An in situ study of CD34+ cells in human fetal bone marrow

Authors


Dr Janet E Allen, H. H. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 1TL, UK. E-mail: janet.allen@bris.ac.uk

Abstract

The purpose of this study was to characterize the spatial distribution, number and size of CD34+ cells in fetal bone marrow. Thin sections of normal fetal bone marrow from lumbar vertebrae were stained using CD34 antibody QBend/10. Sections were used under light microscopy with various eyepiece graticules to make measurements of CD34+ cells in situ. Results showed that at mid- and late gestation, approximately 2% and 0·5% of fetal bone marrow cells were CD34+ respectively. The mean distance of CD34+ cells from the nearest trabecular bone surface was 61 ± 4 and 46 ± 4 μm, respectively, for mid- and late gestation. The mean distance to the nearest neighbour was 46 ± 5 and 105 ± 15 μm, and the mean distance to the nearest blood vessel was 13 ± 1 and 17 ± 2 μm respectively. The concentration of CD34+ cells in the peripheral region was 6·5 times greater than that at the centre of the sections. Overall, the percentage number of CD34+ cells decreased with gestational age. The cellular and nuclear diameters of CD34+ cells remained unchanged throughout mid- and late gestation at 5·4 ± 0·1 and 3·8 ± 0·1 μm respectively. This information will be used to calculate the natural background alpha-radiation dose to haemopoietic stem cells.

The motivation for this study came from the field of radiation research and protection, in which the calculation of radiation dose to bone marrow and its implications for leukaemia risk are of prime importance. The cells at risk of leukaemogenesis in the fetus are the haemopoietic stem cells (Korsmeyer et al, 1983; Foa et al, 1984). These cells originate in the yolk sac, colonizing other haemopoietic organs during fetal development and beginning in the bone marrow at around 14 weeks gestation (Custer, 1974). Several studies have inferred the in utero initiation of childhood acute lymphoblastic leukaemia (ALL) (Wasserman et al, 1992; Ford et al, 1993, 1997; Steenbergen et al, 1994; Gale et al, 1998). Using information from the Collaborative Group Study (Greaves et al, 1985), it is possible to calculate that up to 50% of all childhood leukaemias may be initiated in utero. A recent study by Weimels et al (1999) concludes that childhood ALL is almost always initiated before birth.

We are interested in the role of natural alpha-radiation in the initiation of childhood leukaemia. In fetal bone and marrow, the significant natural alpha-radionuclides are lead-210 and its alpha-emitting decay product polonium-210, although the levels present are very low (Bradley & Ewers, 1995; Henshaw et al, 1995). Lead-210 follows calcium into the fetal bone during ossification (Oyedepo & Henshaw, 1997; Purnell et al, 1999). Alpha-particles have a typical range of 37 μm when emitted from bone and travelling in marrow. We are interested in the frequency of alpha-particle hits to haemopoietic stem cells in utero in order to estimate the proportion of childhood ALL resulting from natural alpha-radiation. Consequently, we need to know the spatial distribution, size and number of these cells in normal fetal bone marrow. We may also be interested in the spatial distribution of stem cells because there is evidence, known as the bystander effect (Azzam et al, 1998; Mothersill & Seymour, 1998), in which a cell that is hit can induce DNA damage in neighbouring cells over a range of at least 150 μm (Lorimore et al, 1998; Prise et al, 1998).

Haemopoietic progenitor cells may be identified by the monoclonal antibody surface marker CD34. Using FACS (fluorescence activated cell sorting) techniques, Baum et al (1992) found that CD34+ cells constitute 2–10% of fetal bone marrow cells compared with 1% in the adult (Civin et al, 1984; Terstappen et al, 1991; Chen et al, 1994). Craig et al (1993) characterized Thy-1+CD34+ cells as functionally ‘the most primitive’ cells. Expression of Thy-1 (CD90) decreased with commitment and differentiation and 0·1–0·5% of total human fetal bone marrow cells were designated Thy-1+CD34+ (Baum et al, 1992).

The phenotype of the most primitive stem cells and the importance of CD34 as a stem cell marker has been challenged in recent years. Using a novel monoclonal antibody against murine CD34 (RAM34), Morel et al (1996) found a total of 4–17% of adult murine bone marrow cells expressed CD34 at intermediate to high levels. Of these CD34+ cells, 60% lacked the lineage (Lin) markers expressed on mature lymphoid or myeloid cells. In addition, among cells that were highly enriched in haemopoietic stem cells (designated Sca-1+Thy-1(10)Lin-/10 cells), 85% expressed intermediate, but not high, levels of CD34 antigen. The remainder of these phenotypically defined stem cells were CD34. Goodell et al (1997) also suggested the existence of a hitherto unrecognized population of murine haemopoietic stem cells that lack the CD34 surface marker. Stem cells with long-term engraftment capabilities were shown to be CD34 and CD34 expression reflected the activated state of haemopoietic stem cells and was reversible (Sato et al, 1999). This activation may represent self-renewal or proliferation linked to differentiation. More recently, a subset of murine pluripotent haemopoietic stem cells (PHSCs) with the phenotype LinSca+kit+CD38+CD34 has been isolated, which the authors believe fulfils the criteria for the ‘most primitive PHSCs’ (Zhao et al, 2000).

Secondary transplantation and limiting dilution studies have confirmed the presence of cells with long-term engraftment potential in the CD34 populations in human bone marrow cells (Zanjani et al, 1998). The CD34 fraction contained cells capable of differentiation into CD34+ progenitors and multiple lymphohaemopoietic lineages. However, it is still possible that the human expression of CD34 does not follow precisely that in the mouse. Goodell (1999) considered that the populations may not be interconvertible, or the CD34 population may not represent an appreciable natural reservoir for CD34+ cells.

There are numerous studies in the literature concerning the different characteristics of the CD34+ populations and subsets, the majority being performed on murine bone marrow cells. Of the few that elucidate human bone marrow cell populations, only one (Baum et al, 1992) has been found that enumerated CD34+ cells in human fetal bone marrow. A literature search has not revealed any more recent values for the percentage of primitive stem cells in human fetal bone marrow. In any case, as FACS techniques use cells in suspension, there have been no studies which examine the spatial distribution of CD34+ cells in situ in human fetal bone marrow.

In terms of this present study, identification of ‘the most primitive haemopoietic stem cell’ contained in the CD34 fraction would constitute the ideal situation. However, we were of necessity bound by current immunological capabilities with paraffin-thin sections. Many of the immunological reagents used with FACS techniques are not suitable for use with paraffin sections. We were therefore restricted to the use of CD34 monoclonal antibody to identify the CD34+ fraction, realizing that we were probably highlighting the activated haemopoietic stem cells and progenitor cells, and that this constituted the best estimate according to current technologies.

Materials and methods

Sample collection and preparation

Cases of fetal vertebra were obtained at autopsy according to Hospital Ethical Committee post-mortem regulations, covering an age range of 16–40 weeks gestation. Information pertaining to each case was obtained on age, sex, weight, crown–heel length, crown–rump length, cause of death and mother's age. From the total number of cases, six were selected to represent the age range, whose cause of death was thought not to have affected the haemopoietic marrow. Fetal bone samples were prepared using standard histological techniques. Processing chemicals, solvents and stains were supplied by Merck (BDH), monoclonal antibodies and associated biological solutions were supplied by Serotec, Oxford, and Dako, High Wycombe, UK. Autopsy samples were immediately fixed in fresh 4% paraformaldehyde solution. The samples were decalcified in EDTA solution for 2–3 months, testing for completion approximately every fortnight using X-ray techniques at the School of Veterinary Science, Bristol, UK. Samples were processed in ethanol and xylene, and embedded in paraffin wax. Thin sections were cut at 4 μm and mounted on APES (3-aminopropyltriethoxysilane)-coated glass slides.

Immunoperoxidase staining using QBend/10 monoclonal antibody (mAb)

The slides were dewaxed in xylene (3–5 min) and rehydrated through 90% and 70% ethanol to water. The slides were incubated in 3% aqueous hydrogen peroxide for 15 min to block endogenous peroxidase activity within the fetal bone marrow. The slides were rinsed by placing in phosphate-buffered saline (PBS) for 15 min. Normal rabbit serum was diluted with PBS (100 μl in 10 ml). The sections were incubated in antisera (QBend/10) diluted 1:50 with dilute normal rabbit serum for 1 h at room temperature in a humidified box. Negative controls were incubated simultaneously in diluted normal rabbit serum. After rinsing with PBS from a wash bottle, the sections were incubated for 30 min in biotinylated rabbit anti-mouse antibody (diluted 1:200 with PBS), rinsed in PBS and incubated in peroxidase-conjugated streptavidin (diluted 1:500 with PBS) for 30 min. DAB reagent (3′-3-diaminobenzidene hydrochloride) was prepared by dissolving a 1 × 10 mg tablet in 10 ml of distilled water and adding 10 μl of hydrogen peroxide (100 vol). The slides were incubated in the DAB solution for 10 min at room temperature and then rinsed in water. The sections were counterstained in methyl green for 1 h, rinsed in water, counterstained in 0·5% eosin for 1 min, dehydrated through graded ethanols to xylene and mounted in DPX mounting medium. The more intimate details of the laboratory procedure may be found in histological reference books such as Bancroft & Stevens (1990) and Beesley (1993). With this regimen, CD34+ cells were identified by their green-stained nucleus and brown-ringed surface. Granulocytes were identified by a green nucleus with relatively large pink-stained granular cytoplasm (Fig 1).

Figure 1.

 Illustration of the staining combination of CD34, methyl green and eosin. The arrow shows a CD34+ cell near to a bone surface. A blood vessel is lined with brown-stained endothelial cells. B, bone; BM, bone marrow; G, granulocyte; RBC, red blood corpuscles.

Sample selection

After preparation and staining as described above, two cases were selected to represent mid- and late gestation lumbar vertebrae at 24 and 35 weeks. Methods used by obstetricians to compare fetal age make use of different body measurements, i.e. crown–rump length (CRL), crown–heel length (CHL) and, to a lesser extent, body weight. In order to examine the likelihood of differences in the two samples being as a result of sample or age variability, we plotted 42 measurements of CRL, CHL, fetal weight and bone sample weight (two vertebra) against their gestational age (Fig 2). Between fetal age and measurements of length there was a strong positive linear correlation (R = 0·96, P < 0·001) and a positive exponential correlation between fetal age and measurements of weight, indicating that differences between the 24 week and 35 week sample will be strongly time variable.

Figure 2.

 Investigation of the extent to which differences between the sample at 24 weeks gestation and the sample at 35 weeks gestation might be time variable.

Sample shrinkage during processing

All measurements of fetal bone, bone marrow and CD34+ cells described below were made using these immunohistologically stained sections under light microscopy with a Leica DM RB microscope (Leica UK, Milton Keynes, UK) and various eyepiece graticules (Graticules, Tonbridge, Kent, UK).

Bone Four lumbar vertebrae were selected and measured in two dimensions using a micrometer with a vernier scale. They were first measured as wet tissue. The bone samples were measured again after 18–24 h in fixative solution; after 4, 8 and 12 weeks in EDTA decalcifying solution; after completion of decalcification; after processing in alcohol; after processing in xylene; and after processing in paraffin wax. Twelve measurements were made and the average reading taken in each case. After embedding in paraffin wax, the mean of 150 measurements of thin sections from progressive slicing through each vertebra, stained with haematoxylin and eosin, was taken as the final stage of processing. The percentage bone shrinkage was calculated from the measurements.

Cells Lumbar vertebrae from one case were immediately halved and the inside surface, i.e. the marrow, was gently blotted onto microscope slides coated in APES adhesive. Nucleated fresh marrow cells were visualized at random under the light microscope. The diameters of 150 lymphocyte-like cells were measured and the mean diameter calculated. The halved vertebrae samples were then processed and embedded as described above. After embedding in paraffin wax, the whole vertebra was progressively sliced into sections of 4 μm thickness. The sections were stained with haematoxylin and eosin and the mean diameter of 150 lymphocyte-like cells was taken as the final stage of processing. The percentage cell shrinkage was calculated from the before’ and ‘after’ mean cell diameters.

Spatial distribution of CD34+ cells in fetal bone marrow

For the measurement of spatial distribution, number and size of CD34+ cells, a ‘map’ of each thin section was made from photomicrographs. In this way, each trabecular space and CD34+ cell could be individually identified. The sections were arbitrarily divided into three concentric regions on the map (peripheral, mid- and central regions). As complete sections differ in size, owing to the structure and age of the vertebra, the compartments were divided in proportion according to the ratio 2:7:7 respectively (Fig 3). Equal numbers of trabecular spaces were selected from each region and the CD34+ cells were located within each space. To quantify the spatial distribution of CD34+ cells in the fetal bone marrow, three different distances were measured for each CD34+ cell using a calibrated eyepiece measuring graticule: (i) the distance to the nearest trabecular bone surface, in order to determine the percentage number of CD34+ cells within range of a possible hit from an alpha-particle emitted at a bone surface; (ii) the distance to the nearest blood vessel, as the CD34+ cells have a common origin and affinity with endothelial cells; and (iii) the distance to the nearest CD34+ neighbour. The latter measurement was chosen because of its possible relevance to the observed bystander effect.

Figure 3.

 ‘Map’ of a vertebral thin section to show the division of sections into compartments for microscopic analysis: 1. peripheral, 2. mid- and 3. central regions. As complete sections differ in size owing to structure and age of the vertebra, the compartments were divided in proportion according to the ratio 2:7:7 respectively.

Number of CD34+ cells in fetal bone marrow

The number of CD34+ cells in each selected trabecular space was determined using a ‘numbered indexed square grid’ eyepiece graticule. The area of each chosen trabecular space was estimated using the grid at 100× magnification to give the number of CD34+ cells per unit area. The number of CD34+ cells, granulocytes and mononuclear cells were counted in each square of the grid overlaying the marrow and totalled to give the number of CD34+ cells and number of granulocytes, expressed as number per 1000 mononuclear (MN) cells. The number of granulocytes was recorded as an index of the degree of differentiation which had taken place at 24 and 35 weeks gestation. This also gave a comparison of the degree of differentiation between the three different regions.

Diameter of CD34+ cells

The major and minor axes of each CD34+ cell and its nucleus were measured using the measuring eyepiece graticule at 1000× magnification. Cellular and nuclear diameters were taken as the mean of their major and minor axes, corrected for shrinkage. As the diameter of fetal CD34+ cells was expected to be approximately the same as the thickness of the thin sections (4 μm), any slicing through the cell would pass either through the diameter of the cell or very close to it. Consequently, any correction for mean true diameter of the cells was considered unnecessary.

Results

Sample shrinkage during processing

Bone The mean measurements at each stage of processing were used to calculate the percentage shrinkage values. The final shrinkage factor for fetal trabecular bone was 7·6 ± 0·2%. This value was used to apply a correction to all measurements of distance made in fetal bone sections, except cellular and nuclear diameters of the CD34+ cells.

Cells The mean cellular diameters of fresh and processed cells was 4·5 ± 0·1 and 3·8 ± 0·1 μm, respectively, giving a shrinkage factor of 14·9%. This value was used to correct the measurement of cellular and nuclear diameters of the CD34+ cells.

Spatial distribution of CD34+ cells

The distances from CD34+ cells to the nearest bone surface were plotted against number (Fig 4A and B). At 24 weeks gestation, the mean distance from the nearest bone surface was 61·0 ± 3·8 μm and 45·8 ± 4·1 μm at 35 weeks (Table I). The mean distances from CD34+ cells to a nearest CD34+ neighbour at 24 weeks was 46·4 ± 4·8 μm and at 35 weeks was 104·8 ± 14·5 μm. The comparatively large distances at the older age reflected the larger size and volume of the bone marrow spaces in the growing vertebra. Around 80–99% of CD34+ cells were within 150 μm of a nearest neighbour.

Figure 4.

 The distance from CD34+ cells to the nearest bone surface in fetal vertebra (A) at 24 weeks gestation and (B) at 35 weeks gestation.

Table I.   Spatial distribution of CD34+ cells in fetal bone marrow of lumbar vertebra.
 Gestational age
 Week 24Week 35
Distance to nearest bone surface, μm (mean)61·0 ± 3·845·8 ± 4·1
CD34+ cells within 37 μm range, (%)24·1 ± 2·542·6 ± 1·9
Distance to nearest neighbour, μm (mean)46·4 ± 4·8104·8 ± 14·5
CD34+ cells within 150 μm, (%)98·7 ± 3·479·6 ± 5·5
Distance to nearest blood vessel, μm (mean)12·8 ± 1·116·5 ± 2·1
CD34+ cells within 10 μm, (%)48·1 ± 0·438·9 ± 0·6

The mean distance to the nearest blood vessel was 12·8 ± 1·1 μm and 16·5 ± 2·1 μm at 24 and 35 weeks gestation respectively (Fig 5A and B; Table I). The most striking feature was that the mode was 2·2 μm in both cases. In the two-dimensional measurement, approximately 17% of CD34+ cells were immediately adjacent to the blood vessels, and around 40–48% were within 10 μm of the blood vessels. This has important implications for the spatial distribution of the CD34+ cells within a trabecular space because the distribution is then largely dependent on the distribution of the blood sinuses within the trabecular spaces. Figure 6 illustrates the effect of the position of the blood vessels on the distance of CD34+ cells from the bone surface.

Figure 5.

 The distance from a CD34+ cell to the nearest blood vessel in the bone marrow of fetal lumbar vertebra (A) at 24 weeks gestation and (B) at 35 weeks gestation.

Figure 6.

 Bone marrow in trabecular spaces of fetal lumbar vertebra, to show the effect of the position of blood vessels (lined with brown-stained endothelial cells) on the distance of CD34+ cells from a bone surface. Arrows indicate CD34+ cells. B, bone; BV, blood vessel; BM, bone marrow.

Number of CD34+ cells in fetal bone marrow

The mean numbers of CD34+ cells and granulocytes sited in different regions of bone marrow in fetal vertebra at 24 and 35 weeks gestation are shown in Table II. Results were based on either a count of CD34+ cells pooled from different trabecular spaces of measured area and expressed as number per unit area, or else based on a count of CD34+ cells or granulocytes pooled from different trabecular spaces of known total area and expressed as number per 1000 mononuclear cells in the same area. The values in parenthesis represent the smallest and largest number (per unit area or per 1000 MN cells) in an individual trabecular space. At 24 weeks gestation, the number of CD34+ cells per unit area decreased from peripheral to central regions of the section. The overall value was 13 CD34+ cells per unit area of 1 × 105μm2. At 35 weeks, the highest number of CD34+ cells was found in the peripheral region, so that the number per unit area in the central region was almost one half that in the peripheral region.

Table II.   The mean number of CD34+ cells and granulocytes sited in different regions of the bone marrow in fetal vertebra at 24 and 35 weeks gestation.
 Marrow regions 
 PeripheralMid-CentralOverall
  • *Based on a count of CD34+ cells pooled from different trabecular spaces of measured area.

  • Based on a count of granulocytes pooled from different trabecular spaces of measured area.

  • The values in parenthesis represent the lowest and highest value in an individual trabecular space.
    Total nucleated cells examined at 24 and 35 weeks gestation were 11 × 103 and 5 × 103, respectively.

CD34+ cells
(per unit area of 1 × 105μm2)
 24 weeks16 (9–22)15 (5–35)10 (4–36)13 (4–36)
 35 weeks9 (5–22)3 (1–6)5 (3–6)5 (1–22)
CD34+ cells
(per 1000 MN cells)*
 24 weeks32 (21–37)20 (6–52)15 (6–39)19 (6–52)
 35 weeks21 (12–38)3 (1–8)3 (3–7)5 (1–38)
Granulocytes
(per 1000 MN cells)
 24 weeks1 (0–2)10 (7–11)10 (6–11)8 (0–11)
 35 weeks8 (0–21)53 (30–85)69 (61–72)55 (0–85)

At both 24 and 35 weeks gestation, the concentration of CD34+ cells (number per 1000 MN cells) in the peripheral region of lumbar vertebra was greater than that in the central region (Fig 7A and Table II). From the overall figures, about 2% and 0·5%, respectively, of fetal bone marrow cells were CD34+. The concentration of CD34+ cells (per 1000 MN cells) in vertebra at 24 weeks gestation overall was 3·8 times greater than at 35 weeks. For granulocytes, their concentration increased from the peripheral to the central region. Overall, about 0·8% and 5·5% of cells were granulocytes at 24 and 35 weeks gestation respectively (Fig 7B and Table II). Within the sample at 35 weeks gestation, the marrow displayed a greater degree of differentiation in the central region than the peripheral, with a central/peripheral ratio of increase for granulocytes of 8·6.

Figure 7.

 The number of (A) CD34+ cells and (B) granulocytes per 1000 mononuclear cells in fetal bone marrow of lumbar vertebra. 1, peripheral; 2, mid-; 3, central. 24, 24 weeks gestation; 35, 35 weeks gestation.

Diameter of CD34+ cells

Cellular and nuclear diameter at both ages 24 and 35 weeks gestation are given in Table III. The mean cellular and nuclear diameter of the CD34+ cells did not change with increasing fetal age.

Table III.   Comparison of diameters of CD34+ cells at 24 and 35 weeks gestation.
 24 weeks gestation35 weeks gestation
Cellular diameter, μm  
 Mean5·35 ± 0·105·45 ± 0·13
 Range3·2–7·23·7–7·5
Nuclear diameter, μm  
 Mean3·80 ± 0·083·80 ± 0·10
 Range2·3–5·72·6–5·9

Discussion

Good quality stained sections of fetal bone marrow were obtained with using CD34 monoclonal antibody QBend/10 and counterstains methyl green and eosin.

The CD34 molecule, also found on endothelial cells, functions as an adhesion molecule which is important in the ‘homing’ of stem cells to the bone marrow (Fina et al, 1990; Healy et al, 1995). It is not surprising therefore that the CD34+ cells have been found closely associated with the blood vessels, with 48% and 39% within 10 μm of a blood vessel at 24 and 35 weeks gestation respectively (Table I). This arrangement has consequences for the spatial distribution of the CD34+ cells within the trabecular spaces and as the blood vessels were observed to lie mainly around the periphery of each space in fetal trabecular bone, the CD34+ cells were also taken in that direction. The number of CD34+ cells within alpha-particle range of a bone surface in fetal bone marrow in the vertebra was 24% and 43% at 24 and 35 weeks gestation, respectively, the higher value at the older age being as a result of the blood vessel distribution. The shape of the histogram revealed a dip in the measurements (Fig 4B) at 35 weeks gestation that did not disappear when the results were plotted with three different bin sizes. This arrangement of the blood vessels is probably linked to the physiology of fetal bone and marrow, supplying oxygen and nutrients and transporting away waste products for the very active osteoblasts at fetal trabecular bone surfaces. The majority of marrow spaces were observed to possess blood sinuses that curve in a fashion parallel to the bone surfaces and up to 20 μm away. Figure 6 illustrates the effect of the position of a blood vessel on the distance of CD34+ cells from a bone surface.

In a study of rat femoral bone marrow using frozen sections and double immunofluorescence techniques, Hermans et al (1989) found that the subendosteal area of the marrow was twice as rich in pre-B cells (primitive cells, target cells for cALL) as the central area. The authors also found that, within the subendosteal area, a positive gradient towards the bone was evident. Results from this present study in human fetal bone marrow from lumbar vertebra were similar to those of Hermans et al (1989); when individual trabecular spaces were disregarded, the bone marrow showed a positive gradient of CD34+ cells from the central region towards the periphery of the section (Fig 7A). However, in contrast to the femur, the periphery of the vertebra is composed of cartilage and not bone. At 24 weeks gestation, the concentration of CD34+ cells (per 1000 MN cells) in the peripheral region was almost twice that in the central region, while at 35 weeks gestation, the concentration of CD34+ cells in the peripheral region was approximately 6·5 times that in the central region. The percentage of CD34+ cells overall was 1·9% and 0·5% of the mononucleated cells in fetal bone marrow at gestational ages 24 and 35 weeks respectively. At 24 weeks, this approximates to the range 2–10% found by Baum et al (1992) in fetal bone marrow (at 18–24 weeks) using in vivo stem cell assays. For comparison, adult bone marrow contains ≈1% CD34+ cells (Terstappen et al, 1991; Chen et al, 1994).

Several authors have proposed that the increased concentration of progenitor cells [CD34+ or colony-forming units (CFU)-S] in the endosteal region of murine bone marrow compared with central regions is indicative of maturing cells moving towards a central sinus (Lord et al, 1975; Hermans et al, 1989; Lord, 1990). Together with the high proliferation rate found for CFU-S subendosteally, a radial developmental sequence has been suggested in which the early haemopoietic progenitors proliferate actively in the peripheral regions of the bone marrow, while their progeny differentiate and migrate simultaneously, before being discharged through the sinus walls into the bloodstream. Hermans et al (1989) concluded that pre-B cells must also move away from the peripheral bone marrow before differentiating to B cells that move to the bone marrow sinuses, otherwise there would be elevated levels of pre-B cells accompanied by elevated levels of B cells in the peripheral bone marrow.

The more recent work with mouse femur shaft (Cui et al, 1996) highlighted the fact that the progenitor cells (defined by the authors as the more primitive cells’) were more concentrated towards the bone surface, while the stem cells (defined as the most primitive cells’) were more concentrated around the central sinus. Bone trabeculae are absent in femur shaft, while the vertebra is almost entirely trabecular bone. If we take each trabecular space in the fetal vertebra individually, there are insufficient CD34+ cells per trabecular space to discover whether the CD34+ cells are more concentrated towards the trabecular bone surfaces. However, if we take the vertebral section as a whole, ignoring the presence of the trabeculae, this study has shown that CD34+ cells are more concentrated around the periphery of the vertebra, which appears to agree with Cui et al (1996) if CD34+ cells may be considered equivalent to the more primitive cells’. It has been postulated that the bone surface may be an important factor in the maintenance of the CFU-S population (Maloney & Patt, 1969a, b; Patt & Maloney, 1972). However, this is regarded here as improbable in fetal vertebra as the periphery of the whole section largely consists of cartilage and calcified cartilage.

A possible explanation for the decrease in number of CD34+ cells towards the central region may be owing to the architecture and mode of growth in three dimensions of the fetal vertebra. As the vertebra grows, minute arterioles deliver progenitor cells via the blood circulation to the peripheral growing region of the bone, giving a higher value in this area. A possible interpretation is that, once the CD34+ cells have been delivered to and housed in the peripheral marrow, their absolute number remains approximately the same. Then, as the vertebra increases in size and the number of marrow cells increases by proliferation and differentiation to fill the enlarging marrow spaces, the relative numbers of CD34+ cells decreases. In addition, as growth of the vertebra continues outwards, the stem cells that were originally deposited in the periphery are by now positioned in the mid- and central areas.

The number of granulocytes in fetal bone marrow was shown to increase both with gestational age and from the peripheral region through the mid- to the central region of vertebra as differentiation progressed, particularly at 35 weeks gestation (Table II). This is consistent with the findings of previous authors (Lord & Schofield, 1980; Lord, 1990), who demonstrated increasing differentiation of stem cells towards the centre of mouse femur shaft. By dividing the vertebra thin sections into peripheral, mid- and central regions and ignoring the presence of the trabeculae, a similar pattern of concentration of CD34+ cells has been demonstrated in human fetal vertebra. However, it is generally considered doubtful as to whether results for mouse femur may be compared with those for human fetal vertebra especially with regard to individual trabecular spaces. As each space possesses its own pattern of blood sinuses, as opposed to one central sinus, it differs in form and function from the pattern revealed for murine femur shaft.

The measured cellular and nuclear diameters of CD34+ cells in fetal bone marrow were 5·5 ± 0·1 μm and 3·8 ± 0·1 μm, respectively, at both 24 and 35 weeks gestation. This is in contrast to the mean nuclear diameter of stem cells (CD34+CD38 cells) in adult bone marrow which exhibited a trimodal distribution of 5·7, 11·6 and 14·8 μm (Utteridge et al, 1997).

Chromosomal instability has been demonstrated in the descendants of non-irradiated bone marrow stem cells at distances of at least 150 μm from irradiated cells (Lorimore et al, 1998). Killing of non-irradiated epithelial cells by culture medium transferred from irradiated cells has also been demonstrated (Mothersill & Seymour, 1998). This observation implied the release of a cytotoxic substance by the irradiated cells. Results also suggested that a signal transduction mechanism may control death or survival by the ‘bystander effect’ rather than the release of a factor which is directly cytotoxic (Azzam et al, 1998). The percentage of CD34+ cells within 150 μm of nearest CD34+ neighbour was 99% and 80% at 24 and 35 weeks gestation respectively. It would therefore be of interest to examine the possible existence of a bystander effect in human fetal bone marrow cell cultures.

It has been estimated that 14% of childhood ALL may be induced by natural alpha-radiation (COMARE, 1996). Estimates of partitioning of bone marrow dose between the natural alpha-emitters suggest that 9% and 5%, respectively, of incidence may be attributed to radon-222 and its short-lived decay products polonium-218 and polonium-214, and to polonium-210 (Simmonds et al, 1995; table 5·6). The majority of polonium-210 measured in autopsy bone samples derives from the radioactive decay of lead-210 at bone surfaces (Henshaw et al, 1988; Salmon et al, 1994). The alpha-radiation dose received by fetal bone marrow and by the haemopoietic stem cells from polonium-210 supported by lead-210 at bone surfaces on a cellular scale is currently unknown. This present study has provided important new information in characterizing and enumerating the CD34+ cells in fetal bone marrow in situ. These measurements are not only of interest in themselves, but are also important for use in Monte Carlo simulations of alpha-particle tracks in bone and marrow. With this information, we may calculate the alpha-radiation dose to bone marrow and to the haemopoietic stem cells in the fetus from natural alpha-emitters. This will be presented in a separate paper.

Acknowledgments

This work was supported overall by Medical Research Council Programme Grant number 8319972 with additional support by Department of Health Grant number RRX27.

Ancillary