Clonal analysis in vitro of osteogenic differentiation of marrow CFU-F

M. E. OWEN, J. CAV£ and C. J. JOYNER

MRC Bone Research Laboratory, Nuffield Department of Orthopaedic Surgery, Xuffield Orthopaedic Centre, Oxford 0X3 7LD, England

Summary

Fibroblastic colonies, each of which is derivedfrom a single precursor cell (CFU-F), are formedwhen suspensions of marrow cells are cultured invitro. The ability of marrow CFU-F to differen-tiate in vitro was investigated using the ex-pression of alkaline phosphatase activity as amarker for osteogenic differentiation. In culturesof rabbit marrow cells the colonies formed var-ied in size, morphology and expression of en-zyme activity, indicating that marrow stromalCFU-F are a heterogeneous population. Growthand differentiation of marrow CFU-F can be

modified in vitro. Epidermal growth factorincreased average colony size and reduced clonalexpression of alkaline phosphatase activity tovery low levels. Hydrocortisone activated theosteogenic differentiation programme within thecellular progeny of a wide spectrum of CFU-F.The results support the possible development ofin vitro clonal methods for the study of differen-tiation and regulation of the osteogenic and otherfibroblastic cell lines of the marrow stromal sys-tem.

Key words: marrow CFU-F, clonal analysis, osteogencsis.

Introduction

The two main cell systems associated with bone andmarrow are the haemopoietic and stromal systems. Bycomparison with the haemopoietic system relativelylittle is known about cell lineage in the stromal system.According to current hypothesis the fibroblastic com-ponents of bone marrow stroma contain stromal stemcells that are able to generate several cell lines, includ-ing osteogenic, adipocytic and other fibroblastic lines(Owen, 1985). Whereas clonal methods for assay ofearly haemopoietic precursors have been used exten-sively in investigating cell lineage in haemopoiesis(Metcalf, 1984), they have scarcely been applied instudies of the stromal system. The present workdemonstrates the possibility of developing clonal analy-sis in vitro for the study of differentiation of marrowstromal precursors.

Fibroblastic colonies were formed when suspensionsof dispersed marrow cells were cultured in vitm(Friedenstein, 1973; Castro-Malaspina et al. 1980;Ashtone/a/. 1980, 1984). Friedenstein (1976) demon-strated that each colony is derived from a singlefibroblastic colony-forming cell (FCFC) or colony-forming unit-fibroblastic (CFU-F). Previous obser-vations suggest that marrow CFU-F may containputative stem cells of the stromal system. When

Journal of Cull Science 87, 731-738 (1987)Printed in Great Britain © The Company of Biologists Limited 1987

individual fibroblastic colonies were transplanted invivo under the renal capsule a proportion of themformed the mixture of stromal tissues that provides themicroenvironment necessary to support haemopoiesis,indicating pluripotency of the initiating CFU-F(Friedenstein, 1980). The fibrous and osteogenic con-nective tissues formed from marrow cells cultured indiffusion chambers in vivo are generated by a smallnumber of stromal precursors with stem cell character-istics (Bab et al. 1986). The sequence of events in thechambers resembles the normal developmental processfor osteogenesis (Mardon et al. 1987).

The influence of marrow stromal tissue on haemo-poiesis is well known (Trentin, 1976; Dexter, 1982).Variation in the number of marrow CFU-F in haemo-poietic diseases has been reported (Castro-Malaspina etal. 1982; Nagao et al. 1981; Haworth & Testa, 1983),but knowledge of the nature of CFU-F and of factorsregulating their differentiation is lacking. The purposeof the present study was to investigate whether CFU-Fare able to differentiate in vitro and if this can bemodified by environmental factors. Expression of alka-line phosphatase (AP) activity is used as a marker forosteogenic differentiation. Rabbit marrow cells arecultured and fibroblastic colonies are formed that varyin size, morphology and level of expression of theenzyme activity. Receptors for epidermal growth factor

731

and hydrocortisone have been demonstrated in osteo-genic cells (Ng et al. 1983; Rodan & Rodan, 1984).Addition of these factors to the culture medium modu-lates growth and expression of AP activity in thepresent system. The results support the feasibility ofdeveloping in vitro clonal assays for investigating celllineage and regulatory mechanisms in the marrowstromal system.

Materials and methods

In vitro culture

Ten New Zealand White male rabbits, 800g to 1-1 kg weight,were used in the experiments. Marrow cells were flushedfrom the mid shafts of the femur with a syringe containing0-5 ml of serum-free medium and a suspension of single cellswas prepared by drawing the cells into the syringe through19, 21 and 23 gauge needles in sequence. Any remaining cellaggregates were removed by passage of the cells through90 f.im nylon mesh (Bolting cloth, H. Simon Ltd, Stockport,UK). Samples of the cell suspension were diluted with 0 5 %(w/v) Trypan Blue prepared in 016M-ammonium chlorideand the number and viability of nucleated cells were countedin a haemocytometer. About 2x10 cells8 were obtained perfemur. The cells were plated out in 25 cm2 plastic tissueculture flasks (Nunc, Gibco Ltd) 107 cells/flask. Culturemedium was BGJh (Gibco Ltd) supplemented with 10%foetal calf serum (FCS) (Flow Labs, Irvine, UK), glutamine(lOmg/lOOml), vitamin C (5 mg/lOOml), penicillin (5000units/100ml) and streptomycin sulphate (5 mg/lOOml). Themedium was prepared 16— 24 h before plating out the cellsand before each change of medium. Incubation took place at37°C in a gassed incubator, 5 % CO2 in air. The medium wasfirst changed at 7 days and then every 3 or 4 days.

Experiments with epidermal growth factor (EGF) andhydrocortisone (HC)

EGF (rat submaxillary gland, Sigma Chemical Co. Ltd,Poole, UK) was dissolved in distilled water and HC (hemi-succinate salt, Sigma) was dissolved in absolute ethanol. Bothwere stored in samples until required, EGF at —20°C andHC in the refrigerator. EGF and HC were added eitherseparately or together to treated flasks at final concentrationsof 40ngml~' and 10~6M, respectively. Control flasks withoutadded factors were cultured under identical conditions.

Colony counting

All flasks were fixed with 95 % ethanol and stained foralkaline phosphatase (AP) activity (Sigma kit no. 85). A fewflasks were also stained with Mayer's haematoxylin and foracid phosphatase activity (Sigma kit no. 386A). Two controlflasks and two from each treatment were fixed and stained at 7days and examined under the microscope. The remainingcultures were stopped at 16 days and the total number ofcolonies formed and the number positive for AP activity werecounted by eye using an Anderman colony counter (Ander-man & Co. Ltd, Kingston-on-Thames, UK) consisting ofa light box and magnifying glass, 1-2X. Colonies wererecorded according to diameter in four size ranges:

10-2 mm, 2-4 mm, 4 to 6 mm and >6mm. Staining forAP activity was designated by eye in four categories (+ to+ + + +) according to intensity. Addition of vehicle to controlflasks had no effect on formation of colonies or expression ofenzyme activity. Counting of the flasks was done blind by twoindependent observers.

Statistical analysis of results

For each experimental schedule the results for several rabbitswere pooled. Comparison between treated flasks and controlswas performed using the Wilcoxon rank-sum test; ^ ^ 0 0 5was considered to be significant. All data are expressed asmean ± S.E.M.

Results

Total colonies formed and alkaline phosphatase-

positive colonies per number of cells inoculated

The total number of fibroblastic colonies formed andthe number positive for AP activity in control culturesat 16 days are directly proportional to the number ofnucleated cells inoculated. A typical result is shown inFig. 1. For 10 marrow cells per flask the colony-forming efficiency, CFE (total colonies per number ofcells inoculated), and percentage of colonies positivefor AP activity in control cultures ranged from

160

120

oU

40

0-5 1-0 1-5Cells inoculated per flask, x 107

2-0

Fig. 1. Total colonies ( • ) and number of AP-positivecolonies ( • ) per flask plotted against number of nucleatedmarrow cells inoculated (mean ± S.E.M.) for 5 flasks foreach inoculum.

732 M. E. Owen et al.

.. /,

*'...**

Fig. 2. Colonies in control culture fixed and stained forAP activity at 7 days. Colonies vary in morphology and sizeand are negative for the enzyme; Zeiss I CM 405microscope and M 35 camera. X16.

1-6X10"6 to 8-5XlCT6 and from 10 to 45%, respect-ively, for the ten rabbits.

Light-microscope observations

At the first change of medium at 7 days small fibro-blastic colonies are visible under the microscope. Theyvary in morphology and size, from a few cells to about ahundred cells (Fig. 2) and rarely contain cells that stainfor AP activity at this stage. Haemopoietic cells do notsurvive well in BGJb medium and die or are removedwith medium changes. At 16 days there was noevidence of acid-phosphatase-stained cells (macro-phages), and other haemopoietic cells in the cultures

and the background of cellular debris seen at 7 days haddisappeared.

In both control and treated cultures fixed at 16 days,there is a wide variation in colony size and in the levelof expression of AP activity (Fig. 3). Colonies arceither completely unstained for the enzyme or containvarying numbers of stained cells that predominate inthe centre or oldest part of the colony and appear tospread from there (Fig. 3). Two adjacent colonies froma control flask, one negative and the other positive forthe enzyme ( + + + ), are shown in Fig. 3A and anotherpositive at the + + level in Fig. 3B. A typical largecolony in an EGF-treated flask positive at the + level isshown in Fig. 3C. Fig. 3D,E illustrates colonies fromHC-treated cultures that differ in morphology and arcstrongly positive for the enzyme ( + + + + ).

Quantitative measurements of the effect of EG I'' andHC

Results for total colonies formed, the number express-ing AP activity, colony size and staining intensity aregiven in Figs 4, 5, 6, and Tables 1 and 2. EGF and HCwere added to the cultures either from day 7 whensmall fibroblastic colonies are already established orfrom day 0.

When EGF is added at day 7, the total number ofcolonies formed is about 70% of control, P=S 0-01, andthe number of AP-positive colonies is decreased toabout 12%, P=S 0-001 (Fig. 4). When HC is added atday 7 there is no effect on total colonies formed but thenumber of AP-positive colonies is increased by about50%, PsSO-01 (Fig. 4). When H C + EGF are addedtogether at day 7 there is no effect on total coloniesformed but the number of AP-positive colonies issignificantly reduced, P^O-01 (Fig. 4). Hence, EGFappears largely to override the effect of HC on enzymeexpression.

When EGF is present throughout the culture periodthere is a decrease of about 30% in total coloniesformed, P^0 -05 , and the number expressing APactivity is reduced to about 5 % of control, P^O'001(Fig. 5). When EGF is added at day 0 only, there is noeffect on total colonies and the number positive for APactivity is reduced to 50%, P^O-05 (Fig. 5). WhenEGF is added at day 0 and EGF + HC and HC fromday 7, there is no effect on total colonies, but there is adecrease in the number expressing the enzyme activitywhen EGF and HC are added together, PsSOOOl,which is intermediate between the effect of the continu-ous presence of EGF and control (Fig. 5).

When HC is present throughout the culture periodthere is a significant increase of about 30 % in totalcolonies formed, P ^ 0 - 0 1 , and in the number express-ing AP activity of about 70%, Ps£0-01 (Fig. 6). WhenHC is added at day 0, and HC + EGF and EGF fromday 7, the increase in total colonies formed is sustained

In vitro differentiation ofmartvw CFU-F 733

Fig. 3. A. Two adjacent colonies in control culture: colony on left completely negative for enzyme, colony on right positiveat + + + level. B. Colony in control culture positive at + + level. C. Colony in culture treated with EGF from day 0,positive at + level. D,E. Colonies in culture treated with HC from day 0: both positive at + + + + level. Note differentmorphologies: in D the majority of the cells right to the edge of the colony are well stained, in E colony is relatively smallwith cells intensely stained piled up in the centre. Flasks fixed and stained for AP activity (black) at 16 days. Zeiss SRstereomicroscope and MC 63 camera. X12.

734 M. E. Ozven et al.

but the addition of EGF significantly reduces thenumber positive for AP activity (Fig. 6).

Colony size and expression of AP activityTotal colonies and percentage positive for AP activityin the different colony size ranges for control andtreated cultures are shown in Table 1. The percentagepositive for AP activity is higher for larger than forsmaller colonies. In EGF-treated cultures the sizedistribution of colonies differs from that of controls andthe average colony size is greater. The number ofcolonies in the smallest size range is less than incontrols, P =S 0-01, and in the largest size range greater,

4. Da\

50-

40-

=^ 30

I 20 J

10-

ControlDay 7 "'-'"' + EGF

T+ HC +HC+EGF

T

(24) (21) (20) (21)

though significant only for EGF added from day 7,P^Q-05. In EGF-treated cultures the decrease incolonies expressing the enzyme occurs in all size ranges(Table 1). In HC-treated cultures the size distributionof colonies is similar to that in controls and the averagecolony size is comparable. When HC is present fromday 0 total colonies are increased over control and theincrease is in all size ranges (Table 1). In HC-treatedcultures the increase in the number of colonies express-ing the enzyme occurs in all size ranges but is relativelygreater for smaller colonies (Table 1).

The distribution of AP-positive colonies amongst thefour staining levels for the same cultures as in Table 1 isshown in Table 2. In both EGF- and HC-treatedcultures the effect on expression of AP activity occursat all staining levels. The largest number of coloniesexpress the enzyme at the lower levels. Colonies in thetwo higher staining categories are very few in number,2-5 % and 5-7 % of total colonies or 6-7 % and 11-4% ofAP-positive colonies, for control and HC-treated cul-tures, respectively, calculated from data in Tables 1and 2.

Discussion

5- DayO & i n - I r o , +EGFDay 7 - + EGF

50-

40-

I 30-

3 2 ( H

+ EGF + EGF+ EGF + HC

+ EGF+ HC

T

(18) (18) (8) (13) (13)

Day 0Day 7

70

60

50

1 40'

1 30-

Col

20'

1 0 •

0

P — + HC + HC-"" ' +HC

T

T

T

y

T

T

+ HC+ HC + EGF

f

+ HC+ EGF

•

(18) (17) (12)No. of flasks

(13) (11)

It has been shown previously that fibroblastic coloniesformed in cultures of marrow cells are each derived

Figs 4, 5, 6. Vertical axis; total colonies per flask (D) andAP-positive colonies per flask (•) (mean ± S.E.M.).4 0 n g m r ' EGF, 1(T6M-HC. Medium changes at 7, 10 and13 days. Flasks fixed and stained for AP activity at 16 days.Comparison with control: (•) Ps=0-05, ( • • ) Ps£0-01,( • • • ) P s ; 0-001; comparison with HC, (AAA) P«£ 0-001.

Fig. 4. Data for five rabbits; four or five flasks perrabbit in each control and treated group. From left toright, control flasks and three groups of treated flasks:EGF, HC and HC + EGF added at day 7 and atsubsequent medium changes.

Fig. 5. Data for four rabbits; two-five flasks per rabbitin each control and treated group. From left to right,control flasks and four groups of treated flasks: EGF addedat day 0, day 7 and at subsequent medium changes; EGFadded at day 0 and not at subsequent medium changes;EGF added at day 0, EGF + HC added at day 7 and atsubsequent medium changes; EGF added at day 0, HCadded at day 7 and at subsequent medium changes.

Fig. 6. Data for four rabbits, three-five flasks per rabbitin each control and treated group. From left to right,control flasks and four groups of treated flasks: IIC addedat day 0, day 7 and subsequent medium changes; HCadded at day 0 and not at subsequent medium changes;HC added at day 0, HC + EGF added at day 7 and atsubsequent medium changes; HC added at day 0, EGFadded at day 7 and at subsequent medium changes.

In vitro differentiation of marrow CFU-F 735

Table 1. Total number of colonies /flask (TC), mean ± S.E.M., percentage of colonies positive for AP activity(% AP) in each colony size range

Control

+ EGFday 7

+ HCday 7

Control

+ EGFday 0

Control

+ HCday 0

TC% APTC

% APTC

% AP

TC% APTC

% AP

TC% APTC

% AP

1-2 mm

18-8 ± 1-77-8

•••9-4 ± 1-10-9

18-7 ±2-720-5

15-4± 166-4

" 6 1 ±0-90-7

12-7 ± 1-39-9

15-6 ± 1-626-8

2—4 mm

18-5 ± 1-535-7

12-1 ±1-39-8

18-6 ± 1-560-9

15 2 ± 1928-9

9-6 ± 1-81-6

14-2 ± 1-940-3

•17-8 + 0-857-4

4—6 mm

10-2 ± 2-171-8

ll-9± 1-9190

12-3 ±2-166-0

13 5±2-671-6

11-1 ±2-913-6

9-9 ±2-779-5

16-1 ±2-478-9

>6mm

l-4±0-568-0

•2-8 ±0-729-5

1-7 ± 0-770-3

1-9 + 0-664-0

4-5 ± 1-44-3

1-3 + 0-654-3

2-3 ±0-862-1

Averagecolony size

(mm)t

3 0

3-6

3-1

3-3

4-0

3-2

3-3

n

24

21

20

18

18

18

17

Control cultures and cultures treated with EGF and HC from day 7 and day 0; first three groups (Fig. 4), first two groups (Figs 5, 6).n, no. of flasks. Comparison with control: • P«0-05 , • • P « 0 - 0 1 , • • • P*; 0-001.•[Calculated assuming average values of 1-5 mm, 3 mm, 5 mm and 7mm for each size range, respectively.

from a single cell, using thymidine labelling, chromo-some markers and time-lapse photography (Frieden-stein, 1976; Friedenstein et al. 1987). In the presentstudy the number of colonies formed is directly pro-portional to the number of cells inoculated, which isconsistent with clonally derived colonies.

Expression of alkaline phosphatase activity is widelyaccepted as a marker of osteogenesis in bone-formingsystems (Rodan & Rodan, 1984). The appearance ofthe enzyme just prior to mineralization in developingosteogenic tissue formed when rabbit marrow cellswere cultured in diffusion chambers in vivo (Ashton etal. 1980; Bab et al. 1984) is the justification for its use

Table 2. Number of AP-positive colonies jflask ineach staining category, mean ± S.E.M., for same

cultures as Table 1

Control+ EGF

day 7+ HC

day 7

Control+ EGF

day 0

Control+ HC

day 0

n, no.

+

10-5 ± 1-72-0 ±0-5

13-2±l-0

9-8 ±2-31-210-6

9-7 ±2-216-1 ±0-9

of flasks.

+ +

4-9 ±0-901 ± 0 1

8-3 ±0-9

5-3 ± 1-10-2 ±0-9

5-1 ± 1-18-8 + 2-1

+ + +

0-9 ±0-30-0

1-7 ± 0-5

1-1 ±0-40-0510-06

1-310-43-1 10-4

+ + + +

0-00-0

0-210-1

0-00-0510-06

0-00-410-1

n

2421

20

1818

1817

in the present study. Furthermore, the rabbit waschosen because marrow reticular cells in this species arealmost entirely negative for the enzyme in situ, whereasin rat, mouse and human these cells are positive,though less intensely than osteogenic cells (Ashton etal. 1985; Beckstead et al. 1981; Westen & Bainton,1979).

In the current hypothesis of lineage in the marrowstromal system, stem cells give rise to committedprogenitors and different cell lines (Fig. 7) (Owen,1985) the number and hierarchy of which have notbeen fully elucidated. Fibroblastic colonies are formedin culture from CFU-F and differentiation along one orseveral lineages may occur within the colonies. Clonalexpression of AP activity indicates differentiation in anosteogenic direction. Markers for differentiation alongfibroblastic and reticular lines are not available and

Stromalstem cells

Committedprogenitors

-CFU-F-

Fig. 7. Hypothetical diagram for lineage in the marrowstromal system. In analogy with the haemopoietic system itis proposed that: (1) stromal stem cells generateprogenitors committed to one or more cell lines; (2) thecells form a continuum where capability for self-renewaland multipotentiality decrease as lineage commitmentincreases; (3) CFU-F are components of the stem andprogenitor cell population.

736 M. E. Owen et al.

adipogenesis is rarely seen in cultures with FCS(Dexter, 1979), and this is also true in the presentexperiments.

The stem cell characteristics of CFU-F were firstsuggested by the work of Friedenstein (1976, 1980).Recently it has been shown that CFU-F have a highcapacity for self-renewal in culture and retain theirability, after many passages, for differentiation intodeveloping osteogenic tissues in diffusion chambers invivo (Friedenstein et al. 1987). The variation in size,morphology and level of expression of AP activityamong colonies in the present cultures suggests aheterogeneous stem and progenitor population for thestromal system similar to that found for the haemopoi-etic system (Metcalf, 1984). Much of the heterogeneitymay be explained if, as proposed, CFU-F are atdifferent stages in a tissue developmental system(Fig. 7). Furthermore, since only 2-5 % of colonies incontrol flasks are in the higher staining categories, i.e.approach committed progenitor status for osteogenesis,it seems likely that a large proportion of CFU-F in thepresent cultures are early stem and progenitor cells ofthe stromal system.

The main effect of EGF in the present experimentswas to increase average colony size concomitant withdepression of AP activity to negligible levels. This is inagreement with the stimulation of [ H]thymidine up-take and inhibition of the enzyme expression by EGFin an established osteogenic cell line (Kumegawae/ al.1983). The continuous presence of EGF is necessary inthe present system for its effect on AP expression(Fig. 5) and might suggest that the presence of thefactor keeps the cells in cycle, preventing them frommoving into the differentiation pathway. Stimulatoryeffects of HC on AP activity have been observed inmany osteogenic systems in vitro (Rodan & Rodan,1984; Tenenbaum & Heersche, 1985). In the presentsystem the number of colonies that express AP activityis increased by HC at all levels of expression of theenzyme (Table 2). These results indicate that osteo-genic differentiation is being activated within coloniesfrom a wide spectrum of CFU-F.

The present work demonstrates that differentiationoccurs in vitro within single colonies derived frommarrow CFU-F and that this can be modified byenvironmental factors. The current model of the mar-row stromal stem cell is based primarily on data from invivo assay of CFU-F (Friedenstein, 1980; Bab et al.1986; Friedenstein et al. 1987). The present resultssupport the model and encourage further developmentof in vitro clonal methods with the object of elucidatingthe lineage of the stromal system and identifyingmechanisms involved in commitment of stromal stemcells to a particular lineage and promotion of special-ized functional cell lines. However, production ofspecific markers for identification of the different

stromal cell lines is an urgent requirement for futureapplication of the methods.

References

ASHTON, B. A., ABDULLAH, F., CAVE, J., WILLIAMSON,

M., SYKES, B. C , COUCH, M. & POSER, J. W. (1985).

Characterization of cells with high alkaline phosphataseactivity derived from human bone and marrow:Preliminary assessment of their osteogenicity. Bone 6,313-319.

ASHTON, B. A., ALLEN, T. D., HOWLETT, C. R.,

EACLESOM, C. C , HATTORI, A. & OWEN, M. (1980).

Formation of bone and cartilage by marrow stromal cellsin diffusion chambers in vivo. Clin. Ortliop. ivl. Res.151, 294-307.

ASHTON, B. A., EAGLESOM, C. C , BAB, I. & OWEN, M. E.

(1984). Distribution of fibroblastic colony-forming cellsin rabbit bone marrow and assay of their osteogenicpotential by an in vivo diffusion chamber method. Vale.Tiss. Int. 36, 83-86.

BAB, I., ASHTON, B. A., GAZIT, D., MARX, G.,

WILLIAMSON, M. C. & OWEN, M. E. (1986). Kineticsand differentiation of marrow stromal cells in diffusionchambers in vivo.jf. Cell Sd. 84, 139-151.

BAB, I., ASHTON, B. A., SYFTESTAD, G. T. & OWEN, M.

E. (1984). Assessment of an /;/ I'ivo diffusion chambermethod as a quantitative assay for osteogenesis. Calc.Tiss. Int. 36, 77-82.

BECKSTEAD, J. H., HALVERSON, P. S., RIES, C. A. &

BAINTON, D. F. (1981). Enzyme histochemistry andimmunohistochemistry on biopsy specimens ofpathologic human bone marrow. Blood 57, 1088-1098.

CASTRO-MALASPINA, H., GAY, R. E., JHANWAR, S. C ,

HAMILTON, J. A., CHIARIERI, D. R., MEYERS, P. A.,

GAY, S. & MOORE, M. A. S. (1982). Characteristics ofbone marrow fibroblast colony-forming cells (CFU-F)and their progeny in patients with myeloproliferativedisorders. Blood 59, 1046-1054.

CASTRO-MALASPINA, H., GAY, R. E., RESNICK, G.,

KAPOOR, N., MEYERS, P., CHIARIERI, D., MCKENZIE,

S., BROXMEYER, H. E. & MOORE, M. A. S. (1980).

Characterization of human bone marrow fibroblastcolony-forming cells (CFU-F) and their progeny. Blood56, 289-301.

DEXTER, T. M. (1979). Cell interactions in vitro. In Clinicsin Haematology, vol. 8 (ed. L. G. Lajtha), pp. 453-468.Eastbourne, UK: W. B. Saunders.

DEXTER, T. M. (1982). Stromal cell associatedhaemopoiesis. InJf. cell. Physiol. Suppl. I, Proc. Sywp.cell, molec. Biol. of Henwpoietic Stem Cell Differentiationpp. 87-94.

FRIEDENSTEIN, A. J. (1973). Determined and inducibleosteogenic precursor cells. In Hard Tissue Givtvth,Repair and Remmeralisation, Ciba Fdn Symp., vol. 11,pp. 169-185. North-Holland, Amsterdam: Elsevier-Excerpta Medica.

FRIEDENSTEIN, A. J. (1976). Precursor cells ofmechanocytes. Int. Rev. Cytol. 47, 327-355.

In vitro differentiation of marrow CFU-F 737

FRIEDENSTEIN, A. J. (1980). Stromal mechanism of bonemarrow: cloning in vitro and retransplantation in vivo.In lmmunobiology of Bone Marroiv Transplantation (ed.S. Thienfelder), pp. 19-29. Berlin: Springer-Verlag.

FRIEDENSTEIN, A. J., CHAILAKHYAN, R. K. & GERASIMOV,U. V. (1987). Bone marrow osteogenic stem cells: Invitro cultivation and transplantation in diffusionchambers. Cell Tiss. Kinet. (in press).

HAWORTH, C. & TESTA, N. G. (1983). CFU-F in acuteleukaemia. Expl Hemat. 11 (suppl. 14), 47, abstract 86.

KUMEGAWA, M . , HlRAMATSU, M . , H A T A K E Y A M A , K . ,

YAJIMA, T., KODAMA, H., OSAKI, T. & KURISU, K.

(1983). Effects of epidermal growth factor on osteoblasticcells in vitro. Calc. Tiss. Int. 35, 542-548.

MARDON, H. J., BEE, J., VON DER MARK, K. & OWEN, M.

E. (1987). Development of osteogenic tissue in diffusionchambers from early precursor cells in bone marrow ofadult rats. Cell Tiss. Res. (in press).

METCALF, D. (1984). Hemopoietic Colony StimulatingFactors. Amsterdam, New York, London: Elsevier.

NAGAO, T., KOMATSUDA, M., YAMAUCHI, K. & ARIMORI,

S. (1981). Fibroblastic colonies in mbnolayer cultures ofhuman bone marrow. J. cell. Physiol. 108, 155-161.

NG, K. W., PARTRIDGE, N. C , NIALL, M. & MARTIN, T.

J. (1983). Epidermal growth factor receptors in clonallines of rat osteogenic sarcoma and in osteoblast-rich ratbone cells. Calc. Tiss. Int. 35, 298-303.

OWEN, M. E. (1985). Lineage of osteogenic cells and theirrelationship to the stromal system. In Bone and MineralResearch, vol. 3 (ed. W. A. Peck), pp. 1-25.Amsterdam, New York, Oxford: Elsevier.

RODAN, G. A. & RODAN, S. B. (1984). Expression of theosteoblastic phenotype. In Bone and Mineral Research,vol. 2 (ed. W. A. Peck), pp. 244-285. Amsterdam, New-York, Oxford: Elsevier.

TENENBAUM, H. C. & HEERSCHE, J. N. M. (1985).

Dexamethasone stimulates osteogenesis in chickperiosteum in vitro. Endocrinology' 117, 2211-2217.

TRENTIN, J. J. (1976). Hemopoietic inductivemicroenvironments. In Stem Cells (ed. A. B. Cairnie, P.K. Lala & D. G. Osmond), pp. 255-261. New York:Academic Press.

WESTEN, H. & BAINTON, D. F. (1979). Association ofalkaline phosphatase-positiVe reticulum cells in bonemarrow with granulocytic precursors. J. exp. Med. 150,919-937.

(Received 9 February 1987 -Accepted 1 April 1987)

738 M. E. Owen el al.