We have created an innovative isocaloric diet approach allowing us to determine the specific effects of cholesterol in mice, by generating 4 murine cohorts, each with a different serum cholesterol level (≈140, 160, 180 & 200 mg/dL; ). As we have shown 
, this approach permits us to study hyper- and hypocholesterolemia simultaneously without affecting liver function, insulin levels, animal weight, or circulating steroid hormone levels. The basic design of our cholesterol-targeting approach combines a diet regimen with a pharmaceutical agent, ezetimibe (Zetia), an FDA approved drug that specifically blocks cholesterol uptake in the intestine, thereby lowering serum cholesterol levels. In this scheme we use a low fat/no cholesterol diet (LFNC) and a high fat/high cholesterol diet (HFHC) (w/o sodium cholate
) ± ezetimibe (30 mg/kg/d added to powdered food). Ezetimibe is a specific antagonist of NPC1L1, the bona fide
gut cholesterol transporter 
. Ezetimibe has no known target other than NPC1L1, and NPC1L1 is expressed only in the intestine and in hepatocytes (in humans), but not by tumor cells.
Serum cholesterol and testosterone levels in murine cohorts receiving cholesterol targeted treatment.
As we previously reported 
mice fed the HFHC diet developed larger tumors (weight and volume), whereas mice fed the LFNC diet had smaller tumors. The addition of ezetimibe to either diet reduced tumor size. The combination of the LFNC diet+ezetimibe had the most significant effect on tumor growth (vs. the HFHC diet w/o ezetimibe). Statistical evaluation demonstrated that both the diet (p
0.048) and ezetimibe (p
0.035) produced significant independent, additive, but not synergistic, effects on tumor growth (both volume and weight). Serology showed no liver dysfunction, no statistical differences in triglyceride (TG) levels (trending higher in the HFHC cohort), & no statistical differences in insulin or IGF-1 levels (not shown). Importantly, the HFHC diet did not increase serum T levels over the LFNC diet (), as expected, since circulating cholesterol does not influence serum androgen levels 
Our prior analysis only concerned whether HFHC and LFNC diets ± ezetimibe were associated with tumor growth, not whether tumor growth was a function of circulating cholesterol, per se. In our current analysis we unbound the groups and determined more precisely the relationship between circulating cholesterol level and tumor characteristics, as cholesterol level does not correspond directly to diet group, i.e. while HFHC raised cholesterol levels for the whole group, not every animal was in the highest cholesterol quartile, and similarly, while the LFNC diet+ezetimibe diet cohort had the lowest cholesterol level as a group not every animal in the cohort was in the lowest cholesterol quartile. Thus, while the HFHC cohort corresponds well to the highest quartile of cholesterol (9/10 were in Q4), the other quartiles contain a mixture of animals/tumors from the other 3 treatment groups (data not shown).
To determine the relationship between circulating cholesterol levels and tumor growth we analyzed tumor data from the 4 cohorts in aggregate, and plotted tumor size (weight and volume) vs. serum cholesterol level. Both tumor weight () and volume () were significantly larger in animals with levels of circulating cholesterol above vs. below the median, and demonstrated significant linear trends over the quartiles of cholesterol (, respectively). For example, tumor weight ranged from a mean of 0.79±0.3 gm in tumors in the lowest quartile of serum cholesterol (Q1) to 1.32±0.6 gm in the highest quartile (Q4; p
0.0057 for linear trend), consistent with the hypothesis that higher cholesterol levels promote tumor growth.
Growth of LNCaP xenograft tumors in relation to serum cholesterol levels.
Although modulating serum cholesterol did not alter circulating T levels (), LNCaP tumor cells express all the enzymes required to synthesize androgens from cholesterol (i.e. de novo
steroidogenesis), giving us the opportunity to determine if circulating cholesterol levels influence intratumoral androgen levels. Snap frozen samples from 52 LNCaP tumors (~n
13 in each treatment group) were analyzed by MS for levels of T and DHT. Overall, mean tumor T levels were 2.71±1.66 pg/mg (range 0.44–7.39 pg/mg), and mean DHT levels were 0.61±0.16 pg/mg (range 0.25–0.96 pg/mg). This ratio of T to DHT is similar to that observed by Locke et al in a previous study of castration sensitive LNCaP tumors prior to castration 
To explore the relationship between cholesterol, tumor androgens, and tumor growth we analyzed all the tumor data in aggregate, and plotted tumor T vs. tumor weight or tumor cholesterol. As demonstrated in , mean tumor T levels were higher in animals with circulating cholesterol levels above compared to below the median (; 2.1±1.2 vs. 3.4±1.8, p
0.0047), as well as higher in the larger tumors (; 2.2±1.4 vs. 3.3±1.8, p
0.025). Similarly, tumor T levels demonstrated significant linear trends over the quartiles of both serum cholesterol (; p
0.0042) and tumor weight (; p
0.037). Cholesterol quartile analysis indicated that DHT levels also trended higher as cholesterol level rose, but did not reach statistical significance (Figure S1
). Notably, linear regression demonstrated significant correlations between serum cholesterol and tumor weight (, R
0.0049), serum cholesterol and tumor T (, R
0.0023), and near significance for serum cholesterol and tumor DHT (, R
Tumor testosterone levels in LNCaP xenografts in relation to serum cholesterol levels and tumor weight.
Correlation of serum cholesterol levels with tumor weight, tumors androgens and expression of CYP17A in LNCaP xenografts.
To determine whether the xenograft tumors expressed the steroidogenic genes necessary to generate T and DHT from circulating cholesterol, we profiled the tumors for expression of transcripts encoding each enzyme in the steroid bio-synthetic pathway, as well as several genes related to cholesterol uptake and mitochondrial transport (). The majority of tumors had demonstrable expression of cholesterol transport genes (), as well as the critical enzymes required for de novo steroidogenesis (). While the majority of these genes did not demonstrate a significant correlation with cholesterol level, their expression levels are consistent with values previously reported in human PCa metastases with elevated tumor androgens 
, and in PCa xenograft models 
. Of particular importance, serum cholesterol levels were significantly correlated with tumoral expression of 17 α-hydroxylase (CYP17A), the central enzyme required for de novo
synthesis of androgens from cholesterol (, R
0.0255). Serum cholesterol levels also demonstrated a significant inverse correlation with the LDLR (low density lipoprotein receptor; R
0.0009), demonstrating that although these tumors accumulated cholesterol, certain normal feedback mechanisms regulating cholesterol uptake were functional (data not shown).
Expression of steroidogenic enzymes necessary for de novo synthesis of androgens from cholesterol in LNCaP xenografts.