Purification of a Mammary-derived Growth Factor from Human Milk and Human Mammary Tumors ”

A growth factor, mammary-derived growth factor 1 (MDGFl), has been purified to apparent homogeneity from human milk. The factor is a pepsin-sensitive, reducing agent-insensitive protein with a molecular mass of 62 kDa and a PI of 4.8. An apparently identical factor has been isolated from human mammary tumors, suggesting that MDGF1 might be made by and act as an autocrine growth factor for mammary cells. High affinity receptors for MDGFl have been detected on mouse mammary cells, normal rat kidney cells, and A431 epidermoid cells (Kn = 2 X 10”’ M). MDGFl at picomolar levels stimulates the growth of mammary cells and greatly amplifies their production of collagen, apparently via elevating collagen mRNA levels, an effect that is demonstrated for normal rat kidney cells. The responsiveness of mammary cells to MDGFl is attenuated when the cells are grown on a basement membrane collagen substratum, a component of the extracellular matrix upon which these cells normally rest in vivo. MDGF1 thus may regulate the production of new basement membrane as mammary epithelium invades the stroma during proliferation.

A growth factor, mammary-derived growth factor 1 (MDGFl), has been purified to apparent homogeneity from human milk. The factor is a pepsin-sensitive, reducing agent-insensitive protein with a molecular mass of 62 kDa and a PI of 4.8. An apparently identical factor has been isolated from human mammary tumors, suggesting that MDGF1 might be made by and act as an autocrine growth factor for mammary cells. High affinity receptors for MDGFl have been detected on mouse mammary cells, normal rat kidney cells, and A431 epidermoid cells (Kn = 2 X 10"' M). MDGFl at picomolar levels stimulates the growth of mammary cells and greatly amplifies their production of collagen, apparently via elevating collagen mRNA levels, an effect that is demonstrated for normal rat kidney cells.

The responsiveness of mammary cells to MDGFl is attenuated when the cells are grown on a basement membrane collagen substratum, a component of the extracellular matrix upon which these cells normally rest in vivo. MDGF1 thus may regulate the production of new basement membrane as mammary epithelium invades the stroma during proliferation.
Basement membrane collagen (type IV collagen) synthesis appears to be important for the growth and/or survival of the epithelium of the normal rodent mammary gland and of welldifferentiated tumors derived from it (1)(2)(3)(4). In attempts to understand the factors regulating production of this protein, we discovered that mammary tumors that produce a basement membrane contain a growth factor that markedly and selectively amplifies the production of type IV collagen in cultures of mammary ducts and alveoli (5). Since poorly differentiated mammary tumors not producing a basement membrane lacked this growth factor, it was postulated that it might autoregulate basement membrane production. This mammary tumor-derived growth factor was highly purified and shown to be an acidic protein with a high molecular mass (68 kDa) ( 5 ) .
Additional evidence for this postulate has been obtained. We have detected a similar type of growth factor in human milk and in human mammary tumors, as detailed in the present report. The latter activity, which is named MDGF1,' * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
has been highly purified. It is a 62-kDa acidic protein which electrophoreses on gels as a single band under denaturing conditions. A survey of the properties of known growth factors in human milk indicates that MDGFl may be a new growth factor. It differs in size, PI, and disulfide bond reducing agent sensitivity from human epidermal growth factor (EGF) (6); it can be distinguished from the three growth factors in milk that Shing and Klagsbrun have described (7) on the basis of differences in size and/or pl and, on the basis of size, from a colony-stimulating factor that Sinha and Yunis (8) have detected in human milk.
MDGFl binds to high affinity membrane receptors which are present on normal mouse mammary epithelium and on two cell lines that have been tested, namely A431 human epidermoid and normal rat kidney cells, which are fibroblasts. All of these cells also differentially increase their production of collagen in response to MDGFl treatment, suggesting that the biological effects of the growth factor might be mediated through interaction with membrane receptors.

EXPERIMENTAL PROCEDURES
Cell Cultures-Normal mouse mammary ducts and alveoli were isolated from virgin female mice by collagenase digestion and differential filtration using the procedure of Wicha et al. (9) as modified by Kidwell et al. (10). The organoids were propagated in a serum-free medium (11) on tissue culture plastic dishes or on dishes coated with type IV or type I collagen (1). Cell number was quantitated following trypsin dissociation using a Biophysics particle counter (1). Normal rat kidney cells (NRK, clone F49) supplied from Dr. J. DeLarco, National Cancer Institute, were grown on tissue culture dishes in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (5).
Collagen Synthesis-The procedure follows that outlined in Ref. 5. Briefly, assays were as follows. Cell cultures were grown for 2-3 days in the presence of [U-"Clproline (Amersham Corp., 280 mCi/mmol, 2 pCi/ml) or [3,4-3H]lysine (Amersham Corp., 85 Ci/mmol, 2-5 pCi/ ml) with or without MDGF1. The growth medium was removed from the dishes, and the cells were scraped into 5 ml of 0.5 M Tris-HC1, pH 7.4, 0.11 M NaCl containing either unlabeled proline or lysine (1 mM). The cells were pelleted by centrifugation for 10 min at 1500 X g. After removing the supernatant, the pellets were resuspended in 50 mM Tris-HC1, pH 7.6, 1 mM proline or lysine and sonicated. Labeled protein was precipitated with 20% trichloric acid containing 1 mM proline or lysine. The precipitates recovered by centrifugation were resuspended in 5% trichloric acid and recentrifuged. This washing sequence was repeated three times. Precipitates were dissolved in 0.2 M NaOH. Aliquots were directly counted or analyzed for collagenase-sensitive radioactivity. In the latter case, 0.2 ml of solution was mixed with 0.1 ml of 1.0 Hepes buffer and 0.16 ml of 0.15 N HCI. Ten p1 of 25 mM CaCL and 0.02 ml of collagenase (Advanced Biofactures, form 111, 3200 units/ml of 0.05 M Tris-HC1, pH 7.6, 5 mM CaC12) were added. Samples were incubated for 90 min at 37 "C. After cooling, 0.5 ml of 10% trichloric acid, 5% tannic acid was added, and the acid-insoluble material was centrifuged down at 4000 X g. Pellets were resuspended in trichloric acid/tannic acid and recentrifuged. The combined supernatants (collagen) were counted. For control 5745 samples, the collagenase was replaced with collagenase buffer. In most cases, the differential stimulation of collagen synthesis by MDGFl is presented. This was calculated by dividing the per cent counts/min incorporated into collagen in MDGF1-supplemented cultures by the per cent counts/min incorporated into collagen in cultures without MDGF1.
Laminin Assays-Laminin accumulation in mammary epithelial cell was estimated by enzyme-linked immunoassay using monospecific antilaminin antibodies and purified laminin as a standard. Cells were cultivated for 3 days in serum-free medium in the presence or absence of MDGF1. The medium was removed and the cell layer was washed with phosphate-buffered saline containing Tween 80 (12). Fifty pl of 1:1,000 diluted rabbit anti-mouse laminin antiserum or prebleed serum were added to the dishes which were incubated for 2 h a t room temperature. Aliquots of this solution were added to microtiter wells coated with pure mouse laminin, and the wells were incubated a t 4 "C for 16 h. The wells were washed with phosphatebuffered saline/Tween 80, and peroxidase-conjugated goat anti-rabbit IgG antiserum (Cappel Laboratories, 1:20,000 diluted) was added. Three h later, the dishes were washed with the same buffer, and the amount of bound peroxidase was quantitated using o-phenylenediamine and Hz02 as substrates (12).
Crude MDGFl Preparation-Pooled milk (2 liters) from 5-10 donors was adjusted to pH 5.6 with acetic acid (13). The samples were centrifuged for 10 min at 15,000 X g a t 4 "C. The floating lipids were removed and the supernatant fluid was recovered and dialyzed overnight a t 4 "C against 1,000 volumes of distilled water. Precipitates that formed during dialysis were removed by centrifugation (15,000 X g for 10 min at 4 "C). Acid/ethanol extracts from pooled primary human breast tumors were prepared based on the technique of Roberts et al. (14). A complete description of the extraction procedure for the tumors is given in Refs. 5 and 14. The acid/ethanol-solubilized material was precipitated with ether, and the precipitate was recovered by filtration onto Whatman No. 1 paper filters. This material was extensively dialyzed against 1% acetic acid and then lyophilized to dryness. The dry material was solubilized in water (-3 mg/ml), and the insoluble residue was removed by centrifugation (5). The soluble fraction was used for further purification of collagen synthesis stimulating activity.
Isoelectric Focusing-A maximum of 200 mg of protein from milk or tumor extracts was adjusted to 10 mM isoelectric focusing buffer, pH 3-10 (Buffalyte, Pierce Chemical Co.). The sample was transferred onto a focusing column (LKB, 110 ml) in a 5-50% sucrose gradient and focused for 24 h a t 4 "C a t which point the terminal current had fallen to about 3 mA. Fractions (0.5-3 ml) were then recovered from the column and the pH was measured. Following dialysis against 500 volumes of 200 mM ammonium formate, pH 5.2, the fractions were lyophilized to dryness and resuspended in phosphate-buffered saline (0.15 M NaC1, 5.6 mM NazHP04, 1.06 mM KH2P04, pH 7.4) (PBS) and assayed for collagen synthesis stimulating activity using NRK cells. Active fractions from multiple runs were pooled and refocused before further purification. Where lyophilization was not performed, the samples were dialyzed against phosphate-buffered saline or water instead of ammonium formate buffer.
Gel Filtration High Performance Liquid Chromatography (HPLCJ-Approximately 1 mg of protein in 200 pl of PBS from the pooled isoelectric focusing fractions, PI 4.8, was injected onto a 7.5 X 30-cm TSK-3000 SW gel filtration HPLC column (Beckman). The proteins were eluted isocratically with the same buffer a t a flow rate of 0.5-1.0 ml/min a t 25 "C. One-ml fractions were collected, and 50-p1 aliquots were screened for their effects on collagen synthesis in NRK cell cultures. Multiple runs were performed, and the peak fractions containing the collagen synthesis stimulating activity were pooled and concentrated by ultrafiltration or lyophilization. These samples were adjusted to 1 M NaCl in PBS and re-chromatographed on the same column in 1 M NaCl/PBS. The peak fraction was then rechromatographed on the same column and again eluted with 1 M NaC1. Fractions of 1.0 ml were collected over the expected peak elution position and retested for collagen synthesis stimulating activity. These fractions, following reduction with @-mercaptoethanol(15), were subsequently analyzed on SDS-polyacrylamide slab gels (15). After this series of purification steps, a single protein band was visualized on the gels stained with a silver reagent (16). Biological activity in this band was directly assessed by eluting the protein (without fixation or staining) from the slab gels (5) and assaying for its effects on mammary epithelial cell growth and collagen synthesis.
Amino Acid Analysis-Twenty-five pg of the purified protein were hydrolyzed in constant boiling 6 N HCl for 18 h a t 110 "C in a nitrogen atmosphere. The sample was decolorized with charcoal and evaporated to dryness in vacuo. A human serum albumin standard (Calbiochem-Behring) was similarly hydrolyzed to verify the analytical procedure. Amino acids were separated by high performance liquid chromatography and post-column derivatized with ninhydrin for quantitation (17).
Radioreceptor Assays-The purified MDGFl was iodinated with lZ5I using a modification of the method of Thorell and Johansson (18). Approximately 1 mCi of Na lz5I (Amersham C o p ) was mixed with 5 pg of MDGFl in 10 pl of 0.4 M sodium acetate, pH 5.6. Five pl of lactoperoxidase (Sigma, 0.8 mg/ml) were added followed by 5 pl of Hz02 (30% solution diluted 1:25,000). The HzOz addition was repeated 15 s later, and after a second incubation, the reaction was terminated by adding a drop of saturated tyrosine (5 mg/ml). The fraction was then chromatographed on a 1.25 X 22-cm Sephadex G-25 column equilibrated in PBS containing 0.1% bovine serum albumin. The peak fraction was further purified by gel filtration HPLC (TSK-3000 SW) in PBS containing 1 M NaCl. Receptors were quantitated on cell monolayers or with membranes prepared from A431 human epidermoid carcinoma cells according to Kimball and Warren (19). In the latter case, aliquots of membranes containing 5 pg of membrane protein were incubated with 0.5-10 ng of '251-labeled MDGFl (specific activity 8 pCi/pg) f a 500-fold excess of unlabeled MDGFl in a total volume of 300 11 of 20 mM Hepes buffer, pH 7.4. Following a 30-min incubation at room temperature, the assay was terminated by adding 2 volumes of 25 mM Tris-HC1, pH 7.4, 10 mM MgCl,, 0.1% bovine serum albumin, and 25% polyethylene glycol in 20 mM Hepes buffer, pH 7.4. Tubes were incubated for 15 min a t 4 "C and then centrifuged a t 100,000 X g for 1 min. Pellets were washed twice with the above buffer, and the samples were counted. Triplicate assays were made on each sample. Scatchard analyses were performed as previously described (20).
Nick Translation and Blot Hybridization-Approximately 100 ng of al(R1) cDNA were incubated in 50 pl of buffer containing the following components: 2.5 p1 each of Tris-HC1, pH 7.4, 0.1 M dithiothreitol, 0.1 M MgClZ, and bovine serum albumin (nuclease-free, Bethesda Research Laboratories, 1 mg/ml); 5 p1 of DNase I (Boehringer Mannheim, 0.1 ng/pl); and 2 p1 each of the four deoxynucleotide [cu-32P]triphosphates (Amersham Corp., 410 Ci/mmol, 1 mCi/ml). After a 15-min incubation a t 37 "C, the sample was cooled on ice and 2 p1 of DNA polymerase I (Boehringer Mannheim, 5 unitslpl) were added. Incubation was performed for 45 min a t 15 "C, and the reaction was terminated by the addition of 50 pl of 0.2 M EDTA containing 100 pg of calf thymus DNA/ml. Unincorporated radioactivity was removed by dialysis against 10 mM NaCl, 10 mM Tris-HC1, pH 7.5, and 1 mM EDTA. Ten pg of salmon sperm DNA were added, and the DNA was boiled for 5 min and then quick-cooled for denaturation. Cytosols were prepared from cells grown with and without MDGFl using the procedure of White and Bancroft (21). Briefly, the procedure was as follows. Cell layers were trypsinized, an aliquot of cells was counted, and then a volume equivalent to lo6 cells was pelleted. The supernatant was discarded, and the cell pellet was resuspended in 45 pl of 10 mM Tris-HC1, pH 7.4,l mM EDTA at 4 "C. Five p1 of Nonidet P-40 (Shell Chemicals) were added, and lysis was performed for 5 min in the cold. An additional 5 pl of the detergent was added, and the suspension was centrifuged for 15 s at 15,000 rpm in a Beckman Microfuge. Fifty pl of the supernatant were mixed with 30 d of 3 M NaC1, 0.3 M trisodium citrate, pH 6.8. Twenty p1 of 37% formalin were added, and the sample was heated for 15 min a t 60 "C. Samples were stored at -70 "C until ready for assay. Aliquots of this solution were applied to nitrocellulose filters, and the filters were baked and prehybridized exactly as described (21). Hybridization with the nicktranslated probe was performed for 20 h under stringent conditions, and the filters were washed as described (22). Autoradiograms were developed for 24 h using RP1 film from Afga-Gevaert. The cul(R1) cDNA against type I collagen was kindly supplied by D. Rowe, University of Connecticut (22). In some cases, RNA blot hybridization was performed. In this case, RNA was prepared with guanidinium thiocyanate following published procedures (23).

RESULTS
Purification and Properties of MDGFl--To determine whether a collagen synthesis stimulating activity was present in human breast tumors and human milk, which is known to contain a variety of hormones and growth factors (24), sam-ples of delipidated, casein-free human milk or acid/ethanol extracts of tumors were subjected to isolectric focusing. The fractions obtained were analyzed for collagen synthesis stimulating activity using NRK cells (Fig. 1). In both preparations, there was a major peak of collagen synthesis stimulating activity which focused with a PI of 4.8. In the assay, the dialyzed, lyophilized fractions were tested for the ability to stimulate labeled amino acid incorporation into collagenase hydrolyzable protein in NRK cells. In this analytical assay (Fig. l ) , 15 ml of milk and 25 mg of tumor protein extract were applied to the isoelectric focusing column. No significant effect of the fractions on total, non-collagen protein labeling was observed, indicating a differential stimulation of net collagen synthesis by 8-and 6-fold with the tumor and milk preparations, respectively.
Preparative isoelectric focusing runs were then performed with human milk. Fractions with the appropriate pH (4.8) were combined for further purification. Milk was initially chosen as a preferred starting material owing to the large amounts of material which could be obtained. For each isoelectric focusing run, 100-200 mg of milk protein were routinely processed. Pooled fractions were dialyzed against ammonium formate and then lyophilized to dryness. Residues were dissolved in minimum volumes of 0.15 M NaCl and 0.01 M NaHP04, pH 7.4, and the samples were chromatographed on a preparative HPLC TSK-3000 SW gel filtration column in the same buffer. Aliquots of the column fractions were directly red Growth Factor 5747 tested for their effects on collagen synthesis using NRK cells. As shown in Fig. 2, a major and minor peak of collagen synthesis stimulating activity was detected with apparent molecular weights of -60,000 and less than 10,000, respectively. The high molecular weight activity represented less than 5% of the total protein applied to the column as estimated from the Azso profile of the fractions (Fig. 2). Radioreceptor assays of the fractions indicated that the small molecular weight species was human epidermal growth factor since there was potent competition with lZ5I-EGF for binding to A431 cell membranes in the presence of this fraction (data not shown). The high molecular weight species was further purified as described below.
Fractions from multiple HPLC runs exhibiting an elution time of 11.5-12.5 min (Fig. 2) were pooled, dialyzed against ammonium formate, lyophilized, and then re-chromatographed on the TSK-3000 SW column using phosphate-buffered saline containing 1 M NaCl as the elution buffer. The fractions eluting between 11.5 and 12.5 min from this step were re-chromatographed on the same column in the high salt buffer. One-ml fractions were collected between 11 and 15 min. These were dialyzed against water, lyophilized to dryness, resuspended in PBS, and assayed for collagen synthesis stimulating activity using NRK cells or analyzed by SDS-gel electrophoresis following reduction and denaturation. The results presented in Fig. 3 demonstrate that a single protein band of approximately M , = 62,000 was present in the fraction eluting from the TSK-3000 SW column at 12 min. All of the biological activity was confined to this fraction.
Purification of MDGFl was monitored by assessing the extent of differential stimulation collagen synthesis in NRK cells. Using this assay, an estimate of the purification was made. Optimal stimulation with crude decaseinated, delipidated milk required 250 pg of protein/ml of medium in the NRK cell cultures. After the isoelectric focusing step, this value was reduced to 35 pg/ml, while after the first HPLC TSK-3000 SW column run, the value was about 0.25 pg/ml. Following the final gel filtration step on the TSK-3000 SW column (1 M NaCl buffer), a maximal response was observed a t 10 ng of MDGFl/ml (Fig. 4), indicating an overall purifi- Tumor extracts or milk preparations were focused as described. Threeml fractions were recovered and dialyzed against PBS, and then 50pl aliquots were added to NRK cell cultures. Following a 3-day cultivation in the presence of 2 pCi/ml [I4C]proline, the cell layers were harvested. The presence of radioactivity converted from trichloroacetic acid-insoluble to an acid-soluble form by protease-free collagenase digestion was tested. 0, per cent total counts/minute in collagen; 0, pH. Top, human tumor extract; bottom, human milk. The arrows denote the focusing position of mouse EGF. Fractions were tested for effects on collagen synthesis ( g r n p h ) and analyzed by SDS-gel electrophoresis after reduction with Ii-mercaptoethanol. Inwt A , analysis of fractions 11-13 on 7.5% gels: inscf 13, annlysis of fraction 12, the biologically active fraction, on 15"; gels. The gels were fixed and stained with a silver reagent as described under "Experimental Procedures." cation was at least 25,000-fold. When cell counts were performed on NRK cells grown in absence or presence of purified milk MDGFl (25 ng/ml) for 3 days, no effect of the factor on NRK cell number was observed (8.4 X 10'' cells/dish in the absence or presence of MDGF1).
An activity present in human tumor extracts apparently identical to MDGFl was also purified using the methods outlined for the milk factor. The electrophoretic property of the purified preparation was compared to MDGFl obtained from human milk (Fig. 5). As indicated, a single silver stained hand (M, = 62,000) was present in both the tumor and milk samples when analyzed on 7.5% SDS slab gels under denaturing and reducing conditions. Starting with 5 liters of milk (15 g of protein), approximately 115 pg of MDGFl were recovered, while approximately 80 pg of the tumor factor were recovered from 100 g of the human tumor (wet weight). The specific activities of the milk-and tumor-derived factors were about equal in a collagen synthesis assay with NRK cells (not shown).
T o ascertain that biological activity was present in the protein band on the gel, 5 pg of MDGFl from human milk were electrophoresed along with a fluorescent serum albumin marker. Slices of gel just below the fluorescent marker (local- NRK cells were incubated with MI)(;FI at the indicated conrentrations, and the differential effects on collagen synthesis were determined as described in the legend of Fig. 2. ('ells were grown on tissue culture plastic dishes. ized by UV) were cut from the gel and eluted in PRS/hovine serum albumin. Control gel slice extracts were also prepared from the same gel regions. These fractions were analyzed for their effects on collagen synthesis in mammary cell cultures. The results presented in Table I demonstrate that biological  TABLE I Recovery of biological activity in the 62-kDa band following gel electrophoresis Five pg of MDGFl were electrophoresed on 7.5% gels. Gel areas correspondig to the MDGFl band or control gel lanes were cut out and solubilized in 6 ml of PBS/bovine serum albumin for 48 h at 4 "C. The solubilized material was recovered by centrifugation and filter-sterilized. Aliquots were added to cultures of mammary ducts and alveoli along with [3H]lysine (5 pCi/ml). Following 3 days of incubation, the amount of incorporation into total protein and into collagen was measured.  Chemical stability of MDGFl The factor was treated according to the protocols described and tested for its effects on collagen synthesis in NRK cell cultures. The -fold differential stimulation was calculated as described for Fig. 2. Twenty-five ng of MDGFl were added per ml of growth medium in each case. Effect of Disulfide Reduction and Heating-One hundred ng of MDGFl from milk were preincubated in the absence or presence of 40 mM P-mercaptoethanol for 2 h a t 25 "C in 1 ml of PBS, pH 7.4. MDGFl was then added to NRK cells, and the incorporation of [14C]proline into collagen was assessed after 3 days. The reducing agent treatment had no appreciable effect on the potency of MDGFl as shown in Table 11. There was a 340% differential stimulation of collagen synthesis using untreated MDGFl and a 300% differential increase with MDGFl treated with P-mercaptoethanol.
In a similar experiment, 100 ng of MDGFl in PBS were heated for 10 min a t 90 "C followed by bioassay. The factor was found to be essentially unaffected by this treatment as shown in Table 11.
Protease Sensitivity of MDGFl-To determine whether MDGFl was protease-sensitive, 100 ng of MDGFl from milk were digested with pepsin (100 ng) in 1% acetic acid for 1 h at 37 "C. Controls including pepsin only and growth factor without pepsin but incubated in acetic acid were also included. Following incubation, all the samples were neutralized and tested for their effects on collagen synthesis. As shown in Table 11, the biological activity of MDGFl was found to be completely destroyed by pepsin digestion.
Amino Acid Composition-The amino acid composition of milk MDGFl was determined on acid-hydrolyzed samples. As anticipated on the basis of PI, there was an excess of acidic over basic amino acids (Table 111).
Cell Membrane Receptors-To determine whether milk MDGFl might be acting via specific membrane receptors, purified milk MDGFl was iodinated and radioreceptor assays were performed on intact NRK and mouse mammary cells Amino acid composition of MDGFl Acid hydrolysates were prepared as described. A human serum albumin standard was prepared for assay in an identical manner, and corrections for losses were made using the known composition of the standard. The molar ratio was calculated from the number of picomoles of amino acid divided by the number of picomoles of histidine.  (11) or with human A431 epidermoid carcinoma cell membranes (19). Specific binding was detected in all cases with approximate saturation of binding with A431 cell membrane occurring at 10 ng/ml (Fig. 6, inset). A Scatchard analysis using the membrane preparation is depicted in Fig. 6. The affinity constant was calculated to be 2 x 10"O M with 1.33 pmol of milk MDGFl specifically bound per mg of membrane protein. MDGFl binding appears to be specific since the inclusion of 2.5 pg of EGF in the radioreceptor assay failed to inhibit the binding of lZ5I-MDGF1 (5 ng). Likewise, 2.5 pg of MDGFl did not affect the binding of '251-EGF (2.5 ng) to EGF receptors on these membranes (data not shown). The receptors for MDGFl on A431 cells are functional since these cells treated with MDGFl (50 ng/ml) for 3 days exhibited a 50% differential increase in net collagen synthesis (data not shown).
Mechanism of MDGFl Action-The effect of milk MDGFl on collagen production by NRK cells was due to an amplification of collagen synthesis rather than to a decrease in Growth Factor ]proline and then either they were harvested for measurements of amino acid incorporation into collagen or the medium was changed to medium free of ["C] proline, and incubation was continued for 24 h. Note the lack of effect of MDGFl on collagen turnover. There were 4 X lo5 cells present when pulse labeling was initiated. The data are expressed as counts of collagen/minute/dish. collagen turnover, as illustrated by pulse-chase studies shown in Fig. 7. Milk MDGFl (100 ng/nl) or control buffer was added to cells with a labeled precursor amino acid. After 15 min of labeling, either the cells were harvested or the medium was changed to remove the unincorporated amino acid, and the cells were grown an additional 24 h. The amount of labeled collagen associated with the cells was then measured using the collagenase assay method. Within 15 min, MDGFl differentially stimulated collagen labeling by approximately %fold. There was little difference in collagen turnover following a 24-h chase period whether MDGFl was present or not. The abundance of d ( 1 ) collagen mRNA-related sequences in NRK cells treated with and without the milk-derived growth factor was then assessed using cDNA probes against type I collagen, the major collagen species made by these cells.' Cytoblot analyses using the al(R1) probe indicated a difference of approximately 4-fold in the amount of collagen I mRNA-related sequences detected in cells after 2 days of culture with MDGFl (Fig. 8). A similar dot blot hybridization assay was also performed with 15 pg of total cellular RNA with comparable results (data not shown). Both assays were performed under stringent hybridization conditions (21).
Effect of MDGFl on Mammary Epithelial Cells--In contrast to NRK cells, mouse mammary duct and alveolar cells treated with milk MDGFl exhibited an increased rate of growth. MDGFl also enhanced the level of collagen production. The dose-response relationship between growth stimulation or collagen synthesis effects and MDGFl concentration is shown in Table IV. Optimal stimulation of both cell division and collagen synthesis was produced at or below a MDGFl concentration of 5 ng/ml.
The effect of MDGFl was tested on collagen synthesis in mouse mammary ductal and alveolar cells propagated on different substrata, a condition which has been previously demonstrated to influence the responsiveness of these cells to W. R. Kidwell,unpublished observations. FIG. 8. MGDFl effects on collagen mRNA levels in NRK cells. Cells were exposed to 0 or 10 ng of MDGFl/ml for 2 days.
Cytosols were prepared, and aliquots were tested for their ability to hybridize to collagen cDNA probes that were nick-translated with 32P-labeled nucleotides. A , cytosol from MDGF1-treated cells; B, cytosol from control cells. Left to right: lo4, 2.5 X lo4, 5 X lo4, and lo5 cell equivalents of cytosol applied to the nitrocellulose filters.
Hybridization probe: a l ( R l ) , 0.05 pg, 20 X lo6 dpm. other growth factors (11). Fig. 9 demonstrates that when cells were plated on type IV collagen-coated dishes there was no effect of MDGF1, whereas cells which had been plated on type I collagen-coated dishes or on tissue culture plastic dishes differentially increased their production of collagen by 3-and d-fold, respectively, in response to MDGF1. This difference in responsiveness could not be accounted for by a differential absorption of the MDGFl to any of the types of culture substratum as shown by incubating medium containing lZ5I-MDGFl on these dishes in the absence of cells. Less than 1% of the MDGFl was bound to any of the substrates when 20 ng of lZ5I-MDGF1 were added in 2 ml of medium and the dishes were incubated for 24 h at 37 "C. Laminin, a second basement membrane component, was also increased in relative amount/cell following MDGFl stimulation of the mammary cells. Using an enzyme-linked immunoassay, laminin was found to increase from 18.5 & 0.9 growth rate, and this was further enhanced by about 50% by MDGFl addition to the cultures. Thus, MDGFl acts synergistically with estrogen by some mechanism. The effect of estrogen may be indirect since it was not reproduced in the culture system.

DISCUSSION
A mammary-derived growth factor, MDGF1, has been purified from human milk and an apparently identical protein isolated from human mammary tumors. MDGFl appears to be distinct from other growth factors in milk. It differs from EGF, a known component of milk (26) as follows. MDGFl is stable to P-mercaptoethanol reduction, whereas EGF is not (6). MDGFl is about 10 times as large as EGF and differs by about 0.3 pH units in its isoelectric point (6). It differs from the major fibroblast growth-promoting factor of human milk, FIG. 9. Substratum-dependent responsiveness of mammary weight (62,000 versus -5,000-6,000). MDGFl also differs cells. Ducts and alveoli were isolated and cultivated in serum-free frorn HMGFI and HMGFII which this author has partially medium f MDGFl(50 ng/ml) for 3 days. The amount of ['4ClProline purified from milk (7). The former is about one-half the size incorporated into collagen and non-collagen protein was determined, 0 HMGFIII, described by Shing and Klagsbrun (7) in molecular and the relative responsiveness was calculated as described for Fig. Of  distinguishable from a bone marrow colony-stimulating factor found in milk. This factor has a molecular weight of 250,000 and a PI of 4.4-4.9 and is reducing agent-insensitive (8).
Preliminary studies with MDGFl have thus far failed to demonstrate any effects of MDGFl on bone marrow cell growth i n vitro.
A major biological response of mammary epithelium to MDGFl is a differential increase in collagen biosynthesis relative to total protein synthesis. Mammary cells produce only type IV collagen, the collagen type found in the basement membrane on which these cells rest in uiuo (27). There is considerable evidence the collagen production is tightly coupled to a mitogenic response of these cells (lo), and there are also strong indications that collagen IV biosynthesis is a requisite for mammary cell growth and survival both i n v i m and in vitro (1)(2)(3)(4).
Responsiveness of the mammary epithelium to MDGFl is + MGF conditional. The cells respond if plated on a plastic or a type + €2 FIG. 10. Estrogen effects on MDGFl responsiveness of mammary cells. Mammary epithelium was isolated from virgin female mice that had been ovariectomized. One group of animals was given an estrogen implant (20-mg pellet composed of estradiol-17P and cholesterol, 2:l) immediately after ovariectomy. Nine days after ovariectomy, the epithelium from both groups was isolated and cultured in serum-free medium for 3 days MDGFl (25 ng/ml). At the time of plating, there were 2.5 X lo5 cells/dish in 2 ml of serum-free medium.
ng/105 cells in the control cell cultures to 38.4 k 6.1 ng/105 cells in MDGFl (10 ng/ml)-treated cells after 3 days of culture. Whether or not the increased accumulation of laminin evoked by MDGFl results from enhanced biosynthesis or decreased breakdown has not been ascertained yet.
Estrogens are known to influence the growth of mammary epithelium i n vivo (25). Experiments were therefore conducted to determine whether the response of these cells to MDGFl might be influenced by estrogens. The growth of mammary ductal and alveolar cells in culture was found to be very low if the cells were isolated from animals depleted of estrogens by ovariectomy (Fig. 10). MDGFl had no significant effect on the growth of these cells. However, if the cells were derived from ovariectomized animals given replacement estrogen (estrogen pellet implant), there was about a 100% greater basal I collagen substratum but not on a type IV collagen substratum. We presume this indicates that cell shape differences are manifest on the various substrata and these affect signal transduction when MDGFl interacts with its receptors on the cell surface. A counterpart to this mechanism may exist in uiuo as the mammary epithelium invades through the basement membrane in response to a proliferative stimulus (2). In this new locus, the epithelium would be dislocated from the basement membrane and consequently be activated to respond to MDGFl by elaborating a new basement membrane between itself and the stroma. Such a process would be ratelimiting. It is also consistent with the process of generating a basement membrane i n vivo since this structure is formed largely at a stromal-epithelial interface and since nonproliferating epithelium which rests on a basement membrane appears to be synthesizing much less type IV collagen than proliferating epithelium (2). Cell shape changes have previously been shown to affect collagen production by smooth muscle cells in culture (28).
Four observations suggest that MDGFl may be physiologically relevant for the mammary gland. First, human mammary tumors contain a factor which is probably identical to the MDGFl from milk. This fact suggests that MDGFl is made by the epithelium. Second, this factor has been shown to stimulate the development of the glandular epithelium in whole mammary glands in ~u l t u r e .~ Third, MDGFl appears to act synergistically with estrogens since the growth factor does not stimulate mammary cells depleted of estrogen in vivo by ovariectomy. Fourth, a MDGF1-like activity has been detected in a variety of rodent mammary tumors, and its abundance is in rough proportion to the amount of basement membrane present in the tumors (5). However, since NRK cells, a fibroblastic cell line, also respond to MDGF1, it is clear that this factor is not selective for the mammary epithelium. Further experimentation will be required to determine the true target issue for MDGFl in uiuo.