ATM is required for SOD2 Expression and Homeostasis within the Mammary Gland

Abstract

ATM, the gene mutated in the human disorder Ataxia-Telangiectasia (A-T), is a protein kinase activated in response to genotoxic and oxidative stress. Dysregulated response to redox imbalance is thought to drive many of the pathologies observed in A-T patients. Among the downstream pathways activated by ATM is the NF-κB transcriptional complex, and we have recently documented, in several cultured cell lines, that basal NF-κB activity is ATM-dependent. Here we show that expression of the NF-κB target gene and critical cellular antioxidant SOD2 (MnSOD) in cultured mammary epithelium is also dependent upon ATM, and that SOD mimetics partially rescue the pro-oxidant sensitivity displayed by ATM knockdown cells. To determine if ATM controls SOD2 expression in vivo, we engineered a mouse line with conditional deletion of Atm within the mammary epithelium using Cre-lox technology as mice with germline deletion of Atm fail to develop mature mammary glands. Confirming results observed in cultured lines, we determined that deletion of Atm resulted in significantly diminished expression of Sod2 within the mammary gland. We also observed that these mice (termed Atm^Δ/Δ) displayed a progressive lactation defect as judged by a reduction in pup growth rate, aberrant lobulo-alveolar structure, diminished expression of milk protein genes, and increased apoptosis within lactating mammary gland. This phenotype appears to be linked to dysregulated Sod2 expression as mammary gland-specific deletion of Sod2 phenocopies defects observed in lactating Atm^Δ/Δ dams. We conclude that ATM is required to promote expression of SOD2 within the mammary epithelium, and that both of these proteins play a crucial role in mammary gland homeostasis.

INTRODUCTION

Ataxia-telangiectasia mutated (ATM) is a protein kinase activated in response to DNA damage [1]. Germ-line mutation of the ATM gene causes a rare pleiotropic recessive disorder termed Ataxia-Telangiectasia (A-T). The most consistent A-T phenotype is early-onset ataxia caused by progressive cerebellar neurodegeneration [2]. Other symptoms of A-T are ocular telangiectasias, immunodeficiency, radiosensitivity, and increased predisposition to the development of cancer, primarily hematological malignancy [3]. Also, obligate female ATM heterozygotes have a modest, but well documented, increase in the relative risk of breast cancer development [4]. The tumor suppressive activity of ATM stems from its ability to activate appropriate responses to genotoxic and other forms of cellular insult [5].

One of the ATM-dependent responses triggered following DNA damage is activation of the NF-κB transcriptional complex. In this signaling axis, ATM phosphorylates NEMO (IKKγ) and relocalizes with NEMO as a complex from the nucleus to the cytoplasm [6]. Within the cytoplasm, the ATM/NEMO complex promotes linear ubiquitination of the ERC1 (ELKS) protein creating a docking platform for TAK1 kinase and IKK α/β protein kinase subunits [7]. TAK1 subsequently phosphorylates/activates the IKK complex which, in turn, phosphorylates the NF-κB inhibitor IκBα. IκBα activates proteosome-mediated degradation of IκBα and entry of NF-κB into the nucleus where it promotes transcription of a wide-spectrum of target genes [8,9].

Recently, we and others reported that constitutive NF-κB activity in cultured cell lines requires ATM [10–13] and that basal NF-κB transcriptional activity was significantly diminished by culturing cells on the anti-oxidant N-acetyl cysteine [13]. This finding is consistent with previous studies from the Paull lab indicating that ATM is catalytically activated in response to reactive oxygen species (ROS) [14]. This finding strengthens the emerging view that ATM acts not only in response to DNA damage, but also as a sensor of cellular redox status [15,16]. In support of a cytoprotective function for ATM during response to oxidative stress, it has been proposed that neuronal degeneration in A-T patients likely stems from unchecked oxidative damage [17–20]. Similarly, dysregulated response to oxidative stress is responsible for loss of hematopoietic stem cells in Atm−/− mice [21].

As both ATM and NF-κB are required for pro-survival signaling in response to oxidative stress [22–24], we sought to uncover genes important in oxidative stress response that are controlled by ATM/NF-κB signaling. Although NF-κB controls transcription of many genes [9], one attractive candidate is superoxide dismutase 2 (SOD2, MnSOD) [25–27]. Nuclear encoded SOD2 resides within the mitochondrial matrix where it catalyzes the dismutation of the toxic superoxide anion, a bi-product of respiration, into hydrogen peroxide [28]. Several groups have documented that SOD2-deficiency leads to high levels of oxidative damage [29–32] underscoring the critical role that SOD2 plays in redox homeostasis.

As outlined herein, we studied the potential link between ATM and SOD2 in cultured mammary epithelial cell lines and murine mammary glands for a number of salient reasons. First, based on both epidemiologic and molecular evidence, reduced ATM function has been linked to breast cancer [33,4,34]; however, causal links have not been well established. Second, oxidative damage is commonly thought to be an initiating lesion in breast tumorigenesis [35,36] suggesting an important role for antioxidant defense mechanisms in mammary tumor suppression. Lastly, studies in dairy cows [37] and rodents [38] have demonstrated increased oxidative stress within the lactating mammary gland, and administration of dietary antioxidants to dairy cows has been shown to limit the development of common post-parturition pathologies such as mastitis (for review see [39,40]). These findings suggest an important role for antioxidant defenses in maintaining mammary gland function and homeostasis.

MATERIALS AND METHODS

Cell Culture and RNA Interference

MCF10A cells were obtained from ATCC (Manassas, VA) and hTERT-immortalized human mammary epithelial cells (HMEC) were the generous gift of Dr. Jianrong Lu (Univ. of Florida) and validated by short tandem repeat (STR) analysis. For RNAi-mediated knockdown of ATM and RelA, shRNA sequences cloned into lentiviral vectors were obtained from Open Biosystems (Huntsville, AL). (Clone TRCN0000014684 was used to repress RelA expression, and clone V2LHS_192880 was used to knockdown ATM.) Lentivirus encoding shRNAs or empty vector were packaged in LentiX cells (Clontech, Mt. View CA) by co-transfection with the packaging plasmids psPAX2 and pMD2.G (Addgene, Cambridge, MA) as previously outlined [10]. Selection with 2 μg/mL puromycin was conducted for approximately 2 weeks prior to analysis of the resultant polyclonal cell populations.

Immunoblot Analysis

SDS-PAGE and immunoblotting was performed using established protocols [41]. Nitrocellulose membranes were probed with anti-ATM (2C1, GeneTex, Irvine, CA), anti-RelA (sc-109, Santa Cruz Biotechnology, Santa Cruz, CA), anti-SOD2 (sc-30080, Santa Cruz Biotechnology) or monoclonal anti-tubulin (E7) obtained from the Developmental Studies Hybridoma Bank (Univ. of Iowa).

PCR Analysis

Genomic DNA was harvested from freshly dissected mammary glands by incubation in DNA extraction buffer (100 mM Tris-HCl pH 8.0, 5 mM EDTA, 0.2% SDS, 200 mM NaCl) containing 100 μg/mL proteinase K (Sigma Aldrich, St. Louis, MO) and overnight incubation at 50°C. DNA was subsequently isolated by a phenol/chloroform extraction followed by ethanol precipitation. Primers P1 and P3, used to examine WAP-Cre mediated excision of Atm exon 58, as well a primers used to amplify a portion of Atm exon 4 are listed in Supplemental Table I.

Total RNA was isolated from mammary tissue and cultured cells using TRI Reagent (Ambion, Austin, TX) per manufacturer’s instructions. RNA was then used in first strand cDNA synthesis reactions using the Go Script Reverse Transcription System (Promega, Madison, WI). Q-PCR was carried out using SYBR Green master mix (Applied Biosystems, Norwalk, CT) in an Applied Biosystems StepOnePlus theromocycler, fold changes in transcript abundance were calculated by the ΔΔCT method [42] using Keratin 18 (Krt18) or Gapdh as internal standards. Primers used for Q-PCR measurements are listed in Supplemental Table I.

Chromatin Immunoprecipitation (ChIP)

ChIP with anti-RelA antibody was performed as previously outlined [10]. Briefly, indicated cells were harvested by trypsinization, washed extensively, proteins were cross-linked with 1% formaldehyde (RT, 10 min), and then cross-linking stopped by the addition of 125 mM glycine for 5 min. After washing, cells were pelleted and resuspended in ice-cold TEG buffer (10 mM Tris, 1 mM EDTA, 0.5 mM EGTA, pH 8.0). Cells were then sonicated on ice for 8 X 30s with a Branson Sonifier (Danbury, CT), cell debris was removed by centrifugation, and the supernatant (soluble chromatin) collected. After preclearing with protein A agarose beads, chromatin was immunoprecipitated with anti-RelA or non-specific rabbit IgG. DNA was isolated from pelleted immunocomplexes by phenol/chloroform extraction and Q-PCR was carried out as outlined above with primers specific flanking the consensus NF-κB binding site (5’-GGGAATACCC-3’) within intron 2 of the human SOD2 gene (see Supplemental Table I).

Alamar Blue Cell Viability

8.0×10³ cells/well were plated into 96-well plates and allowed to adhere overnight. The next day, media was removed, cells were rinsed with 1X PBS, and complete media containing 900 μM paraquat (or vehicle only) was added to each well. After 6 hr, media was removed, rinsed with 1X PBS, and fresh media was added containing 900 μM paraquat and, if indicated, either EUK-8 or TEMPOL at indicated concentrations. After a 48 hr incubation at 37^oC, media was removed, rinsed with 1X PBS, and fresh media added containing 1/10 volume of AlamarBlue reagent (AbDSerotec, Raleigh, NC) was added and cells were incubated at 37°C for an additional 4 hr. After this, 100 μl of media from each well was pipetted into a 96-well plate and fluorescence measured. Paraquat was purchased from Spectrum Chemicals (New Brunswick, NJ) and a 100 mM stock solution prepared in DMSO was stored at −80°C prior to use.

Genetically Engineered Mouse Lines

Atm−/− mice have been previously described [43] as has the mouse line harboring an Sod2 gene with LoxP sites flanking exon 3 [31]. A mouse line with LoxP sites flanking exon 58 of the Atm gene was developed using conventional gene targeting as previously described [44,45]. Transgenic mice harboring a Cre recombinase transgene under control of the whey-acidic protein (WAP) promoter [B6.Cg-Tg(WAP-Cre)11738Mam; strain#01XA8] were obtained from the NCI Mouse Repository (mouse.ncifcrf.gov). Mice were maintained in a mixed genetic background and PCR primers used for genotyping are indicated in Supplemental Table I. All mouse husbandry and experimentation was conducted in accordance with protocols approved by the University of Florida IACUC.

Histological Analysis

At indicated developmental time points, Atm−/−, Atm+/+, Atm^flox/flox and Atm^Δ/Δ females were euthanized with CO₂ gas followed by cervical dislocation. Abdominal (#4) mammary glands were surgically removed and fixed in freshly prepared 4% paraformaldehyde overnight. Tissues were then paraffin embedded, sectioned, deparaffinized in xylene, rehydrated in a graded ethanol series, and stained with hematoxylin and eosin (H&E).

For immunohistochemical (IHC) analysis, sections were deparaffinized and treated with proteinase K (20 μg/mL) for 2 min at RT prior to extensive rinsing and application of anti-ATM antibody (Millipore, Billerica, MA). Tissue staining was performed using VECTASTAIN Elite ABC system (Vector Labs, Burlingame, CA) per manufacturer’s instructions and developed with 3,3’-Diaminobenzidine. Following immunostaining cells were counterstained with hematoxylin.

For whole mount analysis, thoracic (#3) mammary glands were dissected, spread thin onto nitrocellulose membrane, placed in a labeled cassette, and fixed in 10% buffered formalin overnight. The next day, cassettes were incubated in 3 changes of acetone for 1 hr, and then immersed in 100% ethanol for 30 min, 95% ethanol for 30 min, and subsequently stained by immersion in Mayer’s Hematoxylin (Lillie’s Modification) (ScyTek Laboratories, Logan, UT) overnight. The cassettes were then rinsed with tap water until clear and destained in acidic 50% ethanol. Subsequently, cassettes were carried through a graded ethanol series, immersed in xylene overnight, and stored in methyl salicylate prior to microscopy.

Terminal Deoxynucleotidyl Transferase dUTP Nick End Labeling (TUNEL) Assay

TUNEL assays were performed using a Roche in situ Cell Death Detection kit according to the manufacturers’ guidelines (Roche, Indianapolis, IN). Briefly, sectioned tissues were deparaffinized and antigen retrieval performed by citrate-steam. Slides were subsequently rinsed twice with 1x PBS and incubated with terminal transferase and fluorescein-labeled dUTP for 60 min at 37°C. After this, slides were rinsed twice with 1x PBS and counterstained with DAPI (Vector Labs, Burlingame, CA). Slides were imaged on a Leica DM6000B microscope (Leica Microsystems, Buffalo Grove, IL) equipped with Openlab software (Agilent Technologies, Santa Clara, CA). DAPI and TUNEL positive cells were counted using ImageJ software (ver 1.44). For each indicated genotype and developmental time point, micrographs of 6–8 randomly chosen fields were counted for both TUNEL and DAPI positive cells. A total of 4,000–5,000 DAPI positive cells luminal cells were counted for each analysis and the percentage of TUNEL positive cells within this population calculated.

RESULTS

Atm knockdown in cultured mammary epithelial cells results in reduced expression of the NF-κB target gene Sod2 and increased sensitivity to redox imbalance

To initially address if SOD2 expression was linked to ATM we used RNAi-mediated gene knockdown to diminish expression of ATM in immortalized human mammary epithelial cells (HMEC) and the mammary epithelium-derived line MCF10A. To reduce NF-κB activity we knocked down the NF-κB subunit RelA (p65) as the RelA/p50 heterodimer has been previously shown to drive transcription of the SOD2 gene in a neuronal cell line [46]. Cells were transduced with lentivirus encoding shRNA sequences specific for either ATM, RelA, or virus made from empty pLKO.1 vector (shRNA-Control). Immunoblot analysis indicated substantial diminishment in ATM protein abundance associated with specific shRNA expression in each line (Fig 1A). Q-PCR measurement of ATM transcript abundance also measured a coordinate multi-fold decrease in ATM mRNA levels (data not shown). Additionally, we observed that knockdown of ATM resulted in measurable diminishment of SOD2 protein expression in each line as judged by immunoblotting. Likewise, we observed that knockdown of RelA also resulted in a measurable decrease in SOD2 abundance in both lines (Fig 1B). Q-PCR analysis indicated a statistically significant reduction in SOD2 mRNA levels in both MCF10A and HMEC lines following knockdown of ATM (Fig 1C) indicating that loss of SOD2 expression was associated with reduced transcript abundance.

Figure 1. ATM and RelA are required for SOD2 expression in cultured mammary epithelial cells.

A) Cultured HMEC or MCF10A cells were transduced with either control (empty) lentivirus (shRNA-Control) or virus encoding ATM-specific shRNA (shRNA-ATM). Transduced cells were selected using puromycin for 2–3 weeks and resultant polyclonal populations were analyzed by immunoblotting for ATM (top), SOD2 (middle), or Tubulin (bottom) as loading control. B) HMEC or MCF10A cells were transduced with either shRNA control virus or virus encoding RelA-specific shRNA. Following drug selection, lysates were immunoblotted with anti-RelA (top), anti-SOD2 (middle), or anti-tubulin (bottom). C) Q-PCR analysis of relative SOD2 mRNA abundance in HMEC and MCF10A cells transduced with either shRNA-control (unfilled bars) or shRNA-ATM virus (filled bars). (error bars = 1.0 s.d.) D) HMEC or MCF10A cells were treated with 10 mM NAC (filled bars), 500 μM TEMPOL (shaded bars), or untreated (unfilled bars) for 18 hr. After this cells were harvested, total RNA isolated, and Q-PCR analysis conducted to measure relative SOD2 transcript abundance. E) Chromatin was harvested from shRNA control or ATM knockdown HMEC or MCF10A cells, sheared, and either immunoprecipitated with anti-RelA or non-specific control rabbit IgG. Precipitated chromatin was subsequently subjected to Q-PCR analysis using primers flanking the consensus NF-κB site within intron 2 of the SOD2 gene. Graphed is quantitative amplification relative to input chromatin. All Q-PCR assays were conducted, at a minimum, in triplicate. Error bars = 1.0 SD; ** = p<0.01.

Our group previously reported that the co-culture of breast cancer and non-tumorigenic mammary epithelial cells with antioxidants results in diminished ATM and NF-κB activity [13]. To test if antioxidants similarly promoted reduced SOD2 levels, we cultured HMEC and MCF10A with the antioxidants N-Acetyl Cysteine (NAC) or 4-Hydroxy-TEMPO (TEMPOL) and measured SOD2 mRNA levels by Q-PCR (Fig 1D). We observed that culture of HMEC or MCF10A cells with 10 mM NAC or 500 μM TEMPOL for 18 hr resulted in a significant reduction in SOD2 transcript, consistent with the requirement for ATM/NF-κB signaling to support steady state levels of SOD2 in these cell lines.

Several groups documented that the human and rodent SOD2 gene contains an NF-κB-responsive enhancer within intron 2 [25–27]. To determine if knockdown of ATM reduced RelA association within this locus, we performed Chromatin Immunoprecipitation (ChIP) analysis on chromatin harvested from control and ATM-knockdown cells. For this assay we designed PCR primers that flank the NF-κB consensus site within intron 2 of the human SOD2 gene (5’-GGGAATACCC-3’ [25]) (see Supplemental Table I). Q-PCR measured negligible amplification of the 100 bp amplicon in chromatin samples from either HMEC or MCF10A (Fig 1E) when immunoprecipitated using non-specific rabbit IgG. In control cell lines, clear amplification of SOD2 intron 2 was measured in chromatin precipitated using anti-RelA. A statistically significant reduction in relative SOD2 intron 2 amplification was measured in each line following shRNA-induced knockdown of ATM. These results indicate that knockdown of ATM in mammary epithelial cell lines results in a loss of RelA binding within intron 2 of the human SOD2 gene.

SOD2 is responsible for the dismutation of superoxide anion (O₂•⁻) to hydrogen peroxide (H₂O₂). To determine if ATM knockdown resulted in heightened sensitivity to the superoxide radical, ATM knockdown and shRNA-Control HMEC and MCF10A cells were treated with 900 μM paraquat, a potent redox cycler that stimulates superoxide production [47]. This experiment indicated that ATM knockdown in both HMEC and MCF10A cells resulted in a statistically significant decrease in viability 24 hr after paraquat exposure when compared to shRNA-Control cells treated with 900 μM paraquat (Fig 2A).

Figure 2. ATM knockdown results in pro-oxidant sensitivity in mammary epithelial cells.

A) HMEC or MCF10A cells transduced with either shRNA control virus or shRNA-ATM virus were treated with 900 μM paraquat and along with untreated cells assayed for viability using an AlamarBlue assay 24 hr after drug treatment. B) shRNA-Control or shRNA-ATM HMEC cells were treated with 900 μM paraquat only, 900 μM paraquat and 100 μM TEMPOL, 900 μM paraquat and 500 μM TEMPOL, or 900 μM paraquat and 500 μM EUK-8. Cells were incubated for 24 hr and viability determined using Alamar Blue. Error bars = 1.0 SD; ns= not significant, ** = p<0.01 (Student’s t-test).

We next sought to determine if synthetic compounds that mimic the activity of native SOD by converting superoxide anion into hydrogen peroxide could rescue the sensitivity of HMEC with ATM knockdown to paraquat [48]. For these experiments we chose two such SOD mimetics: TEMPOL and EUK-8, a salen manganese complex [49,50]. When HMEC shRNA-Control cells were simultaneously treated with 100 μM or 500 μM TEMPOL and 900 μM paraquat we observed a significant and dose-dependent increase in cell viability compared to cells treated with 900 μM paraquat alone (Fig 2B). When ATM knockdown HMECs were simultaneously treated with 100 μM TEMPOL and 900 μM paraquat we observed no increase in viability compared to HMEC shRNA-ATM cells treated with 900 μM paraquat alone. However, 500 μM TEMPOL was able to significantly suppress paraquat toxicity in this cell line. In both shRNA-Control cells and ATM-knockdown HMEC, 500 μM EUK-8 equally suppressed paraquat toxicity compared to cells treated with paraquat alone. From these studies we conclude that SOD2 mimetics can suppress pro-oxidant toxicity in mammary epithelial cells with reduced ATM.

Atm-deficient mammary glands display reduced Sod2 expression

We next sought to determine if ATM was required for SOD2 expression in the mammary gland. Several groups have developed Atm−/− mice and various developmental defects have been reported [51,52]; however, mammary gland development in this model has not been documented. Mammary gland structure was examined in glands dissected from post-pubertal (12-week old) Atm−/− and wild type (Atm+/+) littermate females. Whole mount analysis revealed that wild type littermates displayed normal ductal morphogenesis as demonstrated by secondary and tertiary ductal side-branching, and distinct alveolar buds (Fig 3A). In contrast, Atm−/− mammary glands exhibited limited mammary gland development. Hematoxylin and eosin (H&E) staining supported the striking differences between these mice (Fig 3B). Wild type littermates displayed abundant ductal structures and alveolar buds, while Atm−/− mice showed limited presence of these structures. To examine whether there were structural abnormalities in the ducts present within the Atm-deficient gland, H&E stained sections were analyzed at higher magnification (Fig 3C). No obvious differences were noted between genotypes as both had a layer of luminal epithelium lining the ducts and a thick surrounding layer of stroma within abundant adipose tissue.

Figure 3. Atm−/− females display blunted mammary gland development.

A) Whole mount analysis of mammary glands dissected from 12-week old virgin wild type (Atm+/+) and Atm-deficient (Atm−/−) female littermates. Evident is extensive ductal tree side branching in the wild type gland and notably reduced ductal structure in the gland from the Atm−/− female. Alveolar buds are also absent in the Atm−/− gland but plentiful in the wild type gland. Also denoted is the centrally located mammary lymph node (LN), scale bar = 500 μm. B) Mammary glands from wild type (left) and Atm−/− (right) females were sectioned and H&E stained. Low power micrographs depict the presence of ducts (D) and alveolar buds (A) in the wild type mammary gland and a clear reduction of these structures in the Atm−/− gland. Scale bars = 100 μm. C) Higher magnification of H&E stained wild type and Atm−/− mammary glands. Evident in the glands from both genotypes is the ductal epithelium, surrounding stroma, and abundant adipose tissue. Scale bars = 100 μm.

The severely blunted mammary gland development observed in 12-week virgin Atm−/− mice may be, at least in part, linked to the lack of estrous cycling secondary to ovarian dysregulation in these mice [51]. Nevertheless, the lack of development in Atm−/− mice restricts our ability to study Atm function within this gland. Thus, we developed a mouse line with conditional deletion of Atm by using conventional gene targeting to place LoxP sites flanking Atm exon 58 (Fig 4A). These mice were subsequently mated with a transgenic mouse line expressing Cre-recombinase under the control of the Whey Acidic Protein (WAP) promoter [53] to target gene deletion within the mammary epithelium. Mice with the genotype (Atm^flox/flox; WAP-Cre) are designated Atm^Δ/Δ, show normal fertility, and were maintained in a mixed genetic background [C57Bl/6 X 129SvEv].

Figure 4. Characterization of Atm^Δ/Δ mice.

A) Architecture of the floxed exon 58 locus within the Atm gene and location of PCR primers used. B) Genomic DNA was harvested from mammary glands dissected from virgin and lactation day 1 (L1) Atm^flox/flox and Atm^Δ/Δ mice. PCR was conducted with primers P1 and P3 and amplification of the recombined Atm allele (Atm Δexon 58) is indicated (top). Amplification of Atm exon 4 was conducted as a DNA loading control (bottom). C) Mammary glands were dissected from an Atm^flox/flox (#148) and an Atm^Δ/Δ dam (#166) at L1, fixed, sectioned, and immunostained with anti-Atm or non-specific IgG. Positive immunostaining is detected within the alveolar epithelium in Atm^flox/flox mammary gland but notably reduced in the Atm^Δ/Δ gland. Scale bars = 100 μm. D) Q-PCR analysis of relative Atm and Sod2 mRNA abundance in L1, L5, and L10 mammary glands dissected from sacrificed Atm^flox/flox (open bars) and Atm^Δ/Δ (filled bars) dams. All Q-PCR reactions were run, at minimum, in triplicate using Krt18 as the internal standard for a mammary gland dissected from each indicated dam. Graphed is the average relative mRNA abundance (error bars = 1.0 s.d.). Relative transcript abundance measured in Atm^Δ/Δ mammary glands was compared to mean values measured in Atm^flox/flox glands using a Student’s t-test (* = p<0.05; ** = p<0.01; *** = p<0.001). E) Total protein extracts were formed from L10 Atm^flox/flox and Atm^Δ/Δ mammary glands and immunoblotted with anti-Sod2 (top) or Tubulin (bottom).

The WAP promoter is activated at pregnancy day 13 (P13), persists through lactation, and ceases as the mammary gland undergoes involution after weaning [53]. To determine if Cre mediates excision of Atm exon 58 in the mammary glands of uniparous Atm^Δ/Δ mice, control (Atm^flox/flox) and Atm^Δ/Δ mammary glands were dissected from 10-week old virgin females and dams at lactation day 1 (L1) and genomic DNA subsequently isolated. PCR was conducted using primers that flank Atm exon 58 (P1 and P3, see Fig 4A). Excision of Atm exon 58 was observed in uniparous Atm^Δ/Δ L1 females, while none was detected in uniparous Atm^flox/flox mice or virgin Atm^Δ/Δ mammary glands (Fig 4B). To examine expression of the Atm protein in Atm^Δ/Δ mice, mammary glands of L1 Atm^flox/flox and Atm^Δ/Δ females were dissected, sectioned, and immunostained with anti-ATM. As shown in Figure 4C, positive immunostaining was recorded in Atm^flox/flox mammary glands but staining was reduced within the L1 Atm^Δ/Δ mammary gland.

We next examined expression of Atm and Sod2 within the mammary gland of Atm^Δ/Δ dams at various points during lactation by Q-PCR. Here we observed a steady progression of Atm mRNA loss in mammary glands dissected from L1, L5 and L10 dams when compared to Atm^flox/flox dams (Fig 4D). Paralleling results obtained in cultured cells, we measured statistically significant reductions in Sod2 mRNA in Atm^Δ/Δ mammary glands at each lactational time point. To independently confirm reduced Sod2 expression, extracts were prepared from L10 mammary glands and immunoblotted with anti-Sod2 (Fig 4E). These results demonstrate a dramatic decrease in Sod2 abundance compared to glands dissected from Atm^flox/flox dams. In sum, our findings indicate that Atm is required for Sod2 expression both within cultured mammary epithelium and intact mammary glands.

Atm^Δ/Δ dams display defects in lactation

During the initial development of the Atm^Δ/Δ line we noticed that pups born of Atm^Δ/Δ dams commonly grew slower than the pups of Atm^flox/flox dams, and that this was not linked to pup genotype. To examine this in greater detail, 5 matings each of Atm^Δ/Δ and Atm^flox/flox females were set up, litters (normally 6–8 pups) were culled to 6 pups per dam at the date of birth without consideration to pup sex. Aggregate litter weight was recorded daily. Litters born of either genotype were of similar weight at birth and remained so through early time points in lactation (Fig. 5A). However, this experiment revealed that three (Atm^Δ/Δ dams 66, 59 and 61) of the five litters born of Atm^Δ/Δ dams displayed reduced average pup weight compared to Atm^flox/flox litters at a mid-lactation time point (lactation day 10 (L10)). The average weight gain of these three affected litters from birth to weaning when compared to weight gain measured in the litters born of control dams was statistically reduced (0.30 g/day vs. 0.42 g/day, p=0.013, respectively). By adolescence (5–6 weeks of age) the weights of offspring born of Atm^Δ/Δ and control dams were indistinguishable.

Figure 5. Litters of Atm^Δ/Δ dams display reduced growth rates.

A) Average pup weight was determined from birth to 21 days of age in litters born of 5 Atm^flox/flox and 5 Atm^Δ/Δ dams. Individual Atm^Δ/Δ litters are shown, the average pup weight in Atm^flox/flox litters is graphed (error bars= 1.0 s.d.). B) Average pup weights at L10 from litters of 21 indicated Atm^Δ/Δ dams is graphed (filled bars). Also graphed is the average L10 pup weight measured in 10 litters of Atm^flox/flox dams (unfilled bar, error bar= 1.0 s.d.). C) Q-PCR analysis of α-lactalbumin (Lalba) or β-casein (Csn2) transcript abundance was performed on RNA extracted from mammary glands dissected from indicated Atm^flox/flox (open bars) and Atm^Δ/Δ (filled bars) dams at L5 or L10. All Q-PCR reactions were run, at minimum, in triplicate using Krt18 as the internal standard. Graphed is the average relative mRNA abundance (error bars = 1.0 s.d.). Relative values obtained from Atm^Δ/Δ mammary glands were compared to the average of values measured in Atm^flox/flox glands (** = p<0.01; *** = p<0.001, Student’s t-test)

Following this initial experiment, we tracked the weights of litters born of 16 additional (21 total) Atm^Δ/Δ and 5 additional control (Atm^flox/flox) dams (10 total) from birth to weaning. The largest difference in average pup weight between the two genotypes was measured from L8 to L10. Data gathered indicates that 13/21 (61.9%) of litters born of Atm^Δ/Δ dams display an average L10 pup weight that is lower than 1.0 SD from the mean weight of L10 pups born of Atm^flox/flox dams (Fig 5B). The basis of the incomplete penetrance of this phenotype among Atm^Δ/Δ dams remains unknown; however, we measured significant heterogeneity in Atm mRNA abundance among mammary glands dissected from post-parturition Atm^Δ/Δ mice compared to Atm^flox/flox controls (data not shown) suggesting that the partially penetrant nature of this phenotype could be attributable, at least in part, to heterogeneity in Atm knockout between individual Atm^Δ/Δ dams.

As pup growth is a sensitive marker for lactational performance [54], we next examined the expression of two milk protein genes (α-lactoalbumin (Lalba) and β-casein (Csn2)) in Atm^Δ/Δ and control dams. Results indicated there was no quantitative difference in Lalba expression between L5 Atm^flox/flox and Atm^Δ/Δ dams (Fig 5C). Similarly, Csn2 also showed no significant difference in expression between the two groups at this time point (data not shown). However, a clear decrease in relative abundance of both Lalba and Csn2 transcript was measured in L10 Atm^Δ/Δ dams (145, 150, 201, and 202) compared to L10 Atm^flox/flox dams (Fig 5C). Of note, we have documented similar defects when Atm^Δ/Δ dams are maintained on an FVB background (data not shown), suggesting that the mixed C57Bl/6 X 129SvEv background of the mice analyzed in this study has no observable impact on phenotype. Taken together, these findings support the conclusion that Atm^Δ/Δ dams display lactation defects at mid-lactation time points and, since this phenotype is not apparent at earlier points in lactation, this defect appears to be progressive in nature.

Atm-deficient mammary glands display structural defects during lactation

To examine if lactational deficit was attributable to a defect in mammary gland structure, glands of Atm^flox/flox and Atm^Δ/Δ (Fig 6A) females were dissected at various developmental time points. Mammary glands were subsequently fixed, sectioned and H&E stained. Representative photomicrographs of each developmental stage are shown. In 10-week-old virgin females, a developmental point prior to activation of Cre-mediated gene deletion, mammary glands of both genotypes showed normal post-pubertal development as ductal development was evident and alveoli were sparsely distributed and quiescent. At mid (P10) and late pregnancy (P16.5) time points, both genotypes showed abundant adipose tissue and alveoli were small and quiescent, but more plentiful. At L1 both genotypes displayed less abundant adipose tissue and increased alveolar cell proliferation. By L5, adipose tissue is minimal and alveoli are dense, large in size, extremely dilated, discrete lobules are observable, and milk secretions are clearly observed in the lumens. As no observable difference between the structure of Atm^Δ/Δ and control mammary glands was noted at any of these developmental time points, we conclude that Atm-deficiency has no discernable effect on pregnancy and early lactation-associated development within the mouse mammary gland.

Figure 6. Atm^Δ/Δ mammary glands display structural defects in mid-lactation.

A) Mammary glands were dissected from Atm^flox/flox (top row) or Atm^Δ/Δ females (bottom row) at the indicated developmental time point, fixed, sectioned, and subjected to H&E staining. Scale bars = 100 μm. B) L10 mammary glands were dissected from Atm^flox/flox dams 140 and 137 and Atm^Δ/Δ dams 145, 150, 201 and 202, and analyzed by H&E staining. Scale bars = 100 μm. C) Atm^flox/flox and Atm^Δ/Δ mammary glands dissected at L5 and L10 were subjected to TUNEL staining (green) and counterstained with DAPI (blue). D) TUNEL positive cells in L10 mammary glands dissected from Atm^flox/flox (unfilled bars) and Atm^Δ/Δ (filled bars) dams were counted using ImageJ software and calculated as a percentage of total DAPI-positive luminal cells (4000–5000 DAPI positive cells per mammary gland were analyzed).

In contrast, clear histological differences were observed between Atm^flox/flox and Atm^Δ/Δ mammary glands at the L10 time point. At this stage, Atm^flox/flox mammary glands (dams 140 and 137) displayed moderate to large alveoli that were dense, abundant, and milk secretions were seen in majority of alveolar and ductal lumens (Fig 6B). Cells lining alveoli were generally uniform high cuboidal to columnar with occasional multiple layers that were actively proliferating. In contrast, Atm^Δ/Δ dams (145, 150, 201, 202) generally displayed small alveoli that were scattered throughout the mammary fat pad and no distinctive lobules were evident (Fig 6B). Most striking, no milk secretions were observed within the alveolar lumens. Further, the luminal epithelium also lacked characteristic blebbing, an indicator of active lipid secretion [54], and adipocyte infiltration was clearly evident.

Reduced milk production during weaning and involution is due to mammary epithelial cell apoptosis [55]. To examine apoptosis within the mammary gland of Atm^Δ/Δ dams, TUNEL staining was performed on histological sections of Atm^flox/flox and Atm^Δ/Δ mammary glands at L5 and L10. TUNEL-positive cells were subsequently calculated as a percentage of total cells (minimum of 4000 DAPI stained luminal epithelial cells) in 6–8 randomly chosen fields. TUNEL-positive cells were sparse in mammary glands of postpartum L5 and L10 Atm^flox/flox mice, as well as L5 Atm^Δ/Δ mammary glands (Fig 6C). In contrast, the measured levels of apoptotic cells were greater than 2-fold increased in Atm^Δ/Δ L10 dams 145, 150, 201, and 202 (Fig 6D) when compared to mammary glands from L10 Atm^flox/flox dams.

Prolactin (Prl) / Stat5a signaling is required to promote alveologenesis and lactogenesis within the developing mammary gland [56]. To determine if Atm-deficiency results in dysregulation of this signaling we examined Stat5a phosphorylation (p-Stat5a) and localization by immunohistochemical (IHC) analysis of L10 Atm^flox/flox and Atm^Δ/Δ (dam 145) mammary glands. IHC revealed p-Stat5a staining within the nuclei of both Atm^flox/flox and Atm^Δ/Δ luminal mammary epithelium (Supplemental Fig 1). These findings suggest no significant dysregulation in Stat5a phosphorylation or localization in the mammary gland from Atm^Δ/Δ dam 145. As this dam exhibits severely reduced milk protein gene expression (Fig 5C) and low mean pup weight at L10 (Fig 5B), reduced lactation in Atm^Δ/Δ dams apparently occurs independent of Prl/Stat5a signaling. Moreover, as Atm-deficiency results in increased apoptosis within the mammary gland, we propose that Atm is not required to support milk production during lactation per se but rather to maintain mammary epithelial cell viability.

Sod2 deficiency within the mammary gland phenocopies Atm deficiency

We proposed that the lactational, histological, and molecular defects observed in Atm^Δ/Δ dams were linked to loss of Sod2 expression within the mammary gland. To test this we mated a mouse line harboring LoxP sites flanking exon 3 of the Sod2 gene [31] with mice harboring the WAP-Cre transgene. Female Sod2^Δ/Δ mice (Sod2^flox/flox ; WAP-Cre) and controls (Sod2^flox/flox) were mated, allowed to give birth and nurse, sacrificed at L10 and mammary glands dissected. As expected, we measured a statistically significant decrease in relative Sod2 mRNA abundance in the mammary glands from Sod2^Δ/Δ dams compared to controls (Fig 7A).

Figure 7. Sod2-deficiency within the lactating mammary gland results in lactation defects.

A) Relative Sod2 mRNA abundance was measured in two Sod2^flox/flox (unfilled bar) and three indicated Sod2^Δ/Δ dams (filled bars) at L10. Graphed is the average relative mRNA abundance using Krt18 as the internal standard (error bars = 1.0 s.d.). B) Average pup weights were determined in litters born of two Sod2^flox/flox dams (S84, S79) and three litters born of Sod2^Δ/Δ dams (S72, S150, S115) at the L1 and L10 developmental time points. C) L10 mammary glands were dissected from Sod2^flox/flox dams (S84, S79) and Sod2^Δ/Δ dams (S72, S150, and S115) and subsequently fixed, sectioned, and H&E stained. Scale bars = 100 μm. D) Q-PCR analysis of α-lactalbumin (Lalba) or β-casein (Csn2) transcript abundance was performed on RNA extracted from mammary glands dissected from Sod2^flox/flox (unfilled bars) and Sod2^Δ/Δ (filled bars) dams at L10. E) Sod2^flox/flox and Sod2^Δ/Δ L10 mammary glands were subjected to TUNEL staining (green) and counterstained with DAPI (blue). F) Mammary glands from L10 Sod2^flox/flox and Sod2^Δ/Δ dams were sectioned and subjected to TUNEL staining. Cells were counted using ImageJ software and calculated as a percentage of total DAPI-positive luminal cells (3000–5000 DAPI positive cells per mammary gland were analyzed). All Q-PCR reactions were run, at minimum, in triplicate using Krt18 as the internal standard. Where indicated, the average relative mRNA abundance is graphed (error bars = 1.0 s.d.). Values obtained from Sod2^Δ/Δ mammary glands were compared to the mean value measured in Sod2^flox/flox glands using a Student’s t-test (*=p<0.05; **= p<0.01; *** = p<0.001)

We also recorded individual pup weights at L1 and L10 in litters born of both Sod2^Δ/Δ and Sod2^flox/flox dams. ANOVA analysis indicated no statistically significant variance in pup weights among litters born of four different Sod2^flox/flox dams (p=0.7). We found that the average weight of pups born of Sod2^Δ/Δ dams S72 and S115 showed no significant difference in average pup weight at L1 when compared to controls (Fig 7B). Pups born of Sod2^Δ/Δ dam S150 displayed a slightly lower average weight at L1 compared to the mean of control mice at this time point (p=0.03). Similar to Atm^Δ/Δ litters, at L10 we measured a highly statistically significant (p<0.01) reduction in mean pup weight in litters born of Sod2^Δ/Δ dams S72, S115 and S150. Of note, we observed that ~30% of Sod2^Δ/Δ dams had litters with average L10 pup weight within the range of control dams (data not shown) indicating that, like Atm^Δ/Δ dams, Sod2^Δ/Δ dams also display a partially penetrant lactation defect at this time point.

We next examined the structure of Sod2^Δ/Δ and control dam L10 mammary glands by H&E staining. The glands dissected from two L10 Sod2^flox/flox dams (S84 and S79) displayed large alveoli with milk secretions clearly evident within the majority of alveoli and ducts (Fig 7C). In contrast, L10 mammary glands from Sod2^Δ/Δ dams (S72, S115 and S150) showed an absence of milk secretions and increased presence of adipocytes within the mammary gland (Fig 7C). Analysis of milk gene transcript abundance by Q-PCR indicated significant reductions in both Lalba and Csn2 mRNA in each L10 Sod2^Δ/Δ dam when compared to L10 Sod2^flox/flox dams (Fig 7D).

Finally, TUNEL staining was conducted on both L10 Sod2^flox/flox and Sod2^Δ/Δ mammary glands (Fig 7E). Similar to mammary glands analyzed from L10 Atm^flox/flox dams, TUNEL-positive cells were scarce in L10 Sod2^flox/flox mammary glands. In contrast, TUNEL-positive luminal epithelial cells in Sod2^Δ/Δ dams were more commonly observed. When TUNEL-positive cells were counted in glands from both genotypes, we measured a multi-fold elevation of TUNEL-positive cells in L10 Sod2^flox/flox dams (Fig 7E), indicating that loss of Sod2 in the lactating mammary gland triggers epithelial cell death. In sum, data gathered in this initial analysis supports the conclusion that Sod2-deficiency within the lactating mammary epithelium appears to phenocopy Atm-deficiency.

DISCUSSION

Evidence indicates that ATM is required for basal NF-κB activity in cultured cells [10,11,13]. Further, numerous groups have demonstrated that the nuclear-encoded SOD2 gene is transcriptionally controlled by NF-κB in response to a wide range of stimuli [57–61,46]. Here we show that knockdown of either ATM or the NF-κB subunit RelA results in reduced basal expression of the SOD2 gene in non-tumorigenic mammary epithelial cells. As SOD2 is critical in the dismutation of superoxide anion (O₂•⁻) to hydrogen peroxide (H₂O₂), it is unsurprising that we observed increased sensitivity to paraquat, a potent producer of superoxide anion, in ATM knockdown cells. Further, using the SOD mimetics TEMPOL and EUK-8 we were able to partially reduce sensitivity to paraquat in ATM knockdown cells. In sum, these results support the notion that sensitivity of ATM knockdown cells to the pro-oxidant paraquat is due, at least in part, to diminished SOD2 activity.

ATM has been implicated in protecting neurons from cytotoxicity associated with oxidative stress [62–64] and dysregulation of this response is believed to underlie the progressive neurodegeneration observed in A-T patients [65]. As this study focused on the mammary gland we cannot comment if ATM is required for SOD2 expression in other cell and organ types; however, we have observed that both ATM and RelA are required for SOD2 expression in a limited panel of breast and non-breast derived tumor lines (see [13] and data not shown). Of note, ATM was found to be required for the NF-κB-dependent upregulation of SOD2 in the leukemia line U937 in response to the histone deacetylase inhibitor LBH-589 [66], underscoring the importance of ATM in the expression of this key antioxidant. It is tempting to speculate that ATM-dependent expression of SOD2 may account, at least in part, for other aspects of A-T pathology.

We noted a prominent lack of development within the mammary gland of post-pubertal 12-week Atm−/− female mice. This phenotype is likely linked to hormone dysregulation secondary to defects in ovary development previously described in Atm-deficient mice [51,67]. This notion is supported by the observations that knockout of estrogen receptor α (ERKOα) produces a similar disruption of postnatal mammary gland development, and deletion of the progesterone receptor B isoform (PR-B) results in defects in ductal side-branching and alveolar bud formation [68,69]. Development of secondary sex characteristics and menstrual periods are commonly observed in female A-T patients [70]; thus, whether mammary gland development in A-T patients parallels the defects in development we have observed in Atm−/− mice awaits further clarification.

Another intriguing question raised by this work that will require additional study is a potential role for ATM in mammary gland development. In support of this possibility, several lines of investigation support the view that NF-κB functions in normal mammary development. For example, NF-κB displays transcriptional activity during pregnancy and post-weaning mammary gland remodeling (i.e., involution) [71,72]. Moreover, knockout of the NF-κB inhibitor IκBα results in increased epithelial cell proliferation and ductal branching [73] and, conversely, knockin of a non-activatable form of the IκB kinase subunit IKKα or a non-inactivatable form of IκBα blunted lobuloavelolar development during pregnancy [74,75]. To understand a potential role for ATM in mammary gland development at time points that precede the expression of WAP-cre (i.e., late pregnancy), alternate ATM conditional deletion models will need to be developed. MMTV-based conditional models such as MMTV-cre or MMTV-CreER^T2 [76] are often used to study the mammary gland; however, MMTV-driven transgene expression commonly shows expression in non-mammary secretory and lymphoid tissues [77]. As ATM deletion in lymphoid cells results in high incidence of lymphoma [51], for this initial study we chose to use the WAP-cre model where gene recombination is largely restricted to the mammary gland [53]. Nevertheless, the development of inducible Atm gene knockout models would be of tremendous benefit to examine the role that ATM plays in various aspects of mammary gland development.

We measured reduced post-parturition pup growth in litters born of Atm^Δ/Δ dams compared to control dams and our findings support that Atm-deficiency in the mammary gland results in a lactation defect. Moreover, as Atm-deficient glands show marked reductions in Sod2, and Sod2^Δ/Δ dams also display lactation defects that mirror those observed in Atm^Δ/Δ dams, we conclude that the lactation phenotype observed in Atm^Δ/Δ dams stems, at least in part, from decreased expression of Sod2. The lactation defect in both Atm^Δ/Δ and Sod2^Δ/Δ dams is associated with heightened levels of apoptosis within the luminal epithelium indicating that both Atm and Sod2 are required to promote epithelial cell survival and organ homeostasis within the mammary gland.

An important role for Atm in response to redox imbalance is supported by work conducted using Atm−/− mice that determined the bone marrow failure observed in these mice was caused by elevated levels of reactive oxygen species (ROS) [21]. Moreover, supplementation of Atm−/− mice with antioxidants corrected this phenotype and others such as tumor latency, neurobehavioral effects, and the constitutively active stress response observed in ATM-deficient Purkinje cells and fibroblasts [78–82]. As others have measured elevated ROS within the mammary gland during both pregnancy and lactation [37,83], we propose that Atm is required for survival of the mammary epithelium by responding to oxidative stress by, in part, driving Sod2 expression. Further, we observed that Atm deficiency led to a progressive phenotype that was prominent during mid-lactation. It is currently unclear if the progressive nature of this phenotype is due to the time required to drive Atm and Sod2 levels to below a threshold required to maintain mammary gland homeostasis or, alternatively, if Atm and Sod2 are required to respond to conditions uniquely present within the mammary gland during mid-lactation.

ATM is a tumor suppressor gene and reduced ATM function in humans is associated with increased breast cancer risk [84,85]. To date, we have maintained ~40 multiparous and uniparous Atm^Δ/Δ dams until 1 to 2 years of age. Mice were monitored for the presence of palpable tumors; however, no mammary tumors developed in any of these Atm^Δ/Δ females. Moreover, H&E staining revealed a lack of epithelial cell hyperplasia within the mammary ducts or the residual alveolar buds of aged parous Atm^Δ/Δ females (data not shown). As we observed high rates of apoptosis within the mammary epithelium of mid-lactation Atm^Δ/Δ dams it is probable that many of the Atm-deficient cells that would potentially give rise to tumors otherwise die during lactation. In support of this possibility, preliminary studies indicate that conditional homozygous deletion of p53 in the mammary gland of Atm^Δ/Δ mice rescues the lactation phenotype and results in a high incidence of mammary tumor development (data not shown). We are currently characterizing these specimens to determine molecular features associated with lost Atm expression in mouse mammary tumors.

Oxidative damage is thought to be of especial importance in breast cancer etiology [36] and several lines of evidence suggest that dysregulation of SOD2 is associated with the disease process in breast cancer [86]. Specifically, SOD2 expression is commonly downregulated in breast tumor lines and primary neoplasms [87,88], and overexpression of SOD2 in cultured cancer cells can limit their tumorigenicity [89–91]. In mice, several lines of evidence support a causal link between reduced SOD2 expression and cancer development. Specifically, Sod2+/− mice show both increased levels of oxidative damage and a 2-fold increase in tumor incidence [32]. Further, mammary tissue from Sod2−/− mice transplanted into wild-type recipients displayed hyperplasia and other alterations consistent with early neoplastic changes [92]. As we have demonstrated a key role for Atm in responding to oxidative stress and regulating Sod2 expression within the mammary gland, it is tempting to speculate that this function for ATM is critical in its ability to suppress breast cancer development.