Gel Electrophoresis of Gold-DNA Nanoconjugates

Gold-DNA conjugates were investigated in detail by a comprehensive gel electrophoresis study based on 1200 gels. A controlled number of single-stranded DNA of different length was attached specifically via thiol-Au bonds to phosphine-stabilized colloidal gold nanoparticles. Alternatively, the surface of the gold particles was saturated with single stranded DNA of different length either specifically via thiol-Au bonds or by nonspecific adsorption. From the experimentally determined electrophoretic mobilities, estimates for the effective diameters of the gold-DNA conjugates were derived by applying two different data treatment approaches. The first method is based on making a calibration curve for the relation between effective diameters and mobilities with gold nanoparticles of known diameter. The second method is based on Ferguson analysis which uses gold nanoparticles of known diameter as reference database. Our study shows that effective diameters derived from gel electrophoresis measurements are affected with a high error bar as the determined values strongly depend on the method of evaluation, though relative changes in size upon binding of molecules can be detected with high precision. Furthermore, in this study, the specific attachment of DNA via gold-thiol bonds to Au nanoparticles is compared to nonspecific adsorption of DNA. Also, the maximum number of DNA molecules that can be bound per particle was determined.

The concentration of the samples was determined by their absorption at the plasmon peak. The used molecular extinction coefficients are enlisted in Table SI-1: The extinction coefficients had been derived in the following way: The absorption spectra of the as-delivered particle solutions were recorded (1 cm pathlength) and the absorption at the plasmon peak A was extracted. Now the extinction coefficient ε was calculated as ε = A⋅c -1 ⋅cm -1 , whereby c [M] was the particle concentration as provided in the data sheet of the vendor of the particles. The same procedure was applied for different batches of particles and the resulting mean values were used as extinction coefficients.  Typical concentrations for the concentrated phosphine-stabilized Au particles were 10 -40 μM, 1-5 μM, and 0.1 -0.5 μM for 5nm, 10 nm, and 20 nm Au nanoparticles, respectively.
The citrate shell was exchanged for a phosphine shell because phosphine-stabilized particles turned out to be more stable in aqueous solution than citrate stabilized ones. In particular higher particle concentrations without agglomeration of the Au particles could be achieved with the phosphine coating. We speculate that this should be due to a higher binding-affinity of the phosphine compared to the citrate to the Au surface. We have to mention that in contrast to citric acid shells phosphine shells have a small fluorescence, which can be undesirable for fluorescence measurements (personal information Eric Dulkeith). The phosphine shell could not be completely replaced by competitive binding (e.g. by adding DNA molecules with -SH modification). Always some fluorescence of the phosphine adsorbed to the particles remained (personal information Eric Dulkeith).

I.2) Attachment of DNA to phosphine-stabilized Au nanoparticles
Single stranded DNA (with and without thiol modification) was added (as received from the supplier) to phosphine-stabilized nanoparticles. The reactions were performed in 50 mM NaCl, 25 mM phosphate buffer (pH = 7.3), and in 25 mM NaCl, 12 mM phosphate buffer (pH = 7.3) for 5nm, 10 nm, and for 20 nm Au particles, respectively. The absolute Au particle concentrations in the reactions were 0.4 -4 μM, 0.05 -1 μM, and 0.005 -0.01 μM for 5 nm, 10 nm, and 20 nm Au particles, respectively. The typical volume for one reaction mixture was 10-20 μl. Series with different amounts of added DNA were prepared. The maximum amount of added DNA molecules per Au nanoparticle was 15000, 30000, and 60000 for 5 nm, 10 nm, and 20 nm particles, respectively. For each series reaction mixtures with a different amount of added DNA molecules were used. The sequences of the used DNA molecules are enlisted in Table SI-2.
As an example a series of 10 nm Au particles with maximum 10700 DNA molecules added per nanoparticle was prepared in the following way. 30 vials were filled with 5 μl of 0.41 μM Au solution. To each vial 5 μl DNA mix was added. The DNA mix comprised DNA dissolved in 100 mM NaCl and 50 mM phosphate buffer of pH 7.3. To the first vial a DNA mix with a DNA concentration of 4800 μM was added, to the second one a DNA mix with a DNA concentration of 4800 μM / 2 = 2400 μM, to the third one a DNA mix with a DNA concentration of 2400 μM / 2 = 1200 μM, etc. In this way all vials had the same concentration of Au particles, the same buffer conditions, the same volume, but a different DNA/Au ratio. In this case the DNA/Au ratio in the first, second, third,....vial was ca. 12000, 5800, 2900, ...., respectively. This means a series has been generated in which the DNA/Au ratio is reduced by a factor of two from vial to vial.
DNAsequence sequence of bases from 5' to 3' end   3' end. In some case two different oligonucleotides with the same number of bases but with different sequences were used. This is indicated by "a" or "b" in the name of the sequence.
The reaction mixtures were incubated for one hour to 6 days at room temperature. The actually chosen conditions depended from the expected result. For the experiments in which a saturation of the Au nanoparticles with DNA was desired the reaction time was 6 chosen long and the DNA/Au ratio as high as possible. For the experiments in which discrete bands during gel electrophoresis were expected, the reaction time was only about 1 hour (because this yielded sharper bands due to reduced nonspecific adsorption) and only as much DNA was added as was needed to observe the bands.
We want to point out the following observations. The binding efficiency of thiolmodified DNA (as received from the supplier) to the Au nanoparticles is relatively low. This might be in part due to the protection of the -SH group of the DNA by the supplier. Sometimes around 10000 DNA molecules had to be added until the Au surface was saturated with DNA. However, much less than 10000 DNA molecules can stick to the surface of the Au particles. This means that a significant fraction of the added DNA remained unbound in solution. This might be explained by the fact that the Au nanoparticles were dissolved in phosphine-containing solution (3 mg /10 ml). This means that the thiolated DNA and the phosphine molecules had to compete to be adsorbed to the Au surface. Due to its importance we have to point it out again: the number of added DNA molecules per Au particle is not equal to the number of DNA molecules that are actually bound per Au particle. The reaction also seemed to be quite slow. Only after approximately one day the solutions had reached equilibrium and the number of actually attached DNA molecules per Au particle did not increase further. Again this could be explained with a competition of thiolated DNA and phosphine molecules for the adsorption to the Au particles.
For gel electrophoresis experiments glycerol was added to the samples to a final glycerol content of about 10%. Around 10 μl of glycerol-modified sample was loaded in each well of the gel. Gels were run 1 hour at 100 V in 0.5 x TBE buffer (corresponds to 44.5 mM TBE, pH around 8.3 -8.5). For very slow running samples the running time was doubled and the mobilities were afterwards divided by a factor of 2 to have normalized conditions. Together with the samples, on each gel, we have always run two reference samples: the plain phosphine-coated gold (without DNA added) and free single and double stranded DNA of different lengths.
Although the distance between the two electrodes in the gel-electrophoresis set-up is bigger than the length of the tray we consider that due to the law of a voltage divider the applied voltage drops only along the tray, since there the resistance is higher due to the thinner layer of electrolyte. The length of the tray is 10.1 cm and 100 V were applied. This corresponds to an electric field of 100V / 10.1 cm = 9.9 V/cm. This is in contrast to our own previous studies, where we wrongfully assume an electric field of 7.1 V/cm [1][2][3].
To visualize the Au particles on the gel after gel electrophoresis digital pictures were taken by using an Eagle Eye II digital camera (Biorad, USA). At the chosen gold concentration the gold bands appeared violet under illumination with visible light and could be seen by eye. We have adjusted the photos in such a way that the length of the gel (10.1 cm) corresponded always to 500 pixels.

I.4) Electrophoretic mobility of Au particles
We determined the migration of each band as the distance (in pixel) of the band after gel electrophoresis to the position where the particles had been loaded. Two examples are shown in Figure  In order to obtain the absolute mobilities m [cm 2 /(Vs)] of the particles the values obtained for the migration l [cm] had to be modified in the following. The velocity of the particles is defined as the absolute migration on the gel (in cm) divided by the time the gel has been run. This time was 60 minutes = 3600 s (or the results have been converted to this normalized running time). The mobility of the particles finally is defined as the velocity per applied electric field. As described above the applied electric field was 9.9 V / cm.   The error for measuring relative quantities instead of absolute one is always lower. In the case of gel-electrophoresis for example small deviations in the gel percentage, in the TBE-buffer concentration, in the exact positions of the electrodes in the gel electrophoresis chamber, and in the height of the TBE-buffer above the gel tray in the gel electrophoresis chamber can affect the values for the absolute electrophoretic mobilities. In order to reduce this error we measured relative instead of absolute mobilities. On every gel a reference sample with know absolute mobility was run. The mobilities of the other samples on the same gel were then related to this mobility. Since all samples of one series were run on the same gel deviations in the absolute mobilities due to experimental uncertainties are cancelled out by considering only the relative mobilities. In order to measure the mobility of Au nanoparticles of different sizes gels with nanoparticles of different diameter were run as shown in Figure SI-1. Typically only the relative mobilities were determined. They are identical to the relative migration: m / m reference = l / l reference Formula SI-2 For example, the relative mobility of the 30 nm Au particles shown in Figure SI-1 relative to the 10 nm Au particles on a 2% agarose gel is m 30nm,2% / m 10nm,2% = l 30nm,2% / l 10nm,2% = 2.22 cm / 3.93 cm = 0.56 . By using the table of the absolute mobilities (Table  SI-3) the relative mobilities can be converted to absolute ones: m 30nm,2% = (m 30nm,2% /  m 10nm,2% ) × m 10nm,2% = 0.56 × 1.07 × 10 -3 cm 2 V -1 s -1 = 0.60 × 10 -3 cm 2 V -1 s -1 . We want to stress this idea again. First, the absolute mobilities for reference particles were recorded (see Table SI-3). This was done for many gels (30-300) for each gel percentage. Therefore the obtained values are highly significant (as mean values from 30 -300 experiments). In all the following experiments on each gel also reference particles were run. The mobilities were then always referred to the mobilities of the reference particles. Therefore much fewer gels (3-10) had been run, since only relative data were extracted from each gel.

I.6) Electrophoretic mobility of Au-DNA conjugates: Extracted Parameters
Au-DNA conjugates with different DNA/Au ratios were prepared as shown in I.2 and run on gels of different percentage as shown in I. 3. An example is shown in Figure SI-5. When DNA is attached to Au particles the Au-DNA conjugate becomes bigger than the plain Au particle. This has been discussed in detail in our previous manuscripts [1,3]. In this way we explain the retardation on the gel of the Au-DNA conjugates on the gel compared to the plain Au-particles as sorting by size. We believe that the addition of DNA to the particles does not significantly change the surface charge density of the particles. This might be explained by the fact that the oligonucleotides are closely packed on the Au surface and that at the used salt concentrations all the charges in the inside of the DNA shell are screened with counter ions. When the particle surface is saturated with DNA, then also a saturation of the retardation of the mobility can be seen ( Figure SI -5).
From each gel the mobility m of the saturated particles is extracted. This is described in I. 7. When only few DNA molecules are added per Au particle discrete bands in the gel can be seen. These bands are referred to Au particles with exactly one, two, three, etc... DNA molecules bound per particle manuscripts [1,3]. From each gel also the mobility m of these bands is extracted, as will be described in I. 8. For each different DNA sequence Au-DNA conjugates with variable DNA/Au ratio have been made and run on a gel. From each gel the following data can be extracted: the saturation mobility and in the case of long-enough DNA the mobility of the conjugates with exactly one, two, etc... DNA strands attached per each particle.
For all the data shown in I.7 and I.8 the mobility of the conjugates m was always referred to the mobility of the plain, unconjugated Au nanoparticles m x,y (with the same Au x size and gel percentage y). For each DNA sequence, particle size, and gel percentage at least 3 different gels were run. The data shown correspond to the mean values and their standard deviations.   0.14 ± 0. 03  --8  ---9 ---

I.9) Some comments about the mobilities of Au-DNA conjugates
As already reported in previously published work [3] the mobility of the nanoparticles decreases the more and the longer DNA has been attached specifically via thiol-gold bonds. The results for Au particles of different diameter and different agarose concentrations are qualitatively the same. Consequently the degree of retardation for attachment of the same number of DNA molecules of the same length to nanoparticles of different diameter is highest for the smallest nanoparticles, since here the relative increase in the effective diameter is the highest. This is of particular importance for the extraction of discrete bands with an exactly known number of DNA molecules attached per nanoparticle [1]. In the case of 10 nm diameter Au particles the binding of one single single-stranded DNA molecule of 43 bases yields a sufficient increase in the effective diameter so that this conjugates migrates on a band that is retarded enough on a 2% agarose gel that it can be clearly resolved from the band of plain Au nanoparticles. The increase in size upon attachment of one single DNA molecule with fewer bases is not big enough to yield a band that is well resolved from the band of plain Au particles. For 20 nm particles longer DNA molecules have to be attached in order to resolve discrete bands, whereas for 5 nm diameter particles discrete bands should be observable also for DNA with fewer bases. This number is important for the calculation of the minimum distance reachable for DNA-mediated assemblies of Au nanoparticles. Upon hybridization between Au-DNA conjugates of complementary DNA sequences the DNA between the Au particles is double stranded and thus relatively stiff (persistence length 50 nm [4]). By using the contour length of double stranded DNA (0.34 nm per base pair [4,5]; under neglecting the thiol-anchor and spacer molecules between the DNA and the thiol group) 43 bases correspond to around 15 nm (in case more sophisticated structures in which part of the linker DNA would be single stranded and thus could be bent). Thus, the minimum distance between the surfaces of two DNA-mediated Au particles in an Au dimer is 15 nm. This is due to the fact that for the isolation of Au nanoparticles with exactly one strand of DNA the length of the DNA has to be at least 43 bases in order to be resolved on an agarose gel. In previous work a Au to Au distance of 29 nm in dimers of 10 nm Au nanoparticles could be directly observed with transmission electron microscopy (TEM) using 50 base pair DNA [2].

II.1) Effective diameter of Au-DNA conjugates
In I) the mobilities of Au-DNA conjugates are investigated in dependence of the diameter of the Au particles, of the length of the DNA, of the number of DNA molecules attached per particle, and of the gel percentage. However, the mobility is not an illustrative quantity. Therefore we intended to convert the mobilities in effective diameters of the Au-DNA conjugates. The more DNA is attached per Au particle, the bigger conjugate will be and the slower its mobility is. For the conversion we have applied two different methods, which are described in II.2 and II.3.

II.2) Calibration curve for the size dependent of the electrophoretic mobility
The idea of this approach is to make a calibration curve that relates mobility to size. For this purpose phosphine-coated Au particles of different size have been run on gels with different percentage and their mobilities have been obtained, see Tables SI-3  In order to inter-and extrapolate these data we fitted empirically with a monoexponential function with 2 fit-parameters A y and T y to the size-dependent mobilities: m/m 10nm,y (d eff (Au)) = A y * exp(-(d eff (Au) -6 nm) / T y ) (Formula SI-4) We have used an exponentially decaying fit function as it has turned out that our data can be interpolated well with a function of this type. Since the effective diameter of the smallest nanoparticles we used (5 nm Au core + 2 • 0.5 nm phosphine shell = 6 nm) was 6 nm we used a shifted exponential function with 6 nm shift in the exponent. For the fit we have taken into account only the data for Au particles from Table SI-4    The (m x,y / m 10nm,y ) values can be taken from Table SI-4. For example: m 5nm,1% / m 10nm,1% = 1.07 From Figure SI-9 it is obvious that for gels with different percentage only effective diameters within certain rages can be extracted, because data m/m x,y << 0.1 are not very reliable. For low percentage gels the size range is biggest. For high percentage gels only small particles can be investigated, but the resolution is better. With gels of y% agarose the following particle size can be analyzed: 0.5%: 0-120 nm, 1%: 0-100 nm, 2%: 0-70 nm, 3%: 0-50 nm, 4%: 0-30 nm, 5%: 0-20 nm, 6%: 0-15 nm.
Since the fit function Formula SI-5 does not exactly go through the point m 10nm,y (11 nm) = 1 for Au particles with a core diameter of 10 nm (the fit function had been determined 38 in a way that it hits this point as close as possible), the mobility m = 1 is not converted into an effective diameter of exactly d eff = 11 nm by the inverse fit function. In other words: m 10nm,y (d eff =11 nm) = 1 has been used as point to obtain the fit function d eff = d eff (m) Formula SI-5. This function has been chosen to include this data point as close as possible. However, since this data point will not lie exactly on the function, the inverse function does not return exactly the effective diameter that has been the "input" for the fit: d eff (m 10nm,y ) will not be exactly 11 nm, but only close to it.
We have to point out that Formula SI-5 can in a strict way be only applied for objects that are in their nature as similar as possible to phosphine-coated Au particles and which are bigger than 5 nm. This is due to the fact that the fit functions have been obtained with data recorded on Au particles. Rigid objects certainly show different electrophoretic properties than soft ones. In this way Formula SI-5 cannot be used for example to convert the electrophoretic mobilities obtained with DNA molecules (see Chapter I.5) into effective diameters for two reasons: First the calibration function has been obtained for objects bigger than 5 nm and therefore the DNA particles are out of the range of extrapolation. Second the calibration curve has been obtained for rigid objects and can't be directly applied for soft objects.
The difference in electrophoretic behavior between soft and rigid objects can directly be seen in Figure SI-10. Whereas the logarithm of the mobilities obtained for different gel percentages scales linearly for soft objects such as DNA, there is different behavior for rigid objects as Au particles and objects with rigid core and soft shell as DNA-Au conjugates.
10 28.9 --7 30. 2  --8  ---9 ---  Table SI-2) with thiol modification. The DNA/Au ratio was chosen in a way that single bands could be observed with gel electrophoresis. The Au-DNA conjugates were run on agarose gels with different percentages. The mobility of the discrete Au-DNA conjugates (e.g. the mobility of the individual bands that correspond to Au particles with exactly 1, 2, .... DNA molecules bound per particle) was measured relative to the mobility of the free Au nanoparticles (Table SI- 9.4). From these data the effective diameters have been obtained using Formulas SI-5 and SI-6. The data are graphically displayed in Figure SI-

II.4) Ferguson analysis
We also intended to obtain absolute numbers for the effective diameters of the conjugates by applying Ferguson analysis [6]. For the Ferguson analysis, absolute mobilities m y obtained for the same sample run on gels of different percentage y are required. In Tables  SI-4 to SI-9 the relative mobilities of many different Au-DNA conjugates are listed. By using the absolute mobilities of plain Au nanoparticles (Table SI-3) the relative mobilities were converted to absolute ones.
In a Ferguson plot the decadic logarithm of the absolute mobility m y (at a certain gel percentage) is plotted versus the gel percentage y. The plot results in linear curves for DNA. However, for Au nanoparticles deviations from the linear behavior can be seen [2,6], see Figure  The absolute mobility m 0 corresponds to a gel at percentage y = 0%. K r is the so called retardation coefficient which is related to the characteristics of the gel and to the effective diameters d eff of the sample by the following equation: Here a is a constant, and l and r define the length and the radius of the gel fiber. As abbreviations α = a(πl) 1/2 and β = a (πl) 1/2 r are used.
Ferguson plots have been made for plain Au particles of different size (see Figure SI-14, relative mobility data see Table SI -4), and the retardation coefficient K r has been obtained for each particle size, see Table SI-14. As particle size we use the effective size d eff of the Au particles that includes their phosphine shell (see Formula SI-3). In this way a calibration curve in which K r 1/2 is plotted versus d eff is obtained. According to Formula SI-8 this is a linear curve. This calibration curve correlates mobilities to effective diameters. For any Au-DNA conjugate that has been run on several gels with different percentage a Ferguson plot m y (y) can be obtained from which the retardation coefficient K r can be derived. From the calibration curve K r 1/2 (d eff ) finally the effective diameter of the Au-DNA conjugate d eff can be obtained.
An example and the resulting effective diameters for the Au-DNA conjugates in our study are shown in the following. In Figure SI-14 the logarithm to the basis 10 log(m y ) of the mobility of gold particles with diameters d from 5 and 60 nm is plotted versus the gel percentages in the range between 1, 2, 3 and 4. The mobility values are taken from Table SI-4. For each particle size a linear fit was performed. According to Formula SI-7 the slope of the fit is called retardation coefficient K r . The retardation coefficients that have been obtained are reported in Table SI-  According to Formula SI-8 the square root of the retardation coefficient K r 1/2 scales linearly with the effective diameter d eff , see Figure SI-15. By fitting the K r 1/2 (d eff ) data with a linear function (K r 1/2 (d eff ) = α d eff /2 + β, Formula SI-8) a calibration curve is obtained, that relates the measured mobilities (in terms of K r ) to an effective diameter d eff . For this the Formula SI-8 is transformed to yield d eff : d eff (K r ) = (K r 1/2 -β) / α Formula SI-9 By using the fit parameters α and β the effective diameter d eff can be derived with Formula SI-9 from the experimental data (the retardation coefficients K r that have been obtained from the mobilities for different gel percentages). To decide in which the best linear fit can be performed, two ranges were examined. α and β were obtained by fitting the K r 1/2 (d eff ) data in the range of 5 nm ≤ d ≤ 30 nm, and 5 nm ≤ d ≤ 40 nm, respectively, see Figure SI-15. As results of the fit the parameters α = 0.0286, β = 0.237, and α = 0.0238, β = 0.270 were obtained, respectively. In order to evaluate the quality of the fit we used Formula SI-9 with these fit-parameters to obtain the effective diameters d eff for Au nanoparticles from their measured K r data. The K r values are taken from the third column of Table SI-14 and the so obtained results for both sets of fit parameters are displayed in the forth and fifth column. By comparing the diameters obtained from the fit (d eff,cal (Au)) (forth and fifth column of Table SI-14) with the "real" values d eff (second column of Table SI-14) we can judge the quality of the fits. It is obvious from the data that in the case of the first set of fit parameters (α = 0.0286, β = 0.237, derived in the fit range from 5 nm ≤ d ≤ 30 nm) the fit cannot be extrapolated to particles of sizes bigger than 30 nm. The second set of fit parameters α = 0.0238, β = 0.270, derived in the fit range from 5 nm ≤ d ≤ 40 nm) yields slightly better results for big particles, but worse 57 results for small particles. For this reason we decided to use in the following only the first set of fit parameters α = 0.0286, β = 0.237. However, in this case on has to be aware that any effective diameters that are obtained from K r data by using Formula SI-9 are only reliable if they are not bigger than approximately 30 nm. The size of bigger particles is underestimated.

Figure SI-15:
The square root of the K r data that has been obtained from the mobilities of plain phosphine coated gold particles at different gel percentages is plotted versus the effective diameter of the gold particles (the diameter that is known from TEM images for the Au core + twice the estimated thickness of the phosphine shell). The data are taken from With the same procedure the effective diameters for a series of Au-DNA conjugates have been derived. We show the example of 10 nm Au particles that have been conjugated with thiolated DNA of 43 bases length (sequence SH-43b, see Table SI-2). The mobilities of Au particles with 0, 1, 2, 3, 4, 5, and 6 attached DNA molecules per particle and the one of Au particles saturated with DNA were measured for agarose gels with 1%, 2%, and 3% agarose content. For the individual DNA molecules attached per particle the mobility data m / m 10nm,y are displayed in Table SI From Figure SI-17 it is obvious, that by using different methods of evaluation significantly different values for the derived effective diameters are obtained. As already mentioned before, the Ferguson analysis reported here only yields reliable data for particles smaller than around 30 nm, because the range for the extrapolation was chosen this way. The size of all bigger particles is severely underestimated. Also the mobilitydiameter master curves for 1, 2, and 3% gels yield significantly different effective diameters for the Au-DNA conjugates for big diameters. The effective diameters are smaller for gels of higher percentage. This direct comparison shows the limits of size determination with gel electrophoresis. It always has to be taken into account, that the data obtained with a certain gel percentage are only reliable for particles within a certain size range. In the following the Ferguson analysis is applied as described above for different Au-DNA conjugates and the retardation coefficients and yielded effective diameters are reported in Tables SI-16

III) Determination of the Maximum Number of Bound DNA Molecules per Nanoparticle
We have quantified the maximum number of DNA strands that can be attached per gold nanoparticle for particles with 5 nm and 10 nm diameter and single stranded DNA with 8 (sequence Cy5-8a-SH, see Table SI-2) and 43 bases (sequence Cy5-43a-SH, see Table  SI-2). For this purpose single stranded DNA that was modified with a thiol group on one and a Cy5 dye on the other end has been attached via formation of thiol-Au bonds to the surface of Au particles. The DNA was added in different DNA to Au ratios and the conjugates were run on 2% agarose gels. The more DNA bound per Au nanoparticle the more the band of this conjugate was retarded on the gel {Parak, 2003 #8595}. At a certain amount of added DNA the retardation of the band of the conjugates did not increase further, which indicates that the Au surface is fully saturated with DNA [3], see Figure SI-20.   For each band the relative mobility m / m 10nm,2% was determined as described in II). In total for each sample four gels were run in order to get a sufficient statistics. The bands of the Au-DNA conjugates were then extracted from the gel by cutting out the agarose piece that contained the band and immersing it in 0.5x TBE buffer solution.
After two days the Au-DNA conjugates had diffused out of the gel in the buffer. The cutting and extraction procedure ensured that the Au-DNA conjugates were separated from any unbound DNA, that migrates much faster on the gel. Then UV/vis absorption spectra were recorded of all the extracted Au-DNA conjugates, see Figure  SI-21. For each of the conjugates the DNA concentration was determined by the Cy5 absorption and the Au concentration was determined by the absorption at the plasmon peak and from both concentrations the number of attached DNA molecules per particles was derived. The absorption spectra were then treated in the following way. First the background absorption at 1100 nm was subtracted. The spectra were then normalized, so that the absorption at the plasmon peak is equal to one. Then the absorption spectrum of plain Au particles was subtracted from all the other absorption spectra. This was done graphically to adjust for the shape of the curves. The residual spectrum corresponds to the absorption of the Cy5 dye that is attached to the DNA. We assume that each DNA molecule is labeled with exactly one Cy5 molecule. In this way we determined the Cy5 absorption for each Au-DNA conjugate. From the absorption values the concentrations were determined. We assumed the extinction coefficient of Cy5 to be ε(Cy5 at 650 nm) = 250 000 M -1 cm -1 . The extinction coefficients of the Au particles at their plasmon peaks are shown in Table SI-1. In this way the concentration of DNA was determined by the Cy5 absorption at 650 nm (after removing the Au signal) and the Au concentration was determined by the Au absorption at the plasmon peak. Cy5 was chosen as dye because of its large shift in absorption in relation to the plasmon peak of the Au (at the plasmon peak at around 520 nm there is almost no Cy5 absorption, the Cy5 absorbs around 650 nm). By dividing the DNA concentration by the Au concentration the actual DNA/Au ratio was finally derived. Now the mobility of each band can be correlated with the DNA/Au ratio of the conjugates that formed the band. It has to be noted, that here the DNA/Au ratio describes the number of DNA molecules per particle that are actually bound per particle. Unbound DNA has been removed on the gel. The results are shown in Tables  SI-18 Figure SI-20). The bands from the conjugates were extracted from the gel and UV/vis absorption spectra were recorded (examples are shown in Figure SI Figure SI Figure SI-20). The bands from the conjugates were extracted from the gel and UVVIS spectra were recorded (examples are shown in Figure SI Figure SI-20). The bands from the conjugates were extracted from the gel and UVVIS spectra were recorded (examples are shown in Figure SI-21 When the mobility of the bands does not further decrease upon the addition of more DNA we assume that the surface of the particles is saturated with DNA. From Tables SI-18.1-18.4 and Figure SI-22 the maximum number of DNA molecules that can be bound per Au particle can be estimated: ca. 53 Cy5-8a-SH per 10 nm Au particle, ca. 43 Cy5-43a-SH per 10 nm Au particle, and ca. 13 Cy5-8a-SH per 5 nm Au particle. Particles with a nominal diameter d of 10 nm and 5 nm have a surface area of 314 nm 2 and 79 nm 2 , respectively. In this way we can determine the surface density of the bound DNA molecules, i.e. the number of molecules bound per nm 2 , see