The lungs are a frequent target of metastatic breast cancer cells, but the underlying molecular mechanisms are unclear. All existing data were obtained either using statistical association between gene expression measurements found in primary tumors and clinical outcome, or using experimentally derived signatures from mouse tumor models. Here, we describe a distinct approach that consists of using tissue surgically resected from lung metastatic lesions and comparing their gene expression profiles with those from nonpulmonary sites, all coming from breast cancer patients. We show that the gene expression profiles of organ-specific metastatic lesions can be used to predict lung metastasis in breast cancer. We identified a set of 21 lung metastasisassociated genes. Using a cohort of 72 lymph node-negative breast cancer patients, we developed a 6-gene prognostic classifier that discriminated breast primary cancers with a significantly higher risk of lung metastasis. We then validated the predictive ability of the 6-gene signature in 3 independent cohorts of breast cancers consisting of a total of 721 patients. Finally, we show that the signature improves risk stratification independently of known standard clinical variables and a previously established lung metastasis signature based on an experimental breast cancer metastasis model. [Cancer Res 2008;68(15):6092-9]

BACKGROUND: New models continue to be required to improve our understanding of colorectal cancer progression. To this aim, we characterised in this study a three-dimensional multicellular tumour model that we named colospheres, directly obtained from mechanically dissociated colonic primary tumours and correlated with metastatic potential. METHODS: Colorectal primary tumours (n ¼ 203) and 120 paired non-tumoral colon mucosa were mechanically disaggregated into small fragments for short-term cultures. Features of tumours producing colospheres were analysed. Further characterisation was performed using colospheres, generated from a human colon cancer xenograft, and spheroids, formed on agarose by the paired cancer cell lines. RESULTS: Colospheres, exclusively formed by viable cancer cells, were obtained in only 1 day from 98 tumours (47%). Inversely, nontumoral colonic mucosa never generated colospheres. Colosphere-forming capacity was statistically significantly associated with tumour aggressiveness, according to AJCC stage analysis. Despite a close morphology, colospheres displayed higher invasivity than did spheroids. Spheroids and colospheres migrated into Matrigel but matrix metalloproteinase (MMP)-2 and MMP-9 activity was detected only in colospheres. Mouse subrenal capsule assay revealed the unique tumorigenic and metastatic phenotype of colospheres. Moreover, colospheres and parental xenograft reproduced similar CD44 and CD133 expressions in which CD44 þ cells represented a minority subset of the CD133 þ population. CONCLUSION: The present colospheres provide an ex vivo three-dimensional model, potentially useful for studying metastatic process.

We assessed the reliability of prognostic biologic markers by means of immunohistochemistry on cell blocks obtained from diagnostic fine-needle cytopunctures of breast carcinomas and their lymph node metastases.Immunohistochemical studies of MIB-1 (Ki-67), estrogen receptors (ER), progesterone receptors (PR), p53, and c-erb-B-2 were performed in 55 cases of primary breast carcinoma on cell blocks (cytoblock technique) and on their corresponding tissue samples (46 mastectomy specimens and 9 Trucut biopsies) and in 38 cases on cell blocks from fineneedle cytopunctures of both the primary breast tumors and their concurrent lymph node metastases.Interobserver reproducibility ranged from 87 to 100%, depending on the marker. A good correlation was observed between immunostaining assessment on cell blocks and on the corresponding tumor tissues as follows: Ki-67 (85%), ER (96%), PR (82%), p53 (76%), and c-erb-B-2 (84%). An excellent correlation was observed between cell-block results for primary tumors and node metastases; however, a far higher percentage of Ki-67-positive nuclei was observed in the nodes than in the corresponding tumors in seven cases. All nodes corresponding to ER-or PR-negative tumors were also negative, whereas the nodes corresponding to two ERpositive and one PR-positive tumor were negative. Marked discrepancies were also noted with p53 in two cases and with c-erb-B-2 in two cases. Most discrepancies occurred with Trucut biopsies and with breast tumors that contained a large intraductal component.

One of the main genetic abnormalities associated with breast carcinogenesis is the loss of genetic material from chromosome arm 16q. Different groups have identified two regions (16q22.1 and 16q24‐ter) that are frequently deleted in primary tumors, suggesting the presence of tumor suppressor genes in these regions. Little is known about the late stages of tumor progression in this respect, and we, therefore, analyzed biopsy specimens of breast cancer metastases for deletions in these critical regions of 16q. We examined fine needle cytopunctures from 24 metastases, each with lymphocyte DNA, for allelic imbalance on 16q by means of polymerase chain reaction (PCR) with 15 highly polymorphic markers. All the metastatic samples showed deletion of at least one informative locus on 16q. The loss of heterozygosity (LOH) pattern often indicated the loss of a complete long arm of chromosome 16 (13 cases); nevertheless, in the remaining 11 samples, partial LOH patterns were observed. A small region of overlap (SRO2) in 16q22.1 was frequently involved, whereas another (SRO1) in 16q24‐ter was affected in only two cases. A third region of LOH in 16q22.2‐q23.2 was found in 6/11 samples. These results suggest that at least three different regions are involved in allelic imbalance on chromosome arm 16q in breast cancer. Loss of material from the third region could be a major event in the genesis of metastases. Genes Chromosom. Cancer 19:185–191, 1997. © 1997 Wiley‐Liss Inc.

Thirty-five patients with large but operable breast carcinoma (T3, NO-N1, MO) received before surgery three cycles of preoperative Adriamycin (doxorubicin), vincristine, cyclophosphamide, methotrexate, 5-fluorouracil (AVCMF) chemotherapy (CT). All patients had sequential cytopunctures for both cytologic examination and flow cytometric DNA analysis (FCM): one before treatment and the three others after each cycle of CT. At cytologic examination, 20 carcinomas showed CT-induced cytomorphologic changes. These changes in malignant cells were also evaluated on histologic sections after surgery. A significant relationship was found between cytomorphologic changes in cytopunctures and in the corresponding tumor tissue. The lobular carcinomas did not reveal changes either on cytologic or on histologic study. At FCM analysis before treatment, ten carcinomas were diploid and 25 were nondiploid. When initial tumor ploidy was compared to cytomorphologic changes, none of the diploid carcinomas showed changes. An objective tumor regression was observed in 12 of 20 tumors with cytomorphologic changes and in four of 15 tumors without changes on cytologic examination. But, a significant relationship appeared only when cellular changes were evaluated on histologic analysis. When tumor regression was compared to ploidy before treatment, the rate of objective regression was significantly higher in nondiploid tumors (15/25) than in diploid tumors (one of ten). From these findings, initial tumor diploidy could be a predictor of tumor chemoresistance, whereas cytomorphologic changes during chemotherapy could be an indicator of tumor cell chemosensitivity.