Skip to Content

Nailing Down Small Proteins

I found this to be a remarkable paper. Making synthetic peptides into drugs has been something that people have been trying for decades now, but it’s a really hard way to make a living. Proteins get degraded. They get degraded in the gut, in the blood, and in every tissue you can name. That may sound a little odd if you’re just picking up the subject of biochemistry, because proteins themselves are such a key part of the physical and chemical structure of any living creature – how (and why?) would they be so likely to be broken up? There are several answers. The proteins being used by living systems are often tailored to their environments, and they generally have an expiration date on them, anyway – they last as long as they need to last in the particular compartment they hang out in. Some have a half-life of minutes, some last for weeks, and a few (like the lens proteins in the eye) are there for the duration, but in general, proteins are there to be turned over. Their sequences, versus the proteases and other degradation pathways around them, are balanced out evolutionarily.

So when you throw a new one into the soup, it’s on its own, and it’s facing an environment that’s biased towards tearing unknown visitors into pieces. That’s not to say that there are no protein/peptidic drugs out there, far from it, but it does present a challenge. There have been many, many tricks developed over the years to toughen peptides sequences up and produced peptidomimetics that can simultaneously hit their targets and survive degradation, but it’s not a straightforward process. It’s more like “Oh, that didn’t work? Well, try this”. Cyclic peptides, to pick one example from the bag of ideas, can be unusually stable, but making them can present some significant synthetic challenges, and you’re also faced with a huge number of potential cyclic possibilities to explore. So a title like “Accurate de novo design of hyperstable constrained peptides” does call attention to itself.

The authors, a large multicenter team from the University of Washington and several other institutions, have a lot to prove after an opening like that, but I have to say that it’s an impressive paper. It shows a dozen examples of cyclized peptides designed through these techniques, which admittedly is not a very long list compared to the universe of possibilities. But they hit a pretty good variety, though – natural and unnatural amino acids and stereochemistries, several types of linkers, and a number of completely different geometries. It’s computationally fairly intensive work, but that’s nothing compared to the alternative:

Large numbers of peptide backbones were stochastically generated as described in the following sections, combinatorial sequence design calculations were carried out to identify sequences (including disulfide crosslinks) stabilizing each backbone conformation, and the designed sequence–structure pairs were assessed by determining the energy gap between the designed structure and alternative structures found in large-scale structure prediction calculations for the designed sequence. A subset of the designs in deep energy minima were then produced in the laboratory, and their stabilities and structures were determined experimentally.

The disulfide-stabilized natural-amino-acid proteins that they designed could be made in cell culture, and they obtained a crystal structure of one of these, and NMR structures for several others. In every case, it seems like the designed structure matched quite well with the real one, and the proteins had greatly enhanced thermal and chemical stability. The unnatural ones, of course, had to be synthesized by hard-working chemists, so the group added an extra layer of simulation to the calculations to make sure that this wasn’t going to be time wasted. One of the four design classes still managed to go off on its own a bit, on one end of the structure, but the others matched up well. Further tests were backbone-cyclized peptides and one class that has no analogs in nature at all.

This could all be very interesting – it’s sort of the second coming of the “stapled peptide” idea, but that one was best suited to presenting a helical peptide of a certain size, whereas these designs seem to be able to deliver various loops, beta-sheets, and other surfaces.  There are a lot of protein-protein targets out there that one would like to apply these to – I’d be very interested in seeing a test case (like the Bcl system) tried out with several of these new constrained geometries to see what the partner protein makes of them. It looks like one should be able to design plausible binding partners to such targets, but we’ll see what happens when someone actually tries that – and after this paper, I think we can assume that someone will. Say the authors:

The hyperstable molecules presented in this study provide robust starting scaffolds for generating peptides that bind targets of interest using computational interface design or experimental selection methods. . .it should be possible to re-engineer the peptide surfaces, incorporating target-binding residues to construct binders, agonists or inhibitors. There has been considerable effort in both academia and industry to use small, naturally occurring proteins as alternatives to antibody scaffolds for library selection-based affinity reagent generation. Our genetically encoded designs offer considerable advantages as starting points for such approaches. . .

Going beyond the re-engineering of our hyperstable designs to bind targets of interest, the methods developed in this Article can be used to design new backbones to fit specifically into target binding pockets. Such ‘on-demand’ target-specific scaffold generation is likely to yield scaffolds with considerably greater shape-complementarity than that of scaffolds generated without knowledge of the target. More generally, our computational methods open up previously inaccessible regions of shape space, and, in combination with computational interface design, should help unlock the pharmacological potential of peptide-based therapeutics.

I hope that they’re right about this, of course. Time to go find out!


40 comments on “Nailing Down Small Proteins”

  1. Kelvin says:

    Previously I developed a new class of N-methylated retro-inverso peptides to treat AD. We showed that the peptides were completely resistant to proteolysis, soluble in both water and organic solvents, and could also wizz across the blood-brain barrier and cell membranes without any problem.

    So why did the project fail? Because despite our data showing great promise contrary to expectations, the CEO (ex-Pfizer) and investors decided that the drugs couldn’t possibly work because they were technically still “peptides”.

    Some things fail not because of what they are, but because of what people believe them to be. Hubris.

    1. John Wayne says:

      As I’ve transitioned into management I’ve also noticed that most people have odd beliefs relating structure to usefulness. The medchem ‘rules of thumb’ were very useful back in the day when our ability to ask questions was very limited. We have a lot more resources now, so you can often throw things that may be useful against the wall and see what sticks. Even knowing that most major discoveries are serendipitous, most researchers are still unwilling to design experiments that challenge the current almost-certainly-wrong way of doing things.

    2. AM says:

      Kelvin – was the N-methylation in regard to making cyclic peptides, stabilizing “linear” peptides to hydroysis (while improving permeability/bioavailability), or all the above. Any key literature you’d recommend on the topic? Sorry about the pfhubris – I wish that stuff was more amendable to degradation. thanks.

      1. Kelvin says:

        No these were short (5-residue) N-methylated all-D amino acid peptides. The N-methylation alone was sufficient to make the peptides both rigid as extended beta-strands (due to steric constraints as per Ramachandran plots) and, more strangely, completely soluble in both water and organic solvents (due to their inability to aggregate (i.e., self-associate) into extended beta-sheets. Furthermore, their retro-inverso nature, as well as their N-methylation rendered them completely resistant to proteolysis.

        There are now actually quite a few papers out which demonstrate the real potential of these peptides to block amyloid aggregation as well as neurodegeneration in vivo. Just Google “amyloid N-methylated peptides”. Which kind of makes me sick that our CEO and investors didn’t pay any attention to our original data when they decided that these drugs could never work just because they are peptides, so I got kicked out, the original IP was abandoned, and the company went into liquidation. Idiots!

  2. Curious Wavefunction says:

    This is a nice paper with potential applications, but it’s worth keeping in mind that companies like PeptiDream have had the capability to make billions, if not trillions, of such peptides for years now. The bottleneck is *always* good druglike properties. From what I understand, you can almost always get impressive affinity if you make enough of them, but getting good permeability and especially good clearance and other related pharmacological properties is extremely hard, and numbers don’t seem to help you much in that regard.

    1. Isidore says:

      I am not a pharmacologist, so this may be naïve, but wouldn’t increased stability adversely affect good clearance? I mean if a synthetic peptide is, say, chock-full of D-amino acids the normal protein/peptide degradation pathways wouldn’t work well at all. Another fine balance to consider.

      1. kjk says:

        The problem is they usually get degraded too fast to be able to be pharmo-active, and increased stability just means degradation at the right speed rather than too fast. A larger problem would be different people having vastly different kinetics and having to tailor the individual doses differently.

      2. Chris Bahl says:

        This is a great point, and one worth considering carefully as we work to build therapeutic function into these peptides.

        Generally, peptides that aren’t degraded enzymatically are rapidly cleared from serum by the kidneys:
        Indeed, renal clearance may pose a significant challenge for therapeutic strategies requiring a long serum half-life.

        1. ScientistSailor says:

          Chris, why would renal filtration be a problem? That’s the lowest clearance rate you can have, without some sort of recycling. Many antibiotics are cleared at GFR: Aminoglycosides, vancomycin, many beta-lactams.

          1. Druid says:

            You are right in a sense as clearance by renal filtration should not exceed GFR with a max of about 2mL/min/kg in a young patient. However, peptides usually have a very small volume of distribution, not much bigger than the plasma volume around 40mL/kg. That gives a half-life around 14 minutes. Even if this is expanded to include most extracellular fluid, the half-life estimate is just over 1 hour.

  3. HOBT says:

    I was making disulfide bonded cyclic peptides back in the 90’s…….

    I never had a “stapler” though.

  4. Chrispy says:

    It is very interesting that they are able to make peptides that fold up as predicted, but that’s still very far from being a drug. Do we suppose that these will be orally bioavailable, that they can get into cells, and that they won’t immediately be removed from circulation by glomerular filtration? Because that’s what you’re up against if you are trying to make these into drugs. In academia it is quite common to ignore the obvious issues, say you’re making drugs (or “unlocking the pharmacological potential”), and focus on what problems you can solve (and, hey, get a big publication while you’re at it!). But that doesn’t make the problems you can’t solve go away. Practical concerns will probably leave this platform stuck with injectable formulations, Fc-fusions for half-life (and fermentation-based production, with only natural amino acids), and extracellular targets.

    So the question might well be asked: why not just use an antibody?

    I hope to be proven wrong, and maybe these peptides are as magic as Aileron’s stapled peptide.

    1. Chris Bahl says:

      You’re right, of course, that academia tends to address solvable problems; it can be remarkably difficult to convince funding agencies to support work on unsolvable problems (as well as publish these efforts). Figuring out how to make peptides into a commonly used drug platform is a challenge that will require many incremental advances in technology to overcome. While this manuscript does not directly address many of these hurdles, we believe the high degree of structural control provided by these computational tools will greatly assist downstream pharmacokinetic engineering efforts.

      Why not just use an antibody? This is an astute question, and it’s one I’ve spent quite a bit of time considering. There are many applications where antibodies make poor drugs. For example: enteric or topically applied therapeutics where stability is key. Furthermore, antibodies require refrigeration, which can make them cost-prohibitive for use in developing nations. Constrained peptides are certainly not a “one size fits all” solution to drug discovery, but we believe there are specific applications where traditional small-molecule or antibody approaches are less amenable.

      1. Chrispy says:

        Ah, you must be the C. Bahl who is third author on the paper. Cool!

        The issue that peptides, even constrained ones, have fundamental problems when it comes to drug discovery is no reason to stop doing this kind of work. Some of these problems may get solved, after all, and a potential drug with problems would be a good incentive to get people working on the solutions (like a potent Myc inhibitor that can’t get into cells, for example).

        You’ll have to forgive a cynicism among drug discovery people, though, when early research is touted as a game-changer is drug discovery. We’ve seen too many of these (combichem, in silico docking, fragment-based drug discovery, etc.). All of these things do advance our efforts, but it is all a lot more incremental and a lot less revolutionary than the early hype. I can think of only one recent platform technology (humanized mice) that really did revolutionize drug discovery (at least for antibody drugs). Frankly, what you have is really compelling even if it doesn’t add up to unlocking the pharmacological potential of these peptides.

        One thing maybe you can give some clarity on: David Baker is pretty well known for saying that most of what his lab does fails, and that 1% of designed proteins may work. This paper leaves the reader with the idea that everything works, though — what is the success rate? For every one of these 12 is there a bunch of others that didn’t work?

        1. Chris Bahl says:

          Yep, that’s me!

          You make a very fair point; early-stage platform technologies are frequently over-hyped, and the current academic reward system does little to discourage such behavior.

          It is difficult to summarize our success rate into a single number that’s broadly representative, as this is highly dependent upon the individual protein and design goal. We reported that 29 of 130 genetically-encoded designs exhibited a CD spectrum consistent with the designed structure (and we were able to determine structures for 5 of these). However, these peptides were engineered and tested in batches, and the design protocol was improved for each successive batch. The success rate is also highly dependent on topology; beta sheets are significantly more difficult to engineer than alpha helices, and the success rates reflect this.

          For the peptides incorporating non-canonical amino acids, these systems are small and rigid enough to enable a more exhaustive computational assessment of alternative conformations, and the success rate of these designs was quite high as a result (almost every design folded as intended).

  5. gippgig says:

    Note that biological production of peptides (i.e., by fermentation) is not limited to those only containing natural amino acids; it is now almost routine to create altered aminoacyl-tRNA synthetases and tRNAs to incorporate unnatural amino acids in genetically coded peptides. (In one case a biosynthetic pathway was also introduced so the bacteria could incorporate “tyrnsine” (p-aminophenylalanine) without having to add it to the growth medium.)
    On an unrelated note, I just attended a fascinating talk at the U. of Md. – Quest for the next generation of antibiotics: a bacterial mechanosensor, MscL, as a novel drug target. MscL acts as an emergency relief valve if the pressure inside the cell gets too high; it opens a huge channel to let ions out quickly. It turns out that knocking out mscL reduces sensitivity to streptomycin; streptomycin binds to MscL increasing its activity (explaining a long-mysterious potassium ion leakage upon exposure to streptomycin) and also appears to use MscL as a major way of getting into the cell (a topic that deserves far more study than it has received). Several other compounds were found that inhibit growth by overactivating MscL, including some sulfa drugs. Unfortunately, there does not seem to be any evidence that knocking out mscL reduces pathogenicity.

  6. Mr Drug Developer says:


    How close is this to being an actual therapeutic that can be used in humans? Seems like this more of research tool right now.

    1. Chris Bahl says:

      This is the key question! We’re working to generate drug candidates for a variety of targets at the moment (mostly for infectious disease), and we’re making good progress. However, few of these projects have advanced to the point of testing in animal models; there is still a lot of work to be done before one of our designed peptides shows up in a clinic.

  7. Skeptic says:

    This work is interesting but academics have no idea how to create drugs that have the attributes to be used in humans. The rigor of science they do is to publish papers, not to de-risk drug candidates to take them through the rigorous testing that goes on industry.

    1. Chris Bahl says:

      Indeed, I think you have highlighted the critical importance of collaboration between academia and industry.

      1. MTB says:

        Thank god you realize that you are not fit for industry. Academic entrepreneurs make terrible business leaders. Sticking to your competency is the best.

        1. Anon says:

          @MTB, likewise, I wouldn’t apply to become a diplomat if I were you.

        2. Science Investor says:

          From my experience, many faculty members who start their own companies have little credibility when they say the want to become CEO. I believe academics are great scientific advisor board members, but as company leaders, the success rate is very low. It’s best to have season industry experienced folks run the business. Investors like me are more comfortable because season industry folks understand how to drive to value inflection points. Academic scientist leaders sometimes follow interesting science that has little or no commercial value. Many don’t understand that investors want returns and they expect a certain amount by specified time.

          Investors don’t care about interesting science. We care about how interesting science creates a competitive advantage in solve problems that have high commercial value. I can tell you the number of times I’ve sat on pitches where the academic entrepreneur thought his cool science could fix a problem in small market that cost $500M to fully commercialize that would only make up $50M max. If the economics don’t make sense, you have no business, period.

          1. John says:

            I hope Chris Bahl is smart enough to realize the infectious disease area he is going after better have a high unmet need, and large patient volumes in the western world. Drugs that may help third world countries may be extremely impactful but since there is no money in them, investors don’t care unless they are truly altruistic.

  8. John says:

    Global Data is not a very well respected piece of market data. I suggest Decision Resources or even DataMonitor is still better than Global Data. Perhaps you should visit antibiotic policy but the reimbursement for novel antibiotics is quite poor. Antibiotics don’t get investors excited like immuno-oncology does where $150k treatment costs and applicability across many large tumors means billion dollar products.

  9. Watcher says:

    Can anyone tell me why the third author on this paper is answering for it? And why you all are so intrigued by his insight? Third authors typically are responsible for <2% of the work. Just sayin…

    1. Jay says:

      Don’t be so harsh on Chris Bahl. He is from Maine and Seattle is his large city that he’s lived in. He’s practically a country boy so there is no way he knows that Global Data is garbage. He probably is idealistic and thinks industry-academia collaborations are all fine and dandy like African Americans in St. Louis. VCs give money like they are Santa with no strings attached. He sounds like one of those “gee wilikers” nice guys who is completely naive of the outside world beyond his bubble in the lab.

      1. KP says:

        I looked at his background and definitely a country boy. Undergrad in Orono in Maine, which is basically 1990’s America still and Hanover, NH where Dartmouth is like summer camp. Seattle looks like it’s the first major modern city he’s been in. It probably seems like NYC to him after being in towns where people still ride tractors and use pay phones.

        1. Passerby says:

          Simply don’t understand how discussing the personal details of the author (with some ugly ad hominem comments) is relevant to the discussion here. I thought the level of discourse in the comments section of this blog was more elevated than that. Stick to the science.

  10. Moses says:

    Jay & KP: ad hominem comments are scuzzy. Attack the science or the business case , but not the man.

    There’s a small Cambridge UK company called Bicycle Therapeutics which seems to be doing well at developing peptide drug candidates so the idea is getting established.

    1. Bob says:

      Bicycle Therapeutics is smart in knowing that investors want to see a platform that can be applicable to oncology. They’ve also been savvy to pair their technology with drug conjugates and prove their peptides have drug like properties. Until Dr. Bahl can do that, this is just interesting science because no translation into the clinic has occurred.

  11. sgcox says:

    The comment by Watcher is arrogant to the extreme. Each author has a right to talk publicly about his work.
    Besides, according to the paper, Chris contributed equally to the work as a first author.

  12. ChrisBahlonlyHiresAsianFemaleUndergradsforhisLab says:

    Chris Bahl – Does your protein engineering cure “yellow fever”?

  13. Watched says:

    Umm, three co-first authors from the same lab is odd. From multiple labs with unique expertise, sure. It’s not arrogant.

  14. Watched says:

    That being said I appluad the paper. Not everything is a drug from the start, doesn’t mean it can’t contribute to moving the field.

  15. Kevin says:

    It seems Chris Bahl is now to afraid or too scared to reply to anything now.

    How can you transport these peptides into cells?

    1. Weary of message board machismo says:

      Dude, leave the insults to cafepharma and the like! Can we please stick to civil scientific debate on this forum?

  16. biologics guy says:

    These guys are much, much further along:

  17. Chrispy says:

    I probably started this pile-on and I’m frankly embarrassed by some of the comments here. And it turns out that Chris Bahl is one of three first authors, not third, as I stated (sorry, Chris, on both counts).

    One of the things I love about this blog is that the comments are almost never mean spirited, and they are often by highly knowledgeable people contributing positively to the discussion, The authors frequently comment, themselves. Thanks, Chris!

    Having been on both sides, I’ve observed that industry folks often think little of the work done by academics. While it is true that drug discovery is over-hyped in academia, all of us scientists gain from good work done by others. A bit of hype is absolutely required to run a big, academic lab. And the Baker lab has pumped out consistently excellent, game-changing work for many years, including this paper.

Comments are closed.