杏吧原创

WHEN the lottery balls fall, anyone who鈥檚 bought a ticket can鈥檛 help willing the outcome to match their numbers. The reality is that the odds are heavily stacked against any particular combination coming up. In Britain鈥檚 Lotto, for example, players pick six numbers between 1 and 49, so the chance of any given ticket hitting the jackpot is about 1 in 14 million.

Chemistry has a lottery of its own: finding new medicines. Organic chemists in the pharmaceuticals industry synthesise vast numbers of new molecules in the hope that one of them will become a blockbuster drug. But the odds are stacked against them. Only one compound in many tens of thousands will turn out to have the right biological properties to become a treatment for patients and win rewards for its discoverers.

But just as in a real-life lottery, there鈥檚 a way of increasing your chances of winning: buy lots of tickets. The equivalent of this in drugs research is to synthesise as many different novel molecules as possible, as rapidly as you can. Putting this into practice is what combinatorial chemistry is all about, and over the past few years it has revolutionised the search for new drugs. Because it takes at least 10 years to take a new drug through development and onto the market (see 鈥淭he long road to a winner鈥�), there aren鈥檛 yet any well-known drugs that owe their existence to combinatorial chemistry. However, there are many promising compounds in development.

Combinatorial techniques have also been used in other areas of synthetic chemistry, notably to discover new materials such as polymers, semiconductors and catalysts.

Before combinatorial chemistry was invented, medicinal chemists had to synthesise new molecules one at a time. First they would take two simple and readily available precursor molecules and react them together to build a more complex molecule. Then they would take this structure and react it with another simple precursor to generate an even more complex molecule, and so on. As the molecule鈥檚 complexity increased, so did the likelihood that it had never been synthesised before and might be a new drug.

Using this step-by-step process, chemists were able to create a large number of different structures guided by design principles and their knowledge of the drug target. But it was slow going. Each step involved setting up a chemical reaction with a carefully controlled combination of reagents, solvent and temperature. After each reaction, the product had to be isolated and purified and its structure analysed to ensure that it was indeed the right compound. A full-time lab chemist typically made just two or three new compounds a week. Given that estimates suggest that there are 10¹⁸ drug-like molecules out there, only a few of which will fulfil their potential, it鈥檚 clear that even a full-time chemist won鈥檛 hit the jackpot very often.

But with combinatorial techniques, that same chemist can now synthesise hundreds, even thousands of new compounds a week. How come? Recall that new compounds are really just a combination of simple precursors. Rather like lottery balls, there are huge numbers of possible combinations, each one structurally distinct and therefore potentially having different biological activity. For pharmaceutical purposes, it鈥檚 the structure of the drug that is important rather than its chemical reactivity. Drugs generally work by binding to receptors or enzymes in the body through weak interactions (hydrogen bonds, for example) rather than through the creation of strong and permanent covalent bonds. To increase the chances of finding the biologically active molecules among all the other variants, chemists in universities and industry have developed techniques that allow them to make every combination of a set of precursors. This is the essence of combinatorial chemistry. Chemists call such a full set of products a library.

As an example, take the reaction between a primary amine (RNH₂, where R is any organic group) and an aldehyde to form a secondary amine (R₂NH). Peruse any catalogue of commercial chemicals and you will see hundreds of primary amines and a similar number of aldehydes (organic compounds containing the group -CHO). Given the right conditions, almost any combination will react together. So by choosing any 10 amines and any 10 aldehydes, you can generate 100 new secondary amines. You could then take this mixture of secondary amines and carry out another reaction鈥攚ith an acyl chloride, for example, to form an amide (containing -C(O)NR₂). With 10 different acyl chlorides you can now make 1000 different amides.

In practice, the term 鈥渃ombinatorial chemistry鈥� covers a vast range of different techniques, but they can be divided into two distinct classes. The older and more established uses solid-phase synthesis to make new compounds. The other class is based on solution-phase synthesis.

Solid-phase synthesis is a neat way of solving the main problem with reactions between mixtures of starting molecules, such as the one described above鈥攈ow do you separate out the products at the end? To make this possible, the reactions in solid-phase synthesis take place on a polymer support, usually polystyrene beads approximately 100 micrometres across. Because of the way the reaction is set up, each bead ends up carrying just one compound. This means you can carry out all the possible reactions with a given set of precursors and easily separate the products at the end.

The first step is to covalently bind a chemical precursor to the polymer itself by means of a linker bond that can be broken at the end of the synthesis. Then you immerse the beads in a solvent containing another precursor. This penetrates throughout the bead, allowing a coupling reaction to take place between the precursors.

Solid-phase synthesis was invented in the 1960s as a way of making short chains of amino acids called peptides (see Figure). Its pioneer, Bruce Merrifield of the Rockefeller Institute in New York, won the 1984 Nobel Prize for chemistry for his work. Merrifield wanted to perform the step-by-step linkage of amino acids to form a peptide chain with a predetermined sequence鈥攍et鈥檚 call it ABCDE. He started by covalently attaching amino acid A to polystyrene beads. The molecules of A had been chemically 鈥渃apped鈥� so they wouldn鈥檛 react together to form chains of AAAAA. Next, Merrifield washed the beads thoroughly to remove any unused A, 鈥渄eprotected鈥� the attached As (chemically removed their caps) and then exposed the beads to amino acid B under the right conditions for a coupling reaction. The result was the dipeptide AB. Then he repeated the wash/ reaction cycle with the amino acid C, then D, and so on, until he got the sequence he wanted. Finally he liberated the peptide chains from the polymer beads.

Clever, but not yet combinatorial chemistry. That had to wait until 1982, when chemist 脕rp谩d Furka of the E枚tv枚s Lor谩nd University in Budapest was searching for a way to prepare all the possible peptides that are four amino acids long, made up from the 20 naturally occurring amino acids. That鈥檚 20⁴, or 160,000 different sequences in all. Preparing every single one would take years of work. Was there a way of preparing the entire set of compounds much more quickly?

The method Furka came up was an adaptation of Merrifield鈥檚 solid-phase synthesis. First he took a large quantity of polymer beads and split them into 20 equal portions. He tagged each portion with one of the 20 amino acids. After washing the portions thoroughly to remove any unreacted amino acids, he recombined the beads and mixed them together. Then he split them back into 20 portions and carried out coupling reactions to form the full set of 400 dipeptides. Carrying out the cycle twice more produced the full set of 160,000 tetrapeptides鈥攚ith such a large number of beads, there was sure to be at least one of each different sequence. Furka called this method split-mix (see Figure).

At the end of the split-mix process Furka鈥檚 library consisted of a large number of polymer beads, each carrying a pure sample of one of the tetrapeptides. He could take any bead and either test the compound while it was still attached to the polymer or release its sample using a decoupling reaction and then test it in solution. Any compounds that proved interesting could then be analysed to find out their structure.

The ingenuity of the split-mix method is that it can generate huge numbers of different peptides using only a few chemical steps. It remains one of the most important techniques in combinatorial chemistry, and is particularly valuable for finding the peptide sequences that bind most strongly to enzymes, receptors or antibodies. For example, in 1991, Kit Lam at the Arizona Cancer Center used this method to discover the peptide sequence recognised by an antibody to the hormone &bgr;-endorphin.

Solid-phase synthesis makes it easy to separate the resulting compounds from one another. Each bead contains only one, often in sufficient quantities to assay for biological activity. But if a compound shows interesting activity, how can you tell which one it is out of the thousands in the library without analysing its structure?

One solution is to label each bead in a way that unambiguously identifies the compound on it. To do so, chemists use a parallel reaction process to tag the polymer bead with a code that can be deciphered if the compound turns out to be interesting.

One tagging process devised by chemists at Columbia University in New York in 1993 works like this. After each coupling step, a new tagging molecule is attached directly to the bead in a separate chemical reaction. This molecule identifies the previous coupling reaction which that particular bead has been through. At the end of the synthesis, the bead carries its library compound plus a series of ID tags. If the library compound turns out to be interesting, you can 鈥渋nterrogate鈥� its bead by chemically removing the ID tags and identifying them using a separation technique such as chromatography. The particular combination of ID tags is like a fingerprint, describing the sequence of steps that bead went through, and thus the structure of the compound. Chemists can then resynthesise that compound in large quantities.

However, from the perspective of drug discovery, peptides are not all that useful. They do not usually make good drugs because they are broken down in the stomach. So in the early 1990s, organic chemists started to look for other types of synthesis that could be done on polymer beads.

The first solid-phase synthesis of non-peptide compounds was developed by Jon Ellman at the University of California, Berkeley. In 1992, he achieved the synthesis of the benzodiazepinone class of anti-anxiety drugs by reacting three different precursor chemicals in a stepwise way, just as Merrifield had done with amino acids. Ellman subsequently showed that he could 鈥渟plit-mix鈥� the reaction to generate large libraries of different benzodiazepinones. This work sparked enormous growth in the development of solid-phase chemistry, providing chemists with a large repertoire of reactions beyond peptide synthesis.

Solid-phase synthesis has moved on in other ways too. Polymer beads are not the only support that can be used; there are now a number of different techniques using pins, crowns, plugs or sheets. In most of these cases the purpose is to make the solid support bigger or increase the density of reagents to produce more of the final library compound. Some types of solid support have been designed simply to be compatible with laboratory equipment or automated processes. For example, pins are usually fixed in a standard array that allows them to be dipped simultaneously into different chemical solutions, whereas porous plugs a bit like cigarette filters can be inserted into tubes to allow reagent solutions to be pumped through them.

Despite the advances in solid-phase synthesis, it is not the most efficient way of making libraries of compounds. Obvious drawbacks are the additional synthetic steps required to attach the starting materials to the support and to remove the products at the end of the synthesis. What鈥檚 worse, despite considerable effort by chemists around the world, the repertoire of reliable solid-phase reactions is much smaller than that of traditional solution chemistry. For example, before they鈥檝e even started, chemists have to choose chemical reactions that will not interact with the solid support.

Even so, these solid-phase combinatorial methods alerted chemists to the possibility of vastly increasing their productivity. And that has led to a considerable amount of interest in developing solution-phase methods for combinatorial chemistry.

The main advantage of solution-phase synthesis is that it can draw on the rich heritage of organic chemistry to generate novel molecules in unprecedented numbers. However, with no polymer support to rely on, separating the products is a big challenge. This problem has largely been solved through the use of robotic methods that allow many individual chemical reactions to be run in parallel rather than in the same vessel.

A better solution

Chemists go back to the test tube

Most of the robotic systems use a standard 96-well laboratory plate鈥攁 plastic tray with 96 small, round indentations arranged in a 12 by 8 grid (see Figure). The robots accurately dispense chemical solutions into the individual wells, making it possible to run 96 reactions in parallel, with each well ultimately containing a different end product. For example, chemists can react one precursor molecule with 96 different reagents, or make every possible combination of one of 12 precursors with one of 8 others. This type of automated parallel synthesis is often referred to as high-speed chemistry.

Separate wells, however, don鈥檛 entirely solve the problem of retrieving the desired end product. Very few chemical reactions鈥攁nd even fewer reaction sequences鈥攇enerate their end products without leaving by-products and unreacted starting materials, so a purification step is needed. This has also been solved through automation. The purification of compounds using automated chromatographic separation is rapidly becoming the norm in solution-based combinatorial chemistry.

Ironically, high-speed techniques have largely removed the need to do the actual reactions. Instead, medicinal chemists now make virtual combinatorial libraries. These are computer simulations of all of the compounds that could be made using combinatorial techniques. When a new drug target is identified, chemists scan these virtual libraries and select groups of compounds that appear to have the most promising properties. This is based on calculated structural properties, a knowledge of what structures have proved successful in the past, plus sometimes a large dose of intuition. These compounds are then prepared individually. Because this selection process often identifies a miscellaneous selection of compounds, the most efficient way to make them is with high-speed chemistry. The aim of combinatorial chemistry has clearly changed: quality rather than quantity is now the goal.

But once a promising molecule is found, combinatorial chemistry comes to the fore once again. One of the technology鈥檚 great successes has been to develop many standard reaction pathways that allow the synthesis of large numbers of similar compounds. If a molecule shows promising activity in biological tests, chemists can be confident of making dozens of closely related structures. High-speed chemistry can then be employed to rapidly synthesise them all, and hopefully produce a new drug molecule.

In the past 20 years, combinatorial chemistry has blossomed from a laboratory curiosity into a technique that has allowed chemists to achieve previously unimagined levels of productivity. In the future, combinatorial chemistry will become a mainstay of drug discovery. The odds are still stacked against winning the pharmaceuticals lottery, but this technique has vastly improved the chances of hitting the jackpot.

The main use for combinatorial chemistry is in generating vast numbers of potential drug molecules for the pharmaceuticals industry to experiment with. But it鈥檚 just one small step in the long, complex and expensive process of drug discovery (鈥淗ow a drug is born鈥�, Inside Science No. 65).

The first step is to find out about the biological pathway that causes the disease you鈥檙e interested in. Then you look for a point in the pathway鈥攑erhaps a vital receptor or an enzyme鈥攚here it might be possible to intervene to block it. 鈥淚ntervention鈥� usually means finding a compound with low molecular weight鈥攂etween 250 and 500 atomic mass units鈥攁nd a strong and specific affinity for the receptor or enzyme of choice. In the case of infections, the target may be a protein that occurs in the virus or bacterium, or it could be a protein in the host cells that provides a toehold for infection.

The next step is to develop a biological test or assay that can identify potential drugs. This is the first point where combinatorial chemistry comes into its own, generating vast numbers of compounds to test. In fact, it has been so successful at producing new drug candidates that the industry initially had trouble trying them all out. In order to cope, researchers use high-throughput screening, in which thousands of compounds are rapidly tested against a biological target using automated methods.

If the assay identifies a compound with some affinity for the biological target, this becomes the lead compound. At this early stage, however, the lead compound typically has a number of flaws. It might have low affinity for the target, it might bind to two different targets, or it might be toxic. So it鈥檚 over to the medicinal chemists to design and synthesise structurally similar compounds with better combinations of properties. Combinatorial chemistry can help here too. Once a lead compound has been identified, high-speed chemistry can generate libraries of structurally related molecules to accelerate the process towards nominating a candidate compound for clinical trials.

One particularly innovative method in solid-phase combinatorial chemistry borrows a trick from the semiconductor industry to prepare peptide libraries on a glass slide. The technique is called photolithography and it involves using light to etch microscopic patterns on the slide.

In photolithographic combinatorial chemistry, standard reactions are used to coat a glass slide with amino acids. But there is one crucial difference: the chemical groups that cap the amino acids to stop them reacting prematurely are sensitive to ultraviolet light rather than chemicals. That means you can remove the caps by exposing them to UV, allowing you to couple a second amino acid to the first, but only on the areas that were exposed to the UV.

Using 鈥渕asks鈥� to selectively expose portions of the slide to UV, it is possible to uncap tiny patches with incredible precision. The masking/coupling procedure has been extended to 10 steps to generate a library of 1024 peptides in a 32 脳 32 array of squares, each just 400 micrometres on each side.

Photolithographic combinatorial chemistry was developed by scientists at biotechnology company Affymax of Palo Alto, California, in the early 1990s. Although it is not now widely used for synthesis, the basic process is essential for some techniques in DNA sequencing.

The advantage of photolithography is that you can screen the entire library in one go, and if one of the products shows interesting activity, you can identify it by its position on the slide.