Problem outline
Marc Juul
problems that may or may not need solving independently:
1. Exactly controlling when the polymerase adds precisely one
nucleotide.
2. Controlling which nucleotide is added at each step.
If we can solve problem one, we can probably anchor the polymerases
and simply wash in the next nucleotide before allowing the polymerase
to continue. This might be a slow process, and a crude solution to
problem two, but it should work.
It doesn't seem entirely unreasonable that we should be able to create
a polymerase that can be activated and deactivated by external
factors, yet even if we can accomplish that, it will probably not be
possible to reliably active the polymerase for exactly long enough to
add precisely one nucleotide, given the stochasticity of polymerase
activity. We need a polymerase that gets stuck in a wait-state after
adding a single nucleotide, where it can only proceed when given a
signal. Even then, we may get into a situation where the minimum
reliable time for the activation signal to be 100% effective will
sometimes allow a polymerase to activate twice and add two
nucleotides. We may need two different signals to control the
polymerase, or figure out some other clever way of avoiding this
scenario. If we're lucky maybe it won't be a problem at all, but we
should do some research on the variability of time taken for photo-
isomerization of enzyme to predict is this will be an issue we need to
plan for.
The stochasticity of polymerases has another nasty side-effect: We
will have to wait long enough between adding nucleotides that the
slowest polymerase has finished adding the current nucleotide before
proceeding to the next step. This could mean that controlled enzymatic
synthesis will always be significantly slower than synthesis by
natural polymerases.
In order to get something like fast whole-genome synthesis, we will
also need to solve problem two in some way that is fast and reliable.
Modifying the nucleotides themselves to be deactivated or activated by
pulses of light would be an interesting way of accomplishing this, if
such a thing is possible in a reliable way.
In single-molecule real-time sequencing, Pacific Biosciences uses
nucleotides with a fluorescent molecules attached to the phosphate
chain:
https://en.wikipedia.org/wiki/Single_molecule_real_time_sequencing
Their technique is sequencing by synthesis, and the attached
fluorescent molecules are automatically decoupled from the nucleotide
as the polymerase attaches it to the DNA, apparently without affecting
synthesis.
So we know that it's possible to have something linked to the
nucleotide, and the question becomes if it's possible to attach a
photo-isomerizable protein that will block the nucleotide from binding
to the polymerase in one isomer and allow it in another.
--
Marc Juul
Bryan Bishop
> I think we can divide the problems that need solving into two main
> problems that may or may not need solving independently:
>
> 1. Exactly controlling when the polymerase adds precisely one
> nucleotide.
>
> 2. Controlling which nucleotide is added at each step.
I think if you have #2, you don't necessarily need #1. As long as you
can switch the polymerase into another state before it adds/matches
the next nucleotide.
> If we can solve problem one, we can probably anchor the polymerases
> and simply wash in the next nucleotide before allowing the polymerase
Eeek! I hope we can avoid wash steps. Those take time. :-/
> to continue. This might be a slow process, and a crude solution to
> problem two, but it should work.
>
> It doesn't seem entirely unreasonable that we should be able to create
> a polymerase that can be activated and deactivated by external
> factors, yet even if we can accomplish that, it will probably not be
> possible to reliably active the polymerase for exactly long enough to
> add precisely one nucleotide, given the stochasticity of polymerase
> activity. We need a polymerase that gets stuck in a wait-state after
> adding a single nucleotide, where it can only proceed when given a
Well, a wait-state would definitely be nice. But is it necessary?
> The stochasticity of polymerases has another nasty side-effect: We
> will have to wait long enough between adding nucleotides that the
> slowest polymerase has finished adding the current nucleotide before
> proceeding to the next step. This could mean that controlled enzymatic
> synthesis will always be significantly slower than synthesis by
> natural polymerases.
Maybe. Wouldn't you need only one of these polymerases to get an
essentially-guaranteed sequence?
> Modifying the nucleotides themselves to be deactivated or activated by
> pulses of light would be an interesting way of accomplishing this, if
> such a thing is possible in a reliable way.
You mentioned an enzymatic method for capping/attaching nucleotides,
but there's also the photocleavable chemically-capped nucleotides that
could work there. (But, wash step.)
> Their technique is sequencing by synthesis, and the attached
> fluorescent molecules are automatically decoupled from the nucleotide
> as the polymerase attaches it to the DNA, apparently without affecting
> synthesis.
>
> So we know that it's possible to have something linked to the
> nucleotide, and the question becomes if it's possible to attach a
Just a caveat- not a very important one- but PacBio did some directed
evolution work on that enzyme, according to their 2008 paper. Nothing
too difficult, but the work would have to be repeated if we were to
get an enzyme that handles the chemically modified nucleotides.
> photo-isomerizable protein that will block the nucleotide from binding
> to the polymerase in one isomer and allow it in another.
Maybe your protein+nucleotide complex could be reversed. Instead of
using light to dissociate the complex, you could have light cause a
change in the enzyme that allows it to hook up to a hacked polymerase,
then passing on the nucleotide cargo to the polymerase, then
dissociating. I can't think of an enzyme that does this sort of
physical handoff of cargo, but it's probably no more crazy than all
this other crap we're smoking.
Bryan Bishop
>> photo-isomerizable protein that will block the nucleotide from binding
>> to the polymerase in one isomer and allow it in another.
>
> Maybe your protein+nucleotide complex could be reversed. Instead of
> using light to dissociate the complex, you could have light cause a
> change in the enzyme that allows it to hook up to a hacked polymerase,
> then passing on the nucleotide cargo to the polymerase, then
> dissociating. I can't think of an enzyme that does this sort of
> physical handoff of cargo, but it's probably no more crazy than all
> this other crap we're smoking.
Now that I think about it, this is probably exactly what you were talking about.
Marc Juul
> On Fri, Feb 24, 2012 at 7:08 PM, Marc Juul wrote:
>> I think we can divide the problems that need solving into two main
>> problems that may or may not need solving independently:
>>
>> 1. Exactly controlling when the polymerase adds precisely one
>> nucleotide.
>>
>> 2. Controlling which nucleotide is added at each step.
>
> I think if you have #2, you don't necessarily need #1. As long as you
> can switch the polymerase into another state before it adds/matches
> the next nucleotide.
The problem is, with a stochastic process, will there be an amount of
time in each state that is sufficient for most or all of the
polymerases to attach one nucleotide, but never two?
>> If we can solve problem one, we can probably anchor the polymerases
>> and simply wash in the next nucleotide before allowing the polymerase
>
> Eeek! I hope we can avoid wash steps. Those take time. :-/
>
>> to continue. This might be a slow process, and a crude solution to
>> problem two, but it should work.
>>
>> It doesn't seem entirely unreasonable that we should be able to create
>> a polymerase that can be activated and deactivated by external
>> factors, yet even if we can accomplish that, it will probably not be
>> possible to reliably active the polymerase for exactly long enough to
>> add precisely one nucleotide, given the stochasticity of polymerase
>> activity. We need a polymerase that gets stuck in a wait-state after
>> adding a single nucleotide, where it can only proceed when given a
>
> Well, a wait-state would definitely be nice. But is it necessary?
We should look into how much the time to add a nucleotide varies for
the same polymerase between different nucleotide addition steps.
>> The stochasticity of polymerases has another nasty side-effect: We
>> will have to wait long enough between adding nucleotides that the
>> slowest polymerase has finished adding the current nucleotide before
>> proceeding to the next step. This could mean that controlled enzymatic
>> synthesis will always be significantly slower than synthesis by
>> natural polymerases.
>
> Maybe. Wouldn't you need only one of these polymerases to get an
> essentially-guaranteed sequence?
Well yes, but the stochasticity isn't between different polymerases
(they should all be the same) but for the same polymerase. You never
know how long it's going to take to add one nucleotide. So as you make
longer sequences there is an increasing risk that none of the
sequences will be correct if you ignore this.
>> Modifying the nucleotides themselves to be deactivated or activated by
>> pulses of light would be an interesting way of accomplishing this, if
>> such a thing is possible in a reliable way.
>
> You mentioned an enzymatic method for capping/attaching nucleotides,
> but there's also the photocleavable chemically-capped nucleotides that
> could work there. (But, wash step.)
Hmm yeah. Wash step :-/
Dan Bolser
> On Fri, Feb 24, 2012 at 7:08 PM, Marc Juul wrote:
>> I think we can divide the problems that need solving into two main
>> problems that may or may not need solving independently:
>>
>> 1. Exactly controlling when the polymerase adds precisely one
>> nucleotide.
>>
>> 2. Controlling which nucleotide is added at each step.
>
> I think if you have #2, you don't necessarily need #1. As long as you
> can switch the polymerase into another state before it adds/matches
> the next nucleotide.
>
>> If we can solve problem one, we can probably anchor the polymerases
>> and simply wash in the next nucleotide before allowing the polymerase
>
> Eeek! I hope we can avoid wash steps. Those take time. :-/
>
>> to continue. This might be a slow process, and a crude solution to
>> problem two, but it should work.
>>
>> It doesn't seem entirely unreasonable that we should be able to create
>> a polymerase that can be activated and deactivated by external
>> factors, yet even if we can accomplish that, it will probably not be
>> possible to reliably active the polymerase for exactly long enough to
>> add precisely one nucleotide, given the stochasticity of polymerase
>> activity. We need a polymerase that gets stuck in a wait-state after
>> adding a single nucleotide, where it can only proceed when given a
>
> Well, a wait-state would definitely be nice. But is it necessary?
I think so, even a very low error rate, like 0.001% is disastrous when
you want to synthesize more than a few MB. You either need one dNTP at
a time in solution (very hard) or an 'add one and wait' strategy.
Bryan Bishop
> change in the enzyme that allows it to hook up to a hacked polymerase,
> then passing on the nucleotide cargo to the polymerase, then
> dissociating. I can't think of an enzyme that does this sort of
> physical handoff of cargo, but it's probably no more crazy than all
> this other crap we're smoking.
One class of enzymes that does this hand-off operation is the
elongation factor class.
"During protein synthesis, tRNAs are delivered to the ribosome by
proteins called elongation factors (EF-Tu in bacteria, eEF-1 in
eukaryotes), which aid in decoding the mRNA codon sequence. Once
delivered, a tRNA already bound to the ribosome transfers the growing
polypeptide chain from its 3’ end to ...."
"This complex transiently enters the ribosome, with the tRNA anticodon
domain associating with the mRNA codon in the ribosomal A site. If the
codon-anticodon pairing is correct, EF-Tu hydrolyzes guanosine
triphosphate (GTP) into guanosine diphosphate (GDP) and inorganic
phosphate, and changes in conformation to dissociate from the tRNA
molecule."
http://en.wikipedia.org/wiki/EF-Tu
http://en.wikipedia.org/wiki/EEF-1
Animation:
http://en.wikipedia.org/wiki/File:Translation.gif
So the requirements:
(1) four engineered enzymes that bind to a specific NTP whenever they
don't have cargo
(2) in their default state, these carrier enzymes do not bind to polymerase
(3) when stimulated with light, a specific carrier enzyme (w/ cargo)
changes shape to bind to polymerase. Non-cargo carrying enzymes need
to be excluded with some additional conformational change? i.e. they
should only bind to polymerase when they have an NTP and a signal has
been given to change it into the "bind to polymerase" state.
(4) polymerase does not bind to free NTPs (no "nucleotide pocket" at
least as they currently exist), but rather only binds to this carrier
enzyme
(5) the carrier enzyme binds to and plugs polymerase, dissociates from
the NTP releasing the NTP inside a pocket on polymerase
(6) polymerase incorporates the NTP, waits for a step signal; if we do
some fluorophore stuff we could theoretically detect incorporation
(7) at this point, we switch all the carrier enzymes to "do not bind
to polymerase". Maybe this step dissociates the carrier enzyme from
the polymerase, or maybe it happens in the next step.
(8) when there's a step signal, polymerase moves (and maybe
dissociates the carrier enzyme during this step?)
(9) switch the next carrier enzyme into the "bind to polymerase" state.
(10) our previously dissociated carrier enzymes go and pick up some
free nucleotides from solution somewhere
Conformational states of this carrier enzyme:
(1) default - unable to bind to polymerase, can bind to one type of
nucleotide (may or may not have nucleotide)
(2) bindable-inactive - unable to bind to polymerase, has nucleotide
Note: "change to a state where you can bind to polymerase" signal + no
nucleotide should /not/ allow it to bind to polymerase.
(3) bindable-active - able to bind to polymerase, has nucleotide
("able to bind to polymerase" caused by light/laser)
(4) polymerase-bound -- bound to polymerase, released/releasing its nucleotide
After #4, it should stay bound to polymerase until a signal switches
it back to the "default" state. So there needs to be a way to switch
it "on" (per type) and a way to switch all "off" regardless of whether
or not the enzyme is already "off" (unless we want 8 wavelengths to
deal with?).
So the "bind to polymerase" state should only occur when (A) it
already has bound to a nucleotide and (B) the bind-to-polymerase
signal has been given. When the "bind to polymerase" signal is given
and it has not found a free nucleotide, the carrier enzyme should be
unable to bind to polymerase. If it's easy to do, then in this
scenario (bind signal + no nucleotide) making it unable to bind to
polymerase *until* it finds itself a nucleotide is okay. I don't see
any reason to force the sequence of steps to be locked in order. But
the real determinant of this is which way the protein design is
easiest, of course.
It is not immediately obvious to me how EF-Tu avoids binding to a
ribosome when it has no tRNA. But there seems to be lots of good
studies on this elongation factor :-).
* Dynamics of Recognition between tRNA and Elongation Factor Tu
http://www.sciencedirect.com/science/article/pii/S0022283608001290
"""
Elongation factor Tu (EF-Tu) binds to all standard aminoacyl transfer
RNAs (aa-tRNAs) and transports them to the ribosome while protecting
the ester linkage between the tRNA and its cognate amino acid. We use
molecular dynamics simulations to investigate the dynamics of the
EF-Tu·guanosine 5′-triphosphate·aa-tRNACys complex and the roles
played by Mg2+ ions and modified nucleosides on the free energy of
protein·RNA binding. Individual modified nucleosides have pronounced
effects on the structural dynamics of tRNA and the EF-Tu·Cys-tRNACys
interface. Combined energetic and evolutionary analyses identify the
coevolution of residues in EF-Tu and aa-tRNAs at the binding
interface. Highly conserved EF-Tu residues are responsible for both
attracting aa-tRNAs as well as providing nearby nonbonded repulsive
energies that help fine-tune molecular attraction at the binding
interface. In addition to the 3′ CCA end, highly conserved tRNA
nucleotides G1, G52, G53, and U54 contribute significantly to EF-Tu
binding energies. Modification of U54 to thymine affects the structure
of the tRNA common loop resulting in a change in binding interface
contacts. In addition, other nucleotides, conserved within certain
tRNA specificities, may be responsible for tuning aa-tRNA binding to
EF-Tu. The trend in EF-Tu·Cys-tRNACys binding energies observed as the
result of mutating the tRNA agrees with experimental observation. We
also predict variations in binding free energies upon misacylation of
tRNACys with d-cysteine or O-phosphoserine and upon changing the
protonation state of l-cysteine. Principal components analysis in each
case reveals changes in the communication network across the
protein·tRNA interface and is the basis for the entropy calculations.
"""
* The specific interaction between aminoacyl-tRNAs and elongation
factor Tu (2011)
http://www.springerlink.com/content/v0k77625342u0262/
"""
EF-Tu couples the hydrolysis of GTP with the accurate delivery of
aminoacyl-tRNAs (aa-tRNAs) into the encoded ribosomal A site. The
well-studied catalytic cycle of EF-Tu (Figure 1) can be subdivided
into five phases: (1) the binding of EF-Tu•GTP to elongator aa-tRNAs;
(2) the binding of the resulting ternary complex to the ribosomal A/T
site where codon sampling occurs; (3) a conformational change of both
the ribosome and the ternary complex with subsequent hydrolysis of
GTP; (4) disruption of the ternary complex with release of phosphate,
accommodation of the aa-tRNA into the A site, and release of EF-Tu•GDP
from the ribosome; and (5) GDP-GTP exchange catalyzed by EF-Ts. Many
of these phases have been dissected into several discrete steps using
a variety of biochemical and biophysical methods (Pape et al., 1998;
Gromadski et al., 2002; Blanchard et al., 2004). The mechanistic
details of this EF-Tu-dependent decoding pathway are the focus of
several articles in this volume. Here we discuss how the thermodynamic
details of the interaction between EF-Tu and aa-tRNA differ for each
tRNA species. We will summarize data showing that the sequence of
three base pairs in the T stem of tRNA “tunes” its affinity for EF-Tu
in a way that compensates for the variable contribution of the
esterified amino acid to the overall binding affinity. This ensures
that any correctly aminoacylated tRNA can initially bind to EF-Tu•GTP
tightly enough for delivery to the ribosome but weakly enough that it
can be released from EF-Tu•GDP during decoding. It appears that this
sequence-specific tuning is highly conserved in bacteria and can
largely explain the complex pattern of sequence conservation in the T
stems of all bacterial tRNAs.
"""
diagram: http://diyhpl.us/~bryan/papers2/bio/elongation-factor-tu-catalytic-cycle.jpg
* Tuning the affinity of aminoacyl-tRNA to elongation factor Tu for
optimal decoding
http://www.pnas.org/content/108/13/5215.short
"""
To better understand why aminoacyl-tRNAs (aa-tRNAs) have evolved to
bind bacterial elongation factor Tu (EF-Tu) with uniform affinities,
mutant tRNAs with differing affinities for EF-Tu were assayed for
decoding on Escherichia coli ribosomes. At saturating EF-Tu
concentrations, weaker-binding aa-tRNAs decode their cognate codons
similarly to wild-type tRNAs. However, tighter-binding aa-tRNAs show
reduced rates of peptide bond formation due to slow release from
EF-Tu•GDP. Thus, the affinities of aa-tRNAs for EF-Tu are constrained
to be uniform by their need to bind tightly enough to form the ternary
complex but weakly enough to release from EF-Tu during decoding.
Consistent with available crystal structures, the identity of the
esterified amino acid and three base pairs in the T stem of tRNA
combine to define the affinity of each aa-tRNA for EF-Tu, both off and
on the ribosome.
"""
* Construction of a fully active Cys-less elongation factor Tu:
Functional role of conserved cysteine 81
http://www.sciencedirect.com/science/article/pii/S157096391100032X
"""
In order to study the structural and functional requirements of the
essential translational GTPase elongation factor (EF) Tu for efficient
and accurate ribosome-dependent protein synthesis, construction of a
cysteine-free (Cys-less) mutant variant allowing for the site-directed
introduction of fluorescent and non-fluorescent labels is of great
importance. However, previous reports suggest that a cysteine residue
in position 81 of EF-Tu from Escherichia coli is essential for its
function. To study the functional role of cysteine 81 and to construct
a fully active Cys-less EF-Tu, we have analyzed 125 bacterial
sequences with respect to sequence variations in this position
revealing that in a small number of sequences alanine and methionine
can be found. Here we report the detailed comparative biochemical
analysis of three Cys-less variants of EF-Tu containing these
substitutions as well as the isosteric amino acid serine. By
characterizing nucleotide binding, EF-Ts interaction, aminoacyl-tRNA
binding, and delivery to the ribosome, we demonstrate that only
alanine (or cysteine) can be tolerated in this position and that the
serine and methionine substitutions significantly impair
aminoacyl-tRNA, but not nucleotide binding. Our findings suggest a
critical functional role of the amino acid residue in position 81 of
EF-Tu with respect to aminoacyl-tRNA binding. Based on structural
considerations, we suggest that position 81 indirectly contributes to
aminoacyl-tRNA binding through the accurate positioning of helix B.