#MicroTwJC part 2: The Rise and Fall of LJ-001

Yesterday, I wrote about the discovery of LJ-001, a brand new drug which can fight against a broad spectrum of viruses, such as HIV, Flu and Ebola. These viruses have lipid membrane envelopes, which allow them to fuse with their host cells during an infection.
 LJ-001 is a compound that can insinuate itself within lipid membranes and cause some level of low level damage. Living cells have natural repair mechanisms that allow them to resist the damage. Viruses are definitionally not alive during the infective portion of their life cycle, and thus can't do anything to stop the damage caused by LJ-001. Thus, LJ-001 kills viruses.
But there were still questions. What exactly is LJ-001 doing in the membrane. Is there some specific lipid target it acts on ?
It is really important to work this out. LJ-001 was discovered through a combination of bloody minded determination and sheer luck. If we can work out how it acts, we can try to make it better. Even more importantly, it will hopefully allow us to predict how quickly viruses can adapt to it if eventually becomes a widely used treatment.



At what point does LJ-001 stop the virus infection cycle ?

In the last paper, the researchers found that LJ-001 stopped viruses from fusing their lipid envelopes with the host. But this fusion event is a complex process, which in itself encompasses a set of shorter steps.
I will quickly describe these general steps *:

1. The enveloped virus has a series of receptors on its surface  that allow it to bind to it's host cell. The first step is when the receptor on the virus binds to the receptor on the host cell.
2. The binding causes the structure of the viral receptor to change. Often it forms an "extended intermediate", in which the viral receptor unfurls and elongates. This is known as the pre-hairpin intermediate, and it is fairly unstable **
3. The intermediate collapses, and in the process of this collapse pulls the attached receptor much closer, bringing the viral membrane and the host membrane into close proximity, such that they begin to merge.
4. The fusion of the membranes begins, by forming a pore structure between the virus and the host membrane, which the allows the contents of the virus to enter it's host target.

So at what step is LJ-001 interrupting this process ?

The researchers performed a time of addition assay.

Panel A: To find this out, the researchers focussed on the effects of LJ-001 on HIV fusion. To investigate this, they modelled an HIV infection, and introduced specific inhibitors that disrupt each step of this process. Leu3a inhibitors act to prevent the initial binding of the HIV receptor to its target. T-20  prevents the virus forming the pore structure. They also had AZT, which acts to kill the virus off at a far later stage of the process. They added these molecules at all of the time points during the first two hours of infection.
The Leu3a inhibitor is the one to act earliest, in the initial interaction. Introducing it after the HIV has already performed this interaction means that it is ineffective, an so cells die off. The T-20 acts much later, and so is only ineffective at the later time points. The longer it takes for a compound to lose its effectiveness, the later it acts in the process. The researchers could quantify this by extrapolating the half life of the effectiveness of each compound i.e. the time at which half of the maximum number of dead cells have been killed.
So the lower the half life demonstrated by these compounds, the earlier they act in the infection cycle. The researchers subjected LJ-001 to the same test, and based on the graph it acts after the final stage of fusion.
Panel B is a repeat of the experiment performed in panel A, but using a different virus (Nipah Virus) and a different set of inhibitors, which also suggest that LJ-001 acts in the later stages, or after the fusion event.
Panel C shows an experiment with the Semliki forest virus. This virus fuses via a slightly different method, using protein known as E1, which performs the final stage of the process of fusion for this virus if it is exposed to a low pH. In this experiment, the viruses were mixed with host cells, and then exposed to a low pH to induce fusion. When this happens, the E1 proteins on the surface of the virus link up to form a homotrimer, which is resistant to attack from enzymes like trypsin, and chemicals like Sodium dodecyl sulphate (SDS). The results suggest that LJ-001 has no effect on the formation of these homodimers.

What molecules within the lipid membrane of the virus are changed after the addition of LJ-001?

We know from previous work that LJ-001 is causing damage to the viral membrane. So if we examine the contents of this membrane, we can see which components are being changed by the virus, and thus work out what it's doing. The first thing they did was to deprive the viruses of cholesterol, only to find that it had no effect on infection, and thus that cholesterol had no effect (Panel A)
The researchers took the Influenza virus, exposed it to LJ-001 and an inactive control (LJ-025) and analysed the viral "lipidome" through liquid chromotography and mass spectroscopy.This allowed the researchers to look at all of the lipids produced by the virus. There weren't any changes in the basic classes of phospholipid (Panel B). But when the researchers took a closer look, the realised that there were changes occurring to these fatty acids. The phosphatidycholines were becoming oxidized (Panel C). They needed to rule out any virus effects on this composition, so they tried out this same experiment using artificially created liposomes (Panel D), and found that LJ-001 caused the oxidation of lipids here as well.

This all suggests that LJ-001 oxidizes the phospholipids within the viral membrane, through generating single oxygen molecules which behave as free radicals.

Is the LJ-001's oxidation of the viral membrane the cause of its anti-viral properties

If LJ-001 is responsible for the oxidation of these lipids, and this oxidation is responsible for the antiviral effects of this drug, then anti-oxidants should be able to completely block its effects.
They added LJ-001 to Dimethyl-anthracene (DMA), a chemical that naturally sucks up singlet oxygen free radicals when they are generated. If a large amount of  DMA is oxidised upon contact with LJ-001, it suggests that LJ-001 is generating "singlet oxygen".  When they added a different antioxidant alongside DMA (alpha-toco), this stole all of the singlet oxygen from the DMA, stopping it from being oxidised, and when they placed the reaction in an argon atmosphere, the reaction stopped as well. (Panel A)

If singlet oxygen is responsible for LJ-001's killing of viruses, then the addition of these antioxidants during an infection will prevent LJ-001 from affecting the viruses. Which is what we see in Panel B. The researchers added a number of different antioxidants to three different species of virus during treatment with LJ-001. The antioxidants allowed the viruses to pursue their infection cycle even in the presence of LJ-001.

In the previous study, we saw that LJ-001 absorbs light in the visible spectrum, and fluoresces. The researchers hypothesised that the absorption of light by LJ-001 helps it to create oxygen radicals.
In Panel C, they mixed the LJ-001 at three different concentrations with Herpes Simplex virus, and then exposed it to light for different lengths of time, before adding this mixture to the host cells for 1-2 hours in the dark. The longer these viruses were exposed to the light, the more singlet oxygen is generated by the LJ-001 incorporated in their membranes. This means that the lowest doses of LJ-001, which are generally non-lethal to the virus, can become lethal if exposed to enough light.
In Panel D, the researchers perform infections in light and in the dark to see if that renders LJ-001 ineffective. Remember, in all of the experiments we have seen previously, even those performed in the previous paper, the LJ-001 has been mixed with the virus in a petri dish, presumably in the light. Even in the animal infections, they mixed the virus and the LJ-001 before they put it in the animal. we see here that without light, LJ-001 is rendered ineffective against viruses.

So how does light affect LJ-001, and how does the oxidation caused by this compound stop viruses from working ?

This is where the paper gets slightly complicated. We know that LJ-001 gets activated by light. when this happens, the molecule goes into an "excited" high energy state. It is thought that this is what gives the molecule the energy to generate singlet oxygen.
But what if we stopped the molecule entering this excited state ?
We can do this using "quenchers". The molecules can collide with LJ-001 in its excited state, and steal all of that excess energy from it, preventing it from working. This simple model of quenching is known as the Stern-Volmer relationship.
We can measure how effective quenching is through measuring the Intensity of fluorescence with the quencher present (I), and the intensity of fluorescence when the quencher is not present (I0).  The ratio of these two measurements can tell us a lot about the properties of the fluorescent molecule we are investigating. If we know the specific properties of the quenching molecule, then it is possible to derive the quantum yield of the molecule's fluorescence, i.e. the number of photons emitted per photons absorbed.
In a reaction where all of the molecules are able to interact with eachother, the I/I0 value will increase in a linear fashion. We can see this occurring when the researchers added rising concentrations of an acrylamide quencher to a reaction in Figure S6 of the paper.

In Panel A below, we see what happens when the researchers add rising concentrations of Large Unilammelar Vesicles made of either an HIV like mixture, or from POPC (1-palmitoyl-2-oleyl-sn-glycero-3-phosphocholine) which are used to represent the host and viral membranes respectively.
In this case, the quencher used here is water soluble, and the LJ-001 (and LJ-025 control) can enter the lipid membrane, hampering the quencher's ability to collide with it. Thus we see when the concentration of the vesicles reaches an upper limit, no more quenching can occur. I = I0, and the gradient of the line becomes zero.
This relationship is described by a specific equation, which can mathematically predict the degree to which LJ-001 enters the lipid membrane. If you want to understand how this happens, you can find the mathematical underpinnings here.
In really simplified terms, the point at which all of the fluorescent compound enters the membrane is determined by it's intrinsic ability to enter the membrane. If it doesn't enter the membrane very well, the point at which the line goes horizontal will be at the higher lipid concentrations. If it enters the membrane readily, the line will level out much earlier. By the looks of things, LJ-001 partitions into POPC membranes slightly more quickly than into HIV like mixture, but I haven't done the stats, so I can't really tell whether it is an important effect or not.

In the next two panels (B & C) the researchers created monolayers of either HIV-like membrane, or the POPC membrane using a piece of equipment known as a Langmuir Blodgett trough.
To understand how this works, imagine you put a drop of oily surfactant on the surface of water. It naturally starts to spread out on the surface to form a film. Surfactants generally have "a head" that is attracted to water, and "a tail" that is repelled by it. So these molecules naturally align with their tails stuck in the air when they form a film. The farther the drop of oil expands, the more room these tails have to move around.
If we limit the spread of this oily layer, by putting down barriers, the oil is forced to form a "monolayer". As we push molecules within this layer further together, the tails of the surfactants can no longer move as much, and they end up lining up together in a continuous layer over the water. This changes the surface tension of the water, and this change is described as "Surface pressure". The more tightly packed a surfactant can become, the higher its surface pressure. This is important for living things like us, because it determines the fluidity of our membranes. If they are too fluid, they fall apart, and if they are too rigid, they can no longer move and function well enough to allow for proteins to float through them.
The Langmuir-Blodgett trough suspends a monolayer in between a set of plates that allow for it to be packed as tightly as possible, and for measurements of surface pressure to be obtained.
We know that LJ-001 oxidises lipids and changes there structure, but we don't know how this structural change affects the whole membrane. The researchers propose that the change in the structure of these lipids caused by the oxidation makes them able to stack closer together.
So when LJ-001 was added to a trough with a layer of POPC in the presence of light, we see that the surface pressure increases (Panel B). We also see a similar effect when HIV-like lipid layers are used (Panel C).
One effect of this increased surface pressure could be a decrease in membrane fluidity. To measure this, we can infuse the membrane with fluorescent markers, and measure the degree to which they move in the membrane. We can do this by taking advantage of a phenomenon known as "Fluorescence anisotropy".
Fluorescence occurs when a photon hits an atom and stimulates it into a higher energy state. Every photon has an electric and a magnetic component (hence why light is also known as electromagnetic radiation), and a photon can only "excite" an electron when it hits it in the right orientation.
Let us imagine that we have a set of fluoroscent molecules stuck down onto a surface, and that we then use polarized light to excite them. Polarized light is light which is oriented in the same direction. When the photons of polarized light hit the electron orbitals in the right orientation, the electrons within those orbitals get excited to a higher level. When they return to the lower energy state, the emit light. Since we excited these fluorescent molecules with polarised light, they emit polarised light as a result.
Let us run the same experiment again, but this time the fluorophores are allowed to move around. This time, when we stimulate them with polarized light, they all get excited, but they are also moving around. So when they move out of position and emit light, it exits at a different orientation to when it initially hit the molecule. Since the fluorescent molecules are all moving in different directions, we find that the light emitted by these molecules comes in all sorts of directions. It is no longer polarised.
In effect, the degree to which the incoming light from the fluorescent molecules is depolarised reflects how well they can move. A low level of polarisation indicates that the fluorescent molecules are free to move. A high degree of polarisation indicates that they are stationary. The degree of polarization of light is known as "Anisotropy".
The researchers attached fluorescent markers to the surface of a lipid membranes and used fluorescence anisotropy to measure whether these molecules were less able to move around after the addition of LJ-001 (Panel D & E). The LJ-001 increased the fluorescence anisotropy more than the control compounds, suggesting that it reduced membrane fluidity.
This is a key issue for viruses, because they need to be able to manoeuvre their receptors through their lipid coat to interact with the host. Decreasing membrane fluidity is like hitting the virus with a freeze ray.

Can we improve on LJ-001

Now that the researchers have a bead on how LJ-001 works, they can now try to create better examples of this drug. A key issue for this drug is that it is reliant on light. But the sun does not shine on the inside of most creatures. However, longer wavelengths like red and infra red can move through the body relatively easily, so it would be useful if we could change the molecule to absorb those wavelengths better. Furthermore, they wanted to increase the molecules potency.
They ended up with a molecule dubbed JL-103 (Shown in Panel A above). They took this new compound, and tested it on a panel of different viruses, and measured the concentration of drug needed to inhibit viral infection by 50% (IC50). The JM103 seems to be more effective against all of the viruses.(Panel B)
They took a look at its absorbance spectrum, and saw that JM-103 had been shifted into the red spectrum (Panel C) If you delve through the supplementary, you'll see that the researchers characterised this drug all over again, from looking at its properties and finding that it generally works in the same way as LJ-001, only deadlier. When it was added to liposomes, it resulted in more lipid oxidation (Panel D).  

Can this drug be used ?

The ultimate test of this drug will be whether it can effectively protect people from viral infections. But there are a number of problems here which need to be understood before we can move forward.
One of the biggest problems is the reliance of this drug on light in order for it to work. Haemaglobin is the enemy in this situation. It is responsible for absorbing most light, except for red and far red. Haemoglobin can absorb light of similar wavelength to the JL-103, and can stop the right wavelengths of light getting to it.
 They tested this by incubating JL-103 with virus, host cells and increasing concentrations of haemoglobin. The haemolobin reduced the effectiveness of the drug by fifty percent when it was present at physiological levels (Panel A).
So once again, they had to go to the drawing board, and created two new compounds (Panel B) that were shifted further towards the infrared end of the spectrum (Panel C). These drugs fared much better when they were tested for effectiveness in the presence of haemoglobin (Panel D).

In the next step, the attempted to see how this drug fares within an animal. So they looked at the drug half life, which measures how long the drug sticks around after it is given. They also looked at Cmax, which is the highest levels of the drug within the blood. The AUC (Area under the curve) to some extent represents a little of both of those two measurements, as it shows the "total drug exposure" over time.  They were all compared to LJ-001 (Panel E) and they all compared better. Which is good, considering that LJ-001 had a half life of about 6 minutes. Luckily, with JL-118 at least they managed to bump it up to 16 hours, which is a much better prospect if we want to think about treatments.
In the next step, they gave these drugs to mice infected with rift valley fever. If you look in the supplementary data, we see the researcher's own attempt at this, and they initially found that the drug had no effect. So they paid SRI laboratories to do the rest of the experiments for them.
Panel F shows how JL-118 and JL-122 fare in preventing mice dying from rift valley fever. The good news is that they extended the life of the mice through the experiment, but the bad news is that all of the mice still died from the disease. Panel G looks at JM-103 vs JM-122, and finds that JM-122 does better, but the mice all still die at the end of the experiment.

Summary

Now we know how LJ-001 works. it enters the membrane of an enveloped virus, and when light strikes it, it creates singlet oxygen, a free radical that attacks the lipids of the viral envelope. This makes the viral envelope more viscous, and thus makes the virus unable to mobilise its membrane proteins, and inactivates it.
With these facts in mind, the researchers developed a new set of drugs that act in a similar way, only better. They changed the light absorbing properties of the molecule to prevent it having to compete with haemoglobin for light when it is used in a living organism. They improved its activity against viruses, and its pharmacokinetic effects. Thus they created drugs which could extend the life (although not save the life)  of mice that had just been infected with Rift valley fever.

Reader Comment

I should note that there is a reader comment on  this paper regarding the structure of LJ-001. The concern here is that whilst the authors have shown that LJ-001 can produce singlet oxygen, this is may not be all it's doing. It has other properties that mean that it could be reacting with proteins, and doing a number of other things which the researcher's haven't really looked at. The comment then cites a paper that is stuck firmly behind a Paywall, which is unfortunate, but an inevitable event in any scientific discussion. So I did a little bit more digging, so that I could at least level an educated guess at what the comment may be referring to. I believe that in the paywall locked article (http://www.future-science.com/doi/abs/10.4155/fmc.10.237) there is a description of a few specific compounds which mess with most of the standard methods of measurement. These "problematic" compounds react with anything. The reader is suggesting that some of these compounds have similar structures to the LJ-001 shown in this article, and that the structure could potentially form covalent bonds with proteins.
 If this was the case, then we would expect to see binding of LJ-001 to viral proteins. This was one of the key things they investigated in the previous paper, and they found nothing. I don't want to dismiss this comment out of hand, but the evidence I've been presented with doesn't support it***.

EDIT: An astute anonymous commenter pointed out a very good reason why they found nothing in their previous study. The researchers used western blots to detect their proteins, in order to see whether any of them had gained any extra weight from covalently bonding with LJ-001. The aforementioned anonymous commenter provided a very good reason why we wouldn't detect the covalent binding of LJ-001 to any of the viral proteins using this method. The key point is that the sulphide double bond hanging off of the blue region of this protein (above) could potentially react with thiol groups hanging off the sides of cysteine groups on viral proteins, forming a bond that we refer to as a "disulphide" bridge. However, when preparing proteins for western blots, we need to get them to unspool into straight line molecules. In order to get them to do this, we add a strong reducing agent, the function of which is to break up bonds like disulphide bridges.  So if LJ-001 did bind to the viral proteins, the manner in which it is processed for western blots would cause it to unbind, and we would never know it happened.


Summary

That being said, I have a few comments and criticisms of the paper myself.

Photo activation:
The moment I learned that this drug needed light in order to work, I winced. This raises a ton of problems for making this treatment applicable for humans. Even though with their later molecules they managed to red shift them out of the range of haemoglobin, the body will always absorb some of the photons on their way to the site of infection. This is not even taking the scattering of photons by the bodies tissues into account. Every millimetre of body tissue that separates your light source from its target exponentially decreases the light signal reaching your target. The smaller the creature you test on, the more effective your drug will appear. As you may have noticed, humans are much bigger than mice. The Beer Lambert law and the Inverse Square law team up to ruin the effectiveness of your photo activated drug.  You really need to test them out on organisms that are closer to the size of humans in order to see how effective they will be in reality. Seeing how effective this drug is in mice, I really am struggling to work out how it could work better in humans.

Regression Abuse:
For once, I don't have a problem with the way the statistics were done. I have a big problem with how they were described, especially the regression analyses. To understand what irritates me here, imagine yourself walking into a pasties shop.

"Hello" you say to the vendor over a series of unlabelled pasties, "What pasties do you have today ?". 

The vendor gives you a wry grin and says "Well we don't have cheese and bacon".

"Thank you for telling me" you say, then one of the pasties catches your eye. "What's in that one ?" you say to the man.

"It's definitely not cheese and bacon" he says again, grin frozen in place.

"We've established that, what is it ?"

"Not cheese and bacon".

You point to another pasty, differently shaped, with a pastry chicken emblazoned on it. You point at it and ask him what it is.   

"Not cheese and bacon " he repeats, his forehead starts to glisten and you see a mad gleam in his eye. 

"Do you have cheese and bacon at least ?" 

 The next sentence he says in a slow monotone and a drop of blood issues from his left nostril.

"All of our pasties are not-cheese and bacon"

You get the impression he may be holding a sharp and deadly under the counter. You hope it's just a knife. You back away.

Now consider what the word "Non-linear" means. It means the analysis is not linear. There are tons of analyses out there that are non-linear. You could be using an nth order polynomial, but I would never have any idea about it because it looks exactly like a different graph I am expecting. Regression isn't just about drawing pretty lines, it is about getting an equation to explore a specific trend in the data. But if I don't know what kind of equation is generating it, then I must ignore it. That is a shame, especially because the authors have definitely created these equations, and they know what they are. 

They also don't show their R-square values, and this is quite important if we want to extrapolate data from a regression line. The R-square value tells us the degree to which a regression line represents the actual data from a study. In theory, you could generate regression lines which have low R-square value which don't represent the data well at all, and still get a value out of them, because those values only come from the regression line, not from the data itself.

Conclusion

This has been a challenging paper to get through, but rewarding nonetheless. The two papers we have in this journal club trace the career of one drug, showing the highs and the lows of drug discovery. 

The reliance of these "membrane based oxidizers" on the interaction of light is something that I personally don't see ever working for severe infections in humans. It may work on skin lesions, like those caused by Herpes, but I am sceptical of it being used to treat diseases like hepatitis, or systemic diseases like Ebola and HIV. 

Even if LJ-001 and it's brethren don't end up working in humans, it has established a crucial weakness in the defences of whole classes of viruses. Now that we have a working hypothesis as to how it freezes and traps viruses within their own envelopes, we can try to develop drugs that do the same thing that aren't reliant on light. This doesn't have to be the end, this could be a new beginning.


Vigant F., Lee J., Hollmann A., Tanner L.B., Akyol Ataman Z., Yun T., Shui G., Aguilar H.C., Zhang D. & Meriwether D. & (2013). A Mechanistic Paradigm for Broad-Spectrum Antivirals that Target Virus-Cell Fusion, PLoS Pathogens, 9 (4) e1003297. DOI: 10.1371/journal.ppat.1003297.s017

Join in the Microbiology Twitter Journal club this Tuesday at 8pm BST, follow the #MicroTwJC hashtag.

*I'm being fairly general in this description. I strongly suspect that many viruses have different variations on this process. Feel free to pipe up in the comments if you have any examples, especially if those differences mean that "membrane based oxidizers" won't work on that specific virus.
**Although the intermediates of  HIV and other viruses have pre hairpin intermediates with a long half life.
*** Hopefully that will change if someone could e-mail me a pdf, or describe why the western blots of the previous paper cannot be trusted EDIT: This literally happened an hour after I posted this, check this comment