PRENATAL DIAGNOSIS, VOL. 15: 9 13-9 19 (1 995)

ENRICHMENT OF FETAL NUCLEATED CELLS FROM MATERNAL BLOOD: MODEL TEST

SYSTEM USING CORD BLOOD KATHRYN ANDREWS*, JOHANNES WIENBERG*, MALCOLM A. FERGUSON-SMITH* AND DAVID c. RUBINSZTBIN~

*Department of Pathology, Cambridge University, Tennis Court Road, Cambridge, CB2 I QP, U. X; ?East Anglian Regional Genetics Service Molecular Genetics Laboratory, Addenbrooke’s NHS Trust, Hills Road, Cambridge,

CB2 2QQ, U.K.

Received 28 June 1995 Accepted 18 July 1995

SUMMARY The presence of small numbers of fetal nucleated red cells in the maternal circulation has been a stimulus for the

development of technologies for non-invasive prenatal genetic analysis. Our laboratory has been assessing the feasibility of density gradient centrifugation followed by magnetic activated cell sorting (MACS) of cells expressing CD32 and CD45, to deplete maternal nucleated blood cells. We have examined the efficiency of each of the steps of this procedure using cord blood from term pregnancies as a source of nucleated red blood cells. Cord blood was shown to contain highly variable numbers of nucleated red cells. Three different density gradients were examined. There was no major difference in the performances of the double and triple gradients. Density gradient centrifugation resulted in enrichments of nucleated red blood cells of about 1000-fold relative to the total cell count. However, it was apparent that the selection of the cell layers which were most enriched for these cells would result in significant losses of nucleated red cells in other layers. MACS sorting of cells using CD45 resulted in white cell depletions ranging from 7 to 34-fold. These data provide a foundation for comparison with other methods and for optimization of the MACS technique. KEY WORDS: nucleated erythroblast; prenatal diagnosis; magnetic activated cell sorting; non-invasive; maternal blood

INTRODUCTION

The presence of small numbers of fetal cells in the maternal circulation has been a stimulus for the development of technologies for non-invasive prenatal genetic analysis (Simpson and Elias, 1994). At least three fetal cell types have been detected in the maternal circulation: syncytiotro- phoblast, fetal lymphocytes, and fetal nucleated red cells. Efforts to use antibodies to syncytiotro- phoblast to isolate these cells have been frustrated by the tendency of maternal cells to adsorb fetal antigens onto their surfaces and fetal lymphocytes may persist from previous deliveries (Covone

Addressee for correspondence: David C. Rubinsztein, East Anglian Regional Genetics Service Molecular Genetics Lab- oratory, Addenbrooke’s NHS Trust, Hills Road, Cambridge, CB2 ZQQ, U.K.

0 1995 by John Wiley & Sons, Ltd. CCC 01 97-385 1/95/1009 13-07

et al., 1988). Accordingly, we have focused on attempting to isolate fetal nucleated erythroblasts (Zheng et al., 1993).

Our laboratory has been assessing the feasibility of the magnetic activated cell sorter (MACS) to deplete maternal nucleated cells. As a first step, density gradient centrifugation was used to enrich maternal blood for nucleated cells. The maternal cells were partially depleted from the nucleated cell fraction isolated from the density gradient by first using mouse antibodies to cell surface antigens present on mature leukocytes (CD45) and mature granulocytes (CD32) to mark maternal nucleated cells. These cells were selectively removed by incubating with rat anti-mouse antibodies coupled to magnetic beads, followed by separation on a magnetic column. Fetal nucleated red cells were ultimately identified using a mouse anti- fetal haemoglobin antibody coupled to alkaline

914 K. ANDREWS ET AL

phosphatase. This allowed cells carrying fetal haemoglobin to be identified with the alkaline phosphatase Vector Red substrate (Zheng et al., 1993).

Attempts to use this method in preliminary trials have been frustrated by very low yields of nu- cleated red cells or absence of such cells on many occasions. It was clear that we had to improve the efficiency of our technique before a trial could be started. In this paper we report the results of experiments that attempt to quantitate the enrich- ment that occurs at each step in the procedure. These provide a reference to allow us to modify and improve our current protocol.

MATERIALS AND METHODS

Cord blood samples from full-term deliveries were collected in lithium heparin tubes. Samples were stored at 4°C and processed within 72 h of delivery. Prior to use, samples were diluted I:3 in phosphate-buffered saline (PBS). We recorded cell counts of diluted cord blood both before and after separation on the gradients (see below for methods). Three density separation gradients were set up for each sample: a single density gradient (Histopaque-l077), a double density gradient (Histopaque-I109 and 1077), and a triple density gradient (Histopaque-1119, 1107, and 1077). After centrifugation, cells were collected from the gradi- ent interfaces, cytospin slides were prepared and stained with Hema Gurr rapid stain, and the relative proportions of cell types were identified. Cytospin slides of the diluted sample prior to gradient separation were also prepared, stained, and profiled.

Density gradients Single density gradient-Six millilitres of

Histopaque-1077 (Sigma, Poole, U.K.) was added to a 15 ml centrifuge tube and 6 ml of diluted cord blood was carefully overlayered using a Pasteur pipette. Tubes were centrifuged at 600 g for 30 min at room temperature. Cells from the interface layer were removed and washed in PBS.

Discontinuous double density gradient-A double density gradient was set up (Zheng et al., 1993) using 3 ml of Histopaque-I109 (3 parts Histopaque- 1119 and 1 part Histopaque-1077) in a 15 ml centrifuge tube carefully overlayered with 3 ml of

Histopaque-1077. Following this, 6 ml of diluted cord blood was overlayered using a Pasteur pipette and the tubes were centrifuged at 600 g for 30 min at room temperature. Cells from the upper inter- face and the 1109/1077 interface were removed, placed in separate tubes, and washed in PBS.

Discontinuous triple density gradient-A triple gradient was set up (Bhat et al., 1993) using 2.5 ml of Histopaque-I119 added to a 15 ml centrifuge tube, overlayered with 2.5 ml of Histopaque-I107 (2 parts Histopaque-I119 and 1 part Histopaque- 1083), and 2-5 ml of Histopaque-1077 added as a final layer. Finally 6 ml of diluted cord blood was carefully added using a Pasteur pipette and the tubes were centrifuged at 600 g for 30 min at room temperature. Cells were removed from each of the three resulting interfaces into separate tubes and washed in PBS.

Slide preparation Cell counts of diluted cord blood both before

and after separation on the gradients were recorded using an improved Neubauer haemo- cytometer. Slides were made using a cytocentrifuge (Shandon Southern, Runcorn, U.K.). The propor- tions of different cell types were scored following staining using Hema Gurr rapid stain (Merck Ltd, Poole, U.K.).

Cell counts The stained slides from the gradient interfaces

were examined under a light microscope ( x 100 magnification) and the cell types identified and classified into three categories (mature red cells, white cells, and nucleated red cells) according to their morphology. A minimum count of 500 cells were recorded for each gradient fraction.

Slides prepared from the whole cord blood were examined under the light microscope ( x 25 mag- nification) and the numbers of white cells and nucleated red cells for the entire cytospin area were recorded.

Magnetic cell sorting Cells isolated from cord blood by density gradi-

ent separation were washed and resuspended in PBS/BSA/EDTA/azide buffer (0.1 per cent bovine serum albumin, 0-01 per cent sodium azide, 5 m~ EDTA in PBS) to a concentration of lo6 cells per

METHOD TO ISOLATE FETAL NUCLEATED RED CELLS FROM THE MATERNAL CIRCULATION 915

Table 1-Cell types found in six cord blood samples (NRBCs=nucleated red blood cells; WBCs=white blood cells. See the Materials and Methods section for a description of how cell numbers were assessed)

NRBCs WBCs Red cells Sample (number) YO NRBCs (number) YO WBCs (number)

5 0.0005 516 0.05 lo6 14 0.0014 730 0.07 1 o6 5 0.001 208 0-04 5 x 105 4 00008 784 0.16 5 x 105 5 0~0001 797 0.16 5 x lo6 0 0 555 0.1 1 5 x 10’

1OOpl. After storing on ice for 30 min, the cells were incubated with mouse IgG2a anti-CD45 (25 pl per lo6 cells) (Eurogenetics, Teddington, U.K.) for 30min on ice. The cells were then washed, resuspended at a concentration of lo7 cells per 80pl of buffer, and incubated with mag- netic microbeads conjugated to rat anti-mouse IgG2a antibody (20pl per lo7 cells) (Eurogenetics, Teddington, U.K.) for 15 min at 8°C. The stained cells were washed, resuspended in buffer to 2 d , and immediately separated on the MACS column, using a type BS column and a flow rate of 1.5 ml/ min. The column was prewashed using PBS/BSA/ EDTA (0.1 per cent bovine serum albumin, 5 m~ EDTA) (column buffer). Cells from the negative and positive fractions were recovered and washed in column buffer, a cell count was taken, and cytospin slides were made. The slides were then stained with Hema Gurr rapid stain and the cell types counted and identified according to their morphology using the light microscope.

RESULTS

In order to investigate the efficiency of the steps of our published protocol (Zheng et al., 1993), we have had to revert to a model system where there are sufficient nucleated red blood cells to ensure a quantifiable yield at each stage of the procedure. We chose cord blood samples from term infants as a source of starting material. Cells were identified according to morphology, as we considered this to be more direct than alternatives such as quan- titative PCR (Hamada et al., 1993) or antibody labelling followed by flow cytometry (Bigbee and Grant, 1994).

How many nucleated red blood cells are there in term cord blood samples?

As a preliminary step in our investigations, we assessed the numbers of nucleated red cells, white blood cells, and non-nucleated red cells in cord blood samples from six infants (Table I). The percentage of nucleated red cells varied from 0-001 per cent to less than 0.0001 per cent. Nevertheless, as our later experiments show, this sample source was generally sufficient to allow some type of quantitation of the efficiency of our procedure.

Assessment of the density gradient centrifugation of cord blood samples

Six experiments are presented where the cell yields in different fractions are compared. Three different gradients were compared using the same starting cord blood sample in each experiment:

a single density gradient, where- cord blood diluted with PBS was overlaid on a 1107 density solution. The layer that was collected after centrifugation contains mononuclear cells and nucleated red blood cells (Bhat et al., 1990); a double density gradient, where cord blood diluted in PBS was overlaid onto a discon- tinuous 1109/1077 density gradient (Zheng et al., 1993). After centrifugation, the upper cell layer was expected to contain white cells and the lower cell layer was expected to contain nucleated red cells; a triple gradient, where cord blood diluted in PBS was overlaid onto a discontinuous 11 19/ 1107/1077 density gradient. After centrifu- gation, the upper cell layer was expected to

916 K. ANDRE WS ET AL.

contain mononuclear cells, the cell layer below was expected to contain nucleated red blood cells, and the third cell layer from the top was expected to contain granulocytes (Bhat et al., 1993). In all cases, the mature, non-nucleated erythrocytes were expected to pellet at the bottom of the tubes.

The results of each of the six experiments are shown in Table 11. Results for each experiment are presented, as the cell numbers collected in various fractions varied. (Presentation of summary results alone would be possibly misleading.) In all cases, the cell layers of each of the gradients contained a large proportion of mature red cells. However, these results suggest that a major enrichment of nucleated red cells (of the order of 1000-fold) is occurring at this step. We found the greatest proportion of nucleated red cells to white cells in the layer that was expected to contain the highest proportion of nucleated red cells (the lower layer of the double gradient and the middle layer of the triple gradient). However, there were significant numbers of these cells in other layers. These data highlight an important problem that arises when trying to enrich for nucleated red cells. A compro- mise often has to be made between choosing a relatively pure fraction where there are high pro- portions of nucleated red cells relative to white cells, and choosing a combination of fractions where the loss of nucleated red cells is minimized. In the context of these density gradients, it is clear that the selection of the fractions with the highest proportion of nucleated red cells to white cells will result in major losses of nucleated red cells, irre- spective of the gradient that is used. In practice, losses of nucleated red cells will have to be mini- mized at almost all costs, as these cells may only account for 1 in lo7 or lo8 cells in the maternal circulation (Simpson and Elias, 1994).

Preliminary results investigating the enrichment occurring using MACS columns

Earlier experiments performed in our laboratory showed that the degree of depletion of white cells that occurred on the MACS column varied from 18- to 180-fold (Zheng et al., 1993). We have confirmed that the performance of the MACS column is highly variable in our hands (see Table 111). The experiments that we have reported have only used CD45 to deplete maternal white cells, as our preliminary experiments suggest that the addition of CD32 does not seem to improve the

enrichment (results not shown). However, further experiments still have to be performed in order to validate this finding. Degrees of depletion were assessed by comparing the percentage of white cells applied to the column to the percentage harvested while the magnet was applied. These ranged from depletions of 7- to 34-fold. Our results suggest that no major losses of red cells were occurring on the column.

DISCUSSION

The paucity and variable numbers of fetal nucle- ated red cells in the maternal circulation (Simpson and Elias, 1994) present a number of challenges to investigators who are trying to isolate fetal cells for prenatal diagnosis. First, fetal cells should be isolated from other nucleated cells with sufficient purity to facilitate their location after they have been labelled with a fetal cell-specific marker. Second, the purification must not sacrifice a size- able proportion of the fetal cells, as this may deplete the already limited starting material to a degree that is unworkable.

Thus, we believe that the quantitative assess- ment of the limits of these techniques is a pre- requisite to their optimization. We have used cord blood samples from term infants as a model system as they contain comparatively more nucleated red cells than maternal blood. In this paper we report the efficiency of each of the purification steps in the method that we have been using.

Examination of a number of cord blood samples demonstrated that these also exhibit at least a ten-fold variation in their nucleated red cell con- centrations. This may on rare occasions result in apparent failures, when attempting to purify nucleated cells. An alternative option that has been used by Klinger and Bianchi (Klinger, 1994) is to use fetal livers as a source of nucleated red cells. It is clearly important to assess the starting concen- trations of nucleated red cells, irrespective of the model system that is used.

The density gradient centrifugation is an impor- tant step in the enrichment procedure. It is a step where the nucleated red cell concentration (as a proportion of total cells) is increased about 1000- fold. However, as our experiments show, there is a serious risk of incurring major losses of nucleated red cells if one only selects the fraction of the gradient that contains the highest proportion of these cells. This seems to be a problem with both

METHOD TO ISOLATE FETAL NUCLEATED RED CELLS FROM THE MATERNAL. CIRCULATION 917

Table 11-Six experiments to compare the yields in different cell layers of density gradients after centrifugation of cord blood. See the Materials and Methods section and the text for a description of the methods. Percentages of cells refer to the percentage of the particular cell type in the specific cell layer after centrifugation. Layers that are supposed to be particularly enriched for nucleated red cells are indicated with asterisks. (NRBCs =nucleated red blood cells; WBCs=white blood cells; RBCs=mature red blood cells. Means and standard deviations (in parenthe- ses) of the percentages of cells in each layer of the different gradients are indicated at the bottom of the table)

% RBCs Yo WBCS Yo NRBCS

Expt. 2

Expt. 3

Expt. 4

Expt. 5

Expt. 6

Means f SD

Expt. 1 Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

Cord blood Single Double: Upper

Lower* Triple: Upper

Middle* Lower

99.95 38.4 32.3 95.9 29.4 90.2 99.3 99.93 39.5 19.2 95.6 22.1 91.0 99-2 99.96 48.4 64.6 98.6 62.4 96.3 99.4 99.98 62.8 61.8 98.6 49.2 99.3 99.9 99.94 75.6 54.6 98.8 57.2 98.7 99.9 99.97 46.1 34.6 97.8 35.0 97.8 98.9 99.96 (0.02) 58.1 (14.6) 44.5 (18.4) 97.6 (1.4) 42.6 (16.1) 95-6 (4.0) 99.4 (0,4)

0.05 66.1 66.8 3.4

70.3 9.0 0.5 0.07

60.2 80.1 3.5

77.6 8.1 0-7 0.04

50.7 34.5

1 .o 36-4 3.1 0.6 002

36.9 37.8

1.3 50.4 0.6 0.1 0.06

23.9 45.2

1 .o 42-6

1.3 0.1 0.03

53.0 65.1 2.2

64.6 2.2 1.1 0.05 (0.02)

48.5 (15.6) 54.9 (18.3) 2.1 (1.2)

57.0 (16.3) 4.1 (3.6) 0.52 (0.38)

0.0005 0.4 0.8 0.6 0.3 0.8 0.2 0.0014 0.3 0.3 0.9 0.3 0.9 0.1 0.0004 0-9 0.9 0.4 1.2 0.6 0 0.0004 0.3 0.4 0.2 0 4 0.1 0 0.0003 0.5 0.2 0.2 0.2 0 0 0.0002 0-2 0.3 0 0.4 0 0 0.0005 (0.0004) 0.4 (0.25) 0.48 (0.29) 0.38 (0.33) 0.5 (0.4) 0.4 (0.4) 0.13 (0.29)

918 K. ANDREWS ET AL.

Table 111-Experiments to determine the efficiency of CD45 and h4ACS columns to deplete samples of white blood cells (WBCs). Cell counts were determined by estimating the volumes applied to and recovered from the column and multiplying by the cell concen- trations. Experiments 1 and 2 were performed on cord blood, while experiments 3 and 4 used blood from a non-pregnant female. The cells collected while the magnet was applied are indicated under ‘+magnet’

Before selection After selection

+magnet Experiment No. O h No. YO

~~

1 RBCs 1.3 x 107 44.7 1.1 x 107 944

NRBCs 4.8 x 105 1 -6 3.7 x 105 3.1 2 RBCs 3 x 107 59.4 1.9 x 10’ 91.3

WBCs 1.9 x lo7 37.6 1.1 x 106 5.5 NRBCs 1.5 x lo6 3.0 6.9 x 105 3.3

WBCs 2.4 x lo6 71.6 1.1 x 105 9.9 4 RBCs 2.8 x lo6 28-4 2.2 x lo6 97.9

WBCs 7.2 x lo6 71.6 4.7 x 104 2.1

WBCs 1.6 x lo7 53.7 2.5 x 105 2.1

3 RBCs 9.4 x lo5 28.4 1.0 x lo6 90.1

double and triple gradients, which performed simi- larly. These results suggest that a reasonable com- promise would be to pool both layers of the double gradient before proceeding to the MACS step. We do not consider that the apparent distribution of nucleated red blood cells in many of the cell layers of the gradient was purely due to technical prob- lems, as white blood cells were distributed accord- ing to expectations. This situation may have arisen if the nucleated red blood cells in cord blood were at different stages of maturity and had different specific gravities. Alternatively, it is possible that the use of mechanical devices to set up the gradi- ents and remove the desired layers may improve this situation.

Previously we showed that the MACS column can result in at least a ten-fold variation in the enrichment of nucleated red cells (Zheng et al., 1993). The great variation in enrichment is sup- ported by the more recent results that we have presented. This is certainly an area that we will attempt to improve by making modifications to the existing protocol. The MACS experiments were performed with columns that had been regenerated after use, by washing through with 100 per cent ethanol and drying. These may have contributed to the variable success of the procedure.

In principle, variable binding of the CD45 and goat anti-mouse antibodies may also affect the

success of the procedure. This aspect will be closely monitored in future experiments designed to improve the performance of the MACS columns.

In conclusion, we have reported the efficiency of each of the steps of the enrichment protocol that is being investigated in our laboratory. This is the first step towards a detailed investigation that will aim to optimize each of these procedures. We believe that it is important to report all related methods in quantitative terms. This will provide a reference for cross-comparison with other methods in the research phase and serve as a means of maintaining quality control on a procedure that is ultimately aimed at clinical use.

ACKNOWEDGEMENTS

We thank Drs Peter Milton and Gerald Hackett for helping to organize access to cord blood samples. This work was supported by Digital Scientific.

REFERENCES Bhat, N.M., Bieber, M.M., Teng, N.N.H. (1990). One

step separation of human fetal lymphocytes from nucleated red blood cells, J. Immunol. Methods, 131,

Bhat, N.M., Bieber, M.M., Teng, N.N.H. (1993). One-step enrichment of nucleated red blood cells-

147-149.

METHOD TO ISOLATE FETAL NUCLEATED RED CELLS FROM THE MATERNAL CIRCULATION 919

a potential application in perinatal diagnosis, J. Zmmunol. Methods, 158,277-280.

Bigbee, W.L., Grant, S.G. (1994). Use of allele-specific glycophorin A antibodies to enumerate fetal erythroid cells in maternal circulation, Ann. NY Acad. Sci., 731,

Covone, A.E.P., Kozma, R., Johnson, P.M., Latt, S.A., Adinolfi, M. (1988). Analysis of peripheral maternal blood samples for the presence of placental-derived cells using Y-specific probes and McAb H315, Prenat. Diugn., 8, 591-607.

Hamada, H., Arinami, T., Kubo, T., Hamaguchi, H., Iwasaki, H. (1993). Fetal nucleated cells in maternal

128-132.

peripheral blood: frequency and relationship to gestational age, Hum. Genet., 91, 427-432.

Klinger, K. (1994). FISH: sensitivity and specificity on sorted and unsorted cells, Ann. NY Acad. Sci., 731,48-56.

Simpson, J.L., Elias, S. (1994). Fetal cells in maternal blood-overview and historical perspective, Ann. N Y Acad. Sci., 731, 1-8.

Zheng, Y .-L., Carter, N.P., Price, C.M., Colman, S.M., Milton, P.J., Hackett, G.A., Greaves, M.F., Ferguson- Smith, M.A. (1993). Prenatal diagnosis from maternal blood: simultaneous immunophenotyping and FISH of fetal nucleated erythrocytes isolated by negative magnetic cell sorting, J. Med. Genet., 30, 1051-1056.