Baubles, Bangles and Beads April 21, 2000

There are a lot of similarities between biotechnology and information technology (i.e. computers). Both are driven by new technology and new techniques. Both can make an investor lots of money, although the time frames are somewhat different for capital gains (years in the case of Immunex; days in the case of many dot coms). Capital losses, however, seem to take about the same amount of time.

And the power of these technologies is constantly increasing. Sixteen years ago, when I made oligonucleotides by hand, it took an entire day to make a 10-mer, assuming I made no mistakes during the extremely monotonous procedure. Automation now permits us to make one almost ten times as long today in that time. Billy Clevenger and Ginny Price performed the first PCR here at Immunex at the bench, entirely by hand. There was no thermostabile Taq polymerase at that time, so they had to add fresh enzyme at every cycle. Imagine, no machine to program. Nothing to control the temperature except a small water bath. The power of the procedure was demonstrated because it worked the first time. Molecular biologists love ANY procedure that works the first time you use it. However, the drudgery of the protocol made automation necessary for easy acceptance.

Everything is getting smaller and faster. Computers have completely altered the scientific landscape. When I started here we used a pretty difficult program called UWGCG running on a VAX. Now we use a pretty difficult program called GCG running under Unix. Okay, cheap shot. Today's software does much, much more, though , allowing us to examine sequences in ways that were unobtainable 10 years ago.

Now the best technology is liberating. It allows us to answer questions that were impenetrable before. Phosphoramidite chemistry allowed the rapid synthesis of oligonucleotides, that permitted PCR to flourish, that accelerated dideoxy sequencing, that will provide us with the sequence of the human genome. Now, these technologies also allow us to ask tougher questions but that is what science is all about.

Thanks to new technology, we can synthesize maybe 100 times more DNA than by hand, we can sequence over 100,000 more bases and, as I will discuss below, we can detect such small amounts of DNA that new units of measure must be applied.

We can make DNA to order. We can make more of it by PCR. We can sequence it. But, detecting the presence of specific sequences in a complex mixture still has its problems. DNA arrays are an extremely popular new technology. Put small amounts of defined sequences down on a solid support and use them to probe for complementary sequences. Or create an array of defined sequences by directly synthesizing the oligonucleotides on a solid support. Both of these approaches are being used to examine gene expression. Simply compare the mRNAs being expressed under different circumstances and you might get a hint as to which genes are important. These technologies are going to provide important answers but they do have problems, particularly with false positives and negatives. Reproducibility is another major problem. But I am sure ways around them will be discovered.

I want to talk a little bit about other adaptations of this approach. Instead of using a glass slide or piece of silicon to attach the DNA sequence to, these technologies use plastic microspheres as the solid support and fluorescence as the detection procedure. They have some exciting possibilities and are potentially very useful additions to our research repertoire.

These beads have several useful and unique properties. They are extremely small, 3-5 um in diameter. One milliliter of solution can contain 10¹⁰ beads. But they have a relatively large surface area, so that they can carry something like 10⁶ molecules attached to them.This creates a very high local concentration of oligonucleotides. If 1000 labeled target molecules hybridize to the probe molecules on a microsphere, concentrations of 1 uM will be seen. This is pretty easy to detect if fluorescent dyes are used. Image analysis algorithms and redundancy can be applied to increase signal-noise ratios, allowing femtomolar concentrations to be detected. The absolute detection limit approaches zeptomoles of DNA (who came up with this term??). This is 10^-21 moles of DNA. Not very much.

So, let's look at some approaches using this technology. I mentioned one of these a few weeks ago. It has been developed by a company called Lynx Technologies, the company Sydney Brenner is associated with. The short explanation of the protocol starts by performing oligonucleotide synthesis on the microbeads. The microbeads were split into 8 fractions and eight different four base 'words' were synthesized on each fraction. The beads were then recombined, split up into eight fractions and the synthesis was done again. This was done for 8 rounds, so that there are 8⁸, or 16,777,216, possible combinations of 'words' in the tag sequence, each 32 bases long. The composition of the words were chosen so that the A:T and G:C composition of each 32 base pair stretch was the same. Also, each tag sequence differs from its nearest neighbor by at least 1 word, or 4 bases, simplifying dissociation of mismatches. Control experiments indicated that over 85% of the beads has only a single tag sequence on them.

Using some nifty recombinant technology, they created a plasmid librry with the same set of tags in it. They carefully arranged restriction enzyme sites to allow a cDNA library to be constructed, placing a different sequence tag upstream from each cDNA. They could then use appropriate PCR primers (e.g. containing fluorescent tags) to amplify the library, only now the cDNAs carried the complementary sequence for the tags on the microbeads. Some further manipulations would then allow each cDNA carrying a tag to anneal to a bead carrying an anti-tag sequence. Ligation would then covalently attach each cDNA to a specific bead. And each bead would contain over 100,000 copies of a specific DNA sequence.

The ability to load billions of beads with specific DNA sequences means that they can be used to capture specific targets. In this paper, they used FACS analysis that could sort 20,000 beads per second. They did some nice control experiments to demonstrate that they could detect different levels of signal on the sorter. Finally, to prove the usefullness of this approach, they looked at induced and uninduced THP-1 cells. They first made a microbead library of cDNAs pooled from both uninduced and PMA-induced THP-1 cells. They then made two sets of probe libraries from this same pool, each set being labeled with one of two fluorophores. They took the 2 probe libraries, mixed them together and allowed them to hybridize to the microbead library. They then examined 100,000 beads with two-parameter fluorescence. Since both the target and the probe libraries were the same, the profiles fell along a line with slope of one.

They made a new library with just the cDNAs from uninduced cells and labeled it with one fluorphore. They also made a library using cDNAs from induced cells and labeled it with the other fluorophore. So, cDNAs that are overexpressed in induced cells will generate more signal along one axis, since there is now more of them to bind to on the beads. cDNAs that are underexpressed will generate less signal. By sorting for beads that fall off the slope determined by the control experiment, they could isolate cDNAs whose expression was altered following induction, WITHOUT doing any further cloning.

They sorted over 1.6 million beads. (What is that...something like 80 seconds!!). They isolated about 14,000 beads from the upregulated region and about 17,000 beads from the downregulated. They recovered the DNA by PCR and sequenced almost 1000 from each region. A large number of the sequences in each group were known genes, such as IL-8, TNFa, etc. But each group also had several novel sequences.

So, this approach looks like a nifty way to isolate novel genes from cells that have been differentially treated. You do not need to really know anything about the nucleotide sequences you are working with. This next approach does require some knowledge of the seqeunces, but with the coming flood of information, this may not be too difficult. This approach also uses microbeads that are loaded up with a target sequence. In this case, the target sequence contains a Molecular Beacon (MB). The sequence forms a hairpin, bringing together a flurophore and a quencher of fluorescence. Under these conditions, there is no signal. However, when a complementary sequencebinds to the target DNA, the hairpin is linearized, the quencher is moved away from the fluorophore and signal is then generated. So, you only get a fluorescent signal when a probe sequence hybridizes to the target sequence. No need to label the probe sequences at all. The signal comes entirely from the conformational change of the target.

This approach also uses another aspect of the high signal gebnerated by microbeads. A single microbead can be placed into a well etched in the end of a optic fiber. And the signal generated by DNA hybridizing to the targets on that one bead can be visualized. By bundling the fibers, each one containing a different bead, a large array can be generated, say 5000 to 50,000 individual fibers. This gives an array with a diameter less than 1 millimeter in diameter. Pretty small.

And it can be used in small volumes, say 10 microliters and concentrations to about 100 pM. Of course, which bead ends up in which fiber is random. So you have to use control sequences to address each fiber. But once you know which sequences light up each fiber, you can melt off the probes and have at it. The sensitivity of this approach is still not quite as good as other approaches but it can be used very rapidly to screen for common mutations, such as cystic fibrosis, or for bacterial DNA sequences. Wouldn't it be neat to stick a small tube in your mouth and find out if you had a strep infection? I think this approach holds a lot of promise in diagnostics.

As time has gone on, more and more toys have appeared for biologists to use. We can now sequence faster than before, identify novel sequences faster, create oligonucleotides faster. All using nice new machines. Technologies using microbeads hold exciting new potentials. And even more nice new machines. Biologists today have to like shiny new machines. It seems to be our destiny. Call it Kismet.