6098
Y. Jin et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6094–6099
Table 3
The IC50 values have also been previously determined for the
Escherichia coli RNAP.21 It was observed that both MTB and E. coli
enzymes exhibited similar trends, with IC50 values in the 10À9
PXR activity of selected rifamycin analogues
b
M)a
EMAX (fold increase)
M
No.
EC50
(l
(nM) range for commercially available rifamycins tested. This
observation is consistent with previous postulates that the lower
sensitivity of Gram-negative bacteria to rifamycins is due to the
removal of rifamycins from the interior of the cell via TolC-depen-
dent efflux pumps. To probe that same efflux activity with our
novel analogues, we have determined the MIC90 values of a subset
1
5.6
9.5
6.9
nda
8.9
8a
9a
10a
2.1
ndac
4.1
a
EC50 is defined as the half maximal effective concentration.
ECMAX is the maximal effective concentration of the compound. (The experi-
mental conditions are reported in the Supplementary data).
b
of our compounds using WT E. coli strains (TG2 and DH5a) and a
c
nda: no detectable activation of hPXR at concentrations up to 100 lM.
mutant E. coli strain (EC2880-‘permeable’ strain, tolCÀ and impÀ).
Generally, the mutant E. coli strain is more sensitive than the WT
strains to tested rifamycins. Interestingly, the MIC90 value for com-
pound 3 is only four-fold greater than that observed for Rif S. This
parallels compound 3’s IC50 value, which is 3–4 higher than that for
Rif S. The results in Table 2 indicate that the rifamycins tested are
subject to Gram-negative efflux pump activity.
To gain better understanding of the SAR, we carried out struc-
ture-based modeling studies. Modeling was based on the 2.5 Å res-
olution structure of rifabutin complexed with the Thermus
thermophilus RNAP holoenzyme (PDB ID: 2a68).22 Since the rifamy-
cin binding site is highly conserved among bacteria, this structure
provides a good foundation for understanding how proposed rifa-
mycin analogues may interact with the MTB RNA polymerase.
Preparation of the structure before modeling is described in the
Supplementary data and was conducted using MOE.23
Given the size of the rifamycins, size of the binding site, and the
flexibility of the ansa ring, accurate docking of these inhibitors to
the RNAP complex would be challenging using standard docking
approaches. Since most of the rifamycin structure remains
unchanged, modeled poses of the analogues were generated by
mutating rifamycin to the analogue and then relaxing the complex
through a series of energy minimizations as described in the
Supplementary data.
Modeled poses were generated for each of the structures listed
in Table 2. A QSAR model (not shown) using interaction energies
between the inhibitor and RNAP complex and two other ligand
descriptors produced a very good R2 of 0.90 and a cross validation
R2 of 0.79. The good fit of the QSAR model based on energies from
the structure-based modeling supports the accuracy of the latter.
The robustness of our structure-based modeling has also been sup-
ported by experimental studies where sensitivity to rifamycins
against binding site mutants qualitatively correlates with our mod-
els.21 The QSAR model will be presented elsewhere.
Figure 2. Selected rifamycin analogues versus PXR activation.
aerobic and anaerobic conditions show that our modification of the
8-OH position of the parent scaffolds results in diminished activity.
The enzymatic and microbiological data are consistent with mod-
eling and computational studies which support the C-8 hydroxyl
acting as a hydrogen bond acceptor with S450 and that Rif resis-
tance in the S450L mutant is due to loss of this hydrogen bond.
Data on representative analogues of this study suggest that only
the pyrazole modification to the rifamycin parent scaffold mini-
mizes hPXR activation.
Acknowledgments
Modeled poses for Rif S (2) and amino analogue (4i) are shown
in Figure 1. RNAP is shown as a molecular surface within 5 Å of the
inhibitors and shaded to indicate areas of lipophilicity (green),
hydrogen bonding (magenta) and mild polar (blue). This figure
illustrates how the loss of the C-8 hydroxyl is a potential loss of
a hydrogen bond contact with RNAP.
We acknowledge generous support by the University of Michi-
gan College of Pharmacy Ella and Hans Vahlteich and UpJohn
Research Funds. We would also like to acknowledge additional
funding by the University of Michigan Office of the Vice President
for Research, and the Rackham Graduate School.
One of the clinical liabilities of RMP (1) includes its induction of
cytochrome P450 3A4 (CYP3A4), which is mediated by the Human
Pregnane X Receptor (PXR).12 Therefore we utilized a commercially
available assay kit24 to examine if representative analogues (8a, 9a,
and 10a) activate the PXR. The results (Table 3, Fig. 2, S-Fig. 2)
show that the methylamino derivatives of RMP S (8a) and RMP
(10a) do activate the PXR to an extent very similar to that for
rifampin, whereas the RMP S pyrazole derivative (9a) does not
exhibit significant hPXR activation at concentrations up to
Supplementary data
Supplementary data (details of the chemical syntheses, the
in vitro RNAP assays, molecular modeling, and the human preg-
nane X receptor (PXR) activation assay) associated with this article
100 lM. From these data, we cannot yet address whether these
analogues are Cyp inhibitors or substrates.
References and notes
In summary, we have synthesized a novel series of rifamycin S
and rifampin analogues incorporating substituted 8-amino, 8-thio,
and 1,8-pyrazole substituents. Screening the compounds for
inhibition of WT M. tuberculosis (MTB) RNAP and rifamycin-resis-
tant MTB RNAP (S450L) as well as antitubercular effects under both
1. WHO. WHO Report 2008.
2. Mitchison, D. A. Front. Biosci. 2004, 9, 1059.
3. Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.; Jensen,
J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S. J.; Windus, T. L.; Dupuis, M.;
Montgomery, J. A. J. Comput. Chem. 1993, 14, 1347.
4. Calvori, C.; Frontali, L.; Leoni, L.; Tecce, G. Nature 1965, 207, 417.