By reducing ambient oxygen exposure in three separate de novo
cancer models, our study shows the potential importance of oxygen in influencing the initiation of tumorigenesis. Although large human epidemiologic datasets have correlated exposure to higher altitudes with lower cancer incidence, our study provides the first controlled demonstration that lower ambient oxygen levels delay tumorigenesis, independent of other variables such as barometric pressure and ultraviolet exposure associated with high altitude 
. Our data also provide physiological evidence for the critical role of oxidative stress in increasing genomic instability, an essential driver in the multi-step process of tumorigenesis 
. Most importantly, our study provides proof of concept that it is possible to alter the time course of tumor development by the simple modulation of ambient oxygen, the essential factor for oxidative stress.
We have previously shown that increased intracellular oxygen levels caused by the disruption of mitochondrial function may contribute to increased ROS generation and genomic DNA damage 
. Thus, the current in vivo
data support the in vitro
observation that oxygen consumption via mitochondrial respiration may serve a fundamentally important function to guard against oxygen-associated genotoxicity. A previous study has reported that the antioxidant N-acetyl cysteine (NAC) can delay tumorigenesis in p53-deficient mice, underscoring the importance of p53's antioxidant activities in tumor suppression 
. Our study provides important complementary data to validate this prior observation which used a pharmacologic agent. Furthermore, the current study is unique in that through a physiological manipulation, we have demonstrated the importance of oxygen in modulating genomic instability.
Although our results show that lowering oxygen reduces genomic instability and tumorigenesis in vivo
, hypoxic conditions in various systems have also been shown to promote tumor cell growth through HIF1-α induction 
. Conversely, established cancer cells engineered to over-express myoglobin, thus containing higher intracellular oxygen and lower HIF1-α levels, exhibit a decrease in xenograft tumor growth 
. However, in our experiments tissues obtained from mice exposed to 10% oxygen did not show a significant increase in HIF1-α activity at this level of physiologically adaptable hypoxia (Figure S4
). Thus, our observations are more likely to stem from a delay in tumor initiation due to reduced genomic instability in low oxygen rather than reflect the growth activities known to be stimulated by HIF1-α in established tumor cells.
Supraphysiologic levels of oxygen, including hyperbaric conditions, have been shown to inhibit tumorigenesis 
. However, oxidative stress responses such as p53 stabilization may contribute to inhibiting tumor growth 
. As demonstrated by many elegant studies, ROS plays multiple roles depending on various factors. ROS as a signaling molecule is necessary for cell proliferation 
, but at higher levels it can also induce cell death and prevent tumor growth 
. Together with these observations, our results reveal the complexity of tumorigenesis and the dual nature of oxygen. Oxygen can promote cancer initiation by increasing genomic instability through oxidative stress while also having the potential to inhibit the growth of established tumor cells through specific signaling mechanisms 
The lessons from our current study may be applicable to human health. Supplemental oxygen is ubiquitously employed in clinical medicine because of its immediate benefits for energy production while the less apparent potential for genotoxicity can be neglected. Our work may provide a biological mechanism for important clinical observations such as the increased cancer risk of neonates exposed to supplemental oxygen or of babies conceived through in vitro
fertilization, which may be performed under 21% oxygen whereas the oxygen concentration in the uterus is 5- to 10-fold lower 
. Although it is difficult to control for the many variables that are associated with the requirement for oxygen therapy, our experiments reveal the potential genotoxicity of oxygen which may be more immediately relevant to the clinics. Thus, minimizing oxygen exposure at early developmental stages such as in neonates or of in vitro
fertilized oocytes prior to uterine implantation may potentially decrease the risk of cancer over a lifetime. Similarly, although antioxidants do not appear to have tumor suppressive effects in the general population 
, targeting individuals who have specific inherited cancer susceptibility syndromes due to defects in DNA repair to strategies of reducing oxygen exposure or antioxidant therapy may yield marked benefits.