This post is really frustrating to write.

Its frustrating because this paper could (should) be FANTASTICALLY COOL, but it was just really frustrating to read:

Mosaic HIV-1 vaccines expand the breadth and depth of cellular immune responses in rhesus monkeys

You all know there is a TON of genetic diversity in HIV-1. That has been a huge barrier for vaccine design. You put one HIV-1 variant in a vaccine, people are exposed to another, they still get sick. Our vaccines are eliciting a very narrow immune response. Our immune systems are only seeing exactly what our vaccine tells it to see.

We would like our HIV-1 vaccines to make ‘broad’ immune response– Youre vaccinated with Variant A1, but you are protected against all Variant A and B1/2/3/4 and C3/6/9, etc. We can find antibodies in AIDS patients that are ‘broadly neutralizing’ (can shut down lots of subtypes of HIV-1), but we cant force your immune system to make that antibody, short of gene therapy.

This group tried a new approach. They put different variants of HIV-1 into adenoviruses:

  1. Consensus M– They took every HIV-1 Group M sequence in the Los Alamos National Laboratory HIV Sequence Database and averaged them.
  2. Consensus B + Consensus C– They took all the Subtype B sequences and averaged them, and all the Subtype C sequences and averaged them, and put them both in one vaccine
  3. Optimized C– They took all the Subtype C sequences and broke them into chunks of nine amino acids. They then computationally ‘optimized’ the amino acid sequences, trying to get the most putative T-cell epitopes covered
  4. Optimized mosaics– They took ALLLLLLL the HIV-1 sequences in the database, broke them up into nine amino acid chunks, and did the same thing as Optimized C. Since there is not just one potential optimal sequence, they put the top two optimal ones into one vaccine. This sequence might contain one optimal Subtype A epitopes and 50 optimal B epitopes and 25 optimal C– its a mosaic! This would hopefully be better than just putting in averaged sequences, or the optimal epitopes for just one subtype.

This is a really friggen neat idea! And it might have worked okay, um, kinda?

Here is where things get frustrating: There are literally no methods in this paper. It was almost unreadable, to the point where I physically cannot judge (peer review) the science of what they did. Im basically ‘taking their word for it’, whole cloth.

I can understand this on one level, as if they think they are making THE HIV-1 vaccine, they dont want to tell everyone right now how they are generating the mosaics (its one thing to make a computer sequence, its another thing to actually generate that sequence for experimental use in the lab). But there was no, not even basic, biochemical characterization of these mosaics.

For instance, they made this entirely artificial envelope sequence, and apparently made the envelope protein well enough that they could get (minor) but see-able neutralizing antibodies. Putting together a sequence of amino acids that is going to get chopped up and presented in MHC is very different than putting together a sequence of amino acids and getting an envelope protein that is structurally sound enough to be appropriately folded, glycosylated, processed, and presented on the surface of cells infected with the vaccine (but not cleaved or fusogenic!). I can make small, minor amino acid changes within my constructs and accidentally kill the envelope… but they put these sequences together piece-meal and got a structurally normal Env?

That is just one example– the whole paper, a Nature paper, is a question mark for me. No details for anything they did, even trivial things. So why did they publish at all? Why did they publish in Nature, if they werent really going to tell anyone anything?

I wish I could tell you all more about this (maybe) cool avenue for an HIV-1 vaccine.

*shrug*

Comments

  1. #1 Jason Dick
    May 13, 2010

    Isn’t the better question how did they publish? In nature, no less?

    I guess it just goes to show that if you’ve got a sensationalist idea, Nature will publish anything.

  2. #2 Sili
    May 13, 2010

    Here is where things get frustrating: There are literally no methods in this paper. It was almost unreadable, to the point where I physically cannot judge (peer review) the science of what they did. Im basically ‘taking their word for it’, whole cloth.

    Well, duh! It’s Nature.

  3. #3 ERV
    May 13, 2010

    Just to be clear, the logic behind it is 100% sound– ie this Big Dog Paper published a few years back. I just dont know how anyone is supposed to judge their methods.

  4. #4 Brian Foley
    May 13, 2010

    Well, to some degree it is like publishing a paper today that includes a DNA sequence. You don’t need to describe exactly how it is all done because it is pretty standard operating procedure for most of the steps. If a prior paper described exactly how to PCR amplify the gene region you sequenced, you refer to that paper rather than re-writing the method.

    The mosaic vaccine design is very difficult to understand until you hit the “aha!” moment when you understand that it is not producing anything really “artificial”. I’ve linked to the MOSAIC make tool at the HIV Databases in my name here.
    There is quite a bit of documentation with the tool.

    Korber et al are definitely NOT trying to hide the method of making these types of vaccines. The tool is free to anyone to use, for HIV or any other organism. Bette Korber is very much against patenting and profiting from HIV vaccine designs.

  5. #5 ERV
    May 13, 2010

    I dont fault Korber– I know she didnt do the wet-lab stuff. But I do think her good name is one of the reasons they could get away with such a… ‘limited’ method section.

    I do fault the contributing authors who did the wet-lab stuff. Its not just the real-world construction of the mosaics that have sparse/no methodology described, and Im not aware of any paper where that is described (nor was it cited).

    Id love to read it if you know what it is!

  6. #6 Brian Foley
    May 13, 2010

    You wrote: “but they put these sequences together piece-meal and got a structurally normal Env?”

    The proteins produced by the Mosaic tool contain no “un-natural” 9mers if the tool was set to nine amino acids, or 13-mers if you set it to 13, etc.

    If subtype A tends to be CTRPRNnNTRkRIRIQRGPG
    and subtype B tends to be CTRPRNaNTRrRIRIQRGPG

    using lowercase to show where they are different.
    And no strain in the database has PRNaNTRkR then this peptide is excluded from the vaccine and only PRNnNTRkR or PRNaNTRrR can go into the vaccine.

    Likewise all the n-mers (9-mers, 13-mers, whatever you ask for) are overlapping so your vaccine contains 2, or 3 or 4 (you choose) essentially “natural” genes. It is not just a mish-mash of random peptides or anything like that. It is similar to choosing 2, or 3 or 4 real HIV strains to make the vaccine from, but instead pretending that we have sequenced a hundred million or so recombinant strains and then picked the best 2 or 3 or 4 from all those recombinants.

    The recombinants are generated in the computer, from real existing HIV sequences. Any n-mers that did not exist in one of the real sequences get excluded, just throw away that sequence and keep going, there are millions to choose from.
    The subtype C mosaic and the M group mosaic were both made the same way, the only difference is the input data set for C was all subtype C in the database, and input for M was all HIV-1 M group sequences in the database.

    My name this time is linked to the help manual page for the MOSAIC tool.

  7. #7 ERV
    May 13, 2010

    Thanks for your help, Brian, with awesome links and insider info!

    But how did they take a sequence on a computer and turn it into a real-world DNA sequence that can be packaged into an adenovirus? And how do you know when you put a series of X-mer together that the resulting protein will be folded/glycosylated/presented properly? A sequence that is fine in the genetic background it evolved with might not work at all in a different genetic background. How am I supposed to know whats up as a reader?

  8. #8 Brian Foley
    May 13, 2010

    This Nature Medicine paper was not about making the mosaic vaccines. The vaccines were made a couple years ago and tested in mice etc, with several papers resulting. This paper is about testing those same mosaic vaccines in rhesus macaques.

    Don’t feel bad for misunderstanding any aspect of the mosaic vaccine idea. I have had it explained to me several times and I still don’t quite “get” some aspects of it. Bette Korber explained it in her talk at the Keystone meeting this year in Banff, and several people who asked questions at the end also clearly had some of the same misconceptions you have, being surprised that the protein was functional, folded properly etc. because it was “synthetic”. A key point it that there are no n-mers in the sequences that do not occur in nature.

  9. #9 ERV
    May 13, 2010

    *runs off to PubMed*

  10. #10 Brian Foley
    May 13, 2010

    You wrote: “sequence might contain one optimal Subtype A epitopes and 50 optimal B epitopes and 25 optimal C– its a mosaic! ”

    Not quite. The MOSAIC tool puts out just one or two or three or four (you choose how many) complete gp160 sequences (for envelope). It’s just that for each n-mer (you choose something like 9-mer to 13-mer) it picks the top one or 2 or 3 or 4 real existing peptides in the database, based on how often they occurred in subtype C, or M group (or whatever your input set was). It is not breaking all the sequences up into n-mers, either. It is recombining sequences to make complete gp160 genes, and then throwing out any that contain non-natural n-mers in them. It runs through many iterations of that process.

    HIV-1 M group is not really all that variable. The pol genes are all 90% identical to each other, env more like 80% identical. Many sites are 100% conserved. Other sites can only vary between 2 amino acids like Leucine and Isoleucine. One problem is that not all n-mers are immunogenic.

    Anyway, the whole point of this vaccine design approach is to NOT put in 25 different C variants and 12 different B variants etc… The tool picks just 1 or 2 or 3 or (you choose) sequences made to best cover all observed variants. There are several reasons why it is best NOT to put in a “swarm” of dozens or more variants, and to go with somewhere between one and 3 or 4 constructs.

  11. #11 Brian Foley
    May 13, 2010

    You asked: “But how did they take a sequence on a computer and turn it into a real-world DNA sequence that can be packaged into an adenovirus?”

    People have been putting “synthetic DNA” into adenovirus, vaccinia virus, and a few dozen other “vectors” for some 20 years now. Also, they have been making SHIVs, attenuated SIVs and many other types of vaccines. Most viruses will package any DNA that has the right length within some reasonable range, and the right packaging signal. For example the packaging signal for HIV-1 genomes are called Psi elements, and a part of the Gag protein (matrix) binds to the Psi elements of the HIV genome, and then other parts of the Gag protein (capsid) package up the matrix-genome complex and transport it to the budding virus.

    Cutting out a chunk of non-essential Adenovirus and replacing it with a HIV-gag or HIV-env gene is just like using Lambda phage to package some yeast or human DNA. Or using m13 phage to get single-stranded human DNA for DNA sequencing by the Sanger method.

  12. #12 Brian Foley
    May 13, 2010

    One more key point is that it is not the goal to make “the most immunogenic” vaccine. It does not do any good to use the peptide SLYNTVATL in your vaccine if every human HLA type recognizes that peptide and attacks it if HIV has already “escaped” this immune response such that none of the viruses out there in the world that the people will be exposed to contains this epitope any longer. The vaccine should contain the epitopes that exist in the viruses that people will be exposed to, not the epitopes that excite their immune system the most.

  13. #13 Brian Foley
    May 13, 2010

    Sorry about the “run on sentence” there. I am sure my sixth grade English teacher is rolling in her grave over that…

    Anyway, my last post has me thinking about “dominant epitopes” and “decoys”. I forget where I read about it now, but I saw the idea that some pathogen(s) put out “decoy epitopes” to which the host produces an overwhelming immune response. The pathogen is adapted to easily evade this response. The host immune system is so busy attacking the decoys that it ignores, or has less energy available for fighting, the epitopes that are more conserved.

    I just did a GOOGLE search for “pathogen decoy epitope” and linked on of the hits to my name below.

  14. #14 JohnV
    May 14, 2010

    I just wanted to say that this is the most amazing comment thread I’ve ever seen. Thanks for all the info Brian. I think I even almost know what the hell is going on now. (This is why I work with bacteria).

  15. #15 Optimus Primate
    May 14, 2010

    For true. Brian, if you don’t have a blog, you need one. Now. My brain hurts so good at the moment.

The site is currently under maintenance and will be back shortly. New comments have been disabled during this time, please check back soon.