Surface densities of DNA carpets with confocal microscope 73

As already discussed in the Ph.D. thesis of Lehner [8], it is hard to resolve the surface end-grafting density from freely fluctuating λ - DNA carpet with fluo-rescence microscopy when the density is higher than 0.14 molecules/µm². This end-grafting density was determined by counting the molecules using a commer-cial software⁷. This method is rather inaccurate since it needs high amount of interactivity from the user such as the setting the binarization level. Further-more it is unclear whether all the counted molecules are really end-grafted and additionally the distinction between separate molecules is difficult. We tested the method of hydrophilic combing, presented above, in order to have an additional method to estimate the surface density directly using a microscope. First of all we made a low density sample and combed it in a flow as presented in Fig. 5.3 (a). After that we measured average length of roughly 50 molecules to be 12.1 µm±4.8µm. This is in a good agreement with a theory developed by Stigter et al [60]. Stigter et al. predicts extension of 17−18µm for B - DNA, with contour length⁸ of 21.8µm, in a hydrodynamic flow when liquid velocities are higher than 50µm/s. The longest molecules we measured were also 17−18µm long but the average length of molecules is pressed down by short fragments which are present due to shearing during pipetting. The flow velocities in our experiment were not controlled but were estimated by looking at the experimental films where the response of the molecules to the spermidine added buffer was over within 500 ms (two frames). The field of view in the microscope is roughly 70×90 µm so we estimated the flow velocities⁹ to be at least or higher than 50 µm/s.

Since the lateral resolution of the microscope is roughly 500 nm [8], it is not expected that we resolve single molecules which are close to each other. Therefore

6Model: PDC - 32G, Harrick Plasma.

7PerkinElmer, UltraView version 5.2.

822µm is sometimes given as the contour length of YOYO-1 loadedλ- DNA molecule [61].

9This is a very conservative estimate as 90µm/0.5 s = 180µm/s>50µm/s.

we decided to use the intensity profiles, which are measured perpendicular to the combing direction, to estimate the grafting density. This was done so that we first measured 20 intensity profiles for single molecules perpendicular to the combing direction in the high density sample presented in Fig. 5.3 (b). From these profiles were then background removed (see example of single intensity profile in Fig. 5.3 (c)) and the remaining intensity was integrated. As a result we got that the integrated intensity of a single molecule perpendicular to the combing direction is (146.5±38.5) Gray levels × µm. After that we measured 10 intensity profiles for longer distances perpendicular to the combing direction. Such a profile is presented in Fig. 5.3 (d) which is the intensity under the red line seen in Fig.

5.3 (b). From the longer profiles the background was also removed by fitting a second order polynomial to the local minima. With the integrated intensity of a single-molecule it was possible to determine the number of molecules under the longer intensity profile since we are comparing intensities within a single frame.

Result of this analysis was then divided with the area limited by the length of the corresponding long intensity profile and the average molecule length. This analysis gave us end-grafting density of∼0.3 molecules/µm² whereas the method used by Lehner for this same sample gives 0.10 molecules/µm². Where does the difference come from? Our analysis apparently overestimates the end-grafting density since the intensity profiles for the single-molecules were determined from the high density sample and there it was impossible to distinct exactly single molecules. Therefore in this analysis we used molecules with the lowest intensi-ties that means molecules where it was more or less sure that they were single molecules. However, it is naturally possible that not all the molecules are fully extended in the flow and therefore they appear in the image more intense. This is also why we give the error for the density only to lower surface end-grafting den-sities, giving a result that the carpet in Fig. 5.3 (d) was 0.2−0.3 molecules/µm² dense. The error is estimated from the single molecule intensity standard devia-tion and from the single molecule length standard deviadevia-tion. On the other hand the method used by Lehner apparently underestimates the density as at high density samples it is rather difficult to distinguish independent molecules.

5.3.2 Surface densities of DNA carpets with microarray reader

First of all we studied the kinetics of biotin- and Cy3-modified λ - DNA car-pet formation on a surface coated with streptavidin (for surface preparation see chapter 2.2.2). This was done by varying the incubation time of the DNA be-tween 0 and 140 minutes at room temperature. The incubation stock solution concentration was kept constant at 2 ng/µl and as a buffer we used TBE with 0.5 M NaCl. At every slide on which the DNA carpets were prepared we had also

(a) (b)

(c) (d)

Figure 5.3: a) The average length of molecules was determined from a low end-grafting density sample combed with spermidine in a flow. b) The combing was done also for a high surface end-grafting density sample. c)The intensity profile perpendicular to the combing direction for a single molecule in a high density sample. The background was interpolated from the surrounding background.

After background reduction the area under the curve was determined by a trape-zoidal integrationd) Furthermore we measured a intensity profile perpendicular to the combing direction for several molecules (red line). Background was deter-mined by fitting a second order polynomial to the local minima in the intensity profile. With the integrated intensity of a single-molecule it was then possible to determine the number of molecules under the longer intensity profile. Result of this analysis was divided with the area limited by the length of the correspond-ing long intensity profile and the average molecule length. This analysis gives end-grafting density of0.2−0.3 molecules/µm².

one reference point of fluorescent streptavidin¹⁰, having almost the same spec-tral characteristics as Cy3. The reference points were used in order to be able to compare measurements with each other since we noticed that the measured intensities varied considerably from measurement to measurement. The reason for the intensity variation arises apparently from the fact that the optics of the microarray reader are not adjustable and the sample from measurement to mea-surement is sitting differently in the focus. For the reference point we incubated fluorescent streptavidin at a concentration of 50 µg/ml for 15 minutes and after that we washed away the unbound streptavidin. After the incubation of DNA and fluorescent streptavidin, the thoroughly washed and dried samples were then measured with microarray reader. The results are presented in Fig. 5.4 where the data is gathered from four different slides all having 4−6 spots. The measured DNA carpet intensities are divided with the corresponding fluorescent strepta-vidin intensity so that different slides can be compared. Each data point is an average of two independent measurements and the error bars are the averaged standard deviation of the measured intensity. Additionally we have subtracted the background so that we were able to fit to the data the first order reaction kinetics (σ(t) =σ∞(1−e^−t/τ)), whereσ(t) is the surface end-graft concentration.

The use of first order reaction kinetics means that we are relating the end-graft concentration to the highest possible end-graft concentration σ_∞¹¹. The use of this classical model is not completely correct as pointed out by Kopelman [62] as our reactants are spatially constrained by the wall but as we are interested only to get an idea of the reaction kinetics we accept this model. The first order re-action kinetics give us a characteristics grafting time of the rere-action (t1/2 =τ ln2

∼26 minutes). The long characteristics grafting time of 26 minutes is somewhat surprising since it takes only∼0.15 seconds forλ- DNA to diffuse over its radius of gyration Rg and so to provide new DNA molecules from the bulk for the sur-face to bind. Here we have assumed R_g of YOYO-1 loadedλ - DNA to be ∼0.8 µm [8] and diffusion constantD_{λ−DN A} to be ∼2·10⁻⁸ cm²/s [63]. However, the slow occupation of the surface can be readily understood as an interplay of at least three effects. Firstly there is free oligos in the incubation solution. This is because these experiment were performed before we learned about the use of PEG precipitation for filtering (see chapter 3). The free oligos diffuse faster¹² than λ - DNA and so occupy the free binding sites at the surface. Secondly as a DNA molecule attaches to the surface and takes its ”parking place” it hinders through its own size other molecules to bind which then again reduces the bind-ing rate. Thirdly there is an electrostatic repulsion between negatively charged DNA and negatively charged streptavidin which the molecules have to overcome in order to establish an end-graft. This is because we are using THE with pH 9

10Alexa Fluorr555 having excitation at 555 nm and emission at 565 nm.

11dσ(t)/dt= 1/τ(σ∞−σ(t)).

12Rg(λ−DN A)∼800 nm andRg(oligo)∼4 nm thereforeDoligo∼200×Dλ−DN A.

to incubate DNA and in such a conditions streptavidin is negatively charged¹³. As these measurements were done before we were aware of the free oligo problem one could argue that we see only the kinetics of the free Cy3-modified oligos binding to the surface. This is, however, not the case as this was tested by incubating biotin- and Cy3-modified λ - DNA solution with free oligos on a surface coated only with BSA. Here as a result we got that the free Cy3-modified oligos do not bind to the surface considerably (data not shown).

Furthermore we measured the binding kinetics of λ - DNA on gold surfaces (data not shown). As a result we found that the characteristic grafting time of the reaction is∼9 minutes. The binding of the thiol-modified DNA to the gold surface is then almost three times faster as the binding of the biotin-modified DNA to the streptavidin surface. This can be understood by the fact that thiol group can react with gold atoms at the surface whereas the biotin molecule has to find the complementary binding site in the streptavidin molecule. In addition to this, as already mentioned in chapter 2.2.2, the gold surfaces are inactivated very rapidly outside the evaporation chamber. This experiment was also performed before we were aware of the free oligo problem.

To give an estimate of the surface end-graft densities as a function of DNA stock solution concentration, we prepared a surface covered with sulfo-SIAB and incubated it with various concentrations of Cy3- and thiol-modifiedλ- DNA. The measurement consisted out of nine spots: for the background we had one spot without any fluorescent dye, as a reference we used three spots of DNA incubated for 30 minutes at room temperature with 2 ng/µl of DNA and as measurement spots we incubated 2 ng/µl, 3 ng/µl, 4 ng/µl, 5 ng/µl and 6 ng/µl for one hour at 37^◦C. The reason for three reference points was that we wanted to rule out the spot to spot variations as much as possible. The average surface grafting density for the three reference spots was determined to be∼0.07 molecules/µm² and this was done with microscopy by using the method of Lehner (discussed above). The increase in DNA concentration and the increase in temperature did not seem to have any influence on the surface density ofλ - DNA (see Fig. 5.5).

This can be understood, as already discussed, as an effect of free oligos blocking the surface. The end-grafting densities compared with the reference point gives a result that the carpets were ∼ 0.08 molecules/µm² dense. As this was only a single measurement, it is difficult to give an error estimate for the determined density. Therefore it would be important to redo these experiments, with a oligo free sample, to get better statistics and to see how reproducible the microarray actually is.

In conclusion the use of a microarray reader for the characterization of λ -DNA carpets was problematic. This is mainly due to the fact that the sample preparation is done by hand and this results in poor statistics and certain varia-tion from spot to spot. Furthermore, in the microarray reader that we used, the

13Streptavidin has its isoelectric point at pH ∼5 [64].

Figure 5.4: With biotin from one end and with Cy3 from the other end modi-fied λ - DNA binding kinetics on streptavidin surface. The DNA carpets were prepared by incubating with 2 ng/µl of λ - DNA stock solution on streptavidin coated surfaces. The incubation time of the DNA was varied between 0 and 140 minutes and after that the thoroughly washed samples were dried and mea-sured with microarray reader. The intensities on a single coverslip were divided by a corresponding reference fluorescent streptavidin spot value so that different measurements were comparable. Each data point is an average of two indepen-dent measurements and the error bars are the averaged standard deviation of the measured intensity. Furthermore the background was subtracted so that we were able to apply the first order reaction kinetics to the data. The analysis gives us a characteristics grafting time of ∼ 26 minutes showing the slow formation of end-grafted DNA carpet. This measurement was done before we learned about the use of PEG precipitation for filtering meaning that the DNA stock solution contained free oligos (see chapter 3).

Figure 5.5: Thiol- and Cy3-modified λ - DNA incubated as a function of bulk concentration for one hour at 37^◦C on sulfo-SIAB surface and measured with a microarray reader. The red spot is the average intensity of three reference spot which were incubated for 30 minutes at room temperature with 2ng/µl. The end-grafting densities were determined for the reference spots by microscopy to be ∼0.07 molecules/µm² and this gives for the other spots grafting densities of

∼0.08molecules/µm². The error bars are the standard deviation of the measured intensities in the spots scaled to the end-grafting density.

optics was not adjustable so small variations in the sample height caused notice-able variations in the reader signals. Nevertheless the results show that surface end-graft densities and kinetic behavior can be measured semi-quantitatively with the microarray reader.

5.3.3 Entropy driven grafting density enhancement

First we studied the unspecific binding of DNA to the streptavidin coated sub-strate. This was done by using λ - DNA which did not have biotin as end-modification and so the specific end-graft to the streptavidin surface was not possible. Addition of PEG (PEG 8000 with average molecular weight of 8000 g/mol) made the solutions more viscous and that was the reason why we always incubated overnight so that DNA had enough time to diffuse to the surface. The incubations were done so that the samples were left to stay few hours at room temperature and then stored overnight in the fridge. The evaluation of the sam-ples was done by using confocal fluorescence microscopy. Here we found that PEG induced very strong unspecific binding of DNA to the substrate and this was the case even if the surfaces were blocked with BSA. It was clearly seen that with the increasing PEG concentration the amount of unspecifically bound DNA increased. Furthermore, when the same experiments were done with biotin-modified DNA on streptavidin coated surfaces, the formation of ”DNA islands”

was found as presented in Fig. 5.6. The island formation was observed to start at PEG concentrations of ∼6 M. This is roughly the same concentration where the coil-globule transition also occurs. On BSA-blocked surfaces the islands were ob-served to be smaller than on unblocked surfaces. This fact would hint that at least part of the interaction of DNA with the surface is unspecific. The samples were always good washed before microscoping them and since they survived the wash-ing it seems that DNA is not only entangled but is really attached to the surface.

Furthermore we noticed that we were able to destroy the islands by increasing the illuminating laser power. This is apparently partly due to YOYO-1 induced photocleavage of dsDNA [38]. The use of free oxygen scavengers (antibleaching) reduces the amount of photocleavage but when the surface concentrations are so high like in our case in ”DNA islands” not all free oxygen can be scavenged.

The use of PEG often resulted in also strong background fluorescence signals.

This is understood by the fact that at relatively high dye concentrations (1 : 5 dye:base) DNA – YOYO-1 complex is not formed only through bis-intercalation but YOYO-1 binds also on the surface of the DNA [21]. Now, as PEG pushes the DNA molecules together and against the surface, it provides a lot of free YOYO-1 for the surface to uptake. As non-target materials (here: surface attached oligos, streptavidin) can also bind YOYO-1 and create increased fluorescence [22], the high surface signals are not surprising. The high surface fluorescence makes it difficult to analyze the samples quantitatively with microscopy. Additionally the YOYO-1 which is bound to the surface is away from DNA reducing the signal

Figure 5.6: Addition of PEG to the incubation solution induces ”DNA island”

formation. Here we had unblocked streptavidin surface, biotin-modified DNA and∼6.5M PEG in the incubation solution which was later washed away. As a result we got concentrates of DNA which survived the thorough washing of the samples but not high intensities in the laser illumination. The snapshots size is 42µm× 55µm.

and bringing an additional difficulty to the analysis.

In conclusion in our experiments with PEG on streptavidin-coated surfaces blocked with BSA we have not seen any positive effect that would be outside the experimental variation estimated by microscopy (the island formation was mostly observed with unblocked streptavidin surfaces). As already discussed several times the free oligos have definitively had an effect on the surface end-grafting density and therefore to redo these experiments with a ”clean” DNA stock solution would be very interesting.