Reduced Basal Nitric Oxide Production Induces Precancerous Mammary Lesions via ERBB2 and TGFβ

One third of newly diagnosed breast cancers in the US are early-stage lesions. The etiological understanding and treatment of these lesions have become major clinical challenges. Because breast cancer risk factors are often linked to aberrant nitric oxide (NO) production, we hypothesized that abnormal NO levels might contribute to the formation of early-stage breast lesions. We recently reported that the basal level of NO in the normal breast epithelia plays crucial roles in tissue homeostasis, whereas its reduction contributes to the malignant phenotype of cancer cells. Here, we show that the basal level of NO in breast cells plummets during cancer progression due to reduction of the NO synthase cofactor, BH4, under oxidative stress. Importantly, pharmacological deprivation of NO in prepubertal to pubertal animals stiffens the extracellular matrix and induces precancerous lesions in the mammary tissues. These lesions overexpress a fibrogenic cytokine, TGFβ, and an oncogene, ERBB2, accompanied by the occurrence of senescence and stem cell-like phenotype. Consistently, normalization of NO levels in precancerous and cancerous breast cells downmodulates TGFβ and ERBB2 and ameliorates their proliferative phenotype. This study sheds new light on the etiological basis of precancerous breast lesions and their potential prevention by manipulating the basal NO level.


Results
Basal level of No production in cultured mammary epithelial cells plummets during normal-to-precancerous progression. We previously demonstrated that non-malignant MECs produce the basal steady-state level of NO immediately after contacting laminin, but not collagen, in the extracellular matrix (ECM) 24 . This observation is in agreement with a notion that NO production is largely influenced by chemical and mechanical properties of the tissue microenvironment such as the ECM 53,[65][66][67] .
In our previuos studies, to test whether basal NO production is altered in breast cancer progression, we used the HMT-3522 breast cancer progression series composed of non-malignant S1 and malignant T4-2 cells 68,69 . S1 was derived from a benign mammary fibrocystic lesion and became spontaneously immortalized in culture 68,[70][71][72] . S1 cells retain non-malignant characteristics, requiring EGF to grow in culture and being unable to form tumors in nude mice 68,71,72 . (The use of S1 cells is restricted to passages below 60 because of genotypic drift at higher passages 71,72 .) S1 cells were utilized to generate the malignant counterpart (HMT-3522-T4-2) by withdrawing EGF from the growth medium and serially transplanting cells into animals, without overexpressing oncogenes 72,73 . We showed that basal NO production in response to laminin is pronounced in non-malignant S1, but is impaired in malignant T4-2 cells 24,53 . This observation was confirmed using normal vs. malignant human breast tissues 24 .
In the present study, we attempted to test whether basal NO production is altered stepwise during malignant progression of MECs. We utilized the MCF10A breast cancer progression series, composed of four isogenic cell lines: MCF10A, AT1, DCIS.COM and CA1d [74][75][76][77][78][79][80] . MCF10A, harboring many characteristics of normal breast epithelium, was derived from a mammary fibrocystic lesion and was spontaneously immortalized in culture 74 . AT1, showing the features of atypical hyperplasia, was generated by transfecting mutant H-Ras into MCF10A cells and serially transplanting them into nude mice 75,76 . AT1 cells do not initially grow out as carcinomas in nude mice; however, within 2 years 25% of them become cancerous 76 . DCIS.COM, showing dysplastic comedo DCIS phenotype, was derived from cells of a hyperplastic lesion formed by AT1 cells in nude mice 77 . DCIS.COM cells do not initially grow out as carcinomas in nude mice; however, after 9 weeks half of them form tumors 77 . CA1d was derived from a tumor formed by AT1 cells after one year in animals 78 . 100% of CA1d cells form metastatic tumors in nude mice 78 . www.nature.com/scientificreports www.nature.com/scientificreports/ We pulse-treated these cell lines with a drip of reconstituted laminin-rich ECM (lrECM, a.k.a. Matrigel) and determined the level of intracellular NO using a fluorescent probe DAF-FM 24,81 . While non-malignant (MCF10A) cells produced an appreciable level of DAF FM signal, the level was dramatically (~−60%) reduced in precancerous (AT1) cell and remained low in cells with more advanced stages (DCIS and CA1d) (Fig. 1A). As a complementary approach, we measured the level of the NO metabolite (nitrite) in the conditioned media of cells cultured in 3D lrECM using a fluorescent probe DAN 24,82 and observed similar results (Fig. 1B). Furthermore, we analyzed the level of S-nitrosocysteine (SNOC, a NO-dependent protein modification) as an indicator of NO production. Consistently, the cytosolic, but not nuclear, SNOC level progressively declined along with cancer progression (Fig. 1C).
To test the generality of this phenomenon, we compared basal NO production in response to lrECM among a panel of normal/non-malignant vs. cancerous breast cell lines using the NO probe DAF-FM. The specificity of the signal was confirmed by quenching it with the NOS inhibitor, L-NAME (2.5 mM), an L-arginine analog that competitively inhibits the substrate binding to NOS (IC 50 ~ 70 μM) 83 [and to arginase (IC 50 ~ 27 mM) 84,85 ]. DAF-FM signals were generally much higher in normal/non-malignant MECs than breast cancer cells ( Supplementary  Fig. 1A,B), consistent with previous reports 24,53,86 . These results altogether demonstrate that the levels of basal NO become lower in MECs along with cancer progression.
NOS-1 and -3 levels remain high, while NOS-2 level remains low, in all cell lines of the breast cancer progression series. NO is produced by three isoforms of NO synthases (NOS 1-3) 21 . To determine which isoform is involved in NO production in MECs, we examined the level of each NOS isoform in the MCF10A progression series. In non-malignant MCF10A cells, NOS-1 and -3 were both expressed at high levels, whereas NOS-2 was undetectable ( Fig. 2A). We then examined the level of each NOS isoform in normal mouse mammary glands. The expression patterns of NOS isoforms were the same as those in MCF10A, consistent with previous reports (Fig. 2B) [26][27][28] . Interestingly, throughout the entire progression series, NOS expression patterns Reduced No production in cell lines of breast cancer progression series correlates with increased acidity and oxidative stress that depletes the Nos cofactor BH 4 . We further examined why NO production dramatically declines in cell lines of the breast cancer progression series, while NOS levels remain unchanged (Figs 1A, 2A). We postulated that this might be attributed to certain physiological traits associated with cancer progression that could debilitate NOS functions. To test this, we measured the intracellular pH and oxidative stress (superoxide and total reactive oxygen species). Acidification and oxidative stress are hallmarks of cancer metabolism 87 and also known to attenuate NOS functions 49,88-90 . As expected, the acidity and oxidative stress were significantly elevated along with cancer progression ( Fig. 2A-C). www.nature.com/scientificreports www.nature.com/scientificreports/ In particular, oxidative stress has been shown to deplete the essential NOS cofactor, BH 4 . BH 4 helps tether two NOS monomers to form the functional NOS homodimer, allowing for coupling two reactions required for NO production: (1) reduction of molecular oxygen and (2) oxidation of arginine 91 . When BH 4 (or arginine) is deficient , however, NOS remains as monomers, and these two reactions are uncoupled, releasing superoxide (from the former reaction only) instead of producing NO -a phenomenon termed 'NOS uncoupling' 49,88 . To test this possibility, we measured the intracellular BH 4 level in cell lines of the breast cancer progression series. As expected, BH 4 level was dramatically (~−40%) reduced in normal-to-precancerous transition and remained low thereafter (Fig. 2D). This paralleled the dramatic increase of superoxide levels, suggesting the possible occurrence of NOS uncoupling (Fig. 2B). We additionally tested for the potential involvement of known endogenous NOS inhibitors, asymmetric dimethylarginine (ADMA) and dynein light chain LC8-Type 1 (DYNLL1) 92 , in reduction of NO level in the progression series. ADMA and DYNLL1 levels did not significantly change throughout the progression series (except for some increase of ADMA in DCIS) (Fig. 3E, Supplementary  Fig. 3), excluding their major contributions. These results altogether strongly suggest that deficiency of the NOS cofactor, BH 4 , is a critical contributor to reduction of basal NO producion in MECs during cancer progression. www.nature.com/scientificreports www.nature.com/scientificreports/ pharmacological deprivation of No impairs mouse mammary gland development, but induces desmoplastic eCM due to overactivation of tGFβ. To determine whether reduction of NO level contributes to breast cancer pathogenesis in vivo, we pharmacologically inhibited NO production in inbred wild-type mice (BALB/c) for a duration of 6 weeks (from 4 to 10 weeks old) by intraperitoneal (i.p.) injection of L-NAME (NOS antagonist, 20 mg/kg), in comparison to control (vehicle) and L-arginine (NOS substrate [agonist], 20 mg/ kg) treatments. L-NAME-treated mouse mammary glands completely lacked alveoli, demonstrating the clear developmental defect, consistent with NO's critical roles in mammary gland development (Fig. 4A) 27,29,[31][32][33][34] . To confirm altered NO levels, drug-treated mammary glands were stained for SNOC. Control glands exhibited the basal level of SNOC, whereas L-arginine-treated glands showed dramatically elevated SNOC level. On the other hand, L-NAME-treated glands exhibited almost undetectable SNOC level (Fig. 4B, Supplementary Fig. 4A). Reduced NO bioavailability in patients with chronic conditions, such as diabetes, cardiovascular disease and obesity, often leads to formation of stiff, desmoplastic (fibrotic) ECM which is directly linked to increased cancer risk [93][94][95] . To test this possibility, we measured the density of collagen fibers in drug-treated mammary tissues using second harmonic generation (SHG) technique 96 . As expected, periductal collagen level was dramatically elevated in L-NAME-treated mammary tissues, compared to control or L-arginine-treated tissues (Fig. 4B, Supplementary Fig. 4B). To determine the cause of the desmoplastic ECM in L-NAME-treated mammary tissues, we tested for the involvement of transforming growth factor β (TGFβ), the major activator of collagen biosynthesis 97 . We stained mammary tissues for phospho-SMAD3 (active form), the downstream effector of TGFβ signaling. Phospho-SMAD3 level, in fact, was dramatically (>5-fold) elevated in L-NAME-treated mammary glands (Fig. 4B, Supplementary Fig. 4C), suggesting strong activation of TGFβ signals upon NO depletion. These results demonstrate that NO deprivation in mammary glands of prepubertal to pubertal mice impairs gland development, but induces desmoplastic ECM along with overactivation of TGFβ signaling.
No-deprived mouse mammary glands form precancerous lesions that overexpress ERBB2. Formation of desmoplastic ECM indicates the occurrence of certain pathological conditions in NO-deprived mammary tissues 95 . Histological examination strikingly revealed that every L-NAME-treated animal (n = 6) formed multiple (peripheral) papillomas -precancerous mammary lesions 54-58 -visualized by H&E staining of serial sections. In contrast, neither control nor L-arginine-treated animals formed such lesions (Fig. 4C). We then tested for the possible involvement of the ERBB2 oncogene, highly linked to precancerous progression of MECs [98][99][100][101] . While a change in the ERBB2 gene locus is found in 20~30% cases of invasive breast cancers, it is even more prevalent in premalignant breast lesions (~50%), suggesting that a change in ERBB2 gene is an early event in breast carcinogenesis [102][103][104] . As expected, ERBB2 levels were dramatically (>3-fold) elevated in L-NAME-treated mammary glands (Fig. 4D, Supplementary Fig. 4D). It is reported that overexpression of ERBB2 in normal or cancerous breast cells induces cellular senescence as an anti-carcinogenic mechanism [105][106][107] . In addition, senescence is shown to be the most prevalent in precancerous lesions, compared to normal or invasive lesions [59][60][61] . Therefore, to test whether the L-NAME-treated glands had undergone senescence, we determined the levels of well-established senescence markers: heterochromatin protein 1 (HP1), β-Galactosidase (β-Gal), p21 and p27 60,61,108 . All these senescence markers were highly elevated in L-NAME-treated glands, whereas they were almost undetectable in control or L-arginine-treated glands (Fig. 4D, Supplementary Fig. 2E-H). These results demonstrate that NO deprivation in mouse mammary glands induces precancerous lesions that strongly express ERBB2, accompanied by the occurrence of senescence.

Inhibition of NO in non-malignant human mammary epithelial cells in 3D ECM cultures impairs acinar morphogenesis and upregulates tGFβ and ERBB2.
To determine whether the effects of NO modulation on mouse mammary glands were intrinsic to MECs or due to stromal and systemic influence, we treated mono-cultures of non-malignant human MECs (MCF10A cells) with L-NAME, in comparison to control (PBS) or L-arginine, in 3D lrECM for 3 weeks. Consistent with our previous report 24 , L-NAME impaired formation of mammary acini, but induced formation of disorganized, proliferative aggregates ( To determine the generality of the effects of modulating NO level, we applied L-arginine or L-NAME to ERBB2-positive (amplified) SKBR3 breast cancer cells 110 . Surprisingly, even after a short-term (overnight) treatment, L-NAME dramatically (>3-fold) elevated the level and membrane-localization of ERBB2 over control, whereas L-arginine almost abrogated them ( Supplementary Fig. 5D). These results suggest that the effects of NO modulation on mammary glands we observed are largely intrinsic to MECs and might have manifested soon after treatment.

Inhibition of No in breast epithelial cells induces bi-lineage and stem cell-like phenotype in both mammary glands and 3D ECM cultures.
It has been reported that activation of certain oncogenic pathways, such as ERBB2, in normal or cancerous MECs could induce stem cell-like properties 101,111 . These cells are characterized by luminal (CK8/18)/basal (CK14) bi-lineage phenotype, as well as high expression of stem cell markers such as CD44, and are also found to be prevalent in precancerous lesions 62-64 . Since NO deprivation dramatically elevated ERBB2 in wild-type mouse mammary glands (Fig. 4D, Supplementary Fig. 4D), we tested whether this might have induced stem cell-like properties 101 . First we tested for the occurrence of bi-lineage phenotype by co-staining drug-treated mammary glands for CK8/18 (luminal) and CK14 (basal) 62 . As previously reported, control and L-arginine-treated mammary glands showed mutually exclusive pattern of CK8/18 vs. CK14 112 . In contrast, L-NAME-treated glands showed a significant (>10-fold) increase in CK18 high /CK14 high bi-lineage cells (Fig. 6A, Supplementary Fig. 6A) 62 .
Next, we tested for the induction of stemness by staining the drug-treated mammary glands for stem cell markers, alkaline phosphatase (AP, a pluripotent stem cell marker) 113 , CD44 (another stem/progenitor marker) and CD24 (the luminal lineage marker) [114][115][116] . In both humans and mice, CD44 high /CD24 low phenotype is the predominant marker for multipotent stem/progenitor cells that have the highest repopulating (i.e., tumor-initiating) potential, while CD44 high /CD24 high and CD44 low /CD24 high phenotypes are the markers for cells "committed" to luminal differentiationin [114][115][116][117] . (CD44 high /CD24 high phenotype, nevertheless, is linked to increased drug-resistance 118 ). In control and L-arginine-treated glands, the three stem cell markers, AP, CD44 and CD24, were virtually all absent. In contrast, in L-NAME-treated mammary glands, both AP and CD44 were www.nature.com/scientificreports www.nature.com/scientificreports/ strongly expressed in all the epithelia, whereas CD24 was mostly expressed in the luminal layer and expressed at low to moderate levels in the basal layer. (Fig. 6B. Supplementary Fig. 6B-D). These results demonstrate that L-NAME-treatment of mammary glands induces the stem cell population (CD44 high /CD24 low ) on the basal layer, as well as another population of cells committed to luminal differentiation (CD44 high /CD24 high ) on the luminal layer 117 .
We tested whether this in vivo phenotype could be recapitulated in mono-cultures of human MECs in 3D lrECM. Similar to our in vivo results, 3D colonies of control and L-arginine-treated non-malignant MCF10A cells showed distinct bi-layers of CK8/18-positive (luminal) vs. CK14-positive (basal) cells (Only 5~7% of cells were CK18 high /CK14 high bi-lineage). In contrast, L-NAME-treated cells were virtually all (>70%) CK18 high /CK14 high bi-lineage (Fig. 6C, Supplementary Fig. 6E). Stem cell markers, AP, CD44 and CD24, were almost undetectable in control and L-arginine-treated colonies. Conversely, in L-NAME-treated colonies, the expression of AP and CD44 was strongly positive throughout; CD24 expression was highly elevated on the periphery, but remained low in the center (marking the peripheral cells as CD44 high /CD24 high vs. central cells as CD44 high /CD24 low ) (Fig. 6, Supplementary Fig. 6F-H). These results altogether demonstrate that L-NAME-treatment of normal/ www.nature.com/scientificreports www.nature.com/scientificreports/ non-malignant MECs of humans and mice, both in culture and in vivo, respectively, induces a stem cell-like population with CK18 high /CK14 high bi-lineage phenotype and CD44 high /CD24 low expression.

Inhibition of No in breast epithelial cells promotes formation of mammospheres enriched for stem cell-like cells.
To further validate the induction of stemness by L-NAME treatment, drug-treated MCF10A cells were cultivated on PolyHEMA-coated (non-adherent) plates, which allows for detection of self-renewal capacity through formation of mammospheres [119][120][121][122][123][124] . For mammosphere formation assay as described previously 119,120,123,124 , cells were seeded at the densities of 1,250~10,000 cells/48-well. Both L-NAME and L-arginine-treatments increased the size, number and formation efficiency of mammospheres over control; however, L-NAME yielded the values twice as much as L-arginine ( Fig. 7A-D, Supplementary Fig. 7A). The proportions of CK18 high /CK14 high bi-lineage cells within mammospheres were significantly (>4-fold) higher in L-NAME and L-arginine-treated spheroids than control spheroids (Fig. 7E, Supplementary Fig. 7B), attesting to the increase of stem cell-like populations 62,112,125 .
Aiming to differentiate the effects of L-NAME from those of L-arginine, we performed a limiting dilution assay on drug-treated MCF10A cells that determines the abundance of cells capable of forming spheroids at clonal densities (200~1600/96-well) 126,127 . L-NAME treatment elevated the abundance of spheroid-forming cells by 5-fold over control and 8-fold over L-arginine treatment (Fig. 7F). These results altogether demonstrate that L-NAME treatment greatly increased the abundance of stem cell-like cells that possess clonal expansion capacity.
To further confirm the stemness of drug-treated spheroids, we measured the levels of stem cell markers, CD44 and CD24 128 , by immunofluorescence imaging of paraffin-embedded/sectioned spheres 129 . L-NAME-treated spheroids were strongly positive for CD44, whereas L-arginine-treated spheroids showed little or no expression. CD24 expression, on the other hand, was low in all conditions (Fig. 7G, Supplementary Fig. 7C,D).
As a complementary approach, we performed FACS analysis on cells dissociated from mammospheres. Consistent with the result from immunofluorescence imaging, L-NAME-treated mammospheres showed a dramatic increase in CD44 level. About 50% of cells were CD44 positive, and 1/4 of these CD44-positive cells (11.5% of the total cells) were CD44 high /CD24 low , cells with the highest repopulating potential [114][115][116][117] (Fig. 7H). On the other hand, there were almost no (0.2~0.6%) CD44-positive cells in control and L-arginine-treated mammospheres. While CD24 level increased in L-arginine-treated (+6.5%) and L-NAME-treated spheroids (+24.1%), a significant fraction of cells (46~65%) showed low expression levels (Fig. 7H). These results altogether demonstrate that L-NAME-treatment elevated stem cell-like cells (CD44 high /CD24 low ) which have the highest self-renewal ability and clonogenicity 128 .

Normalization of No levels with the BH 4 precursor, sepiapterin, ameliorates the malignant phenotype of breast cancer cells.
To further test whether dysregulated NO levels contribute to the phenotype associated with breast cancer progression, we sought to normalize NO levels in precancerous and cancerous MECs. We showed above that reduced basal NO production in cultured MECs along with cancer progression was linked to oxidative depletion of the NOS cofactor, BH 4 , which triggers NOS uncoupling (Figs 1A-C, 2B-D) 49,88 .
In an attempt to ameliorate the conditions of diseases including breast cancer, the BH 4 precursor, sepiapterin, has been successfully utilized for 'recoupling' NOS in a number of cell culture and preclinical studies 49,[130][131][132][133][134] .
We tested whether the application of sepiapterin could help normalize NO levels in precancerous and cancerous MECs in 3D lrECM cultures and whether this could ameliorate the malignant phenotype. After cultivation of the MCF10A progression series in the presence of sepiapterin at 20 μM 49 , NO levels of precancerous and cancerous cell lines (AT1, DCIS and CA1d) were restored to the levels comparable to (or at least half of) the level of non-malignant MCF10A cells. In contrast, NO level in non-malignant MCF10A cells did not change after sepiapterin treatment (Fig. 8A). Importantly, 20 μM of sepiapterin dramatically (>50%) reduced proliferation indices (i.e., colony size and Ki67 positivity) of precancerous and cancerous cell lines, but not of non-malignant cells (Fig. 8B). This was accompanied by a dramatic reduction in the levels of phospho-SMAD3 and ERBB2, which otherwise were pronounced in precancerous and cancerous cells (Fig. 8C). These results confirm that deficiency of BH 4 is a major contributor to reduced basal NO production and malignant phenotype of precancerous and cancerous MECs.
To test whether these effects of sepiapterin are dose-dependent, we treated the MCF10A progression series cultivated in 3D lrECM with a different concentration of sepiapterin (0, 20 or 100 μM). The increasing concentrations of sepiapterin (20 to 100 μM) progressively normalized NO levels in precancerous and cancerous cells, without affecting that of MCF10A cells (Supplementary Fig. 8A). Nevertheless, the growth-suppressive effects (by colony size and Ki67 positivity) of sepiapterin did not differ between the concentrations of 20 and 100 μM ( Supplementary Fig. 8B-D). This suggests that the activity of sepiapterin is threshold-dependent, rather than concentration-dependent, at least within this range.
Furthermore, sepiapterin (20 and 100 μM) helped "normalize" the cellular phenotype of precancerous and cancerous cells of the MCF10A progression series also in a threshold-dependent manner. In particular, AT1 cells restored the apico-basal polarity, indicated by the proper localization of integrin α6 (the basal marker) and cleaved caspase 3 (a marker of lumens) 24,135 analogous to non-malignant MCF10A cells ( Supplementary  Fig. 9A,B). In DCIS and CA1d cells, apico-basal polarity was only partially restored by sepiapterin treatment (Supplementary Fig. 9A,B). Nevertheless, the expression of CK14 (basal cell marker) and the central localization of CK8/18 (luminal cell marker) were almost completely restored (Supplementary Fig. 9C).
We applied sepiapterin to ex vivo 3D cultured mammary tumors from MMTV-PyMT mice (18 weeks old) and tested for its anti-tumor effects. Sepiapterin greatly reversed the proliferative phenotype of the epithelia and even partially restored normal-like glands within tumors in a week. Nevertheless, it needs to be noted that sepiapterin also significantly elevated the vascular density in tumors in support of a notion that the increase of NO promotes antiogenesis 20,136,137 (Fig. 8D). www.nature.com/scientificreports www.nature.com/scientificreports/ To test the generality of the effects of sepiapterin on breast cancer cells, we pulse-treated different breast cancer cell lines (basal type: MDA-MB231 and MDA-MB468; luminal type: MCF7 and SKBR3) with a different concentration of sepiapterin (0, 20, or 100 μM) in lrECM cultures. Even after two hours of treatment, both 20 and 100 μM of sepiapterin equally increased NO levels of these cancer cell lines by 2~3-fold ( Supplementary  Fig. 10A). Surprisingly, sepiapterin decreased Ki67 positivity of cancer cells by 60~100% (luminal cells showed better responsiveness than basal cells) (Supplementary Fig. 10B). This result augments a previous report of the growth-inhibitory effects of sepiapterin on breast cancer cells in culture and in animals after 2 days of treatment 49 .
As a complementary approach, we administered a low concentration (2.5 μM) of a NO donor, SNAP or GSNO, to another breast cancer progression series, HMT-3522 S1 (non-malignant) and T4-2 cells (malignant) 24,68,69 , cultivated in 3D ECM. This concentration of NO donors were chosen in an attempt to induce production of the physiological level (1-4 μM) of NO previously reported by us and others 24,138,139 . Similar to sepiapterin, NO donors at 2.5 μM significantly inhibited growth of tumor cells and restored their apico-basal polarity to the level analogous to that of non-malignant cells (Supplementary Fig. 11A-C). To test whether these effects of NO donors are dose-dependent, we administered a different concentration of SNAP or GSNO (0, 2.5 or 10 μM) to the HMT-3522 progression series cultivated in 3D lrECM. Interestingly, while the lower level (2.5 μM) of SNAP and GSNO restored the S1 cell-like polarity and growth-arrested phenotype in T4-2 cells, the higher level (10 μM) of the same NO donors showed no such effects ( Supplementary Fig. 12A-C). This result is in line with the well-documented concentration dependence of NO's bioactivities in cancer [17][18][19][20] .
The findings altogether suggest that normalization of basal NO level may have a therapeutic potential for breast cancer.

Discussion
Advances in imaging and screening technologies have enabled the detection of precancerous, early-stage breast lesions, the majority of which are ductal carcinoma in situ (DCIS) 2 . Precancerous lesions arise from clonal expansion of a single cell 140 . Precancerous lesions are the precursors to invasive breast cancers, and about 40% of them could progress to invasive forms, if untreated 2 . It is, however, not fully understood what drives formation of precancerous breast lesions, making the management of the disease challenging 2 .
In the study, we hypothesized that aberrant NO production might contribute to the formation of precancerous breast lesions and tested this possibility. Utilizing the MCF10A breast cancer progression series, we showed that the basal NO level in MECs plummeted along with breast cancer progression, consistent with our previous finding using another breast cancer progression series, HMT-3522 24 . In particular, such decline was the most notable in non-malignant to precancerous transition (Fig. 1A). Reduction of basal level of NO in MECs along with cancer progression was largely attributed to reduction of the NOS cofactor BH 4 under increased oxidative stress (Fig. 3B-D). In contrast, the levels of NOS-1 and -3, the enzymes responsible for production of the basal level of NO in the breast, remained unchanged during cancer progression and, thus, were irrelevant to reduction of NO ( Fig. 2A).
To confirm the pathological relevance of reduced basal NO production to breast carcinogenesis, we pharmacologically inhibited NO production in prepubertal to pubertal mice using L-NAME, an unreactive L-arginine analog. Previous studies report somewhat conflicting observations for the effects of L-NAME on cancer. Some studies report anti-cancer effects of L-NAME on already established cancer cells [141][142][143] , while others report the opposite [144][145][146][147] . Such discrepancy may reflect the complex activities of NO and NOS in cancer 18,20,24,[35][36][37][38][39][40][41][42][43][47][48][49][50]145,148,149 as well as possible co-suppression of NOS and arginase by L-NAME at high concentrations (>5 mM) 84,85,150 . L-NAME's effect on the pathophysiology of normal cells, on the other hand, has yet to be explored.
L-NAME treatment of developing mouse mammary glands impaired alveologenesis, inducing precancerous lesions and desmoplastic ECM (Fig. 4A-C). Consistently, these mammary glands showed overactivation of ERBB2, closely linked to formation of precancerous lesions 100,101 , and fibrogenic TGFβ (Fig. 4B,D) 97 . Activation of these pathways were accompanied by the induction of senescence and stem cell-like properties, both prevalent in precancerous lesions (Figs 4D, 6B) 59,62,63,101 . We further showed that normalization of NO levels, by the use of the BH 4 precursor (sepiapterin) or the physiological level of NO donor (SNAP or GSNO), suppressed TGFβ and ERBB2 signals and ameliorated the malignant phenotype of precancerous and cancerous MECs (Fig. 8A-D,  Supplementary Fig. 11A-C).
NO's bioactivities are largely dependent on its concentration, timing and context [17][18][19] . In healthy tissues, NO production is tightly regulated to attain the right condition 151 . In disease states, on the other hand, NO production is often dysregulated, leading to a deficient or excessive level of NO that, in either case, could contribute to the disease pathogenesis 39,40 . Such a complex, paradoxical role of NO in disease conditions, especially in cancer, have led to conflicting reports and a notion that NO plays a double-edged role as both a cancer-promoter and -inhibitor 17,18,20 . The enigma of NO's role in cancer might be partly resolved by clarifying how NO is involved in a specific stage of a particular type of cancer under a certain context.
The present study demonstrates that maintenance of the physiological NO level in the breast exerts protective effects against formation of precancerous lesions. NO production, however, could be impaired by oxidative stress that depletes the NOS cofactor BH 4 , contributing to pathogenesis. Under such a condition, NOS is dysfunctional (uncoupled) and produces superoxide instead of NO, exacerbating oxidative stress 49,88 . In line with this notion, we observed that NO and BH 4 levels both dramatically declined along with an increase in oxidative stress in precancerous and cancerous MECs cultivated in 3D lrECM (Figs 1A and 3B-D).
One of the most compelling findings in this study is that NO deprivation simultaneously upregulated TGFβ and ERBB2 in MECs both in vivo and in vitro (Figs 4B,D and 5A,B). TGFβ and ERBB2 are both under tight regulation in healthy tissues, whereas they become dysregulated during carcinogenesis 152,153 . TGFβ activity is primarily regulated post-translationally, especially through formation of the latent complex along with LTBP1 (2019) 9:6688 | https://doi.org/10.1038/s41598-019-43239-x www.nature.com/scientificreports www.nature.com/scientificreports/ and LAP 154 . ERBB2 elevation, in contrast, is primarily ascribed to gene amplification 155 . However, other mechanisms that do not involve changes in DNA sequence also play major roles in regulation of ERBB2 level. These mechanisms include transcriptional and post-transcriptional activation [156][157][158] as well as protein stabilization 159 . In fact, it is shown that ~20% of ERBB2-positive tumors detected by IHC (score of 3+) do not harbor amplification of the ERBB2 gene 160,161 . We are postulating that the basal steady-state level of NO produced in the healthy breast tissue might negatively regulate proteins involved in upregulation of TGFβ and ERBB2 via S-nitrosylation (SNO). SNO plays a large part of NO signal, regulating structures and functions of >3000 proteins 162 . In healthy cells, a subset of proteins are constitutively S-nitrosylated at the basal level to remain under control 25 . In disease states including cancer, on the other hand, SNO level could be dysregulated, contributing to disruption of homeostasis 25 . Currently, we are examining the potential role of SNO in suppressing TGFβ and ERBB2 expression and activity in non-malignant MECs.
Our study demonstrates that normalization of NO level with the BH 4 precursor, sepiapterin (20 or 100 μM), or with a low level (2.5 μM) of a NO donor, SNAP or GSNO, reduced ERBB2 and TGFβ signals and suppressed the proliferative phenotype of precancerous and cancerous MECs in 3D lrECM. NO normalization also restored cell polarity and lineage markers in these cells (a process termed 'phenotypic reversion' 24 ) (Fig. 8, Supplementary  Figs 8-12). Surprisingly, application of sepiapterin to ex vivo 3D cultured mammary tumors partially restored normal-like glands and largely reversed the proliferative phenotype of tumor epithelia in a week (Fig. 8D). However, it is critical to note that modulation of NO levels should aim at restoring the physiological basal level of NO (1-4 μM) in MECs 24,138,139 . Too much or too little NO supplementation, on the other hand, could result in pathogenic effects because of the concentration-dependent NO's bioactivities 18,20 . In fact, while NO donors at the lower concentration (2.5 μM) exerted anti-cancer activity, the same donors at the higher concentration (10 μM) showed no such activity ( Supplementary Fig. 12A-C). In contrast, sepiapterin exerted the anti-cancer activity equally at both 20 and 100 μM, and thus was threshold-dependent at least within this range. These results suggest that sepiapterin might have a wider therapeutic window than NO donors.
Our present findings strongly suggest the translational potentials of sepiapterin (and possibly the physiological concentrations of NO donors) for treatment of early-stage breast cancer. However, the advancement of such approach would depend on the development of tumor-restricted drug delivery systems, because systemic administration of these agents could have potential drawbacks 17,[163][164][165] . While systemic administration of sepiapterin to animals is shown to reduce tumor cell proliferation 49 , it also enhances vascularization of tumor tissues 20,136,137 . Consistently, when we applied sepiapterin to ex vivo 3D cultured mouse mammary tumors, their proliferative phenotype was dramatically suppressed, whereas the vasculature was significantly enhanced (Fig. 8D). Such enhanced vasculature by sepiapterin treatment might eventually help tumor cells re-grow and reduce the efficacy of the drug. To circumvent these drawbacks, we are currently developing liposome-based systems that could specifically deliver sepiapterin to the breast epithelia of lesions via specific homing peptides 166 . Moreover, future studies could possibly determine whether certain diet and exercise routines help maintain the physiological NO level in the breast and prevent formation of precancerous lesions. Cell culture and reagents. The MCF10A breast cancer progression series comprising non-malignant MCF10A, premalignant AT1, in situ carcinoma DCIS.COM and metastatic CA1d, was maintained as described 78 . The isogenic cell lines of the HMT-3522 human breast cancer progression series, comprising non-malignant S1 and malignant T4-2 cells, were maintained as described previously 24 . All the other breast cancer cell lines were maintained in DMEM/F12 with 10% FBS and 1% penicillin/streptomycin. For 3D culture experiments, cells were seeded at the density of 2.5 × 10 4 cells/cm 2 for non-malignant cells and 1.8 × 10 4 cells/cm 2 for malignant cells in growth factor reduced Matrigel (BD Biosciences) and maintained for 10-21 days with addition of fresh medium on alternate days. For inhibition of NO production, cells were treated with 2.5 mM L-NAME (N ω -Nitro-L-arginine methyl ester hydrochloride, Sigma-Aldrich); for induction of NO production, 2.5 or 10 μM SNAP (S-Nitroso-N-acetyl-DL-penicillamine,) or 2.5 or 10 μM GSNO (S-nitrosoglutathione, Sigma-Aldrich) was used. To compensate for the reduced BH 4 level in cancer cells, 20 or 100 μM L-sepiapterin (BH 4 precursor, Sigma-Aldrich) was used. The NOS inhibitors were all obtained from Cayman Chemical. NOS1 inhibitor (Nω-Propyl-L-arginine hydrochloride) was used at 10 µM; NOS2 inhibitor (1400 W) at 50 µM; and NOS3 inhibitor (L-NIO) at 10 µM 167 .

Materials and Methods
Antibodies. The following antibodies were used. Anti

No measurement in live cells.
To capture the snap shot of NO production in live cells after Matrigel addition, a dye DAF-FM DA (4-Amino-5-Methylamino-2′,7′-Difluorofluorescein Diacetate, Life Technologies) was used according to the manufacturer's protocol. Briefly, cells were seeded at 1.0 × 10 5 cells/24 well-plate maintained in phenol red free growth media overnight and pre-treated with the dye for 1 h. After the drip (5%) of Matrigel (phenol red-free) was added, cells were incubated in dark for 1 h and washed in fresh media. Micrographs were taken on live cells by fluorescent microscope with FITC filter, and the signal intensity/cell was measured with ImageJ.
Nitrite measurement. To quantify the cumulative level of nitric oxide produced by cells, more stable nitric oxide metabolite, nitrite, was measured based on the reaction of a dye DAN (diaminonaphthalene) by using Measure-IT High-Sensitivity Nitrite Assay Kit (Life technologies), according to the manufacturer's protocol. Briefly, cells were plated at 1 × 10 6 cells/60 mm-plate and maintained overnight. Cells were maintained in 2 ml of the fresh phenol red-free medium containing 5% Matrigel (phenol red-free, Corning, #356237) for 24 h. The conditioned medium was harvested and spun to remove the Matrigel. Ten μl of the cleared conditioned medium was reacted with the assay reagents in dark, and the signal intensity was measured using nitrite standards at the excitation/emission maxima of 340/410 nm.  Nos knockdown by shRNA. To knock-down each NOS isoform, lentiviral NOS shRNA vectors that target respective sequences were obtained from Origene (Supplementary Table 1). As a control, lentiviral vector that expresses scramble sequence (CAT No: TR30021) was used. Lentiviral particles were produced as described below and applied to MCF10A cells to make stable cell lines.
Lentivirus production and transduction. Lentivirus production and transduction of target cells were conducted following the guideline by Origene. Briefly, lentivirus vector and packaging plasmid mix (Origene) were transfected into 293FT cells (Invitrogen) using Lipofectamine ® 3000. After 48 hrs, medium was harvested, filtered and used to infect target cells with the addition of polybrene (10 μg/ml). After 24 hrs, medium was replaced. At 72 hrs post-infection, puromycin (0.5 μg/ml) was added for selection and maintained throughout the culturing period.

Immunohistochemistry.
To determine the expression of specific markers, paraffin-embedded sections of mouse mammary tissues were analyzed by immunohistochemistry. Briefly, sections were deparaffinized, hydrated, and treated with antigen unmasking solutions (Vector Laboratories, Inc.) or with Tris-EDTA Buffer (10 mM Tris Base, 1 mM EDTA Solution, 0.05% Tween 20, pH 9.0) heated to 95-100 °C in a pressure cooker. After being blocked with nonimmune goat serum, sections were processed for immunofluorescence staining as described below.
Immunofluorescence staining and imaging. Immunofluorescence staining/imaging was performed as described previously 24 . Samples were incubated with primary antibody for overnight at 4 degree in a humidified chamber. After intensive washing (three times, 15 min each) in 0.1% BSA, 0.2% Triton-X 100, 0.05% Tween 20, 0.05% NaN3 in PBS, fluorescence-conjugated secondary antibodies (Molecular Probes) were added for 1 hr at room temperature. Nuclei were stained with 0.5 ng/ml DAPI. After mounted with anti-fade solution, epi-fluorescence imaging was performed on Olympus IX70 microscope using CellSens software. Confocal fluorescence imaging with Second Harmonics Generation (SHG) module 96  Image Analysis. Quantification of fluorescence signal in micrographs was performed with ImageJ software (NIH) referring to the owner's manual (http://imagej.net/docs/guide/146.html). Briefly, a region of interest (ROI) was determined in reference to an image of DAPI-stained nuclei. For quantification of signal in individual cultured cells, the whole cell was selected as ROI. For quantification of signal in individual organoids in cultures or tissues, each organoid was selected as ROI. For quantification of second harmonics generation (SHG) signal in mammary tissues, the ROI was defined as the periductal ECM/stromal area between the epithelial and adipose layers. For quantification of S-NOC signal in mammary tissues, epithelial layers were selected by setting a threshold range, and the intensity was measured for each gland. For each ROI, the average intensity per pixel was measured, and background intensity was subtracted. For each sample group, at least 50 to 200 measurements were performed. Furthermore, measurement of each sample set was repeated by at least three people, and the results were combined for the final data. The mean value was represented as arbitrary units (AU). The statistical significance of the data was further evaluated using Graphpad Prism Version 5 software (see statistics section).
Animal studies. All animal experiments conformed to The Guide for the Care and Use of Laboratory Animals (National Research Council, National Academy Press, Washington, D.C., 2010) and were performed with the approval of the Institutional Animal Care and. Use Committee of the University of Toledo, Toledo, OH (Protocol No: 108658). Three-weeks old female BALB/c (n = 18) mice were obtained from the Jackson Laboratory (Bar Harhor, MN) and housed under a 12 hr light-dark cycle and given regular chow. Starting at the age of 4 weeks old, mice were given intraperitoneal injection of either drug (vehicle: PBS (100 µl), L-arginine (20 mg/kg, 100 µl), or L-NAME (20 mg/kg, 100 µl)) every other day for 6 weeks. Body weight and morbidity of animals were monitored throughout the treatment period. At the end of treatment period, mice were euthanized, and inguinal mammary glands were harvested. Number 4 mammary glands were processed for whole mounting as described 168 , while number 5 glands were processed for paraffin-embedding and sectioning. Whole mount was imaged by Cytation ™ 5 Cell Imaging Multi-Mode Reader (BioTech Instruments). To determine the gross morphology of glands, paraffin sections were deparaffinized, hydrated and stained with eosin/hematoxyline. Other sections were analyzed by immunohistochemistry and SHG imaging.
Mammosphere assay. MCF10A cell lines were pretreated for 7 days with either vehicle, L-arginine (2.5 mM) or L-NAME (2.5 mM), and then subjected to mammosphere assay as described 120,121,129 . Briefly, the day before experiment, culture plates were coated with poly-HEMA (#3932, Sigma) as previously described 171 . Single cell suspension of 4 × 10 4 cells were resuspended in 2 ml media from Mammocult Human medium kit (#5620, Stemcell Technologies) with 4 μg/ml heparin and 0.48 μg/ml hydrocortisone, and plated into each well of 6-well plates. Media were replaced every other day, and mammospheres were harvested after 7 days and processed for immunostaining/imaging (spheroids in suspension or paraffin-embedded/sectioned 129 ) or FACS analysis (see below). For the mammosphere formation efficiency assay previously described [119][120][121][122][123][124] , the starting cell density was 10,000 cells per well of 48 well plate (n = 4), and was serially diluted to the level <100 cells per well (10,000, 5000, 2500, 1250, 625, 312, 156, 78; spheres were countable when seeded at the densities ≥1250/48-well for all the treatment groups). After one week of culturing, microscopy images were taken on phase I filter, and spheres larger than 75 μM in diameter were counted using ImageJ. The data are presented as both the number of spheroids formed at each seeding number (Fig. 7C) and mammosphere forming efficiency [(%) = (# of spheroids)/(# of cells seeded) × 100] for each seeding density ( Supplementary Fig. 7B) [119][120][121][122][123][124] . For limiting dilution analysis previously described 126 , the starting cell density was 1,600 cells per well of 96 well plate (n = 8), and was serially diluted to 200 cells per well. Each well was examined under light microscope and scored as either " + " (spheres present) or "−" (no spheres present). The fraction of "−" wells (1 − P) was calculated as the number of "−" wells/total wells for each condition. Ln (fraction of "−" wells) was plotted in the Y-axis against the number cells seeded in the X-axis, where the of best fit was drown to intersect at the origin. The sphere forming frequency corresponds to −(slope) of the line (see the equation below 126 ). www.nature.com/scientificreports www.nature.com/scientificreports/