Skip to main content


Into the Clinic. And Right Back Out.

Here’s a good example of why all of us in the industry tiptoe into Phase I trials, the first-in-man studies. A company called SGX, recently acquired by Eli Lilly, has been developing a kinase inhibitor (SGX523) targeting the enzyme cMET. That’s a well-known anticancer drug target, with a lot of activity going on in the space.
SGX’s specialty is fragment-based design, and they’ve spoken several times at meetings about the SGX523 story. The starting point for the drug seems to have come out of X-ray crystallographic screening (the company has significant amounts of X-ray synchrotron beamline time, which you’re going to need if you choose this approach). They refined the lead, in what (if you believe their presentations) was a pretty short amount of time, to the clinical candidate. It seems to have had reasonable potency and pharmacokinetics, very good oral bioavailability, no obvious liabilities with metabolizing enzymes or the dreaded hERG channel. And it was active in the animal models, however much you can trust that in oncology.
So off to the clinic they went. Phase I trials started enrolling patients in January of last year – but by March, the company had to announce that all dosing had been halted. That was fast, but there was a mighty good reason. The higher doses were associated with acute renal failure, something that most certainly hadn’t been noticed in the mouse models, or the rats, or the dogs. It turns out that the compound (or possibly a metabolite, it’s not clear to me) was crystallizing out in the kidneys. Good-looking crystals, too, I have to say. I can’t usually grow anything like that in the lab; maybe I should try crystallizing things out from urine.
Needless to say, obstructive nephropathy is not what you look for in a clinical candidate. There’s no market for instant kidney stones, especially when they appear all over the place at the same time. The patients in the Phase I trial did recover; kidney function was restored after dosing was stopped and the compound had a chance to wash out. But SGX523, which was (other than its unlovely structure) a perfectly reasonable-looking drug candidate, is dead. It didn’t take long.

38 comments on “Into the Clinic. And Right Back Out.”

  1. Lucifer says:

    That does not sound too bad.. Sulfanilamide had similar issues, but it (and it’s analogues) changed drug research forever.
    They just have to play around with analogues and isosteric replacements.

  2. darwin says:

    Another example of the crapshoot element of the business. Only strong science, persistence and a little luck can bring medicines to market.

  3. Ty says:

    Wonder what strong science could have predicted this particular clinical finding.

  4. weirod says:

    “Wonder what strong science could have predicted this particular clinical finding.”?
    How about a simple solubility test?
    When you dose bricks, stuff like this will happen.

  5. DLIB says:

    Actually that seems eminently testable in a test tube. Can you think of a reason not to test for crystallization in urine??? In fact, why not a urine crystallization screen including Asparagusic acid and it’s derivatives, and whatever else might come out during a super bowl party. Very cheap reagents, unless you buy them and the standards from Perkin-elmer.

  6. Muruga says:

    As Derek said, SGX523 is a highly selective kinase inhibitor with reasonable PK and preclinical efficacy. It is a bad luck that the compound promotes acture renal failure through direct crystallisation in the kidney. This is probably compound specific / class specific. We can expect that this problem will be addressed through the back up of SGX523 (with different chemotype?).

  7. Daniel Newby says:

    The kidneys have many active transporters and passive diffusion channels, which they use to filter all sorts of junk out of the blood and then put some of it right back. If some transporter grabs the drug and tosses it out like gangbusters, or if a reuptake pathway ignores it, the drug will rapidly reach a high concentration in the urine. It isn’t the composition of the urine that causes crystallization, but the process of making it.

    Animal tests don’t help as much as we would like because humans have major differences in our kidneys. For one thing, we use much higher levels of uric acid than most animals. Problems like this is why Phase I trials use low doses and exhaustive tests to uncover problems.

  8. MarkM says:

    “Here’s a good example of why all of us in the industry tiptoe into Phase I trials”
    I think a better statement would have been “why all of us *should* ….tiptoe…….”
    I have known my fair share of firms that rush into pI solely to satisfy the timelines of investors and secure more funding.

  9. CMCguy says:

    #3 Ty’s comment is a good one, largely due to factors covered by #7 DN, and goes to why drug development can be subject to high failure rate. The only true model of Humans are Humans, and even then have to deal with degrees of individual variability. Models can aid but will all have limits, particularly with ADME considerations. Appears like SGX did/does have a back-up compound in fairly advanced level which always reflects sound strategy (both for science and business).
    BTW I know Milkshake has an excellent story related to this conversation which I trust he will share here or provide exact link.

  10. Jose says:

    Rapid, simultaneous deposition of crystals throughout the kidneys? Ouch, that has got to sting!

  11. anon the II says:

    Does anybody have a quick link to a structure of this thing? I don’t have library access.

  12. David says:

    Ouch! What an unpleasant surprise for a drug to cause acute renal failure. And you would think that such a potent drug would likely have a low dose for the start of a phase I study.
    I recall that Astra Zeneca’s early antifol from about 9 years ago caused renal impairment and was abandoned. They eventually tried a different analog which did not harm the kidneys. (Tomudex, which passed muster in Europe but not the US for the treatment of colon cancer.)
    I wonder if they can tweak the structure of this one just so to maintain activity but greatly reduce crystallization in the urine. Might be worth a try.

  13. Wavefunction says:

    So what was the exact difference between Homo, Canis and Mus?

  14. anon says:

    I was at a recent meeting and saw this work presented:
    They believe the crystals are from a metabolite, not the parent molecule.
    Suspected metabolite apparently is a human and monkey specific metabolite, explaining why they didn’t see it in their preclinical rat and dog tox.
    The metabolite has a very high melting point, giving further evidence of it’s propensity to crystallize.
    My guess is that it was a low level metabolite in their met id studies and they were in such a rush to get to the clinic, they only did rat and dog tox. Also probably trying to save the cost of a monkey tox study.
    After it fell out of the clinic, they’ve gone back and done a monkey study and see the crystals in the the monkey kidneys.
    The speaker declined to identify the metabolite.

  15. RB Woodweird says:

    “Actually that seems eminently testable in a test tube. Can you think of a reason not to test for crystallization in urine??? In fact, why not a urine crystallization screen including Asparagusic acid and it’s derivatives, and whatever else might come out during a super bowl party. Very cheap reagents, unless you buy them and the standards from Perkin-elmer.”
    Is PKI in that business?

  16. bio btech says:

    No big surprise, SGX had about the stupidest and most arrogant management team around. These knobs blew threw hundreds of millions of VC and public money, only to unload the whole pile of turd on LLy for $3/share (was ~$40 mil net of cash).
    The sheer number of these garbage biotechs in the past has made funding for companies with real prospects, and less arrogant management, tough. Pity.

  17. opsomath says:

    Given that post Milkshake did on how pissing in the copper sulfate batch reactor facilitated the production of glorious blue crystals, you’d think the urine-cryst test would be standard by now.

  18. Retread says:

    The normal kidney is very good at concentrating things. Probably you all know this, but in the 70 kiloGram man (154 pounds) beloved of physiologists everywhere , a normal glomerular filtration rate is over 60 cds./minute. This is the fluid entering both kidneys for further processing. 60 * 60 * 24 is 86,400 cc. or 190 pounds. So even if you put out 3 liters of urine/day your kidney has done a lot of concentrating.

  19. I’d be extremely surprised if SGX don’t know whether the crystals (for which they have determined melting point) are parent or metabolite so their apparent reluctance to reveal this information raises all sorts questions. When this sort of thing happens it’s best to come clean and be open because it will be very difficult to gain trust the next time around.
    With respect to renal precipitation, there may be species differences in pH. I recall some species differences in Phase 2 metabolism possibly involving cats (inability to glucuronidate or sulfate?). It was a while back.

  20. anon the II says:

    Thanks to Kinase Pro,
    for the structure of SGX523.

  21. jose says:

    Looking at the struc, is *anyone* surprised? what did they dose in? hot DMSO? Acetic acid?
    “Into the Kidneys. And not right back out.”

  22. anon III says:

    this reminds me the Melamine added to milk to boost the N concentration in china

  23. befuddled says:

    I’m sure they know if it’s the parent compound or a metabolite. After all, they can always crystallize the offending substance from the patients’ urine.

  24. T says:

    Hmmm. It seems as if kinase inhibitors are universally appalling structures. Are there many examples that don’t have an overabundance of heteroaromatic rings in them?

  25. MarkM says:

    thanks Anon II for the structure link; sure looks like a brick to me.

  26. CMCguy says:

    Why is this structure so much worse/”unsurprising” that many known drugs or other candidates? I see handles/functionality that should aid formulation/aqueous solubility. Till you tell me explorations for possible salts and a logP value I would not be quick to jump on this. There does appear to be potential for metabolite generation but again without some actual results it just guess work, albeit experience based.
    All too frequently the med chemists do seem to come up with molecules that are “bricks” because when try to add groups that might aid physical properties the activity goes away. Sometimes requires excellent Formulator’s effort to progress but like almost everything deal with constant compromises and rarely get the ideal but get to the most workable solutions. Nature of the beast.

  27. Tom Womack says:

    @T : you mean, there are drugs which don’t look like octopi made out of small heterocycles stuck together?
    I’m in crystallography software development, so run through the PDB depositions each week; I suppose this biases towards structures developed by crystallography, so to fragment-based design, and so to collections of things with unpronouncable names starting ‘pyr’ connected together in unaesthetic ways.

  28. milkshake says:

    kinase slit-like binding sites like heterocycles. Since the binding sites extend quite a distance, it is relatively easy to build up a potency and selectivity by Suzuki/Buchwald N-arylation, by makind amides and ureas etc.
    Example of a not-so-awful kinase drug: Fasudil

  29. Pragmatist says:

    Isn’t this the exact reason you would want to get compounds into the clinic as soon as possible???
    If you tip-toe anymore you waste a lot more time, effort and money. Plus, any test you could possibly run would not definitively prove that the drug would crystallize out in the kidney’s.
    Ensure patient safety with proper TOX analysis, then into the human!!

  30. Wijit says:

    Any news on a very similar looking compound from J&J? Replace the S with a CF2 to get it. Last I heard it was in phase I during 2008, wonder if that had similar issues? May provide a clue…..

  31. milkshake says:

    We actually found this class of potent Met inhibitors at Sugen, we just did not get enough time for getting good properties in the last few months of the post-merger slaughter, the JNJ compounds are based on stuff disclosed in our patents.
    The main difficulty was that this class of compounds was that they were too selective: in wild c-Met they were single digit nanomolar but thy got inactive in more than half of most-commonly found mutant forms of c-Met seen in tumors. It has to do with the non-classical binding mode of these compounds (it does not bind in the hinge region and instead interacts with Tyr in the activation loop and puts a lid on the ATP binding site). These interactions are quite easily disrupted by a single point mutation, even quite far from the ATP binding site.
    We had unrelated series of potent benzyloxy-substituted aminopyridines that were broadly active on mutant forms of c-Met, those were advanced in San Diego in former Agouron for awhile before dying quietly. You see, the Pfizer folks there had their own series for which that cared more.

  32. Orgmed says:

    I am sure JNJ -CF2 should not exhibit any such issues.
    If SGX pre-clinical model is Rabit (slightly alkaline urine) then it would have confirmed much earlier.

  33. Shop says:

    Why are you so sure of the fate of the J&J compound?

  34. Orgmed says:

    Gem-difluoro is metabolically stable site. Thats the only reason. Otherwise not sure if some thing else is an issue in SGX-523?.

  35. shop says:

    So that assumes the thio linker is the site of metabolism? That, of course, may not be the case

  36. Norepi says:

    Jeez, with all of the nitrogens, this thing looks like it might explode if heated too hot.

  37. molkshake says:

    you need all those nitrogens because the key interaction is pi-stacking with Tyr in the activation loop. See, the more el-deficient the system is the stronger this interaction is. You get charge-transfer complexes best between el-rich+el poor partner, like trinitrobenzene and naphtalene.

  38. Anonymous says:

    # 37- Trust me. I work with DNA-intercalating compounds, I know all about the pi-stack. 🙂

Comments are closed.