Delayed mammary tumor progression in Muc-1 null mice.

The mucin gene, Muc-1, encodes a high molecular weight integral membrane glycoprotein that is present on the apical surface of most simple secretory epithelial cells. Muc-1 is highly expressed and aberrantly glycosylated by most carcinomas and metastatic lesions. Numerous functions have been proposed for this molecule, including protection of the epithelial cell surface, an involvement in epithelial organogenesis, and a role in tumor progression. Mice deficient in Muc-1 were generated using homologous recombination in embryonic stem cells. These mice appeared to develop normally and were healthy and fertile. However, the growth rate of primary breast tumors induced by polyoma middle T antigen was found to be significantly slower in Muc-1 deficient mice. This suggests that Muc-1 plays an important role in the progression of mammary carcinoma.

The mucin gene, Muc-1, encodes a high molecular weight integral membrane glycoprotein that is present on the apical surface of most simple secretory epithelial cells. Muc-1 is highly expressed and aberrantly glycosylated by most carcinomas and metastatic lesions. Numerous functions have been proposed for this molecule, including protection of the epithelial cell surface, an involvement in epithelial organogenesis, and a role in tumor progression. Mice deficient in Muc-1 were generated using homologous recombination in embryonic stem cells. These mice appeared to develop normally and were healthy and fertile. However, the growth rate of primary breast tumors induced by polyoma middle T antigen was found to be significantly slower in Muc-1 deficient mice. This suggests that Muc-1 plays an important role in the progression of mammary carcinoma.
The MUC1 protein is a heavily glycosylated cell-associated mucin glycoprotein that is highly expressed and aberrantly glycosylated by the majority of carcinomas and, in particular, by Ͼ92% of primary and metastatic breast cancers. This protein was initially identified as a component of the human milk fat globule membrane (1). Subsequently, it was found to be expressed by the majority of simple secretory epithelial cells in addition to being highly expressed by most human carcinomas, including those of the breast, colon, ovary, pancreas, lung, and stomach (2,3). This widespread expression pattern is in contrast to the other human mucin genes that have been isolated, which appear to have a more restricted spatial expression pattern (reviewed in Ref. 4). MUC1 is also known as PEM, episialin, H23 antigen, DF3 antigen, PUM, and epithelial membrane antigen (or EMA) and by a variety of other names.
Mucins, or mucin-type glycoproteins, can be defined as large extended molecules with a high percentage (50 -90%) of their molecular mass made up of carbohydrate that is attached via O-glycosidic linkage through N-acetylgalactosamine to serine and/or threonine. They can be subdivided into the classical secretory or soluble mucins and the membrane-associated mucins. The secretory mucins constitute the viscous mucus of the tracheobronchial, gastrointestinal, and reproductive tracts and typically form extremely large complexed oligomers through linkage of protein monomers via disulfide bonds. These proteins are secreted from the cell and remain at the apical surface in the form of a mucus gel. This secretion serves as a selective physical barrier between the extracellular milieu and the plasma membrane and cell interior. The membrane-associated mucins are intimately associated with the plasma membrane through a hydrophobic transmembrane domain and have not been observed to form oligomeric complexes. However, these proteins can be secreted from the cell surface, most probably through a proteolytic cleavage mechanism (5,6).
Isolation of cDNA and genomic clones for the human MUC1 gene suggested that the gene coded for a typical type I membrane glycoprotein (7)(8)(9)(10)(11)(12). At the amino terminus there was a short signal peptide, and at the carboxyl terminus there was a 28 -31-amino acid transmembrane domain and 69 -72-amino acid cytoplasmic tail. The majority of the external domain of the protein was made up of multiple copies of a 20-amino acid repeat (up to 125 copies) that was rich in serine, threonine, and proline. Serine and threonine residues act as the attachment sites for O-linked carbohydrate, and it is this region of the protein that is highly O-glycosylated. Isolation of the mouse Muc-1 gene indicated that the most strongly conserved regions corresponded to the serine, threonine, and proline residues within the repeat domain and to the transmembrane and cytoplasmic domains (13,14). This would suggest that the primary function of the repeat domain is to act as a scaffold for carbohydrate attachment. The sequence conservation of the transmembrane and cytoplasmic domains would suggest an important function for this part of the protein. In this respect, MUC1 has been demonstrated to be associated with elements of the actin cytoskeleton (15), and presumably it is the cytoplasmic domain of MUC1 that is involved in this interaction. An investigation of the expression pattern of Muc-1 during mouse embryogenesis revealed that expression was regulated both spatially and temporally and correlated well with the onset of epithelial differentiation in each organ (16). In addition, the pattern of Muc-1 expression in the adult mouse and in mouse mammary carcinomas correlated well with the pattern of expression observed for the human MUC1 gene (17,18).
Traditionally, mucins have been thought of as protective molecules. However, the fact that MUC1 is expressed early during epithelial organogenesis and is highly expressed by tumors and metastatic lesions suggests other functions. In general these functions may be a direct reflection of the large size of the molecule. According to the model of Jentoft (19), the fully glycosylated human MUC1 protein may extend as much as 300 -500 nm above the cell surface, far above the glycocalyx. Thus, due to its large extended conformation, one potential function for MUC1 is as an antiadhesive protein, possibly blocking cell-cell and cell-matrix interactions (20 -22). Muc-1 in normal polarized epithelia is expressed only on the apical surface. However, in many adenocarcinomas polarization is lost and the protein can be found on basolateral cell surfaces, where it could interfere with cell-cell and cell-substratum adhesion. Alternatively, MUC1 may play an adhesive role by presenting carbohydrates as ligands for selectin-like molecules and thus aiding metastatic dissemination (23)(24)(25). This may be particularly relevant as MUC1 has recently been shown to express sialyl Lewis x and sialyl Lewis a , ligands for P-and E-selectins (26,27). Clearly both antiadhesive and proadhesive functions could play significant roles in normal development, tumor progression, and disease.
In order to investigate the biological function of the mouse Muc-1 protein, we generated mice deficient in Muc-1 through homologous recombination in mouse embryonic stem (ES) 1 cells. These mice appeared to develop normally and were healthy and fertile. Muc-1-deficient mice exhibited no differences in survival rate and appeared phenotypically normal in all respects.
Direct evidence of a role for MUC1 in the development and/or progression of breast cancer has not been demonstrated previously. To investigate the role of Muc-1 in tumor development and/or progression, we compared the growth of mammary tumors in Muc-1 Ϫ/Ϫ and ϩ/ϩ mice. To generate mammary tumors, we utilized mice transgenic for the polyoma virus middle T antigen (28). In female mice, the middle T antigen is under the control of the mouse mammary tumor virus promoter, and expression is specific for the mammary gland and to a lesser extent the salivary gland. Virgin female mice of this strain have been shown to develop multifocal breast tumors by 2 months of age, and by 4 months of age greater than 50% of these mice will have developed lung metastases. Interestingly, the middle T antigen has been demonstrated to require the presence of the src oncogene for its ability to transform mammary cells (29). Similarly, it has been demonstrated that the neu oncogene, implicated in up to 30% of human breast cancers (30), binds to and activates src tyrosine kinase activity (31). We have employed the middle T oncogene in this study due to its rapid time course of tumor induction, reliable production of spontaneous tumor metastases, and the possible commonality of signal transduction pathways with the neu protooncogene.
The results indicate that Muc-1 does indeed play an important role in the progression of mammary carcinoma, as primary tumor growth rate was significantly slower in Muc-1-deficient mice as compared with their wild type counterparts.

MATERIALS AND METHODS
Muc-1 Gene-targeting Vector Construction-An isogenic Muc-1 genomic clone was obtained through screening a 129SV/J mouse cosmid library (Stratagene Cloning Systems, La Jolla, CA) with the Muc-1 cDNA probe, pMuc2TR, previously described (13). Approximately 5 ϫ 10 5 colonies were plated onto Biodyne nylon membrane (Pall Biodyne, Glen Cove, NY) on LB-agar plates supplemented with ampicillin to 100 g/ml. Double lifts were taken from each membrane and screened with pMuc2TR labeled with [␣-32 P]dCTP (Amersham Corp.) by random priming (32) under conditions recommended by the manufacturer. Positive clones were taken through two further rounds of colony purification to yield pure clones. An 11-kb EcoRI fragment containing the entire mouse Muc-1 gene (13,14) was subcloned into pBluescript SKII ϩ (Stratagene Cloning Systems, La Jolla, CA) and designated 129Muc-1E2.
The cloning strategy is diagrammed in Fig. 1. An Escherichia coli ␤-galactosidase (LacZ) gene (33) lacking the first 7 codons, including the translation initiation codon, designated plasmid 839 (kindly provided by R. Krumlauf), was subcloned into the BamHI site of pBluescript KSII ϩ . The modified LacZ gene, pBS-LacZ, was subcloned further into pBluescript SKII ϩ as a SmaI-EcoRI fragment. This cloning step resulted in the loss of the ␤-galactosidase termination codon and was necessary in order to place the appropriate restriction sites at the 5Ј end of the gene to allow an in-frame ligation between Muc-1 and LacZ to be made and to place sites at the 3Ј end of the gene to allow further manipulations. A 2-kb SmaI fragment of the Muc-1 gene was cloned into the SmaI site created at the 5Ј end of the LacZ gene to construct the plasmid pMuc-LacZ. This step created a Muc-1/LacZ fusion protein designed to be under the control of the Muc-1 transcription and translational machinery. This plasmid incorporated the promoter and first three codons of the mouse Muc-1 gene ligated in-frame, at the SmaI site in the first exon, to the eighth codon of the LacZ gene. A neomycin (neo) resistance gene under the control of the phosphoglycerate kinase (PGK) promoter and poly(A) site (34,35) was released from pKJ-1 (kindly provided by M. McBurney) and cloned into the EcoRI-HindIII sites of the plasmid pMuc-LacZ in the same transcriptional orientation as the Muc-1 and LacZ genes to create pMucLacNeo. To restore the LacZ stop codon and also to place the poly(A) signal of SV40 at the 3Ј end of the LacZ gene, a 0.8-kb EcoRI fragment was subcloned from the plasmid pPGK (E/T)LacZ (kindly provided by M. McBurney) into the EcoRI site between the LacZ and Pgkneo genes. This plasmid was designated pMucLacZSVneo.
The final 129 Muc-1-targeting vector was created as follows: the plasmid 129Muc-1E2 was digested with SmaI, and a 10-kb fragment, containing the cloning vector in addition to 7 kb of the Muc-1 gene locus, was gel-purified on DEAE (NA45) paper (Schleicher and Schuell). The 8-kb MucLacZSVneo cassette was removed from pMucLacZSVneo as a NotI-HindIII fragment. 3Ј overhanging ends were filled in using Klenow DNA polymerase, and the blunted fragment was gel-purified. Ligation of the 10-and 8-kb fragments and selection for plasmids containing the two fragments ligated in the correct orientation resulted in the creation of the final Muc-1-targeting vector, designated 129Muc-1GT. All ligation junctions were confirmed by sequencing by the dideoxy chain termination method (36) using Sequenase version 2.0 (U. S. Biochemical Corp.) and synthetic oligonucleotides.
Subconfluent ES cells at passage 9 were fed 2-3 h prior to electroporation, trypsinized, spun down, and washed with phosphate-buffered saline. Cells were resuspended in phosphate-buffered saline to an approximate concentration of 5 ϫ 10 7 ES cells/ml. The 129Muc-1GTtargeting vector was linearized with NotI, extracted once with an equal volume of Tris-buffered phenol-chloroform-isoamyl alcohol (25:24:1), precipitated, and dissolved in sterile TE buffer, pH 8. DNA was added to the cell suspension to a concentration of 5 nM. The cells were electroporated in 800 l of phosphate-buffered saline with a BTX electroporator at a capacitance of 500 microfarads and 250 V at room temperature. Cells were placed on ice for 10 min immediately after electroporation and then plated onto 10 90-mm dishes of fresh STO-neo feeders. Twenty-four hours after electroporation, G418 (400 g/ml active constituent) (Life Technologies, Inc.) was added to the culture medium. After 10 -11 days colonies were picked and expanded for freezing and DNA analysis.
Genomic DNA was prepared from each clone, digested with EcoRI, size-separated by agarose gel electrophoresis, transferred to Hybond-N nylon membrane (Amersham), and screened by Southern analysis utilizing a 32 P-labeled 2-kb EcoRI-SmaI 5Ј-flanking probe. Conditions used for hybridization were as recommended by the manufacturer. Homologous recombinants were further investigated using the 5Ј-flanking probe and internal probes. The chromosome constitution of the parental ES cell line GK129 and correctly targeted clones was determined by analysis of colcemid disrupted metaphase spreads as described (38).
Generation of Chimeric Mice and Screening Progeny-C57BL/6J blastocysts were isolated at day 3.5 and injected with between 8 and 20 ES cells. An average of nine injected blastocysts were transferred to each uterine horn of pseudopregnant recipient females of mouse strain CD1 or CBA/C57BL/6J F1 at 2.5 days post-coitum. Chimeras were identified on the basis of agouti/chinchilla pigmentation in the coat. Both male and female chimeras were back-crossed with C57BL/6J mice to screen for germ line transmission. Agouti F1 progeny were genotyped by PCR analysis of tail DNA utilizing oligonucleotides specific for the Muc-1 gene, 5Ј-ACCTCACACACGGAGCGCCAG-3Ј (corresponding to bp ϩ7 to ϩ27 (13)) and the LacZ gene, 5Ј-TTCTGGTGCCGGAAAC-CAGGC-3Ј (corresponding to bp 201-181, antisense strand (33)). Tail DNA was prepared as described (39) and dissolved in water to a final volume of 100 l. Five l of each DNA preparation were utilized in a 100-l PCR reaction under standard buffer conditions with 2% (v/v) deionized formamide, 1 M each oligonucleotide primer, 200 M dNTPs, and 2.5 units of Taq polymerase (Boehringer Mannheim). DNA samples were heated for 10 min at 95°C before the addition of 95 l of a PCR mix containing all of the necessary components. Amplification proceeded for 40 cycles of 95°C for 1 min, 62°C for 30 s, 72°C for 1 min, followed by a single final extension of 5 min at 72°C. Amplified products were analyzed by agarose gel electrophoresis. Chimeras demonstrating 100% germ line transmission were further bred with 129SV/J mice in order to establish an inbred line of mutant mice.
Heterozygous agouti offspring were intercrossed to generate mice homozygous for the Muc-1 mutation. Muc-1 ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ mice were identified by PCR analysis utilizing two sets of oligonucleotide primers under the conditions described above. The first set comprised the 5Ј Muc-1 primer, previously mentioned, in combination with a 3Ј Muc-1 primer, 5Ј-TCCCCCCTGGCACATACTGGG-3Ј (corresponding to bp ϩ268 to ϩ248, antisense strand (13)). Together, these primers amplified a 262-bp product from the wild type allele only. The second set of primers were identical to the pair utilized to screen for the presence of agouti F1 heterozygotes, as described. The validity of the PCR genotyping results was initially confirmed by Southern analysis of EcoRIdigested tail DNA utilizing the 5Ј-flanking probe described above.
Immunohistochemistry-Immunostaining with the polyclonal antiserum, CTI, raised to a synthetic peptide corresponding to the 17 C-terminal amino acids in the cytoplasmic tail of human MUC1 (17) was performed as described (16,17). As a negative control, the sections were incubated with CTI antiserum previously blocked with 5 mg/ml of the synthetic peptide. Immune complexes were detected using fluorescein isothiocyanate-conjugated swine anti-rabbit immunoglobulins (DAKO Corp, Carpinteria, CA). Photomicrographs were taken on a Nikon FXA photomicroscope with Kodak T-Max 400 film. In addition, sections were routinely stained with hematoxylin and eosin according to standard procedures to investigate morphology. In all cases three ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ animals were analyzed.
Gene Expression Analysis-RNA samples isolated from tissues of ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ littermates were investigated for the expression of a variety of genes utilizing a slot-blot approach. 1 g of total RNA from ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ tissues was loaded into adjacent wells of a vacuum slot-blot apparatus (Schleicher and Schuell) and vacuum-blotted onto nylon membrane (Hybond-N) under conditions recommended by the manufacturer. Duplicate blots were prepared and hybridized with probes to a large panel of genes (Table II). Genes that were identified as being potentially up-regulated by this approach were further investigated by Northern analysis. Antisense oligonucleotide probes were 5Ј end-labeled with [␥-32 P]ATP (Amersham) utilizing T4 polynucleotide kinase (New England BioLabs, Beverly, MA). All probes were hybridized as described previously for Northern analysis.
Tumor Generation System-To study the role of Muc-1 in tumor formation, FVB mice transgenic for the polyoma virus middle T antigen under the control of the mouse mammary tumor virus promoter (a generous gift of Dr. W. Muller) were crossed with 129SV/J:C57BL/6J outbred mice homozygous for either the Muc-1 mutant or wild type alleles. Resulting offspring were screened for the middle T antigen transgene by PCR utilizing oligonucleotides specific for the polyoma virus middle T gene, 5Ј-AGTCACTGCTACTGCACCCAG-3Ј (corresponding to bp 282-302 (43)) and 5Ј-CTCTCCTCAGTTCTTCGCTCC-3Ј (corresponding to bp 817-837 antisense strand (43)). Positive males were bred back on 129SV/J:C57BL/6J outbred mice homozygous for either the Muc-1 mutant or wild type alleles, respectively. Female offspring of the second cross were screened for the middle T antigen transgene and for homozygosity of either the Muc-1 mutant or wild type Muc-1 alleles. For the study, female mice were housed in groups of three in a low stress barrier housing facility. From 68 days of age through 124 days of age, mice were palpated 3 times/week for the presence of mammary tumors. Palpable tumors were measured by calipers, and tumor weight was calculated according to the following formula (44): grams ϭ (length ϫ (width) 2 )/2, where length and width are measured in centimeters.
Tumor weights plotted are the sum of the weights of the individual tumors. At the 124-day end point, tumors were dissected, and a portion was fixed in methacarn for histochemical analysis. Lungs were dis-sected, fixed in methacarn, and examined under a dissecting scope for visible metastatic lesions.
Comparison of the tumor weights at each time point between the two groups (Muc-1 ϩ/ϩ and Ϫ/Ϫ mice) was made with the two-sample t test. All calculated p values were two-sided, and p values less than 0.05 were considered statistically significant.

RESULTS
Targeted Inactivation of Muc-1-It has been demonstrated that the use of isogenic gene-targeting constructs increases the efficiency of homologous recombination in ES cells (45). Therefore, we cloned the Muc-1 gene locus from a 129SV/J cosmid library (Stratagene Cloning Systems). The mouse Muc-1 gene consists of seven exons and is a compact gene, spanning approximately 5 kilobase pairs (13) and located on mouse chromosome 3 (46). In addition, this gene appears to be part of an unusually tight cluster of genes (47) that are conserved between mouse chromosome 3 and human chromosome 1q21-1p22. Human MUC1 has been localized to chromosome 1q21 (48). The polyadenylation signal (poly(A)) of the mouse thrombospondin-3 (Thbs-3) gene is located within 3 kilobase pairs immediately upstream of the Muc-1 gene (49), and the poly(A) of a third gene (Gene Y) is located approximately 1 kilobase downstream of the Muc-1 poly(A) (47). Although these genes are tightly linked, their patterns of expression are unique, and thus they appear to be regulated by distinct promoters. We constructed a replacement-type targeting vector, 129Muc-1GT, in which the Muc-1 gene was disrupted by the insertion of a cassette containing the E. coli ␤-galactosidase (LacZ) gene (33) and SV40 poly(A) sequence in addition to the PGK neo gene (Figs. 1 and 2A).
Plasmid 129Muc-1GT was linearized prior to electroporation into the ES cell line GK129 (37). Electroporated cells were cultured in medium containing the antibiotic G418 for 10 -11 days. Resistant colonies were picked and expanded for freezing and DNA analysis. DNA was isolated, and homologous recombinants were identified by Southern analysis of EcoRI-digested DNA using a 5Ј-flanking probe. Correctly targeted clones were obtained at a frequency of 1 in 18 G418 resistant colonies. The wild type (endogenous) Muc-1 allele displays an 11-kb EcoRI band when hybridized with the 5Ј-flanking probe (probe 1). In contrast, the correctly targeted allele displays a novel 7-kb EcoRI band ( Fig. 2A). Clones identified in this way were further analyzed using EcoRI and BamHI digests in combination with a variety of internal and flanking probes. This was especially important considering the close proximity of genes 5Ј and 3Ј of Muc-1. In addition to clones displaying the expected novel hybridizing band at 7 kb (#56 and #132, Fig. 2A), a number of aberrantly targeted clones were obtained (#8, Fig. 2A). These aberrant clones appeared to be the result of the insertion of a concatemer of the construct into the Muc-1 locus followed by a complex rearrangement event.
Generation of Muc-1 Mutant Mice-Three correctly targeted clones were injected into C57BL/6J blastocysts and transferred to the uteri of pseudopregnant recipients. All three lines generated chimeric mice with ES contributions up to 100% as judged by agouti/chinchilla coat color. Chimeras derived from all three clones transmitted the ES cell genome through the germ line as evidenced by the agouti coat color of the offspring after back-crossing with C57BL/6J mice (Table I) Approximately 50% of the agouti offspring of the chimeras were heterozygous for the designed mutation. These mice were indistinguishable from their wild type littermates. Heterozygous (ϩ/Ϫ) animals were intercrossed to generate animals homozygous for the Muc-1 mutation. Homozygous (Ϫ/Ϫ) animals were obtained from clones 56 and 31. Homozygous animals were identified through a PCR-based screening procedure (Fig.  2B), and results were initially confirmed through Southern blotting of EcoRI-digested tail DNA utilizing the 5Ј-flanking probe (Fig. 2C). In all cases, animals homozygous for the disrupted Muc-1 allele (Ϫ/Ϫ) were obtained at the expected Mendelian frequency. In addition, inbred 129SV/J heterozygotes were obtained from the original chimeric animals and intercrossed to derive an inbred line homozygous for the Muc-1 mutation. Similarly, inbred C57BL/6J heterozygotes are being derived through a series of back-crosses onto C57BL/6J. These mice are currently at N8 (99% inbred with respect to C57BL/6J).
Homozygous Mutant Mice Do Not Express Muc-1 mRNA and Protein-In order to determine whether the insertion of the LacZ-Pgkneo cassette resulted in the efficient disruption of Muc-1 transcription and subsequent translation, total RNA was prepared from a panel of tissues isolated from ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ littermates. Approximately equivalent amounts of total RNA were subjected to Northern analysis with the previously characterized Muc-1 cDNA probe, pMuc2TR (13). Expression of Muc-1 was found to be reduced in the tissues of heterozygous mice and undetectable in homozygous mice (Fig. 3). In addition, immunohistochemistry indicated no detectable Muc-1 protein on the apical surface of Muc-1 Ϫ/Ϫ secretory epithelial tissues (Fig. 4). Thus, targeted inactivation of the Muc-1 gene by the replacement vector, 129Muc-1GT, resulted in the creation of a null Muc-1 allele.
An investigation of LacZ expression in Muc-1 ϩ/Ϫ and Ϫ/Ϫ mice indicated that the expression of LacZ in the context of the endogenous Muc-1 promoter did not faithfully reflect the previously described expression pattern of Muc-1 in the developing embryo or adult. This was found to be due to a block in transcription of the Muc-1/LacZ fusion gene. 2 Muc-1 Null Mice Develop Normally and Are Viable and Fertile-Mice deficient in Muc-1 were obtained at the expected frequency from all crosses. These mice appeared to develop normally and gained weight at the same rate as their heterozygous and wild type littermates (data not shown). Examination of hematoxylin-eosin stained sections prepared from all the major organs revealed no obvious differences between Muc-1deficient mice and their corresponding littermates (data not shown). Similarly, whole mounts of virgin mammary glands of 12-week old wild type and Muc-1 null animals showed no obvious differences in glandular morphology (data not shown). All possible pairwise crosses of genotypes indicated no differences in the fertility of the parents, subsequent litter size, growth rate, and survival of the litters (data not shown). This would suggest that Muc-1 present in milk is not important for the growth and survival of neonates under pathogen-free conditions.
A Screen for Potential Compensating Genes-We explored the possibility that the up-regulation of expression of one or more mucin-like genes or membrane glycoproteins may have accounted for the apparent lack of a phenotype in Muc-1-deficient mice. Probes were obtained either as antisense oligonucleotides (50 mers) or, alternatively, cloned cDNAs were utilized (Table II). Total RNAs isolated from ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ littermates were investigated by slot-blot analyses with the various probes. No difference in expression levels was observed for Muc-2 or Muc-4, although high levels of Muc-4 expression were observed in lactating mammary gland, salivary gland, lung, stomach, kidney, and colon (Fig. 5). Similarly, no difference was observable for other mucin-like genes, including ASGP-2, CD34, CD43 (leukosialin), glycophorin, and Mad-CAM-1. The expression level of GlyCAM-1 was elevated in several outbred homozygous animals, but this apparent increase in expression was not consistent in inbred homozygotes, nor did it appear to correlate with an increase in GlyCAM-1 protein levels in milk. In addition, no difference was observable in the expression of Thbs-3, although the expression of this gene did appear to be highly variable in the tissues tested (Fig. 5).
Primary 35 female Muc-1 ϩ/ϩ mice were utilized. All mice were virgin females positive for the polyoma virus middle T antigen (MTag) transgene. Fifty percent of mice in this study developed palpable lesions of the mammary gland by 68 days of age. There was no significant difference in the rate of appearance of palpable lesions between Muc-1 mutant and wild type mice. Tumors appeared in 100% of wild type mice and in 98% of mutant mice. Tumors in Muc-1 mutant and wild type mice had similar histological appearances and were poorly differentiated adenocarcinomas (Fig. 6A). Pathological analysis showed that the tumors were high grade, based on the high mitotic rate, the solid growth pattern, and the presence of necrosis, and appeared to be typical of tumors generated using the MTag transgene. Immunohistochemical analysis using antiserum directed to the Muc-1 cytoplasmic tail showed that tumors that developed in Muc-1 ϩ/ϩ animals expressed high levels of Muc-1 (Fig. 6B). Interestingly, tumor growth rate differed significantly between the two groups (Fig. 6C). As early as 104 days of age, Muc-1 Ϫ/Ϫ mice had significantly smaller tumors than did mice with wild type Muc-1 alleles (p Ͻ 0.05), and by the 124-day end point, differences in tumor size were highly significant (p Ͻ 0.001) (two-sample t test).
Although a lack of Muc-1 protein would be expected to affect the metastatic cascade, no significant differences in rates of metastasis were observed. Overall, 58% of mice developed grossly observable lung metastases, with 53% of Muc-1 Ϫ/Ϫ mice and 67% of Muc-1 ϩ/ϩ mice developing metastases (Fig.  6D). Although this difference suggests a trend toward de-   creased rates of tumor metastasis in Muc-1 Ϫ/Ϫ mice, it was not statistically significant as assessed by -square analysis (p Ͼ 0.12). However, based on the sample sizes in this study, the power to statistically detect the observed difference was only 33%. It is possible that with a large sample size, this difference in metastatic rate would be statistically significant. It should be noted that although the rate of metastasis observed in this study was lower than that previously reported in MTag mice (28), this may be a consequence of the alternate type of analysis used to detect metastasis, microscopic examination versus the RNase protection assay.

Absence of a Developmental Phenotype in Muc-1 Ϫ/Ϫ Mice-
The ability to create mice that possess deficiencies in specific genes is providing important insights into the physiological role played by specific proteins both during embryonic and postnatal development and during adult life. To investigate the biological function of the Muc-1 membrane glycoprotein, we created a null mutation in the Muc-1 gene using homologous recombination in mouse ES cells. Mice were subsequently derived that were deficient in Muc-1 (Figs. 3 and 4). We and others had postulated that Muc-1 on the apical surface of an aggregation of differentiating epithelial cells may repel adjacent cells or mask adhesive molecules, thus promoting the formation of a lumen (16,21). However, we were surprised to find that, despite the widespread expression of Muc-1 during epithelial organogenesis, Muc-1-deficient mice were obtained at the expected frequency and appeared normal in all respects. In addition, Muc-1-deficient mice were fertile and produced and weaned litters of average size. Pathological analysis of hematoxylin-eosin-stained sections of all the major organs of Muc-1 mice failed to indicate any significant differences between homozygous mice and their heterozygous and wild type littermates. Furthermore, crossing of the disrupted Muc-1 gene into inbred 129SV/J and C57BL/6J backgrounds produced mice that appeared normal in all respects.
Other Mucins and Mucin-like Genes May Compensate for Muc-1 Deficiency-We hypothesized that the lack of a specific phenotype in Muc-1-deficient mice might be due to the specific up-regulation of expression of another mucin or mucin-like gene. In humans, eight mucin genes have now been identified (reviewed in Ref. 4), and it is quite likely that there are other mucin genes yet to be discovered. To date, MUC1 is the only human epithelial membrane-spanning mucin gene that has been cloned. Of the human mucin genes that have been identified, only two definite rodent homologues have been isolated, those for MUC1 and MUC2, respectively (13,14,(51)(52)(53)(54). However, a rat intestinal mucin, designated M2 has been isolated (55,56), which may be the rodent homologue of the human intestinal MUC3 gene, and a mouse gastric mucin gene has recently been reported that may represent the mouse Muc-6 gene. In addition, we have recently isolated genomic clones for the mouse homologue of the human MUC4 gene. 3 In an attempt to determine if a mucin-like molecule could potentially compensate for Muc-1 function, we obtained probes for a large panel of mucin and mucin-like genes. Slot-blot analyses of RNA were utilized to screen for their expression in Muc-1 ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ mice. Muc-2 was strongly expressed in the colon, and Muc-4 appeared to be strongly expressed throughout epithelial tissues (Fig. 5). The structure of the human MUC4 and mouse Muc-4 genes and their encoded proteins has not been fully determined, and it is therefore possible that Muc-4 may be structurally and functionally similar to 3 S. J. Gendler, unpublished data. Muc-1. Other genes that did not show any consistent alteration in expression included CD34, CD43 (leukosialin), glycophorin, ASGP-2, GlyCAM-1, and MadCAM-1. In addition, the expression levels of genes closely flanking the Muc-1 gene (Thbs-3 and Gene Y) were unaffected by the targeted inactivation of Muc-1 (Fig. 5). We are currently employing the differential display PCR approach (57) on mRNAs isolated from inbred mice in an attempt to identify potential compensating genes.
Muc-1: a Role in Tumor Development and/or Progression-It has long been known that the human MUC1 gene is highly expressed and often up-regulated in adenocarcinomas and metastatic lesions. In many carcinoma cells, polarization of the epithelial cells is lost, and the MUC1 protein can be detected on all cell surfaces, including those facing the stroma and adjacent cells. Under these circumstances, the antiadhesive property of MUC1 may destabilize cell-cell and cell-substratum interactions, thus promoting the disaggregation of a tumor site, leading to tumor spread and metastasis. Previous studies have suggested various roles for the MUC1 glycoprotein in facilitating tumor growth, including inhibition of cell-cell contacts (20,22), protection from recognition, and destruction by immune cells (58,59) and also serving as an E-selectin ligand to facilitate extravasation of metastatic cells from the bloodstream (25). In human colon cancers, levels of mature MUC1 mucins were significantly higher in primary tumors from patients having metastasis than in primary tumors from patients without F IG. 5. A Screen for genes potentially up-regulated in Muc-1deficient mice: slot-blot analysis to detect the expression of known genes. Equivalent amounts of total RNA were transferred to a nylon membrane utilizing a slot-blot apparatus. Duplicate membranes were hybridized with radioactively labeled probes for a large panel of mucin and mucin-like genes (Table II). Results are shown for Muc-1, Muc-2, Muc-4, and Thbs-3. At 104 days, Muc-1 Ϫ/Ϫ mice had significantly smaller tumors than did Muc-1 ϩ/ϩ mice (p Ͻ 0.05). By the 124-day end point, differences in tumor size were highly significant (p Ͻ 0.001). Asterisks indicate statistical significance. D, graph showing the percentage of Muc-1 ϩ/ϩ and Ϫ/Ϫ mice with metastatic lesions in the lung at 124 days. The trend toward decreased rates of tumor metastasis in Muc-1 Ϫ/Ϫ mice suggested that the lack of Muc-1 was showing some effects. Our sample size was not sufficiently large to reach statistical significance. metastasis (60). However, this is the first study to directly assess the role of the Muc-1 glycoprotein in spontaneous tumor development in the mammary gland. We demonstrated that the presence of the Muc-1 molecule is beneficial to the developing tumor, resulting in a tumor with a significantly faster growth rate than that seen in tumors that do not express the Muc-1 molecule. It is not clear at present how the Muc-1 molecule affects tumor growth rate in the primary tumor. However, it is likely that Muc-1 modulates the immune system in some way. It will be important to determine differences in natural killer and cytotoxic T cell activities at the tumor sites that might underlie the differential tumor growth rates. MUC1 protein can be detected in the circulation of patients with breast and pancreatic carcinomas (61)(62)(63)(64), and free MUC1 protein appears to inhibit the cytotoxic T cell lysis of target cells (65) and may be immunosuppressive (66). High levels of circulating Muc-1 might therefore block the specific T cell activity and thus aid the cells in escaping from T cell-mediated lysis. Studies to define the mechanism whereby Muc-1 expression facilitates growth of the primary tumor are currently under way.
The effect on metastasis in the Muc-1 Ϫ/Ϫ mice was less than expected. The trend toward decreased rates of tumor metastasis in Muc-1 null mice suggested that the lack of Muc-1 was showing some effects. However, our sample size was not sufficiently large to reach statistical significance. In addition, the rapid kinetics of the middle T tumor phenotype may have rendered some immune responses impotent. The middle T mice exhibit hyperplastic mammary glands as early as at 3 weeks of age and the rapid production of multifocal mammary adenocarcinomas by 35 days. In many animals tumors developed in every mammary gland. Alternatively, middle T may be modulating the immunogenicity of the tumor cells independently of Muc-1, or up-regulation of other molecules may compensate in the absence of Muc-1. Further analysis of the role of Muc-1 in metastasis may await a tumor model more relevant to human breast cancer. Such a study is in progress, utilizing mice with c-neu under control of the mouse mammary tumor virus LTR, which develop focal mammary tumors and metastasize after long latency (67).
In conclusion, Muc-1 null mice developed normally. The lack of an obvious developmental phenotype suggests that Muc-1 is not required for mouse development or that functionally similar proteins have been recruited to compensate for Muc-1 deficiency. Studies investigating this area are currently in progress. As mouse lines specifically deficient in other mucin and mucin-like molecules are created it will be important to analyze the possible phenotypic effects in double and multiple mucindeficient mice. Although normal development was not affected by a lack of Muc-1, mammary tumor growth was significantly reduced. This is a meaningful observation, as MUC1 was initially thought to be important through its role in breast cancer in humans. Clearly, the expression of Muc-1 by mouse mammary tumor cells confers upon them a selective advantage. Although a clear mechanism cannot be invoked at this time, Muc-1-deficient mice are expected to be critical in our future understanding of the role of Muc-1 in the development and progression of breast and other cancers. In addition to the work described herein, mucins have been postulated to be involved in adhesion and replication of viruses and bacteria, in receptivity of the uterus to blastocyst implantation, and in diseases such as asthma and cystic fibrosis. Muc-1-deficient mice will provide important models in the study of these aspects of mucin biology.