Cell-Based Noninvasive Prenatal Testing: A Promising Path for Prenatal Diagnosis

Introduction: The Changing Landscape of Prenatal Diagnosis

In recent years, increased interest in noninvasive prenatal testing (NIPT) methods led to the commercialization and clinical implementation of NIPT based on the analysis of “fetal” cell-free DNA (cfDNA) isolated from maternal blood samples; this has caused a drastic transformation of the field of prenatal screening and diagnosis. Since the initial report on “fetal” cfDNA being present in maternal plasma by Dennis Lo and colleagues in 1997 , the clinical potential of this finding has been elaborately described and validated, with a substantial decrease in amniocentesis and chorionic villus sampling (CVS) diagnostic procedures as a consequence .

However, one inherent disadvantage is the overwhelming amount of maternal cfDNA that is inevitably analyzed together with the “fetal” fraction. The “fetal” cfDNA only comprises 5%–15% of the total plasma cfDNA pool. cfDNA-based NIPT can thus currently only be recommended to be offered as a screening test for trisomy 21, 13, and 18, and any positive result should be confirmed by diagnostic methods. For these common autosomal aneuploidies, cfDNA-based NIPT has an undeniably higher positive predictive value (PPV) than serum marker analysis using a combination of 3, 4, or 5 markers (alpha-fetoprotein [AFP], human chorionic gonadotropin [hCG], and unconjugated estriol [uE3] for a triple screen, with the addition of dimeric inhibin A [DIA] for a quad screen and hyperglycosylated hCG as well for a penta screen). The detection of other autosomal and sex chromosome aneuploidies, and subchromosomal copy number abnormalities is less adequate, and multiple laboratories have reported high false-positive rates and low PPVs . cfDNA-based prenatal testing is thus not recommended for these indications by the leading professional organizations , although multiple commercial providers do offer such testing.

About 6000 children per year are born in the United States with Down syndrome. Trisomy 21 is the most common genetic disability; however, it makes up only about 5% of all genetic disabilities. Other autosomal and sex chromosome aneuploidies and microdeletions and duplications on their own are significantly rarer but adding them all together leads to a much higher incidence than Down syndrome. Additionally, a whole range of both inherited and de novo Mendelian disorders can occur, adding yet another large population of affected pregnancies deserving adequate prenatal testing and diagnosis. Taking this into consideration, it is thus clear that current cfDNA-based prenatal testing only provides few pieces of the total landscape ( Fig. 1 ).

Additionally, the interfering maternal cfDNA makes the incidental discovery of maternal findings unavoidable, which is one of the main causes for false-positive cfDNA-based test results. Furthermore, sometimes no results can be generated in case of low fetal fraction, due to fetal or maternal characteristics such as a very small placenta or a high maternal body mass index (BMI) . Also, samples from multiple pregnancies pose challenges for cfDNA-based NIPT.

Prenatal testing starting from fetal cells, on the other hand, provides a better alternative. The trafficking of intact fetal or trophoblast cells into the maternal circulation is a natural phenomenon during pregnancy, and these cells provide a DNA source of purely fetal or placental origin without any maternal contamination. The isolation and analysis of multiple single fetal or trophoblast cells gives multiple results per sample, which contributes to a high confidence test result and quality. This would also enable a more nuanced detection of potential placental mosaicism. Specifically determining the individual genetic profile of each analyzed cell makes it possible to obtain a result for every fetus within a multiple pregnancy.

Fetal Cells in the Maternal Circulation

Fetal cells can be detected in the mother's circulation already very early on; Ariga et al. report the detection of fetal cells already at 4 weeks’ gestational age (GA) . Altogether, there are four different groups of fetal cells that can be distinguished: (i) trophoblastic cells, (ii) fetal nucleated red blood cells (fnRBC), (iii) lymphocytes, and (iv) stem cells and progenitor cells. The life span of these cell types is determining their use in prenatal testing: fetal lymphocytes and stem/progenitor cells may persist in the maternal circulation for even years after delivery , making them unsuitable for prenatal testing/diagnosis. The reasons behind their persistence are multifold and they are likely to have several beneficial roles such as promoting maternal tolerance against the fetus, but are contrastingly also involved in the development of maternal autoimmune diseases—the immunological implications of pregnancy-related microchimerism are reviewed elsewhere .

Trophoblasts and fnRBCs, on the other hand, disappear after delivery and thus only reflect the pregnancy ongoing at the time of testing, without confounding of previous pregnancies. Fetal cytotrophoblasts from anchoring villi invade the maternal endometrium and migrate to remodel the spiral arteries into larger vessels . The migration of these extravillous trophoblasts is a prerequisite for normal fetal and placental development, and occurs in two waves (one at 8–10 weeks and one at 16–18 weeks GA) . In contrast to the multinucleate syncytiotrophoblasts, cytotrophoblasts are mononucleate and an attractive target for prenatal testing. However, these cells are technically derived from the placenta instead of the fetus itself, and as a consequence concerns have been raised that they are susceptible to confined placental mosaicism. Fetal trophoblastic cells can be obtained both from maternal blood and through cervical sampling .

Also, fnRBCs escape into the mother's blood stream and are considered to be “truly fetal” instead of placenta derived. In the first trimester, these fnRBCs consist of both primitive and definitive erythroblasts, which differ both in morphology and in location where they are produced . The definitive fnRBCs are produced in the fetal liver and bone marrow, but before 6 weeks’ GA erythropoiesis still takes place in the yolk sac, where thus the primitive cells are formed. Precautions need to be taken for the isolation and analysis of fnRBCs, as the simultaneously present maternal immature RBCs might be isolated along with them: the portion of fnRBCs is estimated to be 30% of the total nRBC population, although this may vary between patients and across gestation . It has been suggested that fnRBCs only contribute a very small fraction of the total fetal cell population .

Independent of which cell type is focused on, the biggest obstacle to overcome remains the very low number of fetal cells available. The publications by Hamada et al. and Krabchi et al. describe some of the most laborious efforts to quantify the total number of fetal cells. Both groups used a similar approach of only minimally manipulating blood samples collected from male pregnancies, prior to an elaborate microscopic analysis. In short, after maternal RBC lysis and fixation, the obtained nucleated cell suspensions were spread on microscopy slides for fluorescent in situ hybridization (FISH) specifically for the X and Y chromosome. No specific enrichment was done, nor were the microscopy slides analyzed for any characteristics specific for a certain fetal cell type. By simply counting the cells with an XY profile, the total number of fetal cells was obtained. Hamada et al. reported a range of 1 fetal cell in 144,000 maternal to 1 in 4000, while Krabchi et al. found 2–6 fetal cells per mL of maternal blood for a normal pregnancy. The latter group confirmed all fetal cells by use of reverse FISH, in which the same X and Y probes were used as the first step but now with reversed color.

Different factors might influence the number of fetal cells such as GA, fetal genetic abnormalities, maternal and placental health, physical activity of the mother before the blood draw, and normal interindividual biological variation. It is important to keep in mind that differences in blood collection conditions, sample size, targeted fetal cell type, isolation protocol with or without enrichment, detection method, and so on, might substantially contribute to the variability between study results.

Multiple studies have been reported on the association of the number of fetal cells with GA, with varying results. The study by Hamada et al. reports a significant increase in total fetal cells from < 1/100,000 in the first trimester to about 1/10,000 at term. Ariga and colleagues collected samples from 4 weeks’ GA until 39 days after delivery, and noticed a temporary peak in fetal cells around 12 weeks and another more steady increase after 20 weeks. The samples taken after delivery showed a quick clearing of the fetal cell population after birth. Other groups have focused on one specific fetal cell type. Bianchi and colleagues did not detect any fnRBCs after 16 weeks , while Takabayashi et al. described that fnRBCs could be retrieved at earliest at 8 weeks but no more after 24 weeks GA . Rodríguez de Alba et al. found a significant increase in fnRBCs from first to second trimester . Shulman and colleagues found no significant association between the number of fnRBCs recovered and the GA , and neither did Lim et al., although they noted a trend of decreasing numbers for both fnRBCs and trophoblast cells with increasing GA . For trophoblast cells retrieved from the cervix, no correlation with GA was seen in a large series of samples collected between 5 and 20 weeks’ gestation . The variability in results makes it difficult to draw a definitive conclusion on the potential association between GA and number of fetal cells. The total fetal cell population seems to increase over the course of a pregnancy, but fnRBCs might only contribute to this increase until about 20–25 weeks’ GA. The number of trophoblastic cells appears to be rather decreasing over time, although no significant correlation was found so far. Most importantly, all above-mentioned studies show that it should be feasible to obtain a sufficient amount of fetal cells noninvasively in the first or early second trimester, which is the time frame aimed for in prenatal genetic diagnosis ( Table 1 ).

Also, fetal genetic abnormalities are known to affect the number of fetal cells. Bianchi and colleagues reported a sixfold increase in fetal cells in blood of women carrying a male trisomy 21 fetus, with an increase from a mean of 1.2 cells/mL of blood to 6.9 cells/mL, as measured by Y-PCR quantitation . Also Krabchi et al. described that the total number of nucleated fetal cells was 2–5 times higher in aneuploid compared to euploid pregnancies around the same GA .

Fetal-maternal cell trafficking and even more so the release of placental cfDNA might be substantially altered in other pregnancy-related pathologies as well, as was reviewed by Hahn and colleagues . They, for example, refer to one study reporting a 37-fold elevation of fetal trophoblasts in uterine vein blood of preeclamptic women, and also the number of erythroblasts went up drastically. An interesting observation was that the number of circulating fetal cells was most often already elevated before the onset of any clinical symptoms. Therefore both Hahn et al. and van Wijk et al. proposed to use fetal cells and fetal cfDNA as a tool to investigate abnormal placentation . This trend was not confirmed for endocervical trophoblast sampling, as Fritz et al. described a nonsignificant decrease in numbers of cells in samples collected from women with preeclampsia or a fetus with IUGR .

Contrary to fetal cfDNA-based testing , cell-based NIPT is not impacted by maternal BMI, neither for trophoblasts collected from maternal blood (Vossaert et al., unpublished observation), nor for cells obtained via cervical sampling .

Another factor to consider is the potential influence of an invasive procedure. Christensen and colleagues compared the yield in fnRBCs before and after CVS, which clearly causes fetal bleeding into the maternal circulation as they obtained many fnRBCs post-CVS but practically none prior to the procedure . This thus emphasizes the importance of working with preprocedure samples during the process of cell-based NIPT protocol optimization.

Evolution of Cell-Based NIPT

Early Reports: A Century of Research

Fetal cells in the maternal circulation have been studied for decades. Schmorl wrote one of the first reports on fetal-maternal cell trafficking in 1893, describing the presence of trophoblasts in the maternal pulmonary capillaries in women who died of eclampsia . He had found large thrombi in maternal lung tissues, which partially consisted of large multinucleated cells, most likely originating from the placental villi and invading the lungs during eclampsia. These thrombi were not present in the lungs of women who had died shortly after giving birth but not due to eclampsia, leading Schmorl to conclude that the number of fetal cells exchanged in noneclamptic women were probably much lower. Several studies from the 1950s and 1960s describe the presence of fetal cells in the maternal circulation, although often these findings were based on samples collected after delivery . In 1969, the usefulness of circulating fetal cells for prenatal genetic testing was proposed for the first time by Walknowska and colleagues , who reported on the presence of male cells in lymphocyte culture preparations from maternal blood collected from 14 to 37 weeks of gestation. Fetal cells have been studied ever since, and the evidence gathered over the years has only corroborated their diagnostic potential.

As mentioned previously, the two fetal cell types suitable for prenatal diagnosis are trophoblastic cells and fnRBCs, and various protocols have been published for the specific isolation of each cell type, with the early focus mainly on fnRBCs. The earliest studies relied solely on the detection of an XY profile among the vast pool of maternal cells, independent of any cell type-specific characteristics. One of the first reports including an extra selection step was published in 1979 by Herzenberg and colleagues . They incorporated fluorescence-activated cell sorting (FACS) based on paternal human leukocyte antigens (HLA) absent in the mother. The selected fetal candidates were consequently confirmed via microscopic detection of the Y chromosome.

Dozens of other reports followed over the years, further consolidating the existence of circulating fetal cells. Generally, the basic protocol for enrichment included removing the excess of maternal cells, a specific enrichment step for fetal cells and finally applying an adequate detection method.

The initial blood-processing step consisted of bulk separation or more selective density gradient centrifugation (DGC), selective lysis of maternal RBCs, or a combination of aforementioned techniques.

For bulk separation, the blood sample is centrifuged in a specialized tube to separate the maternal RBCs and plasma with a layer of nucleated cells, the so-called “buffy coat” in between. This buffy coat includes the fetal and trophoblast cells . This method was further refined by adding a specific density component, such as Ficoll or Percoll . Ficoll is a hydrophilic polysaccharide (polymerized sucrose) solution, which is osmotically inert but yet toxic for cells. Percoll solution contains colloidal silica particles that are coated with polyvinylpyrrolidone, which makes them nontoxic for cells. They serve as a means to better mark the interface between the RBC layer and the plasma and make it easier to isolate the layer of nucleated cells including the fetal and trophoblast cells. Several studies have used and compared these reagents, and generally conclude that adding a higher density component results in a higher fetal cell yield .

The selective lysis of maternal but not fetal RBCs had already been suggested in 1976 by Boyer et al. . This method was based on the fact that fetal cells only have carbonic anhydrase II while maternal cells have both anhydrase I and II activity. After adding the appropriate substrate (potassium bicarbonate and ammonium chloride), the ammonium bicarbonate that is formed subsequently will attract water into the cells, until they eventually burst. Given the higher enzyme concentration in maternal cells and with the help of acetazolamide, a selective blocker for carbonic anhydrase II, maternal cells are selectively lysed over fetal cells .

Additional depletion was often included by means of magnetic-activated cell sorting (MACS) with specific maternal white blood cell (WBC) markers. Surface antigens such as cluster of differentiation (CD) markers are good targets for this purpose: CD45, also known as leukocyte common antigen, and CD14, or monocyte differentiation antigen, are widely used to deplete maternal WBCs .

The confirmation methods applied after cell isolation were however not ideal: XY FISH can only be used for male fetuses and HLA genotyping is patient specific. Extensive efforts were put into finding specific markers for fnRBCs and trophoblastic cells, to allow for specific fetal cell selection by FACS or more often MACS, and optimized confirmation by immunostaining ( Table 2 ).

Table 2

Markers and Characteristics for Fetal Cell Selection and Identification

Fetal Cell Type		Marker or Characteristic		References
fnRBCs	Surface markers	CD71	Transferrin receptor, or cluster of differentiation 71, expressed during different stages of erythrocyte maturation
		CD36	Thrombospondin receptor, expressed on erythrocytes during early differentiation
		GPA	Glycophorin A, a RBC-specific marker
		Antigen i	Only expressed on fetal and newborns erythrocytes
		HAE9	Specific antibody to cell surface antigen of human nucleated erythroid cells
		CD34	Hematopoietic progenitor cell antigen, expressed on progenitor cells in very early developmental stages
	Cytoplasmic markers	HbF	Fetal gamma hemoglobin is the predominant cytoplasmic protein in fnRBC after 7 wks GA
		Epsilon Hb	Embryonic epsilon hemoglobin, more specific for fetal cells than HbF but only detected at very early gestation
		Zeta Hb	Embryonic zeta hemoglobin, expression decreases even earlier than epsilon Hb
	Charge flow separation		Separation based on the difference in surface charge density of different cell types
Tropho	Surface markers	FD0161G	Antibody specific for extravillous trophoblastic cells
		hPL	Human placental lactogen or human chorionic somatomammotropin, a placental hormone
		HLA-G	Human leukocyte antigen-G, a nonclassical major histocompatibility complex class I antigen
		Mab340	Monoclonal antibody 340, specifically for cytotrophoblasts
	Isolation based on size (ISET)		The slightly larger trophoblasts are caught on a filter, while the smaller maternal WBCs flow through

Abbreviations used: fnRBCs , fetal nucleated red blood cells; GA , gestational age; Tropho , fetal trophoblastic cells; WBC , white blood cell; wks , weeks of gestational age.

CD71 is the most extensively tested fnRBC selection marker, although studies showed it is insufficient on its own for adequate cell isolation . Bianchi described the use of anti-CD71 or anti-CD36 antibodies in combination with anti-glycophorin A (GPA) . When analyzing the samples by flow cytometry, they noticed two different groups of CD71 +/GPA + cells, on one hand, and a considerably larger group of CD36 +/GPA + cells, on the other hand. The addition of GPA to the other antibodies significantly brought up the fetal gender prediction accuracy: from 57% correct with CD71 alone or 88% with CD36 alone, up to 100% correct prediction when the combination was used. These markers also were popular targets for fnRBC selection in multiple other studies (e.g., Refs. ), as was reviewed by Jackson . Additional strategies included targeting antigen i , the use of specific antibody HAE9 , or fnRBC selection based on CD34 , although these did not prove to be as useful as GPA.

In the meantime the suitability of cytoplasmic fetal (gamma) and embryonic (epsilon and zeta) hemoglobins (Hb) was investigated as well, both for cell sorting and detection/identification of fetal cells . Fetal gamma hemoglobin (HbF) is significantly higher expressed in fnRBCs than in adult nRBCs , but the maternal HbF expression levels in blood can be upregulated under certain conditions, which hinders its use as a fetal cell marker . Choolani et al. studied the levels of epsilon Hb by making mixtures of fetal blood and blood of women who had never been pregnant and by studying samples of women carrying a male fetus, collected at 7–14 weeks’ gestation . They described a linear decrease in epsilon Hb + cells with increasing GA, to an almost insignificant level at 14 weeks. Christensen and colleagues found that staining for epsilon Hb showed a higher fetal specificity than staining for HbF in samples collected post-CVS at 9–14 weeks’ GA, but did not find any epsilon + cells pre-CVS at that same time frame . Given that the expression of zeta Hb is lost even earlier than epsilon Hb , both epsilon and zeta Hb are not useful as fetal cell markers.

Most reports showed a rather low fnRBC recovery ranging from 0 to 20 cells per blood sample . In contrast, Wachtel and colleagues reported a far higher recovery of fnRBCs by applying charge flow separation (CFS), even claiming a recovery of about 2000 fnRBCs per sample as confirmed by XY FISH . This method uses specific surface charge densities of each cell type to separate the different blood cell types into individual compartments.

Also for trophoblasts multiple studies were done in the search for more specific markers. Mueller et al. developed an antibody, FD0161G, that is specific to extravillous trophoblasts , and the same group found later that also human placental lactogen can be used as a target in MACS-based positive selection . van Wijk and colleagues adopted the marker HLA-G into their workflow, which is a nonclassical major histocompatibility complex class I antigen present on extravillous trophoblasts, where the classical HLA-A and HLA-B are not expressed . They only recovered few cells, but as their laborious workflow resulted in about 17 microscopy slides per sample, they only analyzed one slide per sample; the amount of fetal cells reported is thus not a definitive number. Durrant et al. developed yet another antibody specifically for cytotrophoblasts, Mab 340, but this antibody cross-reacted with syncytiotrophoblasts as well.

Vona and colleagues took a slightly different approach and set out to select for trophoblastic cells based on their size . In the isolation by size of epithelial tumor/trophoblastic cells (ISET) protocol, diluted blood is filtered through a polycarbonate filter with 8-μm pores, onto which larger cells including the trophoblasts are retained. After hematoxylin and eosin staining or immunostaining (cytokeratin +, placental alkaline phosphatase +, CD45 −), the cells are removed from the filter by laser capture microdissection (LCM) to undergo further analysis. They found 0.5–3.5 trophoblastic cells/mL of maternal blood.

As illustrated here, a myriad of variably successful protocols for blood processing and fetal cell selection has been published so far. It can be appreciated that less manipulation of the sample is better for preserving intact fetal cells. None of the described markers has 100% fetal cell specificity; CD71 whether or not in combination with GPA, and HLA-G seemed among the better ones for, respectively, fnRBC and trophoblast selection.

The question has also been raised whether it would be possible to culture the fetal cells after isolation in order to increase the yield. Theoretically, CD34 + progenitor cells are still immature and have potential for growth and differentiation. Wachtel and colleagues set out to prepare RBC progenitor cultures, as they achieved a > 99% viability of the nucleated cell suspension obtained with their CFS method . After CFS enrichment of the blood samples, maternal RBC lysis and further lymphocyte depletion was performed. Thereafter the surviving cells were cultured in methylcellulose. The investigators succeeded in culturing several tens of clones in all samples, each consisting of about 100–200 cells. However, during validation, only a minority of the clones turned out to be Y-positive in male pregnancies, showing that the majority of the formed clones were of maternal origin. Also other attempts were unsuccessful .

Another interesting finding was the presence of apoptotic fetal cells in maternal plasma. When studying cells isolated from the plasma fraction after Percoll DGC, van Wijk et al. found between 1 fetal in 500 maternal cells to 1 in 2000 . These cells had both morphological features suggesting apoptosis and a positive terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) staining, indicative of ongoing apoptosis. Nevertheless, despite these cells being in an apoptotic state, XY FISH and PCR analysis still yielded successful results. Simultaneously, Sekizawa reported on the incidence of apoptosis among fnRBCs isolated from maternal blood samples . After DGC and subsequent fnRBC flow cytometric sorting based on HbF, the fetal cells obtained were assessed microscopically for potential positive TUNEL staining. Fourth 3% of the fnRBCs had undergone apoptotic change. This phenomenon can explain some of the lower numbers sometimes reported, and undoubtedly provides a source for fetal cfDNA ( Fig. 2 ).