Glutathione-Dependent and -Independent Oxidative Stress-Control Mechanisms Distinguish Normal Human Mammary Epithelial Cell Subsets

Glutathione-dependent and -independent oxidative stress-control mechanisms distinguish normal human mammary epithelial cell subsets

Nagarajan Kannana, Long V. Nguyena, Maisam Makarema, Yifei Donga, Kingsley Shiha, Peter Eirewa, Afshin Raoufa,1, Joanne T. Emermanb, and Connie J. Eavesa,c,2

aTerry Fox Laboratory, British Columbia Cancer Agency, Vancouver, BC, Canada V5Z 1L3; and Departments of bCellular and Physiological Sciences and cMedical Genetics, University of British Columbia, Vancouver, BC, Canada V6T 1Z3

Edited by Tak W. Mak, The Campbell Family Institute for Breast Cancer Research at Princess Margaret Cancer Centre, Ontario Cancer Institute, University Health Network, Toronto, ON, Canada, and approved April 9, 2014 (received for review March 5, 2014) Mechanisms that control the levels and activities of reactive oxy- mechanisms operative in the human cells and their associated gen species (ROS) in normal human mammary cells are poorly un- differential sensitivity to treatments that elevate ROS levels. derstood. We show that purified normal human basal mammary epithelial cells maintain low levels of ROS primarily by a glutathi- Results one-dependent but inefficient antioxidant mechanism that uses Differing ROS Levels in Different Subsets of Primary Human Mammary + + − − mitochondrial glutathione peroxidase 2. In contrast, the matching Epithelial Cells. CD49f CD90(THY1) EpCAM /lowMUC1 BCs + − + + purified luminal progenitor cells contain higher levels of ROS, mul- and CD49f CD90(THY1) EpCAM MUC1 LPs were purified tiple glutathione-independent antioxidants and oxidative nucleo- by fluorescence-activated cell sorting (FACS) at ≥98% purity tide damage-controlling proteins and consume O2 at a higher rate. from normal human reduction mammoplasty samples depleted The luminal progenitor cells are more resistant to glutathione de- of hematopoietic cells, endothelial cells, and stromal cells (SCs) pletion than the basal cells, including those with in vivo and in (Fig. 1 A and B). These LP and BC subsets were highly enriched vitro proliferation and differentiation activity. The luminal progen- in cells with lineage-restricted and bilineage clonogenic activity itors also are more resistant to H2O2 or ionizing radiation. Impor- in vitro, respectively (∼20%; Fig. S1A). Some BCs can also re- tantly, even freshly isolated “steady-state” normal luminal generate complete bilayered gland structures containing clono- progenitors show elevated levels of unrepaired oxidative DNA genic progeny in xenotransplanted immunodeficient mice (8, 9). damage. Distinct ROS control mechanisms operating in different The inner luminal epithelial layer of the normal human mam- subsets of normal human mammary cells could have differentia- mary gland also contains many cells devoid of clonogenic activity. tion state-specific functions and long-term consequences. These cells, known as luminal cells (LCs), can be isolated as a phenotypically distinct population based on their lack of CD49f human epithelial stem and progenitor cells | mammary differentiation | and CD90(THY1) expression and high EpCAM and MUC1 ex- 3D clonogenic assay | superoxide dismutase | peroxiredoxin pression. LPs are thus assumed to represent an intermediate MEDICAL SCIENCES stage of differentiation between BCs and LCs. ellular synthesis of different reactive oxygen species (ROS) FACS analysis of dihydroethidium (DHE)-stained cells Cresults primarily from the incomplete reduction of molecular showed a higher free-radical superoxide anion (O2°) content in oxygen in mitochondria to generate free radical superoxide the purified LPs compared with the purified BCs (Fig. 1C). anions. ROS also are produced when cells are exposed to dif- Parallel measurements of intracellular H2O2, generally assumed ferent environmental sources of oxidative stressors, including to be the major intracellular nonradical ROS produced by ionizing radiation. Together, these regulate many normal cellular processes and also contribute to DNA damage, tumorigenesis, Significance and cell death (1–3). In mice, the inner layer of “luminal” epithelial cells of the Our study reveals lineage-specific mechanisms of ROS control normal adult mammary gland have been found to contain higher and associated sensitivity to oxidative DNA damage in the levels of ROS than the outer “basal” layer of epithelial cells (4). basal and luminal progenitor-enriched subsets of normal hu- The basis for these differences in ROS levels in the two cell types man mammary cells. We show that the primitive luminal cells has been attributed to differences in their content of mito- contain more mitochondria, show greater uptake of O , sustain chondria; however, their stress response mechanisms have not 2 and withstand higher levels of ROS, and have mechanisms that been defined. Even less is known about ROS levels and their allow them to accrue mutagenic levels of oxidative DNA control in the normal adult human mammary gland, which damage. These findings support a growing body of data sug- consists of a similar continuous bilayered epithelial network of gesting the involvement of primitive luminal cells in the gen- ducts and terminal alveolae. MCF10A cells are an immortal but eration of human breast cancers. nontumorigenic human mammary cell line that, when grown in

3D Matrigel cultures, generates multilayered spheres in which Author contributions: N.K. and C.J.E. designed research; N.K., L.V.N., Y.D., and K.S. per- a lumen forms owing to the acquisition of lethal levels of ROS by formed research; J.T.E. organized accrual of primary tissue; N.K., M.M., P.E., A.R., J.T.E., the inner cells (5). These studies have suggested that ROS reg- and C.J.E. analyzed data; and N.K. and C.J.E. wrote the paper. ulation plays an important role in the structural morphogenesis The authors declare no conflict of interest. and homeostasis of normal adult human mammary tissue. This article is a PNAS Direct Submission. Here we report the results of experiments designed to investigate Freely available online through the PNAS open access option. the levels of ROS and their control in highly purified populations of 1Present address: Department of Immunology, Faculty of Medicine, University of viable luminal progenitors (LPs) and basal cells (BCs) isolated from Manitoba, Winnipeg, MB, Canada R3E 0T5. normal human breast tissue (6–8). Our results confirm the different 2To whom correspondence should be addressed. E-mail: [email protected]. levels of ROS identified in the corresponding subsets of mouse This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. mammary cells and further characterize the distinct antioxidant 1073/pnas.1403813111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1403813111 PNAS | May 27, 2014 | vol. 111 | no. 21 | 7789–7794 Downloaded by guest on September 27, 2021 DHE (O °) DCFDA (H O ) A B C 2 D 2 2 (~1 in 5) LC LP LP

(~1 in 5) MFI MFI MUC1 CD49f EpCAM

SC BC maximum %

BC maximum of % CD49f SSC THY1 BC LP BC LP

β-F1-ATPase levels MitoDNA M rekcarToti m(ERMT ) O uptake E F G H ψ I 2 BC BC LP LC Sample 1 23123123 MFI MFI (mm Hg)

H3 2 % maximum % % maximum % O Fold-change LP 0 SPECTRAL UNITS 65535 BC LP BC LP BC LP 0 60 120 240 Time (seconds)

Fig. 1. Isolation of normal human mammary cells at different stages of differentiation and their changing levels of ROS. (A and B) FACS profiles and gating − − − + strategies used to isolate the three different viable (DAPI ) human mammary (CD31 CD45 ) subpopulations studied. BCs: CD49f EpCAM-/low (28) or CD49fhiTHY1+MUC1− cells (clonogenic cell frequency, 18 ± 3%; n = 10). LPs: CD49f+EpCAM+ (28) or CD49fhiMUC1+THY1− cells (clonogenic cell frequency, 22 ± 5%; n = 10). LCs: CD49f-/lowEpCAM+ cells. Mammary SCs: CD49f−EpCAM− cells. (C) Representative FACS histograms of DHE-stained BCs and LPs showing their

different O2° levels and comparing the fold increase in median fluorescent intensity (MFI) values of LPs vs. matching BCs. n = 7; P = 0.001. (D) Representative FACS histograms of 2′,7′ –dichlorofluorescein diacetate (DCFDA)-stained BCs and LPs and the fold increase in MFI values of LPs vs. matching BCs. n = 9; P < 0.0001. (E) Western blot analysis of mitochondria-specific β-F1-ATPase protein in each of the three mammary epithelial subpopulations studied (n = 3), with histone H3 as the internal loading control. (F) Comparison of the mitochondrial DNA content relative to genomic DNA as measured by qRT-PCR in extracts of paired isolates of BCs and LPs. Results for LPs are shown relative to the results for BCs in the same sample. n = 6; P = 0.01. (G) Representative FACS histograms of MitoTracker dye-stained BCs and LPs and a comparison of the fold increase in MFI of the LPs vs. matching BCs. n = 9; P < 0.0005. (H) Representative FACS histograms of mitochondrial membrane potential (mΔψ) measured by TMRE staining and a comparison of the fold increase in MFI of the LPs vs. matching BCs. 5 n = 3; P < 0.05. (I) Representative plots of O2 consumption of 10 BCs and matched LPs measured using a Seahorse respirometer in one of four experiments that yielded similar results. Unequal slopes; P < 0.0001.

spontaneous or catalyzed dismutation of mitochondrial O2°, revealed showed higher levels in LPs (Fig. 2D). Transcript levels for the higher levels in the LPs (Fig. 1D).ThehigherROSinLPsiscon- catalytic subunit of γ-glutamyl-cysteinyl ligase (GCLC), a rate- sistent with the higher number of mitochondria that they contain, as limiting enzyme in glutathione biosynthesis (14), also were indicated by their higher levels of mitochondrial-specific β-F1- higher in LPs (Fig. 2E). Elevated glutathione levels in LPs were ATPase seen on Western blot analysis (Fig. 1E), higher levels of confirmed by FACS-based measurement of monochlorobimane- mitochondrial-specific DNA on quantitative RT-PCR (qRT- stained cells (Fig. 2F). Taken together, these findings indicate PCR) analysis (Fig. 1F), and more intense fluorescence when that the LPs in the normal human mammary gland have a stained with MitoTracker dye (Fig. 1G). The LPs also showed stronger antioxidant defense profile than the BCs. greater sequestration of the tetramethylrhodamine ethyl ester In contrast, both qRT-PCR and Western blot analyses showed (TMRE) dye (Fig. 1H) and more rapid O2 consumption (Fig. 1I), that the mitochondrial enzyme glutathione peroxidase 2 (GPX2) indicative of increased mitochondrial activity in the LPs. is present almost exclusively in BCs (Fig. 2 A and G), as suggested by previous serial analyses of gene expression and Different Antioxidant Mechanisms Are Active in Different Subsets of microarray analyses of their transcript levels (Fig. S2A) (7). On Normal Human Mammary Cells. We next asked whether the dif- Western blot analysis, GPX2 was largely undetectable in either ferent subsets of normal human mammary cells would show luminal subset or in the SCs (Fig. S2B). Interestingly, BCs con- differences in the mechanisms used to produce and control sistently showed higher transcript levels of the antioxidant tran- ROS. We first compared the levels of SOD1, SOD2, and SOD3 scription factor NRF2 and its response element, γ-glutamyl- present in FACS-purified BCs, LPs, and LCs, because these cysteinyl ligase modulatory subunit (GCLM) (Fig. 2E). These three enzymes act spatially to maintain steady-state concentra- observations suggest that the BCs of the normal human mam- tions of O2° through conversion to H2O2 (10). Western blot mary gland may rely primarily on their lower, albeit tightly reg- analysis showed consistent elevations of all three SODs in both ulated, glutathione levels and GPX2 to deal with the lower levels LPs and LCs compared with BCs, in agreement with the find- of ROS that they produce compared with the LPs and LCs. In ing that the total SOD activity in LPs was higher than in the contrast, the latter make use of different, more effective mech- BCs (Fig. 2 A and B). We also investigated the levels of both anisms to survive the higher levels of ROS that they sustain. glutathione-dependent and glutathione-independent enzymes that reduce intracellular H2O2. The former included the gluta- Differing Effects of Perturbing Endogenous Control of Oxidative thione peroxidases (GPX) 1–8 and 1-cysteine peroxiredoxin- Stress in Different Subsets of Normal Human Mammary Cells. To test peroxidase (PRDX) 6, and the latter included catalase and 2- these predictions, we first compared the effect of lowering the cysteine peroxiredoxin-peroxidases, PRDX1–5 (11, 12). Western level of intracellular glutathione on the proliferative activity of blot analyses revealed higher levels of many of the antioxidant- BCs and LPs by adding buthionine sulfoximine (BSO), a potent peroxidases in the LPs and LCs compared with the BCs (Fig. and selective inhibitor of glutathione synthesis (15), or vehicle 2A), consistent with the increased total GPX activity detected in directly into the media in which the cells were assayed (Fig. 3A LPs compared with BCs (Fig. 2C). and Fig. S1B). BSO concentrations that had little or no effect on Biochemical determination of intracellular levels of NAD(P)H, LPs markedly, selectively, and specifically reduced the yield from the ultimate electron donor for antioxidant mechanisms (13) BCs of total viable cells in 4-d bulk cultures (Fig. 3B) and of

7790 | www.pnas.org/cgi/doi/10.1073/pnas.1403813111 Kannan et al. Downloaded by guest on September 27, 2021 A BC LP LC B Total SOD C Total GSH Px Sample 1 2 3123123 Activity Activity 3 400 SOD1 * * 300 H3 2 tases / cell / / cell /

u 200 U SOD2 U μ μ 1 H3 100 dismu ° SOD3 2 O 0 0 BC LP BC LP H3 s GPX1 s D NAD(P)H E GCLC GCLM NRF2 H3 2.5 0.1 * * GPX2 2.0 H3 1.5 0.0 Fig. 2. Differences in the antioxidant mechanisms GPX3 1.0

script levels active at different stages of normal human mam- D at 490 nm n n H3 ndent peroxidase * OD 0.5 * mary cell differentiation. (A) Representative Western Tra GPX7 -0.1 0.0 blots showing the relative levels of the enzymes BC LP BC LP BC LP BC LP indicated in all three matching epithelial sub- H3 populations from three samples out of a total of six PRDX6 samples analyzed that all yielded similar results. Monochlorobimane (Glutathione) F Histone H3 served as the loading control. (B) Fold Glutathione depen H3 G 3 * increase in total SOD activity in LPs vs. BCs. n = 3; P < 0.001. (C) Fold increase in total glutathione peroxi- CAT 2 dase activity in LPs vs. BCs. n = 3; P < 0.05. (D) NAD(P) H3 H levels in extracts of equal numbers of BCs and LPs, MFI PRDX1 1 determined from paired samples by spectropho- % maximum % tometry using the Biovision kit. n = 3; P = 0.01. (E) H3 roxidases e e Fold increase in qRT-PCR–determined transcript lev- 0 PRDX2 BC LP els of the catalytic subunit of GCL (GCLC; n = 10; P < H3 0.05), its modulatory subunit (GCLM; n = 10; P < G DN TP63 GPX2 0.0001), and transcription factor Nuclear Factor PRDX3 (Erythroid-Derived 2)-Like 2 (NRF2; n = 8; P < 0.001) H3 in LPs compared with matched BCs. (F) Representa- vels n PRDX4 ne independent p tive FACS histograms of monochlorobimane-stained H3 BCs and LPs showing their different intracellular reduced-glutathione levels and the fold increase in MFI PRDX5 = < MEDICAL SCIENCES

Glutathio of LPs vs. matching BCs. n 7; P 0.0005. (G) Com- H3 –

Transcript lev parison of qRT-PCR determined transcript levels of = < = < BC LP BC LP GPX2 (n 9; P 0.005) and TP63 (n 8; P 0.01) in 0 SPECTRAL UNIT 65535 BCs and LPs.

colonies produced in either 2D (at 20% O2) or 3D (at 5% O2) Differing Effects of Exogenously Derived Oxidative Stress on Different 10-d cultures (Fig. 3 C–F). However, the addition of either Subsets of Normal Human Mammary Cells. To test the possibility N-acetyl cysteine (NAC) or Trolox completely reversed the toxic that LPs might also be more resistant to perturbations that in- effect of BSO on the BCs and had no effect on the LPs (Fig. 3C). crease their intracellular ROS levels, we next compared the ef- We also tested the effect of pretreating purified BCs and LPs fect of increasing concentrations of H2O2 ontheviablecell with BSO for 2 d in vitro (under 2D conditions) on their sub- output of purified BCs and LPs in 4-d bulk cultures and on their sequent clonogenic activity assessed either immediately, in the clonogenic activities in both 2D and 3D assays (Fig. 4 A–D). absence of BSO, or, in the case of the BCs only, at 4 wk after Compared with the BCs, the LPs were more resistant to H2O2 being transplanted under the kidney capsules of immunodefi- exposure in all three assays. Despite this differential toxicity, the G cient mice (Fig. 3 ). The results showed that BCs with both in differentiated features of the 10-d progeny of the H O -treated vivo and in vitro proliferative activity are highly sensitive to even 2 2 H cells were unchanged compared with untreated controls (Fig. short-term exposure to BSO (Fig. 3 ). S3). In addition, the selective H O toxicity in BCs was partially As an alternative approach, we depleted GPX2 by transduc- 2 2 neutralized by pretreatment of BCs with NAC or Trolox (Fig. 4C). ing bulk normal human mammary cells with a lentivirus encoding a Ionizing radiation also mediates its lethal effects primarily by reporter fluorescent protein (VENUS)withorwithoutashRNA producing intracellular ROS, which then cause double-strand directed against GPX2 transcripts (shRNA GPX2)(Fig.3I)and then compared the numbers of colonies obtained from the trans- breaks in the DNA (16). Responses of mammary cells to ionizing + duced (VENUS ) cells (Fig. 3J). The results of these experiments radiation are of particular interest, because this treatment re- also showed a much greater shRNA GPX2-mediated reduction of mains a current mainstay in the management of breast cancer. colony yields from the BCs compared with the LPs. Comparing the effect of increasing X-ray doses on the sub- These findings indicate that the more primitive human mam- sequent clonogenic activity of BCs and LPs (in 2D assays) mary BCs are much more reliant on glutathione-dependent revealed that the LPs were again more resistant (Fig. 4E). In- mechanisms to survive even the relatively low levels of ROS terestingly, LPs also expressed higher levels of OGG1, MTH1 that they possess. In contrast, the LPs’ multiple glutathione- (also known as NUDT1), and MUTYH(MYH) (Fig. 4F and Fig. independent antioxidant mechanisms appear sufficient to ensure S4), enzymes that promote cell survival by reducing the toxic, their survival and clonogenic activity despite their much higher oxidized nucleotides generated by ROS from free intracellular ROS levels. pools or that have been incorporated into cellular DNA (17–19).

Kannan et al. PNAS | May 27, 2014 | vol. 111 | no. 21 | 7791 Downloaded by guest on September 27, 2021 A B MTT assay G FACS NAC/Trolox BC LP Amino acids In vitro LP treatment BSO GCLC BC LP control f GSH

%of Transplant 2D colonies GPX2 BC Fig. 3. Effects of perturbing endogenous ROS con- H2O2 H2O trol mechanisms on normal human mammary cells shRNA at different stages of differentiation. (A) Schematic BSO (μM) 2D colonies showing experimental strategies used to inhibit In vivoo In vitroto reduction of intracellular ROS. (B) Effect of BSO CH(Sigma-Aldrich) on purified BC- and LP-derived cells 100 in 4-d bulk cultures. (C) Purified BCs and LPs were plated in 2D-clonogenic assays with 1.5 mM (two 75 experiments) or 6 mM (three experiments) NAC μ 50 (Sigma-Aldrich) or 50 M Trolox (Calbiochem) and 50 LP μM BSO added 3 h later. Colonies were counted after BC – < < % of control 25 8 10 d (total of five experiments). *P 0.01; **P % of control % 0.001. (D)(Upper) Confocal section of a 3D colony 0 _ _ _ generated from unmanipulated purified BCs or LPs. BSO + + + Cells were stained with a high-affinity F-actin probe BC BC LP (phalloidin) conjugated to tetramethylrhodamine (Molecular Probes). (Lower) H&E-stained section of DE I 3D cultures. (E) Effect of BSO on colony formation by purified BCs and LPs assayed in the 3D system. Values HepG2 cells are percent of controls containing no BSO. n = 5; LP unequal slopes; P < 0.005. (F) Representative pho- – BC GPX2 tomicrographs of day 12 14 cultures from E.(G)Ex- perimental design for testing the effects of an in μ GAPDH vitro 48-h exposure to 100 M BSO on the immediate 50μM clonogenic activity of BCs and LPs (n = 2) and on cells % of control BC CControl t l shRNA hRNA that produce clonogenic cells in transplanted immunodeficient mice (n = 3). (H) Results of experiments performed as shown in G. Black bars represent LP BSO (μM) in vivo regenerated clonogenic cells. (I) Western blot showing the reduction of GPX2 protein expression in BC transduced HepG2 cells exposed to either the shRNA F BC LP J shRNA GPX2 or control lentivirus and analyzed 3 d later. GAPDH 75 served as the loading control. (J)EffectofGPX2sup- pression on the clonogenic activity of mammary 0 * cells transduced with a fluorescent reporter (VENUS)- M) 50 encoded lentivirus with or without shRNA targeting LP GPX2. Transduced cells were selected by FACS at 3 d after virus exposure and plated in 2D-clonogenic

BSO ( μ 25 % of control assays. Colony yields from shRNA GPX2 transduced 102 cells are shown as percent of values obtained in 0 assays of the same number of control transduced BC LP cells. n = 3; P < 0.01.

Given the elevated levels of intracellular ROS characteristic findings establish the normal human luminal epithelium as a of LPs, along with the long-established association between distinct site of oxidative stress resistance in the presence of chronic oxidative stress and oxidative damage to DNA, we then high intracellular ROS levels mediated by a unique enzyma- asked whether the LPs would exhibit evidence of oxidative DNA tic repertoire. These features distinguish LPs from BCs, which damage even without treatments designed to alter their ROS include very primitive bipotent cells with both luminal and levels. H2O2 in the presence of transition metal ions, such as myoepithelial differentiation abilities (24). The unique proper- iron or copper, is known to cause oxidative damage to DNA via ties of the LPs include up-regulated levels and activity of SODs Fenton’s reaction, which generates short-lived hydroxyl radicals and their superoxide substrates, with an accompanying switch in (OH°) continuously at low levels in living cells (20). One in- mechanisms used to control ROS levels and promote cell survival dicator of such oxidative damage is the formation of genomic in the presence of increased ROS levels. Specifically, this “switch” 8-oxo-deoxyguanosine (8-oxo-dG) nucleotides (21–23). FACS involves the replacement of a highly glutathione-dependent mech- measurement of genomic 8-oxo-dG levels in the DNA from anism primarily mediated by GPX2 in the BCs with a largely purified LPs and BCs revealed higher levels in the LPs (Fig. 4G). glutathione-independent mechanism in the LPs. Thus, LPs that survive oxidative insults presumably continue to It is important to note that the observations made here are for accumulate additional oxidative DNA damage (Fig. 4H). cells obtained from “resting” normal adult human mammary glands. Exmaining whether different results will be found for Discussion mammary cells subject to the strong hormonal influences of Delineating homeostatic ROS control mechanisms in human pregnancy and initiation or cessation of lactation will be of in- mammary BCs and luminal epithelial cells may provide new terest. The present findings also set the stage for future inves- insights relevant to both normal and malignant breast tissue. Our tigations of the transcriptional regulation of the switch in ROS

7792 | www.pnas.org/cgi/doi/10.1073/pnas.1403813111 Kannan et al. Downloaded by guest on September 27, 2021 A B C D E 10 2 BC LP 10 2 LP LP LP 10 1 10 1 LP 0 ontrol ontrol ontrol 10 control c c c c c c urvival 10 0 BC BC 10 -1 % su BC % of % of % of BC % of 10 -2 10 -1 050100 050100

H2O2 (μM) H2O2 (μM) H2O2 (μM) X-ray dose (Gy)

FHG 8-oxo-dG nucleotides BC LP

BC LP LC ROS BC Sample 1 23123123 GSH/GPX2- Multiple ROS ROS dependent (damage)

OGG1 um ROS control control MTH1 Survival with LP Death MUTYH MFI higher mutation

% maxim % ROS ROS H3 Differentiation FITC-conjugateFITC conjugate BC LP Oxidative Stress

Fig. 4. Effects of exogenous perturbations of ROS levels on normal human mammary cells at various stages of differentiation. (A) Effect of H2O2 on purified BC- and LP-derived cells in 4-d bulk cultures. (B) Effect of H2O2 on colony formation by BCs and LPs in 2D colony assays. Values are percent of controls containing no H2O2. n = 3; unequal slopes; P < 0.0001. (C) Purified BCs and LPs were plated in 2D clonogenic assays with 50 μM Trolox or 1.5 mM (two experiments) or 6 mM (five experiments) NAC, with 50 μMH2O2 added 3 h later. Colonies were counted after 8–10 d. n = 7; *P < 0.001; **P < 0.0001. (D) Same design as in B, but with assays performed in 3D cultures. n = 4; unequal slopes; P < 0.005. (E) Effect of 300 kVp (1 Gy/min) X-rays on subsequent colony formation by unseparated mammary cells in 2D assays. BC- and LP-derived colonies were distinguished morphologically. Values are percent of controls given a sham treatment. n = 3; unequal slopes; P < 0.0001. (F) Representative Western blots showing the relative levels of OGG1, MTH1, and MUTYH in three matching subpopulations from three samples out of a total of six samples analyzed that all yielded similar results. Histone H3 served as the loading control. (G) Representative FACS histogram of BC and LP cells stained for the 8-oxo-dG DNA adduct using a specific FITC-labeled probe (Left) and fold change in MFIs in LPs relative to matching BCs (Right). n = 4; P < 0.005. Arrows indicate LPs and BCs treated similarly but unstained for the 8-oxo-dG DNA adduct. (H) Model illustrating the mechanisms underlying the changes in production, control, and responses to altered ROS levels in BCs and LPs.

control that demarcates the generation of LPs from BCs. Vari- Tissue was processed and viable single cells were isolated as described pre- MEDICAL SCIENCES ous transcription factors show dramatic differences in expression viously (31). In brief, reduction tissues were first minced with a scalpel and in these two subsets, of which TP63, NOTCH3, and GATA3 are dissociated in Ham’s F-12 medium and DMEM (1:1 vol/vol; Stem Cell Tech- notable examples (7, 9, 25, 26). Of these, TP63 has been reported nologies) supplemented with 2% wt/vol BSA (Fraction V; Life Technologies), 300 U/mL collagenase (Sigma-Aldrich), and 100 U/mL hyaluronidase (Sigma- to increase the tolerance of breast cancer cells to oxidative stress Aldrich), from which an epithelial organoid-rich pellet was obtained after by transactivating GPX2 (27). It is also possible that the gluta- centrifugation at 80 × g for 4 min. Single-cell suspensions were then ob- thione-independent and oxidative DNA damage control mech- tained after further incubation in 1 mM EDTA supplemented with 2.5 mg/mL anisms that confer resistance to oxidative stress described here trypsin (Stem Cell Technologies), 5 mg/mL dispase (Stem Cell Technologies), as characteristic of normal LPs and LCs may operate in their and 100 μg/mL DNase1 (Sigma-Aldrich), followed by filtration of the sus- malignant derivatives, making them more resistant to oxidative pension through a 40-μm strainer. Mammary cells were then stained with treatments. antibodies listed in Table S1, and subsets were isolated as described pre- The glutathione-independence of the LPs and their more viously (7, 8), using a protocol in which hematopoietic and endothelial cells were removed using fluorochrome-conjugated antibodies to CD45 and CD31, differentiated LC progeny is associated with up-regulated ex- + respectively. Cells were also stained with DAPI to eliminate dead (DAPI )cells. pression of numerous enzymes that, in combination, appear able SCs were eliminated based on absent or reduced expression of EpCAM and/ to counteract at least some of the potentially lethally mutagenic or CD49f. Viable subsets of epithelial cells were then isolated at ≥98% effects of the ROS levels that these cells produce. These ob- purity using a FACSAria or Influx II cell sorter (BD Biosciences), as described servations add to a growing body of data revealing other mech- in Fig. 1 A and B. anisms that predispose normal human luminal cells to DNA damage (28). Thus, the up-regulated expression of multiple an- In Vitro and in Vivo Growth Assays. Bulk cultures for toxicity experiments were tioxidant enzymes at the point of mammary progenitor cell re- initiated by seeding 5,000 purified BCs and LPs on collagen-coated wells in striction to the luminal lineage could contribute to multiple 96-well plates in SF7 media supplemented with 5% FBS, followed by color- mechanisms that promote the death of LCs and allow lumen imetric assessment for viable cells after a 4-h incubation with 3-[4,5-dime- formation (5). Consistent with this hypothesis is our present thylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (MTT), as described previously (32). The 2D and 3D clonogenic assays were performed as de- finding of increased expression in LPs and LCs of unique PRDX- scribed in Fig. S1. peroxidases that facilitate a type of “floodgate” control of ROS ’ levels owing to these enzymes innate capability to be inactivated Intracellular Flow Cytometry. For live-cell analyses, cells were either costained selectively in response to various stimuli (29, 30). with antibodies to allow the isolation of subsets or were first purified and then incubated for 30 min at 37 °C with 100 nM MitoTracker, 10 μM DCFDA, Materials and Methods 5 μM dihydroethidium, 400 nM TMRE (all from Invitrogen), or 50 μMmono- Cells. Histologically confirmed normal anonymized tissue from women un- chlorobimane (Sigma-Aldrich). For the detection of 8-oxo-dG DNA adducts, dergoing cosmetic reduction mammoplasty was obtained following proce- FACS-purified cells were secondarily fixed, permeabilized, and then stained dures approved by the University of British Columbia’s Ethics Review Board. with a specific Calbiochem 8-oxo-dG–binding FITC probe (EMD Millipore).

Kannan et al. PNAS | May 27, 2014 | vol. 111 | no. 21 | 7793 Downloaded by guest on September 27, 2021 Stained cells were analyzed using a FACSAria, FACSInfluxII, or Fortessa ma- Lentiviral Transduction. A human GPX2 targeting shRNA (Open Biosystems) chine (BD Biosciences). was cloned into the pLKO.1-VENUS lentiviral vector, and virus-containing supernatants were generated and used to infect test cells as described + 5 – O2 Consumption. For measurement of O2 consumption, 10 purified cells were previously (7). Transduced (VENUS ) cells were isolated 2 3 d later and seeded in Seahorse XF96 cell culture plates (with and without a Matrigel assayed for colony formation. The shRNA activity of the lenticonstruct was coating) and cultured for 24 h in SF7 media supplemented with 5% FBS. confirmed by Western blot analyses of GPX2 levels in extracts of transduced human hepatocarcinoma HepG2 cells cultured for 2–3 d before isolation of On the next day, the medium was changed to 150 μL of bicarbonate-free + the transduced (VENUS ) cells. assay medium (adjusted to 17 mM glucose). O2 levels in the medium were measured at 15-s intervals for 3 min with a Seahorse Biosciences XFe Graphs and Statistics. GraphPad Prism 6 software was used for statistical 96 Analyzer. comparisons and graph generation. Values are reported as mean ± SEM, and P values were generated using the Student t test. The number of indepen- Antioxidant Activity Assays. A total SOD activity kit (Enzo Life Sciences) and dent experiments with cells from different donors is shown in brackets. total glutathione peroxidase activity kit (Cedarlane) were used to measure respective antioxidant activities in purified cells. ACKNOWLEDGMENTS. We acknowledge the excellent help of D. Wilkinson in collecting and processing samples of normal human mammary tissue. NAD(P)H Levels. NAD(P)H was measured on lysed aliquots of 105 FACS-purified We thank Drs. J. Sproul and N. Van Laeken for providing the reduction cells using a commercially available kit (BioVision Research Products). mammoplasty material, and G. Edin, M. Hale, D. Ko, and W. Xu for technical and FACS assistance. N.K. held a Canadian Breast Cancer Foundation (CBCF) British Columbia/Yukon Postdoctoral Fellowship. This work was also sup- qRT-PCR and Western Blot Analyses. Cells were rapidly lysed for RNA and ported by operating grants from the Canadian Breast Cancer Research protein isolation and analyzed by standard methods (33). The antibodies and Alliance, funded by the Canadian Cancer Society, and from the British primers used are listed in Tables S1 and S2, respectively. Columbia/Yukon division of the CBCF.

1. Nathan C, Ding A (2010) SnapShot: Reactive oxygen intermediates (ROI). Cell 140(6): 18. Rai P, et al. (2009) Continuous elimination of oxidized nucleotides is necessary to 951, e2. prevent rapid onset of cellular senescence. Proc Natl Acad Sci USA 106(1):169–174. 2. Cooke MS, Evans MD, Dizdaroglu M, Lunec J (2003) Oxidative DNA damage: Mech- 19. David SS, O’Shea VL, Kundu S (2007) Base-excision repair of oxidative DNA damage. anisms, mutation, and disease. FASEB J 17(10):1195–1214. Nature 447(7147):941–950. 3. Kryston TB, Georgiev AB, Pissis P, Georgakilas AG (2011) Role of oxidative stress and 20. Marnett LJ (2000) Oxyradicals and DNA damage. Carcinogenesis 21(3):361–370. DNA damage in human carcinogenesis. Mutat Res 711(1-2):193–201. 21. Maki H, Sekiguchi M (1992) MutT protein specifically hydrolyses a potent mutagenic 4. Diehn M, et al. (2009) Association of reactive oxygen species levels and radioresistance substrate for DNA synthesis. Nature 355(6357):273–275. in cancer stem cells. Nature 458(7239):780–783. 22. Cheng KC, Cahill DS, Kasai H, Nishimura S, Loeb LA (1992) 8-Hydroxyguanine, an → → 5. Schafer ZT, et al. (2009) Antioxidant and oncogene rescue of metabolic defects caused abundant form of oxidative DNA damage, causes G T and A C substitutions. J Biol – by loss of matrix attachment. Nature 461(7260):109–113. Chem 267(1):166 172. 6. Stingl J, Raouf A, Eirew P, Eaves CJ (2006) Deciphering the mammary epithelial cell 23. Grollman AP, Moriya M (1993) Mutagenesis by 8-oxoguanine: An enemy within. – hierarchy. Cell Cycle 5(14):1519–1522. Trends Genet 9(7):246 249. 7. Raouf A, et al. (2008) Transcriptome analysis of the normal human mammary cell 24. Nguyen LV, et al. (2014) Clonal analysis via barcoding reveals diverse growth and differentiation of transplanted mouse and human mammary stem cells. Cell Stem Cell commitment and differentiation process. Cell Stem Cell 3(1):109–118. 14(2):253–263. 8. Eirew P, et al. (2008) A method for quantifying normal human mammary epithelial 25. Asselin-Labat M-L, et al. (2007) Gata-3 is an essential regulator of mammary-gland stem cells with in vivo regenerative ability. Nat Med 14(12):1384–1389. morphogenesis and luminal-cell differentiation. Nat Cell Biol 9(2):201–209. 9. Lim E, et al. (2009) Aberrant luminal progenitors as the candidate target population 26. Kouros-Mehr H, Slorach EM, Sternlicht MD, Werb Z (2006) GATA-3 maintains the for basal tumor development in BRCA1 mutation carriers. Nat Med 15(8):907–913. differentiation of the luminal cell fate in the mammary gland. Cell 127(5):1041–1055. 10. Oberley LW, Buettner GR (1979) Role of superoxide dismutase in cancer: A review. 27. Yan W, Chen X (2006) GPX2, a direct target of p63, inhibits oxidative stress-induced Cancer Res 39(4):1141–1149. apoptosis in a p53-dependent manner. J Biol Chem 281(12):7856–7862. 11. McQueen CA, ed (2010) Comprehensive Toxicology (Elsevier, Kidlington, UK), 2nd Ed, pp 28. Kannan N, et al. (2013) The luminal progenitor compartment of the normal human – 325 328. mammary gland constitutes a unique site of telomere dysfunction. Stem Cell Rep 1(1): 12. Rhee SG, Woo HA, Kil IS, Bae SH (2012) Peroxiredoxin functions as a peroxidase and 28–37. – a regulator and sensor of local peroxides. J Biol Chem 287(7):4403 4410. 29. Wood ZA, Poole LB, Karplus PA (2003) Peroxiredoxin evolution and the regulation of 13. Kirsch M, De Groot H (2001) NAD(P)H, a directly operating antioxidant? FASEB J 15(9): hydrogen peroxide signaling. Science 300(5619):650–653. – 1569 1574. 30. Jarvis RM, Hughes SM, Ledgerwood EC (2012) Peroxiredoxin 1 functions as a signal – 14. Meister A, Anderson ME (1983) Glutathione. Annu Rev Biochem 52(1):711 760. peroxidase to receive, transduce, and transmit peroxide signals in mammalian cells. 15. Griffith OW, Meister A (1979) Potent and specific inhibition of glutathione synthesis Free Radic Biol Med 53(7):1522–1530. by buthionine sulfoximine (S-n-butyl homocysteine sulfoximine). J Biol Chem 254(16): 31. Stingl J, Emerman JT, Eaves CJ (2005) Enzymatic dissociation and culture of normal 7558–7560. human mammary tissue to detect progenitor activity. Methods Mol Biol 290:249–263. 16. Riley PA (1994) Free radicals in biology: Oxidative stress and the effects of ionizing 32. Kawamura E, et al. (2013) Identification of novel small molecule inhibitors of cen- radiation. Int J Radiat Biol 65(1):27–33. trosome clustering in cancer cells. Oncotarget 4(10):1763–1776. 17. Gros L, Saparbaev MK, Laval J (2002) Enzymology of the repair of free radical- 33. Eirew P, et al. (2012) Aldehyde dehydrogenase activity is a biomarker of primitive s-induced DNA damage. Oncogene 21(58):8905–8925. normal human mammary luminal cells. Stem Cells 30(2):344–348.

7794 | www.pnas.org/cgi/doi/10.1073/pnas.1403813111 Kannan et al. Downloaded by guest on September 27, 2021