I’ve been working on my thesis for the last few months, squirreled away in libraries and coffee shops, but now I’ve submitted and waiting to defend, I’m happy to share what’s happened in the meantime! A research paper I’ve been working on for a long time has been accepted, and published in the journal Structure. You can find it online, here. This paper makes up a bulk of the work in my PhD thesis, containing 10 protein crystal structures, and I’m glad to have it finally available to the world!
In the paper I describe the structure of a protein that causes antibiotic resistance. This protein makes a bacterium resistant to a class of antibiotics called aminoglycosides. They are one of the oldest classes of antibiotics and include some well-known compounds such as streptomycin, kanamycin, tobramycin, and gentamicin. They are effective antibiotics used against a broad variety of bacteria, and resistance factors that make them ineffective are a serious problem in the treatment of infections.
The protein that I work with generates aminoglycoside resistance. It chemically alters the antibiotics, turning them into inactive byproducts, making any bacterium with the protein resistant to aminoglycoside antibiotics. The protein acts as a “resistance factor” – bacteria that carry the gene for this protein use the protein to deactivate the antibiotic. They can easily break it down and go about their bacteria business instead of being killed.
This protein is called APH(2”)-Ia (more on that name later). It inactivates several different aminoglycosides. To learn how it carries out this transformation, I looked at the structure of the molecule and how it changes when it interacts with the antibiotic. To understand why the structure is important, let’s talk about what this type of molecule really is.
Enzymes: biological molecules that make chemistry happen
Proteins that carry out chemical reactions are also called enzymes. They allow a chemical change to happen that normally happens at extremely low or nonexistent rates. The enzymes that act on aminoglycoside antibiotics are collectively referred to as aminoglycoside-modifying enzymes. They deactivate antibiotics by transferring part of a common, metabolic molecule to the antibiotic. This makes the antibiotic worthless, and lets the bacterium survive the toxic effects of the compound.
An enzyme drives a reaction by forcing these chemicals together. They do this by specifically binding to the molecules, in a mechanism sometimes referred to as a lock-and-key interaction. An enzyme is specific for the molecules that it acts on, like a lock only opens for a specific key. The keys of an aminoglycoside-modifying enzyme are the antibiotic and a cellular molecule that donates a chemical group to the antibiotic. In APH(2”)-Ia, that cellular molecule is guanosine triphosphate, or GTP. The enzyme binds the antibiotic and GTP tightly, and undergoes changes in structure to drive a chemical change between these molecules. This all happens in the most important part of the protein, the active site.
The active site is the most important part of an enzyme. This part of the protein is typically a deep pocket that the rest of the molecule wraps around, where the enzyme holds and manipulates the molecules, and where chemical bonds are broken and formed. The enzyme separates these molecules from water and other compounds, and in this isolated state, the enzyme drives the chemical change to occur.
The active site of any enzyme is typically extremely sensitive to the shape and properties of the molecules it acts upon. Evolution has driven enzymes to develop a high degree of specificity for these molecules, known as substrates. An enzyme typically has only a few different substrates that it acts upon. The combination of the specificity of chemicals that an enzyme acts on and the reaction it carries out gives it its name, in this case aminoglycoside phosphotransferase.
I try to avoid saying the name of the protein I work on unless I want to sound important.
APH(2”)-Ia stands for aminoglycoside (2”)-O-phosphototransferase type Ia. It’s usually preferable to just say “the enzyme” and that’s what I’ll mostly do here. As the “type Ia” might suggest, the enzyme is part of a larger group of enzymes that all carry out a similar reaction. They use magnesium atoms, held tightly in the active site, to move a phosphate group, PO43-, from one molecule to another. In this enzyme, it moves a phosphate from GTP to the aminoglycoside antibiotic. GTP is a nucleoside triphosphate molecule, the cellular phosphate source in this reaction. You might be familiar with a similar molecule, ATP, the “energy currency of the cell”.
APH(2”)-Ia is somewhat unusual because most similar enzymes use ATP, but this enzyme uses GTP. Researchers in my lab and in other groups have studied the relationship between these proteins and GTP, and there’s still some interesting unsolved mysteries about the use of GTP in these proteins. However, for this work, I focused on the part of the molecule that is the same between ATP and GTP, the triphosphate.
These enzymes that use nucleoside triphosphates as phosphate donor have a special name: kinases. We know quite a lot about kinases. Their importance for cell biology was discovered in the 1970s-1990s and researchers learned that they are critically important regulators of cellular metabolism, cell division, and many other processes. They add phosphate groups to proteins that generate molecular communication networks in the cell. In mammalian cells kinases are typically involved in important cellular decisions, and because many of these decisions impact how a cell grows and divides, many of these kinases are involved in cancer. When it was found that aminoglycoside phosphotransferases were kinase enzymes, there was already a large amount of research to compare to on similar enzymes. However, comparisons to other enzymes only gets you part of the way. To really learn how any molecule works, you have to look at it directly.
How do you look at something a few nanometers in size?
A mantra in the molecular sciences is that structure dictates function. I argue it needs a little updating, that structure and dynamics dictate function (more here and here), but you need to have a structure before you can study how it moves. Determining the structure is where we start.
Using the techniques of structural biology, we can get a direct look at the molecules that carry out biological functions. Techniques like nuclear magnetic resonance, and electron microscopy can provide excellent structural information about biological macromolecules, but the technique I used for these experiments was X-ray crystallography. Matt Kimber, the professor who taught my undergraduate structural biology course referred to crystallography as a “one-trick pony, but it’s a damn good trick”. Well, I’ve ridden that pony right to the end of my degree.
To determine a structure by crystallography, you purify the protein that you’re interested in, and run an array of experiments in parallel to try to find conditions under which it will form crystals. Not every protein can crystallize, and not every crystal gives you good data, which makes protein crystallography an intimidating technique. The perceived risk of protein crystallography experiments drives many in the molecular sciences to treat protein crystallographers with a sort of reverence. I don’t know that that reverence is particularly well-placed, but I’ll take it all the same.
If you are able to make protein crystals of sufficient size and quality, then you can collect X-ray diffraction data with the crystals. Several companies sell instruments to collect this diffraction data, and there are dedicated facilities that conduct these experiments using high-intensity X-rays from accelerated electrons. The instrumentation for these experiments continues to dramatically improve, allowing us to get more and better data from our protein crystals all the time. The job of the X-ray crystallographer is much easier these days than it used to be.
Using a home-based or synchrotron source, a beam of X-rays are directed at the protein crystal. X-rays interact with the electrons of a molecule. Because of the physics of diffraction of rays from a crystal, the result is a pattern of diffraction from the x-rays that can be recorded. By measuring these diffraction spots, we can apply physical rules about how diffraction works to interpret the distribution of electrons within the crystal of proteins. From the intensity of diffracted X-ray spots, we can reconstruct the shape of the electron density in our molecule of interest.
At this point, the job still isn’t done. In some cases, you have to solve what we call the phase problem, although in this case it wasn’t too much trouble so I’ll jump past it. However, there is still a considerable amount of analysis required to interpret what the electron density means, and what the shape of molecules that fill this electron density truly are. It’s the crystallographer’s job to build a molecular model that recapitulates the observed electron density as closely as possible. The point at which you consider the model “done” is an ongoing struggle, similar to that experienced by artists and writers, where there is always another brushstroke, sentence, or water molecule that could improve the final product, but then at some point you stop and call your model “finished” and interpret what it says about the chemistry of the molecule you’re studying.
In the case of APH(2”)-Ia, a lot of the challenge for me was making the models as good as possible, and after a long struggle, they were of sufficient quality that I could use them to gain some interesting insights about how the protein works and suggested a new feature of an antibiotic resistance enzyme.
Determining the structure of APH(2”)-Ia
When I started working with APH(2”)-Ia, others had already determined the structures of three related molecules: APH(2”)-IIa, -IIIa, and -IVa. These structures give us the shape of the enzymes and some of their interactions with their substrates, but a key factor always missing was a well-defined active site set of the triphosphate ligand.
Without the triphosphate in the enzyme, we can’t get any sense of how the enzyme interacts with that molecule. And if we can’t see it, we can’t predict how it works, or understand how to affect it.
Building the shape of the first versions of APH(2”)-Ia weren’t as hard as I’d expected. I had a few nights up late, excited to carry out my next rounds of model-building and improving the data, and was proud to build models with some pretty excellent statistics for the quality of X-ray diffraction data I was working with. The challenge came when I had to start interpreting what was in the active site of the enzyme.
Probing the active site of an antibiotic resistance enzyme
Remember how I talked about how the enzyme takes a phosphate group from GTP and moves it to the antibiotic? Well, in the first structures I looked at, that really didn’t seem possible. I used a GTP-like analogue molecule called GMPPNP to make the protein crystals, and in the structures, the phosphate groups of the GTP analogue were stuck in a position where they couldn’t react with the antibiotic. I named this the stabilized conformation, because it seems to be sitting in a position where it can’t react with anything.
There were several similar enzymes I could look to for comparison, and none of them show this stabilized triphosphate form. They have a different, activating conformation which directs it toward the other substrate. I was able to make the molecule adopt the activated conformation, but I had to break the protein by removing an important amino acid to make it let go of the stabilized phosphate.
So why didn’t the protein in my structures normally activate the triphosphate?
The answer came when I added antibiotic molecules to the crystals. Introducing the antibiotic after the crystals were grown could let us track the changes driven by the introduction of the antibiotic. I had a good idea from other aminoglycoside kinases that the antibiotic would bind in the same position, and initially was just trying to confirm that the antibiotic bound in the same position. As a fortuitous observation, the addition of aminoglycoside substrate drove changes in the shape of the protein, as well as the GMPPNP molecule. The antibiotic activated the triphosphate by binding to the enzyme.
The flip between these states gives us a clear way to understand how the enzyme could turn itself “on” when it encounters an antibiotic.
A motion sensor switch for antibiotics
So, we see a new shape of the triphosphate group that can’t modify the antibiotic. It is held in an awkward, non-reactive form in the back of the enzyme active site. When an antibiotic comes in, the enzyme shifts its shape to respond. In the process of making this change, the GTP triphosphate changes and becomes active. This catalytic switch helps the enzyme keep the GTP inactive until the antibiotic is bound and then activates its triphosphate to a shape that lets it carry out the reaction.
But why is this necessary?
This is where we have to speculate a bit. In the stabilized position, the enzyme can’t carry out its normal reaction, but it also can’t carry out a second reaction, called hydrolysis. Hydrolysis is the breakdown of a molecule caused by water. As all biological molecules are found in water, it is always available to react with an activated molecule. Normally, the kinase enzyme should transfer the phosphate the antibiotic substrate. However, when there is no antibiotic around, it’s possible that a water molecule can sneak into the active site and react with the GTP instead. The result of this interaction is that the GTP is broken down, and no antibiotic is inactivated. This wastes precious GTP for the bacterium, so any enzyme that cuts back on rates of hydrolysis will be preferable for the survival of the cell. This mechanism may have been developed to reduce this energy wastage by the APH(2”)-Ia enzyme. In environments where there is a lot of competition and resources are scarce, enzymes that conserve energy are an enormous benefit for a bacterium.
This switch between stable and activated forms of the GTP molecule turns parts of the enzyme into a molecular motion sensor for the presence of antibiotics. When they aren’t around, the enzyme hangs out and holds on to the GTP, inactive. It’s only when the antibiotic shows up and sticks to a different part of the protein, that the enzyme undergoes changes that activate the GTP. Like a motion sensor-based system to turn your lights off when there’s no one around, flipping between these states might be an interesting way for the enzyme to turn its activity off and conserve energy when it doesn’t have an antibiotic to modify.
So what’s the big picture?
There’s a lot of places this work leads. I’ve skipped over another interesting finding that two different classes of aminoglycoside interact with the protein despite the fact that only one of those classes can actually be modified. I’ll save that for another blog post. There’s also a lot more detail on the specifics of how APH(2”)-Ia works that I’ve glossed over, which we could also explore some time later.
This mechanism isn’t too different from other mechanisms in proteins that carry out various functions in biology. This switch can be considered a type of induced fit or conformational selection of the enzyme, a well-established model for protein behaviour. The thing that makes it different is that this enzyme is an antibiotic resistance enzyme. Usually antibiotic resistance enzymes aren’t thought to be very complicated. They are thought to be inefficient but highly active machines to turn off antibiotics as fast as possible. This work shows us that an antibiotic resistance factor can be modulated and subject to regulation in a way that reduces its energetic waste.
How about the bigger picture?
Taking the inference a step further, this induced activation of the enzyme indicates that there is a greater complexity in the action of this enzyme than we previously might have expected. However, it isn’t as surprising as we might think when we remember that many antibiotic resistance factors have been around for millions of years, with a very long time to optimize their catalytic activity. We treat antibiotic resistance as something that pops up fresh when we use antibiotics, but the truth is, like we’ve discussed before, there are some forms of ancient antibiotic resistance are highly fine-tuned and regulated to respond to challenges in their natural environment.
APH(2”)-Ia is one resistance factor among many. This paper shows us that resistance factors need not just be catalytically optimized machines, that they can contain a degree of fine-tuning that regulates their activity. This nuanced activity is supported by long periods of evolutionary selection to produce highly effective resistance enzymes, ones that lead to terrifyingly effective antibiotic resistant microbes. We now understand one factor a little bit better, and hopefully that helps us just a little bit more in our efforts to beat back the surge of antibiotic resistance.
Caldwell SJ, Huang Y, & Berghuis AM (2016). Antibiotic Binding Drives Catalytic Activation of Aminoglycoside Kinase APH(2″)-Ia. Structure, 24 (6), 935-45 PMID: 27161980