Eur. J. Biochem. 213, 523-528 (1993) 0 FEBS 1993

Haemonectin, a granulocytic-cell-binding protein, is related to the plasma glycoprotein fetuin Hany WHITE', Nicholas TOTTY' and George PANAYOTOU' ' Department of Oncology, University College and Middlesex School of Medicine, London, England * Ludwig Institute for Cancer Research, London, England

(Received November 16, 1992/January 6, 1993) - EJB 92 1622

Haemonectin, a protein present in rabbit bone marrow extracellular matrix extracts, has been reported to bind granulocytes in a developmentally regulated manner. We have purified haemonectin from such extracts and determined the partial amino-acid sequence. The sequence obtained shows 60-70% similarity with the sequence of the plasma glycoprotein fetuin from other mammal species. This difference is consistent with the difference between fetuins from different species. We conclude that the rabbit haemonectin molecule is related to fetuin. The similarity between haemonectin and fetuin is reinforced by analysis with Western blots of one- and two-dimensional gels. These show that haemonectin, like fetuin, is present in serum and that migration of haemonectin from serum and extracellular matrix extracts, on two-dimensional gels, co-incides with that of human fetuin (a,HS-glycoprotein) from serum, extracellular matrix extracts and in purified form. Also, anti- haemonectin antibodies cross react with human fetuin. These data imolv that the rabbit haemonectin molecule is closely related to fetuin, but do no1 functionally distinct.

The orderly production of haemopoietic cells by the bone marrow relies on many complex interactions between de- veloping blood cells and the extracellular environment. Me- diating these interactions are growth-modulating molecules that can be soluble or bound to cells or the haemopoietic extracellular matrix (ECM), as well as adhesion molecules that may determine or stabilise these interactions. The central role played by the ECM in these processes has been well documented. It has been shown to bind both soluble growth factors [I , 21 and haemopoietic cells [3, 41.

With a view to elucidating the nature of adhesive interac- tions between bone marrow cells and the ECM, a 60-kDa protein, haemonectin, was purified from haemopoietically active ECM [5 , 61. This molecule has been shown to bind preferentially to the morphologically immature cell types of the granulocytic lineage [6, 71. Haemonectin has also been shown to bind specifically to granulocyte-macrophage colo- ny-forming-units and erythroid burst-forming units and to HL60 cells that have not been phenotypically differentiated by the addition of dimethylsulphoxide [7]. By immuno- staining and Western blotting, it has been shown that the expression or localisation of haemonectin corresponds to known areas of granulopoiesis, both in adult mice [6] and developing mouse embryos [ 81. Haemonectin has been further implicated in haemopoiesis by the observation that it is depleted in the bone marrow of SteeVSteel-Dickie hetero-

Correspondence to H. White, Department of Oncology, Univer- sity College and Middlesex School of Medicine, 91 Riding House St., London W1P 8BT, England

Fax: +44 71 436 2956. Abbreviation. ECM, extracellular matrix.

rule out the possi6iiity that these molecules are

zygous mice, deficient at the stem-cell factor locus (SINd mice) as determined by immunostaining, and that addition of it to stromal cell lines grown in vitro from such mice can restore some of the haemopoietic deficiency associated with this strain [9].

To characterise haemonectin and aid the molecular clon- ing of the haemonectin gene, we have purified the protein from rabbit ECM extracts and obtained a partial amino acid sequence from peptides generated by proteolytic digestion. The sequence data obtained suggest that haemonectin is re- lated to the plasma glycoprotein fetuin. We have sub- sequently used, therefore, one- and two-dimensional PAGE and Western blotting to investigate these observations further.

MATERIALS AND METHODS Materials

Leupeptin, soyabean trypsin inhibitor, RNase, DNase, guanidine . HC1, dithiothreitol and Tween-20 were purchased from Sigma Chemicals (Poole, England). Fresh rabbit hind- legs were obtained from Woldsway Foods (Lincolnshire, England). Lysyl endopeptidase was purchased from Wako Chemicals GmbH (News, Germany). Trifluoroacetic acid was purchased from Applied Biosystems (Foster City, Cali- fornia, USA). Acetonitrile was obtained from Rathburn Chemicals (Walkerburn, Scotland). Polyvinylidene difluoride membrane, Immobilon-P, was purchased from Millipore (UK) Ltd (Watford, England). Marvel low-fat skimmed milk powder was obtained from Premier Brands (UK) Ltd (Birmingham, England). Rabbit anti-(human a,HS-glyco-

524

protein) was purchased from Behring Diagnostics (UK) (Middlesex, England). Horse-radish-peroxidase conjugated swine anti-rabbit and rabbit anti-guinea pig antisera were ob- tained from Dako Ltd (High Wycombe, England). The guinea-pig anti-(rabbit haemonectin) antiserum was a gen- erous gift from Dr. Alan Campbell. The chemiluminescence detection system was obtained from Amersham International plc (Buckinghamshire, England).

fluoroacetic acid, 90% acetonitrile) was applied at a flow rate of 0.15 ml/min to elute bound material which was de- tected by a diode array spectrophotometer (Hewlett-Packard) and 0.15-ml samples were collected. Amino acid sequencing of selected peptides was done with an Applied Biosystems 477A protein sequencer with an on-line HPLC.

Polyacrylamide gel electrophoresis

Rabbit bone-marrow ECM preparation Rabbit bone-marrow ECM was prepared essentially ac-

cording to the methods of Campbell et al. [7]. Femoral bone marrow from 8-12-week-old New Zealand White rabbits was scraped into ice-cold homogenisation and washing buffer containing 3.4 M NaC1,50 mM Tris pH 7.4,l pg rn-l leupep- tin, 10pgml-' soyabean trypsin inhibitor. This was hom- ogenised with a Polytron homogeniser for 20 s at top speed. The homogenate was centrifuged for 30 min at 3000g at 4 ° C and the pellet was resuspended in the same buffer. This wash was then repeated several times, over 3 days, with the sample at 4°C. The pellet was then digested with RNase and DNase (100 pg d-' and 25 pg rn-' respectively) for 1 h at 37°C in Iscove's medium with protease inhibitors as above. The pel- let was then extracted with 4 M guanidine . HC1,2 mM dithi- othreitol, 50 mM Tris pH 7.4, with protease inhibitors as above for 24h at 4°C. After centrifugation at 3000g for 30 min at 4"C, the supernatant was dialysed exhaustively against HzO, 0.2 mM dithiothreitol. The volume of the sample was then reduced with poly(ethy1ene glycol) 20000, the sample removed and centrifuged at 15000g for 20 min. The supernatant was then removed and lyophilised.

Human bone-marrow ECM preparation

SDSPAGE was performed according to Laemmli [l l] , with gels run at constant temperature (15"C), always using 12% resolving gels with 5% stacking gels. The first-dimen- sion IEF for two-dimensional PAGE was performed as de- scribed [12], and the second-dimension SDS gels were run as for standard SDSPAGE.

Western blotting PAGE and two-dimensional PAGE gels were electroblot-

ted in 25 mM Tris, 192 mM glycine, 1 mM SDS, 10% meth- anol onto polyvinylidene difluoride membranes for 3 h at 2 mA cm-' at 15°C using the high-field option with a Bio- Rad Trans-blot cell. Membranes were then rinsed in NaCV P, (170 mM NaC1, 3.3 mM KC1, 1.8 mM KH2P04, 10 mM Na,HPO,, pH 7.2) and blocked with 5 % Marvel low-fat skimmed milk, 3 % bovine serum albumin, NaClP, for 2 h at room temperature. The membranes were then washed three times for 5 min in NaCVP, containing 0.05% Tween-20 (NaCVP,/Tween). Primary antibody incubations were done for 2 h in NaCVP,/Tween at room temperature. Rabbit anti- (human fetuin) polyclonal antiserum was used at a dilution of M O O 0 and polyclonal guinea-pig anti-(rabbit haemonectin) antiserum at a dilution of 1/2000. After antibody incubations, membranes were washed four times for 5 min in NaCW,I Tween at room temperature and incubated with a 1/2000 di-

Human femoral bone-marrow samples were obtained from material to be discarded after reaming of the top of the femur shaft to make space for the prosthesis during total hip- replacement operations. These samples were put in ho- mogenisation buffer as above and the ECM extracted in an identical manner. chemiluminescent detection system.

lution of horseradish-peroxidase-conjugated swine anti-rabbit or rabbit anti-guinea-pig second-layer antiserum for 1 h at room temperature. Membranes were then washed once for 15 min and three times for 5 min in NaCVP,/Tween and the signal detected using a horseradish-peroxidase-specific

RESULTS Purification and microsequencing of haemonectin Rabbit bone-mmow ECM extract was run On a 12%

SDSPAGE gel. The position of the haemonectin band in the Identification of haemonectin in rabbit ECM extracts gel was visialised, by negative staining in 4 M NaOAc for 20 min, with strong one-sided illumination. The protein band, and a similar-sized control gel slice from a blank area of the gel, were excised. The gel slices were placed in dialy- sis tubing and electroeluted in standard Laemmli gel running buffer: 25 mM Tris, 192 mM glycine, 0.1 % SDS. The gel slice was removed from the dialysis tubing and 0.2 pg lysyl endopeptidase was added. The tubing was incubated at room temperature for 1 h and then the sample was transferred from the tubing to a microfuge tube. The digestion was continued for 4 days at 30°C with addition of a further 0.2 pg lysyl endopeptidase on the first and third days. The digest was then dialysed exhaustively against 25 mM Tris pH 9.0, 0.1 % SDS. Peptides from the digest were applied to tandemly connected Brownlee columns of Aquapore AX-300 (30X2.1 mm) and Aquapore OD-300 (100X2.1 mm) (Ap- plied Biosystems) equilibrated in 0.1 % trifluoroacetic acid, 1% acetonitrile (buffer A) on a Hewlett-Packard HP 1090 HPLC system. A gradient of 0-60% buffer B (0.1% tri-

Haemonectin has been previously identified in ECM ex- tracts prepared from fresh rabbit bone marrow [6]. These extracts are made by selective solubilisation in 4 M guani- dine . HC1 of proteins remaining in a precipitate after extens- ive high-salt washing of a bone-marrow homogenate (see [7] and Materials and Methods).

To confirm that the rabbit haemonectin molecule was present in rabbit bone-marrow ECM extracts made in this laboratory, and migrated at the expected molecular mass on SDSPAGE, it was necessary to do a Western blot using guinea-pig anti-(rabbit haemonectin) (a generous gift from Dr. Alan Campbell). Fig. 1 shows that the anti-haemonectin antiserum detects a protein of the expected molecular mass [6] in rabbit ECM and that this corresponds to an easily vis- ible protein band on the Coomassie-stained SDSPAGE gel. This result corresponds to that previously published [6] and suggests that, with the available rabbit ECM, it should be possible to use SDSPAGE to purify a sufficient amount of haemonectin for amino-acid sequence determination.

525

a. Human fetuin Haemonectin Sheep fetuin 4

Fig. 1. Detection of haemonectin in rabbit bone-marrow ECM extracts. Lane ECM, Coomassie-blue-stained SDSE'AGE of rabbit bone-marrow ECM extract showing the presence of a protein band at a molecular mass of 58-60 kDa; lane anti-HN, Western blot of rabbit bone-manow ECM extract detected with 1/2000 guinea-pig anti-(rabbit haemonectin) polyclonal antiserum; lane GPS, duplicate Western blot detected with 112000 guinea-pig serum. Western blots were developed with horseradish-peroxidase-conjugated rabbit anti- guinea-pig antiserum and enhanced chemiluminescence. Figures at side refer to molecular masses of protein markers in m a .

Partial amino-acid sequence of haemonectin

A sample of rabbit bone-marrow ECM extract, contain- ing between 3-6 pg haemonectin (50-100 pmol), was sub- jected to electrophoresis and microsequencing as described in Materials and Methods.

The amino-acid sequence was determined for five pep- tides of length 8, 8, 11, 16 and 13 residues, providing a total of 56 residues of sequence. Haemonectin peptide sequences were then used to perform a Fast A search using the GCG software package on the NBRF protein sequence database. These sequences showed significant similarity to a sequence derived from the mammalian plasma glycoprotein fetuin [13]. The sequences of the rabbit haemonectin peptides and their alignment with the corresponding sequence of sheep and human fetuin is shown in Fig. 2a. The rabbit fetuin gene has not been cloned or sequenced. The amino-acid identity and the overall amino-acid similarity between haemonectin and sheep and human fetuin is shown in Fig. 2b. This has been calculated for the available 56 residues of sequence. The overall amino-acid similarity between sheep fetuin and rabbit haemonectin, 71 %, can be considered high.

It should be noted that of the 56 residues of sequence determined, 55 were unambiguously assigned. The last, resi- due 229 of the equivalent sheep fetuin sequence, present in peptide 220-235 has been assigned as X. In all mammalian fetuins sequenced this residue is Cys. Our data are consistent with this possibility, as under the conditions of amino-acid analysis used, Cys residues are not usually modified in a manner allowing positive identification.

Human fetuin HTLNQIDE Haemonectin $1 i r S ~ f AK Sheep fetuin 40 $TihQfl!)S 41

Human fetuin Haemonectin Sheep fetuin 103

Human fetuin &$GTYAYTCTyTQ'p$

Sheep fetuin 220 LGGEDVTVTCTLFQTQ 235 Haemonectin YW5yEEYTXTf FP??

Human fetuin PPDSHVLLAAPPG Haemonectin Sheep fetuin 211 S-WAGPSWAVP 280

b. Haemonectin

/ \ 55(71) 41(61)

/ Human - 48(68) - Sheep

Fetuin Fetuin Fig. 2. Amino-acid sequence of rabbit haemonectin peptides and similarity to fetuin sequences. (a) Peptide sequences. The rabbit haemonectin peptide sequences have been aligned with those of hu- man fetuin and sheep fetuin. The numbers at the sides refer to the positions of the first and last amino acid of each peptide within the sheep fetuin sequence. Bars indicate amino-acid identity, an asterix denotes a conservative amino-acid substitution. X denotes an unas- signed amino acid. (b) Relationship between haemonectin and fetu- ins. Amino-acid sequence similarity has been expressed as a percent- age for all the peptides. Figures in brackets refer to the amino-acid similarity when conservative amino-acid substitutions are included.

Haemonectin and fetuin in serum and ECM extract

If haemonectin is related to fetuin it might be present in serum. It was thought necessary, therefore, to check rabbit serum for material immunoreactive to the anti-haemonectin antibody. Similarly, detection of fetuin in bone-marrow ECM extracts would suggest a relationship between these mol- ecules. Due to availability of reagents, all experiments on fetuin have been done using human fetuin, cr,HS-glyco- protein [ 131.

Fig. 3 shows immunoblots of serum and bone-marrow ECM extract of both human and rabbit origin, and of purified human fetuin. These blots have been probed with anti- haemonectin and anti-(human fetuin) antisera. From these data (Fig. 3a) it is clear that there is an anti-haemonectin- reactive molecule present in rabbit serum that is similar in molecular mass to ECM-derived haemonectin. Fig. 3a also shows, importantly, that the anti-haemonectin anti-serum

526

Fig. 3. Western blots showing presence of fetuin and haemonec- tin species. Anti-haemonectin reactive protein. Rab ECM, rabbit bone-marrow ECM extract (approx. 1 pg); Rab Ser, 10 p1 1/500- diluted fresh normal rabbit serum; Hu FT, 2 pg human fetuin; anti- HN, blot detected with anti-haemonectin antiserum; GPS, guinea- pig serum control. (b) Anti-(human-fetuin) reactive protein. Hu FT, 50ng human fetuin, Hu ECM, human bone-marrow ECM extract (approx. 2 pg); Hu Ser, 10 pl 1/500-diluted fresh normal human serum; anti-FT, blot detected with MOO0 dilution of anti-human fetuin antiserum; RS, rabbit serum control. The conditions used were as for the previous blot. Figures on left refer to molecular masses in kDa.

cross reacts with human fetuin, and that these molecules ap- pear to have the same molecular masses on SDSFAGE.

Fig. 3b shows that the anti-(human fetuin) antiserum de- tects human fetuin in bone-marrow ECM extracts and adult human serum. The smaller faint bands visible in the human ECM track (Fig. 3 b) are due to degradation. A similar pattern can be generated by partial proteolysis of pure fetuin with trypsin (unpublished observation). Anti-(human fetuin) anti- serum was raised in rabbits and so does not detectably cross react with haemonectin (not shown).

Two-dimensional PAGE of haemonectin and fetuin Rabbit bone-marrow ECM extracts have previously been

resolved on two-dimensional gels and the haemonectin mol- ecule has been shown to be present on blots of these gels as a tight cluster of three or four isoforms that have a PI of about 4.5 [6]. The isoelectric point of the fetuin molecule from rat has also been determined as shown to be 4.5-5 [14, 151.

Fig. 4. Western blots of two-dimensional SDSPAGE. (a) 10 pl of 11500 dilution of fresh normal rabbit serum; (b) 10 p1 of 11500 di- lution of fresh normal human serum; (c) rabbit bone-marrow ECM extract (approx. 1 pg); (d) human bone-marrow ECM extract (ap- prox. 2 pg); (e) 3 pg purified human fetuin; (0 1 pg purified human fetuin. (a, c, e) Detected with anti-haemonectin (Anti-HN) anti- serum; (b, d, f) detected with anti-(human fetuin) (Anti-FT) anti- serum. Figures on top refer to the pHof the gel in the horizontal dimension. The top of each panel corresponds to the top of the resolving gel, the bottom to the dye front of the gel. Arrows indicate position of antisera-reactive protein, haemonectin or fetuin, on the blots of serum proteins. Both these molecules electrophoresed in the second dimension with an apparent molecular mass of 59 kDa. Molecular mass markers are not shown.

To confirm these observations and determine whether the serum and ECM-derived forms of both these proteins have related two-dimensional PAGE profiles, a series of immuno- blots was done on fetuin and haemonectin resolved by two- dimentional PAGE. Essentially the experiments in Fig. 3 were repeated as two-dimensional gels, and the immunoblots are shown in Fig. 4. From these data it can be seen that hu- man fetuin and haemonectin have a very similar PI and that this is the same for molecules derived from plasma and bone- marrow ECM extracts. The cross reaction of the guinea-pig anti-(rabbit haemonectin) antiserum can be seen to be to the appropriate molecule, as it reacts with the 59-kDa human fetuin protein at the appropriate PI (Fig. 4e). Another obser- vation is that the pattern of fetuin isoforms present in human bone-mmow ECM extract when resolved on this system

527

Sequenccd-t,d06 - - - - - II. radvn , I I I I I I I I 1

0 40 80 120 160 200 240 280 320 360 Fig. 5. Structural features of the human fetuin molecule. The scale refers to the number of amino-acid residues from the N-terminus of the secreted molecule. SP represents the 18-amino-acid signal peptide that is removed prior to secretion. Cystatin-like domains 1 and 2 are shown stippled darkly and lightly, respectively. The carboxyl-terminal domain is shown with diagonal stripes. Bars below the molecule represent the position of the squences in fetuin that are similar to the sequences of the haemonectin-derived peptides. C indicates the position of the cysteine residues; EF, the position of the EF-hand motif, a potential divalent-cation-binding site; Coll., the position of a three-times-repeated Gly-Xaa-Pro collagen-like motif; Connecting, the connecting peptide sequence thought to be cleaved off in secreted human plasma fetuin; 0-CHO and N-CHO symbolise the sites of 0-linked and N-linked glycosylation, respectively. The jagged arrows represent known or potential proteolytic cleavage sites. The C-terminal site is known to be cleaved in secreted human fetuin.

(Fig. 4d) appears very much as for the haemonectin molecule (Fig. 4c) and as previously published [6].

DISCUSSION We have shown, by deducing the partial amino-acid se-

quence and by analysis of one- and two-dimensional immu- noblots, that the rabbit bone-marrow ECM-derived granulo- cyte adhesion molecule haemonectin is structurally related to the plasma glycoprotein fetuin.

The haemonectin peptide fragments that have been se- quenced are well distributed over the length of the related fetuin molecule (Fig. 5), suggesting that the similarity be- tween the two molecules is contiguous, not being interrupted by a long non-identical tract. The sequence similarity be- tween rabbit haemonectin and human and sheep fetuin is comparable to that between these two fetuins over the se- quence determined. This suggests that the sequence diver- gence seen with haemonectin could be accounted for by species difference [ 161.

These observations, in conjuction with the similarity of the apparent molecular masses, PI and distribution of two- dimensional gel-resolved isoforms of rabbit bone-marrow haemonectin and human bone-marrow fetuin, suggest a close relationship between these two molecules. The detection of a haemonectin species in rabbit plasma with the same immu- nochemical properties as bone-marrow haemonectin and plasma and bone-marrow fetuin is consistent with this con- nection. The presence of fetuin in bone-marrow ECM ex- tracts, as judged by two-dimensional imrnunoblotting, shows that human fetuin and rabbit haemonectin can be obtained from the same tissue by the same extraction procedure.

The anti-haemonectin antiserum that we have used has previously been used in adhesion blockade [6,7], immunocy- tochemistry [6, 81 and immunoblotting [6] and to detect a single species (of four isoforms) on a two-dimensional gel of rabbit bone-marrow ECM [6]. In short, it has been used to define haemonectin. We have shown that it detects a haemo- nectin species in the same position on a two-dimensional gel of rabbit serum, probably rabbit fetuin, and it also detects human fetuin.

Haemonectin is, however, a protein that has been defined functionally as a binding molecule specific for immature cells of the granulocytic lineage. It is prepared by extraction in 4 M guanidine HCl of a protein precipitate present after high salt washing. This suggests that it has a low solubility or is attached strongly to other components of the bone-marrow ECM. Conversely, plasma-derived fetuin has a very high solubility. These observations suggest that these two mol-

ecules are not identical. We have tried to detect myeloid cell binding to commercially available human plasma fetuin but with negative results. This observation is consistent with fetuin and haemonectin, as they have so far been defined, being functionally distinct.

The differences between these molecules can be inter- preted in two ways: either the proteins are products of the same gene but are differentially modified or they are encoded by different but related genes. We show here that haemonec- tin is structurally related to the fetuin molecule, from other species, in a manner not inconsistent with it being the prod- uct of the same gene. It is possible, therefore, that haemonec- tin and fetuin exhibit differences due to alternative splicing of mRNA or post-translational processing. There is evidence showing extensive post-translational modification of fetuin, including glycosylation, proteolytic cleavage and phos- phorylation [17-191. Fetuin in humans is secreted by liver cells as a disulphide-linked heterodimer formed by the re- moval of the connecting peptide (see Fig. 5). In other species fetuin is present as a single-chain molecule in the plasma [16], although size heterogeneity is observed [16]. On this point, it is interesting to note that rabbit haemonectin con- tains a sequence similar to the connecting peptide, suggesting that this has not been removed by proteolysis (see Fig. 5, peptide 5 , 277-288).

If fetuin and haemonectin were encoded by the same gene, the soluble haemonectin-related protein detected in rabbit serum would be rabbit fetuin, as fetuin has previously been defined, and the bone-marrow form would be the ECM- localised, insoluble ‘haemonectin’ cell adhesion molecule.

If fetuin and haemonectin are encoded by different but related genes, our observations firmly place haemonectin as a member of the fetuidcystatin superfamily.

The observation that an abundant plasma glycoprotein is related to an adhesion molecule is intriguing. It has already been suggested that human fetuin is localised around haemo- poietic foci in human long-term bone-marrow cultures [20]. If haemonectin and fetuin are encoded by the same gene, it is important to determine whether the bone-marrow-localised form is selectively absorbed from the plasma and bound to the ECM or whether it is synthesised in situ by bone-marrow cells. The reported synthesis of fetuin by osteoblasts [14] supports the second hypothesis.

A summary of fetuin biology and molecular biology can be found elsewhere [13, 161. The function of fetuin is as yet unknown. The primary site of fetuin biosynthesis is thought to be the liver 1161 although, as mentioned, expression by osteoblasts has been reported [14]. These observations are consistent with the known sites of synthesis and localisation

528

of haemonectin. Haemonectin is found f i s t in the hepato- cytes of mouse embryos [8] but subsequently is found in an extracellular form predominantly in liver, bone-mmow, bone and skin [6, 81.

The reported observations suggest work needs to be done to whether fetuin and haemonectin are products Of the same gene, to define the molecular nature of the plasma and ECM-localised haemonectin and fetuin forms and to de- termine how these differences affect the function of the mol- ecule with respect to haemopoiesis.

extracellular matrix haematopoietic cytoadhesion molecule, Blood 75, 357.

9. Anklesaria, P., Greenberger, J. S., Fitzgerald, T. J., Sullenbarger, B., Wicha M. & Campbell, A. (1991) Haemonectin mediates adhesion of engrafted murine progenitors to a clonal bone- marrow stromal cell line from Sl/Sld mice, Blood 77, 1691.

10. Lebreton, J. P., Joisel, F., Raoult, J. P., Lannuzel, B., Rogez, J. P. & Humbert, G. J. (1973) Clin. Invest. 64, 1118.

ll. Laemmli, u. K. (1970) Nature 227, 680. 12. Dunbar, B. s., G ~ ~ ~ ~ , H, & l-immons, T. M. (1990) Rotein

analysis using high resolution two-dimensional polyacryl- This work was supported by the Leukaemia Research Fund, UK.

We would like to thank Dr. Alan Campbell for the generous gift of guinea-pig anti-(rabbit haemonectin) antisera; also Prof. Peter Be- verley and Katarzyna Dziegielewska for their support and helpful discussion, Professor Mike Waterfield for support, and Arnold San- derson and Woldsway Foods, UK, for the supply of rabbit hindlegs.

REFERENCES 1. Roberts, R., Gallagher, J., Spooncer, E., Allen, T. D., Bloom-

field, F. & Dexter T. M. (1988) Heparan sulphate bound growth factors: a mechanism for stromal cell mediated haemopoiesis, Nature 332, 376.

2. Gordon, M. Y., Riley, G. P., Watt, S. M. & Greaves, M. F. (1987) Compartmentalisation of a haemopoietic growth factor (GM- CSF) by glycosaminoglycans in the bone-marrow micro- environment, Nature 326, 403.

3. Tsai, S., Patel, V., Beaumont, E., Lodish, H. F., Nathan, D. G. & Sieff, C. A. (1987) Differential binding of erythroid and my- eloid progenitors to fibroblasts and fibronectin, Blood 69, 1587.

4. Spooncer, E., Gallagher, J. T., Krisza, F. & Dexter, T. M. (1983) Regulation of haemopoiesis in long term bone-marrow cul- tures N. Glycosaminoglycan synthesis and the stimulation of haemopoiesis by p-D-xylosides, J. Cell Biol 96, 514.

5. Campbell, A., Wicha, M. S. & Long, M. (1985) Extracellular matrix promotes the growth and differentiation of murine haematopoietic cells in vitro, J. Clin. Invest. 75, 2085.

6. Campbell, A. D., Long, M. W. & Wicha, M. S. (1987) Haemo- nectin, a bone-marrow adhesion protein specific for cells of granulocyte lineage, Nature 329, 744.

7. Campbell, A. D., Long, M. W. & Wicha, M. S. (1990) Develop- mental regulation of granulocytic cell binding to haemonectin, Blood 76, 1758.

8. Peters, C., O’Shea, K. S., Campbell, A. D., Wicha, M. S. & Long, M. W. (1990) Fetal expression of haemonectin: an

amide gel electrophoresis, Methods Enzymol. 182, 441. 13. Dziegielewska, K. M., Brown, W. M., Casey, S.-J., Christie, D.

L., Foreman, R. C., Hill, R. M. & Saunders, N. R. (1990) The complete cDNA and amino-acid sequence of bovine fetuin, J. Biol. Chem. 265, 4354,

14. Ohnishi, T., Arakaki, N., Nakamura, O., Hirono, S. & Daiku- hara, Y. (1991) Purification, characterisation and studies on biosynthesis of a 59-kDa bone sialic acid-containing protein (BSP) from rat mandible using a monoclonal antibody, J. Biol. Chem. 266, 14636.

15. Auberger, P., Falquerho, L., Olivier-Contreres, J., Pages, G., Le Cam, G., Rossi, B. & Le Cam, A. (1989) Characterisation of a natural inhibitor of the insulin receptor tyrosine kinase: cDNA cloning, purification and anti-mitogenic activity, Cell 58,631.

16. Brown, W. M., Dziegelewska, K. M., Saunders, N. R., Christie, D. L., Nawratil, P. & Muller-Esterl, W. (1992) The nucleotide and deduced amino-acid structures of sheep and pig fetuin. Common structural features of the mammalian fetuin family, Eur. J. Biochem. 205, 321.

17. Kellermann, J., Haupt, H., Auerswald, E.-A., Muller-Esterl, W. (1989) The arrangement of disulphide loops in human a,HS- glycoprotein, J. Biol. Chem. 264, 14121.

18. Lee, C. C., Bowman, B. H. & Yang, F. (1987) Human a,HS- glycoprotein: the A and B chains with a connecting sequence are encoded by a single mRNA transcript, Proc. Nut1 Acad. Sci. USA 84, 4403.

19. Rauth, G., Poschke, O., Fink, E., Eulitz, M., Tippmer, S . , Kel- lerer, M., Haring, H.-U., Nawratil, p., Haasemann, M., Jahn- en-Dechent, W. & Muller-Esterl, W. (1992) The nucleotide and partial amino-acid sequences of rat fetuin, Identity with the natural tyrosine kinase inhibitor of the rat insulin receptor, Eul: J. Biochem. 204, 523-529.

20. Long, M. W., Williams, J. L. & Mann, K. G. (1990) Expression of human bone-related proteins in the haematopoietic micro- environment, J. Clin. Invest. 86, 1387.