Nucleic Acids Research

DNA preparation from whole blood

Permission to use blood taken from the patients was obtained from the Ethics Committee at our institution. One ml of maternal blood and 2 ml of fetal cord blood were subjected to EC lysis and DNA from the nucleated cells was isolated using DNAzol (Life Technologies, Basel, Switzerland).

Isolation and preparation of nucleated red blood cells (NRBCs)

A mixture of 2.5 ml maternal blood and 0.5 ml cord blood in 15 ml phosphate buffered saline (PBS) was centrifuged at 1600 g for 10 min over ficoll hypac (Pharmacia, Dubendorf, Switzerland) at a density of 1.110 g/ml. The nucleated cell fraction was washed in PBS, after which cells were transferred onto glass slides by cytospin centrifugation (Shandon, Frankfurt, Germany) and stained with hemotoxylin/eosin. NRBCs were identified by their unique morphology, namely a compact darkly staining nucleus smaller in diameter than an erythrocyte. Single cells were picked with glass needles using a Zeiss micromanipulator and inverted microscope at 400× amplification. Cells were individually placed in 0.5 ml microfuge tubes containing 5 µl of 400 ng/µl proteinase K and 17 µM SDS. The solution was overlaid with mineral oil and incubated at 50°C for 1 h, then at 99°C for 30 min.

PCR

Primer pairs for each of the loci listed in Table 1 were purchased from Research Genetics (Huntsville, AL) and contained the Fam fluorescent dye attached to the 5[prime] primer of each pair. The initial genotyping of fetus and mother was done using 100 ng of DNA in a 20 µl reaction volume cycling 94°C for 30 s, 55°C for 30 s, 30× using standard PCR conditions (2 mM MgCl₂).

Table 1. Genotyping of mother, fetus and cell picks
Four heterozygous and one homozygous loci were analyzed. The numbers for each allele refer to the % of total signal at a given locus. NA, no amplification; ND, not done.

Semi-nested multiplex PCR was performed on each NRBC using five primer pairs for the external reactions simultaneously. The additional primers which we used for nested PCR were 5[prime]-GCCTGGCAGGCTAAGAGTTA-3[prime] for D18S535, 5[prime]-AACTCACTCTCTCTCTCTCTT-3[prime] for D21S1270, 5[prime]-AATATCTCTCCAGACACATGA-3[prime] for D13S767, 5[prime]-TTCTAAAAGAAATCAAAATGATGC-3[prime] for D21S1432 and 5[prime]-ATGTGTGTATGCCAGCCTCTG-3[prime] for D21S1440. PCR mix (50 µl) containing each of the 10 primers at 400 nM, 2 mM MgCl₂ and standard PCR buffer (Life Technologies, Basel, Switzerland) were added to each tube containing a cell. The tubes were heated to 94°C before addition of 2 U of Taq polymerase. After 30 cycles of amplification as described above, 1 µl aliquots were removed and used as template for the internal reactions using conditions identical to those used with 100 ng of DNA except that hot start was performed.

PCR products were first checked on a 2% agarose gel and then analysed on an ABI310 prism automated sequencer using the Genescan software program (Perkin Elmer, Foster City, CA). Tamra labeled GeneScan 500 molecular weight standards were included as internal size standards for each run.

Figure 1. Electropherograms of locus 18S535 for maternal DNA, fetal DNA and 10 cell picks.

Results

Identification of informative loci

Fetal DNA isolated from cord blood and maternal DNA from maternal blood were used to genotype the two individuals at five microsatellite loci. The allele sizes are shown in Table 1. Two loci (D18S535 and D21S1270) were used to prove fetal origin in candidate fetal cells by virtue of the unique paternal allele present in the fetus. The other three loci were used to test the ability of single cell PCR to genotype monoallelic (D13S767) and diallelic (D21S1440 and D21S1432) loci.

Identification of fetal cells

Cord blood (0.5 ml), containing NRBCs at a frequency of 4% of total nucleated cells, was mixed with 2.5 ml maternal blood that had no NRBCs detectable by hemotoxylin/eosin staining of whole blood. Nucleated cells were purified by density centrifugation, placed on slides and stained. Since adult blood has a higher leukocyte count than cord blood, NRBCs represented ~0.3% of nucleated cells on these slides, which is similar to the frequency of fetal NRBCs in the blood of pregnant women after enrichment with anti-CD71 coated magnetic beads (11,12). Eight NRBCs, one neutrophil and two erythrocytes were picked and individually subjected to multiplex PCR at the five loci listed in Table 1. The ten primer pairs required for the external amplification step of semi-nested PCR were added to each cell and 30 cycles of PCR were performed. Two 1 µl aliquots of the external PCR products were removed and used as template for two separate amplifications of the two loci used for identification. Those cells proven to be fetal were then analysed at the other three loci.

Electropherograms of locus D18S535 for all eight NRBCs, the neutrophil and one erythrocyte are presented in Figure 1. The neutrophil had the expected maternal genotype, while the erythrocytes were negative for all five loci (Fig. 1 and data not shown). One of the eight NRBCs (NRBC 2) did not amplify at either of the two loci used for identification and was not analysed further. The other seven NRBCs amplified at both loci used for identification, and as can be seen in Figure 1 for locus D18S535, either one or both fetal alleles were amplified from each of these cells.

By examining only one locus (18S535), three NRBCs could be proven to be fetal either by virtue of the amplification of both alleles (NRBCs 4 and 8) or only the paternally inherited fetal allele (NRBC 1). On the other hand, by examining both loci, six of the NRBCs could be shown to be fetal (Table 1). No conclusive identification could be made for NRBC 5, since the paternal allele failed to amplify for both loci. This cell was not analysed further. Cells 1, 3 and 4 had the definitive paternal allele at both loci, while cells 6, 7 and 8 possessed the paternal allele at only one locus, and had ADO of the paternal allele at the other locus. ADO occurred in five out of the 12 amplifications (42%) used to prove fetal origin in these six cells.

Genotyping monoallelic and diallelic loci

We next examined three other loci in these six NRBCs, which we had shown to be fetal based on the presence of a unique paternal allele. Of these three loci, locus D13S767 is homozygous in the fetus, while loci D21S1440 and D21S1432 are heterozygous. All three loci were correctly genotyped using the combined data from the six cells proven to be fetal (Table 1). Furthermore, the two loci used to prove fetal origin can also be considered as test loci if cells are used which are proven fetal at other loci. For example, NRBCs 1, 3, 4, 6, 7 and 8 can be proven fetal using loci D21S1270 and D21S1432. Analysis of D18S535 in these cells clearly demonstrates that the locus is heterozygous. Similarly, locus D21S1270 can be proven heterozygous using cells 1, 3, 5, 6, 7 and 8 proven to be fetal due to paternal alleles at loci D18S535 and D21S1432. Thus all four heterozygous loci tested were correctly shown to be heterozygous by averaging the signals from the six cells analyzed.

Frequency of ADO

Although these results showed that a correct diagnosis can be made despite the problem of ADO, provided that multiple cells are examined by single cell PCR, we wished to quantitate the frequency of ADO by analyzing a sufficiently large number of heterozygous loci amplified from single cells. In Table 2 the results are presented for 92 amplifications done at eight different loci in a total of 50 cells. ADO was considered to have occurred when one allele had <10% the intensity of the other allele, and hence was too close to background to score reliably. This arbitrary 10% minimum may result in an overestimation of the ADO frequency, but it should also reduce the likelihood of scoring background noise as an allele. These data show that the frequency of ADO is 62%, and that ADO is indeed random, with the smaller alleles dropping out with roughly the same frequency as the larger alleles at a given locus.

Table 2. ADO in single cell PCR

Locus	Only smaller allele amplifies	Only larger allele amplifies	Both alleles amplify	Number of cells examined
D18S535	3	1	4	8
D21S1432	6	11	7	24
D21S1437	5	3	4	12
D21S1270	2	1	5	8
D21S1440	3	2	5	10
GATA129D11	6	2	2	10
GATA71H10	3	1	3	7
DXS6785	3	5	5	13
Totals	31 (34%)	26 (28%)	35 (38%)	50a

^aMultiplex PCR was done on some of the cells, thus the total number of cells examined is less than the total number of loci examined (92).

Table 3. The frequency of misdiagnosing a heterozygote as homozygous as a function of cells analyzed with an ADO rate of 62%

Cells analysed	% misdiagnosed
1	62
2	19.2
3	6.0
4	1.8
5	0.57
6	0.18
7	0.055
8	0.017

Discussion

The accurate diagnosis of heterozygous loci that we describe requires the averaging of a small number of single cell PCR, hence it is technically not a single cell procedure. However, for most forensic and prenatal diagnosis applications a small number of cells are usually available. Genetic diagnosis of the preimplanted embryo is exceptional in this respect in that only a single cell is available. Primer extension preamplification (PEP) as method for whole genome amplification should allow multiple amplifications of the same locus from a single cell. Unfortunately, we have found that ADO and spurious products of the incorrect size are even more of a problem when preamplified template from a single cell is used as template for PCR.

Our four out of four success rate in correctly identifying a locus as heterozygous in Table 1 may seem high, especially when considering the problem of ADO. However, it is to be expected given the number of cells analysed and the fact that the frequency of ADO is <100%. The likelihood that all cells suffer ADO at a given locus decreases exponentially as the number of cells tested increases. Furthermore, ADO is random for most loci, hence the chance of the same allele dropping out in all amplifications suffering ADO also decreases exponentially with the number of cells tested. The rate of misdiagnosis equals 2× (frequency of ADO × 5)ⁿ , where n equals cell number. Table 3 shows the calculated rate of misdiagnosing a heterozygous locus as homozygous when the ADO rate is 62%. When six cells are analysed, the diagnosis will be correct at 99.82% of all loci tested. Therefore, when using six cells, we can expect to correctly genotype four out of four loci 99.28% of the time. Accuracy at this level should be sufficient for many clinical applications.

Acknowledgements

This work was supported in part by Swiss National Science Foundation Grant Number: 3200-047112.96 and NIH (USA) Contract Number: N01-HD-4-3202.