Title: Robust analysis of synthetic label-free DNA junctions in solution by X-ray scattering and molecular simulation

Transcription

1 Supplementary Information Title: Robust analysis of synthetic label-free DNA junctions in solution by X-ray scattering and molecular simulation Kyuhyun Im 1,5, Daun Jeong 2,5, Jaehyun Hur 1, Sung-Jin Kim 2, Sungwoo Hwang 1, Kyeong Sik Jin 3, *, Nokyoung Park 1, *, and Kinam Kim 4 1 Frontier Research Laboratory, Samsung Advanced Institute of Technology, Samsung Electronics, Yongin, Gyeonggi-do , South Korea 2 Computational Science Group, Samsung Advanced Institute of Technology, Samsung Electronics, Yongin, Gyeonggi-do , South Korea 3 Pohang Accelerator Laboratory, Pohang University of Science and Technology, Pohang , South Korea 4 Samsung Advanced Institute of Technology, Samsung Electronics, Yongin, Gyeonggi-do , South Korea 5 These authors contributed equally to this work * Correspondence should be addressed to Nokyoung Park (n2010.park@samsung.com) and Kyeong Sik Jin (jinks@postech.ac.kr)

2 Figure S1. Gel electrophoresis analysis of branched DNA. Lane 1: ss DNA (36mer), Lane 2: 3WJ DNA, Lane 3: 4WJ DNA.

3 Table S1. Oligonucleotide sequences Strand Sequence 3WJ DNA Y1 Y2 Y3 CGA CCG ATG AAT AGC GGT CAG ATC CGT ACC TAC TCG CGA GTC GTT CGC AAT ACG ACC GCT ATT CAT CGG TCG CGA GTA GGT ACG GAT CTG CGT ATT GCG AAC HAC TCG X1 CGA CCG ATG AAT AGC GGT CAG ATC CGT ACC TAC TCG 4WJ DNA X2 X3 CGA GTA GGT ACG GAT CTG CGT ATT GCG AAC GAC TCG CGA GTC GTT CGC AAT ACG GCT GTA CGT ATG GTC TCG X4 CGA GAC CAT ACG TAC AGC ACC GCT ATT CAT CGG TCG

4 Figure S2. Guinier plots of scattering curves in response to change in temperature. Log I(q) versus q 2 for 3WJ (a) and 4WJ (b) in PBS buffer as a function of temperature. For clarity, each plot is shifted along the log I(q) axis. The straight lines were obtained from linear regression in the q 2 region.

5 Table S2. Structural parameters obtained from SAXS analysis of 3WJ and 4WJ in response to change in temperature Sample Temperature (ºC) R a g,g (nm) R b g,p(r) (nm) D c max (nm) P 1st /P d 3rd (a.u.) MD ± ± ± ± ± ± ± WJ ± ± ± ± ± ± ± ± ± ± MD ± ± ± ± ± ± ± WJ ± ± ± ± ± ± ± ± ± ± a R g,g (radius of gyration) was calculated from the Guinier analysis of SAXS data. b R g,p(r) (radius of gyration) was calculated from the p(r) function by the program GNOM. c D max (maximum dimension of particle) was obtained from the p(r) function by the program GNOM. d P 1st /P 3rd is the ratio of the 1 st peak intensity value to the 3 rd peak intensity value in the p(r) function.

6 Equilibrium structures of 3WJ and 4WJ DNA at room temperature from MD simulations a. Comparison of two force fields Here, we discuss 2 MD simulation results employing different force field parameters for nucleic acids in comparison with experimental results. The structures averaged over 100 ns for 2 models at T = 300 K, of which parameters parmbsc0 and parm99 were chosen, are displayed in Figure S3. The average structure obtained from the parmbsc0 force field is found to be a symmetric and T-shaped 3WJ DNA; in contrast, the parm99 force field yields a Y-shape. Various geometric quantities computed from the 2 models are compiled in Table S3, by using 3DNA software for structural analysis. In both cases, a few broken base pairs were observed despite the branched structure, and this is the main cause of inconsistency with experimental results that will be discussed later. The radius of gyration and helix radius of the parmbsc0 model is smaller than that of the parm99 model. It is of interest that the width of the major groove of the former is smaller than that of the latter, and the width of the minor groove is opposite. Helix radius is defined as the radial displacement of atom P in the local helix frame of each dimer. In general, the 3WJ from parmbsc0 is smaller than that from parm99. This is also revealed by displaying the averaged WAXD patterns computed from MD simulation results of the two models (Fig. S4). Differences between the 2 models can be understood as a result of the modified potential parameters for dihedral angles α and γ in the backbone, which were added in the parmbsc0 model. The dihedral potentials endow temporal stability on the double helix structure, which is consistent with recent experimental results. Enhanced stacking, originating from the potential function, is believed to induce 2 helices to form a straight double helix of 36 base pairs and a closed junction with the other arm in the resultant structure. It also reduces fluctuations of the global structure, which is revealed as smaller standard deviations of geometric quantities (Table S3).

7 We now turn to the discussion on the validity of the 2 force fields, with reference to the experimental results, by analyzing p(r). The results from SAXS experiments are considered an ensemble average of the equilibrium structures. To obtain the corresponding quantities from the simulations, we obtain the average of p(r), computed from 50 configurations separated by 2 ns in the MD simulation trajectory. P(r) from individual configurations is depicted in Figure S3 with black (thin) lines, whereas their averages are in red (thick) lines. The trajectories of 100 ns might not be sufficient to ensure the ergodicity of the system. However, we consider these trajectories as equilibrium ensembles obtained after performing a sufficiently long equilibration of 100 ns, while monitoring the stationarity of structural quantities, such as the root-mean squared deviation and number of hydrogen bonds (results not shown). The simulation results of p(r) from 2 force fields generally coincide with experimental results, with respect to the positions of main peaks and the decreasing behavior of tails at large r. However, a couple of discrepancies should be addressed in more detail. First, the difference between the heights of 2 peaks is very small for both cases compared with the experimental result. Second, the positions of the first peak in the 2 models are located at a slightly smaller r. Third, the second peak disappears in the averaged p(r) from simulation results. Finally, the tails at large r (>8 nm) do not coincide with each other, particularly in the case of the parm99 model. Considering that the reconstructed structure for 3WJ DNA from the experiments was distorted Y-shape, the parm99 model, which produced Y-shaped structures, would exhibit greater consistency with the experimental result in Figure S3; however, this is not the case. Although some individual curves have the second peak, the position of the third peaks of those curves does not match. In particular, the tail at large r exhibits a marked deviation from the experimental result. On the other hand, in the parmbsc0 model, we do not observe the second peak in p(r) from individual configurations, while the

8 positions of the third peak and tail exhibit a reasonable coincidence with the experimental result (Fig. S3). The 2 models display different level of the variance in the configurational ensemble. As pointed out earlier, the result of the parmbsc0 force field shows less structural heterogeneity compared with that of the parm99 force field. In the latter, the third peak is sometimes markedly enhanced and changes its position to, implying that one interhelical angle should decrease considerably. At this moment, we are unable to determine the proficiency of the 2 force fields because the current experimental method cannot probe the level of structural fluctuation. To summarize, both models with two force field parameters yield ensemble averaged structures, which are generally consistent with the experimental results. The parmbsc0 force field reproduces the experimental result well at large r, while the parm99 force field captures the existence of the second peak at some configurations. The relative heights of the first and third peaks are reproduced only in some instances among the structural ensemble, but not in the sense of ensemble averages for both models. b. Characterization of global structures of 3WJ DNA using p(r) The distribution of interparticle distance, p(r), is a convenient quantity to consider for validating simulation results with experimental results. Simple geometries, such as spheres, disks, and rods, are easily identified with p(r). However, some difficulties exist in specifying the complex shape of branched DNA from p(r). In solution, three helices consisting of 3WJ DNA would not be perfectly ordered as they are in a crystal, due to thermal fluctuation. Moreover, complimentary base pairs in junction regions and helical ends may be partially broken. Here, we examine how p(r) reflects various characteristics of global structures, while excluding disorder in local structures. We separately built three helices using the tool NAB, with the same sequences used in the experiment. The geometry of each helix is based on the

9 generic crystal structure of a double helix. We only considered key features of the global structure, by changing the relative positions and angles of the 3 straight helices. The 3 strands are not completely connected at the junction region in our model structures. Six representative structures are chosen: planar Y-shape, pyramid, symmetric T-shape, asymmetric T-shape, planar Y-shape with open junction, and planar Y-shape with broken base pairing at the edge. For brevity, the 6 model structures will be denoted M1 6, respectively. We analyzed and compared p(r) of model structures in terms of the relative orientation of the 3 helices and the breakage of base pairs. The former case analyzed with M1 4 is presented in Figure S5a, while the latter case with M1, M5, and M6 is shown in Figure S5b. Figure S5c displays the structures of M1 6. p(r), calculated from the model structures, is characterized by three peaks and a monotonic tail at large r (Figs. S5a, b), whereas the second peak is not pronounced and the tail decrease is not monotonic in the experimental result. The position of the first peak is related to the width of the major groove, which is found to be 1.7 nm for M1 6. The size and shape of the whole molecule determines the third peak, around r = 5.5 nm, and the tail. In Figure S5a, the third peak is lower for the planar Y structure (M1) than in the pyramid (M2), while the tail of the former exhibits a more gradual decay than the latter. Meanwhile, the symmetric T-shape (M3) yields a broader peak at the same r and heavier tail at large r (>10 nm) compared with M1, because 2 helices are aligned to form a long and straight rod. In addition, the height of the third peak becomes relatively larger. The overall symmetry also affects p(r), and the third peak for the asymmetric T-shape (M4) is enhanced further. Finally, the second peak, located around r = 3.5 nm, is related to base pairing. Figure S5b corresponds to the cases for different base pairing. The planar Y-shape with an open junction (M5), where the bases near the branch are flipped out, yields an overall broadening of three peaks compared with M1. Similarly, in the case of M6 (i.e., when the bases flip out at the edge of

10 helical arms), the second peak becomes less distinct than that of M1 (Fig. S5b). Our observations provide a valuable reference point for understanding experimental and simulation results in terms of p(r). c. Characterization of global structures in 4WJ DNA using p(r) We generated 6 basic structures representing the global geometry of the 4WJ system. We believe that the number of such structures should be more; however, at this moment our effort was focused on understanding p(r) in terms of the relative orientation of four helices. The 6 model structures denoted by M1 6 (Fig. S6) are present in a stacked, planar, planar cross, Y, tetrahedral, and pyramid shape, respectively. We display p(r) corresponding to M1 6, obtained using the programs CRYSOL and GNOM, together with the experimental results in Figure S6. In the experimental p(r), a major peak is located at, while the shoulders are observed around and. Comparing the experimental p(r) to the computed p(r) reveals a substantial variance in the relative height of the main peaks. No p(r) for M1 6, by itself, agrees with the experimental results, and M1 6 do not reproduce the position of the main peak or the second shoulder. However, it is tempting to consider a combination of appropriately weighted p(r) for M1 6 as a reasonable reproduction of the aforementioned features in the experimental results. This will be a good prototypical system for further developing the ensemble refinement of SAXS in future studies. Although not quantitative, one can assume that the contribution of M4 should be substantial, as two helices are likely to get close to each other during conformational fluctuations.

11 Table S3. Geometry of 3WJ DNA from MD trajectories of 100 ns, employing the parmbsc0 and parm99 force field parameters parmbsc0 parm99 Number of base pairs ± ± Radius of gyration (nm) ± ± Width of major groove (nm) ± ± Width of minor groove (nm) ± ± Helix radius (nm) ± ± 0.041

12 Figure S3. MD results: average structures and p(r) obtained from 2 force field parameters. a b c d

13 Figure S4. Averaged WAXD patterns calculated from the equilibrium structures, employing the parmbsc0 and parm99 force fields at 300 K.

14 Figure S5. Characterization of global structures of 3WJ DNA using p(r).

15 Figure S6. Characterization of global structures of 4WJ DNA using p(r). a b M1 M2 M3 M4 M5 M6

16 Figure S7. Small-angle X-ray scattering (SAXS) profiles and pair distance distribution functions from experiments and MD simulations. (a) Scattering profiles of 3WJ and 4WJ in PBS buffer at 72ºC. The solid colored lines are the experimental data, and the solid black lines are the theoretical SAXS curves for each of the 50 structural models generated from MD simulations. (b) Pair distance distribution functions for 3WJ and 4WJ, based on an analysis of the experimental SAXS data using the program GNOM. The solid colored lines are the p(r) functions from the experimental data, and the solid black lines are the p(r) functions for each of the 50 structural models obtained from MD simulations, labeled as black solid lines in a.

17 Figure S8. MD model structures of single-stranded DNA used to simulate the structure of 3WJ and 4WJ in solution at 400 K. Structural models calculated from MD simulations for a single strand of DNA using simulation conditions in Table S4. Surface rendering in the structural models was achieved using Discovery Studio 1.6 (Accelrys, Inc.). (a) M1, (b) M10, (c) M20, (d) M30, (e) M40, (f) M50

18 Table S4. Structural parameters for MD model structures of single-stranded DNA at 400 K MD R a g,p(r) (nm) D b max (nm) MD R a g,p(r) (nm) D b max (nm) M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± M ± a R g,p(r) (radius of gyration) was calculated from the p(r) function by the program GNOM. b D max (maximum dimension of particle) was obtained from the p(r) function by the program GNOM.

19 Intensity (a.u.) Fig S9. UV melting curve (UV absorption as a function of temperature). Heating runs were performed between 20 and 90 C, heating at a rate of 1 C/min, while continuously monitoring the absorbance at 260 nm.the melting temperature values are 64 and 63 o C for 3WJ and 4WJ, respectively. These were calculated from melting curves of each DNA samples in PBS buffer solutions of ph WJ DNA 3WJ DNA Temperature ( o C)

20 CD (mdeg) CD (mdeg) Fig S10. Circular dichroism spectra of 3WJ and 4WJ DNA. Both BDMs have positive and negative peaks near 280 and 250 nm, respectively, which is typical of spectra of the B- form DNA. 3WJ T20 T50 T68 T80 T30 T60 T70 T85 T40 T65 T72 T Wavelength (nm) 4WJ 10 Temperature T20 T50 T68 T80 T30 T60 T70 T85 T40 T65 T72 T Wavelength (nm)

21 MD simulation results of 3WJ and 4WJ DNA at 400 K We increased the temperature of the system up to 400 K to observe DNA thermal denaturation. As a result, the 3WJ and 4WJ DNA were shown to be stable for hundreds of nanoseconds. The number of hydrogen bonds between base pairs in 3WJ exhibited a slight decrease during a time scale of longer than 500 ns, while those in 4WJ decreased very slowly (Fig. S11). We provide 2 notes regarding this observation: the scale of the temperature and the extent of sampling in the configuration space. Although 400 K is an unphysical temperature above the boiling point of water, the TIP3P water model does not boil, and the helix arms form a stable duplex. This is primarily due to the employed interatomic potential, which has been optimized with respect to the structural and dynamical properties of nucleic acids and water at low temperature in general. The force field parameter sets chosen here for nucleic acids and water are known to enhance base stacking and hydration, respectively. Accordingly, the temperature applied in our model system does not scale consistently with the physical temperature. We believe that 400 K should correspond to an enhanced noise level whose physical temperature is still below the melting point of our model system. The second note is that the time scale reached in the simulation is hundreds of nanoseconds, which might be too short to ensure ergodicity. If the branched DNA has a very rugged landscape of potential energy in the configuration space, the system would exhibit slow relaxation. The slight decrease in the number of hydrogen bonds might correspond to either slow relaxation to a collapsed molten globule or long-term fluctuations of partially disordered states at equilibrium. In addition, the system could approach different configurations (e.g., planar or tetrahedral structures) according to the initial configuration, particularly in the case of 4WJ. The initial condition dependence of the 4WJ system at 400 K is shown in Supplementary Figure S13. The upper panel exhibits the averaged structures at 300 K and 400K starting from a tetrahedral structure, while the lower panel is for the case of a planar

22 extended structure. Figure S11 MD results: p(r) calculated from equilibrium structures of at 400 K. a. 3WJ DNA b. 4WJ DNA

23 Figure S12. MD simulation results for the stability of 3WJ and 4WJ at 400 K.

24 Figure S13. Averaged equilibrium structures at 300 K and 400 K, starting from different initial configurations.

25 Table S5. Overview of the radius of gyration (R g ) and forward scattering intensity I(0) obtained from SAXS analysis of 3WJ and 4WJ in a wide concentration range Sample Conc. (mg/ml) R g I(0) (a.u.) ± E+02 ± WJ ± E+02 ± ± E+02 ± ± E+03 ± ± E+02 ± WJ ± E+02 ± ± E+03 ± ± E+03 ±

26 Table S6. Overview of the radius of gyration (R g ), forward scattering intensity I(0) and estimation of the molecular mass obtained from SAXS analysis of 3WJ and 4WJ in solution Sample MM calculated (kda) Conc. (mg/ml) I(0) Abs R g MM SAXS (kda) 3WJ ± WJ ±