Oncogene (2002) 21, 1009 ± 1016 ã 2002 Nature Publishing Group All rights reserved 0950 ± 9232/02 $25.00 www.nature.com/onc

Expression of the PAX2 oncogene in human breast cancer and its role in progesterone-dependent mammary growth

Gary B Silberstein*,1, Gregory R Dressler2 and Katharine Van Horn1

1Department of Molecular, Cell and Developmental Biology, Sinsheimer Laboratories, University of California, Santa Cruz, California, CA 95064, USA; 2Department of Pathology, University of Michigan, Ann Arbor, Michigan, MI 48109, USA

In this study, we ®rst describe expression of the paired known to regulate several mammotrophic growth factor domain transcription factor PAX2 in the normal and pathways, including insulin-like growth factor, epider- cancerous human breast, then demonstrate in a murine mal growth factor, and transforming growth factor beta model a novel function for PAX2 in the regulation of (Dey et al., 1994; Drummond et al., 1992; Vicanek et al., progesterone stimulation of secondary ductal growth. In 1997; Werner et al., 1993). WT1 is also expressed in the human mammary tissue, PAX2 expression was coin- kidney, where it is positively regulated by the paired- cident with sub-populations of mammary ductal cells, domain transcription factor PAX2 and is required for some of which possessed an undi€erentiated histiotype, the metanephric mesenchyme to respond to signals and was also found in 450% of the human breast inducing cell proliferation (Dehbi et al., 1996; Kreidberg tumors surveyed (n=38). In the mouse, mammary et al., 1993). During embryonic kidney development parenchyma with a targeted deletion of PAX2 developed PAX2 expression is transient whereas persistent expres- normal ductal systems when grafted into wild-type host sion of the gene in Wilms' kidney tumors (Dressler and mammary fat pads, but failed to undergo higher order Douglass, 1992), renal cell carcinoma (Gnarra and side-branching and lobular development in response to Dressler, 1995), and polycystic kidneys (Ostrom et al., progesterone. A previously unsuspected PAX2/WT1 2000) suggests multiple roles in maintaining the dedi€er- (Wilms' tumor suppressor gene) regulatory axis in the entiated or proliferative state and preventing apoptosis mammary gland was also indicated. Using RT ± PCR, a of the renal epithelium. Deregulated expression of PAX2 signi®cant reduction in WT1 mRNA expression was can transform cells in vitro and promote tumor detected in the PAX2 mutant glands compared to wild- formation in nude mice, indicating a direct role for type counterparts and double-antibody immunohisto- PAX2 in tumorigenesis (Maulbecker and Gruss, 1993; chemistry detected the co-localization of PAX2 and Stuart and Gruss, 1996). PAX2 expression in the WT1 in the nuclei of normal and cancerous breast cells. mammary gland and its potential relationship with These data indicate a role for PAX2 (and possibly WT1) WT1 in that tissue has not been previously investigated, in the regulation of the progesterone response of the in part due to the death of PAX2 mutant mice within the mature mammary gland. The potential contribution of ®rst hours after birth. Here we report the unexpected PAX2 to breast tumor pathogenesis is discussed. expression of PAX2 in normal human breast tissue, as Oncogene (2002) 21, 1009 ± 1016. DOI: 10.1038/sj/onc/ well as in a high percentage of breast tumors. To begin to 1205172 understand a functional role, we have rescued the mammary anlagen from PAX2-null mouse embryos to Keywords: breast cancer; mammary; PAX2; progester- wild-type host mammary glands and now report that one; Wilms' tumor PAX2 is necessary for progesterone-dependent growth of lateral ductal branches.

Introduction Results Previously, we demonstrated that the WT1 Wilms' tumor suppressor gene was expressed in normal human Expression of PAX2 in the normal and cancerous human breast tissue as well as in a high percentage of breast breast tumors (Silberstein et al., 1997). The WT1 protein may act as an activator or suppressor of transcription and is The expression of PAX2 in the human breast was investigated by immunohistochemistry and in situ hybridization. Using a polyclonal anti-PAX2 antibody (Dressler and Douglass, 1992), nuclear staining was *Correspondence: GB Silberstein, Sinsheimer Laboratories, detected in cells from the two primary developmental University of California, Santa Cruz, CA 95064, USA; lineages comprising myoepithelial and luminal cells in E-mail: [email protected] Received 11 October 2001; revised 5 November 2001; accepted 7 alveolar (Figure 1a) as well as ductal (Figure 1b) November 2001 elements of the tissue. These cells are arranged in two PAX2 mammary expression and function GB Silberstein et al 1010

Figure 1 PAX2 expression in normal and cancerous human mammary tissue. The nuclei of immuno-positive cells stained brown; counter stain: methyl green. No counter stain was used for the in situ hybridization, (a) Lobule cluster (cross- and grazing sections). PAX2 positive cells with oval as well as polygonal nuclear morphology (arrows within the lumens, L) lined the lobule lumens and were often adjacent to unstained cells. Immuno-positive myoepithelial cells (arrows on periphery of lobules) and immuno-negative counterparts (arrowheads) were also detected (Bar=20 mm). (b) Duct (longitudinal section of one side of a duct; high magni®cation). The arrangement of PAX2 immuno-positive and -negative cells mimicked pattern seen in lobules with strongly staining luminal cells often adjacent to non-staining cells (boxes). Myoepithelial cells (outside dotted line) stained lighter than ductal cells (small arrow) or were negative for PAX2 (arrowhead). (Bar=5 mm). (c) In situ hybridization localization of PAX2 mRNA in normal mammary duct (small arrow) and side-branch (large arrow). Control hybridization with sense-strand probe (inset). (Bar=15 mm). (d) Ductal carcinoma in situ. PAX2 positive cells (arrows); PAX2 negative cells (arrowheads). (Bar=20 mm). Inset: PAX2 mRNA in situ hybridization. Cells expressing PAX2 mRNA (arrows) surround PAX2 non-expressing cells (arrowheads). (Bar=25 mm). (e) In®ltrating ductal carcinoma. In®ltrating tumor cells had large, anaplastic nuclei (large arrows) or small nuclei with compact chromatin (small arrows); both populations were negative for PAX2 protein. Inset: PAX2 positive cells (arrows) in vicinity of main tumor (Bar=20 mm)

Oncogene PAX2 mammary expression and function GB Silberstein et al 1011 circular layers and form a tube-within-a-tube. The Table 1 Survey of PAX2 protein expression in normal and innermost cells line the lumen, forming the ductal or cancerous breast tissue acinar tube proper, and give rise to secretory structures Tissue PAX2 positive a PAX2 negative b PAX2-neg/pos c at pregnancy, while the myoepithelium forms the Tumor 40% (15/38) 47% (18/38) 13% (5/38) epithelial boundary with the stroma and is comprised Vicinal d 20 1 - of contractile cells that facilitate milk extrusion (Silberstein, 2001). aTumors with 80% or more of cells immunopositive for PAX2, b The histiotype of certain PAX2 positive cells with estimated by inspection. PAX2 immunostaining was absent in greater than 90% of the tumor cells. cTumors with an approximately large, oval nuclei and di€use chromatin (Figure 1b) equal mixture of PAX2 negative and positive cells. dTissue that was was consistent with that of putative mammary stem excised from the vicinity of a tumor, contained normal-appearing cells (Smith and Medina, 1988; Stingl et al., 1998; structures and cells that exhibited ductal and lobular PAX2 staining Toker, 1967; Williams and Daniel, 1983). PAX2 patterns as in Figure 1a,b. Tumors studied: n = 38; classes represented: 27 in®ltrating ductal carcinomas, seven in®ltrating positive luminal cells with this nuclear morphology lobular carcinomas, medullary, colloid and adenocarcinoma, one coexisted with cells not ®tting this histiotype (Figure each, and one unclassi®ed 1a). The not-infrequent side-by-side arrangement of PAX2 positive and negative cells (Figure 1b, boxes) may have functional signi®cance as paracrine signaling grade and steroid receptor positive tumors were is known to be important in progesterone-dependent roughly equally divided between PAX2 positive and lateral branching (Brisken et al., 1998, 2000). Myoe- negative tumors. High grade and receptor negative pithelial cells also exhibited a mixed, PAX2 positive tumors were more likely to be PAX2 negative; 86% of and negative staining pattern (Figure 1a,b), with some Grade 3 tumors and 75% of receptor negative tumors myoepithelial cells having less immunoreactivity than were PAX2 negative. WT1 and PAX2 expression were adjacent luminal cells (Figure 1b). The localization of strongly correlated; 72% of PAX2 positive tumors PAX2 mRNA in mammary tissue was detected by in were WT1 positive, while 85% of WT1 negative situ hybridization and re¯ected PAX2 immuno-stain- tumors were also PAX2 negative. This study added ing. In the pictured section of a branched duct PAX2, 18 more cases to the patient population originally for example, PAX2 mRNA was found throughout the screened for WT1 expression (Silberstein et al., 1997). structure, with possibly higher amounts in the tips of All 18 cases showed some degree of WT1 staining, developing branches (Figure 1c). No signal was seen supporting recent ®ndings that a much higher with a sense probe (Figure 1c, inset). percentage of breast tumors are WT1-positive than Once we established the expression of PAX2 in the was originally reported (Loeb et al., 2001). normal breast, mammary tumors were examined. Carcinoma in situ is one of the earliest histologically Transplantation rescue and progesterone insensitivity of identi®able stages in breast tumor progression, in PAX2-null mammary glands which proliferating cells occlude the lumen of a duct (Figure 1d). In this example, the tumor cell mass To investigate the function of PAX2 in the mammary consists primarily of PAX2 positive cells. The pattern gland, we utilized the PAX2 null allele generated by of PAX2 mRNA expression re¯ected the immunohis- homologous recombination. Mice homozygous for the tochemistry, with small groupings of cells exhibiting PAX2 null allele die post-natally due to exencephaly little or no PAX2 mRNA interspersed with cells and renal insuciency (Torres et al., 1996). PAX2 null expressing high levels (Figure 1d, inset). An in®ltrating mammary anlagen were therefore rescued by grafting ductal carcinoma (Figure 1e) consisted of at least two into wild-type mammary fat pads in immune compro- cell populations, one with large, anaplastic nuclei, the mized mice. PAX2 null mammary outgrowths were second with smaller nuclei that were also abnormal; then divided and re-transplanted into multiple hosts for both cell types were PAX2 negative. A nearby, possibly detailed analysis. In order to investigate the e€ects of less advanced element of this tumor, contained a PAX2 ablation on mammary growth and gene mixture of PAX2 negative and positive cells (inset expression, fragments of PAX2 null and homotypic Figure 1e). wild-type mammary tissue were simultaneously trans- In a more comprehensive survey, mammary carci- planted into contralateral mammary glands of indivi- nomas from 38 patients were grouped based on dual hosts. This bilateral transplant experimental whether specimens had high or low numbers of design ensures that mutant and control tissues are PAX2 positive epithelial cells (Table 1). All histo- exposed to identical endocrinological and physiological pathological types and grades of tumor were repre- backgrounds. Transplants were allowed 12 weeks to sented, and patients ranged in age from 19 to 88 years. establish and grow, after which animals were treated PAX2 positive tumors were observed in 53% of these with progesterone and estrogen for 7 days prior to cases, with ®ve tumors having more or less equally sacri®ce. mixed populations of PAX2 negative and positive cells. Progesterone, but not estrogen, stimulates the No correlation was noted between PAX2 staining and growth of the higher order, lateral branching that is ductal versus lobular tumors. Pathological grading and commonly associated with pregnancy in the mouse estrogen/progesterone receptor expression status was (Atwood et al., 2000). When PAX2 null mammary available for a subset of the tumors (Table 2). Lower glands were stimulated with a mixture of progesterone

Oncogene PAX2 mammary expression and function GB Silberstein et al 1012 Table 2 PAX2 expression in breast tumors relative to tumor grade, steroid hormone receptor status, and WT1 expression Pathological Grade a ER/PR Status b WT1 c PAX2 1 ± 2 3 + 7 + 7

Positive 57% (12/21) 14 (1/7) 47% (7/15) 25 (2/8) 72% (18/25) 15 (2/13) Negative 43 (9/21) 86 (6/7) 53 (8/15) 75 (6/8) 28 (7/23) 85 (11/13)

a28 tumors were screened for pathological status; the criteria for immuno-positive and -negative status is de®ned in Table 1, footnote. bER, estrogen receptor; PR, progesterone receptor: 23 tumors were screened for estrogen and progesterone receptor status. Concordance of Grade 3 and receptor-negative status was 100% (®ve out of ®ve of the tumors that comprised an overlapping set, e.g. were scored for each condition.) cWT1 immuno-stained tumors. Tissue sections were cut from the same paran-embedded specimens that were used for the PAX2 study (Table 1)

and estrogen the loss of PAX2 function resulted in epithelium grew into wild-type stroma the absence of stunted to non-existent lateral branching and poor an ampli®cation product also means that the bulk of lobular development, evidenced by the open architec- PAX2 mRNA is epithelial and not stromal in origin. In ture of the mutant gland (Figure 2a) compared to its the PAX2 null glands, there was a signi®cant reduction wild-type counterpart (Figure 2b). In contrast to e€ects in WT1 mRNA apparently con®rming the predicted on lateral branching and lobular development, growth positive regulatory relationship between the two genes of the primary ductal framework was una€ected; (Figure 3a). PCR primers spanning the alternative mutant and wild-type ducts grew the full length and splice site in the WT1 regulatory domain (exon 5) were breadth of the fat mammary fat pad. This point was used and demonstrated that expression of both the plus con®rmed by examination of transplants after 7 weeks, and minus exon 5 isoforms were a€ected by PAX2. at which time it was noted that mutant and wild-type Similar results were seen with primers spanning a splice ducts had extended an identical distance from their donor site (KTS) in the DNA binding domain (data origin and had similar numbers of end buds (ductal not shown). Starting RNA concentrations were growth points) of normal histo-morphology (not approximately equivalent as indicated by b-actin shown). expression (Figure 3c). Examination of a€ected glands under higher magni- The relationship between PAX2 and WT1 was ®cation revealed that duct tips in the mutant tissue had further investigated by double-antibody immuno-¯uor- failed to branch into the fatty stroma (compare boxed escence. In the normal duct, PAX2 and WT1 co- zones in Figures 2c,d). In addition, the larger ducts had expression was detected in luminal and myoepithelial numerous, compact outgrowths (Figure 2c, large cells (Figure 3c,d), and in tumors, the nuclei of arrows) comprised of epithelial cells lining a lumen numerous in®ltrating cells contained both proteins (Figure 2e). While not overtly dysplastic, these (Figure 3e,f). As noted earlier (Table 2), PAX2 and structures were unusual. The tips of PAX2 null ducts WT1 expression were correlated in a high percentage of failed to undergo normal branching development breast tumors. While suggestive, this positive correla- (compare Figure 2c and d, small arrows). This was tion does not comment on same-cell expression because con®rmed at the histological level, where extensive for the tumor survey immunostaining for each protein normal branching at duct termini (Figure 2g, top) was examined in separate tissue sections. contrasted with simpli®ed termini in the mutant tissue (Figure 2f). The accretion of ®brous stroma around the mutant ducts and smaller branches was similar to that Discussion seen in the wild-type tissue. It should be noted that some normal-appearing lobules were found in mutant A function for PAX2 in the mammary gland had not tissue, usually in the vicinity of the transplant origin. been surmised in the previous investigations of gene This probably re¯ects known regional di€erences in expression patterns or mutant analyses. The early ductal sensitivity to hormones, in which the most death of PAX2 mutant mouse newborns precludes di€erentiated ducts (in the vicinity of the nipple in any phenotypic characterization of mammary develop- intact glands) require the least complex mixture of ment directly. Once expression of PAX2 in the normal hormonal and growth factor stimulation (unpublished human mammary gland and in a high proportion of observations and DuBois and Elias, 1984). In a second human breast cancers had been established (Figure 1, host strain (Balb/c, nu/nu), the PAX2 null phenotype Table 1), we investigated the PAX2 mammary was sometimes less pronounced indicating that stromal phenotype by grafting PAX2 null embryonic glands and physiological background play a role. into normal hosts. Whereas PAX2 mutant mammary epithelium underwent normal ductal development, as indicated by elongation and branching that was PAX2 and WT1 expression are coordinated and comparable to the wild-type, the injection of progester- co-localize in human mammary cells one with estrogen failed to stimulate typical secondary The PAX2 null status of mammary tissue derived from branching and lobular growth (Figure 2). The mutant a rescued anlage was con®rmed by RT ± PCR using phenotype is consistent with higher levels of PAX2 RNA isolated from bilaterally transplanted mutant and expression in and around the branch points (Figure wild-type mammary glands (Figure 3a). Since mutant 1b), and indicates an important function for PAX2 in

Oncogene PAX2 mammary expression and function GB Silberstein et al 1013

Figure 2 E€ects of the PAX2 null mutation on mouse mammary ductal and lobular growth. Contralaterally transplanted mutant and wild-type glands were taken from a single, bilaterally transplanted RAG1 mouse that had been treated with progesterone and estrogen for 7 days prior to sacri®ce. Whole-mount images of PAX2 null (a) and wild-type (b) mammary glands. Lateral (secondary) branching and lobular growth was considerably reduced in the mutant gland. Note that both the mutant and wild-type primary ducts ®lled the available fat pad. The mutant gland exhibited regional variation in secondary growth with some lobular growth nearest the site of transplantation (left side of image out of the picture; the direction of growth is from left to right as pictured). (Bar=2.5 mm). Details of branching and lobular growth in mutant (c) and wild-type (d) glands. The in-®lling of available fatty stroma by lateral branches and lobules was attenuated in the mutant (boxes highlight reduced lateral in-®lling in similar spaces in the mutant versus the wild-type gland. The ends of mutant ducts exhibited truncated branching compared with equivalent wild-type structures (small arrows) and unusual, compact outgrowths appeared on mutant ducts (c, large arrows). (Bar=625 mm). Histo- morphology of mutant (e, f) and wild-type (g) ducts. Mutant glands exhibited numerous compact outgrowths that arose from larger ducts (large arrows). The ends of wild-type ducts typically exhibit extensive tertiary branching (g, top half of image) that was absent in the mutant (f, small arrow). Stromal investment of mutant outgrowths (e, small arrow) compared to wild-type (g, small arrow). (Bar=200 mm) mediating the progesterone response in the mature wnt-4 and low levels of PAX2 can be detected in wnt-4 mammary gland. mutants, indicating a regulatory relationship (Dressler, The inhibition of side branching in PAX2 mutant 2002; Stark et al., 1994). Alternative explanations for mammary glands was similar to that seen in mammary the mutant phenotype include possible e€ects on glands in which the progesterone receptor or wnt-4 had mammary stem cell survival and faulty induction of been genetically ablated (Brisken et al., 1998, 2000). the progesterone receptor. In other developing systems Wnt-4 is a secreted glycoprotein that was recently the loss of PAX2 leads to stem cell death and shown to act under progesterone stimulation in a subsequent agenesis of the developing tissue (Brunner paracrine mode driving ductal cells into secretory et al., 1999; Ostrom et al., 2000; Porteous et al., 2000). di€erentiation (Brisken et al., 1998, 2000). In the If some PAX2 positive ductal cells are undi€erentiated, developing kidney, PAX2 appears to act upstream of secretory progenitors (Figure 1a), their absence in the

Oncogene PAX2 mammary expression and function GB Silberstein et al 1014

Figure 3 Expression of PAX2 and WT1 in normal or tumorous mammary tissue. RT ± PCR analysis of PAX2, WT1, and b actin gene expression in PAX2 null and wild-type mouse mammary glands. (a) PAX2 and WT1. No ampli®cation signal was detected for PAX2 in the mutant mammary gland (KO) after two rounds of PCR with nested primers whereas a signal of predicted molecular size was detected in the wild-type (Cont). Ampli®cation of WT1 was reduced in KO compared with wild-type preparations for each of the exon 5 alternative splice isoforms. (b) b-actin. Ampli®cation products of the predicted size for the b-actin primers were detected mutant and wild-type preparations in approximately equivalent concentrations over a 10-fold dilution range. (c ± f) Double immuno-¯uorescence detection of PAX2 and WT1 expression in normal and tumorous human breast tissue: PAX2, (c and e); WT1, (d and f), (c and d) normal duct with myoepithelial cells, to the right of dotted line, and luminal cells expressing both PAX2 and WT1 (arrows); cell expressing PAX2 only (arrowhead, c); L, lumen. (e and f) in®ltrating ductal carcinoma cells strongly expressing both PAX2 and WT1 (arrows). (Bar=20 mm)

mutant could be expected to impede progesterone- The expression patterns of PAX2 and WT1 showed dependent development of lateral branching. If PAX2 a high degree of coincidence in the normal mammary is necessary for normal mammary tissue estrogen gland as well as in tumors, consistent with, but not responsiveness, then it is also possible that the loss of proving, a regulatory relationship between the two PAX2 could result in a failure of estrogen to properly genes. PAX2 expression was seen in a relatively high induce the progesterone receptor thereby accounting percentage of tumors (Table 1) and of these, over 70% for the PAX2 null phenotype. also expressed WT1 (Table 2). If these tumors A role for PAX2 in the pathogenesis and possibly represent clonal expansions of PAX2 or WT1-positive the ontogeny of breast cancer is also suggested by our ductal cells, this would indicate the special vulnerability studies. As seen in polycystic kidney disease, the of these cells to transformation. PAX2 null glands persistent (deregulated) expression of PAX2 contri- showed a marked reduction of WT1 mRNA expression butes to pathogenesis by the combined mechanisms of that is consistent with a relationship that was preventing di€erentiation and interfering with apopto- previously established in the kidney (Figure 3a). Two- sis (Ostrom et al., 2000). Apoptosis is an important antibody immunostaining demonstrated co-localization aspect of breast cancer pathogenesis that presents a of PAX2 and WT1 in the nuclei of luminal and complex and somewhat counter-intuitive picture. The myoepithelial cells of the normal duct as well as in cells highest rates are seen in the most aggressive, highly of an in®ltrating breast tumor supporting a potential proliferative tumors, which were mainly PAX2 nega- direct regulatory relationship between the two genes in tive, while lower grade tumors, which are frequently the mammary gland (Figure 3b). Alternatively, lower PAX2 positive (Table 2), have relatively low rates WT1 mRNA levels (Figure 3a) could also be due to (Lipponen, 1999). If PAX2 does a€ect mammary selective death of PAX2/WT1 double-positive cells, apoptosis, its expression could contribute to the leaving only the WT1 single-positive cells remaining in pathogenesis of some breast tumors by blocking the mutant gland. apoptosis early in progression or, by its absence during The e€ect of progesterone on normal and neoplastic later stages, it could favor the selection of apoptosis- mammary epithelial cells is paradoxical. During resistant cells (Zhang et al., 1998). Thus, PAX2 pregnancy, for example, the hormone simultaneously expression in breast tumors may contribute not only stimulates cell proliferation and lobulo-alveolar di€er- to proliferation, but also to cell survival by preventing entiation and, depending on dosage and cell type, it di€erentiation and apoptosis. Given its persistent can either stimulate or inhibit mammary tumor cell expression in a high percentage of tumors, a role for proliferation. Recent models now suggest that proges- PAX2 in tumor resistance to certain chemotherapies terone modulates a switching mechanism(s) that acts, that depend for their e€ect on the stimulation of in part, through nuclear transcription factors to control apoptosis is certainly worth investigating. the activity of key genes involved in breast cell fate

Oncogene PAX2 mammary expression and function GB Silberstein et al 1015 (Lange et al., 1999). While the elements of this system Immunohistochemistry are coming into focus, the transcription factor(s) that mediate the switching are unknown. In light of its role Using standard procedures, formalin-®xed, paran-em- in cell fate determination in other systems and its bedded tissue was sectioned at 7 mm, deparanized, and dramatic e€ect on progesterone responsiveness, we sections incubated with a rabbit anti-PAX2 polyclonal antibody (PRB-276P, Covance/BAbCo, Berkeley, CA, USA) suggest that PAX2 is a candidate to modulate fate at a 1 : 50 ± 75 dilution in PBS. WT1 single-antibody switching in progesterone-responsive mammary cells immunohistochemistry has been described (Silberstein et al., and thus merits further investigation. 1997). A biotin ± avidin secondary antibody system (Vector Laboratories, Burlingame, CA, USA) was used to produce the brown detection product. For double-antibody immunostaining, sections were in- cubated overnight in a mixture of the PAX2 polyclonal Materials and methods antibody and a WT1 monoclonal antibody, the latter at a 1 : 25 dilution (PharMingen, San Diego, CA, USA; clone Human mammary tissue 2C12). Fluorescein-conjugated avidin was used in conjunction Specimens of ®xed, sectioned breast tumors were provided by with biotinylated goat anti-mouse secondary antibody to the University of Michigan Breast Cell/Tissue Bank, where detect WT1. Vector red dye (Vector Labs, Burlingame, CA, histological grading and steroid hormone receptor status were USA) was used according to the manufacturer's protocol in determined. Additional specimens used for immunohisto- conjunction with biotinylated goat-anti rabbit secondary chemical or in situ hybridization analysis were obtained antibody to detect PAX2. A Leitz Aristoplan microscope directly after surgical excision, transferred immediately to interfaced with a CCD-array camera was used to capture chilled (48C) 4% paraformaldehyde in phosphate-bu€ered ¯uorescent images. saline (PBS) and ®xed for 3 h. Histological and steroid receptor typing of some of these specimens was performed by In situ hybridization Dr KR O'Keefe, Dominican Hospital, Santa Cruz, CA, USA. Pathological grading and steroid hormone receptor A detailed protocol for in situ hybridization to mammary tissue status were available only for a subset of samples. appears elsewhere (Silberstein et al., 1996). Digoxigenin-labeled anti-sense and sense-strand riboprobes were generated from a 527-bp BamHI ± EcoRI fragment of cDNA PAX2 clone c31A Rescue and clonal expansion of PAX2-homozygous recessive (Dressler et al., 1990) using 10X NTP labeling mixture (null) mouse mammary anlagen (Boehringer-Mannheim, Indianapolis, IN, USA). E17 embryos from PAX2 heterozygote crosses in mice were removed from the uterus and PAX2 homozygous recessive RNA preparation and RT ± PCR analysis mutants identi®ed by the characteristic exencephaly of the hindbrain and con®rmed by genotyping (Torres et al., 1996). The isolation of mammary gland RNA, reverse transcription, Dead mutant and companion embryos were then ¯own and PCR analysis was as described (Silberstein et al., 1997). overnight in ice-cold DMEM from the Dressler laboratory to RNA amounts were estimated by spectrophotometry and Santa Cruz. PAX2 null and wild-type mammary anlage were concentrations were equalized by dilution before ampli®cation. surgically removed from the embryos and, without culturing, PAX2 expression was analysed with primers designed to transplanted contralaterally into the inguinal fat pads of amplify mouse or human cDNA. To detect low message levels, immune-ablated, Balb/C nu/nu mice, whose endogenous nested primer sets were used in two rounds of ampli®cation. mammary parenchyma had previously been surgically excised One-sixth (3 ml) of the puri®ed DNA/RNA volume was (Robinson et al., 2000). To guarantee complete `clearing' of subjected to ®rst round PCR ampli®cation (948C for 1 min, the host parenchyma, each of the excised fragments was 588C for 2 min and 728C for 3 min for 30 cycles) using Taq stained and examined in whole-mount to ensure that it polymerase (Fisher Scienti®c) in a 20 ml reaction volume in a contained the host parenchyma in its entirety. To determine Perkin-Elmer model 9600 thermal cycler. For the second transplant success, anlage were given approximately 4 weeks round, a 1 : 10 dilution of the ®rst-round reaction mixture was to grow, after which fragments of each gland were removed made in a fresh reaction mixture containing the second set of without sacri®cing the host and examined in whole-mount. primers and ampli®ed for 30 cycles (948C for 1 min, 618C for Three independent sets of mutant and control transplants 2 min and 728C for 4 min). Primers for the initial ampli®cation were established, and one set was selected for clonal were: (forward) 5'-CGACAGAACCCGACTATGTTCG-3'; expansion by transplantation into RAG1 immune-ablated (reverse) 5'-TGAATCTCCAAGCCTCRTTGTAG-3'. The in- mice. ternal primers ¯anked the exon 10 alternative splice site and were (forward): 5'-GACC/TGGTCGTGAC/TATGA/GCGA- 3'; (reverse): 5'-TGTAC/TGGGTTGCCG/TGAGAACT-3'. Induction of ductal side-branching and lobular development For b-actin ampli®cation (948C for 1 min followed by 35 Transplants were allowed to grow for at least 6 weeks. To cycles of 948C, 558C and 728C for 1 min each concluding with a stimulate branching and lobular development animals were 10 min extension at 728C), primers were (forward): 5'- then treated with estrogen and progesterone (5 mg Depo- GCTGGTCGTCGACAACGGCT-3'; (reverse): 5'-ATGA- estradiol in cottonseed oil plus 5 mg Depo-provera; CCTGGCCGTCAGGC-3'. Pharmacia/Upjohn, Kalamazoo, MI, USA) for 7 days via a long-acting, subcutaneous injection. To investigate morphology, glands were stained with hematoxylin for whole-mount analysis (Silberstein and Daniel, 1982), Acknowledgments photographed, then sectioned at 7 mm and restained with We thank Drs J Snyder and KR O'Keefe (Dominican hematoxylin and eosin. Hospital, Santa Cruz); Mr D Albritton, Ms P Huerta and

Oncogene PAX2 mammary expression and function GB Silberstein et al 1016 Ms E Olson (University of California, Santa Cruz); Dr S supported by National Institutes of Health Grants DK- Ethier (University of Michigan Breast Tissue Bank) for 48883 (to Charles W Daniel, Santa Cruz) and D-54740 (to obtaining breast tissue samples and Drs Charles W Daniel, GR Dressler), Department of the Army Grant DAMD17- Daniel Medina and Charles Roberts Jr. for helpful 94-J-4230 (Charles W Daniel) and a gift from the Milton suggestions and review of the manuscript. This work was Meyer Foundation (to GB Silberstein).

References

Atwood CS, Hovey RC, Glover JP, Chepko G, Ginsburg E, Ostrom L, Tang MJ, Gruss P and Dressler GR. (2000). Dev. Robison WG and Vonderhaar BK. (2000). J. Endocrinol., Biol., 219, 250 ± 258. 167, 39 ± 52. Porteous S, Torban E, Cho NP, Cunli€e H, Chua L, McNoe BriskenC,HeinemanA,ChavarriaT,ElenbaasB,TanJ, L, Ward T, Souza C, Gus P, Giugliani R, Sato T, Yun K, Dey SK, McMahon JA, McMahon AP and Weinberg RA. Favor J, Sicotte M, Goodyer P and Eccles M. (2000). (2000). Genes Dev., 14, 650 ± 654. Hum. Mol. Genet., 9, 1 ± 11. Brisken C, Park S, Vass T, Lydon JP, O'Malley BW and Robinson G, Accili D and Hennighausen L. (2000). Methods Weinberg RA. (1998). Proc. Natl. Acad. Sci. USA, 95, in mammary gland biology and breast cancer research. Ip M 5076 ± 5081. and Asch B (eds). New York: Kluver Academic/Plenum. Brunner E, Brunner D, Fu W, Hafen E and Basler K. (1999). pp 307 ± 316. Dev. Biol., 206, 178 ± 188. Silberstein GB. (2001). Microsc. Res. Tech., 52, 155 ± 162. Dehbi M, Ghahremani M, Lechner M, Dressler G and Silberstein GB and Daniel CW. (1982). Dev. Biol., 93, 272 ± Pelletier J. (1996). Oncogene, 13, 447 ± 453. 278. Dey B, Sukhatme V, Roberts A, Sporn M, Rauscher III F Silberstein GB, Van Horn K, Shyamala G and Daniel CW. and Kim S. (1994). Molec. Endocrinol., 8, 595 ± 602. (1996). Cell Growth Di€er., 7, 945 ± 952. Dressler GR. (2002). Mouse development: patterning, mor- Silberstein GB, Van Horn K, Strickland P, Roberts Jr CT phogenesis, and organogenesis. Tam P and Rossant J. (eds). and Daniel CW. (1997). Proc. Natl. Acad. Sci. USA, 94, New York: Academic Press. pp 395 ± 420 8132 ± 8137. Dressler GR, Deutsch U, Chowdhury K, Nornes HO and Smith GH and Medina D. (1988). J. Cell. Sci., 89, 173 ± 183. Gruss P. (1990). Development, 109, 787 ± 795. Stark K, Vainio S, Vassileva G and McMahon AP. (1994). Dressler GR and Douglass EC. (1992). Proc. Natl. Acad. Sci. Nature 372, 679 ± 683. USA, 89, 1179 ± 1183. Stingl J, Eaves CJ, Kuusk U and Emerman JT. (1998). Drummond I, Madden S, Rohwer-Nutter P, Bell G, Di€erentiation, 63, 201 ± 213. Sukhatme V and Rauscher FJ III. (1992). Science, 257, Stuart E and Gruss P. (1996). Cell Growth Di€., 7, 405 ± 412. 674 ± 677. Toker C. (1967). J. Ultrastruct. Res., 21, 9±25. DuBois M and Elias JJ. (1984). Dev. Biol., 106, 70 ± 75. Torres M, Gomez-Pardo E and Gruss P. (1996). Devel., 122, Gnarra JR and Dressler GR. (1995). Cancer Res., 55, 4092 ± 3381 ± 3391. 4098. Vicanek C, Ferretti E, Goodyer C, Torban E, Mo€ett P, Kreidberg J, Sariola H, Loring J, Maeda M, Pelletier J, Pelletier J and Goodyer P. (1997). Kidney Int., 52, 614 ± Housman D and Jaenisch R. (1993). Cell, 74, 679 ± 691. 619. Lange CA, Richer JK and Horwitz KB. (1999). Mol. Werner H, Re G, Drummond I, Sukhatme V, Rauscher III Endocrinol., 13, 829 ± 836. FJ,DaS,GarvinA,LeRoithDandRobertsJrC.(1993). Lipponen P. (1999). Endocr. Relat. Cancer, 6, 13 ± 16. Proc. Natl. Acad. Sci. USA, 90, 5828 ± 5832. Loeb DM, Evron E, Patel CB, Sharma PM, Niranjan B, Williams JM and Daniel CW. (1983). Dev. Biol., 97, 274 ± Buluwela L, Weitzman SA, Korz D and Sukumar S. 290. (2001). Cancer Res., 61, 921 ± 925. Zhang GJ, Kimijima I, Tsuchiya A and Abe R. (1998). Maulbecker CC and Gruss P. (1993). EMBO J., 12, 2361 ± Oncol. Rep., 5, 1211 ± 1216. 2367.

Oncogene