Aldehyde Dehydrogenase 1–Positive Cancer Stem Cells Mediate Metastasis and Poor Clinical Outcome in Inflammatory Breast Cancer Free

Inflammatory breast cancer (IBC) is an angioinvasive form of breast cancer associated with a high incidence of early nodal and systemic metastasis. In contrast to the recent decrease in breast cancer incidence in the United States, the annual incidence of IBC continues to increase (1, 2) with an attendant increase in mortality (3). Despite advances in the use of systemic chemotherapy, the prognosis of IBC remains considerably worse than that of other locally advanced breast cancers (1).

Several molecular changes have been described in IBC including RHOC overexpression, hypomethylation of caveolin-1 or caveolin-2 promoters, and deletion of the tumor suppressor WISP3 (4–8). In addition, IBCs have been reported to overexpress E-cadherin/α, β-catenin, and angiogenic factors (4, 7, 9–14). Although each of these genetic changes may contribute to the metastatic nature of IBC, no markers have been described that can predict the development of systemic metastasis or survival in IBC patients. Although ERBB2 expression is associated with aggressive behavior in most breast cancers, this is not the case in IBC (15).

There is increasing evidence that human breast cancers are driven by a tumor-initiating “cancer stem cell” (CSC) component that may contribute to tumor metastasis and therapeutic resistance (16–20). Breast CSCs were initially characterized as CD44⁺/CD24⁻/lin⁻ cells that were capable of serial transplantation in nonobese/severe combined immunodeficient (NOD/SCID) mice (21). In addition to these markers, we have recently shown that cells with stem cell properties in both normal and malignant breast samples can be identified by the expression of the enzyme aldehyde dehydrogenase 1 (ALDH1). Using flow cytometry and the ALDEFLUOR assay, which measures ALDH activity, we isolated CSCs from primary human mammary carcinomas grown as xenografts in NOD/SCID mice. In addition, ALDH1 immunostaining identified normal and malignant CSCs in situ in fixed paraffin-embedded sections (22).

The rare occurrence of IBC as well as the small size of tumor specimens contribute to the difficulties of studying the biology of this disease. This development of an immortalized cell line and xenograft model of human IBC has facilitated studies of IBC biology (23, 24). In the present study, we have used in vitro assays as well as mouse models of the SUM149 IBC cell line and early passages of the MARY-X xenograft generated from a primary IBC tumor. We determined whether IBC contains CSCs and whether these cells mediate tumor invasion and metastasis. To investigate the clinical relevance of these findings, we examined the expression of the stem cell marker ALDH1 in tissue sections from patients with IBC. The in vitro studies and mouse xenografts provide evidence that the invasive and metastatic behavior of IBC is mediated by an ALDH1-positive CSC component. Furthermore, expression of this stem cell marker was associated with the development of early metastases and poor clinical outcome in IBC patients.

Additional data are available in the Supplementary Materials and Methods section.

SUM149, a gift from S. Ethier (Karmanos Cancer Institute, Detroit, MI), is a breast cancer cell line derived from a patient with primary IBC and is obtained from early passages only (<6 mo; ref. 25).⁹

The ALDEFLUOR kit (StemCell technologies) was used to isolate the population with a high ALDH enzymatic activity. SUM149 and MARY-X cells were suspended in ALDEFLUOR assay buffer containing ALDH substrate (BAAA, 1 μmol/l per 1 × 10⁶ cells) and incubated for 40 min at 37°C. In each experiment, a sample of cells was incubated with 50 mmol/L of the specific ALDH inhibitor diethylaminobenzaldehyde (DEAB) as negative control. Flow cytometry sorting was conducted using a FACStarPLUS (Becton Dickinson). ALDEFLUOR fluorescence was excited at 488 nm and fluorescence emission was detected using a standard FITC 530/30 band pass filter. The sorting gates were established using the propidium iodide–stained cells for viability and the ALDEFLUOR-stained cells treated with DEAB as negative controls.

Tumorigenicity of ALDEFLUOR-positive, ALDEFLUOR-negative, and unseparated SUM149 and MARY-X cells was assessed in three independent sets of three NOD/SCID mice. Fat pads were cleared at 3 wk of age before puberty and humanized by injecting a mixture of irradiated and nonirradiated immortalized human fibroblasts (1:1 irradiated/nonirradiated, 50,000 cells/100 μL Matrigel/fat pad) as described in ref. (22).

After sorting, the tumorigenicity of the ALDEFLUOR-positive, ALDEFLUOR-negative, and unseparated populations of SUM149 and MARY-X cells was tested by inoculation of limiting dilutions of cells (50,000, 5,000, and 500 cells) mixed with Matrigel (BD Biosciences; 1:1) and implanted in the cleared humanized mammary fat pads 2 to 4 wk later.

For luciferase gene transduction, 70% confluent SUM149 cells and suspension culture of MARY-X single cells were incubated overnight with a 1:3 precipitated mixture of lentiviral supernatants Lenti-LUC-VSVG (Vector Core) in culture medium. The following day, the SUM149 cells were harvested by trypsin/EDTA and subcultured at a ratio of 1:6, and MARY-X cells were maintained in suspension culture. After 1 wk of incubation, luciferase expression was verified by adding 2 μL d-luciferin 0.0003% (Promega) in the culture medium and by counting photon flux by device camera system (Xenogen).

Six-week-old NOD/SCID mice were anesthetized with 1.75% isoflurane/air anesthesia, and the left ventricule of the heart was injected with 50,000 or 100,000 cells (ALDEFLUOR positive, ALDEFLUOR negative, unseparated) in 100 μL of sterile Dulbecco's PBS lacking Ca²⁺ and Mg²⁺. Experiments were done in duplicates for SUM149 and MARY-X.

After cell inoculations, the animals were screened for metastasis using bioluminescence. Baseline bioluminescence was assessed before inoculation and each week thereafter. Mice were anesthetized with a 2% isoflurane/air mixture and were given a single i.p. dose of 150 mg/kg d-luciferin (Promega) in PBS. Animals were then reanesthetized using a 2% isoflurane/air mixture 6 min after the administration of d-luciferin. For photon flux counting, we used a charge-coupled device camera system (Xenogen) with a nose-cone isoflurane delivery system and a heated stage for maintaining body temperature. Results were analyzed after 2 to 12 min of exposure using Living Image software provided with the Xenogen imaging system. Signal intensity was quantified as the sum of all detected photon flux counts within a uniform region of interest manually placed during data after processing. Normalized photon flux represents the ratio of the photon flux detected each week after inoculations and the photon flux detected before inoculation.

IBC patients were selected from computerized clinicopathologic databases of Institut Paoli-Calmettes between 1976 and 2003 as consecutive cases with available paraffin-embedded tumoral specimen. IBC was clinically defined as a T_4d tumor (tumor-node-metastasis, Unio Internationale Contra Cancrum), and metastatic patients at time of diagnosis were not included; the presence of dermal lymphatic emboli was not mandatory for IBC definition. One hundred and nine patients with IBC with a median follow-up of 67 mo were included. All details are in Supplementary Materials and Methods.

Expression of ALDH1, BCL2, E-Cadherin, estrogen receptor (ER), MIB1, ERBB2, MUC1, and progesteron receptor (PR) was measured by immunohistochemistry, and expression of CD24 and CD44 was measured by immunohistochemical double staining as well as by flow cytometry. The characteristics of the antibodies used and the details of the technique are listed in Supplementary Table S1 and in Supplementary Materials and Methods. ALDH1 immunostaining was done as described previously (22), using positive external control for each experiment. For each slide, when internal controls (surrounding fibroblasts or histiocytes located in the stroma or between tumoral cells) were not stained for ALDH1, the slide was not included in the study. Results were expressed in terms of percentage (P) and intensity (I) of positive cells as described previously (22). Results were scored by the quick score (Q; Q = P × I). Tumor presenting at least one ALDH1-positive cancer cell was considered as an ALDH1-positive tumor. Scoring was estimated by two independent breast pathologists (ECJ, JJ) and the mean-value was used. Discrepancies were resolved under multiheaded microscope.

For in vitro experiments and animal models, results are presented as the mean ± SD for at least three repeated individual experiments for each group. Statistical analyses used the SPSS software (version 10.0.5). Correlations between sample groups and parameters were calculated with the Fisher's exact test or the one-way ANOVA for independent samples. A P value of <0.05 was considered significant.

For the IBC retrospective series, data were summarized by frequencies and percentages for categorical variables, and by median and range for continuous variables. No missing data imputation technique was applied. To study the associations among variables, univariate analysis was done using nonparametric Wilcoxon rank-sum test, χ² test, or Fisher's exact test when appropriate. Specific survival (SS) and metastasis-free survival (MFS) rates were estimated by the Kaplan-Meier method using the following first event definitions: death by breast cancer for SS and first metastasis recurrence for MFS. Patients without events were censored at the time of last follow-up or at the date of death if they died for other reason than breast cancer. All survival times were calculated from the date of breast cancer diagnosis. Changes in the relative risk of events according to prognostic factors were assessed by using the log-rank test in univariate analysis. Multivariate analysis was done using Cox's proportional hazard models with a backward stepwise selection of variables to minimize the Akaike Information Criterion. All statistical tests were two sided at the 5% level of significance, using the R 2.9.1 software.

The enzyme ALDH has been a useful marker for isolating primitive stem cell populations. We have shown previously that normal human mammary stem and progenitor cells as well as transformed tumor-initiating stem cells may be isolated by virtue of their expression of ALDH activity, as assessed by flow cytometry using the ALDEFLUOR assay. To determine whether IBC contains a CSC population, we used two different models: SUM149, a breast cancer cell line derived from a patient with primary IBC (25), and MARY-X, a human IBC xenograft (9). Using the ALDEFLUOR assay, we isolated an ALDEFLUOR-positive component comprising 5.96 ± 2.2% and 7.2 ± 1.5% of the total cell population (Fig. 1A-B and G-H) of SUM149 and MARY-X, respectively. The tumorigenicity of the ALDEFLUOR-positive, ALDEFLUOR-negative, and unseparated populations of SUM-149 and MARY-X cells was tested by inoculation of limiting dilutions of cells (50,000, 5,000, and 500 cells) into mammary fat pads of NOD/SCID mice that were humanized by the introduction of irradiated and nonirradiated human mammary fibroblasts (22, 26). For the two models used, the fat pads injected with 50,000, 5,000, and 500 ALDEFLUOR-positive cells generated tumors, whereas the ALDEFLUOR-negative cells failed to generate tumors even when 50,000 cells were inoculated (Fig. 1D-F and J-L). A summary of the fat pad injected and the tumor formation in each case is in Supplementary Table S2. H&E staining of fat pad sections confirmed that tumors formed by the ALDEFLUOR-positive cells contained malignant cells with a histology similar to the initial tumors (Fig. 1E and K). In contrast, only residual Matrigel, apoptotic cells, and mouse tissue were seen at the sites of the ALDEFLUOR-negative cell injections (Fig. 1F and L). As indicated in Fig. 1D and J, the size and latency of tumor formation correlated with the number of ALDEFLUOR-positive cells injected with 500 ALDEFLUOR-positive cells generating tumors within 30 days for SUM149 and 75 days for MARY-X. To show the self-renewal potential of the ALDEFLUOR-positive population, we performed three consecutive serial passages in NOD/SCID mice. With each inoculation, the ALDEFLUOR-positive but not ALDEFLUOR-negative cells were able to form tumors even when 500 cells were inoculated. The ability of a few ALDEFLUOR-positive cells to generate tumors that could be serially passaged shows the self-renewal capacity of these cells.

To determine the differentiation capacity of the ALDEFLUOR-positive population, the distribution of ALDEFLUOR-positive and ALDEFLUOR-negative cells was determined at each serial passage. As shown in Fig. 1C and I, ALDEFLUOR-positive cells generated tumors that contained ALDEFLUOR-positive and ALDEFLUOR-negative cells in a similar proportion to that found in the initial tumor. Using the two models, we showed that in addition to self-renewal, ALDEFLUOR-positive cells from SUM149 and MARY-X are able to differentiate, generating a population of ALDEFLUOR-negative nonself-renewing cells.

We next investigated the overlap between the ALDEFLUOR-positive population and the previously described breast CSC phenotype, CD44⁺/CD24⁻ (21). Flow cytometry analysis of the SUM149 tumors showed that the ALDEFLUOR-positive population was enriched in CD44⁺/CD24⁻ cells, with 13.5% of CD44⁺/CD24⁻ cells in the ALDEFLUOR-positive population compared with only 3.02% in the ALDEFLUOR-negative population (Supplementary Fig. S1). Because >98% of MARY-X cells are CD44⁺/CD24⁻, we did not investigate the overlap between the two phenotypes (Supplementary Fig. S1) in this model.

IBC has a high propensity for the development of lymphagenic invasion and distant metastasis. To determine whether these properties were mediated by the CSC component, we used a Matrigel invasion assay to examine the ability of ALDEFLUOR-positive and ALDEFLUOR-negative SUM149 cell populations to invade. As shown in Fig. 2A, the percentage of ALDEFLUOR-positive SUM149 cells capable of invasion through Matrigel was >3-fold higher than that of the ALDEFLUOR-negative population (P < 0.05). To determine the metastatic capacity of these cell populations, we labeled SUM149 and MARY-X cells with a luciferase lentivirus reporter system. The luciferase-labeled cells were sorted using the ALDEFLUOR assay and ALDEFLUOR-positive, ALDEFLUOR-negative, and unseparated cells were introduced by intracardiac injection into NOD/SCID mice. For SUM149 and MARY-X, a suspension of 50,000 and 100,000 cells from each group were injected and the development of metastasis was assessed by noninvasive luciferase bioluminescent imaging quantified by photon flux (27). As shown in Fig. 2B to G, only ALDEFLUOR-positive and unseparated cells formed metastases in this assay. Histologic sections confirmed the presence of bone and lung metastases (Fig. 2H-I) produced from injection of ALDEFLUOR-positive and unseparated cells. Furthermore, vertebral metastasis formed from intracardiac injection of ALDEFLUOR-positive MARY-X cells contained populations of ALDH1-positive and ALDH1-negative cells recapitulating the heterogeneity of the initial tumor.

In contrast, no metastases were detected in mice inoculated with either 50,000 or 100,000 ALDEFLUOR-negative cells. The absence of systemic metastasis was confirmed by histologic examination of sections of the liver, bones, brain, and lung tissues. These results suggest that invasion and systemic metastasis in IBC are mediated by cells expressing the stem/progenitor cell marker ALDH1.

We have shown previously that monoclonal antibodies to ALDH1 are able to identify normal and malignant breast stem/progenitor cells in fixed paraffin-embedded sections in situ (25). To determine the clinical relevance of expression of this stem/progenitor cell marker, we examined the expression of ALDH1 in a series of 109 IBC patients treated at our institution between 1976 and 2003. ALDH1 was expressed in 34% of tumors with approximately 3% to 5% positively stained cells in these tumors (Fig. 3A). This percentage of ALDH1 expression in this patient population is similar to that found in non-IBC (22). ALDH1-positive cells were also detected in intralymphatic tumor emboli (Fig. 3B-D) consistent with a role for these cells in tumor metastasis. ALDH1 expression correlated with the SBR grade, but there was no correlation with several other clinical and pathologic features including estrogen and progesterone receptors, BCL2, ERBB2, E-cadherin, or MUC1 (Table 1).

To explore another marker of breast CSCs in IBC, and to evaluate its correlation with ALDH expression, we investigated the presence of the CD44⁺/CD24⁻ phenotype in IBC archival specimens. The presence of CD44⁺/CD24⁻ cells was evaluated by immunohistochemistry using double staining on paraffin sections (Supplementary Fig. S2). The CD44⁺/CD24⁻ phenotype was evaluated in 81 cases of the 109 IBC samples, and was present in 51 cases. By comparison with non-IBC tumors, we observed an increase of the percentage of tumors that displayed the CD44⁺/CD24⁻ phenotype in IBC (63% in IBC versus 31% in non-IBC across literature data; ref. 28). This phenotype in IBC was correlated with high proliferation index, with absence of MUC1 expression, and with less lymph node invasion, which differed from previous reports of expression of this phenotype in non-IBC (Supplementary Table S3). More importantly, there was no correlation between the CD44⁺/CD24⁻ phenotype and ALDH1 expression (Table 1).

To determine whether the expression of the stem cell marker ALDH1 correlated with the development of systemic metastasis and survival in IBC patients, Kaplan-Meier survival curves were constructed and compared by log-rank tests. ALDH1 expression correlated with the development of distant metastasis and with decreased survival in these patients. With a median follow-up of 67 months, ALDH1 expression strongly correlated with MFS (P = 0.0152) as well as with tumor-specific survival (SS; P = 0.0337; Fig. 3E-F). The median MFS was 49 months in patients with ALDH1-negative tumors, compared with 20 months in patients with ALDH1-positive tumors (Table 2). Median SS was 63 months for patients whose tumors did not express ALDH1 compared with only 27 months for patients with tumors expressing ALDH1. Among other parameters tested in univariate analysis, BCL2 expression or ER/PR expression were also correlated with an increased MFS and SS (Table 2). In a multivariate analysis using Cox proportional hazard models, ALDH1 was the only prognostic marker of MFS [P = 0.0055; hazard ratio (HR), 2.81; 95% CI (95% confidence interval), 1.355-5.815] and the most powerful prognostic marker of SS (ALDH1: P = 0.0012; HR, 2.7; 95% CI, 1.48-4.93) when factors significant in univariate analysis (hormonal receptors and BCL2 expression) were included in the model (Table 3). BCL2 expression was still significant in the model for SS only (BCL2: P = 0.031; HR, 0.4; 95% CI, 0.224-0.93). In contrast, the expression of the CD44⁺/CD24⁻ phenotype previously shown to be a CSC marker in non-IBC was not associated with either development of metastasis or patient survival (Supplementary Fig. S3).

IBC is among the most angioinvasive and metastatic variants of human breast cancer. Molecular mechanisms have been implicated in IBC clinical aggressiveness. E-cadherin overexpression and dysfunctional, hyposialylated MUC1 may contribute to the metastatic route of IBC (10, 29). The frequent overexpression of RHOC GTPase, hypomethylation of caveolin-1 or caveolin-2 promoters, or deletion of the tumor suppressor WISP3 that belongs to the CCN family illustrates the ability of IBC cells to migrate through putative induction of epithelial to mesenchymal transition process (4–8, 30). Yet, very few markers are available to improve IBC clinical course.

In this study, we have used in vitro and mouse models to show that IBCs contain a cellular component, characterized by the expression of ALDH that displays stem cell properties and is able to mediate IBC aggressive behavior.

We have previously shown the feasibility of using established cell lines for studies of CSC biology (31). Of note, despite the availability of cell lines representing the different molecular subtypes of breast cancer, the establishment of similar models for IBC has been notoriously difficult. Strikingly, of the 50 cell lines and ∼20 xenografts in vivo established in the past decades from breast cancers, only SUM149/190, MARY-X, and WIBC-9 were derived from IBC and are currently available to study this type of cancer (9, 24, 32). It highlights the importance of the material available for stem cell studies in IBC. Indeed, tumor xenografting is mandatory to test stem cell properties including tumorigenicity on serial passages and differentiation (16, 22). In that case, the use of primary human breast tumors is not feasible because breast cancer has a well-documented low xenografting rate. Furthermore, due to the small size of tumor specimen available in IBC at time of diagnosis, these questions cannot be addressed for most inflammatory breast tumors. In this study, we used the SUM149 IBC cell line and the MARY-X xenograft, which was established from an IBC tumor without any in vitro culture step (9, 13).

We used in vitro and in vivo experiments to test stem cell properties in cells that are expressing ALDH, an enzyme implicated in retinoic acid metabolism. These “canonical” stem cell properties include self-renewal as shown by tumorigenicity on serial passages, and differentiation as shown by the ability to reconstitute the phenotypic heterogeneity of the initial tumor. Furthermore, we showed that these ALDEFLUOR-positive cells are able to mediate tumor invasion in vitro and tumor metastasis in mouse xenografts. These properties recapitulate the aggressive behavior of IBC in patients. This is consistent with the ALDH staining in the tumor emboli described here and by others (33).

Although it is clear that IBC has an extremely poor prognosis, there currently are no validated markers that predict outcome in this disease. In fact, well-known markers such as ERBB2 are associated with the aggressive behavior of non-IBC and do not correlate with patient outcome in IBC (15, 34). The antiapoptotic factor BCL2, which expression has been associated with improved SS in non-IBC, is described for the first time as an independent factor for SS in IBC (35). However, BCL2 expression is not significantly associated with longer MFS (Table 2) and is not correlated with the stem cell marker ALDH1 expression (Table 1) or CD44⁺/CD24⁻ phenotype (Supplementary Table S3). In contrast, our work shows that expression of the stem cell marker ALDH1 is associated with development of early metastasis and is an independent prognostic marker for IBC. To our knowledge, ALDH1 is the only independent marker ever described in IBC to predict metastasis. Interestingly, the CD44⁺/CD24⁻ phenotype shown to be associated with CSCs for non-IBC was not associated with outcomes in IBC. This either suggests that ALDH1 and CD44⁺/CD24⁻ are detecting distinct although overlapping cell populations or that the reliability of detection of CD44⁺/CD24⁻ in fixed tumor tissue may be limited (16, 36).

Although cells that expressed ALDH1 have been shown to mediate poor prognosis and metastasis in IBC, the percentage of ALDH1 expression in this patient population is similar to that found in non-IBC and is thus not sufficient to explain by itself the aggressive behavior of IBC. It remains true that expression of ALDH1 is associated with poor outcome and early metastasis in IBC, and we can assume that ALDH-expressing CSC exert different abilities in IBC and in non-IBC. In human hematopoietic system, only a subpopulation of ALDEFLUOR-positive stem cells, expressed CD133, are able to repopulate the murine bone marrow (37). In solid tumor, using a pancreatic cell line xenografted in immunocompromised mice, only a fraction of pancreatic CSCs carry the metastatic potential (38). Consequently, in breast tumor, only a subset of CSC may carry the metastatic potential. Hence, the proportion of metastatic CSC inside the ALDEFLUOR-positive CSC population can differ between IBC and non-IBC and should explain the clinical discrepancies between the two clinical entities. Further studies are needed to isolate potential markers and validate this hypothesis.

We have shown that IBC contained a CSC population that expresses ALDH and that ALDEFLUOR-positive IBC cells are highly invasive and mediate metastasis in mice. It is often thought that aggressive cancers may not be organized hierarchically or may have a predominant rather than rare CSC population (39, 40). Our work shows that this is not the case in IBC. Moreover, we have shown for the first time that the expression of the stem cell marker ALDH1 is associated with the development of systemic metastasis and decreased survival in IBC patients. Our data also suggest that in addition to mediating metastasis, CSCs also have the ability to reconstitute the tumor heterogeneity of the primary tumor at these metastatic sites.

These results suggest that ALDEFLUOR hierarchy in this subset of breast cancer with respect to its clinical behavior and biological diversity plays an important role in mediating the aggressive course of IBC. Recent preclinical and neoadjuvant clinical studies in non-IBC raised the hypothesis that this hierarchy may affect the clinical management of breast cancer patients. Hence, they have suggested that CSCs in these tumors are relatively resistant to chemotherapy compared with the bulk tumor cell populations (41, 42). Because cytotoxic chemotherapy is the current recommended treatment for IBC (43), it will be a major point to determine whether the CSC components of IBC are also resistant to chemotherapy accounting for poor outcome in these patients. If this is the case, then alternative strategies aimed at targeting this CSC population will need to be developed. Based on the successful clinical application of differentiation therapy in acute promyelocytic leukemia, one of these strategies could be the induction of differentiation (44). This underlines the importance of the ALDH1 status in translating CSC research into IBC clinical practice.

M.S. Wicha has financial holdings in and is a scientific advisor for Oncomed Pharmaceuticals. The other authors disclosed no conflicts of interest.

We thank Jean-Paul Borg and Florence Monville for the helpful discussions, and Olivier Cabaud, Jeannine Geneix, Laetitia Repellini, and the CIML flow cytometry core (P. Grenot and M. Malissen) for the technical support.

Grant Support: NIH grant CA101860 (M.S. Wicha), CA29765 and P36CA45692 (M.S. Wicha), support from the Taubman Institute (M.S. Wicha), Institut National du Cancer grants PHRC 2006, PHRC 2007 (E.C. Jauffret), and PL2008 (D. Birnbaum), Institut Lilly grant AO 2006 (E.C. Jauffret), and Ligue Nationale contre le Cancer grant Label DB 2007 (D. Birnbaum).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.