Skip to Content

Beware of Zinc. And of Other Stuff.

Here’s a paper that will give many medicinal chemists a shiver of recognition. A group from Dundee reports a screening hit against the enzyme Ube2T, a ubiquitin-conjugating (E2) enzyme. For those outside the field, ubiquitin is indeed ubiquitous. It’s a short bit of protein that gets hung onto other proteins (or removed, if it’s there already), a process that mostly happens on exposed lysine amino acids. Ubiquitination has evolved into a massive signaling system in living cells, with whole biochemical pathways set up to sense whether a protein has been so tagged, where the ubiquitin has been added, and how many times. There are three stages involved in just the adding-ubiquitin stage of things, catalyzed by E1, E2, and E3 enzymes, and all of these are under investigation, as are the de-ubiquitinating pathways on the other side of the process.

Polyubiquitinated proteins are almost always marked for the trash heap – it’s a biochemical marker for them to disappear into the proteasome to be broken up for amino acid scrap and recycled. Right now, a very hot topic indeed in chemical biology and early-stage drug research is the attempt to hijack and control that garbage-hauling machinery. We’d like to be able to target a given protein for the shredder before its time (especially proteins that we haven’t found any other way to shut down), or keep one around when it would otherwise disappear. Doing this at will and selectively is not quite a real thing yet, although some very interesting examples have appeared, but a lot of effort is going into making it one.

The reason Ube2T shows up as a target is because it’s overexpressed in many tumor lines, and is part of a DNA repair pathway. Interfering with DNA repair in general is a bad idea, since that would probably lead to tumor formation, but doing it temporarily would also make the fast-replicating tumor cells more vulnerable to radiation and some types of chemotherapy, and several of these pathways have been pursued. Targeting an E2 ubiquitin-conjugating enzyme is a tough way to do it (few inhibitors are known for that class), but the Dundee group had already reported discovery of an allosteric pocket on the Ube2T enzyme, and had run a fragment screen to find chemical matter for it. So there’s your disease, your target, your screen, and your small-molecule hits – the medicinal chemist’s equivalent of “All I ask is a tall ship and a star to steer her by”. Time to go earn that salary!

That’s exactly what everyone was trying to do in this latest paper. Their lead compound from the fragment screen was a hit in several orthogonal assays, showing reversible dose-dependent effects on Ube2T. NMR, DSF, ITC – these three-letter acronyms will gladden the heart of any early-stage drug discovery person, because normally you’d expect any compound that shows activity in all three of these to be the real thing. The group set up crystallization experiments with the compound and the protein (in just that try-all-sorts-of-things-at-once style that I referred to yesterday!), and while those were going, they bought or synthesized a series of analogs.

There the worries began. Ideally, you’d want to see the beginnings of a real SAR (structure-activity relationship) – this change did nothing, this one made it worse, this one made it better. But everything they did to the compound wiped out the binding. That sort of flat SAR is not a good sign – there are real singleton compounds out there, for which every change is a bad one. But this is also the sign of a false positive, where something different is going on than what you think.

And so it proved, sadly. They did get crystals suitable for X-ray, which showed extensive structural rearrangement of the protein. But what it didn’t show was any of the small molecule hit in there. Looking at the catalytic cysteine, though, there was a strong electron density peak, a solo atom that diffracted plenty of X-rays. . .and it turned out to be a zinc atom. The team went back and ran their nice-looking ITC experiments in the presence of the classic metal chelator, EDTA, and found that it totally abolished any activity. The starting compound was contaminated with zinc, and that’s where all the activity was coming from.

I have been down this exact same road, exactly twenty years ago. Fortunately, we didn’t get as far as letting an X-ray crystal structure tell us the bad news. We had a hit for a phosphatase enzyme (an unlikely event), but none of the analogs that we bought or synthesize showed any activity at all. I ordered up another sample of the original powder, and at the same time had someone in the group synthesize a fresh batch of it. The freshly made stuff had no activity at all – and the library sample turned out, by elemental analysis, to have a goodly amount of zinc in it. That was, of course, the source of all activity in the enzyme assay.

Always check the purity of your screening hits, folks. In this case, the NMR and mass spec came out fine for the Dundee group, which is the sort of thing that will lead you down the path. More recently, I had another situation like this, in which a compound checked out by LC/MS and NMR, but had flat SAR and batch-to-batch variability. Turned out that if you ran a bit of the sample through a silica gel plug, it looked the same analytically, but now was totally dead in the assay. A tiny amount of purple stuff was stuck to the very top layer of that silica – and that junk (which was some sort of oxidative polymer) was what was lighting up the assay. Always check the purity, and then check it again.

28 comments on “Beware of Zinc. And of Other Stuff.”

  1. Cialisized says:

    I’m sure you will get a whole list of sad stories here, like this one:
    -Screening hit on a kinase enzyme. Good so far.
    -Re ordered, no activity, but screening hit did not match structurally–was the precursor. Still hope.
    -Ordered precursor, no activity.
    -Fractionated on HPLC–All activity was in the polar solvent front.
    -Active fraction happened to be a faint bluish-green color–Some metal in there, Ni or Cu.
    -Great lesson, caught it early at least.
    -Next:

    1. metals wrangler says:

      Metals contamination causing false positives keeps people doing ICP-MS in business.

  2. GutDecipher says:

    As long as it’s not a metalloenzyme, could also do a few of the early-stage assays w/ EDTA in the buffer. Not that you always want to assume contamination, and hindsight’s 20/20, but…that’s a lot to go through to find out so late.

  3. HTS all day says:

    This shouldn’t have been solved this way: ALWAYS run a +/- EDTA buffer when checking cherry-picks from HTS. This work should not have progressed nearly this far – should have dropped out of an EDTA counter-screen a lot earlier.

    1. David Borhani says:

      HTS all day, agree 100%; you beat me to the punch(line)! For most systems, EDTA should be included in the buffer as a matter of course—it helps to prevent a lot of bad things happening to your protein and nucleic acids (especially RNA), and can protect against false hits like this one . Should be (it used to be…) standard practice. Even with metalloenzymes, one can usually craft a metal-buffer to achieve the desired protective effects without compromising genuine activity.

  4. Billy says:

    “I have been down this exact same road, exactly twenty years ago.” It’s like you read my mind and quoted what I was thinking, Derek.

  5. Peter Kenny says:

    I recall that heterocyclic system from when I was designing a small library of fragments for targeting metalloenzymes. They were are bit unlucky that it was diamagnetic zinc because some other metals would have made their presence more obvious in the NMR. I wonder how easily one would pick up impurities like these in the mass spec?

    1. Some idiot says:

      Hmmm… Good point. Lots of metal adducts (Na, Fe etc) are often seen in MS, but I’m not sure that Zn is all that common… could be worth keeping an eye out for…!

    2. Roger Moore says:

      You could always try something like ICP/MS. It’s fast, sensitive, and capable of detecting a wide range of metals, which seems like exactly what you’d want. It’s also something the FDA has started to require for QC of production drugs, so there is probably already some in-house technical capability.

      1. HTSguy says:

        ICP/MS is what these folks did:
        http://dx.doi.org/10.1021/ml3003296

  6. Chrispy says:

    Wow, negative data getting published. That’s great.

    I have to think, of places I’ve been, the Big Pharma would have quietly shelved this, and the biotech would (probably) have tried to raise a bunch more venture capital for their “hit.”

    1. John Wayne says:

      “I have to think, of places I’ve been, the Big Pharma would have quietly shelved this, and the biotech would (probably) have tried to raise a bunch more venture capital for their ‘hit.'”

      That is hilarious … because it is true. Crap.

      I’ve had a few of these goose chases myself: the ones I can remember are (1) palladium acetate was shockingly potent (2) screening hit was a useless decomposition mixture of the intended compound (goes faster in DMSO), (3) wrong chemical from vendor, and more ‘we could hunt this down but it isn’t worth our time’ false hits than I can count. I’ve worked on a lot of phenotypic screening, so this is more of a thing.

      Don’t worry about lessons learned big pharma. All those experienced people you keep letting go in favor of the young and naive definitely will not spend shareholder’s money rediscovering things like this. It will be fine.

      1. Dr. CNS says:

        Agreed.
        Surety and purity are the first things you should check in a compound.
        When I asked a group of academic biologists how they check these in the compounds they buy from external sources, the response was: “well, we don’t. If there’s a problem, we will find out down the road.”

        Boy, do I miss the days of internal chemistry labs…

        1. Some idiot says:

          To quote a friend and former colleague (process chemist), said with a smile: “NMR and LCMS look good? Good start. Now let’s wait for the CHN results to see if it is pure…”

          😉

  7. hn says:

    Yes, thanks for publishing this paper so we can all learn and relearn this lesson!

  8. milkshaken says:

    In the old days, when glycoside forming reactions have to be promoted with a stoechiometric quantity of heavy metal (Hg or Ag), there was plenty of irreproducible activity published on antiviral nucleotides

  9. Gene says:

    In the environmental field, zinc is ubiquitous. If I get an analysis report with extremely low or zero zinc, I suspect it.

  10. Biologist asking? says:

    And an academic to boot! Glad to see this topic though.

    My group has been using NMR, MS3 and HPLC-PDA (important to check normal AND reverse phases). Even so, some junk still manages to get into my screening queues (probably because we’re doing things “in just that try-all-sorts-of-things-at-once style”).

    With so many ways to assay purity and often not enough resources ($$$ / FTEs) to run them all, what do folks like to see before sending samples off for screening? How do you all suppose that might change if you were in an academic v startup v big pharma lab?

    1. Mol Biologist says:

      Elevation of E2 ubiquitin-conjugating enzyme in many cancer is a consequence but not a reason. Falconi Aneimia (FA) characterized by deficiency for the ubiquitin-conjugating enzyme (E2), UBE2T. https://www.ncbi.nlm.nih.gov/pubmed/26119737
      FA is also associated with chromosomal fragility, aplastic anaemia, congenital abnormalities and a high risk of cancer. Ubiquitination is very fascination area of human biology and alter protein cellular location, affect their activity, and promote or prevent protein interactions. In my opinion, the increase in expression UBE2T in advanced cancer is the absence of feedback due to metabolic changes. But at the onset of the cancer, the situation is likely to be reversed and characterized by a deficiency of UBE2T.
      https://www.ncbi.nlm.nih.gov/pubmed/?term=hypoxia+and+UBE2T

  11. Jeff says:

    So, sometimes (rarely) this sort of goes the other way. Back in the day when they still let me in the lab, I was perhaps a little generous with the tertrakis in my suzuki reaction and a little less observant of my NMRs and LCMS’s than I should have been. I too saw funny SAR which culminated in my boss remaking the compound (over Christmas I think) via a different route leading to the same compounds but much less active. Turns out that when you make a HCV NS5a inhibitor accidentally that is picomolar, even in tiny quantities (ie less than 1% as a result of homo-coupling) your “parent” compound looks nM and interesting. Damn you Strem and your beautifully yellow tetrakis. On the plus side, we did figure that all out and so began a very cool project and some other people eventually launched Harvoni, cured lots of people and made a ton of cash. But not us.

    1. drsnowboard says:

      Is that you, Jeff…? The other fortunate thing was the impurity could be isolated as the tetrahydrochloride easily when you made it on purpose. Stuff was brick dust otherwise…

  12. Anonymous says:

    As an undergrad, I was doing so many (biophys) experiments that I decided to switch from using the 500 mL glass bottles of solvent to the 4 L metal cans (all from the stockroom). As the PI incorporated all of my old and new data into a paper, there were a bunch of crazy outliers and he was, um, annoyed. He let me go back to try to figure out what was wrong and I repeated and compared experiments. I did not prove (AA, ICP, etc.) that it was a metal contaminant from the can, but the glass-to-metal switch definitely affected the assay results. I repeated the runs using glass and things fell back into place. It was a lesson about impurities and how to conduct research that I never forgot.

    In re: the biotech comment which, I agree, is too true. One of our compounds run in an inhibition assay, in duplicate, produced 100% and 0% inhibition. Some of you might think that the assay should be repeated or further validated, but that is why you are schlubs and not in management. The math went: “ignore the zero and just announce that we have a potent inhibitor.” (It also ignored data from other control screens that indicated that this was just some non-specific toxic compound.) The ruse successfully suckered in several more millions and the hit turned out to be an impurity (a reagent that had not been completely washed out – I think that one was Ac2O. That did not explain the 100/0 result, but it helped to put a nail in the coffin.).

    One of my favorite chapters from Primo Levi’s Periodic Table is Chromium and I required all new students to read it.

  13. David Stone says:

    Zinc as a contaminant is not terribly surprising to me, given the widespread use of galvanized metal duct work in many buildings. I know someone who wanted to look at some proteins with potential zinc centres here but, given the levels of zinc in the blanks compared to the levels they wanted to test for in their samples, it wasn’t going to work. They would have had to do the whole expression, isolation, and measurement in a clean room if they had wanted to do the analysis in this building.

  14. Nerd v geek says:

    A group from Roche reported on Zn impurities causing false hits in HTS some years ago (10.1021/ml3003296) so not really a surprise. At least the authors from this paper reference the Roche work.

  15. Mister B. says:

    It’s not the first time Derek highlight this kind of bad story. Are there any good routine analysis that can prevent this type of contamination ?

    Elemental analysis is not suitable for large chemical librairies, but still…

    And second question. From a HTS, what is required to prevent to go too far with a “contaminated hit” ? Re-synthesis it and re-test it ?

  16. MrRogers says:

    And occasionally it goes the other way. After identifying the active component of a natural product extract, and determining that it was active in our cell-based assay, we ordered some from a well-known supplier for animal studies. After the first dose all of the animals were dead. A quick question to the animal tech determined that the commercial material was much darker in color (dark grey) vs. the original isolate (white). The chemist had seen a clean NMR and thought nothing of the color difference. One preparative HPLC later, we were back in business and are making analogs now.

  17. Wheels17 says:

    The earliest reference I know of for this type of problem is from 1886, “Strange Case of Dr Jekyll and Mr Hyde” by Stevenson. For him to have based the story on the effect, one would think it must have been known at the time. “My provision of the salt, which had never been renewed since the date of the first experiment, began to run low. I sent out for a fresh supply, and mixed the draught; the ebullition followed, and the first change of colour, not the second; I drank it and it was without efficiency. You will learn from Poole how I have had London ransacked; it was in vain; and I am now persuaded that my first supply was impure, and that it was that unknown impurity which lent efficacy to the draught.”

  18. NovAliX says:

    Moral of the story. Always use native mass spectrometry if possible. Simple, cost-effective, definitive.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.