Louisa Flintoft

Breast cancer: Recurrence

Types and treatment of breast cancer recurrence Recurrence describes the return of breast cancer after primary treatment.1 There are three types of recurrent breast cancer:1 • Local recurrence occurs when cancerous cells reappear at the original tumour site over time. Local breast cancer recurrence is not considered to be a spread of the cancer but rather due to failure of the initial treatment. Even after mastectomy, portions of breast skin and fat remain, making local recurrence possible, albeit uncommon. Rather, it is women treated with breast-conserving therapy and radiation who are at slightly higher risk of this type of recurrence. Treatment of locally recurring breast cancer depends on the initial therapy undertaken at the time of first diagnosis. If breast conserving surgery was originally performed, recurrent breast cancer will usually be treated with mastectomy. • Regional recurrence is more serious than local recurrence because it usually indicates that the cancer has spread past the breast and underarm (auxiliary) lymph nodes. Regional recurrences of breast cancer can occur in the chest muscles, in the internal mammary lymph nodes under the breastbone and between the ribs, in the nodes above the collarbone and in the nodes surrounding the neck. The latter two sites of regional recurrence tend to suggest more aggressive cancers. Overall, regional recurrence is very common, occurring in approximately 2% 5% of all breast cancer cases. Treatment can be complex however, including surgery to remove the cancerous node, radiotherapy, chemotherapy and adjuvant endocrine therapy depending on the previous treatment used. • Distant recurrence, also known as metastasis is the most serious type of recurrence and is associated with significantly lower survival. Having left the confines of the breast tissue, the cancer usually spreads first to the axillary lymph nodes. In 65-75%2,3,4,5 of distant recurrences the breast cancer then spreads from the lymph nodes to the bone. More rarely, the breast cancer may metastasise to other sites including the lungs, liver, brain or other organs. Surgery is rarely an option for metastatic breast cancer because the cancer is not usually confined to one specific site on a given organ. Instead, treatment approaches include chemotherapy, radiation therapy or endocrine therapy.

Bringing mice together

TP53 mutations are associated with many cancers, but are surprisingly rare in acute promyelocytic leukaemia (APL). Work from PierGiuseppe Pelicci, Saverio Minucci and colleagues now shows that an APLassociated oncoprotein provides a direct mechanism of p53 inactivation by promoting its deacetylation and subsequent degradation. Most cases of APL result from a chromosomal translocation that produces PML–RAR, a fusion of the PML and retinoic-acid receptor (RAR) proteins. This oncoprotein recruits histone deacetylases (HDACs) to chromatin to inhibit the expression of RAR target genes, predisposing cells to tumorigenesis. So, could PML–RAR also be responsible for p53 inactivation in APL? Pelicci, Minucci and colleagues showed that exposing mouse haematopoietic cells to genotoxic stress causes p53-dependent cell loss, but expression of PML–RAR protected cells from this effect. By contrast, no protection was provided by a mutant form of PML–RAR — known as the AHT mutant — that is unable to recruit HDACs. In addition, PML–RAR inhibited transcription from p53-responsive constructs in vitro, but transcription was restored using the AHT mutant. But how does PML–RAR inhibit p53 activity? Exposure of wild-type mice to genotoxic stress leads to increased p53 levels because of decreased proteasomal degradation. However, p53 levels in APL mice were shown to remain low after exposure to X-rays, indicating that PML–RAR might promote p53 degradation. Consistent with this, treatment of X-rayexposed APL animals with proteasome inhibitors led to increased p53 levels. The authors also showed that Mdm2, which targets p53 to the proteasome, is required for p53 downregulation by PML–RAR, as PML–RAR is unable to reduce p53 expression in Mdm2 –/– cells. In addition, Arf, an Mdm2 inhibitor, restored p53 expression that was blocked by PML–RAR. Acetylation is known to stabilize p53 in response to stress by inhibiting its proteasomal degradation. So, the authors tested whether HDAC recruitment by PML–RAR leads to deacetylation of p53. When wildtype mice are exposed to X-rays, acetylated p53 accumulates. However, this effect was not seen in APL animals, even when p53 was stabilized using a proteasome inhibitor. This inhibition of p53 acetylation requires HDAC recruitment, as treatment of APL mice with an HDAC inhibitor restored both total p53 and acetylated p53 to levels similar to those seen in wild-type mice. So, PML–RAR stimulates p53 deacetylation and promotes its Mdm2-dependent proteasomal degradation. This is in contrast with the previously known function of PML–RAR in inhibiting gene expression by deacetylating chromatin. Interestingly, this newly discovered function of PML–RAR requires wild-type Pml, which binds p53 and is involved in its normal regulation. The authors showed that PML–RAR binds p53 only weakly and is unable to deacetylate p53 or protect cells from genotoxic stress in the absence of Pml. Interestingly, this is the first known example in which an oncoprotein is dependent on expression of the corresponding wild-type protein for its tumorigenic function. Louisa Flintoft References and links ORIGINAL RESEARCH PAPER Insinga, A. et al. Impairment of p53 acetylation, stability and function by an oncogenic transcription factor. EMBO J. 19 Feb 2004 (doi:10.1038/sj.emboj.7600109) WEB SITES PierGiuseppe Pelicci’s lab: www.ifom-ieo-campus.it/groups/pelicci.html Saverio Minucci’s lab: www.ifom-ieo-campus.it/groups/minucci.html Taking the direct route O N C O G E N E S

Breast cancer: Establishing normality

www.nature.com/reviews/cancer Microarray studies have been used to classify breast tumours according to their expression of markers characteristic of different normal mammary epithelial cells. However, such classifications are limited by the fact that normal breast tissue has rarely been used for comparison with tumour samples. Researchers from the United Kingdom and Italy have now established a ‘baseline’ for gene expression in normal mammary epithelial cells, paving the way for advances in understanding and treating breast cancer. Jones et al. used tissue left over from breastreduction surgery to isolate the two main cell types that give rise to breast carcinomas — the luminal and myoepithelial cells of the ductallobular system — and carried out microarray analysis of gene expression in these cells. Initial analysis using an unsupervised hierarchicalclustering method revealed sets of genes that are expressed differentially between the two cell types. Supervised analysis using a trained algorithm then identified the genes with the greatest predictive value, establishing a set of 33 genes that can be used as markers to distinguish between myoepithelial and luminal cells. These differential expression patterns were confirmed by PCR after reverse transcription of RNA and antibody staining of tissue sections. To evaluate the prognostic usefulness of some of these markers, the authors used microarrays of breast tumour samples for which the clinical outcome was already known. Several showed potential for future use in predicting disease outcome. For example, nuclear expression of galectin 3, a protein with several proposed roles in cancer, and increased expression of SPARC, an extracellular-matrix protein associated with tumour invasiveness, both correlated with decreased survival rates. This study also raises concerns about the use of cultured cell lines to represent normal breast tissue in microarray experiments. Commercially available human mammary epithelial cells (HMECs) — derived from unsorted breast epithelium — have been used for comparisons with tumour cells in breast cancer microarray studies. However, it has been suggested that HMECs are derived mainly from myoepithelial rather than luminal cells, which could bias the interpretation of results. This was confirmed by comparing gene expression in HMECs from previous studies with the data obtained for normal luminal and myoepithelial cells. HMECs were found to express mainly myoepithelial markers, indicating that results from previous gene-expression studies using these cells should be re-interpreted. The gene-expression profiles established in this study now provide an accurate baseline for future microarray experiments. This should speed up progress in determining important changes during breast cancer development and identifying new targets for anticancer drugs. Louisa Flintoft References and links ORIGINAL RESEARCH PAPER Jones, C. et al. Expression profiling of purified normal human luminal and myoepithelial breast cells: identification of novel prognostic markers for breast cancer. Cancer Res. 64, 3037–3045 (2004) Although many targeted anticancer drugs prove effective when tested in biochemical assays, it is another matter to determine how they affect their molecular targets in vivo. In the June issue of Nature Medicine, William Kaelin’s lab reports a bioluminescent method of monitoring the in vivo efficacy of drugs designed to inhibit the cyclin-dependent kinase CDK2. CDK2 regulates cell-cycle progression and is therefore a potential anticancer-drug target. Among its many substrates, it phosphorylates the cyclin-dependent kinase inhibitor p27 (also known as KIP1). This phosphorylation leads to ubiquitylation and eventual proteolytic degradation of p27. Kaelin’s group created a plasmid vector that expressed a fusion of p27 and the enzyme luciferase, which can be tracked in vivo with bioluminescent imaging. This protein, termed p27Luc, behaves in a manner similar to p27 and can therefore be used as a marker of CDK2 activity. The authors transfected several tumour cell lines with the p27Luc-expressing vector. They observed that the luciferase activity in these cells increased when cells were treated with CDK2-inhibitory proteins, peptides or small inhibitory RNA, meaning that p27Luc was no longer degraded. Treatment of p27Luc-expressing cells with CDK2inhibitory drugs such as flavopiridol and Rroscovitine caused a dose-dependent increase in luciferase activity in these cell lines, whereas they had no effect in cells that simply expressed the luciferase gene. But can this system be used to monitor CDK2 activity in vivo? Kaelin’s group injected p27Luc-expressing lung carcinoma cells subcutaneously into nude mice, and imaged the resulting tumours 6 weeks later (see figure). They observed that flavopiridol induced p27Luc-mediated luminescence at the xenograft site. A particularly useful feature of this approach is the ability to take repeated measurements over extended time periods, allowing analysis of tumour growth and spread without sacrificing animals. So, luciferase reporters will be useful in monitoring the pharmacokinetics of CDK2 inhibitors, as well as other targeted therapeutics. Kristine Novak

Statins branch out

Statins — better known for their use in cutting cholesterol levels — might also significantly decrease the risk of developing some cancers, according to the results of two studies presented at the 40th annual meeting of the American Society of Clinical Oncology.

Carcinogenesis: Know your enemy

In order to defend yourself, it’s important to know what weapons your enemy has at their disposal. In the fight against cancer, information about the mechanisms by which carcinogens promote tumour development is essential for developing treatments that can protect against or counteract their effects. A paper by Barcellos-Hoff and colleagues now indicates that ionizing radiation can cause cancer by epigenetic mechanisms as well as by mutagenesis, providing a new insight into how best to protect against this carcinogen. Although the ability of ionizing radiation to induce DNA damage is the accepted basis for its carcinogenic properties, there is evidence to indicate that it might also contribute to tumour progression by disrupting normal cell–cell and cell–matrix interactions. The loss of such interactions is crucial for the development of tumours, as it removes the constraints on growth and motility that are usually imposed on a cell by its surroundings. Barcellos-Hoff and co-workers tested whether exposing human mammary epithelial cells (HMECs) to ionizing radiation induces morphological changes that might promote tumour progression independently of DNA damage. They exposed HMECs to sublethal doses of ionizing radiation and studied the effects on cellular interactions and morphogenesis in colonies that were produced from the irradiated cells. The authors saw a severe disruption of the organization of these cells into the acinar structures that are usually formed in mammary epithelial tissue. Importantly, this was seen in the progeny of irradiated cells for several generations after irradiation. The expression patterns of several proteins that are crucial for mediating normal cellular interactions were also abnormal in these colonies. E-cadherin and β-catenin, which are vital for cell–cell adhesion, and connexin 43, a component of gap junctions, were present at decreased levels at their normal sites of localization, indicating that ionizing radiation disrupts normal cellular contacts. These changes are unlikely to be due to the mutagenic effects of irradiation, as they were seen in almost all surviving cells — far more than would be expected to show these effects because of mutation. This indicates that ionizing radiation induces heritable changes in cellular organization that could promote tumour development independently of its mutagenic properties, predisposing cells to further malignant progression. These results indicate that therapies that are used to prevent tumour development caused by exposure to ionizing radiation should include agents that prevent or counteract the disruption to cellular organization, providing a new line of defence against this carcinogen. Louisa Flintoft

Staying active: Oncogenesis

www.nature.com/reviews/cancer One important aspect of tumour development is the ability of cancer cells to escape detection and destruction by the host immune system. Gabriel Rabinovich and colleagues provide evidence that mouse melanoma cells express a T-cell inhibitor called galectin-1 (Gal1), which could be a useful therapeutic target. Gal1 — a carbohydrate-binding protein with regulatory functions — is expressed by many different tumour types. Increased expression of Gal1 has been correlated with tumour aggressiveness and metastasis. The authors studied its function by transfecting B16 mouse melanoma cells — which express high levels of Gal1 — with Gal1 antisense cDNA to establish knockdown clones. These clones expressed either low or intermediate levels of Gal1. Conditioned medium from wild-type B16 cells induces high levels of apoptosis when it is applied to T cells in vitro, whereas conditioned medium from the Gal1 knockdown clones induced only low levels of T-cell apoptosis. In mice, tumour growth from clones that expressed intermediate levels of Gal1 was substantially delayed, compared with wild-type melanoma cells. Furthermore, cells that expressed low levels of Gal1 were rejected. As all clones grew at the same rate in immunodeficient mice, components of the immune system must be responsible for these different responses. The authors went on to show that tumour rejection induced by Gal1 blockade required intact CD4 and CD8 T-cell responses. In previous studies the authors had shown that Gal1 specifically suppresses the T-helper 1 (T H 1) immune response, so they looked to see if this response occurred in mice injected with the Gal1-knockdown melanoma cells. Tumourdraining lymph nodes were removed from these mice and ex vivo stimulation of their immune cells resulted in increased levels of the T H 1 cytokines interleukin-2 and interferon-γ, compared with controls. So, blocking Gal1 production by melanoma cells restores the ability of the host immune system to initiate a T H 1 response. The Gal1-knockdown melanomas were also observed to be infiltrated with mononuclear cells, and T cells from these mice were less susceptible to apoptosis. These mice eventually rejected the Gal1knockdown cells. Furthermore, when they were rechallenged with wild-type B16 melanoma cells, tumour growth was delayed, compared with naive mice challenged with B16 cells. So, inhibition of Gal1 expression not only enhances T H 1 tumour-specific immune responses, it also provides resistance to subsequent challenge with Gal1-expressing tumour cells. This work provides an important new link between Gal1 and tumour progression, and identifies an important mechanism that contributes to immune privilege of tumours. Emma Croager

The PUMA effect: Tumour suppressors

www.nature.com/reviews/cancer Hypoxic areas of tumours, and particularly of metastases, are notoriously resistant to chemoor radiotherapies. Jiwu Wei et al. have developed a scheme to specifically deliver a cytotoxic gene product to hypoxic lung metastases, through embryonic endothelial progenitor cells (EPCs) that usually contribute to tumour vasculogenesis. EPCs arise in the bone marrow and are recruited to sites of active neovasculogenesis by vascular endothelial growth factor (VEGF) and other factors. As they have been shown to contribute to the growing tumour vasculature, Wei et al. investigated if they could be used to deliver therapeutics to hypoxic tumours. They specifically chose to use EPCs isolated from mouse embryos, rather than adult EPCs, because they can be easily grown in culture and genetically manipulated. Also, as embryonic EPCs do not express major histocompatibility complex class I molecules, they are not rejected by the host’s immune system. Wei et al. showed that when these cells were injected into the tail veins of mice, they localized primarily to lung metastases that developed from transplanted osteosarcomas or Lewis lung carcinomas, but were also found in liver and kidney metastases. The EPCs homed mostly to poorly vascularized metastases, which were found to be hypoxic and expressed high levels of VEGF. So could these EPCs be used to deliver a cytotoxic ‘suicide’ gene to these metastases? The authors stably transfected EPCs with the yeast cytosine deaminase gene, fused to uracil phosphoribosyl transferase. This fusion enzyme converts the prodrug 5-fluorocytosine (5-FC) into the cytotoxic compound 5-fluorouracil (5-FU), which can diffuse into the interstitial space and mediate a cytotoxic effect on surrounding tumour cells. Wei et al. found that mice with established multiple lung metastases lived significantly longer after treatment with suicide-geneharbouring EPCs and 5-FC, compared with controls. Up to 90% of all lung metastases were targeted by the cells, and the treatment was not found to cause any toxic effects or embryonic tumours. The mice, however, eventually succumbed to metastases formed in other organs, as well as the non-hypoxic, well-vascularized metastases that were not efficiently infiltrated by the EPCs. Another complication of the system was that it did not kill tumour cells immediately — it required time for the EPCs to incorporate into the tumour vasculature, to express the suicide gene and to kill bystander cells. Furthermore, control EPCs that did not carry the suicide gene actually promoted tumour vascularization and growth. Therefore, one crucial aspect for future clinical use will be the inclusion of safeguards to ensure that the cytotoxic system becomes activated in all EPCs. Kristine Novak References and links ORIGINAL RESEARCH PAPER Wei, J. et al. Embryonic endothelial progenitor cells armed with a suicide gene target hypoxic lung metastases after intravenous delivery. Cancer Cell 5, 477–488 (2004) FURTHER READING Rafii, S. et al. Vascular and haematopoietic stem cells: novel targets for anti-angiogenesis therapy? Nature Rev. Cancer 2, 826–835 (2002) The p53-regulated protein PUMA is a proapoptotic member of the BCL2 family and is required for p53-activated apoptosis in some contexts. However, whether it also has a role in preventing tumorigenesis is unclear. Scott Lowe and colleagues now reveal that PUMA is required for tumour suppression by p53 in response to specific oncogenes. PUMA is crucial for p53-activated cell death in mouse embryonic fibroblasts (MEFs), indicating that it might also be required for tumour suppression by p53 in these cells. To determine this, the authors tested whether loss of Puma expression mimics the increase in tumour formation in the absence of p53 that is seen in MEFs expressing the oncogenes Ras and E1a. Lowe and colleagues used an RNA interference (RNAi) method in which Puma-specific short hairpin RNAs (shRNAs) are expressed from retroviral vectors, allowing sustained inhibition of gene expression. MEFs expressing Puma shRNAs and either Ras alone or both Ras and E1a were injected into mice. When cells expressed both oncogenes, inhibition of PUMA expression resulted in high levels of tumour formation — similar to those seen for cells lacking p53 — whereas cells transduced with a control retrovirus failed to form tumours. However, for cells expressing Ras alone, blocking Puma expression did not stimulate tumour formation. So, although PUMA is required for tumour suppression by p53 in response to a combination of Ras and E1a, it is not essential for the response to Ras alone. Is PUMA required for tumour suppression in response to other oncogenes? Eμ-Myc haematopoietic stem cells (HSCs) form B-cell lymphomas when injected into mice as they express Myc under the control of an immunoglobulin heavy-chain promoter. This lymphomagenesis is accelerated when p53 expression is inhibited by RNAi. The authors tested whether the same effect is seen when PUMA expression is blocked. All mice injected with Eμ-Myc HSCs expressing Puma shRNAs developed lymphomas, compared with 40% of mice injected with cells carrying a control retrovirus. Similar to the effects of p53 inhibition, the cells lacking PUMA expression also formed lymphomas at an increased rate. Lowe and colleagues confirmed that PUMA inhibition cooperates with Myc expression to promote tumorigenesis, as only one-sixth of mice injected with wild-type HSCs expressing Puma shRNAs developed lymphomas. So, PUMA is required to suppress tumorigenesis in some contexts, such as expression of MYC or a combination of E1A and RAS, but not in others, such as expression of RAS alone. Further studies should increase our understanding of how different p53 targets contribute to tumour suppression in response to specific oncogenic stimuli, with implications for how this key pathway could be manipulated in the treatment of different types of cancer. Louisa Flintoft

Tumorigenesis: Hero or villain?

www.nature.com/reviews/cancer Conflicting evidence has led to confusion about whether peroxisomeproliferator-activated receptor-γ (PPARγ) has an inhibitory or stimulatory effect on tumorigenesis. Ronald Evans and colleagues now show that Pparγ signalling promotes the formation of mammary carcinomas in mice, but only in animals that are genetically predisposed to developing these tumours. PPARγ is overexpressed in several human cancers, including mammary tumours. Activators of this receptor inhibit tumour development in rat models of mammary carcinoma, indicating that stimulating PPARγ signalling might be a useful anticancer treatment. However, PPARγ activators can suppress proliferation even in cells in which the receptor is not expressed and can cause increased tumour development in mouse models of colon cancer. This confusion necessitates a greater understanding of how PPARγ signalling contributes to the development of different tumour types. Evans and colleagues generated transgenic mice that express a constitutively active form of Pparγ in breast epithelium. Mammarygland development was normal in these animals and they showed no increased tendency to develop tumours. The authors then crossed these animals to mice that express the polyoma virus middle T antigen (PyV-MT) in mammary tissue, which rapidly develop tumours with an average time to detection of 57 days in female mice. Those that expressed both activated Pparγand PyV-MT showed accelerated development of mammary tumours, with an average time to appearance of just 37 days. So, although increased Pparγactivation does not initiate tumour formation in normal mammary tissue, it promotes tumorigenesis on a tumour-susceptible background. Increased Wnt signalling is implicated in the development of mammary carcinomas, so could Pparγ contribute to tumour development through activation of this pathway? The Wnt target genes cyclin D1 and c-Myc were shown to be upregulated in mice expressing PyV-MT and constitutively active Pparγ compared with mice expressing PyV-MT only. Increased expression was also seen for β-catenin, a component of the Wnt pathway, and for the Wnt receptor frizzled homologue 4, whereas Wnt5a, a negative regulator of Wnt signalling, was downregulated. This study has made important progress in understanding the role of Pparγin tumorigenesis, but several key questions remain. For example, why does Pparγ signalling only promote mammary tumour development in genetically susceptible cells and how does Pparγ interact with the Wnt signalling pathway? This work also strengthens the suggestion that the inhibition of tumorigenesis by Pparγligands in previous studies is due to receptor-independent effects of these proteins. Clearly, further investigation will be needed before a safe verdict on the connection between PPARγand cancer can be reached. Louisa Flintoft

Hero or villain?: Tumorigenesis

www.nature.com/reviews/cancer Conflicting evidence has led to confusion about whether peroxisomeproliferator-activated receptor-γ (PPARγ) has an inhibitory or stimulatory effect on tumorigenesis. Ronald Evans and colleagues now show that Pparγ signalling promotes the formation of mammary carcinomas in mice, but only in animals that are genetically predisposed to developing these tumours. PPARγ is overexpressed in several human cancers, including mammary tumours. Activators of this receptor inhibit tumour development in rat models of mammary carcinoma, indicating that stimulating PPARγ signalling might be a useful anticancer treatment. However, PPARγ activators can suppress proliferation even in cells in which the receptor is not expressed and can cause increased tumour development in mouse models of colon cancer. This confusion necessitates a greater understanding of how PPARγ signalling contributes to the development of different tumour types. Evans and colleagues generated transgenic mice that express a constitutively active form of Pparγ in breast epithelium. Mammarygland development was normal in these animals and they showed no increased tendency to develop tumours. The authors then crossed these animals to mice that express the polyoma virus middle T antigen (PyV-MT) in mammary tissue, which rapidly develop tumours with an average time to detection of 57 days in female mice. Those that expressed both activated Pparγand PyV-MT showed accelerated development of mammary tumours, with an average time to appearance of just 37 days. So, although increased Pparγactivation does not initiate tumour formation in normal mammary tissue, it promotes tumorigenesis on a tumour-susceptible background. Increased Wnt signalling is implicated in the development of mammary carcinomas, so could Pparγ contribute to tumour development through activation of this pathway? The Wnt target genes cyclin D1 and c-Myc were shown to be upregulated in mice expressing PyV-MT and constitutively active Pparγ compared with mice expressing PyV-MT only. Increased expression was also seen for β-catenin, a component of the Wnt pathway, and for the Wnt receptor frizzled homologue 4, whereas Wnt5a, a negative regulator of Wnt signalling, was downregulated. This study has made important progress in understanding the role of Pparγin tumorigenesis, but several key questions remain. For example, why does Pparγ signalling only promote mammary tumour development in genetically susceptible cells and how does Pparγ interact with the Wnt signalling pathway? This work also strengthens the suggestion that the inhibition of tumorigenesis by Pparγligands in previous studies is due to receptor-independent effects of these proteins. Clearly, further investigation will be needed before a safe verdict on the connection between PPARγand cancer can be reached. Louisa Flintoft

Hero or villain

www.nature.com/reviews/cancer Conflicting evidence has led to confusion about whether peroxisomeproliferator-activated receptor-γ (PPARγ) has an inhibitory or stimulatory effect on tumorigenesis. Ronald Evans and colleagues now show that Pparγ signalling promotes the formation of mammary carcinomas in mice, but only in animals that are genetically predisposed to developing these tumours. PPARγ is overexpressed in several human cancers, including mammary tumours. Activators of this receptor inhibit tumour development in rat models of mammary carcinoma, indicating that stimulating PPARγ signalling might be a useful anticancer treatment. However, PPARγ activators can suppress proliferation even in cells in which the receptor is not expressed and can cause increased tumour development in mouse models of colon cancer. This confusion necessitates a greater understanding of how PPARγ signalling contributes to the development of different tumour types. Evans and colleagues generated transgenic mice that express a constitutively active form of Pparγ in breast epithelium. Mammarygland development was normal in these animals and they showed no increased tendency to develop tumours. The authors then crossed these animals to mice that express the polyoma virus middle T antigen (PyV-MT) in mammary tissue, which rapidly develop tumours with an average time to detection of 57 days in female mice. Those that expressed both activated Pparγand PyV-MT showed accelerated development of mammary tumours, with an average time to appearance of just 37 days. So, although increased Pparγactivation does not initiate tumour formation in normal mammary tissue, it promotes tumorigenesis on a tumour-susceptible background. Increased Wnt signalling is implicated in the development of mammary carcinomas, so could Pparγ contribute to tumour development through activation of this pathway? The Wnt target genes cyclin D1 and c-Myc were shown to be upregulated in mice expressing PyV-MT and constitutively active Pparγ compared with mice expressing PyV-MT only. Increased expression was also seen for β-catenin, a component of the Wnt pathway, and for the Wnt receptor frizzled homologue 4, whereas Wnt5a, a negative regulator of Wnt signalling, was downregulated. This study has made important progress in understanding the role of Pparγin tumorigenesis, but several key questions remain. For example, why does Pparγ signalling only promote mammary tumour development in genetically susceptible cells and how does Pparγ interact with the Wnt signalling pathway? This work also strengthens the suggestion that the inhibition of tumorigenesis by Pparγligands in previous studies is due to receptor-independent effects of these proteins. Clearly, further investigation will be needed before a safe verdict on the connection between PPARγand cancer can be reached. Louisa Flintoft