Samples flagged by these methods as being high-risk have thus generally been forwarded for examination by the oldest of the genetic testing methods, cytogenetics, through the direct examination of fetal cells. Cytogenetic methods are good at detecting gross genetic changes such as aneuploidies and some translocations, although less (or un-) able to detect smaller-scale mutations. Further, cytogenetic methods suffer from their underlying need for fetal cells to examine, which by nature would seem to require an invasive sampling method. Two such sampling approaches, chorionic villi sampling (CVS) and amniocentesis, have been well established and can serve in this role. Both, however, come with some finite risks of loss of the fetus, with studies indicating these risks can range as high as one percent—which is far from insignificant.
The topic of this month’s “Primer” can address these risks, improve on classical cytogenetics’ capability to detect small molecular defects, and have better diagnostic accuracy than serum marker screening or ultrasound imaging. Molecular diagnostics of fetal cells started with the clear demonstration in 1979 that fetal cells were detectable in maternal peripheral blood during pregnancy. (That is, the assumption that fetal cells are available only by [uterine] invasive methods was not strictly correct.) These fetal cells include erythroblasts, trophoblasts, granulocytes, and lymphocytes, all of which are nucleated and could theoretically be amenable to cytogenetic analysis. The challenge, however, is that these fetal cells make up a very small fraction of cells in a maternal peripheral blood specimen. Published estimates are that fetal cells range from one per 10,000 to one per 1,000,000 maternally derived cells in these samples, or “one cell per ml whole blood” as a (very) crude rule of thumb. Regardless of which end of these estimates holds true for a given sample, the truth remains that trying to selectively recover just the fetal cells from maternal blood is challenging in the extreme.
This doesn’t mean it can’t be done, however. Protocols exist, usually based on gradient centrifugation (to get mononuclear cells, both maternal and fetal) and then either fluorescence-activated cell sorting (FACS) or magnetic separation approaches, utilizing tagged antibodies specific to fetal cell surface markers. Fetal cells enriched through this approach can now be subjected to classical cytogenetics, or have DNA extracted for analysis by any molecular approach such as comparative genomic hybridization (CGH) or single-nucleotide polymorphism (SNP) arrays, targeted Sanger sequencing, next-generation sequencing (NGS), or targeted locus-specific polymerase chain reaction (PCR) assays. While the power of this approach is obvious, it’s hampered in practice by the unfortunate truth that the underlying sample preparation process (regardless of specifics) remains too costly, time consuming, and low-yield to be ideal for routine use.
A simpler route to obtaining fetal DNA became apparent in 1997 when research was published indicating that cell-free circulating DNA of fetal origin had been detected in maternal peripheral blood. While this too occurs in a background of maternally derived cell-free DNA, the proportion of fetal to maternal contribution is much more favorable, with estimates suggesting it represents three percent of total free circulating DNA in the first trimester, increasing to possibly in excess of 10 percent by end of term. While this has gone a great way toward improving on the amount of fetal DNA available over that in isolated fetal cells, it has also presented the challenge of how one goes about distinguishing which bits of free-floating DNA originated from the mother as opposed to the child.
One clever approach to detection of aneuploidies simply ignores this problem. Imagine, for example, that we have a mother with a trisomy 21 fetus, and we take the circulating free DNA from maternal peripheral blood and just throw all of it on an NGS platform. These platforms all work, in effect, by sequencing millions of short sequence elements in parallel, and since we have full human reference genomes to compare against, many of these “reads” can be uniquely mapped back to a specific chromosome and position. Barring intricacies of differential detection of chromosomal regions (which can be compensated for with bioinformatics data processing to some degree), the number of “reads” mapping to a chromosome relates to the physical linear amount of each chromosome present in the sample. That is, more reads are generated from a long chromosome than a short one—but also, more reads are generated from three copies of a chromosome, than two. Since our hypothetical sample here has a very small fractional excess of chromosome 21 (50 percent extra in the small total fraction of DNA arising from the fetus), if we do a large enough number of reads, eventually this excess will give rise to a small but statistically relevant overabundance of chromosome 21 reads compared to all the other chromosome reads in the sample.
As described, this method can detect essentially any aneuploidy condition; however, it requires an enormous number of reads per sample, which translates to high cost and low throughput. An improvement on this is to carry out a targeted NGS, whereby, for instance, only representative regions of the highest-risk chromosomes X, 13, 18, and 21 might be read. This vastly reduces the total number of reads needed per sample and allows cheaper assays with higher throughputs (as many more samples can be included on a single NGS run). The trade-off is that only aneuploidies impacting the queried chromosomes will be detected; however, in most cases this is an acceptable balance.
A still simpler approach to this basic method has been suggested through the approach of digital PCR, targeted to a similar set of informative chromosomal markers. This approach promises simplicity both of performance and data interpretation, as it is inherently highly accurate for quantitation.
What if we wanted to take advantage of the (relative) prevalence of free fetal DNA in maternal circulation, but still wanted some way to identify and examine it as distinct from maternal? That is, can we have our proverbial cake and eat it too? It turns out that we can, at least sometimes. The simplest and most immediately obvious case would be when we’re interested in a holandric (Y-chromosome) fetal locus. If we assume that any circulating Y-chromosome DNA in the maternal circulation is fetal in origin, then all we need do is apply PCR, Sanger sequencing, or whatever our method of choice is to the locus of interest to get our desired information (more about the validity of this assumption, below).
Unfortunately, not that many loci of interest reside on the Y chromosome. Recall, however, that every person has a unique set of single nucleotide polymorphisms (SNPs) scattered across their genome, meaning that for every SNP loci there is a finite chance that the father and mother will have a differentiable marker. If we know the SNP genotypes of both parents and they differ at a specific locus, it is then possible by analogy to the Y chromosome example to identify the fetal genotype for that locus. (This is probably most intuitively obvious in the case where a read contains a uniquely paternal-derived SNP; it’s presumably from the fetus, and allows us to make assumptions about fetal carriership of paternal gene alleles closely linked to the SNP.) Note that it’s not even a requirement that the parents each be homozygous (and different) at the associated SNP; for example, if the maternal SNP is “A/A” homozygous and the paternal SNP is “A/C” heterozygous, then with deep enough sequencing, if no “C” SNP reads are found, one can assume the fetus to have gotten the paternal “A” allele.
The exact amount of information which can be obtained from SNP linkages in this sort of analysis will vary from case to case, but with proper context and bioinformatic analysis it can be helpful in determining risk of particular Mendelian inherited conditions. Mass readout of SNP genotype can be done by NGS or array-based methods, or for a smaller selected allele set of interest by classical PCR approaches.
In addition to the SNP-based methods for selectively identifying non-maternal (and thus presumably fetal) content in circulating peripheral free DNA, DNA methylation state (that is, imprinting) may have some current or future applicability as well.
In all of the above, we’ve been working with the tacit assumption that if DNA (cellular or free-floating) in maternal peripheral circulation isn’t maternal, it’s from a current fetus. Is this an unassailable truth? Probably not. One case study has detected fetal cells in maternal peripheral circulation up to an astounding 27 years postpartum (suggestive that during pregnancy, some of the fetal cells have actually engrafted within the mother and continued to propagate in a form of microchimerism). Additionally, examples of male microchimerism (that is, trace amounts of male DNA) have been reported in women without a known history of pregnancy with a male fetus, suggesting other possible sources such as an in-utero absorbed fraternal (male) twin, or possibly even from a mother’s male older sibling. While these are intriguing, it’s important to note that these anomalies generally have been found at very low levels—much lower than the 3%+ of total circulating free DNA which is well associated with a current fetus. For the types of applications described here, these vanishingly small template amounts from “other sources” will generally not alter the results obtained, although they present a warning not to rely solely on an ultrasensitive qualitative detection of any marker as irrefutable evidence of fetal genotype.
In the context of proper bioinformatics analysis and genetic counselling, prenatal genetic diagnosis from maternal peripheral blood is a useful tool and likely to become increasingly commonly applied as NGS methods become cheaper and more widely available in clinical settings.