compounds (12, 19, 20) > dianionic (21) > monoanionic (22, 23,
4). The same trend was seen when these compounds were
tested against TPP-dependent enzymes. For the trianionic com-
pounds, replacing the bridging oxygen of the pyrophosphate
with a methylene group significantly improves the binding (40-
required for the stabilisation of the OFF conformation of the
riboswitch and therefore gene regulation.
Several factors must be borne in mind when comparing the
2
9
thermodynamic K value for a compound binding to the aptamer
d
with its effect on expression of a reporter enzyme: there could be
significant differences in affinity between full length mRNA and
the aptamer alone; switching could be kinetically rather than
fold lower K ), whereas replacing it with a difluoromethylene
d
group has a much smaller effect (2-fold lower K ). The oxygen
d
16
2+
is not directly involved in forming any interactions to the RNA
and the effect of the bridging group on the electronics of the
other phosphate oxygens seems more important. There is a clear
thermodynamically driven; the Mg ion concentrations in the
systems are likely to be different, etc. Nevertheless, our results
suggest that there is a trend between the ligand binding
affinity of the TPP analogues and the degree of gene repression.
The most potent analogue, 20, binds to the thiM aptamer with a
very similar affinity to that of TPP and represses the expression
of the Renilla luciferase to the same extent (approximately 60%
reduction). Whereas, compounds 12–14, 19 and 21, which have
correlation between the pK of the phosphates/phosphonates and
a
the binding affinity, with the best ligand being the methylene
phosphonate (K = 9 nM), which has the least acidic protons
d
11
(
approximate pKa3 value for the parent acid is 7.45 ), followed
by the difluoromethylene phosphonate (K = 170 nM, pK
≈
d
a3
5
.80) and then the pyrophosphate (K = 370 nM, pK ≈ 5.77).
K values approximately 20–250-fold higher than TPP, exert a
d
d
a3
The ability of the phosphate/phosphonate to stabilise electronic
charge will obviously have an effect on its interaction with the
magnesium ions and the RNA. Due to the presence of the two
divalent cations, it is likely that all three of these compounds
would be bound in the trianionic form. In this situation, the least
acidic phosphonate has the least stabilised negative charge and
would presumably form the strongest interaction to the oppo-
sitely charged magnesium ions.
lesser effect on gene expression, only reducing it by 21–38%.
Interestingly, TMP shows the same behaviour as these analogues.
The concentrations of all these compounds in the IVTT mixture
is much higher than their measured K values, so at equilibrium
d
the riboswitch should be almost fully bound to the compound.
The fact that translation is not turned off as fully with these com-
pounds as with TPP, therefore, suggests either that the system
does not have time to reach equilibrium (as already suggested for
TPP) or that the pyrophosphate-sensing helix is not so tightly
held in place in the bound state and a small proportion of it
exists in other conformations that allow binding of a ribosome
(or a combination of these two effects). Further experiments to
follow the conformational changes in the pyrophosphate-sensing
helix would be needed to distinguish between these two
possibilities.
The results obtained with the alcohol analogues (thiamine,
pyrithiamine and 8T) show that some regulation of expression
does still occur, even in the complete absence of a pyrophos-
phate group. This is not what would be predicted from the model
described above for the switching mechanism. This may hint at a
more subtle model for switching whereby the pyrimidine-
sensing helix, while preformed, is still flexible and docking of a
ligand helps to rigidify it and thereby enhances the formation of
the L5–P3 tertiary contact. This in turn shifts the equilibrium
towards the OFF conformation. Although further studies will be
necessary to clarify this point, our in vitro transcription–trans-
lation assay allowed for investigation of TPP analogues that
would have not been possible with an in vivo system. Therefore,
it should prove a valuable tool to complement crystallographic
and in vivo expression techniques.
The analogues that showed binding to the riboswitch aptamer
were also tested for their ability to reduce gene expression, using
an IVTT reporter gene assay. The assay monitored the riboswitch
activity indirectly by measuring the luminescence produced by a
catalytic reporter protein (Renilla luciferase) produced by IVTT.
This system differs from previously published thiM reporter gene
10
assays because it is performed in vitro using cell-free E. coli
extracts, thus avoiding any doubts about cell-permeability, and it
employs Renilla luciferase as the reporter gene, which gives high
sensitivity. It is notable that with this system even TPP at
1
6
00 μM only decreases expression of the reporter gene by ca.
0%. This is possibly because the concentrations of the com-
ponents in the IVTT mixture mean that the TPP only has a
limited amount of time to bind to the riboswitch before a ribo-
some binds to the ribosome binding site (RBS) and it is too late
to prevent translation from starting.
The conformational changes that occur during switching have
1
2–15
been studied in detail.
The present model proposes that, in
the absence of TPP, the pyrimidine-sensing helix (P2 and P3) is
mostly preformed while the pyrophosphate-sensing helix (P4
and P5) is thought to have more conformational freedom allow-
ing the riboswitch to exist in a fine equilibrium between two
alternately folded states. In one, the two strands that make up the
P4–P5 helix base-pair, forming the pyrophosphate binding
pocket, leaving the RBS and start codon to be sequestered in
another stem–loop within the expression platform. In the other, a
strand of P4–P5 helix base-pairs with a portion of the expression
platform, leaving the RBS and start codon single-stranded,
thereby allowing initiation of translation (see Fig. S1‡). The
binding of TPP stabilises the pyrophosphate sensing helix and
other tertiary interactions within the aptamer (especially a distal
contact between P3 and loop L5 at the end of the P4–P5 arm),
thereby favouring the former of the two states. This model
suggests that, while the pyrimidine ring of TPP is important for
molecular recognition, the pyrophosphate moiety of TPP is
Conclusions
In conclusion, the studies reported here show that the thiM ribo-
switch of E. coli is highly selective for the 4-aminopyrimidine
ring (A ring) of TPP, although other heterocycles can bind if
they can offer the same pattern of hydrogen bond donors and
acceptors. The riboswitch is not very selective for the thiazolium
ring (B ring) of TPP, but positively charged rings bind much
better than neutral ones and open-chain compounds are poor
ligands. The pyrophosphate group of TPP is essential for tight
binding but can be replaced by other trianionic groups of similar
This journal is © The Royal Society of Chemistry 2012
Org. Biomol. Chem., 2012, 10, 5924–5931 | 5929