Specific effects of antitumor active norspermidine on the structure and function of DNA

We compared the effects of trivalent polyamines, spermidine (SPD) and norspermidine (NSPD), a chemical homologue of SPD, on the structure of DNA and gene expression. The chemical structures of SPD and NSPD are different only with the number of methylene groups between amine groups, [N-3-N-4-N] and [N-3-N-3-N], respectively. SPD plays vital roles in cell function and survival, including in mammals. On the other hand, NSPD has antitumor activity and is found in some species of plants, bacteria and algae, but not in humans. We found that both polyamines exhibit biphasic effect; enhancement and inhibition on in vitro gene expression, where SPD shows definitely higher potency in enhancement but NSPD causes stronger inhibition. Based on the results of AFM (atomic force microscopy) observations together with single DNA measurements with fluorescence microscopy, it becomes clear that SPD tends to align DNA orientation, whereas NSPD induces shrinkage with a greater potency. The measurement of binding equilibrium by NMR indicates that NSPD shows 4–5 times higher affinity to DNA than SPD. Our theoretical study with Monte Carlo simulation provides the insights into the underlying mechanism of the specific effect of NSPD on DNA.


CD measurements
: CD spectra of DNA at different concentrations of SPD and NSPD. Figure S1 shows CD spectra of DNA with the addition of different concentrations of polyamines, where calf thymus (CT) DNA was adopted for the measurements. For both NSPD and SPD, there seems to be no apparent change in the CD spectra, indicating that the secondary structure retains the B-form under these conditions. CD spectra of CT DNA upon the addition of polyamines were measured at 25°C in 1 mM Tris-HCl buffer (pH 7.5) on a J-720W spectropolarimeter (JASCO, Tokyo, Japan). The DNA concentration was 30 µM in nucleotide units. Polyamine concentrations varied from 10 to 500 µM. The cell path length was 1 cm. Data were collected every 1 nm between 220 and 340 nm at a scan rate of 200 nm/min, and accumulated 3 times.

Details of numerical modeling
The SPD and NSPD conformations shown in Fig. 6 correspond to the lowest-energy conformer for each molecule, and can be obtained by using the energy minimization scheme of Avogadro software under UFF (Universal Force Field) 1 . To address the excluded volume of ammonium and methylene groups, each of these groups is modeled as a hard sphere of diameter d = 0.39 nm in Fig. 6a, similar to the coarse-grained polyethylene model 2 . Each of the three ammonium groups in SPD and NSPD is assigned +1 unit charge.
To model a DNA segment in the coarse-grained model, we adapt a charged DNA segment of one pitch of about 3.4 nm 3 . Figure 6 shows the model DNA segment consisting of 10 pairs of charged spheres (-1 unit charge for each sphere) around a soft cylinder of length H = 3.4 nm and radius RDNA = 1 nm. The charged spheres represent the phosphate groups of typical dimeter s = 0.476 nm. The two charged spheres of each pair are 180° apart around the DNA cylinder (RDNA = 1 nm). The first pair starts at an axial location z = 0.17 nm above the bottom of the cylinder corresponding to the DNA segment (z = 0), and the remaining pairs are oriented 36°and elevated 0.34 nm sequentially. The final pair ends at a height 0.17 nm below the top of the cylinder of the DNA segment (z = 3.4 nm).
In the investigated DNA model, 20 charged sites on the DNA segment are grouped into 10 pairs. The two charged spheres in each pair are separated by 180° around the cylinder and their separation is set at 2 nm, close to the width of a DNA molecule. These 10 pairs of charged groups are arranged from the bottom to the top of the cylinder by rotating 36° for each pair, and any two adjacent pairs are separated by 0.34 nm.
The top and bottom pairs are 0.17 nm away from the ends of the cylindrical cell. In our Monte Carlo simulation, we choose translation and rotation motion randomly with equal probability to sample configurations of the polyamine with temperature at T = 298 K. Also, the step sizes for the above two types of sampling schemes are adjusted to achieve an acceptance ratio around 50-60% under the Metropolis algorithm 6 . For each parameter set, a total of 9 ×10 8 moves are conducted in the simulation, and the first 10 8 moves are discarded to ensure convergence in calculations.
The electrostatic interaction between the monomers of a polyamine and a phosphate group is treated at the level of the primitive model with the screened Coulomb potential to incorporate water, counterions and coions implicitly into the interaction potential, given by where kB is the Boltzmann constant; T is the temperature; r is the separation between a monomer in a polyamine and the coarse-grained phosphate group in the DNA; i is the i-th monomer in the polyamine with charge qi (0 for a methylene group; + 1 for an ammonium group); j is the j-th phosphate group in DNA with charge Qj ; G is the interaction strength; k is the inverse Debye screening length; and rc is approximated to be 0.31 nm, a typically closest distance between a charged amine and a phosphate group in DNA 7 . In this work, G ranges from 0.68 to 1.7 nm, about 1 to 2.4 Bjerrum lengths at 25°C to mimic the low dielectric regime around DNA with low water content, and the inverse Debye screening length k is chosen to be 2.4 nm -1 . This inverse Debye screening length is set to reflect the low-water and low-electrolyte region close to the DNA surface.
In this work, we are particularly interested in the density distribution function of charged monomers (i.e., ammonium groups) in polyamine. We compute the reduced density distribution function of the i-th ammonium group ri(r) (defined in Fig. 6) as a function of its radial distance r measured from the center of the DNA axis for different G. To compute ri(r), the cylindrical simulation cell is divided into 100 layers along the radial direction, and then the average number of the charged monomer i in the k-th layer Bk(r) where vi(r) is the volume of the i-th layer, and Vcyl is the total volume of the cylindrical simulation cell to keep the density distribution function dimensionless. Moreover, the calculation of ri(r) is further extended to the two-variable density distribution function r5(r, cos q) for the ammonium group at the middle of the polyamine (#5 defined in Fig. 6) where q is the angle between the direction of the DNA axis and the unit vector drawn from the ammonium at position #1 to that at position #5 (shown in Fig. 6), given by r5(r, cos q) = Vcyl B5(r, cos q)/(vi(r)Dµ) where µ = cos q and Dµ is the interval of µ in the simulation. In the calculation, we divide the range of µ (between -1 and 1) into 50 intervals with Dµ = 0.04, and the radial distance r (between RDNA -s /2 and Rcell -d/2) into 50 intervals with a total of 2500 intervals in the histogram B5(r, cos q).  indicating that, beyond this distance, the electrostatic interaction decreases significantly due to a more marked separation between oppositely charged ammonium and phosphate. Also, Fig. S3a shows that the two end ammoniums (#1 and #9) in NSPD are identical because they are symmetric to each other against the central monomer (#5) in the chain molecule. In contrast, the two end ammoniums (#1 and #10) in SPD are asymmetric. The density distribution function of #1 ammonium r1 in SPD has a similar magnitude as that of NSPD, and it becomes slightly lower than that of NSPD at smaller r. At the #10 ammonium of SPD, the density distribution function r10 is significantly lower at r < 1.3 nm due to the additional methylene group on the side of the #10 ammonium in SPD. This extra methylene likely reduces the local charge density as well as the electrostatic interaction between the #10 ammonium and the charged phosphates. In Fig. S3b, r5 for the middle ammonium (#5) of SPD is significantly lower than that of NSPD below r < 1.2 nm, like r10 in SPD, and r5 has a similar magnitude for both SPD and NSPD at around r = 1.3 nm, like r 1 in SPD. Namely, r 5 exhibits features in-between r 1 and r 10 for SPD, and serves as a good physical quantity to differentiate the distinct behavior between SPD and NSPD. Figure S3c is the radial density distribution function of all ammonium groups: the density distribution of NSPS is greater than that of SPD below r = 1.4 nm or so, but this trend reverses for r > 1.4 nm.   1 nm), whereas for G = 1.02 nm, r5 shows that the middle ammonium distributes preferentially outside of the soft DNA boundary. For both G values, the r5 of SPD tends to be smaller than that of NSPD should be at around r < 1.4 nm, suggesting that the electrostatic attraction between SPD and DNA is effectively weaker than the interaction between NSPD and DNA.

Additional explanation of the results of the numerical calculation
To further test the effect of electrostatic interaction, in Fig. S4, we plot the r5 of both NSPD (red symbols and lines) and SPD (blue symbols and lines) for G in (Fig. S4a)  respectively, in (Fig. S4b). For weak electrostatic interaction G = 1.02 nm, both NSPD and SPD display a greater tendency to be away from the DNA surface, whereas under strong electrostatic interaction G = 1.7 nm, both NSPD and SPD are more likely to be near the DNA surface. For both G values, when r < 1.4 nm or so, the probability of finding NSPD near DNA is greater than that for SPD, but for r > 1.4 nm, we obverse an opposite trend, in that SPD exhibits a greater probability density distribution than NSPD. This feature is basically similar to that in Fig. S3b. Note that the local peak near r = 1.3 nm is present for the density distribution function r5 for both NSPD and SPD in Figs S3 and S4. The density distribution r5 can clearly show a distinct difference due to electrostatic interaction between NSPD and SPD.
In   Figure S5 is the same plot as in Fig. 7c except that G is increased to 1.7 nm. The qualitative features are the same between Fig. 7c and S5, but for the greater G, the hills and valleys in the free energy difference landscape become more pronounced. Near 1.3 nm, a significant increase in probability is observed for different q-angles due to rotational entropy, which increases rotational degrees of freedom. These features help us understand Fig. S4a, in which, for a greater G, r5(r) has a maximum near DNA due to the more energetically favorable process, but for a smaller G, r5(r) has a maximum away from DNA because polyamine gains entropy through rotation and translational motion along the direction of the DNA axis.  In general, N(binding) increases as G is increased due to stronger electrostatic attractions between polyamine and DNA. For a given Rcyl, the N(binding) of NSPD is greater than that of SPD for all G, consistent with the picture that NSPD interacts more strongly with DNA. For a greater Rcyl, N(binding) tends to be smaller for a given type of polyamine due to competition from configurational entropy. But for the very high G (= 1.7 nm), the effect of Rcyl becomes less important because strong electrostatic attractions suppress configurational entropy.