Introduction: 3DM applied to the nuclear receptors


A: 3D-numbers make life easy

Login at 3dmjs.bio-prodict.nl with your 3DM account. If you don't have a 3DM account you can request one via the "Sign-up" button. To be able to do this course you need at least a course login. After you have requested an account you can request a course login by sending an email to Joosten@bio-prodict.nl

For this exercise you need either Pymol or Yasara installed with the 3DM plugin functional. If you don't have one of these or you are missing the 3DM functionalities, please consult the installation instructions. Before you start this exercise make sure you have the latest version of Yasara or Pymol installed.

After entering the login details you will see a link to the “Nuclear Receptors Ligand Binding Domain 2012” database. Open this 3DM system.

At the starting page of each 3DM database you see the 3DM data cycle. The icons in the circle represent links to the most important 3DM options. The same options can be found on the left side of the 3DM website.

Let’s say you found a paper that reports a mutation (Y401A) of the Pig vitamin D3 receptor to have an effect on specificity. Your research is actually about the human androgen receptor. You are wondering if a mutation at the same position in your protein would have the same effect. The first step would of course be searching the literature.

Note the overall structure of 3DM: The menu on the left, containing the icons for the 3DM tools, is available at all 3DM pages. In the menu the highlighted option tells you what type of data is currently displayed. Often the data is split up over several subpages. If you click on the 'Alignment statistics' item, for instance, the menu expands and now you can see 'Data plots' and 'Custom plot selection' options.

Finding articles describing structural equivalent residues in homologues sequences with 3DM is very easy. Open the alignment pages by clicking on the alignment icon. If you mouse over the green button above the alignment several options will appear to customize visualization of the alignment. The sequences displayed in the default view ("consensus alignments”) give a nice quick overview about what trends can be seen in the alignment that the evolutionary pressures have left behind in the alignment. Click on alignment position 183 in the overall consensus. Consult the “Mutations” tab. Here you see a list of mutations that were retrieved from the literature.

There are many papers describing mutations in the human androgen receptor and we still haven’t found a paper reporting a mutation that has effect on specificity. You might just read all these papers, but there is an easier way to do this using Yasara or Pymol:

B: Easy visualization of data in structure files

Structural biologists will confirm that the visualization of data in structures is time consuming. 3DM offers many different ways to visualize many different data types in any of the available structures. 3DM uses Yasara for this purpose. Let’s have a look:

In the other tabs (e.g. “Correlated mutation”, “conservation”) you can select different data types that can be visualized in the selected structures. Have a look at these tabs. By default the correlated mutations and the conserved residues are selected. Let’s leave it like this.

You will see a loading bar below the 'Visualize selection' buttons and after a few seconds you can download a Yasara or Pymol scene. Save the file and open it with Yasara or Pymol.

Note that the first time you probably need to manually select Yasara of Pymol to open the file as your computer doesn't know yet that .sce files of .pse files are scene files. You can tell your computer to always use Yasara to open .sce files or Pymol to open .pse files.

In Yasara the 3DM module is always active. If you are a Pymol user you first need to start the plugin. To start the 3DM plugin use "plugin -> "initialize plugin system". Then "plugin -> "legacy plugins" -> "3DM". A new window should start. Use your 3DM credentials to login. At the left top of this window the 3DM menu should appear.

What are correlated mutations? In large superfamily alignments correlated mutations (also called co-evolution of residues) are almost always functionally related. Residues that mutate simultaneously often share a function. These can be different functions. Sometimes correlated mutations that are related to enzyme activity, or enantioselectivity, or co-factor binding, but most of the time they are related to changes in specificity. Being important for a certain function created an evolutionary pressure that resulted in restricted mutation rates. If a function changes during evolution (e.g. the specificity of the enzyme changed) than the residues involved in this function need to mutate to facilitate this change (e.g. the binding of a new substrate).

Take home message for protein engineers: Correlated mutations can often be found surrounding the ligand/substrate pocket. When they do they often are correlated with specificity changes and are therefore specificity hotspots. If you want to change specificity make a mutant library at these positions. If your library becomes too large, take only residues that are coming in the alignment.

3DM is connected to Yasara and Pymol. The 3DM menu contains different options to select different data types to visualize in protein structures. The "literature hotspots" option is a really powerful tool. Use "Specificity" as keyword in this option and click OK (login with your 3DM login data if asked). It will select positions for which mutations have been reported in literature to have an effect on specificity.

One article that according to 3DM contains mutation data reporting specificity changes at position 183 is: "Broadened ligand responsiveness of androgen receptor mutants obtained by random amino acid substitution of H874 and mutation hot spot T877 in prostate cancer". Open the paper here. If you read the first sentence of the abstract you can see that the mutation indeed has an effect on specificity. The title reveals that T877 is a prostate cancer hotspot. You wonder if there are mutations at other positions known to cause prostate cancer. How would you normally solve this problem (don't actually do it)?

Look in Yasara or Pymol to see if position 183 makes a contact with the ligand. Now you would like to know if position 183 is also a hotspot where ligands bind. The best way to find out is to open all 789 available structure files and count the number of contacts this position makes with co-crystalized ligands, right?

Nuclear receptors can either be activated or inhibited by small molecules (3DM calls these ligands). Activating compounds are called agonists and inhibiting compounds are called antagonists. Say you would like to know where activating compounds bind, where inhibiting compounds bind, and if there is a difference.

C: The search options and subset generation.

In 3DM many different search options are available to select a subset of sequences (consult the different tabs of the search module and see if you understand all the search options). The resulting sequences of a search can be saved in a subset with the subset window. To display the subset window click on the 'SUBSETS' button on the right hand-side of the page header. With this subset window a mini 3DM can be generated for a subset. All 3DM functionalities will work in all 3DM webpages and even in Yasara if you switch to using subset data. How to generate a subset and use the subset window will be discussed in more detail below.

Nuclear receptors can be inhibited or activated by ligands (small organic molecules). We want to investigate the difference between inhibiting-, and activating ligands. The differences between these two might reveal the mechanism of inhibition/activation. To investigate the different binding mode two subsets need to be generated. One containing only PDB files that have an activating compound in the ligand binding pocket and a second containing PDB files with an inhibitor. Comparing the positions where ligands bind in the one group to the other could reveal a different binding mode.

Because the generation of the mini 3DM systems for these two subsets takes time, we pre-generated an activator and inhibitor subset. To generate the subsets we used the 3DM search option (structures subpage). Because some PDB files contain both an agonist and an antagonist we have to use a trick, which nicely indicates how to use the subset window. Try the searches yourself to see if you find the same number of sequences. To make the antagonists subset we first used the keyword “antagonist”. By opening the subset generation window a plus sign will appear next to the search result. The antagonist search results in 90 structures. Note that you can select and deselect sequences manually with the check boxes in front of each sequence.

Next we ran a search for keyword "agonist" - with the 'Match whole words' option on (to exclude all "antagonists") and subtracted the result by using the – sign. This deletes the overlap of the two searches. So the PDB files that contain both an agonist and an antagonist will be deleted from the set of antagonists that was in our subset window. Next we ran a search for keyword "agonists" and repeat the subtraction step from the previous search. Now we have 62 structures left in the subset. We saved the subset, named this subset “inhibitors for course” and generated a mini 3DM for it (option: save and regenerate). Note that the option “save” will only save the sequence in the subset window for later use without making a mini 3DM for the subset.

The reverse search was done to generate the agonist subset. Searching for “agonist” and "antagonists" gives 372 structures. Deleting the antagonist’s results in a subset of 344 structures. This subset is called “activators for course”.

From the 'Subset' dropdown menu choose the 'activators for course' subset - now all data will include only the sequences that are part of this subset.

To show the effects of subsets first go to the "Alignment statistics" page by clicking the Alignment statistics icon. The histograms plot different data in relation to the 3D numbers (x-axes). Have a look what kind of data is plotted in these histograms by scrolling down. You can switch between subsets using the middle menu item at the top of 3DM. Using this selection menu you can see the "activators for course" and "inhibitors for course" subsets. The data in these histograms depends on the subset that is selected.

Go to the second option (custom plot selection) of the alignment statistics. Here you can select different data types and subsets. In the left box select "activators for course" and "inhibitors for course", in the right box select ligand contacts and click the customize button. Evaluate the resulting histogram.

In Yasara or Pymol load the ligand from 1ERRA using the "3DM" -> "Structures" -> "load structure from 3DM" option.

D: Designing drugs

In 3DM use "Search" -> "Structures" to find a human androgen receptor structure with its natural ligand dihydrotestosterone bound in the ligand binding pocket.

Use the "visualize" option of 3DM to visualize in Yasara or Pymol both the protein chain and the ligand from the first hit from the previous search (1I37A). An easy way to do this is to click on 1I37A in the search result which links to the "protein detail page" of 1I37A. You can use the visualize icon to go directly to the "visualize" module of 3DM where this PDB is then selected. Just click on "visualize selection in Yasara or Pymol" and you are ready (note that on some computers you need to close Yasara first before opening a new one). Once you have this structure loaded in Yasara or Pymol use the 3DM->Literature Hotspot-> Specificity option to show the hotspots for specificity.

Take home message: With 3DM you try to find correlations between different data types and learn from these correlations underlying biological meaning. You have seen above that if correlated mutations are close to the ligand/substrate they are likely important for specificity. You have combined two different data types: e.g. correlated mutation data with structural location. They overlap and therefore you have learned something. If you want to answer a biological question with 3DM you should do these two general steps: 1 Which subsets do I need to generate and 2. What data do I need to compare. Sometimes you don’t need to make subsets and just comparing two data types is already sufficient.

If specificity changes in the androgen receptor T877A mutation are indeed the underlying cause for prostate cancer it is not unlikely that the mutant can now be activated by a different ligand present in humans. It is also not unlikely that it can now accept another nuclear receptor ligand because of the similarity of the ligands that are recognized by the members in this protein family. A scientist working in this area would realize that progesterone is very similar to dihydrotestosterone. So it seems a good hypothesis that a T877A mutated human androgen receptor is overly active because this mutant has changed specificity and can now be activated by progesterone thereby causing cancer, but we need additional proof for this idea.

Use the 3DM->"structures" option in Yasara or Pymol to load the ligand of structure 1A28A (progesterone) and compare this ligand with dihydrotestosterone that should already be loaded in Yasara. If it is not loaded also load the ligand of 1I37A. For easy comparison change the ligands to stick visualization.

After this experiment it seems very likely that progesterone is capable of activating the T877A androgen receptor mutant. Most drugs that are used to treat prostate cancer are general androgen receptor inhibitors but are not specific for treating prostate cancer caused by the T877A mutation. Some drugs are available, such as hydroxyflutamide (also called S-1), and some derivative compounds, but these are not very effective in patents with the T877A mutations. It would therefore be helpful to have an alternative drug that specifically targets the T877A androgen receptor.

We should therefore design a compound that could compete with the binding of progesterone in the T877A mutant but does not activate this mutated androgen receptor. We would have to make a compound that is similar to progesterone (simply because we now know progesterone can bind) but has a larger group around position 183 because we know this is how inhibitors work as we have learned during this exercise (do you agree with this hypothesis?). Luckily, using simply chemistry, it is easy to add groups at the ester group (O=C-C-R) of progesterone that is near position 183.

If this idea would succeed we have designed a drug for treating prostate cancer in patients with the T877A mutation. Woudn’t that be great? Unfortunately, we are not the first to make this type of compounds for treating prostate cancer. Several progesterone derivatives have been published as androgen T877A inhibitors. But the above process nicely shows how combining data from a 3DM system can be used as guidance in the early phases of drug design.