Mammary derived growth inhibitor is not a distinct protein but a mix of heart-type and adipocyte-type fatty acid-binding protein.

The amino acid sequence of the mammary derived growth inhibitor (MDGI) from bovine mammary gland (Böhmer, F.-D., Kraft, R., Otto, A., Wernstedt, C., Hellman, U., Kurtz, A., Müller, T., Rohde, K., Etzold, G., Lehmann, W., Langen, P., Heldin, C.-H., and Grosse, R. (1987) J. Biol. Chem. 262, 15137-15143) revealed 95% identity to bovine heart fatty acid-binding protein (H-FABP), explaining the observed immunocross-reactivity. However, a cDNA encoding MDGI has not been found to date. Artificial MDGI cDNA was expressed in an in vitro transcription/translation assay. Analysis by isoelectric focusing of the immunoprecipitated in vitro translation products of lactating bovine mammary gland mRNA did not indicate a protein corresponding to the in vitro translation product of artificial MDGI mRNA. Moreover, two-dimensional electrophoresis of bovine mammary gland proteins confirmed the absence of a protein with the pI of the in vitro translated artificial MDGI mRNA in bovine mammary gland and instead revealed, apart from H-FABP, an unknown protein that was recognized by anti-H-FABP antibodies. From lactating bovine mammary gland the cDNA for adipocyte fatty acid-binding protein (A-FABP) was cloned. The in vitro translation of recombinant mRNA derived from this cDNA yielded a polypeptide that behaved like the unknown immunoreactive protein. Western blotting and immunofluorescence using monospecific antibodies demonstrated the coexistence of H-FABP and A-FABP in the lactating mammary gland. Taking into account that deviations of the MDGI sequence from the bovine H-FABP sequence correspond with A-FABP we attribute the structure originally reported as MDGI to a mix of these proteins.

A polypeptide designated as "mammary derived growth inhibitor" (MDGI) 1 had been purified from lactating bovine mam-mary gland using a proliferation assay with Ehrlich ascites carcinoma cells as target cells (1) and characterized by amino acid sequencing (2). The amino acid sequence of the 14.5-kDa MDGI revealed no homology to any of the hitherto known growth inhibitors (3); rather, MDGI belongs to a family of fatty acid-binding proteins (FABPs) capable of binding hydrophobic ligands with high affinity (4). The members of this family were named according to the tissue of their first isolation (i.e. heart, adipose, intestine, liver, myelin, epidermis, and brain). From numerous studies a role of FABP in intracellular transport and metabolism of fatty acids (5,6) as well as in differentiation (7) and signal transduction was inferred (4).
In addition to growth inhibition of Ehrlich ascites carcinoma cells and of mammary epithelial cell lines by MDGI (8,9), Yang et al. (10) recently reported that MDGI as well as H-FABP can act as growth inhibitors of primary mouse mammary epithelial cells. In mammary gland organ culture, growth inhibition was associated with functional differentiation in the presence of H-FABP or MDGI (10).
The literature abounds with partially conflicting data about presence and identity of FABP in the mammary gland. MDGI exhibits a striking 95% identity to bovine heart FABP (H-FABP) and on the basis of amino acid sequence differs from H-FABP in only seven positions (2,11). Correspondingly, both proteins show immunocross-reactivity (12). Screening of a cDNA library derived from bovine lactating mammary gland revealed a cDNA sequence (13) that was identical to the H-FABP cDNA sequence from bovine heart (14). Jones et al. (15) isolated two FABPs from a lactating rat mammary gland that differed in their isoelectric points (pI 4.8 and 4.9). The major form (pI 4.9) was partially sequenced, and the amino acid sequences obtained could be aligned with 67 residues of the rat H-FABP (16). An immunopurified FABP from lactating rat mammary gland showed no differences to H-FABP from rat heart on the basis of molecular masses of the protein and of tryptic peptides (17). The cDNA for the FABP from mammary gland of the pregnant mouse (Ref. 18; revised in GenBank TM U02883) was identical with the exon sequences of the H-FABP gene isolated from a mouse genomic library (19). Virgin mouse mammary gland expressed a 14.5-kDa protein with 100% identity to A-FABP over two stretches of 57 amino acids between positions 25 and 90 (20). From pregnant and lactating mouse mammary gland, however, a protein was isolated and sequenced (20) that over a 98-amino acid stretch was only 97% identical to the H-FABP amino acid sequence deduced from the cDNA published by Tweedie and Edwards (21).
In our laboratory numerous attempts to clone the MDGI cDNA from bovine mammary gland failed, and doubts arose about the true nature of the structure originally published by Böhmer et al. (2). In this report we show that H-FABP and A-FABP are simultaneously expressed in lactating bovine mammary gland, and on the basis of this observation we provide evidence that the published structure can be traced back to a mixture of H-FABP and contaminating A-FABP and propose that the MDGI function is exerted by the well characterized H-FABP.
Isolation of mRNA and cDNA Synthesis-Total RNA was isolated either from lactating bovine mammary gland tissue by the guanidine thiocyanate/cesium chloride method as described by Chirgwin et al. (22) or from subcutaneous adipose tissue according to Chomczynski and Sacchi (23). To obtain poly(A)-rich RNA, total RNA was passed through oligo(dT) spin columns according to the protocol supplied by the manufacturer.
Rapid Amplification of cDNA Ends (RACE-PCR)-For rapid amplification of cDNA ends of bovine A-FABP the methodology of Frohman et al. (24) was employed. Primers for RACE-PCR were synthesized using the Gene Assembler Plus (Pharmacia).
3Ј-End Amplification of cDNAs-The amplification was performed using the specific primer (5 Ј -TCGGACAGTCGACATCGGT-3Ј; 23 pmol) identical to the 5Ј-end of the universal primer from the first-strand synthesis of cDNA and the degenerated gene-specific sense primer (5Ј-GGNACCTGGAAGCTNGTNTC-3Ј; 80 pmol). The PCR mixture (10 l of cDNA, 0.2 mM each dNTP, 10 l of 10 ϫ PCR buffer, 23 pmol of specific primer, and 80 pmol of degenerated gene-specific sense primer in a total volume of 100 l, overlaid with mineral oil) was denatured for 5 min at 94°C and cooled to 72°C. After the addition of 2.5 units of Taq polymerase, PCR was performed for 30 cycles (1 min at 58°C, 1 min at 72°C, 1 min at 94°C) followed by a final extension at 72°C for 10 min using a TRIO-Thermalblock (Biometra, Göttingen, Germany).
5Ј-End Amplification of cDNAs-For reverse transcription, 0.7 g of poly(A)-rich mammmary gland RNA and 1.2 pmol of gene-specific antisense primer 1 (5Ј-AAAATAACAATATATGTTTG-3Ј) were used. After reaction the mixture was diluted with 2 ml of 0.1 ϫ TE-buffer, and excess primer was removed by using a Centricon 100 spin filter. The solution was then concentrated to 20 l using a vacuum concentrator. For tailing, 1 l of terminal deoxytransferase (16 units/l), 4 l of 1 mM dATP, 2 l of 10 ϫ buffer, and 13 l of the concentrated solution were mixed and incubated for 10 min at 37°C, heated for 15 min at 65°C, and then diluted with 0.5 ml of water. The compounds for the subsequent PCR reaction were 5 l of template, 0.2 mM each dNTP, 10 l of 10 ϫ PCR buffer, 10 pmol of universal primer, 23 pmol of specific primer, and 23 pmol of nested gene-specific antisense primer 2 (5Ј-CTGATTTAATGGTGACCACAC-3Ј), in a total volume of 100 l. The Taq polymerase (2.5 units) was added after denaturation for 5 min at 94°C and cooling to 72°C. In this case, the PCR program was 2 min at 50°C, 40 min at 72°C, 35 ϫ [1 min at 94°C, 1 min at 58°C, 1 min at 72°C], and 10 min at 72°C.
PCR Amplification-Amplification of the complete coding region from the cDNA used in 3Ј-RACE was done with two specific primers derived from the nucleotide sequence obtained after 3Ј-end and 5Ј-end amplification. The upstream primer (5Ј-CACCATATGTGTGATGCTT-TTG-3Ј) contained an NdeI and the downstream primer (5Ј-GGATCCT-TATGCTCTCTCATAAAC-3Ј) a BamHI restriction site for subsequent cloning into an expression vector. The same protocol was also applied to cDNA from adipose tissue that was reverse-transcribed using the downstream primer.
Cloning and Sequencing of DNA-The RACE products and the PCR products from mammary gland and adipose tissue were purified by electrophoresis on 1% agarose gels (25). The RACE products were cloned into the blunted, dephosphorylated pUC-vector (SureClone TM Ligation Kit), whereas the PCR products of the complete coding regions were cloned into the pCR TM II-vector (TA-Cloning Kit) following the recommendations of the manufacturers. Competent bacteria were prepared and transformed with plasmids as described elsewhere (25). The NdeI/BamHI fragment of an A-FABP clone derived from the mammary gland, which contained the complete coding region and which was sequenced on both strands with Sequenase as described by the manufacturer, was subcloned into the expression vector pET-3a (26) to yield pET-BA-FABP. For comparison, the coding region of an A-FABP clone derived from adipose tissue was sequenced using the digoxigenin system and direct blotting onto a membrane (MWG, Ebersberg, Germany).
In Vitro Transcription-The pET vectors containing the coding region of bovine A-FABP cDNA, bovine H-FABP cDNA (14), and artificial bovine MDGI cDNA (Ref. 10; provided by A. Kromminga and A. Lezius) were transcribed in vitro using the supplier's recommendations. Transcripts were capped by including 0.5 mM of the cap analogue P 1 -(7methyl-5Ј-guanosine)-P 3 -(guanosine-5Ј-triphosphate) in the transcription reaction.
In Vitro Translation-Cell-free translation was carried out in rabbit reticulocyte lysate containing L-[ 35 S]methionine (1000 Ci/mmol) as described by the manufacturer except for the addition of 20 units of RNasin. Typically, 3-10% of the transcripts were translated in a 25-l translation assay, after which the reaction mix was immediately frozen and stored at Ϫ70°C until further use. N-terminal acetylation was inhibited by a preincubation with citrate synthase and oxaloacetate (27).
Immunoprecipitation-Immunoprecipitation was carried out as described by Darley-Usmar et al. (28) with protein A-Sepharose CL-4B and rabbit anti-bovine H-FABP antibodies affinity-purified on a bovine H-FABP-Sepharose column, which later turned out to be cross-reactive with bovine A-FABP.
Purification of FABP from Bovine Heart and Mammary Gland-FABP from bovine heart was isolated according to the protocol of Jagschies et al. (29). For the isolation of FABP from bovine mammary gland, the tissue was homogenized by using an ultraturrax in PBS buffer containing 0.3 mM phenylmethylsulfonyl fluoride. The FABP was purified from the 100,000 ϫ g supernatant by affinity chromatography (rabbit anti-H-FABP antibodies covalently bound to protein A-Sepharose CL-4B) as described by Nielsen and Spener (30).
Electrophoretic Analysis-Prior to electrophoresis, immunoprecipitates for IEF were suspended in a lysis buffer containing 2% Triton X-100, 100 mM dithiothreitol, 2% ampholyte (pH range 7-9), and 9 M urea at 50°C for 10 min. In vitro translation samples (1 l) were diluted with 9 l of lysis buffer and then subjected to IEF. FABP isolated from bovine mammary gland and bovine heart (5 l) was mixed with 5 l of lysis buffer and 5 mg of urea and then applied to IEF.
Isoelectric focusing was performed under denaturing conditions as described earlier (31). Two-dimensional gel electrophoresis, which was also performed under denaturing conditions, and Western blotting were done as described by Nielsen and Spener (30). Proteins were detected by Coomassie staining or specifically immunodecorated after Western blotting using affinity-purified rabbit anti-H-FABP antibodies. For fluorography IEF-gels were soaked in Amplify TM solution. The dried gels were exposed to x-ray films at Ϫ70°C for 2-4 days.
Monospecific Antibodies-Antibodies were raised in New Zealand White rabbits. Monospecific anti-bovine H-FABP antibodies were prepared by double-affinity chromatography. As affinity matrix, 5 mg of recombinant H-FABP (32) and recombinant A-FABP, respectively, were covalently bound to CH-activated Sepharose 4B (Pharmacia). First, antiserum raised against bovine H-FABP was passed over the H-FABP column, and specific antibodies were eluted with a buffer containing 10 mM citrate, 20 mM sodium phosphate (pH 2.8). This antibody fraction was then rechromatographed on the A-FABP column. The flow-through contained monospecific anti-bovine H-FABP antibodies as confirmed by Western blotting. For Western blotting proteins were resolved by 13.5% SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane (Schleicher & Schuell, Dassel). The membrane was incubated either with the double affinity-purified anti-bovine H-FABP antibodies (5 g/ml) or with bovine A-FABP antiserum (1:200) in 0.5% BSA/PBS, 0.05% Tween followed by protein A-peroxidase (1:2500) and stained with 4-chloro-1-naphthol. Recombinant bovine A-FABP for generation of A-FABP antiserum and as standard in Western blotting was obtained by transforming E. coli BL21(DE3)pLysS with pET-BA-FABP. Expression, purification, and characterization of the recombinant A-FABP will be described elsewhere. 2 Immunofluorescence Microscopy-Lactating mammary gland, obtained from a local slaughterhouse, was extensively washed with icecold PBS, fixed in formalin, paraffin-embedded, and hematoxylin/eosinstained. Slides (5 m) were deparaffinized and then blocked for 30 min with 3% BSA/PBS, washed three times with PBS, and incubated for 30 min with double affinity-purified anti-bovine H-FABP antibodies (5 g) or anti-bovine A-FABP antiserum (1:20 dilution), each in 0.5% BSA/ PBS. After washing with 0.5% BSA/PBS (three times), sections were incubated with Cy3-labeled goat anti rabbit IgG (1:300 in 0.5% BSA/ PBS) for 30 min. All steps were performed in a humidity chamber at 37°C. Sections were viewed and photographed under a Nikon DIA-PHOT microscope (Tokyo, Japan). Nonspecific staining was assessed by omission of the primary antibody or by using preimmune serum.

Absence of MDGI in Bovine Mammary
Gland at the mRNA and Protein Levels-Previously, Böhmer and co-workers (2) isolated a 14.5-kDa protein from the bovine mammary gland, which, according to its particular function, was named "mammary derived growth inhibitor" (MDGI). This protein crossreacted with antibodies against H-FABP purified from bovine heart (12) but differed from the latter at positions 12, 14, 40, 43, 93, and 127 and additionally lacked the C-terminal amino acid (see Fig. 8). Since no cDNA for MDGI could be found so far, an artificial cDNA for MDGI was constructed by site-directed mutagenesis starting from the bovine pI 5.1 H-FABP cDNA (10). To identify possible transcripts of MDGI in tissue, the artificial MDGI cDNA was cloned into the pET-3d vector to serve as reference in an in vitro expression assay.
After in vitro transcription of the artificial MDGI cDNA with T7 RNA-polymerase, the corresponding artificial mRNA was translated in vitro in the rabbit reticulocyte lysate using L-[ 35 S]methionine for in situ labeling of synthesized proteins. The in vitro translated protein, analyzed by urea IEF, had an isoelectric point of 4.95 (Fig. 1, lane 3). The coding part of pI 5.1 H-FABP cDNA was also expressed in vitro, and the translation product was analyzed by IEF under denaturing conditions (Fig.  1, lane 2). To detect possibly present MDGI in bovine mammary gland, the poly(A)-rich RNAs of this tissue were translated in vitro, and the translation products were subjected to immunoprecipitation with affinity-purified anti-H-FABP antibodies. The analysis of the immunoprecipitated, in vitro translated proteins revealed two bands (Fig. 1, lane 1) corresponding to the authentic pI 4.9 H-FABP and pI 5.1 H-FABP isoforms isolated from bovine heart (Fig. 1, lane 4). Among the immunoprecipitated proteins no protein with a pI corresponding to the in vitro translation product of artificial MDGI mRNA could be identified.
To investigate the situation in more detail in tissue, a lactating bovine mammary gland was homogenized, and the protein fraction obtained from the 100,000 ϫ g supernatant was used for further experiments. Analysis of the soluble protein fraction by two-dimensional gel electrophoresis resolved several hundred proteins (Fig. 2A). The position of FABP was identified by Western blotting and immunostaining using affinity-purified anti-H-FABP antibodies (Fig. 2B). No protein with a pI value of 4.95 corresponding to the in vitro expressed product of artificial MDGI cDNA was detected (Fig. 2B). Surprisingly, in addition to the pI 4.9 and 5.1 isoforms of H-FABP, a yet unknown protein with a more acidic pI value became immunodecorated. We also isolated FABP from bovine mammary gland by immunoaffinity chromatography and analyzed it by urea IEF. Again the third FABP band with a more acidic isoelectric point was observed along with pI 4.9 H-FABP and pI 5.1 H-FABP, whereas no spot corresponding to artificial MDGI could be detected. Interestingly, analysis of the immunopurified FABP by native IEF revealed only two bands (Fig. 3B, lane  3) corresponding to the pI 4.9 and 5.1 isoforms of H-FABP (Fig.  3B, lane 4). Thus, the unknown immunoreactive protein comigrated either with pI 4.9 H-FABP or with pI 5.1 H-FABP under native conditions. These experiments confirmed the absence of a unique MDGI protein in bovine mammary gland, but the unknown immuno- reactive pI 4.7 protein needed further attention. Since the mammary gland contains also fat and since A-FABP has relatively high homology to H-FABP (63% in the rat), resulting in immunological cross-reactivity, we suspected that the additional protein observed in two-dimensional gel electrophoresis could be the bovine A-FABP. The following experiments were carried out to determine the bovine A-FABP nucleotide sequence and to establish the presence of A-FABP in the lactating bovine mammary gland.
Isolation of the Full-length cDNA Encoding Adipocyte FABP from Bovine Lactating Mammary Gland Using RACE-PCR-To obtain the 3Ј-end of the bovine A-FABP cDNA, purified mRNA from bovine mammary gland was reverse-transcribed. For 3Ј-end amplification of cDNA, a degenerated genespecific sense primer was used together with the specific primer. The degenerated gene-specific primer was designed according to the published cDNA sequences of human A-FABP (33) and mouse A-FABP (34) but would also allow the amplification of a yet unidentified MDGI cDNA (Fig. 4). Care was taken that the two 3Ј-nucleotides of this primer did not match the cDNA sequence of bovine pI 5.1 H-FABP, which is abundantly expressed in differentiated bovine mammary gland (13). The primers for 5Ј-end amplification were designed according to the DNA sequence obtained after sequencing of the 3Ј-end amplified PCR product. Once the 3Ј-and 5Ј-ends had been identified, the complete coding region of the bovine A-FABP cDNA was amplified and sequenced. The deduced amino acid sequence of bovine A-FABP reveals 85 and 87% identity to human A-FABP and mouse A-FABP, respectively (Fig. 8).
Recombinant A-FABP mRNA Encodes the Unknown Immunoreactive pI 4.7 FABP-We transferred the coding part of the A-FABP cDNA into the pET-3a vector containing the T7 promoter. The product of the in vitro transcription by T7 RNA polymerase and subsequent translation in a cell-free system was analyzed by urea IEF (Fig. 5, lane 1). The in vitro expressed protein comigrated with the unknown immunoreactive pI 4.7 protein present in the FABP isolated from bovine mammary gland by immunoaffinity chromatography (Fig. 5, lane 5). Thus, the three bands observed in the urea IEF gel of immunopurified FABP from bovine mammary gland can be explained as the pI 4.9 and 5.1 isoforms of H-FABP and as A-FABP, respectively. This result points to the existence of A-FABP in addition to H-FABP in lactating bovine mammary gland.
The appearance of two additional variants with more basic pI values observed in urea IEF of in vitro expressed products of bovine A-FABP cDNA (Fig. 5, lane 1) is probably due to different N-terminal processing (A-FABP, Ac-Cys-Asp-Ala-; variant 1, Ac-Asp-Ala-; variant 2, Asp-Ala-). By the addition of Nterminal acetylation inhibitors (27), the amount of variant 1 was decreased, whereas the band of variant 2 became stronger (Fig. 5, lane 2). By the same token, the N-terminal acetylation of in vitro expressed pI 5.1 H-FABP (Fig. 5, lane 3) could be partially suppressed as indicated by the appearance of an additional band with a more basic pI value (Fig. 5, lane 4).
Detection of A-FABP in Adipose Tissue and Lactating Mammary Gland-To demonstrate that the protein we had cloned from the lactating mammary gland and termed A-FABP according to its homology to other A-FABPs was indeed the adipocyte-type FABP we analyzed the expression of the corresponding protein in adipose and mammary gland tissue. Western blotting of cytosolic proteins with antiserum directed against the recombinant bovine A-FABP confirmed the presence of a protein of identical size (ϳ14 kDa) in mammary gland as well as in adipose tissue (Fig. 6, lanes 6 and 8). We then by PCR on cDNA from adipose tissue amplified a 400-base pair DNA fragment using the same primers as described above for the mammary gland. The amplified DNA was cloned and sequenced and represented bovine A-FABP cDNA (data not shown).
The question arose whether the A-FABP cloned from the mammary gland originates from the adipocytes that may be present in mammary tissue or from the mammary cells. The staining of mammary gland and adipose cytosol with the anti-A-FABP antibodies in the Western blot (Fig. 6) already indicated that the amount of A-FABP was too large to be attributed to the relatively few adipocytes still present in the lactating mammary gland, in particular since we have tried to free the mammary gland tissue from all adhering adipose tissue. To prove this hypothesis we analyzed the cellular localization of both A-FABP and H-FABP by immunofluorescence. As a prerequisite, the specificity of the antibodies had to be demonstrated because it is known that anti-H-FABP antibodies often cross-react with A-FABP (see Fig. 2 and Ref. 35). Here we show that the A-FABP antiserum specifically recognized recombinant bovine A-FABP (Fig. 6, lane 7), whereas the same amount of recombinant bovine H-FABP was not stained (Fig. 6, lane 9). A monospecific antibody against bovine H-FABP which was obtained by double affinity purification did not cross-react with recombinant bovine A-FABP and detected no H-FABP in adipose tissue (Fig. 6, lanes 4 and 5) but, as expected, stained large amounts of H-FABP in heart and mammary gland (lanes 1  and 3).
Localization of A-FABP and H-FABP in the Mammary Gland-To analyze the expression of both FABP types on cellular level in the lactating mammary gland, these monospecific antibodies were used in immunofluorescence. The typical alveolar structure of the mammary gland is demonstrated in the hematoxylin/eosin stain (Fig. 7A). With monospecific anti-H-FABP antibodies a strong immunofluorescence was observed in the epithelial cells of the terminal acini (Fig. 7B). Reactivity of the myoepithelial cells surrounding the alveolar epithelium could not be judged with certainty. In contrast, the anti-A-FABP antiserum reacted with the myoepithelial cells, omitting the alveolar epithelium (Fig. 7C). Immunofluorescence of fibroblasts, connective tissue, and some smooth muscle cells around blood vessels was negligible with both antibodies compared with the epithelial cells. DISCUSSION Since DNA probe and immunological techniques cannot distinguish between the closely related MDGI and H-FABP, we applied an in vitro expression analysis to address the possible presence of both proteins in the mammary gland. Based on published amino acid sequences, the isoelectric points of MDGI and H-FABP should differ. From the in vitro translation experiments and from two-dimensional gel electrophoresis it became clear that the previously described sequence for MDGI (2) could not be correct. However, an as yet unidentified immunoreactive pI 4.7 protein was observed. For isolation of the cDNA of the unknown immunoreactive pI 4.7 FABP from lactating bovine mammary gland, a degenerated primer was used. The primer sequence was chosen according to the cDNA sequences of hu-man A-FABP (33) and mouse A-FABP (34). By multiple alignment, the deduced amino acid sequence of the isolated cDNA was identified as bovine A-FABP (Fig. 8). The identity of the hitherto unknown bovine A-FABP with mouse A-FABP was 87% and with human A-FABP was 85%, in accordance with typical identities of FABPs of the same type but from different mammalian species of about 80 -90% (4). The in vitro expressed protein of the bovine A-FABP cDNA comigrated with the unknown pI 4.7 FABP in urea IEF (see below), confirming that the unknown immunoreactive pI 4.7 FABP indeed represents A-FABP. This assignment was corroborated by cloning and sequencing of the same cDNA from bovine adipose tissue. Moreover, antibodies raised against recombinant bovine A-FABP in Western blotting recognized a 14-kDa protein in adipose and mammary gland tissue. The three FABPs observed in lactating bovine mammary gland were thus identified as pI 4.9  H-FABP, pI 5.1 H-FABP, and A-FABP. We have earlier traced back the molecular basis for the two H-FABP isoforms, which were originally observed in bovine heart (29), to an Asn-Asp exchange at amino acid position 98. The published H-FABP cDNA (14) codes for Asn 98 and hence represents the pI 5.1 isoform (see also Fig. 1, lane 2). We presented arguments (31) that the isoforms stem from two different mRNAs rather than from posttranslational modification (deamidation).
The published amino acid sequence of MDGI deviated in only seven positions from that of bovine H-FABP (2,11). Each of these seven amino acids was found at the same position in the sequence of bovine A-FABP. Consequently, the amino acid sequence published for MDGI represents a mix of H-FABP and A-FABP. Native IEF of FABP from bovine lactating mammary gland (Fig. 3B, lane 3) only showed the two bands corresponding to pI 4.9 H-FABP and pI 5.1 H-FABP. Apparently, A-FABP had comigrated with pI 4.9 H-FABP or pI 5.1 H-FABP and can be separated from H-FABP only in IEF under denaturing conditions (Fig. 3A, lane 1). In our hands it was not possible to separate A-FABP and H-FABP in an isolation procedure employing native conditions. This behavior likely led to isolation and sequencing of a mixture consisting of H-FABP and A-FABP by Böhmer and co-workers (2).
A critical experiment was the confirmation of the identity of bovine A-FABP and the unknown immunoreactive pI 4.7 FABP by in vitro transcription/translation. The findings were somewhat obscured by the presence of more than one band in the case of in vitro expressed bovine A-FABP (Fig. 5, lane 1). Wold (36) formulated a general model for N-terminal processing of eukaryotic cytosolic proteins that incorporated the presence of an acetylating activity and an acetylaminoacyl-hydrolyzing activity as general components. N-terminal processing of class II actin (Met-Cys-Asp-) in rabbit reticulocyte lysate studied by Rubenstein and Martin (37) is consistent with this model. They found that methionine was removed as free amino acid, followed by acetylation of Cys-Asp-, and eventually the aspartate residue was acetylated after the removal of the acetylated cysteine, giving the final in vivo product. Analyzing the in vitro expressed products of bovine A-FABP cDNA encoding for Met-Cys-Asp-by urea IEF we observed a protein (Fig. 5, lane 1) corresponding to authentic A-FABP (Ac-Cys-Asp-; Ref. 38). Two additional variants were found (presumed N terminus of variant 1, Ac-Asp-Ala-; and of variant 2, Asp-Ala-), which were probably due to N-terminal processing as observed for class II actin (37,39). Formation of fully processed class II actin (Ac-Asp-) in the rabbit reticulocyte lysate could be decreased by inclusion of acetylation inhibitors (39) as also observed by us It remains an open question whether this different N-terminal processing is due to specific recognition of an N-terminal sequence motif or whether N-terminal processing is tissue-specific.
Here we clearly demonstrate for the first time that A-FABP is expressed in cells of the lactating bovine mammary gland as well as H-FABP. Bansal and Medina (20) reported the expression of A-FABP in the virgin mouse mammary gland and the expression of H-FABP in the lactating mouse mammary gland. The differentiation-dependent expression of H-FABP in the rat mammary gland was earlier shown on mRNA (18) and protein level (17). The presence of A-FABP in virgin mammary gland was expected because virgin mammary gland contains considerable amounts of adipocytes. In contrast, in differentiated mammary gland 90% of the cells are epithelial cells. Since A-FABP could be clearly seen in the two-dimensional gel of mammary gland proteins (Fig. 2), in urea IEF of the immunopurified FABP (Fig. 3A, lane 1), and in Western blotting with the A-FABP-specific antiserum we believed that A-FABP was expressed in the mammary cells as well. We could then show by immunofluorescence with the monospecific antiserum that A-FABP was present in myoepithelial cells, whereas H-FABP was detected in mammary epithelial cells as reported earlier (35,40). The cellular localization of both FABPs in the mammary gland will be addressed in more detail in a separate immunohistochemical study. Some reports in the literature describe the presence of two FABP types in one organ (hearttype and intestinal-type FABP in stomach (41), heart-type and liver-type FABP in kidney (42), heart-type and brain-type FABP in brain (43)), as well as in a single cell-type (intestinaltype and liver-type FABP in enterocytes of the intestine (44)).
What would be the reason for the presence of two types of FABP (A-FABP and H-FABP) in the same organ? Coexpression of FABP types is generally interpreted in terms of specialized functions in the metabolism of fatty acids. H-FABP is predominantly found in cells where fatty acids are used as energy source and is probably involved in ␤-oxidation (5-7). Its abundance in the mammary gland, which during lactation has an active triglyceride synthesis, remains to be understood. The role of adipocytes in vivo is the storage and mobilization of triglycerides. The highest concentration of A-FABP was found in adipocytes; therefore, A-FABP could be involved in triglyceride metabolism in the mammary gland as well. The results of this work clearly demonstrate that MDGI as a distinct protein does not exist. However, it had been shown earlier that its function can be exerted by H-FABP (10). H-FABP from bovine heart and bovine mammary gland, recombinant H-FABP, and recombinant MDGI derived from the artificial MDGI cDNA act as growth inhibitors on primary mouse mammary epithelial cells. In a mammary organ culture, growth inhibition was associated with functional differentiation. On the contrary, intestinal-type FABP and liver-type FABP did not exhibit inhibitory activity in this assay (10). In a different approach Huynh et al. (40) increased the intracellular H-FABP content of MCF-7 human breast cancer cells by transfection and observed reduced proliferation rate and reduced tumorgenicity in nude mice relative to nontransfected or mocktransfected controls. The authors suggest that H-FABP is a candidate for a tumor suppressor gene. The inhibitory function of H-FABP does not require binding of fatty acids, because an undecapeptide derived from the C terminus of H-FABP, which does not bind fatty acids, could also mimic the activity of H-FABP (10). The C terminus of A-FABP has a modest similarity to H-FABP (Fig. 8). It remains to be established whether A-FABP can act as growth inhibitor as well. The lacking inhibition by intestinal-type FABP and liver-type FABP may well be related to a low homology in the C-terminal region to H-FABP.