Steric and kinetic effects on the microarray surface

The hybridization of a target to its specific probe is influenced by several steric and kinetic effects. The most important parameters are (I) the distance to the array surface, (II) the

123 probe density, (III) the direction of probe immobilization, (IV) the secondary structure of the target, (V) the sequence composition of the target, (VI) the target labelling, and (VII) the Langmuir adsorption coefficient.

The distance of the probe from the array support strongly influences hybridization. Negative steric effects of the microarray surface on the hybridization signal were previously shown (Shchepinov 1997). Guo et al. examined the influence of different spacer length between 0 and 15 nucleotides on the hybridization efficiency of immobilized probes and found the highest signals for the probes with a 15 dT-spacer (Guo 1994). The optimum for bound probes might be higher and it strongly depends from the array support. A hydrophobic surface, such as polypropylene, is solvent repellent and hinders hybridization. The best hybridization efficiency can be expected from probes in solution. However, a compromise between optimal hybridization efficiency and expense in probe synthesis has to be made. In this work, a spacer length of 11 nucleotides was successfully used. An alternative solution could be chemical linkers as described by Shchepinov et al. (Shchepinov 1997).

The probe density can negatively influence the hybridization signal by either being too low or too high. It influences the accessibility of probes and hybridization kinetics by defining the maximal absorbable amount of target. This is expressed in the Langmuir adsorption isotherm describing the kinetics of an adsorption process on a surface, which is influenced by the intrinsic adsorption coefficient, the maximal absorbable amount of target, and the concentration of the target in solution. The surface density of the probes can be controlled by the probe concentration in the spotting solution, as there is a linear relation between both parameters (Guo 1994). One oligonucleotide occupies around 6-12 Ų of the chip surface (Maskos 1992). Guo et al. found an optimal surface density of 500 Ų/molecule (2*10¹³ molecules/cm²) by applying an oligonucleotide solution of 5 mM concentration (Guo 1994).

However, for a longer fragment to be hybridized the optimal surface density was significantly lower. It can be assumed that optimal density has to be determined for each array system separately. Other investigations resulted in ~10¹¹ molecules/cm² (Michel 2007) or ~10¹² molecules/cm² (Peterson 2002). The main parameter determining the upper limit of the surface density is not steric interference but electrostatic Coulomb repulsion of the DNA strands (Vainrub 2002). The probe density is also a key parameter for correct quantification on microarrays, as described by Michel et al. (Michel 2007). Non-charged peptide nucleic acids provide an interesting alternative to lessen the unfavourable electrostatic interactions (Shakeel 2006), but would also increase assay cost. Only in electronic-based array systems the charge properties of DNA are positively used (Sosnowski 2002; Zhang 2005b). In this work, the probes were diluted and spotted with a concentration of 20 µM. Under the ideal assumption that the spotting solution is put down to the array surface as a hemisphere and that all probes are immobilized, the probe density was 4.8*10¹³ molecules/cm². This is in agreement with previously published optimal probe densities, if one assumes that not all probes are immobilized to the chip surface.

Due to the lower stability of A:T versus G:C pairs, the hybridization efficiency is also influenced by the target sequence. One study suggests that not only the overall base composition is important but also the sequence (Maskos 1993). Sequence effects are expected, as it is known that base stacking interactions, which depend on nearest neighbours, significantly affect duplex stability (Southern 1999).

Peplies et al. showed that the probe immobilization via 3’ and 5’end resulted in different hybridization efficiencies (Peplies 2003). It was supposed that this is due to steric interference of the non-hybridizing overhang of the target strand with the array support.

The previously mentioned effects may play a role in an observation made in this work. Each probe on the microarray was spotted as sense and antisense version resulting in different signal intensities. This observation remains a phenomenon to be investigated. In most cases, these differences were highest when undigested DNA strands were hybridized. The hybridization may have a direction starting from one end depending on the base composition of the probe. Hybridization might then be sterically hindered by the close microarray surface

124 mainly for one probe of each pair due to their immobilisation direction. As the observed effect could not be correlated with either sense or antisense probe, it might have its seeds in the probe-binding site in the target strand and the base composition of each probe strand. In case of hybridization of non-digested PCR products, a 5’-immobilized sense probe sharing sequence identity with a 5’ end and an antisense probe sharing sequence identity with a 3’

end of the target sense sequence should theoretically display higher signals than their opposites, as the overhang of the hybridizing target strand should not interfere with the microarray surface. In contrast, Peytavi et al. found an inverse correlation between hybridization signal intensity and the length of the 5’ overhanging end of the captured strand (Peytavi 2005). The length of the 3’ overhang had no influence on the signal intensity. They assumed that in case of long solution-directed overhangs kinetic effects and re-association of the PCR product’s complementary strand could lead to destabilization of the capture probe/DNA target duplex. The theory of Peytavi et al. was slightly supported in this work, although clear observations were not made in this regard. A strong correlation of the relative signal response (RI) of probes depending on their position in the target strand was not found for the 19 probes, which hybridized with the four undigested PCR products of E. coli, R. intestinalis, P. shigelloides, and C. coli. Nevertheless, for few probes a signal increase was seen if they hybridized to the 5’ end of the target, which resulted in a surface-directed DNA strand overhang. The direct comparison of corresponding sense and antisense probes also disproved the assumption that a surface-directed overhang negatively influences hybridization. With increasing distance of the 5’ immobilized sense probe to the target middle and location closer to the 3’ end (binding the 5’end of the target), the signal ratio of sense vs.

antisense probe also increased. This means the surface-directed target overhang resulted in higher hybridization signals than the solution-directed overhang. Probes in the target middle displayed nearly equal signals. However, if the sense probe was closer to the 5’end and the corresponding antisense probe closer to the 3’ end, the signal ratio of antisense vs. sense did not continuously increase. This result shows, that a simple theoretical consideration of the 3’ and 5’ overhanging ends of the target cannot fully explain hybridization behaviour.

Moreover, it was observed that the target digestion partly equalized differences between response intensities of sense and antisense probes belonging to one probe pair. This could be due to a reduction of secondary structures of the target strands. Therefore, it is supposed that secondary structures within the target strands, which affect accessibility of the binding motif, and the individual base composition, have comparably big influence on the hybridization behaviour as the relative position of the probe.

Although DNA targets are less problematic than RNA targets in terms of accessibility, secondary structures and duplex formation can hinder hybridization, and thus reduction of the sequence complexity is preferable (Southern 1999). In a recent study of genome-scale probe design, up to a third of all probe binding sites were affected by secondary structures, which made target regions inaccessible for hybridization (Ratushna 2005). By comparing the hybridization efficiency of amplified DNA from four bacterial reference strains, it was shown that most fluorescence signals increased after DNase I-digestion. The average signal intensity was 1.5 times higher after digestion. According to this result, all PCR products were digested prior to hybridization. Digestion influences all target-probe pairs differently depending on the intrinsic thermodynamic parameters of each target strand. Two contradictory effects have to be considered, that is secondary structures are reduced when target strands become shorter and by this hybridization is facilitated but the number of fluorophores per hybridizing DNA strand will be reduced as well. Additionally, it was observed that target fragmentation can support false-positive signals in some cases (Peplies 2003). Therefore, the degree of digestion had to be optimized to find a compromise between these effects, which allows a sensitive and specific detection of all species in parallel. The optimal digestion degree was reached with 0.4 mU DNase I per ng DNA for five minutes at room temperature. However, this parameter can be different in case of a more or less active enzyme, as it was seen in the experiments performed in Shanghai. As an alternative to DNA digestion before hybridization, Nick translation was used to incorporate the label and by this disrupt secondary structures in the target (Lane 2004). However, Nick translation should

125 have a comparable effect to label incorporation during PCR, as the target length remains constant. In this work, an additional improvement of the signal intensities was observed after target digestion. This effect was not investigated by Lane. The direct fluorophore incorporation in combination with DNase I digestion, as it was conducted in this work, takes less time than a PCR with subsequent nick translation. The only advantage of the ladder approach could be an increased amplification efficiency of the target DNA. Instead of shortening the target strands, also helper oligonucleotides were applied to open inaccessible rRNA (Fuchs 2001) and rRNA gene (Peplies 2003) target strands. However, in some cases they reduced the discrimination power, which is an important factor for pathogen identification from complex faecal samples.