About 8 years ago, a mammary gland transcription factor was described independently in both our laboratory and that of Berndt Groner and named MPBF (for milk protein binding factor) or MGF (for mammary gland factor) respectively. Subsequently, cloning of MPBF/MGF revealed it to be a member of the family of STAT (signal transducer and activator of transcription) factors originally identified in Darnell. laboratory. This family of transcription factors comprises seven members which are localised in pairs on three different chromosomes (counting STAT5a and STAT5b, which are highly similar, as one gene). In evolutionary terms, STAT5 appears to be the ancestral gene having homologues in other species such as Drosophila and Dictyostelium. STATs are latent transcription factors which are sequestered in the cytoplasm in an inactive form. Upon ligand engagement of a cytokine (or in some cases) a growth factor receptor, the receptor subunits aggregate and initiate a cascade of tyrosine phosphorylation events in which the receptor-associated Jaks (Janus protein tyrosine kinases) become activated. This creates a docking site for the STAT which then becomes phosphorylated on a single tyrosine residue, dimerises and dissociates from the receptor and translocates to the nucleus where is interacts with its recognition site in the promoters of target genes. The specificity of STATs for receptors appears to depend on the cell type and a single ligand can activate a number of different STATs. The target genes are just beginning to be characterised but the milk protein gene promoters were among the earliest targets to be identified. Indeed, the original identification of STAT5 in the mammary gland was through its interaction with the β-lactoglobulin (BLG) and β-casein promoters.

3.  STATS IN THE MAMMARY GLAND     

      The first clue that STAT5 was important for mammary function came from mutational analyses of the BLG and whey acidic protein (WAP) promoters. Transgenic mice were generated which had mutations in combinations of the three STAT binding sites in the sheep BLG promoter (9) or the two sites in the rat WAP promoter. Abolishing STAT binding reduced expression of BLG by 80% and WAP by 90% but did not affect the tissue-specificity of expression. Mutation of a single site in the BLG promoter had little effect whereas mutation of any 2 sites dramatically reduced expression. It has recently been shown that binding of 2 STAT dimers to adjacent sites, where they form tetramers, induces maximal expression , The next step was to determine if other members of the STAT family were expressed in mammary tissue and whether their expression pattern mimicked that of STAT5. Interestingly, STAT5a/b are unique in their induction during pregnancy since both STATs 1 and 3 are expressed at fairly constant levels and STAT4 is barely detectable. STAT2 DNA binding activity could not be detected in extracts from lactating tissue although there is some evidence that STAT6 may have a role in mammary gland.Analysis of protein levels and DNA binding activities showed that STATs 3 and 5 have a reciprocal relationship suggesting that they play very different roles in mammary gland. Of particular interest is the activation of STAT3 at the beginning of involution indicating that it may be a death signal for the epithelial cells. The subsequent knockout of both STAT5 genes showed unequivocally that STAT5a, at least, is required for mammopoeisis and not suprisingly, given its requirement for efficient milk protein gene expression, for lactogenesis. The role of STAT5b is less clear although it is worth noting that a high proportion of STAT5 DNA binding activity in mammary tissue is composed of heterodimers and 5a and 5b despite the lower levels of STAT5b protein. However, the knockout of the STAT3 gene proved to have an embryonic lethal phenotype, with death occurring around day 5 of gestation. The demonstration, in ES cells, that STAT3 is required for stem cell renewal explains this very early lethality. In order to address the role of STAT3 in mammary involution, it was therefore necessary to generate a conditional knockout.

4.  THE LOW/CRE RECOMBINATION SYSTEM AND GENERATION OF BLG/CRE TRANSGENIC MICE

      A number of approaches have been devised to facilitate conditional gene targeting. These include the lox/Cre recombination system from P1 bacteriophage, and the yeast Flp system. The most extensively used of these is lox/Cre. This is summarised in Fig. 1 and reviewed. The essential components of this binary system are a transgenic line harbouring a Cre recombinase/tissue-specific promoter fusion construct and a floxed knockout line containing ideally one null (non-functional) allele of the chosen gene (in this case STAT3) and one allele with 2 loxP sites inserted within introns a suitable distance apart. We first generated a transgenic line with mammary-restricted Cre expression.

Conditional Gene Targeting using Cre/lox Recombination

                                    
                                       Fig. 1 STAT3 deleted only in mammaryepithelial cells

      The BLG promoter has been used extensively, primarily in John Clark’s laboratory, to drive expression of genes specifically to the secretory epithelial cells of mammary gland during late pregnancy and lactation. This promoter was chosen to drive expression of Cre recombinase to mammary gland. Variable expression in the 15 Cre-positive transgenic lines generated was observed and a subset of these were analysed further.

      The specificity of expression of the Cre transgene was examined by RT-PCR and mammary-restricted expression observed in a number of these. The two lines with the highest, mammary restricted expression were crossed to a reporter strain harbouring a floxed DNA ligase I (LigI) allele to determine functionality and specificity of the Cre recombinase. Using Southern blotting and a quantitative PCR based approach to detect the excision event, we observed 70-80% recombination in the lactating mammary gland in crosses with one of the lines (line BLG-Cre 74) with less than 1.1% in all other tissues. Lower levels of recombination occurred in mammary glands of virgin animals (7%) and during pregnancy (20-30%).Immunohistochemistry and in situ PCR show that Cre mediated recombination is very likely restricted to the epithelial cells of the mammary gland which are embedded therefore in a wild-type stroma¹⁹. Conditional gene targeting using this BLG-Cre line will allow thus the role of genes in lactogenesis or apoptosis of the epithelial compartment to be investigated.

5.  CONDITIONAL STAT3 KNOCKOUTDue to the early embryonic lethality of STAT3, a floxed strain was generated, in the laboratory of Shizuo Akira, in which an exon of the essential SH2 domain was flanked by loxP sites. Both floxed and null STAT3 animals were crossed with the BLG-Cre transgenic line 74 to generate animals of the desired genotype. These were then analysed for phenotype during lactation and involution.Given the activation profile of STAT3, it was anticipated that the most dramatic phenotype would be observed during involution and indeed this was the case. Fig. 2 shows H & E staining ofmammary sections from day 3 involution time points. It is striking that involution is not initiated in the conditional knockouts for at least 3 days following the removal of pups and that the extent of apoptosis in the epithelial cells is reduced. Mammary glands in conditional knockout animals do eventually undergo remodelling demonstrating that a compensatory mechanism(s) exists.

                                                                    STAT3 floxed/BLG-Trc 3 day involution
                                                       
Fig. 2. Mammary phenotype of Stat3 knockout mice. H & E stained sections of mammary glands from control (right) and conditional STAT3 knockout (left) mice at day 3 of involution.

         Analysis of changes at the molecular level shed some light on possible mechanisms for initiating this delayed involution process and apoptosis. Of particular interest is the precocious activation of STAT1 , a dimerisation partner for STAT3, which is often activated in concert with it (Fig. 3). Perhaps STAT1 can activate a similar set of target genes. This speculation can be addressed by generating double knockouts of both STATs 1 and 3 and by using microarray analysis to identify downstream target genes of these transcription factors.

6. STATS AND BREAST CANCER     

     What is the relevance of this work for breast cancer? We demon­strated a number of years ago that constitutive activation of STATs 1 and 3, but not STAT5, is a feature of invasive breast carcinoma but not in situ or benign disease . Constitutive STAT activation has subsequently been shown to be associated with transformation by Src, Abl, and other oncoproteins and tumour viruses in a variety of cell types. More recently, STAT factors have been shown to be involved in apoptosis. STAT3 is required for cytokine-induced apoptosis in myeloid leukaemia cells but can suppress apoptosis in a pro-B cell line . Similarly, STATs 1 and 5 have been implicated in cell survival and death . Truncated, dominant-negative STAT mutants can block these effects and in this context it is worth noting that dominant-negative  mutants  can  occur naturally.Cancer may be viewed as a failure of apoptosis. Our observations on the conditional STAT3 knockout demonstrate that STAT3 is required for apoptosis. It will be interesting to determine if these mice are more susceptible to tumour development following pregnancy or if deletion of STAT3 has a protective effect. In light of the opposing roles of different STATs in the mammary epithelial cell, it may be necessary to perturb the function of more than one member of this family to cause cancer. Crosses of knockouts of STATs 1 and 3 will be particularly interesting. The undoubted role of the JaW/STAT pathway in apoptosis and transformation places STAT factors in the range of therapeutic targets for breast cancer and high throughput analysis is being used by drug companies to discover inhibitors of STAT function.

7.  THE WAY AHEAD

     We have presented an example of conditional gene targeting in the mammary gland using the lox/Cre recombination system and a BLG-Cre transgenic line. This approach is applicable to any gene of interest and we are currently generating floxed Jak2 animals to overcome the embryonic lethal phenotype and investigate its role in mammary development. Since Jak2 lies upstream of both STATs 3 and 5 and is implicated in other signalling pathways, a mammary conditional Jak2 knockout can be expected to have a different phenotype from both the STAT3 and STAT5 knockouts.The BLG-Cre transgenics from our laboratory and the WAP-Cre transgenics from the Henninghausen laboratory will be a valuable resource to those interested in the conditional knockout of any one of the number of genes which have already been implicated in breast cancer. However, a further level of sophistication in conditional gene targeting is desirable to provide both tissue-specificity and temporal regulation. Strategies to achieve this have been devised and the new millennium should herald the arrival of finely tuned gene deletion.