Lipopolysaccharide and humans

LPS is bad stuff. But for humans, LPS is far worse.

We humans, it turns out, are unique in a very specific way. Immune system receptors for various things are called toll-like receptors (TLRs). There are quite a few of them, TLRs 1-13.

In most animals, each TLR is paired with a Siglec. Siglecs aren’t numbered for the TLR that they work with. But TLR4, the one for LPS, is normally controlled by Siglec 13.

Each TLR turns the immune system on. The Siglecs damp the TLR response if it gets too high. They don’t work 100%. If they did, no animal would die of sepsis. But they make several thousand-fold difference.

We humans don’t have a working copy of Siglec-13. We haven’t had it for a long, long time, because Neanderthals didn’t have them either. We split with Neanderthals around 1 million years ago. That’s 83,000 grandma’s in the past.

The LPS LD50 for a 25 gram mouse is 150 micrograms. That is 150 millionths of a gram. Not very much, but it works out to around 6 milligrams per kilogram. (1 kg = 2.2 pounds).

If humans had working Siglec-13, we would have an LD50 of around 420 milligrams, based on 6 milligrams per kilogram. But instead, the LD50 for humans is… 300 micrograms.

That’s right. It takes just double the dose of LPS to kill a 25 gram mouse as it does to kill a 70,000 gram man.

That’s why, for injection into a human being, it is even more important to have ultra-clean plasmid preps. And the price for a prep for a human clinical trial reflects the level of cleanliness. Instead of $10,000 or so, human clinical trial material costs around $700,000.

Yes, some of that is “what the traffic will bear.” Some of that is regulatory cost. But it’s also because you just can’t make a mistake.

Well, technically, you can make a mistake and use things like Enbrel, steroids and rapamycin to save the life of a test subject. And anyone doing human clinical trials better have those on hand, along with a well defined protocol for following patients just in case. It should never happen, but that’s always true.

Lipopolysaccharide, the “killer app” of plasmid gene therapy


I’m kidding about the “killer app” of course.


LPS – Section A, on the left is the poly-saccharide part made of sugar molecules. B, on the right, is the lipid, or oil part. Rendered using Rasmol.


TLR4 receptor together with MD-2, which form the fully capable receptor. You can see little LPS molecules in there, lower left. RCSB 3FXI

Diagram of TLR4 operation.

Lipopolysaccharide (LPS), also known as endotoxin, kills. LPS is the broken down cell walls of bacteria.LPS triggers signalling of toll-like receptor 4 (TLR4).  TLR4 is one of a pair of large protein molecules that manage to recognize this pattern in bacteria. If you had no TLR4, you would be in serious trouble, because you couldn’t respond to the presence of infecting bacteria. So animals evolved the TLR4 receptor so we can detect their presence, and every animal has it.

The problem with LPS is that it causes septic shock. Injection with purified LPS is the lab model for sepsis. In animals, the LD50 is around 6 milligrams per kilogram of body weight.

So here is where it gets interesting. Because the most reliable way to make plasmids is to do it in E. coli bacteria. (Oh, dear, oh dear.)

A cluster of Escherichia coli bacteria magnified 10,000 times. (Wikipedia commons) You can see the cell walls of these bacteria. The lipid tails of LPS point inward, toward the center of the bacteria. The visible surface is saccharide (linked sugars).

Thousands of plasmids can be made by each little E. coli bacteria. We culture the bacteria, then break open their little cells. That releases the plasmids that are floating around inside of the bacteria. This mess is centrifuged, and the clear liquid that contains the plasmids is drawn off. So far, so good. But a fair amount of the LPS is also floating in the water along with the plasmids. It’s enough that you can kill a 25 gram lab mouse with one little injection. The amount of LPS varies, but it’s usually at least 1.5 milligrams per milliliter of water, and can be 10 to 50 milligrams. (One milliliter of water = 1 gram of water. There are 1,000 milliliters in a gram.)

A 25 gram mouse only needs 150 micrograms (0.15 milligrams) to get an LD50 dose. Rats are a little better. A lab rat weighs about 500 grams, so it needs 20 times as much, or 3 milligrams. But the injections for rats are also larger, so you will usually kill your rats with uncleaned native preparations.

There are two ways to clean the LPS out. You can use ultra-centrifugation in cesium chloride gradient, but it doesn’t always work. You can also use a methods that has the DNA of the plasmids stick to a surface and wash away the LPS. This loses quite a bit of DNA.

And there are kits for it that are supposed to let you clean LPS on the lab bench – easy-peasy. Just one problem with that. I couldn’t make them work well enough. And nobody I know has done it either. A professor said that it nearly killed his lab mice when he injected them. So, yes, those kits work. Without it, his mice would have croaked. Instead, they just got very sick. They work, but not well enough.

Seriously, I tried to make those kits work for almost a whole year. When I couldn’t do it, I tried creating new cleaning protocols of my own. For instance, I mixed the plasmid preparations (we call them preps) with oil. Then I let them separate, drew off the oil and repeated it. That also worked – some. It was incremental, and removed about 7% of the LPS with each iteration. To remove 99.5% of LPS in the prep, it would take 75 iterations – in theory. But I lost too much water each time. After 10 iterations I had half the fluid I had to begin with, and I had lost a lot of DNA along with it.

It may be possible to further protect against LPS by adding LALF (derived from Limulus polyphemus amebocytes ) to the preparation. But I haven’t tried that method yet. It’s partial. It might work if you combined the clean prep kit with it. Emphasis on might. It should generate an immune response to the LALF protein though, so it may only work the first time you inject.

But even if LPS doesn’t make your animal sick it is highly activating to the immune system. For gene therapy, you don’t want that. In gene therapy, you want the immune system to forget about what just came in, “Nothing to see here. Move along. Move along.” For gene therapy, you do anything you can to evade the immune system’s notice.

And that is why I send my stuff out to companies that specialize in producing clean preps. They either use the ultra-centrifuge method, or they use the method that adheres DNA to a surface and washes away the LPS. When they are done, they test it using Limulus amebocyte lysate (LAL) to certify the LPS level. You get certified material and you won’t kill your mice. You won’t make them sick. You will minimize immune system activation. It isn’t cheap at $10,000 or so, but it’s the way to go once you get past cell culture.

Transfection methods – Which is best?

Several basic types of transfection: gene gun (effective on almost anything), electroporation (oldie but goodie), liposomes, polymer complexes, and viral particles, with or without reverse transcriptase.

Cell culture transfection

Liposomes are sometimes problematic, because they kill cells, but these days there are kits from Sigma-Aldrich and others that are quite good. Polymer complexes like Turbofect from Thermofisher also work well, distinguished from liposomes in that they cling to the DNA, and work just about as well with large DNA strands as with short plasmids of a few thousand base pairs. But nobody really cares about that, you just want your DNA to get into the cells so you can move on. Usually, in cell culture, those are just fine.

There are electroporation systems that work in cell culture for both adherent and non-adherent cells. It typically takes a little more handling, and if you don’t set equipment right, you’ll smoke your little cells. But I’ve found electroporation to work fine, even when things are a bit off. Cells are pretty tough if you keep them cold, etc.

The nice thing about gene guns is that you can use them easily right on a plate. Bang, it’s done. They don’t do so well with suspended cells, because the particles won’t penetrate much into water. But usually you can get around that by spinning and draining cells, or culturing them on plates the way it’s usually done when creating monoclonal antibodies. Sterile technique is a lot trickier then though. You’ll need to take great care with your facility, use UV lamps and wipedowns all the time.

Not so many people use viral particles for cell culture, and that’s something I don’t have experience with, so I won’t say any more about it.

In vivo transfection

All of the above work in-vivo also, although you have to prep a bit differently. For instance, to use a gene gun with a plasmid into muscle tissue, you’ll have to open the skin and lay bare a enough muscle. You’ll need to keep the distance correctly, and may want to protect non-inoculation areas from getting transfected. Aluminum foil will do fine surrounding the area if you care.

I like electroporation into muscle also. This is done with a needle surrounded by fine tines that surround the needle in the muscle. The needle becomes negative terminal and the tines around it positive, to drive DNA into cells. This method has been shown to work well, at least in large animals.

In addition to these methods, I like using microbeads of 2-5 microns composed of PEG, PLG or something like that, with your DNA embedded into it. That’s an optimum size for uptake into cells. The drawback is that this can generate an immune inflammatory response – mild, but that needs to be compensated for. (And Butterfly Sciences has some vehicle formulations for that.)

I think that injection into mouse muscle is difficult. Their muscles are so small, and syringes are large enough, that I think it’s pretty easy to screw up the injection. So I’d prefer not to work with anything smaller than a rat.

Special notes about viral particles

Viral gene therapy is more widely known than bare DNA plasmid therapy. In this method, a recombinant virus is engineered containing a gene of interest. Adenoviruses, adeno-associated virus and lentiviruses are commonly used. Viral vectors can be made specific to a tissue type by careful engineering of the virus’s attachment moities.

This method has several drawbacks:

  • The doses required can cause death through a sepsis-like syndrome, although animals are less likely to suffer this consequence because their immune systems are more robust than humans. Jesse Gelsinger died this way. If you get into this kind of trouble, try a shot of Enbrel.
  • There is some risk of cancer. It may be low, but it is there. AAV has a low rate of integration into chromosomes, very low. But it’s not zero.
  • With retrovirus deliver of genes the risk of cancer is significant. Around half of all the children given gene therapy for SCID died as a result.
  • Viral delivery of genes is irreversible. Viruses disseminate in the body, and cannot be removed surgically. In theory, in some limited cases, like synovial fluid, the viral vector should not get outside the synovium. But what about injection errors? Those happen. With plasmids, there is a small section of muscle where the transfection is virtually entirely localized.
  • If the viral vector is accidentally injected into the wrong person, you have a serious problem. Of course, that shouldn’t happen, but when thinking about products that will be sent out all over, things happen. If it’s an animal product, someone might use it on a human. Medical errors happen all the time, from amputation of the wrong leg, to administering the wrong dose of medication. Count on it happening sooner or later.

Both viruses and bare DNA (plasmids) are workable. I prefer the plasmid system because it’s more forgiving. Transfection efficiency is not as good as viral delivery, but it’s good enough.