Skip to main content

Analytical Chemistry

The Good Stuff Goes One Way. . .

I’ve always like the idea of aptamers – as generally used, that word refers to oligonucleotides that are selected for binding to something else (a protein target, for example). You get to use all the tools of molecular biology, which means that you can start out from insanely huge numbers of possible binders and select your way down (a bit like using DNA-encoded small molecule libraries, and feasible for the same kinds of reasons). The process is usually referred to as SELEX (systematic evolution of ligands by exponential enrichment), and it consists of solid-phase binding experiments to enrich for potential candidates, amplification of those, and further more stringent rounds of selection binding. If you’re using RNA aptamers, there’s a reverse-transcriptase step to send you into DNA space for the amplification by PCR; DNA aptamers can just roll right along. You can incorporate some mutation in the protocol if you like, and overall the number of variations on the SELEX idea would at this point be hard to count.

It can work dramatically well, if you’re careful to watch out for tight but nonspecific binders and so on. The resulting aptamers are useful reagents, although it’s generally hard to imagine them as drugs, being nucleic acids. If you could come up with equivalents for reverse transcriptase, PCR (high and low fidelity), and the rest for rounds of selection for more conventional small molecules, you would be in business for sure, and you’d put the rest of us out of business. DNA-encoded library screening is about the closest we’ve gotten.

The SELEX process, though, can be a lot of painstaking work, and from what I know about it, the majority of the time you don’t actually zoom in on any binders. One problem that’s been suggested is that the PCR step is actually a bit less likely to enrich the binding sequences, which may be more structured than the nonbinders, and there are influences from the composition of the starting library as well. But this new paper appears to be a rather startling improvement. It proposes “ideal filter” enrichment by capillary electrophoresis. In a standard CE experiment, just like with any other column or gel, you enrich by having one population move more quickly than another down the length of the device. The nonbinders wash off, for example, leaving the binders behind – but that’s a relative process, and if you have a bazillion nonbinders in the background with varying mobilities, things can get messy and smeared-out. This new technique actually makes the binders and nonbinders move in opposite directions, though. Which is quite a picture to contemplate, for those of us who are used to standard chromatographic column thinking.

The key is to increase the ionic strength of the running buffer. You’re tuning the mobility of the binders and nonbinders to the electroosmotic flow, and when you hit the right balance, you can get the ideal-filter effect. You also reduce the amount of binding to the sides of the capillary tube, which is one of those nonglamorous issues that can actually complicate assay technology a great deal. The authors showed this with a DNA repair protein called MutS, for which an aptamer (found by three rounds of traditional SELEX) was already available. Experiments with that one versus controls were very promising, so they tried a selection from a random-sequence library. One round of the high-salt conditions pulled out a group of aptamers with double-digit nanomolar affinities, while everything else went away. The signal/noise improvement was dramatic, roughly ten-million-fold better partitioning than the normal conditions.

So that sounds pretty useful, and as the authors point out, you might well be able to use these conditions for any other system where you have DNA attached to things – such as DNA-encoded library screening itself. I hope that this turns out to be as general a technique as it would seem!

10 comments on “The Good Stuff Goes One Way. . .”

  1. anon3 says:

    I’m probably just misunderstanding, the high ionic strength helps remove non-specific aptamers, but with DNA encoded libraries the non-specific binding occurs via the ligands, not the DNA…hard to see why this would work.

  2. The high ionic strength (which is actually very close to that of PBS and is only high for capillary electrophoresis) reduces the mobility of electroosmotic flow, which, in turn, results in target-bound and target-unbound DNA moving in the opposite directions inside the capillary. The movement of the bulk of unbound DNA away from the capillary outlet drastically reduces the selection background.
    It is worth emphasizing that the method is homogeneous; accordingly, there is no an issue of non-specific binding of ligands (or DNA) to the surface, that makes classical bead-based selection so inefficient.

    1. Derek Lowe says:

      Glad to see you here! I was wondering if your group has tried out the application to DEL screening mentioned at the end of the paper, and if you have an idea of how tightly bound such DEL hits might have to be in order for this technique to be successful?

  3. Sergey Krylov says:

    Thank you very much for bringing IFCE to your “discussion table”! We have not tried selection of hits from DELs yet; a random-sequence DNA library was a simpler model for method development than a DEL. However, 1-step selection of hits from DELs, with an expected hit purity of >99%, is, of course, our strategic goal (the current industry standard, to the best of our knowledge, is 3 consecutive rounds with a final hit purity of 1-2%).
    With regards to “how tightly bound such DEL hits might have to be in order for this technique to be successful” I can write the following. IFCE-based selection works identically for random-sequence libraries and DELs. It has intrinsic selectivity for koff – making migration inside the capillary longer biases the procedure towards selection of complexes with lower koff values. However, due to the very low nonhit background, even the hits with high koff values will not be lost if they have large numbers of copies in the starting DEL. To conclude, I do not see any specific limitation that IFCE would impose on Kd and/or koff of the selected hits. I hope this answers your question.
    Let me briefly explain our plans. Ryan Hili (my colleague at York) and I have just started a 3-year NSERC-supported project on the development of 4 technologies to facilitate automated “manufacturing” of validated hits from DELs. This project is our collaborative effort with GSK, SCIEX, and Alphora Research. The 4 technologies include: 1) a non-aqueous approach to production of DELs, 2) 1-step selection of high-purity hits from DELs by IFCE, 3) combined continuous-flow synthesis/ continuous-flow purification of DNA-free hits, and 4) accurate measurement of Kd by “Accurate Kd via Transient Incomplete Separation” (AKTIS).
    AKTIS is a disruptive technology which is free of sources of inaccuracy that inevitably affect current industry-standard techniques (SPR, BLI, and ITC). We completed AKTIS proof-of-principle and archived a full-size manuscript describing this approach ( A concise communication on AKTIS is now under review by Angewandte. We also filed patent applications for both IFCE and AKTIS.
    Overall, we look forward to very interesting research and hope to develop robust and rugged technologies that would make generation of validated hits from DELs work more like “manufacturing” than “discovery”.

  4. Scott says:

    “(the current industry standard, to the best of our knowledge, is 3 consecutive rounds with a final hit purity of 1-2%).”

    Ewwww… That’s … *horrible*. Any improvement on that would be immense! (Though I’m going from a rough knowledge of uranium enrichment, where getting up to 20% is a pain, but getting from 20% to 90+% is much faster.)

  5. Alex Taylor says:

    1-2% is actually not a problem – this is the yield of single sequences within the selection pool. If 1-2% of the pool are the binders you are after, it makes their identification by deep sequencing (or even screening by a plate-based method) fairly straightforward.

  6. Alex Taylor says:

    I do think this technique could be very powerful, the idea of running a selection in a matter of hours is very exciting! My only concern is a general one that applies to partitioning approaches to SELEX – it is likely that the best possible binders resulting from a standard SELEX experiment will be variants not present in the starting pool, but which arise from beneficial mutations accumulated during (imperfect) replication. Nonetheless, even if one were to perform e.g. mutagenic PCR on the separated pool and repeat the selection using this approach, the potential to significantly reduce time to identification of good aptamers seems very promising.

  7. Sergey Krylov says:

    1-2% is not a problem for aptamers, but becomes an issue for DELs, when re-synthesis and testing the re-synthesized molecules for target binding are required and are resource-consuming.
    Indeed, one-step selection makes evolution impossible, and, of course, IFCE can be used in multi-round SELEX with low-fidelity polymerases. The high efficiency of partitioning of IFCE should only improve the evolution process. It would be interesting to compare multi-step IFCE-based aptamer selections using polymerases of considerably different fidelity.

    1. anon says:

      With an initial batch of identically sized probes, could you preferentially separate candidates based on how folded they are? These would have systematically higher mobility, right? One can envision all sorts of MEKC style arrangements based on this work. Great stuff!

  8. The short answer in “no”. Ge-free electrophoresis does not have sufficient resolving power to separate same-length DNA of different conformations.

Comments are closed.