DNA Polymerase:Frankenstein's Monster Made Good
Of obvious importance to cells everywhere is the maintenance of the library of instructions for doing... everything(!).This library is made entirely of DNA. Indeed, the complexity of the human body and nervous system could NOT BE ACHIEVED if the error rate of DNA replication were even 10-fold higher. Amongst the key players in error prevention and correction are the machines that do the actual copying of DNA, the DNA Polymerases. Despite their critical roles, if we look carefully at these machines, we can see that they can be viewed as cobbled together pieces of other machines, and crippled derivatives of the RNA Polymerases from which they may have evolved.

The Big Job:Adding nucleotides
How does Polymerase know which nucleotide to add? The short answer is, it doesn't. There is no reason to believe that DNA Polymerase ever achieves direct knowledge of the nucleotide it is adding! Indeed, it has been elegantly demonstrated that it can add a slew of nucleotides outside of the canonical Big 4 it's used to. How, then, does it achieve "chemically impossible" low rates of error?

The first secret lies in the template--the existing strand of DNA being used to generate a complementary copy. By incorporating the template strand of DNA into its active site, DNA Polymerase uses the template nucleotide to assess the appropriateness of a given dNTP for addition.In short, DNA Polymerase never knows what it's adding, but it does determine whether its candidate 'fits' --as defined by the ability of the newcomer to form a specifically predicted structure with it. Recall that G:::C and A::T basepairs are almost superimposable, despite the different shapes and hydrogen bonding capabilities of their constituents. Thus DNA Polymerase may beignorant of the basepairing 'code' wherein G calls C, A calls T and vice versa, but itdetects whether 'whatever the template nucleotide is' matches 'whatever the incoming nucleotide is'. One direct prediction of this model is that DNA Polymerase can add ANY purine and any pyrimidine as long as the two form productive hydrogen bonds!This prediction is indeed met [Figure of synthetic ntes that Pol will add)...].Indeed, if Martian DNA were to be found, our DNA Polymerase should be able to replicate and sequence it whether or not it uses the same 4 bases that we do--as long as the backbone is deoxyribose and as long as nothing too bulky is hanging off the base rings.

Grabbing nucleotides: Polymerase is lazier than you think!
One common misconception is that DNA Polymerase 'chooses' or 'selects' the desired nucleotide from solution. Quite the contrary, ANYthing can wander into the active site of DNA Pol! To grab a specific nucleotide, DNA Pol would have to 1) form an opinion about what it wanted, and 2) have a machinery that lassoed nucleotides (this machinery would need to take instruction as to what DNA Pol wanted AND could see or smell nearby nucleotides!). To the contrary, the open binding site of DNA Polymerase is like an available parking spot--anything can drive in [DNA Pol with just a funnel as a binding site, with basepairing positions of template visible at bottom]. There will, however, be constraints on what nucleotides stay in the pocket. DNA polymerase must check to make sure it is not polymerizing ribonucleotides. How can it 'check'? By creating a binding pocket that conflicts structurally with the structure of a ribonucleotide. You can imagine designing the pocket in DNA Pol such that a deoxyribonucleotide fit snugly once basepaired, but that the extra hydroxyl at the 2' position of a ribonucleotide bumped into some part of the protein, preventing it from 'assuming the position.' Conversely, expected features of correct molecules could be 'rewarded' by designing the DNA Pol pocket such that favorable interactions occurred with, for example, the ribose, the triphosphate, the 3' OH--IF they had been oriented by favorable hydrogen bonding of the base on the incoming nucleotide to the template [again, having a nucleotide or couple shown in several orientations with the CORRECT one showing interactions of its phosphates, ribose, etc. to sites on Pol]. So all small molecules check in--but only basepairable deoxynucleotide triphosphates stay the night and are added to the growing chain.

Assessing the fit
How can DNA Pol come to know that a nucleotide is in the right position? Consider how an enzyme works. It acts by aligning the key players in whatever reaction it is catalyzing. By taking a somewhat 'hands off' approach, DNA Pol can make the interactions of the candidate nucleotide much more important. The newcomer becomes DEPENDENT upon its own interactions (H-bonding, stacking with other nucleotides), and if these interactions are less than optimal, the staying power of the guest nucleotide suffers. DNA Pol can be built such that its catalytic assistance only comes to bear if the nucleotide assists by presenting its triphosphates for optimal attack by the 3'OH on the primer strand.

Making mistakes
Since the materials it works with are constantly undergoing chemically driven changes that alter the hydrogen bonding surface they present (see section on tautomerization), DNA Pol's strategy of adding bases by fit and feel is doomed fail from time to time. Errors WILL occur. How is it we nonetheless exist with our 30,000-40,000 genes? The answer lies in the parts and properties of Polymerase itself.

In order to deal with the inevitable sidetrips of nucleotides to their tautomeric forms, two 'readings' of the basepairing ability of a nucleotide are made. First is the requirement for 'fit' before a candidate nucleotide is added as discussed above. However, since their is a possibility that either one is 'lying' about its identity--by adopting the uncommon tautomeric form--this step alone is insufficient to achieve high fidelity. By 'checking it twice', DNA Pol takes advantage of the fact that the tautomeric forms are energetically less favorable than their more common counterparts, and thus dependably short-lived. After making an addition to the chain, DNA Polymerase is now in a position to use the new base as the site of the subsequent addition. However, it only makes an addition if the previously added base is well-behaved, i.e. firmly basepaired.If the previous nucleotide is not basepaired, it will occupy a variety of positions, and only rarely simulate a basepaired nucleotide. This pickiness is the first step in error correction by DNA Polymerase.

Sounds good in words. But how to do this mechanically? By making DNA Pol somewhat 'helpless' in terms of positioning the primer nucleotide. Again, weakness is accuracy! If DNA Pol is not too 'pushy' in terms of forcing the primer nucleotide (most critically: the 3' OH of the ribose!), the nucleotide must hold itself in place. It's going to need all the help it can get--including the basepairing H-bonds! Failure to get everything lined up will mean DNA Pol cannot proceed--the key chemical groups just won't be in the right places for catalysis.

But what profit does a DNA Pol gain by 'stalling' after making an incorrect incorporation? If it were working on its own, none--eventually the machine would either fall off the DNA, ending the cell's hopes of replicating its genome, or the incorrect addition might through chance 'sit still' in the appropriate position long enough to allow subsequent additions. But somewhere in the course of evolutionary history, a happy shotgun marriage was performed: the joining of a DNA Polymerase with a non-specific exonuclease (a machine that lops off nucleotides from the end of a chain). [attaching Pac-Man nuclease by a stiff spring to nuclease could capture the lack of specificity required here]. While this seems a silly idea for an addition to a DNA Polymerase molecule, in the context of proofreading it makes brilliant sense. By guaranteeing the presence of nucleotide removal machinery in the vicinity of DNA polymerase, the cell gains the ability to resolve the dilemma presented above. After stalling of polymerase due to an incorrect addition, opportunities arise for the exonuclease to 'get its hands on' the aberrant nucleotide. The nuclease can be thought of as a ravening beast--it eats what it finds--but due to the propensity of DNA Pol to stall after an erroneous addition, a reasonably slick scenario for error correction emerges:
1)DNA Polymerase is fooled by a tautomeric form or by chance apparent 'fit' of an incorrect nte. in the correct position.
2)Upon proceeding to make the next addition, DNA Pol is unable to 'find' a 3' OH in the correct position for further polymerization. Further additions are thus blocked
3)DNA Pol slides back and forth on the 'rails' of DNA aimlessly
4)Periodically, the DNA Pol will slide 'backwards', exposing the recently added nucleotides to solution
5)The DNA 3' end will intermittently 'fray' (separate from the template strand), and the freed strand of DNA (capped by the incorrect nucleotide!) will wander
6)The wandering strand may wander into the maw of the nuclease, which snips off as many ntes. as it has the opportunity to
7)now lacking the offending mis-addition, there will come a time when the DNA strands 're-zip' (again, driven only by chance!) and DNA Pol wanders into position and is happily confronted with a correct pairing, to which it merrily adds and the polymerization process continues.
[A cartoon of the above would be astoundingly excellent; however it'd be pretty complex. Perhaps mini-cartoons of each step? Can have the ole 'thought bubble' coming out of DNA Pol containing a question mark when it can't proceed, an exclamation point in 7. A ticking watch somewhere for pauses could illustrate the VERY KEY point that passage of time drives this]

Evolutionary considerations
The 'discovery' of proofreading by DNA polymerases illuminates several important principles and practices of evolution of protein machines. First, note that the story as told contains an inherent paradox--if DNA stalling came first, then there would have been no DNA synthesis until accidental addition of the exonuclease to the complex--doubtless the organism would be extinct long before the opportunity arose!
A more likely scenario is that original DNA Polymerases were not so picky about placement of the 3'OH. How could this be? Firstly, note that RNA Polymerases are not so picky about template-directed positioning of the 3'OH. Doubtless they contain structures that draw or hold the 3' nucleotide in the correct position for addition and thus do not rely on basepairing to perform this task. Ancient DNA Pols evolved from RNA Pols would be expected to share this property. Accidental addition of a nuclease to the complex at this point would still have benefited DNA Pol. Though stalling may not have been absolute, an incorrect addition will inevitably be 'harder' to position than a correct one. So some stalling is still expected. By creating the capability to resolve the situation as described above, a more accurate and efficient DNA Pol would be created.

Now comes the cool part. To make DNA Pol more accurate once the nuclease has been 'glued on', we actually need to cripple the DNA Polymerase! By decayingits ability to 'grasp' the last-added nte., we render the stalling at mis- additions ever more lengthy, and the events leading up to nuclease action more inevitable. Eventually we would achieve the current day situation in which many DNA Polymerases are extremely picky about behavior of last-added nucleotides and thus extremely accuratesince the inevitable outcome is removal of the offending nucleotide.

This model also explains another property of DNA Polymerases--their inability to initiate nucleotide chains using template alone. Is this pickiness inevitable in a nucleotide polymerase? Clearly not--RNA Polymerases are quite capable of doing the "1+1" reaction, adding the first two nucleotides together to initiate a new nucleotide strand. Why then would DNA Pols, which presumably evolved FROM RNA Pols, lack this capability? First, it may be an inevitable outcome of primer dependance--by loosening DNA Pol's 'grip' on the template nucleotide, you would expect to lose the ability to hold a single nucleotide in the same position (such as would be required to perform the "1+1" initial addition to start the chain).

A second argument for why DNA Pol would lose the ability to initiate synthesis without a primer is this:such an ability would defeat proofreading! Note how the sequence of events in proofreading requires that DNA Pol stall after a mis-addition. This represents a window of opportunity for the nuclease to get involved. If, on the other hand, DNA Pol had the capacity to initiate a new chain, it would simply leave its mistake behind and move on--doing a nice job of moving quickly, and a very poor one of moving accurately.