Chromosome spreading in well slides
We have previously observed that enzymatic DNA amplification reactions in situ may consume sufficient reagents for local depletion to occur, limiting the reaction if it is performed in a standard format in which the liquid phase is a thin film spread under a coverslip [19]. Therefore, after initial experimentation with the standard system, we converted to a well-based system, since it offers several advantages: 1) A larger volume of reaction mixture can be applied per mm2 of sample, to prevent local depletion of reagents. 2) Less reagent per reaction is necessary compared to standard slides. 3) It is easier to multiplex the reactions.
Initially, we encountered problems due to poor spreading of the condensed chromosomes in well-slides. Subsequently, we designed a protocol optimized to well-slides, based on observations by Henegariu et al. [20]. We found that a combination of hot water vapor, low volume of cell suspension, and heating gave the best results (Figure 2), although the result varied from spread to spread. In particular, the temperature following the second incubation over hot water vapor was important to achieve good spreading of the chromosomes. When the chromosomes were dried at 27–37°C they did not spread enough to make a clear distinction between each individual chromosome, whereas too high a temperature (above 57°C) resulted in over-spreading of the chromosomes. The optimal temperature was in general 47°C (Figure 3).
In situ detection of genomic sequences
We wanted to test whether the in situ padlock system presented by Larsson et al. for the analysis of mitochondrial DNA in situ [9] (see also Figure 1) could be used on condensed metaphase chromosomes. A limitation to this technique is that it involves many enzymatic steps. If each step has for example an efficiency of only 50% it will give a final detection level of 0.5X, where × is the number of enzymatic steps. Thus, with four enzymatic steps the success rate will only be 6%, a figure for the overall efficiency that matches well with the 1–10% range estimated in the Larson study. Indeed, in accordance with this estimate, we had disappointing success rates with single-copy genes (data not shown). This indicates that this method is, at present, best suited for targets represented in several copies in the sample, such as mitochondria, chloroplasts, viruses, and, as presented here, repeated sequences in the genome.
The method worked well with two repeated genomic targets, first of which was the Y-chromosome specific part of the satellite I repeat which contains a 2.7 kb dispersed repeat of high copy number (approximately 2000 copies) [17, 21]. While this target is a repeated sequence, the hybridization target consists of 2000 single copy sequences spread over the long arm of the Y-chromosome. Thus, with an overall reaction efficiency of 1–10%, 20–200 probes/chromosome should lead to the formation of visible rolling circle products. These should appear as discrete single products or larger conglomerates of products where multiple reactions took place close enough to each other for the signals not to be discriminated in the microscope. Furthermore, the expected number of signals per chromosome would seem high enough that all Y-chromosomes should appear labeled, despite unavoidable random variation across the preparation. The second target we detected with high efficiency was the kringle IV domain from the LPA gene sequence. The LPA gene encodes the major component of the plasma protein complex lipoprotein(a) (LP(a)) and is in itself a single copy gene. However, it contains a 5.5 kb kringle IV domain which is a repetitive sequence varying in copy number between alleles (12–51 copies per allele) [22]. Thus, while the target is a single copy gene, the hybridization target is contained within a tandem repeat found in 12–51 copies, which with an overall reaction efficiency of 1–10% should give rise to something like 1–5 rolling circle products per chromosome. With the unavoidable random variation across the preparation, some chromosomes (or chromatids) might not harbor a signal, but overall signals should be found at most sites. The copy-number of the kringle IV domains is inversely correlated to the concentration of plasma LP(a) and the risk for coronary heart disease (CHD) [23], so a versatile means of getting a direct measure of the copy number in a particular individual could potentially be of practical interest, and PRINS has previously proven useful for the sizing of telomeric and trinucleotide repeats in situ on individual chromosomes [5, 6].
The target sequences for satellite I repeat and for the LPA gene were extracted from previously published articles [17, 24]. Representative results with padlock probes targeting short sequence elements in the male specific satellite I repeat and in the kringle IV domain in the LPA gene are shown in Figure 4.
Using the Y-chrom probe, which has a 40 nucleotide target sequence, the long arm of the Y-chromosome was clearly labeled. All Y-chromosomes were labeled and the signal on any one Y-chromosome appeared as a grainy coloring, each grain representing a rolling circle product, or a set of rolling circle products, covering almost the entire long arm of the chromosome, consistent with the known distribution of the satellite I repeat (Figure 4A). No signals were detected from any other chromosomes, indicating that the correct recognition sequence was selectively detected. The exact number of signals per chromosome was difficult to establish, since the rolling circle products were quite numerous, overlapping, and of variable intensity, but the overall results seemed well in agreement with the 1–10% efficiency observed on mitochondrial DNA.
The signals from the probe for a 32 nucleotide sequence in the kringle IV domain of the LPA gene were positioned on the appropriate part of the long arm of chromosome 6. Two distinct signals, one on each sister-chromatid, were ideally obtained from each chromosome 6 (Figure 4B), although in a larger screen, sister-chromatids (of chromosome 6) also appeared with 0 and 1 signals (data not shown). In 70 out of a 100 metaphases both long arms on both chromosome 6 were stained, whereas in the remaining 30 metaphases one or – occasionally – two signals were missing. The overall efficiency could not be estimated with any great precision since: 1) The precise number of kringle IV domains positioned on each allele was not known in our study material. 2) In some chromosome spreads the sister-chromatids could not be differentiated. 3) Individual rolling circle products on one chromatid could generally not be differentiated from each other, because they were situated too closely on the chromosomes.
In summary, we have demonstrated stable detection of short repeated genomic sequences in metaphase chromosomes using padlock probes and target primed rolling circle DNA synthesis, a possibility that seemed questionable based on the literature, and with an overall reaction efficiency in the same range as that previously reported for mitochondrial DNA in situ. Since the overall efficiency is still low, the method seems at this stage of development best suited for the detection of motifs within multi-copy sequences such as repeated sequences in the genome (as presented here), viral DNA, chloroplasts and, as previously described, mitochondrial DNA [9]. It follows from the math of the reaction that adapting it for single target detection would require systematic work to significantly improve the efficiency in each step of the reaction. This may be quite laborious and time consuming, and though the outcome of such an effort, if successful, may be very valuable, it was quite beyond the resources of this study. However, in a parallel study we have addressed the issue of improving one step in the reaction, the ligation of the padlock probes to closed circles. For this we have developed a procedure enabling us to replace the chemically synthesized padlock probes with enzymatically produced oligonucleotides, which have been shown to posses an increased ligation efficiency [25].