Aluminium chloride promotes anchorage-independent growth in human mammary epithelial cells

ABSTRACT

Aluminium salts used as antiperspirants have been incriminated as contributing to breast cancer incidence in Western societies. To date, very little or no epidemiological or experimental data confirm or infirm this hypothesis. We report here that in MCF-10A human mammary epithelial cells, a well-established normal human mammary epithelial cell model, long-term exposure to aluminium chloride (AlCl₃) concentrations of 10–300 µ m, i.e. up to 100 000-fold lower than those found in antiperspirants, and in the range of those recently measured in the human breast, results in loss of contact inhibition and anchorage-independent growth. These effects were preceded by an increase of DNA synthesis, DNA double strand breaks (DSBs), and senescence in proliferating cultures. AlCl₃ also induced DSBs and senescence in proliferating primary human mammary epithelial cells. In contrast, it had no similar effects on human keratinocytes or fibroblasts, and was not detectably mutagenic in bacteria. MCF-10A cells morphologically transformed by long-term exposure to AlCl₃ display strong upregulation of the p53/p21^Waf1 pathway, a key mediator of growth arrest and senescence. These results suggest that aluminium is not generically mutagenic, but similar to an activated oncogene, it induces proliferation stress, DSBs and senescence in normal mammary epithelial cells; and that long-term exposure to AlCl₃ generates and selects for cells able to bypass p53/p21^Waf1-mediated cellular senescence. Our observations do not formally identify aluminium as a breast carcinogen, but challenge the safety ascribed to its widespread use in underarm cosmetics. Copyright © 2012 John Wiley & Sons, Ltd.

INTRODUCTION

Breast cancer mainly affects industrialized countries where its incidence has significantly increased in the past decades. A change in the topological distribution of breast cancers has also consistently been reported, with the majority of tumours arising nowadays in the external part of the mammary gland, thus pointing to underarm cosmetics as potential contributors (Darbre, 2003, 2005). Aluminium chloride (AlCl₃) and aluminium chlorohydrate (Al₂Cl(OH)₅), the most common antiperspirants, are tolerated in these cosmetics at concentrations up to 20% (approximately 1  m; http://www.fda.gov/RegulatoryInformation/Dockets/ucm130350.htm). The precise mechanism by which aluminium reduces perspiration is not known, but this effect is believed to result from physical occlusion of the sweat duct, chemical inhibition of the sweat gland, or a combination of both (Exley, 2004).

In aqueous solutions at pH 7.0, both aluminium chloride and aluminium chlorohydrate yield aluminium hydroxide and they are absorbed through human and mouse skin (Flarend et al., 2001; Guillard et al., 2004; Anane et al., 1995, 1997). Transdermal uptake of aluminium contained in antiperspirants is not the only route by which this element accumulates in the body, aluminium-based compounds being found in pharmaceuticals, food additives and various household products (Krewski et al., 2007). However, the daily use of antiperspirants, with the application of aluminium salts that are not washed off, often in the context of hair shaving procedures that may facilitate aluminium absorption through the creation of small wounds, is likely to represent the main source of aluminium for the underarm dermis and its underlying tissues, including the mammary epithelium.

Aluminium is not essential to biological systems. Overall, little is known about its toxic effects at the cellular and systemic levels, although it is generally acknowledged that aluminium is neurotoxic (Krewski et al., 2007; Kawahara and Kato-Negishi, 2011). With respect to the breast, relatively high levels of aluminium have been measured in the normal human breast adjacent to breast tumours (Exley et al., 2007), in the nipple aspirate fluid collected from women affected by breast cancer (Mannello et al., 2011), and in the fluid collected from breast cysts observed in gross cystic breast disease (Mannello et al., 2009), a benign breast disorder believed to facilitate the appearance of breast cancer. The relatively high levels of aluminium in these settings may result from intrinsic metabolic features of the tissue analysed, thus being purely correlative in nature. However, as suggested by the authors of these papers, they also raise the possibility that aluminium plays a causal role in the pathogenesis or progression of breast tumours, thus being present at higher levels in people who developed breast cancer, or at higher risk of it.

Although considered harmless by many, as with other putative carcinogens, aluminium needs to be tested in models more appropriate than conventional toxicological tests using rodents or bacteria because of well-known cell type- and species-specific effects of cancer genes and carcinogens. Accordingly, current procedures of chemical testing and risk assessment have recently been questioned (Hunt, 2011; David and Zimmerman, 2010). In MCF-7 human breast cancer cells, aluminium has been reported to interfere with the binding of estradiol to estrogen receptor (ER) and to enhance transcription from an estrogen-responsive element in a reporter system, thus suggesting that aluminium is a metalloestrogen (Darbre, 2005, 2006). A genotoxic effect of aluminium has been reported in several systems, including human peripheral blood lymphocytes. Although the molecular basis of this effect has not been defined, it has been hypothesized to be mediated by direct interaction of aluminium with DNA, by the induction of oxidative DNA damage, by the release of lysosomal DNase, or by interference with normal microtubule function or with cellular DNA repair mechanisms (Banasik et al., 2005; Lankoff et al., 2006; and references therein). Taken together, these results suggest that aluminium might play a causative role in human cancer. However, the possibility that aluminium promotes morphological and functional changes typical of the first phases of malignant transformation in normal human mammary epithelial cells has not been addressed.

To investigate the possibility that long-term exposure to aluminium – as it occurs in the human mammary epithelium owing to daily use of antiperspirants – promotes carcinogenesis of the human mammary gland epithelium, we performed long-term cultures of human mammary epithelial cells where aluminium chloride was renewed twice a week with fresh culture medium. Since we were interested in a potential process of cellular transformation induced by aluminium in the normal human mammary epithelium, we chose the MCF-10A cell line, a well-established and widely recognized model of normal human mammary epithelium (Soule et al., 1990; Miller et al., 1993; Debnath and Brugge, 2005), as well as primary human mammary epithelial cells. The MCF-10A cell line is closely related to, and was originally isolated from the same patient as, the human mammary epithelial MCF-10F cell line, extensively used to demonstrate the transforming effect of 17β-estradiol, bisphenol A and butyl benzyl phthalate (Fernandez and Russo, 2010). Using MCF-10A cells and primary human mammary epithelial cells, we studied aluminium chloride for its capacity to induce effects typical of the first phases of cellular transformation, including cellular senescence and the capacity to proliferate in the soft agar assay. The latter is the standard method to measure anchorage-independent growth, a key hallmark of cultured tumour cells (Freedman and Shin, 1974).

MATERIALS AND METHODS

Cell Culture

MCF-10A cells were purchased from ATCC (Manassas, VA, USA). The cells were purchased at passage 98 and used between passage 99 and 125. HaCaT spontaneously immortalized human keratinocytes (Fusenig and Boukamp, 1998) were purchased from Cell Lines Service (Eppelheim, Germany). The cells were purchased at passage 40 and used between passages 41 and 60. C26Ci spontaneously immortalized human colonic fibroblasts (Forsyth et al., 2004) were kindly provided by Dr J.W. Shay. The cells arrived at a nonspecified passage number and used between the first and the twentieth passage after arrival. Primary human mammary epithelial cells were purchased from Lonza (catalogue no. CC-2551). The cells were purchased at passage 6 and used between passages 8 and 11. The cells were cultured as described (Mandriota et al., 2010).

Chemicals

Aluminium chloride hexahydrate (Sigma catalogue no. 06232, or Alexis Biochemicals catalogue no. ALX-400-033-G005, purity ≥ 99.0%) was dissolved in H₂O at the concentration of 1  m and immediately diluted at the intermediate concentrations of 300, 100 or 10 m m in H₂O. These stocks were either used immediately or stored at 4 °C until used (no more than 10 weeks after preparation). Under these conditions, precipitation of aluminium owing to polymerization is expected to be minimal and, accordingly, we did not observe precipitates. The stocks, or H₂O as a control, were diluted 1:1000 in fresh culture medium twice a week for long-term culture, or for shorter treatments as indicated. Under the conditions used, the addition of AlCl₃ had little effect on the pH of the MCF-10A cell culture medium (H₂O: pH = 7.50 ± 0.04; AlCl₃ 10 µ m: pH = 7.40 ± 0.06; AlCl₃ 100 µ m: pH = 7.41 ± 0.06; AlCl₃ 300 µ m: pH = 7.33 ± 0.03, n = 2). The three commercially available liquid antiperspirants D, N or R were diluted 1:10 in H₂O and either used immediately or kept at 4 °C until used (no more than 10 weeks after preparation). These intermediate dilutions were further diluted in fresh cell culture medium twice a week for long-term culture at the final dilution of 1:50 000.

Determination of Aluminium in Antiperspirants

Aluminium in the commercially available antiperspirants D, N, R was measured by inductively coupled plasma optical emission spectrometry (ICP-OES), as described in protocol EPA200.7 (Environmental Protection Agency, USA) using an ICP-OES Optima 2100 DV instrument (Perkin-Elmer, Rotkreuz, Switzerland). These analyses were carried out by Scitec-Research Laboratory (Lausanne, Switzerland).

Soft Agar Assay

Soft agar assay with MCF-10A or HaCaT cells cultured in the presence of the indicated AlCl₃, GaCl₃ or InCl₃ concentrations, or diluted antiperspirants, as indicated, was performed as described (Mandriota et al., 2010). AlCl₃, GaCl₃, InCl₃ or diluted antiperspirants were not added to the soft agar assays.

Proliferation Assays

MCF-10A cells were seeded in six-well plates at the density of 5000 cells per well in triplicate, grown in the presence of the indicated concentrations of AlCl₃ for 7 days and counted using a haemocytometer.

Annexin V Staining

MCF-10A cells were seeded in 75 cm² flasks at the density of 42 000 cells per flask and grown in the presence of AlCl₃ 300 µ m for 4 days. Apoptosis was measured by flow cytometry using the FITC Annexin V/ Dead Cell Apoptosis Kit (Invitrogen catalogue no. V13242).

Senescence-associated β-Galactosidase Staining

MCF-10A cells were seeded in six-well plates at the density of 100 000 cells per well and grown for 7 days in the presence or absence of AlCl₃ as indicated. The fraction of senescent cells was counted using the Senescence β-Galactosidase Staining Kit (Cell Signaling catalogue no. 9860).

γ-H2AX Immunofluorescence

Immunofluorescence with phospho-Ser139 H2AX antibody JBW301 (catalogue no. 05–636, Millipore, Zug, Switzerland) was performed according to manufacturer's instructions. For the quantification of γ-H2AX foci decay after irradiation, MCF-10A cells were seeded in 24-well plates in triplicate at the density of 20 000 cells per well, incubated for 16 h in the presence or absence of AlCl₃ 300 µ m, irradiated with X-rays (1 Gy) using an XRAD 320 irradiator (Precision X-Ray Inc., North Branford, CT, USA), and then stained for γ-H2AX at the indicated time points. In the quantification, the number of foci was normalized with respect to the number of foci in unirradiated cells.

Array Comparative Genomic Hybridization

DNA was extracted from cells following standard protocols. The array comparative genomic hybridization (CGH) analysis was performed using Agilent human genome CGH microarray kit 44 K (Santa Clara, CA, USA). Labelling and hybridization were performed following the protocols provided by the manufacturer. All slides were scanned on an Agilent DNA microarray scanner. Data were obtained by Agilent feature extraction software version 9. Data were analysed with Agilent Workbench version 6.0 software, using the statistical algorithms zscore and ADM-2 according to a sensitivity threshold at 2.5 and 6.0, respectively, and a moving average window of 0.2 Mb. Mapping data were analysed on the human genome sequence using the UCSC genome browser hg19. Copy number variations were checked in the Database of Genomic Variants (DGV, GRCh37).

Mutagenesis in Bacteria

Independent cultures of CC105 (Nghiem et al., 1988) a Lac⁻ Rif^SE. coli strain, were scored for the presence of Lac⁺ revertants (indicating AT to TA transversions) and Rif^R forward mutants (detecting all base pair substitutions). Control experiments showed that AlCl₃ had no detectable effect on the bacterial doubling time, even at 1 m m. Strain ML3 (Monod and Audureau, 1946) is an Lac⁻ strain that carries an 11 base pair duplication in lacY. The presence of three repeats is unstable and Lac⁺ bacteria appear at high frequency, precluding an analysis of mutation rate. The indicated values represent the number of Lac⁺ revertants (left) and of total bacteria (right) in 2 ml overnight cultures.

5-Ethynyl-2′-deoxyuridine Labelling

MCF-10A cells were seeded in 24-well plates at the density of 20 000 cells per well in quadruplicate, incubated in the presence or absence of the indicated concentrations of AlCl₃ for 16 h, labelled with 20 µ m 5-ethynyl-2′-deoxyuridine (EdU) for 1 h and counted according to the manufacturer's instructions (Click-iT® EdU Alexa Fluor® 488 Imaging Kit, Invitrogen catalogue no. C10337).

Western Blotting

Cells were lysed using RIPA buffer (catalogue no. R0278, Sigma-Aldrich) supplemented with Halt Protease Inhibitor Cocktail (catalogue no. 78410, Pierce/Thermo Fisher Scientific, Lausanne, Switzerland) and Phosphatase Inhibitor Cocktail (catalogue no. 78420, Pierce/Thermo Fisher Scientific) and sonicated. Aliquots of 50–80 µg of protein were run in SDS–PAGE gels and transferred to nitrocellulose membranes (catalogue no. RPN 303 D, Amersham/GE Healthcare Life Sciences, Glattbrugg, Switzerland). Blots were incubated with antibodies against p53 (catalogue no. 628201, Biolegend/Lucerna Chem AG, Luzern, Switzerland), p53 Ser15-p (catalogue no. PC386, Merck Chemicals Ltd), p21^WAF1/CIP1 (catalogue no. M7202, Dako, Baar, Switzerland), Phospho-pRb (Ser807/811; catalogue no. 9308, Cell Signaling Technology, Danvers, MA, USA) or pRb (4H1; catalogue no. 9309, Cell Signaling Technology) according to the manufacturer's instructions. Horseradish peroxidase-conjugated secondary antibody/primary antibody complexes were revealed on the nitrocellulose membranes using the Lumi-Light (catalogue no. 2015200, Roche Applied Science, Rotkreuz, Switzerland) or Lumi-Light Plus (cat. 12015196001, Roche Applied Science) western blotting substrates.

Quantitative Real-time Polymerase Chain Reaction

cDNA was synthesized from 1 µg of total RNA using random hexamers and Superscript III reverse transcriptase (Invitrogen) according to the manufacturer's instructions. SYBR Green assays were designed using the program Primer Express version 2.0 (Applied Biosystems, Rotkreuz, Switzerland) with default parameters. Amplicon sequences were aligned against the human genome by BLAST to ensure that they were specific for the gene being tested. Oligonucleotides were purchased from Invitrogen. The efficiency of each design was tested with serial dilutions of cDNA. Oligonucleotide primer sequences for p16/INK4a analysis were as follows: forward, 5′-AGAACCAGAGAGGCTCTGAGAAAC-3′; reverse, 5′-GTAGGACCTTCGGTGACTGATGAT-3′. Polymerase chain reactions (PCRs; 10 µl volume) contained diluted cDNA, 2× Power SYBR Green Master Mix (Applied Biosystems), and 300 n m forward and reverse primers. PCR was performed on a SDS 7900 HT instrument (Applied Biosystems) with the following parameters: 50 °C for 2 min, 95 °C for 10 min, and 45 cycles of 95 °C for 15 s and 60 °C for 1 min. Each reaction was performed in three replicates on a 384-well plate. Raw C_t values obtained with SDS version 2.2 (Applied Biosystems) were imported in Excel, and normalization factor and fold changes were calculated using the GeNorm method (Vandesompele et al., 2002).

Statistical Analysis

Values were analysed by two-sided t-test where indicated.

RESULTS

When cultured for 9 weeks in the presence of a 1:50 000 dilution of three commercially available liquid antiperspirants (referred to as D, N, R) containing several chemicals, including Al₂Cl(OH)₅, MCF-10A human mammary epithelial cells formed colonies in the soft agar assay, the standard method to distinguish transformed cells from normal cells in vitro (Fig. 1a, b). ICP-OES revealed that these three antiperspirants contained Al₂Cl(OH)₅ at the concentrations of 653.48, 417.31 and 483.81 m m, respectively, thus resulting in an Al₂Cl(OH)₅ concentration of 13.07, 8.35 and 9.68 µ m, respectively, in the cell culture medium, at the 1:50 000 dilution used. Lower dilutions (1:25 000; 1:10 000) of the antiperspirants were slightly toxic to the cells, i.e. they led to cell vacuolization and death after a few days, and were not studied further.

Figure 1. Commercially available antiperspirants containing Al₂Cl(OH)₅ induce anchorage-independent growth in MCF-10A human mammary epithelial cells. (a) MCF-10A cells cultured for 9 weeks in the presence of a 1:50 000 dilution of the commercially available antiperspirants D, N or R, corresponding to a final Al₂Cl(OH)₅ concentration of 13.07, 8.35 or 9.68 µ m, respectively, in the cell culture medium, or the equivalent volume of H₂O, were grown in soft agar for 14 days in the absence of the diluted antiperspirants. Bar = 100 µm. (b) The growth in agarose gels was quantified by measuring the diameter of the structures (single cells or multicellular colonies) formed after 14 days. Randomly selected structures (single cells or multicellular colonies) from two independent experiments/condition were measured. The number of measured structures (single cells or multicellular colonies) was as follows: H₂O, 227; D, 231; N, 123; R, 111. P-value for D, N or R vs H₂O < 0.01. P-value for D vs N, N vs R or D vs R > 0.1; two-sided t-test.

To directly investigate the possibility that the transforming effect of antiperspirants could be ascribed to aluminium, MCF-10A cells were chronically cultured in the presence of increasing concentrations of pure AlCl₃. Pure AlCl₃ was used in the remainder of this study. At the expected pH of the cell culture medium (pH ~7.2), AlCl₃ and Al₂Cl(OH)₅ yield the same dissociation product, aluminium hydroxide. When cultured for 6–9 weeks in the presence of 10–300 µ m AlCl₃, MCF-10A cells displayed focal loss of contact inhibition (Fig. 2) and formed colonies in the soft agar assay (Fig.3a, b). The time of appearance of the capacity to grow in agar (approximately 9 weeks of continuous AlCl₃ administration, as detailed in Materials and Methods) and the size of the colonies formed in soft agar by AlCl₃-treated MCF-10A cells were consistent with the concentrations of aluminium used in experiments with antiperspirants (Fig. 1a, b). Growth in agar was retained 5 weeks after AlCl₃ had been suspended (Fig. 3a, b), thus showing that growth in agar had become autonomous. AlCl₃ had no effect on p53^−/− HaCaT human keratinocytes or on C26Ci human colonic fibroblasts (an example of these experiments is shown in Fig. 4). Chlorides of gallium or indium that, as aluminium, belong to the thirteenth group of the Mendeleev Table and hence have similar chemical properties had no effect (Fig. 3b). AlCl₃ had no detectable mutagenic effect in bacteria, as assessed on the basis of the number of Lac⁺ revertants or rifampicin resistant (Rif^R) cells in CC105 (Lac⁻ Rif^S) cultures or of Lac⁺ revertants in ML3 (Lac⁻) cultures. Bacteria were grown overnight in LB medium, for at least 20 generations, in the presence or absence of 300 µ m AlCl₃ (Fig. 5). No chromosomal gains or losses were detected in AlCl₃-treated MCF-10A cells compared with controls by CGH array (data not shown), that efficiently detects relatively large genomic imbalances occurring in overt tumours, but not more subtle mutations occurring in earlier tumour lesions (Habermann et al., 2007).

Figure 2. AlCl₃ induces loss of contact inhibition in MCF-10A cells. MCF-10A cells were cultured for 6 weeks in the presence of AlCl₃ 100 µ m (b) or an equivalent volume of H₂O (solvent) alone (a). The treatment was renewed twice a week with fresh culture medium. The cells were photographed under phase contrast. Bar = 100 µm.

Figure 3. AlCl₃ induces anchorage-independent growth in MCF-10A cells. (a) MCF-10A cells cultured for 9 weeks in the presence of the indicated AlCl₃ concentrations or the equivalent volume of H₂O were grown in soft agar for 14 days in the absence of AlCl₃. ‘AlCl₃ 300 µ m-stop’ are MCF-10A cells that were left untreated for 5 weeks before being tested in agar. Bar = 100µm. (b) The growth in agarose gels of MCF-10A cells cultured for 9 weeks (AlCl₃) or 14 weeks (GaCl₃, InCl₃) in the presence of the indicated concentrations of AlCl₃, GaCl₃, InCl₃ or the same volume of H₂O (solvent) as a control was quantified by measuring the diameter of the structures formed after 14 days. The growth of MCF-10A cells that after 9 weeks incubation in the presence of AlCl₃ 300 µ m were left untreated for 5 weeks (AlCl₃ 300 µ m stop) before being tested in agar is also shown. Randomly selected structures (single cells or multicellular colonies) from two independent experiments per condition were measured. The number of measured structures (single cells or multicellular colonies) was as follows: H₂O, 121; AlCl₃ 10 µ m, 100; AlCl₃ 100 µ m, 124; AlCl₃ 300 µ m, 100; AlCl₃ 300 µ m-stop, 50; GaCl₃ 100 µ m, 121; InCl₃ 100 µ m, 121. P-value for AlCl₃ vs H₂O < 0.01 for all AlCl₃ concentrations (including the condition ‘AlCl₃ 300 µ m stop’). P-value for GaCl₃ or InCl₃ vs H₂O > 0.2; two-sided t-test. AlCl₃, GaCl₃ or InCl₃ were not added to the soft agar assays. Few or no differences were observed if AlCl₃ was not discontinued in the soft agar assay (not shown).

Figure 4. AlCl₃ does not induce anchorage-independent growth in HaCaT keratinocytes. HaCaT keratinocytes cultured for 17 weeks in the presence of AlCl₃ 300 µ m or the same volume of H₂O as a control were suspended in low-gelling temperature agarose and grown for 14 days in the absence of AlCl₃. Bar = 100 µm.

Figure 5. AlCl₃ is not detectably mutagenic in bacteria. Bacterial strains CC105 (upper panel) or ML3 (bottom panel) were grown in the presence or absence of 300 µ m AlCl₃ and tested for the indicated mutations through the counting of Lac⁺ revertants or Rifampicin resistants (Rif^R) as indicated; t-test was two-sided.

Although long-term exposure to AlCl₃ clearly confers a transforming growth advantage to MCF-10A cells, in 7 day proliferation assays AlCl₃ slightly decreased the number of cells (Fig. 6a) in the absence of overt cellular toxicity or apoptosis. Annexin V staining revealed no differences in apoptosis between controls and AlCl₃ treated MCF-10A cells after 4 days of treatment (Fig. 6b). In contrast, AlCl₃ increases the percentage of senescence-associated β-galactosidase-positive cells in proliferating MCF-10A cultures after 7 days (Fig. 6c). Similarly, AlCl₃ increases the expression of p16/INK4a, a well-defined marker of cellular senescence (Collado and Serrano, 2010), in proliferating primary human mammary epithelial cells after 7 days (Fig. 7).

Figure 6. AlCl₃ induces cellular senescence in MCF-10A cells. (a) MCF-10A cells were seeded in six-well plates at the density of 5 000 cells per well in triplicate, grown in the presence of the indicated concentrations of AlCl₃ for 7 days and counted using a haemocytometer. AlCl₃ was renewed with fresh medium 3 days after the initial seeding. The graph shows the mean number of cells ± SEM from six wells from two independent experiments. P-value for AlCl₃ 100 µ m vs H₂O = 0.07; P-value for AlCl₃ 300 µ m vs H₂O < 0.001, two-sided t-test. (b) MCF-10A cells were seeded in 75 cm² flasks at the density of 42 000 cells per flask and grown in the presence of the indicated concentrations of AlCl₃, or the same volume of H₂O as a control, for 4 days. Apoptosis was measured by flow cytometry by measuring the fraction of Annexin V positive cells. The graph represents the average fraction of annexin V positive cells ± SEM from three independent experiments. P-value for AlCl₃ 100 µ m vs H₂O = 0.72; P-value for AlCl₃ 300 µ m vs H₂O = 0.87, two-sided t-test. (c) MCF-10A cells were seeded in six-well plates at the density of 100 000 cells per well, grown for 7 days in the presence or absence of AlCl₃ as indicated, and stained for senescence associated (SA) β-galactosidase. The graph shows the percentage of SA β-galactosidase positive cells ± SEM in at least seven photographic fields; at least 2100 cells per condition from two independent experiments were counted. P-value for AlCl₃ vs H₂O < 0.0001 for all the AlCl₃ concentrations shown in the figure; two-sided t-test.

Figure 7. AlCl₃ increases p16-INK4a expression in primary human mammary epithelial cells. Proliferating primary human mammary epithelial cells were incubated for 7 days in the presence of the indicated concentrations of AlCl₃, or the same volume of H₂O. AlCl₃ was renewed in fresh culture medium 3 days after the initial seeding. At the end of the incubation, the cells were analysed for the expression of p16-INK4a by quantitative real-time PCR. The graph shows the average ± SD from two independent experiments.

Increased DNA double strand breaks (DSBs) resulting from defective DNA repair or oncogene-induced proliferation stress are key triggers of cellular senescence (Collado and Serrano, 2010). By counting phosphorylated histone H2AX (γ-H2AX) nuclear foci, a validated marker of DSBs, we found that AlCl₃ increases DSBs in a dose- and time-dependent manner in proliferating MCF-10A cells (Fig. 8). Importantly, AlCl₃ also increased γ-H2AX nuclear foci in proliferating primary human mammary gland epithelial cells (number of γ-H2AX foci per cell: H₂O: 4.06 ± 0.82; AlCl₃ 100 µ m: 6.89 ± 2.97; AlCl₃ 300 µ m: 9.50 ± 2.18; P-value for AlCl₃ 100 µ m vs H₂O = 0.01, P-value for AlCl₃ 300 µ m vs H₂O < 0.0001, two-sided t-test; at least 50 cells from at least seven independent photographic fields per condition were counted; incubation with H₂O or AlCl₃ was for 16 h; see also Fig. S1 in Supporting Information), but had little or no effect on proliferating HaCaT keratinocytes (number of γ-H2AX foci per cell: H₂O: 5.75 ± 3.10; AlCl₃ 100 µ m: 7.65 ± 3.23; AlCl₃ 300 µ m: 7.86 ± 2.42; P-value for AlCl₃ 100 µ m vs H₂O = 0.41, P-value for AlCl₃ 300 µ m vs H₂O = 0.21, two-sided t-test; at least 100 cells from at least seven independent photographic fields per condition were counted; incubation with H₂O or AlCl₃ was for 16 h).

Figure 8. AlCl₃ increases DNA double-strand breaks (DSBs) in MCF-10A cells. (a) Number of γ-H2AX nuclear foci in MCF-10A cells incubated in the presence of the indicated concentrations of AlCl₃ or the same volume of H₂O for the times indicated. Values represent the mean ± SEM from three independent experiments. At 1 and 16 h, P-value for AlCl₃ vs H₂O < 0.002 for all AlCl₃ concentrations, two-sided t-test. (b) An example of the γ-H2AX immunofluorescence after 16 h incubation.

The lack of detectable mutagenic effect in bacteria (Fig. 5) and the slow kinetics of DSB induction in MCF-10A cells exposed to AlCl₃ (Fig.8) suggested that AlCl₃ might increase DSBs in MCF-10A cells indirectly, for example by interfering with the repair of naturally occurring DSBs, thus resulting in persistent γ-H2AX foci. Experiments counting γ-H2AX foci in X-rays irradiated MCF-10A cells in the presence or absence of AlCl₃ ruled out this hypothesis (Fig. 9).

Figure 9. AlCl₃ does not delay the decay of γ-H2AX nuclear foci in MCF-10A cells after ionizing radiation. MCF-10A cells were seeded in 24-well plates in triplicate at the density of 20 000 cells per well, incubated for 16 h in the presence of AlCl₃ 300 µ m (+) or the same volume of H₂O (−), irradiated with X-rays (1 Gy) and then stained for γ-H2AX at the indicated time points. The graph shows the number of γ-H2AX nuclear foci ± SEM from at least five independent photographic fields where at least 70 cells per condition from two independent experiments were counted. In the quantification, the number of foci was normalized with respect to the number of foci in unirradiated cells. At 8 h: P-value for AlCl₃ vs H₂O = 0.91; at 24 h: P-value for AlCl₃ vs H₂O = 0.53; two-sided t-test.

In contrast to proliferating MCF-10A cells, AlCl₃ did not increase γ-H2AX foci in post-confluent MCF-10A cells (number of γ-H2AX foci per cell: H₂O: 6.20 ± 1.47; AlCl₃ 300 µ m: 6.34 ± 1.31; P-value for AlCl₃ 300 µ m vs H₂O = 0.89, two-sided t-test; at least 100 cells from at least four independent photographic fields per condition were counted; incubation with H₂O or AlCl₃ was for 16 h). In addition, AlCl₃ increased the fraction of EdU-positive cells (Fig. 10a) and retinoblastoma protein (pRb) phosphorylation (Fig. 10b) in proliferating MCF-10A cells after 16 h.

Figure 10. AlCl₃ increases DNA synthesis and retinoblastoma (pRb) phosphorylation in proliferating MCF-10A cells. (a) MCF-10A cells were seeded in 24-well plates at the density of 20 000 cells per well in quadruplicate, incubated in the presence of the indicated concentrations of AlCl₃ or an equivalent volume of H₂O for 16 h, labelled with 20 µ m 5-ethynyl-2′-deoxyuridine (EdU) for 1 h and counted. The graph shows the percentage of EdU positive cells ± SEM from three independent experiments. P-value for AlCl₃ 100 µ m vs H₂O = 0.003; P-value for AlCl₃ 300 µ m vs H₂O < 0.0001; two-sided t-test. (b) Proliferating MCF-10A cells cultured for 16 h in the presence of 100 µ m AlCl₃ or the same dilution of H₂O were lysed in the presence of phosphatase inhibitors and analysed by western blotting for the levels of pRb or pRb phosphoSer807/811. Numbers on the left indicate kilo-daltons. One of two experiments with similar results is shown.

Taken together, our results suggest that AlCl₃ induces proliferation stress in MCF-10A cells, thus resulting in increased DSBs and senescence. Long-term AlCl₃ administration selects for cells that bypass senescence. Accordingly, MCF-10A cells morphologically transformed by long-term AlCl₃ administration display strong upregulation of the p53/p21 pathway (Fig. 11), a key mediator of growth arrest and cellular senescence (Collado and Serrano, 2010). AlCl₃-treated MCF-10A cells did not form tumours in unirradiated Nod/Scid mice (data not shown). However, only a fraction of human tumour cells grow in mouse models (Kelly et al., 2007).

Figure 11. Morphologically transformed MCF-10A cells cultured in the presence of AlCl₃ show upregulation of the p53/p21 pathway. MCF-10A cells cultured for 10 weeks in the presence of the indicated concentrations of AlCl₃ or the same volume of H₂O as a control were lysed in the presence of phosphatase inhibitors and analysed by western blotting for the levels of p53, p53 phosphoSer15, p21Waf1/Cip1, or β-actin as a loading control. AlCl₃ was renewed twice a week with fresh culture medium starting from 1000× concentrated stocks. Numbers on the left indicate kilo-daltons. Two different western blots are shown.

DISCUSSION

Epidemiological and clinical data show that the incidence of breast cancer has increased in Western countries over the past few decades, with the majority of tumours being detected nowadays in the upper outer quadrant of the breast (Darbre, 2003, 2005), an anatomic region close to the axilla, where antiperspirants are applied on a daily basis by the large majority of the population. Based on this observation, on the large array of effects by which aluminium has been reported to affect cellular physiology, including genotoxicity (Kawahara and Kato-Negishi, 2011; Banasik et al., 2005; Lankoff et al., 2006), on its well-established neurotoxic profile (Kawahara and Kato-Negishi, 2011; Krewski et al., 2007), and on its high concentration in antiperspirants, it has been hypothesized that aluminium might play a causal role in the carcinogenesis of the human mammary epithelium (Darbre, 2003, 2005). However, very little or no experimental data supporting this hypothesis exist. Aluminium has been reported to interfere with estrogen binding to its receptor and to enhance transcriptional activity from an estrogen receptor-regulated reporter (Darbre, 2005). Although this activity might be relevant to breast carcinogenesis, processes related to cellular transformation of the normal human mammary epithelium could not be assessed in this study since it used MCF-7 cells, a well-established breast cancer – and therefore already transformed – cell line.

Aluminium is absorbed through the human (Flarend et al., 2001; Guillard et al., 2004) and mouse (Anane et al., 1995, 1997) skin, and it is conceivable that the daily application of antiperspirants, containing high concentrations of aluminium, represents a major route through which the human mammary epithelium is exposed to aluminium. Interestingly, although aluminium concentrations in the body are generally low, levels measured in the breast area (including milk) are relatively high (Mannello et al., 2009, 2011; Exley et al., 2007). Whether this is due to antiperspirant use or other routes of exposure, and whether it reflects intrinsic metabolic specificities of the breast, is currently not known. Independent of the route of exposure, these measurements provide a reliable estimation of the amount of aluminium to which the human mammary epithelium is exposed in vivo, and therefore a relevant range of concentrations to test for potential deleterious effects in the experimental and toxicological settings.

In this study we report that, within this range of concentrations, long-term exposure of MCF-10A human mammary epithelial cells to aluminium results in anchorage-independent growth, a key hallmark of cultured tumour cells and of cells on the way to malignant transformation. Further analysis showed that shorter (7 days) exposure to aluminium results in diminished cell numbers owing to cellular senescence. Since one of the best-characterized causes of cellular senescence is an increase in DSBs (Collado and Serrano, 2010), we studied aluminium for its capacity to induce DSBs. We found that in both MCF-10A cells and primary human mammary epithelial cells aluminium increases DSBs in a dose-dependent manner. Interestingly, compared with more direct mutagens as neocarzinostatin or ionizing radiations that increase DSBs within minutes, induction of DSBs by aluminium occurred slowly, thus raising the possibility that this effect is indirect and possibly cell-type specific, i.e. not attributable to a direct or general mutagenic effect of aluminium at the DNA level, although other possibilities (e.g. slow uptake) could be not excluded. This hypothesis was supported by experiments showing that aluminium is not detectably mutagenic in conventional mutagenic assays in bacteria, and that it does not induce loss of contact inhibition or anchorage-independent growth in human keratinocytes or fibroblasts, despite the fact that the keratinocytes we used (the HaCaT cell line) are prone to cellular transformation because of biallelic inactivation of p53 (Fusenig and Boukamp, 1998). Subsequent experiments showed that aluminium does not delay the decay of IR-induced γ-H2AX nuclear foci, a well-established marker of DSBs, thus suggesting that it does not increase DSBs by delaying the repair of naturally occurring DSBs. Finally, aluminium increases DSBs in proliferating but not post-confluent MCF-10A cells, thus suggesting that it alters normal ongoing proliferation by inducing proliferation stress. Consistent with this hypothesis, aluminium increased DNA synthesis (as assessed by EdU incorporation) in proliferating MCF-10A cells.

This effect of aluminium on cellular proliferation, cellular senescence and DSB numbers is strikingly similar to that of activated oncogenes like ras, which increases DSBs and cellular senescence by inducing proliferation stress when transfected into normal cells (Collado and Serrano, 2010). We found this effect surprising for aluminium, present in antiperspirants at high concentrations because, presumably, it is considered harmless for the breast area. In contrast to aluminium, gallium or indium, which belong to the same group as aluminium in the Mendeleev Table, did not induce anchorage-independent growth in MCF-10A cells, showing that the effects reported in this study are specific for aluminium.

A genotoxic effect of aluminium has been reported in several systems, including human peripheral blood lymphocytes. In the latter cell type, this effect correlated with the induction of apoptosis (Banasik et al., 2005; Lankoff et al., 2006). Although the molecular basis of the genotoxic effect of aluminium has not been defined, it has been hypothesized to be mediated by direct interaction of aluminium with DNA, by the induction of oxidative DNA damage, by the release of lysosomal DNase, or by interference with normal microtubule function or with cellular DNA repair mechanisms (Banasik et al., 2005; Lankoff et al., 2006; Kawahara and Kato-Negishi, 2011). As discussed above, the effect of aluminium on human mammary epithelial cells reported in this study seems to be mediated in part by proliferation stress. However, we cannot exclude the possibility that one or more of the mechanisms hypothesized in the studies mentioned above contribute to the effects we observed in human mammary epithelial cells, with the exception of the interference with DNA repair, which does not seem to occur in the mammary epithelial cells we used. As already mentioned, when looking for a potential direct mutagenic effect of aluminium by means of classical mutagenic assay in bacteria, we did not find evidence for it. Taken together with the other results reported in this study, including the lack of transforming effect of aluminium on human keratinocytes or fibroblasts, and with the results on the genotoxic effect of aluminium reported in the literature, aluminium seems to have a complex effect on cellular physiology, and to act at the level of several different molecular targets regulating several distinct cellular functions, possibly including aluminium cellular uptake, cellular proliferation, apoptosis, cellular senescence or DNA repair. The outcome of this interaction is likely to depend on the particular gene expression profile and thus on the particular cell type considered, but also on the specific parameters studied. For example, the senescent response to aluminium of human mammary epithelial cells reported in this study, or the apoptotic response of human lymphocytes following exposure to similar concentrations of AlCl₃ (Lankoff et al., 2006) might simply reflect different responses to the same DNA damage, as it is known that different cell types may choose either pathway to respond to the same DNA damage. Of the DSBs induced by aluminium in human mammary epithelial cells, it is likely that some are not repaired properly, thus resulting in stable genetic defects. Accordingly, MCF-10A cells treated with aluminium for several weeks grow in agar several weeks after aluminium treatment has been interrupted. The proliferation pathway with which aluminium interferes in human mammary epithelial cells remains to be identified. Whereas it could be argued that it corresponds to or is related to the estrogen receptor pathway on the basis of results previously reported in MCF-7 cells (Darbre, 2005), it must be pointed out that, in contrast to MCF-7 cells, which express both ERα and ERβ, MCF-10A cells, like the related cell line MCF-10 F, express ERβ but not ERα (Fernandez and Russo, 2010; and our unpublished results). Therefore the potential involvement of the ER pathway in the results presented in this study requires further investigation.

In summary, this is the first report showing that aluminium recapitulates the first steps of malignant transformation and increases DSBs and senescence in MCF-10A and primary human mammary epithelial cells at concentrations up to 100 000-fold lower than those commonly used in deodorants (http://www.fda.gov/RegulatoryInformation/Dockets/ucm130350.htm) and in the range of those recently found in the human breast (Exley et al., 2007; Mannello et al., 2009, 2011). Our findings do not provide conclusive evidence that aluminium is a breast carcinogen. However, they unveil an unsuspected biological activity of this element that, similar to activated oncogenes transfected into primary cells (Collado and Serrano, 2010), increases DSBs and senescence in normal human mammary epithelial cell models (this study). This effect raises concerns about the wide use of aluminium in underarm cosmetics, and together with the aluminium's lack of mutagenic or transforming effect in bacteria, HaCaT keratinocytes or fibroblasts, contributes to the recent questioning on the choice of appropriate models for risk assessment of chemicals (Hunt, 2011; David and Zimmerman, 2010) in specific cell types or organisms.

Acknowledgement

The authors declare no conflict of interest. This work was supported by grants from the Ligue Genevoise contre le Cancer; from the Fondation pour la Lutte contre le Cancer et pour des Recherches Médico-Biologiques; the Fondation André et Cyprien; the Fondation Prévot; and the Fondation Meyer. We thank Drs P. Chappuis, R. FitzGerald, O. Sorg and M. Wilks, and Professors R. Montesano, C. Piguet, T. Halazonetis and F. Fracassi for discussions, Professor T. Halazonetis for help with cell irradiation, B. Foglia and F. Silva for technical assistance and Professors S.E. Antonarakis and J.D. Vassalli for reading the manuscript and for support.