask you to find other examples. We therefore try another approach: If the information we are looking for is not part of the "Subcellular location" annotation, it might be a part of the description (the protein name). Tip: you can always discard a search term in the Advanced interface by clicking the Remove button.

Searching by protein name

2f) How many Pf proteins contain erythrocyte in their Protein Name [DE] field? How many of these are from Swiss-Prot (reviewed)?

2g) How many of these erythrocyte proteins also have membrane in their name?

Some of the hits you find in this way are very short (you can try to sort them by length by clicking the Length heading). These short proteins might be fragments.

2h) How many of the hits are complete (not annotated as fragments)? (Tip: see question 3.7 in the UniProt exercise).

Searching for known 3D structures

2i) Starting with the results from Question 2h, how many of these proteins have a determined 3D structure? In other words: Do any proteins from the previous search have a cross-reference to the database PDB? (Tip: you should look for Cross-references in the menu, and again place a * in the field). If yes, what are their names and accession codes?

As a last step in this part of the exercise, you should save all sequences from the last search in FASTA format. This is most easily done by clicking Download above the results list and choosing FASTA (canonical). You should download them in uncompressed format (remember to choose No under Compressed) and then open them in a text editor; we will need them later in the exercise.

 

Analysis of membrane protein domain structure

Background: Why are PfEMP1 proteins interesting vaccine targets?

In the previous section, we identified a group of proteins called PfEMP1 (Plasmodium falciparum Erythrocyte Membrane Protein 1).

These proteins are produced by the malaria parasite and transported to the surface of infected red blood cells (erythrocytes). In the figure, the red/orange structures represent PfEMP1 proteins.

PfEMP1 proteins help infected red blood cells attach to receptors on endothelial cells, which line the inside of blood vessels. This attachment allows infected red blood cells to remain in small blood vessels rather than circulating through the spleen, where abnormal red blood cells would normally be removed.

This mechanism helps the malaria parasite evade the immune system and contributes to the development of severe malaria. If a vaccine could stimulate the production of antibodies that bind to PfEMP1 and prevent this attachment, it could help the body clear infected red blood cells and reduce disease severity.

In this part of the exercise, we will investigate the domain structure of PfEMP1 proteins to better understand which regions are exposed on the cell surface and may be accessible to antibodies.

Exploring protein families and domains with InterPro

Return to the protein entries you identified at the end of Section 2. Select only those originating from P. falciparum strain 3D7. You should find three proteins. If you find a different number, revisit your search from Section 2. :Backup: If you did not find the proteins in Question 2h, you can continue this part of the exercise using the protein with UniProt accession: Q8I639.

Open each of these three protein entries in UniProtKB. Under Family & Domains, find the section Family and domain databases. Here, you will find links to databases that classify proteins into families and identify domains within their sequences. We will use InterPro, which combines information from several protein family and domain databases, including Pfam.

Open the link View protein in InterPro in a new browser tab.

In InterPro, locate the graphical overview under Entry matches to this protein. The colored bars represent regions of the protein that match known protein "families" or "domains". Hover over a bar to see its name and position in the protein sequence.

InterPro uses identifiers beginning with IPR, while its member databases have their own identifiers. For example, Pfam identifiers begin with PF. You may therefore encounter both an InterPro identifier and a Pfam identifier for the same domain.

What are protein families and domains?

According to the InterPro FAQ:

• A protein Domain is a distinct region of a protein that can have a particular structure or function. Similar domains may occur in different proteins, sometimes more than once within the same protein. Common examples of protein domains are the PH domain, Immunoglobulin domain or the classical C2H2 zinc finger.
• A protein family is a group of evolutionarily related proteins that share similarities in their sequences, structures, or functions.

The distinction between protein families and domains is not always clear-cut.

3a) Identifying repeated domains

In InterPro, you should find a domain called the Duffy binding domain, which occurs several times in each of the three PfEMP1 proteins.

• What are its InterPro and Pfam identifiers?
• How many times does the domain occur in each of the three proteins?

Click the domain identifiers to read more about the Duffy binding domain.

3b) Identifying membrane-spanning regions

Proteins located in cell membranes often contain transmembrane segments, which are regions of the protein that pass through the lipid bilayer.

In InterPro, change the *Feature Display Mode from Summary to Full and scroll down to find the annotated transmembrane segment.

For each of the three proteins:

• Which amino acid positions correspond to the transmembrane segment?
• Which part of the protein (N-terminal or C-terminal) is located **inside** the infected erythrocyte, and which part is located **outside**?
• Are the Duffy binding domains located on the intracellular or extracellular side of the membrane?

Look (in UniProt) at the PDB cross-references under 3D structure databases (under Structure). Focus on X-ray structures only. Compare the coordinates (positions) for the structures to the coordinates for the domains denoted in Pfam.

3c) Comparing domains with experimentally determined 3D structures

• Which positions are structurally determined by X-ray in each of the three proteins?
• If you number the occurrences of the known Duffy binding domain from 3a (1, 2, 3, and so on, starting from the N-terminus), which domains are covered by the structurally determined region(s) in each of the three proteins?

Now read what is said about the function and location of our proteins according to Gene Ontology (GO - Molecular function, GO - Biological process and GO - Cellular component) in UniProt.

3d) Using Gene Ontology to evaluate vaccine targets

Return to UniProtKB and find the Gene Ontology (GO) annotations for each protein. GO annotations describe proteins using three categories:

Molecular function: What the protein does at the molecular level.

Biological process: Which biological activities or processes the protein is involved in.

Cellular component: Where the protein is located.

• Based on the GO annotations, do these proteins appear to be suitable candidates for further investigation as malaria vaccine targets?
• Give at least three examples of GO annotations that support your answer.

 

Prediction of B-cell epitopes in a membrane protein

Background: VAR2CSA and malaria during pregnancy

In the previous section, we investigated the domain structure of PfEMP1 proteins. We will now focus on one particular PfEMP1 protein, VAR2CSA (UniProt accession Q8I639), which plays an important role in pregnancy-associated malaria (PAM).

One of the challenges in developing malaria vaccines is that Plasmodium falciparum has approximately 60 genes encoding different PfEMP1 proteins. By switching between these proteins, the parasite can evade antibodies produced by the immune system. VAR2CSA is particularly interesting because it is the main PfEMP1 variant associated with the accumulation of infected red blood cells in the placenta. This makes it a promising target for vaccines specifically aimed at preventing pregnancy-associated malaria.

For our bioinformatics analysis, we will focus on **the Duffy binding domain of VAR2CSA for which an experimentally determined 3D structure is available, as identified in the previous section.

Epitope prediction

A potential approach to vaccine development is to identify epitopes: regions of a protein that can be recognized by the immune system.

In this exercise, we will focus on B-cell epitopes, which are regions of an antigen that can be recognized by antibodies. There are also T-cell epitopes, but we will not investigate those here.

We will use 'BepiPred 2.0, a bioinformatics tool developed at DTU, to predict which regions of the VAR2CSA protein might be B-cell epitopes.

Important Note: Before running the B-cell epitope prediction it will be needed to create a hotspot with your own MobilePhone as all the submissions from DTU will be runnning on a single queue, and not starting until the previous has finished.

https://services.healthtech.dtu.dk/services/BepiPred-2.0/

Step 1: Retrieve the amino acid sequence Before running BepiPred 2.0, we need the amino acid sequence of the structure-determined Duffy binding domain, rather than the complete VAR2CSA protein.

We will obtain this sequence from the Protein Data Bank (PDB), using the link provided in UniProt.

• Open the VAR2CSA entry (Q8I639) in UniProt.
• Scroll down to the Structure section and find the PDB cross-references.
• Locate the X-ray structure of the Duffy binding domain identified in 3c).
• Open the corresponding PDB link in a new browser tab.
• On the PDB page, select Display Files → FASTA Sequence to view the amino acid sequence.
• You can also download the sequence using Download Files → FASTA Sequence.

Make sure you obtain the sequence corresponding to the domain of interest, not the full-length VAR2CSA protein.

4a) What is the name (title) of the PDB entry, and which experimental method was used to determine its structure: X-ray crystallography, electron microscopy (EM), or NMR spectroscopy?

4b) Report the FASTA sequence of the structure-determined Duffy binding domain in VAR2CSA.

Step 2: Identify the sequence coordinates

The FASTA sequence you have extracted represents only part of the full VAR2CSA protein. Notice that amino acid positions in the extracted FASTA sequence start at position 1. However, this domain is located further along the original full-length protein sequence. We therefore need to keep track of both numbering systems:

4c) Using the information in the UniProt entry, report:

• The start and end positions of the structure-determined domain in the original full-length VAR2CSA sequence.
• The position in the full-length VAR2CSA sequence that corresponds to position 1 of your extracted FASTA sequence.

Keep these values for later, as you will need them to map the predicted epitopes onto the 3D structure.

Step 3: Run the B-cell epitope prediction

We are now ready to predict potential B-cell epitopes using BepiPred 2.0.

Open the [BepiPred 2.0 prediction server](https://services.healthtech.dtu.dk/services/BepiPred-2.0/).

• Paste your FASTA sequence from Question 4b into the input field, or upload the FASTA file.
• Submit the sequence for prediction. Use only the extracted domain sequence, not the full-length VAR2CSA protein.
• Once the prediction is complete, open the results page.
• Set the Epitope Threshold to 0.53.

The threshold determines which amino acid positions are classified as potential epitope residues. Consecutive residues that exceed the threshold form predicted epitope regions.

We will focus on predicted epitopes that are at least 5 amino acids long. Record the start and end positions of all such predicted epitopes.

Hint: With a threshold of 0.53, you should identify 6 epitopes containing at least 5 amino acids. The second epitope begins at L23 (leucine at position 23).

If the prediction takes more than 5 minutes. You can use the pre-calculated BepiPred 2.0 results instead.

Step 4: Convert the epitope coordinates

The positions reported by BepiPred 2.0 refer to the extracted domain sequence, not the original full-length VAR2CSA protein.

In the next section, we will visualize the predicted epitopes on the 3D protein structure using PyMOL. For this, we need to determine where each epitope is located in the original VAR2CSA sequence.

Use the information from 4c to convert the predicted epitope positions into full-protein coordinates.

4d) Create a table containing the following information for each of the predicted epitopes (at least 5 amino acids long):

• Start and end positions in the extracted domain sequence.
• Epitope length (number of amino acids).
• Start and end positions in the original full-length VAR2CSA sequence.

Visualization of epitopes

Finally, we will use PyMOL to visualize the predicted B-cell epitopes in the 3D structure of the VAR2CSA Duffy binding domain. We want to see whether the epitopes are exposed on the protein surface, where antibodies could potentially recognize them.

Checking the PDB structure

Return to the PDB entry from the previous section and open the Sequence tab. Note that the residue numbering in this structure follows the original full-length UniProt sequence.

Question 5a)

• Which sequence intervals are missing from the 3D structure? (Hint: Look for the UNMODELED feature.)
• Do any of these missing regions overlap with our predicted epitopes?

Visualizing epitopes in PyMOL

Load the PDB structure into PyMOL (using fetch or by opening the downloaded file). Refer to the earlier PyMOL exercise if needed.

Start with a grey cartoon representation:

color gray80
hide all
show cartoon

Select one protein chain

The structure contains two copies of the protein (chains A and B). We only need one for our analysis.

To keep chain A, remove chain B using:

remove chain B

Highlight the epitopes

Create named selections for the six epitopes identified in Question 4d, using the residue positions in the original full-length protein.

For example:

select epitope_1, ka and resi START-END
color red, epitope_1

Replace START-END with the correct positions, and use kb instead of ka if you selected chain B.

Give each epitope a different colour.

Hint: Turn on the Sequence Viewer to see your selections and colours directly in the sequence.

3. Explore the protein surface

Try switching between these representations:

• Show as → Surface to examine surface accessibility.
• Show as → Cartoon, followed by Show → Mesh, to see the protein fold together with its surface.

Rotate the structure to examine the epitopes from different angles.

Question 5b)

Create one or more figures for your report showing:

• The positions of the predicted epitopes, each in a different colour.
• A legend explaining the colours.
• Which epitopes are partially or completely missing from the structure.
• Whether the remaining epitopes are exposed on the protein surface.

Remember to indicate which chain (A or B) you used.

Epilogue

Now all that remains is to ship off the sequences of the surface accessible epitopes to the lab, to start the long process of constructing an expression vector with the gene fragments, with the right linker sequences, getting it expressed in a production host, follow up with animal testing and phase 1, 2 and 3 clinical trials, and the vaccine should be ready for the market.