Research article

Netropsin improves survival from endotoxaemia by disrupting HMGA1 binding to the NOS2 promoter

Marianne A. Grant, Rebecca M. Baron, Alvaro A. Macias, Matthew D. Layne, Mark A. Perrella, Alan C. Rigby


The inducible form of nitric oxide synthase (NOS2) plays an important role in sepsis incurred as a result of infection with Gram-negative bacteria that elaborate endotoxin. The HMGA1 (high-mobility group A1) architectural transcription factor facilitates NOS2 induction by binding a specific AT-rich Oct (octamer) sequence in the core NOS2 promoter via AT-hook motifs. The small-molecule MGB (minor-groove binder) netropsin selectively targets AT-rich DNA sequences and can interfere with transcription factor binding. We therefore hypothesized that netropsin would improve survival from murine endotoxaemia by attenuating NOS2 induction through interference with HMGA1 DNA binding to the core NOS2 promoter. Netropsin improved survival from endotoxaemia in wild-type mice, yet not in NOS2-deficient mice, supporting an important role for NOS2 in the beneficial effects of MGB administration. Netropsin significantly attenuated NOS2 promoter activity in macrophage transient transfection studies and the AT-rich HMGA1 DNA-binding site was critical for this effect. EMSAs (electrophoretic mobility-shift assays) demonstrated that netropsin interferes with HMGA1 NOS2 promoter binding and NMR spectroscopy was undertaken to characterize this disruption. Chemical shift perturbation analysis identified that netropsin effectively competes both HMGA1 DNA-binding AT-hooks from the AT-rich NOS2 promoter sequence. Furthermore, NOESY data identified direct molecular interactions between netropsin and A/T base pairs within the NOS2 promoter HMGA1-binding site. Finally, we determined a structure of the netropsin/NOS2 promoter Oct site complex from molecular modelling and dynamics calculations. These findings represent important steps toward refined structure-based ligand design of novel compounds for therapeutic benefit that can selectively target key regulatory regions within genes that are important for the development of critical illness.

  • endotoxaemia
  • high-mobility group A protein
  • lipopolysaccharide
  • minor-groove binder
  • netropsin
  • nitric oxide synthase (NOS)


Sepsis represents a devastating clinical condition in which systemic infection leads to multi-system organ failure and high mortality [1]. A subset of patients with sepsis suffer from infections with Gram-negative bacteria that elaborate endotoxin, leading to propagation of a downstream inflammatory cascade, production of pro-inflammatory cytokines, and transcriptional up-regulation of the inducible form of nitric oxide synthase (NOS2, also abbreviated iNOS) that results in production of nitric oxide (NO) and development of refractory hypotension [2]. Complete inhibition of NOS2 induction has not proved to be beneficial in animal models or human trials of sepsis, and numerous studies have supported a beneficial role for some level of NOS2 expression (e.g. in wound healing and intracellular bacterial killing) [3]. Therefore the ability to administer a compound that predictably attenuates NOS2 expression in vivo represents an important avenue of novel therapeutic investigation in critical illness.

Our group demonstrated an important role for the architectural TF (transcription factor) HMGA1 (high-mobility group A1) in up-regulating NOS2 expression in vitro [46]. HMGA1 binds to the AT-rich Oct (octamer) site (bases −61 to −54) within the NOS2 promoter and, through formation of a ternary complex, recruits binding of the p50 subunit of NF-κB (nuclear factor κB) to its cognate binding site upstream of the Oct site (bases −84 to −75). Optimal induction of NOS2 promoter activity requires HMGA1 DNA binding to the AT-rich region of the core NOS2 promoter. Therefore the ability to interfere with HMGA1 binding to the NOS2 promoter represents a potential mechanism for targeted regulation of NOS2 expression as a therapeutic strategy to improve survival during endotoxaemia.

HMG proteins are non-histone chromosomal proteins that function as architectural TFs in organizing the assembly of nucleoprotein–DNA transcriptional complexes for the activation and/or repression of transcription [7,8]. Whereas most TFs recognize and bind their cognate DNA promoter sites through the major groove of DNA, HMGA1 (formerly HMG-I/Y) is one of the few TFs that preferentially bind AT (adenine thymine)-rich sequences through the minor groove [9]. A key feature of the HMGA1 TF is the presence of three DBDs (DNA-binding domains), referred to as AT-hooks, which mediate selective interaction with AT-rich DNA sequences in the minor groove [9]. A single AT-hook preferentially binds a stretch of 4–6 base pairs within an AT-rich DNA sequence. The NMR structures of the second and third AT-hooks of HMGA1 in the presence of DNA identified that the central Arg-Gly-Arg core of each AT-hook is responsible for penetrating and anchoring these DBDs into the minor groove, whereas a pair of lysine and arginine residues at either end of this core mediate electrostatic and hydrophobic contacts with the DNA backbone [10]. The second AT-hook of HMGA1 demonstrates high-affinity DNA binding through a network of residues C-terminal to the core (Gly-Ser-Lys-Asn-Lys) that interact specifically with the sugar–phosphate backbone of adjacent DNA bases in the minor groove [10].

MGB (minor-groove binder) drugs represent a class of small molecules, including the antibiotics netropsin, distamycin A and berenil, that recognize AT-rich sequences within the minor groove of DNA through a sequence- and conformation-dependent mechanism similar to that proposed for the HMGA1 AT-hooks [10]. We and other groups have used distamycin A in vitro to disrupt HMGA1 DNA binding and thus prevent the formation of a competent transcriptional complex [46,11,12]. In fact, we demonstrated that distamycin A interfered with AT-rich DNA binding to the minor groove of NOS2 promoter DNA, decreased NOS2 induction and resulted in intermediate levels of NOS2 expression during murine endotoxaemia in vivo that proved beneficial for survival and attenuation of hypotension [13]. However, the molecular mechanism of the effect of distamycin A has not been fully characterized and we note that distamycin A is not water-soluble and is poorly tolerated by humans in its nascent form, although derivatives of this drug have been used as vehicles to target chemotherapeutic agents to DNA [14]. Netropsin is an MGB that avidly binds the HMGA-binding sites within promoter elements and is thus capable of effectively competing with HMGAs for DNA binding [11,12,15]. However, to our knowledge, use of netropsin has not been studied in vivo during critical illness, and the molecular understanding of HMGA1 DNA binding to the core NOS2 promoter has not been reported previously. We therefore set out to determine (i) whether the beneficial effects of MGBs in vivo during endotoxaemia extend beyond that of a single agent within this class of drugs, and (ii) the mechanism of MGB interference with AT-rich DNA binding within a particular region of the NOS2 promoter.

In the present paper, we report molecular and structural data demonstrating that the water-soluble MGB netropsin proves beneficial during murine endotoxaemia and interferes with HMGA1 binding to the AT-rich core NOS2 promoter in a sequence- and conformation-specific fashion. We show that netropsin improves murine survival from endotoxaemia in WT (wild-type) mice, but not in mice treated with a NOS2-specific inhibitor. In addition, we show that netropsin attenuates LPS (lipopolysaccharide)-induced NOS2 promoter activity in murine macrophages and that the AT-rich HMGA1-binding site in the NOS2 promoter is critical for this effect. Furthermore, we provide direct evidence that netropsin disrupts interactions between the high- and low-affinity DNA-binding AT-hook motifs of HMGA1 and NOS2 Oct site DNA using EMSAs (electrophoretic mobility-shift assays) and solution NMR spectroscopy. Additional thermal denaturation curve analyses and two-dimensional NMR spectroscopy data demonstrate that netropsin binds A/T base pairs in the minor groove of the NOS2 Oct site. These data provide the first details for the molecular basis of netropsin's interaction with the AT-rich Oct site within the NOS2 promoter. Importantly, our studies demonstrate the feasibility of correlating biological findings with detailed structure–function data as a means of developing small molecules with therapeutic potential.


Mouse model of endotoxaemia

Male C57BL/6 WT mice (Charles River Laboratories) 6–8 weeks of age were treated with PBS (vehicle) or with 20 mg/kg (subcutaneously) of the NOS2-specific inhibitor 1400W (Cayman Chemical) twice daily. All mice were injected i.p. (intraperitoneally) with 40 mg/kg LPS (Escherichia coli serotype O26:B6; Sigma) and with either 25 mg/kg i.p. netropsin (Calbiochem) or vehicle 30 min before LPS injection.

Transient transfections of RAW264.7 cells and reporter assays

Murine RAW264.7 macrophages (A.T.C.C.) were cultured as described in [5]. The generation of the mouse NOS2 luciferase reporter plasmids ‘NOS2-WT’ (−234/+31) and ‘NOS2-Mut’, which contains a mutated AT-rich Oct site (bases −61 to −54, ATGCAAAA to CGTACCCC) have been described previously [6,16]. These plasmids were transiently transfected into RAW264.7 cells (FuGENE 6 transfection reagent; Roche Applied Science), as described previously [17]. At 12 h following transfection of the reporter construct and a β-galactosidase expression vector (to normalize for luciferase activity), the cells were treated with netropsin (10–25 μM) or vehicle (water), followed by treatment with LPS (50 ng/ml in PBS) 30 min later. At 24 h following treatment, cells were harvested and assessed for luciferase activity and β-galactosidase, as described previously [17].


EMSAs were performed as described previously [5] using a 32P-labelled double-stranded oligonucleotide probe encoding the AT-rich Oct-binding site and flanking sequence of the murine NOS2 promoter (bases −65 to −50, AGTTATGCAAAATAGC) or control probe containing the Sp1 (specificity probe 1) consensus binding site (ATTCGATCGGGGCGGGGCGAGC). Either a 43-amino-acid synthetic peptide comprising the second and third AT-hook DNA-binding regions of the HMGA1 protein, HMGA1(2/3) (Peptide Synthesis Core Facility, Tufts University, Boston, MA, U.S.A.) or recombinant Sp1 protein generated as described previously [5] from in vitro transcription and translation of a Sp1 expression vector were used. Given the small size of the HMGA1(2/3) peptide, EMSAs using this peptide were run on a 6% polyacrylamide gel. Netropsin (0.25 μM) or vehicle (water) was incubated with the radiolabelled probe for 1 h before gel electrophoresis.

Competition binding NMR spectroscopy

15N-Labelled glycine (15N, 98%; Cambridge Isotope Laboratories) was selectively incorporated at Gly11, Gly20 and Gly37 during the chemical synthesis of the HMGA1(2/3) peptide (Peptide Synthesis Core Facility, Tufts University). NMR samples contained 0.5 mM [15N]Gly11, [15N]Gly20, [15N]Gly37-HMGA1(2/3) peptide, 10 mM sodium phosphate (pH 5.7) in 1H2O/2H2O (9:1). All NMR spectra were acquired on a Varian INOVA 500 MHz spectrometer at 25 °C. The carrier frequency was set to the water resonance, which was suppressed using presaturation during the pre-acquisition delay. Two-dimensional 1H–15N-HSQC (heteronuclear single-quantum correlation) spectra were acquired with spectral widths of 8000 Hz and 2127 Hz for 1H and 15N respectively, with 2048 real data points in t2 and 256 transients for each of the 380 increments in t1 [18]. Complementary 16 bp oligonucleotides containing the NOS2 promoter Oct site (5′-AGTTATGCAAAATAGC-3′) were annealed following thermal denaturation at 95 °C using a cooling rate of 1 °C/min. Annealed NOS2 Oct DNA was titrated into the HMGA1(2/3) sample to a ∼2:1 stoichiometric equivalence, and netropsin was titrated into the HMGA1(2/3)–DNA complex from a concentrated stock in phosphate buffer. 1H–15N resonance assignments for the DNA-bound HMGA1(2/3) peptide were made from two-dimensional 1H–15N-HSQC spectra of DNA-complexed HMGA1(2/3) peptide with single site [15N]glycine incorporation at either Gly11 or Gly37.

Thermal DNA melting studies

UV DNA-melting curves were measured using a Cary 3E UV–visible spectrophotometer (Varian) equipped with a thermoelectric temperature controller. Stock solutions of complementary 10 bp oligonucleotides containing an AT-rich fragment of the Oct site and its flanking sequence from the NOS2 promoter (5′-GCAAAATAGC-3′) or mutated sequence of the Oct site (5′-GCACCCGAGC-3′; mutated residues are underlined) were annealed as described above. Samples of the dsDNAs (double-stranded DNAs) at concentrations of 50 μM alone or in the presence of netropsin (1:1 molar ratio) were prepared by directly mixing aliquots from each of the stock solutions (dsDNAs and netropsin), followed by incubation for 1 h at 25 °C to ensure equilibration. Samples were heated to 80 °C at a rate of 1 °C/min while continuously monitoring the absorbance at 260 nm. Data were analysed with the program Origin (Microcal).

Two-dimensional 1H–1H NOESY

Proton assignments for netropsin, NOS2 Oct dsDNA, or HMGA1(2/3) peptide were determined from one- and two-dimensional 1H-NMR data recorded at 25 °C on a 500 MHz Varian Unity INOVA spectrometer. Two-dimensional NOESY spectra were recorded with a mixing times of 50–250 ms. A spectral width of 10999.4 Hz was used in both dimensions, and a total of 2048 real data points were acquired in t2 and 256 TPPIs (time-proportional phase increments) were implemented in t1. A total of 160 summed transients were collected with a relaxation delay of 1.5 s. Spectra were processed with Gaussian and sine bell functions for apodization in t2 and a shifted sine bell function in t1 using the VNMR software processing (Varian). All data were zero-filled to 4 K by 4 K real matrices. In the case of the NOS2 Oct DNA–netropsin complex, netropsin was titrated into the 10 bp NOS2 Oct dsDNA sample from a concentrated stock solution.

Netropsin–DNA complex homology structure generation

Initial co-ordinates for netropsin and NOS2 Oct DNA (5′-GCAAAATAGC-3′) were based on the 2.2 Å (1 Å=0.1 nm) resolution X-ray crystal structure of the netropsin–DNA (5′-GCAAATTTGC-3′) complex (PDB code 121D) [19] with divergent bases (not underlined) replaced using InsightII software (Accelrys). The DISCOVER program was used to carry out energy minimization in vacuo using 5000 steepest descent steps keeping the entire complex fixed except for the newly altered bases and then again by 5000 steepest descent steps without constraints. Further energy-minimization refinement from Molecular Dynamics simulations in explicit solvent were carried out using YASARA (Yet Another Scientific Artificial Reality Application; with the Yamber3 force field [20,21] and a three-step protocol: (i) a short steepest descent minimization until the maximum atom speed dropped below 2200 m/s; (ii) 500 steps of simulated annealing; and (iii) a 500 ps Molecular Dynamics run carried out at 298 K and 1 atm (101.325 kPa) using simulated annealing. Multiple time steps of 2 fs for intramolecular forces and 1 fs for intermolecular forces were used. The time-averaged temperature was controlled using the Berendsen thermostat algorithm [22] and atom velocity rescaling. The initial velocities were obtained from a Maxwellian distribution at the desired temperature and rescaled every ∼100 steps whenever the average of the last ∼100 measured temperatures converged. The system pressure was controlled by rescaling the periodic boundaries according to fluctuations in the time-averaged solvent density. Snapshots were taken every 25 ps, and trajectories were analysed and validated iteratively using YASARA–WHATIF Twinset software.


NOS2 is critical for mediating the protective effect of netropsin during murine endotoxaemia

WT mice were administered with LPS+netropsin or LPS+vehicle and assessed for survival for 90 h following injection (Figure 1A). Dose–response experiments were performed to determine the dose of LPS that would result in mortality comparable with that reported in the literature [13,2327] and an effective dose of netropsin that was not toxic for the animals (results not shown). WT mice exposed to LPS+vehicle exhibited 100% mortality (0% survival) during the first 60 h following injection. WT mice injected with LPS+netropsin exhibited a significantly higher survival rate (36% survival) at 70–90 h following injection (P=0.03 compared with LPS+vehicle). Mice treated with the NOS2-specific inhibitor 1400W were similarly administered with LPS+netropsin or LPS+vehicle and assessed for survival for 90 h following injection (Figure 1B). 1400W-treated mice exposed to LPS+vehicle exhibited 100% mortality (0% survival) within 50 h. However, in contrast with the WT mice in Figure 1(A), the 1400W-treated mice receiving LPS+netropsin did not exhibit a significant improvement in survival rates compared with those treated with LPS+vehicle.

Figure 1 Netropsin improves survival from endotoxaemia in WT mice, but not in WT mice treated with a NOS2-specific inhibitor, 1400W

(A) WT mice were injected i.p. with LPS (40 mg/kg)+netropsin (25 mg/kg) (■, WT LPS+Net; n=11) or LPS (40 mg/kg)+vehicle (□, WT LPS; n=10). Mice were evaluated for survival every 10 h following injection. *P<0.05 compared with WT LPS. (B) 1400W-treated WT mice were subjected to the same experimental protocol as outlined for the WT animals (n=6 in each treatment group). No significant difference was found between the two groups of 1400W-treated animals.

Netropsin attenuates LPS-induced NOS2 promoter activity via the HMGA1-binding site

RAW264.7 cells were transiently transfected with the NOS2-WT luciferase promoter–reporter plasmid, then harvested for luciferase activity following treatment with LPS+vehicle or LPS+netropsin (Figure 2A, left-hand side). LPS-induced NOS2 promoter activity was significantly attenuated following the addition of 10 and 25 μM netropsin (reduced to 79.7±0.08 and 44.9±0.08% of the promoter activity observed with LPS+vehicle for 10 μM and 25 μM netropsin doses respectively; P<0.05 for LPS+vehicle compared with both LPS+netropsin treatment conditions). We next examined whether the AT-rich HMGA1-binding site in the NOS2 promoter was required for the attenuation of LPS-induced NOS2 promoter activity by netropsin. RAW264.7 cells were transiently transfected with the NOS2-Mut promoter–reporter construct [6], then harvested for luciferase activity 24 h after treatment with LPS+vehicle or LPS+netropsin (Figure 2A, right-hand side). In contrast with the NOS2-WT plasmid, no significant reduction in LPS-induced NOS2 promoter activity was observed following addition of 10 or 25 μM netropsin.

Figure 2 Netropsin attenuates induction of NOS2 promoter activity through the AT-rich Oct-binding site in LPS-treated mouse macrophages

Netropsin interferes with HMGA1 binding to the NOS2 promoter. (A) RAW264.7 mouse macrophages were transfected with 0.5 μg of the NOS2-WT luciferase promoter–reporter construct (NOS2-WT, left-hand side) or 0.5 μg of the NOS2-Mut Oct promoter–reporter construct (NOS2-Mut, right-hand side), then treated with LPS+vehicle or LPS+netropsin. Cells were harvested for luciferase activity, which was normalized for β-galactosidase, then compared with LPS+vehicle (Veh)-treated cells (normalized to 100%) to determine the percentage reduction with netropsin (Net) treatment (*P<0.05 compared with LPS+vehicle-treated cells). The experiment was repeated a minimum of three separate times, with each condition performed in duplicate during each experiment. NS, not significant. (B) EMSA was performed using HMGA1(2/3) peptide and a radiolabelled probe spanning the NOS2 promoter Oct site or recombinant Sp1 generated by in vitro transcription/translation (TNT) or empty vector (V) subjected to TNT reaction and a radiolabelled probe containing the consensus Sp1-binding site. Then, 100–500 ng of HMGA1(2/3) was incubated with radiolabelled probe and either vehicle (lanes 2–3) or 0.25 μM netropsin (Net; lane 4). Equivalent amounts of empty vector (V) or Sp1 were incubated with radiolabelled probe and either vehicle (lanes 6 and 7) or 0.25 μM netropsin (Net; lane 8). Reaction mixtures were subjected to electrophoresis. As a control for each probe, reaction mixtures were incubated without the addition of protein (lanes 1 and 5). The experiment was repeated three separate times. Prot, protein. *, Nucleoprotein complex represents HMGA1 DNA binding; **, nucleoprotein complex represents Sp1 DNA binding.

Netropsin interferes with HMGA1 binding to the Oct site of the NOS2 promoter

To assess directly the effect of netropsin on HMGA1 binding at the Oct site of the NOS2 promoter, an EMSA was performed using a synthetic HMGA1 peptide, HMGA1(2/3), that possesses two DBD AT-hook motifs, DBD2 and DBD3 (see Supplementary Figure S1 at The HMGA1(2/3) peptide was incubated with a radiolabelled probe spanning the AT-rich Oct-binding site and flanking sequence of the NOS2 promoter (bases −65 to −50) (see Supplementary Figure S1) as well as vehicle or netropsin (0.25 μM). The reaction mixtures were then subjected to electrophoresis (Figure 2B). HMGA1 DNA binding was observed at baseline, represented by a nucleoprotein complex of increasing intensity, with addition of 100 or 500 ng of HMGA1(2/3) peptide (Figure 2B, lanes 2 and 3). We observed a significant reduction in HMGA1 DNA binding following the addition of netropsin (Figure 2B, lane 4). Netropsin does not non-specifically reduce TF binding to DNA as indicated by retained binding of Sp1 to its consensus binding site in the presence of netropsin (Figure 2B, lane 8 compared with lane 7). As expected, no DNA complex was formed in the absence of protein (Figure 2B, lanes 1 and 5) or when empty vector control in vitro translation/transcription reaction product was used (Figure 2B, lane 6).

Netropsin competes away HMGA1 DNA binding

To examine netropsin-specific perturbations of the AT-hook-mediated interaction of HMGA1 with the AT-rich Oct DNA site in the NOS2 promoter, we pursued qualitative NMR studies. We monitored the interaction of each HMGA1 AT-hook (DBD2 and DBD3) with NOS2 Oct DNA using a chemically synthesized HMGA1(2/3) peptide in which [15N]glycine was selectively incorporated at Gly11, Gly20 (in DBD2) and Gly37 (in DBD3) (see Supplementary Figure S1). Aliquots of annealed 16 bp NOS2 Oct DNA were added incrementally to 15N-labelled HMGA1(2/3) peptide in the same buffer. The proton chemical shift frequencies in the amide region of these data were moderately degenerate in the absence of DNA (results not shown), suggesting that the DBD AT-hook motifs are not well-folded in solution. However, in the presence of the NOS2 Oct DNA, these amide proton resonances become well dispersed and demonstrate narrow resonance linewidths (results not shown). These data suggest induction of a structured backbone conformation in the AT-hooks of HMGA1(2/3) as a result of NOS2 Oct DNA binding.

We monitored chemical shift perturbations in the 1H–15N-HSQC spectrum of 15N-HMGA1(2/3) during the course of NOS2 Oct DNA titration (Figure 3). As noted in the absence of DNA, the backbone amide proton resonance for the three [15N]glycine residues (positions 11, 20 and 37) of HMGA1(2/3) are degenerate (Figure 3A). Representative spectra from several titration points (Figures 3B–3H) demonstrate differential chemical shift perturbations of the backbone amide proton resonances of the labelled glycine residues following DNA interaction. Under our experimental conditions, one molecule of HMGA1(2/3) binds two molecules of NOS2 Oct DNA. The DBD2 AT-hook binds in the intermediate exchange regime based on the chemical shift perturbation of the backbone amide proton resonance of Gly11 at ratios of DNA/protein that are less than 1:1. At a molar ratio of 1:0.48, this Gly11 cross-peak is shifted to a frequency representative of the bound conformation and does not shift further following the addition of more DNA. In contrast, the backbone amide proton of Gly37 from DBD3 is in fast exchange and its cross-peak in the HSQC spectrum only begins to shift from its position in the unbound (free) state at a protein/DNA ratio greater than ∼1:0.72, and is in the bound conformation at a ratio of 1:1.2. No chemical shift perturbation is observed for the remaining backbone amide proton resonance of Gly20. Gly20 is positioned at the C-terminal end of the DBD2 AT-hook where a flexible linker begins that joins the two DBD motifs.

Figure 3 Chemical shift mapping of selectively 15N-labelled glycine residues in HMGA1(2/3) upon NOS2 DNA binding and addition of netropsin

1H–15N-HSQC spectra illustrating the chemical shift perturbations of the amide backbone resonances of Gly11 and Gly37 in [15N]Gly11,[15N]Gly20,[15N]Gly37-HMGA1(2/3) upon the addition of increasing amounts of a 16 bp DNA sequence containing the Oct site from the NOS2 promoter (5′-AGTTATGCAAAATAGC-3′). Molar equivalents of NOS2 Oct DNA added are as follows: (A) 0; (B) 0.04; (C) 0.24; (D) 0.32; (E) 0.4; (F) 0.48; (G) 0.72; and (H) 1.2. (I) Addition of netropsin to the HMGA1(2/3)/NOS2 Oct DNA complex (peptide/DNA/netropsin molar ratio 1:1.2:1.2).

To examine netropsin interference with HMGA1 AT-hook binding to NOS2 Oct DNA, we added netropsin to the 15N-labelled HMGA1(2/3)–NOS2 Oct DNA complex and observed this sample using NMR spectroscopy. When netropsin was present at a concentration equivalent to the amount of NOS2 Oct DNA, it effectively competed away DNA-bound HMGA1(2/3) (Figure 3I). In the presence of netropsin, the backbone amide proton resonances of 15N-labelled Gly11 and Gly37 shift from their unique DNA-bound positions back to degenerate resonance frequencies nearly identical with the unbound conformation of HMGA1(2/3). These data demonstrate the ability of netropsin to competitively inhibit the binding of both the high-affinity DBD2 and lower-affinity DBD3 AT-hooks of HMGA1 to NOS2 Oct DNA.

Netropsin binds directly to the AT-rich Oct site from the NOS2 core promoter

To validate targeting the Oct site within the NOS2 promoter with the MGB netropsin, we sought evidence of a direct interaction between netropsin and an AT-rich fragment of the Oct site and its flanking sequence from the NOS2 promoter (5′-GCAAAATAGC-3′). For these data, we used UV spectroscopy thermal melting experiments. In the absence of netropsin, the melting temperature (Tm) of unbound NOS2 Oct DNA, i.e. the temperature at which 50% of the dsDNAs denature, was measured to be 47.7 °C (Figure 4). In the presence of an equal molar amount of netropsin, the Tm of the NOS2 Oct DNA increased to 60.7 °C. This shift to a higher Tm in the presence of netropsin supports that netropsin complexes directly with NOS2 Oct DNA and forms a small-molecule–DNA complex that renders the DNA more resistant to the thermal denaturation. In contrast, netropsin did not significantly affect the Tm of mutated Oct-site DNA (5′-GCACCCGAGC-3′). This finding demonstrates that netropsin's interference with HMGA1 binding to the NOS2 promoter is through direct and specific interaction at the core sequence of the Oct site, argues against non-specific effects on HMGA1 protein and validates targeting the Oct site within the NOS2 promoter with the MGB netropsin.

Figure 4 Thermal UV DNA-melting curves

Thermal denaturation curves of the 10 bp NOS2 Oct dsDNA (5′-GCAAAATAGC-3′) (●) and the complex with netropsin (■), and the 10 bp Oct-mutant dsDNA (5′-GCACCCGAGC-3′) alone (○) and with netropsin (□). All reaction mixtures were in 10 mM Tris/HCl (pH 7), 50 mM KCl and 0.5 mM EDTA. The absorbance values have been normalized to allow for a visual comparison of the curves.

Molecular interactions between netropsin and NOS2 Oct DNA

To elucidate molecular interaction between netropsin and NOS2 Oct DNA we applied NMR methods. One-dimensional 1H-NMR spectra of netropsin were collected in the presence and absence of 10 bp NOS2 Oct DNA (5′-GCAAAATAGC-3′) and were used for netropsin resonance assignments along with NMR data reported previously [28,29]. We monitored chemical shift perturbations of the non-exchangeable netropsin resonances as a result of DNA titration in two-dimensional NOESY data. In the presence of NOS2 Oct DNA, the CH5 and CH11 1H resonances of netropsin (see Supplementary Figure S2 at were conformationally perturbed as noted by their enhanced spectral resolution and downfield chemical shift perturbation (results not shown). We also monitored the chemical shifts for the imino proton resonances of Watson–Crick A–T and G–C base pairs in NOS2 Oct DNA as a function of netropsin. We observed significant downfield chemical shift (δ) perturbation and line-broadening of several A–T base-paired imino protons (δ>13.5 p.p.m. [30]), and significantly less chemical shift perturbation and/or no peak broadening of G–C base-paired imino protons (δ<13.5 p.p.m. [30]) (see Supplementary Figure S3 at Thus the association of netropsin with the NOS2 Oct site changes the environment of proton groups on the concave face of netropsin and perturbs specifically the environment of the imino protons in A–T base pairs.

In addition, we examined 1H–1H NOE (nuclear Overhauser effect) cross-peak correlations between netropsin and NOS2 Oct DNA in NOESY spectra. Peak assignments of non-exchangeable DNA base and sugar H1′ protons were made based on sequential H8 or H6 to H1′ NOE connectivities and sequential adenine H2 to H2 NOE connectivities, as well as NOE cross-peak correlations from CH3 thymine protons to H2 adenine protons. Assignments were supplemented with assessments of H8 or H6 to H2′ and H2″ NOE connectivities, as well as observance of upfield CH3 resonances from thymine bases. Several NOE cross-peaks between DNA base protons and netropsin protons were identified in the NOESY data (Figure 5). For example, the H2 protons of adenine bases A4 and A5 (numbered 5′-G1C2A3A4A5A6T7A8G9C10-3′) exhibit NOE cross-peak correlations to the netropsin concave face pyrrole CH11 proton (peaks 1 and 2 respectively) and the H2 protons of adenine bases A6 and A5 exhibit strong NOE cross-peak correlations to the netropsin concave face pyrrole CH5 proton (peaks 3 and 4 respectively). Further inspection of the netropsin CH11 pyrrole proton frequencies reveals NOE cross-peak correlations to the H1′ protons of thymine bases T17 and T16 (complementary strand numbered 5′-G11C12T13A14T15T16T17T18G19C20-3′) respectively (peaks 5 and 6). Similarly, we observed additional cross-peaks reflecting intramolecular interaction between the H2 proton of A6 and the non-equivalent CH2 methylene protons of position 2 in netropsin (results not shown). We did not observe NOE interactions between convex pyrrole CH7 or CH13 protons of netropsin and H2 protons of adenine bases or H1′ protons of thymine bases. These NOE cross-peak correlation data support that the netropsin concave face pyrrole protons CH5 and CH11 and A/T base protons within the NOS2 Oct DNA are positioned ∼3–5 Å from one another within the netropsin–DNA complex, indicating netropsin binding in the NOS2 Oct minor-groove AT-track and its binding mode. These NOE cross-peak correlations establish points of intermolecular contact between netropsin and NOS2 Oct DNA and suggest that the concave face pyrrole CH5 proton of netropsin lies close to the H2 proton of bases A5 and A6, whereas the pyrrole CH11 proton of netropsin is juxtaposed with the H2 proton of base A5 and H1′ protons of bases T16 and T17.

Figure 5 Two-dimensional NOESY data

A comparison of regions from the NOESY spectrum of NOS2 Oct DNA alone (upper panel) to the 1:1 NOS2 Oct DNA–netropsin complex at 25 °C (lower panel). Sweep width in both dimensions is 10999.2 Hz, pulse width is 6.2 μs, recycle delay is 1.5 s; number of complex points in F2 is 2048; number of increments in F1 is 256; number of transients is 128. Observed intermolecular NOE cross-peaks are: 1, NtCH11-A4 H2; 2, NtCH11-A5 H2; 3, NtCH5-A6 H2; 4, NtCH5-A5 H2; 5, NtCH11-T17 H1′; 6, NtCH11-T16 H1′; denoting netropsin proton–DNA base proton correlations. Numbering is based on the 5′-G1C2A3A4A5A6T7A8G9C10-3′ sequence and its complementary strand 5′-G11C12T13A14T15T16T17T18G19C20-3′.

Molecular modelling of the netropsin–NOS2 Oct DNA complex

Computational examination of netropsin binding to the minor groove of the AT-rich fragment of the NOS2 Oct site was carried out using molecular modelling and Molecular Dynamics simulation. Assessments of the lowest-energy complex structures indicate that hydrogen bonds, van der Waals interactions and electrostatic forces play important roles in stabilizing this netropsin–DNA complex. Consistent with experimentally determined crystal structures of netropsin–AT-rich DNA complexes [19,31,32], netropsin is positioned squarely in the 6 bp 5′-AAAATA-3′ region of the minor-groove NOS2 Oct site, interacting with six bases on one strand and four bases on the other (Figure 6). The NH amide groups of netropsin form bifurcated hydrogen bonds to adenine base N3 atoms and thymine base O2 atoms on opposite DNA strands in the AAAT core (see Supplementary Figure S4 at Several stabilizing van der Waals interactions between O4′ atoms of the DNA sugars and netropsin atoms are observed especially at the guanidinium and amidinium ends of netropsin (see Supplementary Figure S4). The narrow width of the minor groove increases the extent of van der Waals interactions between netropsin and atoms along the floor and walls of the DNA minor groove. Electrostatic interactions between the negatively charged groups of the AT-stretch minor groove and the positively charged end groups of netropsin may also play a role in complex formation and the sequence selectivity of netropsin. This netropsin–NOS2 Oct complex structure is consistent with our biochemical and NMR data, as well as the wealth of structural data available for netropsin interaction with AT-tracks, and provides a powerful tool to understand the molecular basis for netropsin interaction at this AT-rich minor-groove site as well as netropsin's ability to effectively compete for DNA binding with the AT-hooks of the HMGA1 TF. These are the first structural details describing the mechanism of interference in HMGA1 activity within the NOS2 promoter at the Oct site using the MGB netropsin.

Figure 6 Model of the NOS2 Oct DNA–netropsin complex

Space-filling model of the lowest-energy structure resulting from Molecular Dynamics simulations of netropsin with NOS2 Oct DNA (5′-GCAAAATAGC-3′) emphasizing the relative orientation of netropsin within the DNA minor groove. Left-hand panel, both DNA strands are shown as a space-filling model; right-hand panel, one of the two strands is shown in stick rendering.


In the present study, we have demonstrated that: (i) the water-soluble MGB drug netropsin improves survival from murine endotoxaemia; (ii) NOS2 expression plays an important role in this protective effect; and (iii) netropsin modulates NOS2 expression by disrupting HMGA1 DNA binding to an AT-rich region of the NOS2 promoter in a sequence-specific fashion. These findings highlight two important new concepts. First, these results, in combination with our previous in vivo findings using the MGB distamycin A [13], suggest that MGBs as a class of drugs hold promise for targeted therapeutics in controlling gene expression during critical illness. Secondly, our data represent a novel correlative study, in which NMR was used to explore the molecular basis of an observed effect in a murine model of critical illness. Thus this work provides an important proof of concept in supporting the synergy of in vivo animal model data with biophysical studies for structure-based design of molecules that alter transcriptional regulation of key gene targets.

Although MGB drugs and their derivatives have been explored extensively as chemotherapeutic agents [33,34], their toxicity profile and concerns over lack of binding specificity has limited broader application of the agents to other disease processes. We demonstrated previously that another MGB, distamycin A, similarly improved murine survival from endotoxaemia through attenuating NOS2 induction [13]. Although distamycin A has long been used by our group and others as a tool to interfere with HGMA1 DNA binding in vitro, we found that distamycin A also disrupted the binding of another TF, IRF1 (interferon regulatory factor 1), to an AT-rich upstream site in the NOS2 promoter. Moreover, distamycin A has been used as a tool to study the regulation of other genes by HMGA1 as well, suggesting that MGB drugs can have effects on additional genes, although it is also interesting to speculate that families of genes with similar regulatory regions might be targeted for therapeutic benefit [35]. Previous work using MGBs suggests that an AT-rich DNA sequence alone does not predict binding and, rather, that DNA conformation plays a key role in predicting protein–DNA binding [36]. Thus the sequence- and conformation-dependent effects of distamycin A at more than one TF-binding site within more than one gene highlight the importance of reliable in vivo and in vitro readouts of MGB activity and mechanism of action, such as those described in the present paper. The present study demonstrates the use of NMR to gain an increased molecular understanding of interactions between an MGB agent and an AT-rich DNA sequence that has biological relevance to human critical illness. This information provides an important framework for the extension of our work to the development of novel small-molecule therapeutic agents that can alter gene transcription during disease.

We chose to use the MGB netropsin in the present study based upon on the in vivo biological activity of this compound and its notable affinity for HMGA1 DNA-binding sites relative to the other MGB drugs [12]. Moreover, in contrast with distamycin A, netropsin is water-soluble, which removes the need for formulation studies and/or the addition of excipients before administration. Examination of the structures of netropsin and distamycin A reveals that these small-molecules structurally mimic the conformation of the highly conserved Pro-Arg-Gly-Arg-Pro core of the HGMA1 DBD AT-hooks (see Supplementary Figure S2). In addition, all of these compounds reversibly bind to the minor groove of AT-rich sequences with high selectivity. Distamycin A is the tris-N-methylpyrrole analogue of netropsin and is distinguished from netropsin by the presence of a single cationic guanidinium group and a more extended DNA-binding site, as observed in a co-crystal structure with the sequence d(CGCAAATTTGCG)2 [37]. Thus the high affinity of netropsin for minor-groove DNA sites in the NOS2 promoter reflects a greater similarity between the molecular structures of netropsin and the AT-hook motifs of HMGA1 [10,12], as both contain two five-membered rings that are important for presentation of a narrow planar structure and two guanidinium groups that are important in hydrogen-bonding to bases and van der Waals interactions with the narrow minor-groove wall (see Supplementary Figure S4). Studies on the DNA-binding behaviour of MGBs, such as netropsin, have afforded unique insight into the mechanisms whereby these small molecules recognize DNA and thus target the DNA, which may have useful biological and possibly therapeutic effects. One impetus for studies of this sort is to provide feasibility data allowing one to determine whether there might be alternative/novel motifs that are capable of recognizing and thus disrupting TF–DNA interactions in the context of therapeutic development.

Recent years have witnessed an exciting explosion of in vitro minor-groove small-molecule design, primarily with the goal of chemotherapeutic development. For example, a biosensor–SPR (surface plasmon resonance) method was used to monitor HMGA2 binding to immobilized DNA sequences selected by SELEX (systematic evolution of ligands by exponential enrichment) [15]. Whereas HMGA2 is a member of the HMGA family of architectural TFs with properties similar to those of HMGA1, HMGA2 is a distinct gene from HMGA1 and has different gene-regulatory functions [7,8]. Biosensor–SPR demonstrated that netropsin interfered with HMGA2 binding to an AT-rich DNA sequence in vitro, although no effect on a specific target gene or organism in vivo was demonstrated. On the basis of the known binding characteristics of HMGA proteins and an analysis of the promoter structures of genes that are known to be positively regulated by these proteins in vivo, others have proposed that HMGA protein AT-hooks read specific DNA ‘barcodes’ [38]. If so, it should be possible to create MGB compounds that are also capable of recognizing this ‘barcode’ and thus competitively inhibit HMGA binding to genes of interest. A number of studies have addressed targeting minor-groove DNA through the use of molecular docking approaches [13,3942]. In addition, a very recent study has documented important proof-of-concept data describing how one might explore ligand–DNA complexes using computer-aided drug-discovery approaches including pharmacophore modelling [43]. Pharmacophore modelling methods are able to reproduce the experimentally determined preferred binding sites of minor-groove ligands. Furthermore, significant advances have been made in the design and synthesis of small minor-groove-binding polyamide molecules, called lexitropsins, that bind with high affinity to predetermined DNA sequences in the human genome [31,34,44,45]. These compounds are based on an understanding of the structural basis of polyamide–minor-groove DNA interactions provided by the Dervan laboratory, which resulted in the first rationally designed polypyrrole–polyamide compounds that bound specifically to AT-sequences with affinities and specificities comparable with native DNA-binding proteins [14,36,4649]. Triple-helix-forming oligonucleotides (triplexes), synthetic zinc fingers and hairpin polyamides are other sequence-selective representatives of molecules directly targeting DNA [50].

In summary, these results are the first to demonstrate the in vivo effect of an MGB agent in a murine model of critical illness that correlates biological function with sequence-specific inhibition of protein–DNA binding. Moreover, we provide new information regarding the interaction of netropsin with a biologically relevant AT-rich target within the NOS2 promoter. The escalating abundance of genome sequence information is fuelling efforts aimed at the design of compounds capable of modulating the expression of selected genes that have predictable biological effects. Growing information about regulatory regions of genes suggests that such ‘designer’ drug design is becoming increasingly feasible, especially in critical illness during which dysregulated gene expression plays a key pathophysiological role. Moreover, new treatment approaches are desperately needed in the area of critical care medicine, as decades of research have failed to produce any effective targeted treatments for sepsis [1]. The combination of in vivo models of human disease and in vitro assessment of functional DNA–protein binding to target genes will be critical for guiding the rational development of therapeutics in complex disease.


This work was supported by the American Heart Association [grant number SDG 0530348N (to M. A. G.)]; and the National Institutes of Health [grant numbers K08 AI054465 (to R. M. B.), U01 AI061246 (to M. A. P.), R01 GM53249 (to M. A. P.) and R21 AR054442 (to A. C. R.)].

Abbreviations: AT, adenine thymine; DBD, DNA-binding domain; dsDNA, double-stranded DNA; EMSA, electrophoretic mobility-shift assay; HMGA, high-mobility group A; HSQC, heteronuclear single-quantum correlation; i.p., intraperitoneally; LPS, lipopolysaccharide; MGB, minor-groove binder; NOE, nuclear Overhauser effect; NOS2, inducible nitric oxide synthase; Oct, octamer; Sp1, specificity protein 1; SPR, surface plasmon resonance; TF, transcription factor; Tm, melting temperature; VSM, vascular smooth muscle; WT, wild-type


View Abstract