24-Hydroxycholesterol participates in pancreaticneuroendocrine tumor developmentMatias Soncinia,1,2, Gianfranca Cornaa,1, Marta Morescoa, Nadia Coltellab, Umberto Restucciac, Daniela Maggionia,Laura Raccostaa, Chin-Yo Lind, Francesca Invernizzie, Roberto Crocchiolof, Claudio Doglionie,g, Catia Traversarif,Angela Bachic, Rosa Bernardib, Claudio Bordignonf,g,2, Jan-Åke Gustafssond,2, and Vincenzo Russoa,2

aImmuno-Biotherapy of Melanoma and Solid Tumors Unit, Division of Experimental Oncology, Istituto di Ricerca e Cura a Carattere Scientifico (IRCCS)Ospedale San Raffaele, Milan 20132, Italy; bDivision of Experimental Oncology, IRCCS Scientific Institute San Raffaele, Milan 20132, Italy; cFunctionalProteomics Program, Istituto FIRC di Oncologia Molecolare (IFOM), Milan 20132, Italy; dCenter for Nuclear Receptors and Cell Signaling, Department ofBiology and Biochemistry, University of Houston, Houston, TX 77004; eDepartment of Pathology, Scientific Institute San Raffaele, Milan 20132, Italy;fMolMed S.p.A., Milan 20132, Italy; and gUniversità Vita-Salute San Raffaele, Milan 20132, Italy

Contributed by Jan-Åke Gustafsson, August 16, 2016 (sent for review August 9, 2016; reviewed by Vincenzo Bronte and Antonio Sica)

Cells in the tumor microenvironment may be reprogrammed bytumor-derived metabolites. Cholesterol-oxidized products, namelyoxysterols, have been shown to favor tumor growth directly bypromoting tumor cell growth and indirectly by dampening anti-tumor immune responses. However, the cellular and molecularmechanisms governing oxysterol generation within tumor micro-environments remain elusive. We recently showed that tumor-derived oxysterols recruit neutrophils endowed with protumoralactivities, such as neoangiogenesis. Here, we show that hypoxiainducible factor-1a (HIF-1α) controls the overexpression of the en-zyme Cyp46a1, which generates the oxysterol 24-hydroxycholes-terol (24S-HC) in a pancreatic neuroendocrine tumor (pNET) modelcommonly used to study neoangiogenesis. The activation of theHIF-1α–24S-HC axis ultimately leads to the induction of the angio-genic switch through the positioning of proangiogenic neutrophilsin proximity to Cyp46a1+ islets. Pharmacologic blockade or geneticinactivation of oxysterols controls pNET tumorigenesis by damp-ening the 24S-HC–neutrophil axis. Finally, we show that in somehuman pNET samples Cyp46a1 transcripts are overexpressed,which correlate with the HIF-1α target VEGF and with tumor di-ameter. This study reveals a layer in the angiogenic switch ofpNETs and identifies a therapeutic target for pNET patients.

oxysterols | HIF-1α | pancreatic neuroendocrine tumors | angiogenic switch |neutrophils

Recent studies have highlighted the diversity of metabolicpathways altered between normal and tumor cells (1, 2). Acti-

vation of specific metabolic pathways within tumors is believed toderive from an intricate connection among intrinsic and extrinsicfactors, such as oncogenic signaling, stromal-derived molecules, andhypoxia (3). Tumor hypoxia and hypoxia inducible factor-1a (HIF-1α) activation have been linked to increased glucose metabolismand cancer progression in a number of tumor types (4). WhetherHIF-1α signaling regulates other metabolic products in tumor cells orduring tumorigenesis remains only partially understood.The differential regulation of tumor metabolism and the rela-

tive abundance of some tumor-derived metabolites have also beenshown to condition the tumor microenvironment, with particularemphasis on immune cell components (5). For example, metabolicproducts like pyruvic acid and lactic acid induce hypoxia-independent stabilization of HIF-1α in tumor-associated macro-phages (6). These products, especially lactic acid, are products ofthe so-called Warburg effect (aerobic glycolysis) (7) and mainlyrequire the enzymatic activity of the pyruvate kinase M2 (PKM2),an isoform expressed by tumor cells and associated with the pro-duction of high amounts of pyruvate and lactate (8). More recentlycholesterol metabolism, oxysterols, and liver X receptors (LXRs)have been shown to be important players in tumor metabolism(9, 10), due to their dual involvement in tumor and immune cellbiology (11–13). This dual involvement makes the LXR/oxysterolaxis an attractive target for tumor therapies. Whether and how the

LXR/oxysterol axis is governed by tissue determinants presentwithin tumor microenvironments, such as hypoxia and tumor-specific metabolic regulations, remain elusive.We recently showed that oxysterols recruit protumor neutro-

phils within the tumor microenvironment in an LXR-independent,CXCR2-dependent manner (14). Tumor-recruited neutrophils areendowed with proangiogenic activities, as they secrete MMP9 andBv8 proangiogenic factors (14). We asked whether this axis wasalso active in a model of spontaneous pancreatic neuroendocrinetumor (pNET), i.e., the rat insulin promoter 1–T-antigen 2 (RIP1-Tag2) model (15), which is commonly used to investigate neo-angiogenesis and to test the effectiveness of anti-angiogenictherapies (16). RIP1-Tag2 transgenic mice develop pancreaticβ-cell tumors through the progressive transformation of beta cellislets from hyperplastic toward angiogenic islets. Then, a smallfraction of angiogenic islets progresses to tumors (15, 17). MMP9+

myeloid cells sustain the angiogenic switch in this tumor model(18). Accordingly, the depletion of neutrophils releasing theproangiogenic factors MMP9 and Bv8 in RIP1-Tag2 mice greatlyreduces the number of angiogenic islets (19, 20). In the present

Significance

Oxysterols promote tumor growth directly or through thedampening of tumor-infiltrating immune cells. Whether oxy-sterols contribute to pancreatic neuroendocrine tumor (pNET)development and how they are generated within the pNETmicroenvironment are currently unknown. Here, we show thatthe 24S-hydroxycholesterol (24S-HC) oxysterol-generatingenzyme Cyp46a1 is overexpressed during the angiogenicswitch in rat insulin promoter 1–T-antigen 2 (RIP1-Tag2) pNETformation. Moreover, we report that Cyp46a1 overexpressionrequires hypoxia inducible factor-1a (HIF-1α). Importantly, weshow that pharmacologic blockade and genetic inactivation of24S-HC delays angiogenic switch and therefore tumor formationin RIP1-Tag2. Overexpression of Cyp46a1 transcripts in somehuman pNET samples suggests that targeting this axis in pa-tients affected by pancreatic neuroendocrine tumors may be aneffective therapeutic strategy.

Author contributions: M.S., G.C., C.T., C.B., J.-Å.G., and V.R. designed research; M.S., G.C.,M.M., N.C., U.R., D.M., L.R., C.-Y.L., F.I., C.D., A.B., and R.B. performed research; N.C., U.R.,R.C., A.B., and R.B. contributed new reagents/analytic tools; M.S., G.C., J.-Å.G., and V.R.analyzed data; and M.S., G.C., J.-Å.G., and V.R. wrote the paper.

Reviewers: V.B., Istituto Oncologico Veneto; and A.S., University of Piemonte Orientale“A. Avogadro” and Humanitas Clinical and Research Center.

The authors declare no conflict of interest.1M.S. and G.C. contributed equally to this work.2To whom correspondence may be addressed. Email: soncini.matias@gmail.com,bordignon.claudio@hsr.it, jgustafsson@uh.edu, or russo.vincenzo@hsr.it.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1613332113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1613332113 PNAS | Published online September 26, 2016 | E6219–E6227

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

study, we investigated whether the neutrophil-dependent angio-genic switch occurring during RIP1-Tag2 pNET formation wasdependent on oxysterols and attempted to define tissue conditionsregulating oxysterol generation.

ResultsExpression of Cholesterol Hydroxylases During pNET Tumorigenesis.Oxysterols can be generated through autoxidation, by means of re-active oxygen species and through the activity of specific enzymessuch as cholesterol 24-hydroxylase (Cyp46a1), cholesterol 27-hydroxylase (Cyp27a1), cholesterol 25-hydroxylase (Ch25h), Cyp7b1,Cyp3a4, and Cyp11a1 (21, 22). To determine the potential in-volvement of cholesterol hydroxylases in pNET, we first analyzed byquantitative PCR (qPCR) the expression of their transcripts at vari-ous stages of pNET tumorigenesis. Ch25h, Cyp27a1, and steroido-genic acute regulatory protein (StarD1) transcripts (the latterencoding a cellular cholesterol transporter) were expressed by WT

islets and by hyperplastic, angiogenic, and tumor islets at similar levels(Fig. 1A). Cyp11a1 and Cyp7b1 transcripts slightly decreased inall tumorigenic phases compared with WT islets (Fig. 1B), whereasthe Cyp46a1 transcript, whose product forms 24S-hydroxycholesterol(24S-HC), was significantly and increasingly up-regulated duringpNET tumorigenesis (Fig. 1B). The Cyp46a1 transcript was mainlyexpressed by the CD45− tumor fraction, which includes bona fidetumor cells, an observation in agreement with Cyp46a1 mRNA ex-pression by βTC3, a tumor cell line arising from RIP1-Tag2 insuli-nomas (Fig. 1C and see Fig. 5A) (23). We found a statisticallysignificant association between Cyp46a1 and Abcg1 mRNA expres-sion during pNET tumorigenesis (Fig. 1D), the latter being a targetgene of oxysterol-engaged LXRs. In agreement with Cyp46a1 tran-script up-regulation, we also observed overexpression of the Cyp46a1protein in hyperplastic islets from RIP1-Tag2 mice compared withWT islets from age-matched controls (Fig. 1E). Finally, we detecteda higher amount of 24S-HC by LC-MS analysis in RIP1-Tag2

Fig. 1. Expression of Cyp46a1 during RIP1-Tag2 tumorigenesis. (A) qPCR analysis for Ch25h, Cyp27a1, and StarD1 transcripts of WT, hyperplastic, angiogenic,and tumor islets purified from pancreata of 10-wk-old WT or RIP1-Tag2 mice. (B) qPCR analysis for Cyp11a1, Cyp7b1, and Cyp46a1 transcripts of WT, hy-perplastic, angiogenic, and tumor islets purified from pancreata of 10-wk-old WT or RIP1-Tag2 mice. Each symbol corresponds to a single mouse, and the linerepresents the mean value: **P < 0.01; ****P < 0.0001. (C) qPCR analysis for Cyp46a1mRNA of CD45+ and CD45− cells purified from tumor islets of pancreataof 10-wk-old RIP1-Tag2 mice. (D) Positive correlation between Cyp46a1 and Abcg1mRNA expression (P = 0.0203 by Pearson test). (E) Western blot analysis forCyp46a1 protein of WT and hyperplastic islets from 8-wk-old WT or RIP1-Tag2 mice. Brain was used as positive control. (F) Characterization and quantificationof oxysterols purified from 11-wk-old WT, RIP1-Tag2, and RIP1-Tag2 × RIP1-SULT2B1b double transgenic mice; *P < 0.05.

E6220 | www.pnas.org/cgi/doi/10.1073/pnas.1613332113 Soncini et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

compared with WT pancreata, although it was not statistically sig-nificant (Fig. 1F). The other oxysterols detected were negligible (Fig.1F). We have previously showed that 24S-HC oxysterol induces neu-trophil migration (14). By performing immunofluorescence analysesof pancreata from 11-wk-old RIP1-Tag2 mice, we observed Gr1+

neutrophils in close proximity to 24S-HC–producing Cyp46a1+ cells

(Fig. S1), thus indicating a preferential localization of the neutrophilsclose to the cells releasing 24S-HC oxysterol.

Pharmacologic and Genetic Inactivation of 24S-HC Oxysterol Interfereswith the Angiogenic Switch During pNET Tumorigenesis. We first ex-amined the role of 24S-HC in the tumorigenic process by treating

Fig. 2. Pharmacologic and genetic inactivation of oxysterols in RIP1-Tag2 mice inhibits angiogenic and tumor islet formation. (A) Zaragozic acid (ZA) treatmentstudy design (Prevention trial). (B) Number of angiogenic islets in RIP1-Tag2 mice treated with vehicle or ZA; P = 0.008. (C) Percentage of CD31+CD45− endothelialcells in angiogenic islets of RIP1-Tag2 mice treated with vehicle or ZA; P = 0.02. (D) CD11b+Ly6G+ neutrophils in angiogenic islets of RIP1-Tag2 mice treated withvehicle or ZA. FACS analysis of one representative experiment is reported. (E) Percentage of CD11b+Ly6G+ neutrophils in angiogenic islets of RIP1-Tag2 micetreated with vehicle or ZA; P = 0.02. (F) Number of angiogenic islets collected from pancreata of 8-wk-old RIP1-Tag2 and RIP1-Tag2 × RIP1-SULT2B1b (doubletransgenic) mice. Ns, not significant. (G) Percentage of CD11b+Ly6G+ neutrophils collected from angiogenic islets of 8-wk-old pancreata of RIP1-Tag2 and doubletransgenic mice; P = 0.003. (H) Percentage of CD31+CD45− endothelial cells in angiogenic islets collected from pancreata of 8-wk-old RIP1-Tag2 anddouble transgenic mice; P = 0.002. (I) Number of tumor islets collected from pancreata of 8-wk-old RIP1-Tag2 and double transgenic mice. Ns, not significant.(J) Number of angiogenic islets collected from pancreata of 11-wk-old RIP1-Tag2 and double transgenic mice; P = 0.031. (K) Percentage of CD11b+Ly6G+ neu-trophils collected from pancreata of 11-wk-old RIP1-Tag2 and double transgenic mice; P = 0.027. (L) Number of tumor islets collected from pancreata of 11-wk-oldRIP1-Tag2 and double transgenic mice; P = 0.03. Each symbol corresponds to a single mouse, and the line represents the mean value ± SEM.

Soncini et al. PNAS | Published online September 26, 2016 | E6221

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

mice with the cholesterol/oxysterol inhibitor zaragozic acid (ZA)(24). Five-week-old RIP1-Tag2 mice were treated with ZA (16)(Fig. 2A), and we observed a significant reduction in the numberof angiogenic islets following ZA treatment (Fig. 2B) and a lowerpercentage of CD31+CD45− endothelial cells and CD11b+Ly6G+

neutrophils (Fig. 2 C–E) compared with vehicle-treated RIP1-Tag2 mice, suggesting that the blockade of cholesterol/oxysterolformation delays the tumorigenic process due to the reduction ofangiogenic islets. Because we could not rule out the possibility thatZA might induce off-target effects ultimately leading to tumorgrowth control, we generated a transgenic mouse, in which theexpression of the mouse oxysterol-inactivating enzyme sulfo-transferase SULT2B1b (SULT2B1b) (25, 26) is driven by the ratinsulin promoter (RIP1) in pancreatic β cells (Fig. S2A). TheULT2B1b transgene was highly expressed in the pancreas ofRIP1-SULT2B1b transgenic mice compared with WT mice, asevaluated by qPCR (Fig. S2B). We then evaluated pNET tu-morigenesis in RIP1-Tag2 × RIP1-SULT2B1b double transgenicmice (hereafter referred to as double transgenic mice). Doubletransgenic mice analyzed at early time points (8 wk) showed alower number of angiogenic islets, albeit not significantly (Fig. 2F),and a lower percentage of CD31+CD45− endothelial cells andCD11b+Ly6G+ neutrophils in angiogenic islets (Fig. 2 G and H),compared with RIP1-Tag2 mice. We did not notice any significantdifference in the number of tumor islets, possibly due to the lowfrequency of tumor formation at this time point (Fig. 2I). At11 wk, we observed a significant reduction of angiogenic islets (Fig.2J), neutrophils (Fig. 2K), and tumor islets (Fig. 2L). In agreementwith these results, we observed a lower amount of 24S-HC inpancreata of double transgenic mice compared with RIP1-Tag2mice (Fig. 1F). Additionally, we did not observe any difference inthe expression of common chemoattractive and inflammatory cy-tokines between pancreata of RIP1-Tag2 and double transgenicmice (Fig. S3), strongly suggesting a predominant role of 24S-HCin the local recruitment of proangiogenic neutrophils, as evi-denced by immunofluorescence analysis showing neutrophils closeto 24S-HC–producing Cyp46a1+ cells (Fig. S1). Of note, we alsoevaluated the percentage of neutrophils in the blood of 8-wk-oldWT, RIP1-Tag2, and double transgenic mice to rule out gran-ulopoiesis promotion as a consequence of cholesterol/oxysterol

alterations (27) in transgenic and double transgenic mice. How-ever, we failed to detect any difference in the blood of the above-reported groups of mice (Fig. S4), further supporting theinvolvement of the local recruitment of neutrophils to the RIP1-Tag2 neoplastic islets.

Delayed pNET Progression in RIP1-Tag2 × RIP1-SULT2B1b DoubleTransgenic Mice Is Independent of Intrinsic Effects of SULT2B1bExpression. To rule out possible intrinsic cellular effects playedby sulfated oxysterols, we analyzed proliferation and apoptosis,evaluated as Ki67 and cleaved caspase-3 (CCL-3), an activatedform of caspase-3, respectively, in tumors from 8-wk-old RIP1-Tag2 and double transgenic mice. CCL-3 staining rates weresimilar between RIP1-Tag2 and double transgenic mice (Fig. 3A–C), whereas we observed increased proliferation in the tumorsof double transgenic mice (Fig. 3 D–F). These results are inagreement with data reporting a proliferative advantage of mousehepatocytes over-expressing SULT2B1b (28). In addition, giventhe delay of angiogenic and tumor islet formation in the doubletransgenic mice, these results suggest a predominant role of 24S-HCoxysterol on neutrophil-dependent angiogenesis in the RIP1-Tag2 model.

Tissue Determinants Involved in Cyp46a1 Overexpression During pNETTumorigenesis.We sought to identify the determinants responsiblefor the up-regulation of Cyp46a1 enzyme in transformed cells ofRIP1-Tag2 pancreata. We first investigated the role played byhypoxia in the angiogenic switch (29) but failed to detect HIF-1αprotein in pancreata, as well as in isolated pancreatic islets fromRIP1-Tag2 mice, although we used several commercially availableanti–HIF-1α antibodies and different staining methods. Becausetranscriptional regulation of HIF-1α has also been recentlyreported in tumors (30), we carried out qPCR analysis to evaluatethe expression ofHif-1α transcripts at different times during RIP1-Tag2 tumorigenesis. We observed the up-regulation of Hif-1α inhyperplastic islets of 6-wk-old RIP1-Tag2 mice (Fig. 4A). At10 wk, Hif-1α transcripts were increased in hyperplastic, as well asangiogenic and tumor, islets (Fig. 4B). Remarkably, we observedthe up-regulation of the inducible form ofHif-1α (31) (Fig. 4C), anobservation recently reported in mouse macrophages (6). HIF-1α

Fig. 3. Analysis of cell proliferation and apoptosis of tumor islets from RIP1-Tag2 mice and RIP1-Tag2 × RIP1-SULT2B1b double transgenic mice. (A and B)Cleaved caspase-3 (CCL-3) immunohistochemical analysis of pancreata from 8-wk-old RIP1-Tag2 and double transgenic mice. One representative image frompancreata of RIP1-Tag2 (A) and double transgenic mice (B) is shown. (Scale bars, 100 μm.) (C) Quantification of CCL-3+ cells in transformed islets of RIP1-Tag2(n = 6 mice) and double transgenic mice (n = 7 mice). (D and E) Ki67 immunohistochemical analysis of pancreata from 8-wk-old RIP1-Tag2 and doubletransgenic mice. One representative image from pancreata of RIP1-Tag2 (D) and double transgenic mice (E) is shown. (Scale bars, 100 μm.) (F) Quantificationof Ki67+ cells in transformed islets of RIP1-Tag2 (n = 4 mice) and double transgenic mice (n = 6 mice); *P < 0.05.

E6222 | www.pnas.org/cgi/doi/10.1073/pnas.1613332113 Soncini et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

stabilization leads to the transcriptional activation of the bona fideHIF-1α targets Vegf and the enzyme pyruvate kinase isoform M2(Pkm2) (32). Therefore, we analyzed Vegf and Pkm2 transcripts atdifferent stages of RIP1-Tag2 tumorigenesis. Vegf and Pkm2 werefound significantly up-regulated in angiogenic islets of 8-wk-oldRIP1-Tag2 mice (Fig. 4 D and E). The up-regulation of VegfmRNA was also observed in hyperplastic islets in addition toangiogenic and tumor islets of 10-wk-old RIP1-Tag2 mice (Fig.S5). Of note, the up-regulation of Pkm2 transcripts (Fig. 4E)suggests a possible relationship between the glycolytic metabolicswitch and Hif-1α up-regulation during pNET tumorigenesis (33).Because hypoxia is one of the major drivers of HIF-1α stabiliza-tion, we carried out in vivo studies with the hydroxyprobe pimo-nidazole and detected hypoxic transformed islets in pancreatafrom 8-wk-old RIP1-Tag2 mice (Fig. S6). To evaluate whetherCyp46a1 was expressed concomitantly or after Hif-1α up-regula-tion, we carried out qPCR for Cyp46a1 at different times duringpNET tumorigenesis. We failed to detect Cyp46a1 up-regulationat 6 wk (Fig. 4F), whereas it was significantly increased at latertimes (Fig. 4G).

HIF-1α Activates Cyp46a1 Overexpression in pNET-Derived Cells. Theresults reported thus far point to a direct role of HIF-1α in thetranscriptional activation of Cyp46a1 expression. To test this, weused βTC3 cells, a RIP1-Tag2–derived cell line (23). βTC3 cellsexpress higher levels of Cyp46a1 transcripts compared with NIH3T3 and other mouse tumor cell lines analyzed, such as RMAlymphoma, Lewis lung carcinoma (LLC), and 4T1 breast carci-noma (Fig. 5A). We treated βTC3 cells with an shRNA specificfor Hif-1α, as demonstrated by Western blot analysis for HIF-1α(Fig. 5B) and analyzed the expression of transcripts encodingCyp46a1 and Vegf in hypoxic culture conditions. Besides amodest decrease of Vegf expression, Hif-1α–silenced hypoxicβTC3 cells showed significantly reduced Cyp46a1 levels (Fig. 5 Band C), suggesting a direct role of HIF-1α in Cyp46a1 activation.To demonstrate a direct regulation of Cyp46a1 by HIF-1α,chromatin immunoprecipitation (ChIP) assays were performed

in βTC3 cells on expression of a mutant stable form of HIF-1α(34), as confirmed by Western blot analysis for HIF-1α (Fig.S7A) and qPCR analysis for Cyp46a1 transcripts (Fig. S7B). In-creased recruitment of HIF-1α to its putative consensus-bindingsite located in the Cyp46a1 promoter and not to a distal region(Fig. 5D) was observed in βTC3 cells stably expressing HIF-1α,compared with mock-expressing cells (Fig. 5E). Similar experi-ments with 4T1 tumor cells confirmed recruitment of HIF-1α tothe Cyp46a1 promoter in another cell context (Fig. S8).Altogether, these results indicate that in pancreatic islets un-

dergoing large T antigen-mediated transformation, HIF-1α up-regulates the expression of Cyp46a1, which in turn produces theoxysterol 24S-HC that positions neutrophils in close proximity tohypoxic areas before the occurrence of the angiogenic switch.

Expression of Cyp46a1, VEGF, and HIF-1α Transcripts in Human pNETSamples. Finally, we investigated whether Cyp46a1 was alsoexpressed in human pNETs and whether a link among Cyp46a1,VEGF, and HIF-1α expression was present in human pNETs.The distribution of Cyp46a1 expression is shown in Fig. 6A, withmedian, minimum, and maximum values being 2.137, 0.165, and117.367, respectively. As Cyp46a1 expression values in eightspecimens were far from a normal distribution, they were grou-ped together (i.e., “high expressors,” relative expression >10);remaining values were divided into two groups (<1 and ≥1)according to the relative Cyp46a1 mRNA expression above orbelow that of normal samples (P-P plot; Fig. S8A). We thenperformed linear regression analyses to evaluate the associationbetween Cyp46a1 and HIF-1α or Cyp46a1 and VEGF in theabove-reported three groups. We did not find any statistical as-sociation between Cyp46a1 and HIF-1α or Cyp46a1 and VEGF inthe <1 group (P = 0.308 and P = 0.215, respectively). We foundonly a trend in the >1 group both for HIF-1α and VEGF (Fig. S9B and C), which failed to achieve significance. We found a trendbetween Cyp46a1 and HIF-1α (Fig. 6B; R2 = 0.391, P = 0.072,n = 8) and a significant association between Cyp46a1 and VEGFin the high expressors group (Fig. 6C; R2 = 0.667, P = 0.007,

Fig. 4. Tissue determinants associated with the up-regulation of Cyp46a1 transcripts during RIP1-Tag2 tumorigensis. (A) qPCR analysis for Hif-1α mRNA ofWT and hyperplastic islets purified from pancreata of 6-wk-old WT and RIP1-Tag2 mice, respectively; **P < 0.01. (B) qPCR analysis for Hif-1α mRNA of WT,hyperplastic, angiogenic, and islets purified from pancreata of 10-wk-old WT and RIP1-Tag2 mice; *P < 0.05; **P < 0.01; ***P < 0.001. (C) qPCR analysis for theinducible form of Hif-1α mRNA of WT, hyperplastic, and angiogenic islets purified from pancreata of 8-wk-old WT and RIP1-Tag2 mice; **P < 0.01. (D) qPCRanalysis for Vegf mRNA of WT, hyperplastic, angiogenic, and islets purified from pancreata of 10-wk-old WT and RIP1-Tag2 mice; *P < 0.05; **P < 0.01;***P < 0.001. (E) qPCR analysis for Pkm2mRNA of WT, hyperplastic, and angiogenic islets purified from pancreata of 8-wk-old WT and RIP1-Tag2 mice; Ns, notsignificant; **P < 0.01. (F) qPCR analysis for Cyp46a1 transcripts of WT and hyperplastic islets purified from pancreata of 6-wk-old WT or RIP1-Tag2 mice,respectively; Ns, not significant. (G) qPCR analysis for Cyp46a1 mRNA of hyperplastic islets purified from pancreata of 6-, 8-, and 10-wk-old RIP1-Tag2 mice;*P < 0.05; ****P < 0.0001. Each symbol corresponds to a single mouse, and the line represents the mean value ± SEM.

Soncini et al. PNAS | Published online September 26, 2016 | E6223

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

n = 8). We also detected HIF-1α protein by immunohisto-chemistry in one available pNET sample in the high expressorsgroup (i.e., 0511.2993AT; Fig. 6D), thus corroborating the resultsobtained at mRNA levels. As expected, we found a significantassociation between HIF-1α and VEGF (Fig. S7D). Becausemost pNET samples overexpressing Cyp46a1 transcripts weregrade 1 (G1) (Table S1), we analyzed a possible correlationbetween tumor diameter and Cyp46a1 expression in 22 patientsaffected by G1 pNETs. We found a correlation between tumordiameter and Cyp46a1 overexpression (Fig. 6E; R2 = 0.183, P =0.042, n = 22). Altogether, these results suggest a possible role ofCyp46a1 also in human pNET tumorigenesis. In the near future,we will investigate the correlation between Cyp46a1 and HIF-1αat mRNA and protein levels on a larger cohort of pNET patientsand will define the role of Cyp46a1 overexpression in humanpNETs to establish whether this overexpression correlates withtumor progression and metastasis or with response to therapy inpNET patients.

DiscussionThe role of oxysterols and LXRs in tumor development andprogression is greatly debated due to contrasting evidence pos-

sibly related to the multiple roles played by oxysterols and LXRswithin the tumor microenvironment (9, 10, 12). The differentoutcomes observed following the treatment of tumors withmolecules stimulating or interfering with LXR/LXR ligandsshould take into account the tumor models used, the presence ofan intact immune system, the distinct isoforms of LXRs engaged,and the intrinsic biologic ability of oxysterols to bind receptorsdifferent from LXRs (12, 35). Therefore, appropriate modelsdissecting how LXR/LXR ligands operate within the tumor mi-croenvironment are required. We recently studied the functionof LXR/oxysterols during the growth of transplantable mousetumors and observed that they exert protumoral functionsthrough neutrophils recruitment and induction of an angiogenicswitch (14).Here, we analyzed the role of oxysterols in spontaneous tumor

models, in particular during the development of pancreaticneuroendocrine tumors in RIP1-Tag2 mice. We show that duringpNET tumorigenesis the transcript encoding the Cyp46a1 en-zyme is specifically up-regulated. This up-regulation occursconcomitantly to the onset of tumor hypoxia in transformed is-lets of RIP1-Tag2 pancreata and increased expression of bonafide HIF-1α target genes Vegf and Pkm2. PKM2 is activated

Fig. 5. HIF-1α activates Cyp46a1 up-regulation in insulinomas derived from RIP1-Tag2 mice. (A) qPCR analysis for Cyp46a1 mRNA of NIH 3T3, βTC3, RMA, LLC,and 4T1 cell lines. Mean ± SEM of three experiments; ***P < 0.001. (B) Western blot analysis showing a marked decrease of HIF-1α protein levels in Hif-1α–silenced-βTC3 cells compared with scrambled βTC3 cells. One of three independent experiments with similar results is shown. (C) qPCR analysis for VegfmRNAof scrambled- and Hif-1α–silenced-βTC3 cells. Mean ± SEM of three experiments; **P < 0.01. (D) qPCR analysis for Cyp46a1 mRNA of scrambled- and Hif-1α-silenced-βTC3 cells. Mean ± SEM of three experiments; ***P < 0.001. (E) A Hif-1α responsive element (HRE) in the Cyp46a1 promoter is shown. Whitearrowheads represent qPCR primers proximal to HRE region, whereas gray arrowheads represent qPCR primers located in Cyp46a1 promoter distal region(about −800 bp from transcription start). (F) Chromatin immunoprecipitation assay on βTC3 cells transduced with a vector encoding a mutant stable form ofHIF-1α. The selective enrichment of Hif-1α responsive element in the Cyp46a1 promoter of βTC3 cells expressing the mutant stable form of HIF-1α is shown.Mean ± SEM of two experiments; *P < 0.05.

E6224 | www.pnas.org/cgi/doi/10.1073/pnas.1613332113 Soncini et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

during aerobic glycolysis and regulates the production of pyru-vate and lactate, which have been described to stabilize HIF-1α through hypoxia-independent mechanisms in a self-enforcingcircle (6, 36). We find that Cyp46a1 is an HIF-1α target gene, asstabilized HIF-1α binds HREs in the Cyp46a1 promoter in apNET-derived cell line. Increased expression of Cyp46a1 leads toincreased synthesis of the oxysterol 24S-HC in transformed RIP1-Tag2 islets, concomitantly with the expression of other bona fideHIF-1α target genes involved in angiogenesis and metabolic ad-aptation. In addition, consistently with previous results, 24S-HCaccumulation leads to the positioning of neutrophils close tohypoxic regions that require formation of neo-vessels.With our work, we identify a mechanism leading to oxysterols

accumulation in tumors, and we position oxysterols synthesisdownstream HIF signaling in tumors. These findings unveil a

complex network leading to tumor angiogenesis in pancreaticneuroendocrine tumors: on one hand, HIF-1α up-regulation leadsto VEGF transcription and direct induction of neoangiogenesis;on the other hand, 24S-HC accumulation and neutrophils re-cruitment further supports neovessels formation (Fig. 7).The importance of oxysterol accumulation in this tumor model

was demonstrated by pharmacological and genetic studies. RIP1-Tag2 mice treated with the squalene synthase inhibitor ZAshowed a significant reduction of proangiogenic neutrophils andangiogenic islets, confirming the previously described link betweencholesterol metabolism and neutrophil-mediated angiogenesis(14). Moreover, the involvement of oxysterols, specifically 24S-HC,in neutrophil-dependent angiogenesis is further corroborated bythe results obtained in a transgenic mouse model in which theenzyme SULT2B1b is expressed constitutively by pancreatic islets.

Fig. 6. Expression of Cyp46a1, HIF-1α, and VEGF transcripts by human pNET samples and correlation studies. (A) qPCR analysis for the expression of Cyp46a1mRNA in human pNET samples. (B and C) Linear regression analyses between Cyp46a1 expression and HIF-1α (Β) or VEGF expression (C). (D) Immunohis-tochemistry for HIF-1α in a pNET sample in the “high expressors” group. Original magnification, 200×. (E) Linear regression analyses between Cyp46a1expression and tumor diameter.

Soncini et al. PNAS | Published online September 26, 2016 | E6225

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

The choice of this transgenic model is based upon the possibility toinactivate all of the oxysterols generated within the microenvi-ronment of RIP1-Tag2 pNET, including those produced bynonradical reactive oxygen species or by inorganic free radicalspecies (37). Double transgenic mice displayed a delay in pNETtumorigenesis development (reduced number of angiogenic andtumor islets), which was independent of intrinsic effects (apo-ptosis and/or proliferation), whereas it was correlated with thereduced percentage of proangiogenic neutrophils infiltratingtransformed islets.Angiogenic switch in the RIP1-Tag2 pNET model has been

closely correlated to MMP-9+- and Bv8-releasing neutrophils (19,20). In this context, we observed that decreased 24S-HC accu-mulation accompanied reduced numbers of neutrophils in pan-creata of double transgenic mice compared with RIP1-Tag2 mice,whereas we detected similar expression of transcripts encodingCXCR2-driven chemokines, such as Cxcl1, Cxcl2, and Cxcl5,usually involved in neutrophil migration (38). These results alongwith the observations that neutrophils are found close to cells oftransformed islets overexpressing Cyp46a1 and that 24S-HC wasshown to induce neutrophil migration in vitro (14) reinforce theidea that 24S-HC oxysterol might be crucially involved in vivo inthe positioning of proangiogenic neutrophils in the proximity ofcells releasing high amounts of 24S-HC. The concept that oxy-sterols may behave as short-range chemoattractants has recentlybeen described for the oxysterol 7α, 25-HC, which controls cellpositioning within specific areas of lymphoid organs (39). Thepositioning role of 24S-HC in physiologic conditions deservesfurther investigations.Because ZA-treated mice and double transgenic mice showed

only a delay of the tumorigenic process (reduced angiogenic andtumor islets), it is likely that other mechanisms concur to the de-velopment of pNET tumorigenesis or that compensatory mecha-nisms occur during the course of the treatments.Recent findings that some molecular and metabolic features

of human pNETs are also present in the RIP1-Tag2 modelmakes our observations clinically relevant (40). This observationis further corroborated by our preliminary results showingCyp46a1 overexpression in some human pNET samples and

suggesting a linear correlation among Cyp46a1, VEGF, and tu-mor diameter in G1 pNET patients. This model could representa useful tool to test combination therapies of drugs currentlyused in pNET patients and cholesterol-lowering compoundsendowed with a well-established pharmacologic profile (e.g.,statins) (41).Altogether, these results reveal an early mechanism underlying

the angiogenic switch responsible for the subsequent tumor stageprogression in the pNET model RIP1-Tag2. Moreover, these re-sults provide the basis for the exploitation of compounds endowedwith a similar mechanism of action for antitumor therapies.

Materials and MethodsMouse Models and Treatments. RIP1-Tag2 mice were maintained on theC57BL/6N background (Charles River). To obtain RIP1-Tag2 × RIP1-SULT2B1bdouble transgenic mice, RIP1-Tag2 males were crossed with RIP1-SULT2B1bfemales. Offsprings were tested by PCR for both the transgenes using thefollowing primers: forward, gctctgctgacatagaagaatgg; reverse, gtactcatt-catggtgactattccag. Five-week old RIP1-Tag2 mice were treated with ZA A(Sigma Aldrich) with alternating i.v. (200 μg/mouse) and i.p. (100 μg/mouse)infusions. Islet analyses were carried out after 3 wk of treatment. Animalstudies were approved by the Institutional Animal Care and Use Committeeof the San Raffaele Scientific Institute.

Generation of the RIP1-SULT2B1b Transgenic Mouse Model. RIP1-Timer plasmid(Addgene plasmid #15109, DM#285) was linearized by NotI. DsRed1-E5 cDNAwas replaced with SULT2B1b cDNA, obtained by PCR using the In-Fusion PCRcloning kit (Clontech Laboratories). Primers to clone SULT2B1b fromSULT2B1b-ΔNGFr lentiviral vector (42) were designed according to In-Fusionkit instructions. HbaI linearized plasmid was then microinjected into donorFVB pronuclei and subsequently transplanted in pseudopregnant acceptorCD1 mice. Transgenic mice were generated by our Institutional CFCM facil-ity. RIP1-SULT2B1b founders were backcrossed with C57BL/6 mice for fivegenerations. pNET analysis was performed in F4/F5 double transgenic mice.RIP1-SULT2B1b mice were typed with the following primers: RipSult FW5′-gggaatgatgtggaaaaatg-3′ and RipSultREV 5′-gcacgttgctagtgttctca-3′.

ChIP Assay. Harvested cells were fixed for 10 min at room temperature with1% formaldehyde. Cross-linking was terminated by the addition of 125 mMglycine. Cells were rinsed with PBS and centrifuged at 400 × g for 5 min at4 °C. Pellets were resuspended in 50 nM Hepes-KOH, pH 7.5, 140 mM NaCl,

Fig. 7. Schematic representation of the mechanism linking HIF-1α, 24S-HC, and neutrophils in the pNET angiogenic switch. In hyperplastic islets, HIF-1αinduces the overexpression of Cyp46a1 enzyme, which in turn produces the oxysterol 24S-HC that positions neutrophils in close proximity to hypoxic areas,thus favoring the angiogenic switch.

E6226 | www.pnas.org/cgi/doi/10.1073/pnas.1613332113 Soncini et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020

1 mM EDTA, 10% (vol/vol) glycerol, 0.5% Nonidet P-40, 0.25% Triton X-100,and leupeptin-pepstatinA-aprotinin at 5 μg/mL (pH 8.1). Nuclei were col-lected by centrifugation (3,000 × g for 5 min at 4 °C), resuspended in 10 mMTris·HCl, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, and leupeptin-pepstatinA-aprotinin at 5 μg/mL (pH 8.1), and rotated for 10 min at 4 °C. Washed nucleiwere centrifuged, resuspended in 1 mM EDTA, 10 mM Tris·HCl, 100 mMNaCl, 0.5 mM EGTA, 0.1% Na-deoxycholate, 0.5% N-sauroylsarconsin, andleupeptin-pepstatinA-aprotinin at 5 μg/mL (pH 8.1), and then sonicated togenerate DNA fragment sizes of 0.2–0.8 kb, using the Diagenode Bio-ruptor twin (20 cycles, 30 s on/off, maximum power). Samples were clearedby centrifugation at 16,100 × g for 10 min at 4 °C. Ten percent of thecleared supernatant was used as input, and the remaining volume wasimmunoprecipitated with a monoclonal antibody anti–HIF-1α (CloneH1α67; Novus Biologicals). Quantification of the precipitated DNA regionswas performed by qPCR (Table S2).

Statistical Analysis. Data are expressed as mean ± SEM and were analyzed forsignificance by ANOVA with Dunnet’s, Bonferroni’s, or Tukey’s multiplecomparison test, by ANOVA with Pearson correlation, or by Student t test.The analysis was performed with Prism software.

ACKNOWLEDGMENTS. Anti-Cyp46a1 mAb was a generous gift of Dr. DavidW. Russell (Departments of Molecular Genetics, University of Texas South-western Medical Center). We thank Elena Tiziano, Simona Porcellini, andClaudia Asperti for technical help and Lorenzo Piemonti and ValeriaCorti for providing mRNA from donor-derived pancreatic islets. Thiswork was supported from the Italian Association for Cancer ResearchGrants IG 12876 and IG 15452 (to V.R. and C.B.) and from the ItalianMinistry of Health Grant RF2009 (to V.R.). M.S. was supported by grantsfrom the Association International for Cancer Research, UK. J.-Å.G. was sup-ported by the Swedish Research Council and by grants from the Robert A.Welch Foundation (E-0004).

1. Cairns RA, Harris IS, Mak TW (2011) Regulation of cancer cell metabolism. Nat RevCancer 11(2):85–95.

2. DeBerardinis RJ, Thompson CB (2012) Cellular metabolism and disease: What dometabolic outliers teach us? Cell 148(6):1132–1144.

3. Bensaad K, Harris AL (2014) Hypoxia and metabolism in cancer. Adv Exp Med Biol 772:1–39.

4. Semenza GL (2013) HIF-1 mediates metabolic responses to intratumoral hypoxia andoncogenic mutations. J Clin Invest 123(9):3664–3671.

5. Villalba M, et al. (2013) From tumor cell metabolism to tumor immune escape. Int JBiochem Cell Biol 45(1):106–113.

6. Colegio OR, et al. (2014) Functional polarization of tumour-associated macrophagesby tumour-derived lactic acid. Nature 513(7519):559–563.

7. Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the Warburgeffect: The metabolic requirements of cell proliferation. Science 324(5930):1029–1033.

8. Christofk HR, et al. (2008) The M2 splice isoform of pyruvate kinase is important forcancer metabolism and tumour growth. Nature 452(7184):230–233.

9. Lin CY, Gustafsson JA (2015) Targeting liver X receptors in cancer therapeutics. NatRev Cancer 15(4):216–224.

10. Bovenga F, Sabbà C, Moschetta A (2015) Uncoupling nuclear receptor LXR and cho-lesterol metabolism in cancer. Cell Metab 21(4):517–526.

11. Bensinger SJ, Tontonoz P (2008) Integration of metabolism and inflammation bylipid-activated nuclear receptors. Nature 454(7203):470–477.

12. Traversari C, Sozzani S, Steffensen KR, Russo V (2014) LXR-dependent and -independenteffects of oxysterols on immunity and tumor growth. Eur J Immunol 44(7):1896–1903.

13. Spann NJ, Glass CK (2013) Sterols and oxysterols in immune cell function. NatImmunol 14(9):893–900.

14. Raccosta L, et al. (2013) The oxysterol-CXCR2 axis plays a key role in the recruitment oftumor-promoting neutrophils. J Exp Med 210(9):1711–1728.

15. Hanahan D (1985) Heritable formation of pancreatic beta-cell tumours in transgenicmice expressing recombinant insulin/simian virus 40 oncogenes. Nature 315(6015):115–122.

16. Bergers G, Javaherian K, Lo KM, Folkman J, Hanahan D (1999) Effects of angiogenesisinhibitors on multistage carcinogenesis in mice. Science 284(5415):808–812.

17. Folkman J, Watson K, Ingber D, Hanahan D (1989) Induction of angiogenesis duringthe transition from hyperplasia to neoplasia. Nature 339(6219):58–61.

18. Bergers G, et al. (2000) Matrix metalloproteinase-9 triggers the angiogenic switchduring carcinogenesis. Nat Cell Biol 2(10):737–744.

19. Nozawa H, Chiu C, Hanahan D (2006) Infiltrating neutrophils mediate the initial an-giogenic switch in a mouse model of multistage carcinogenesis. Proc Natl Acad SciUSA 103(33):12493–12498.

20. Shojaei F, Singh M, Thompson JD, Ferrara N (2008) Role of Bv8 in neutrophil-de-pendent angiogenesis in a transgenic model of cancer progression. Proc Natl Acad SciUSA 105(7):2640–2645.

21. Björkhem I (2002) Do oxysterols control cholesterol homeostasis? J Clin Invest 110(6):725–730.

22. Russell DW (2000) Oxysterol biosynthetic enzymes. Biochim Biophys Acta 1529(1-3):126–135.

23. Efrat S, et al. (1988) Beta-cell lines derived from transgenic mice expressing a hybridinsulin gene-oncogene. Proc Natl Acad Sci USA 85(23):9037–9041.

24. Bergstrom JD, et al. (1993) Zaragozic acids: A family of fungal metabolites that arepicomolar competitive inhibitors of squalene synthase. Proc Natl Acad Sci USA 90(1):80–84.

25. Fuda H, Javitt NB, Mitamura K, Ikegawa S, Strott CA (2007) Oxysterols are substratesfor cholesterol sulfotransferase. J Lipid Res 48(6):1343–1352.

26. Falany CN, Rohn-Glowacki KJ (2013) SULT2B1: Unique properties and characteristicsof a hydroxysteroid sulfotransferase family. Drug Metab Rev 45(4):388–400.

27. Drechsler M, Megens RT, van Zandvoort M, Weber C, Soehnlein O (2010) Hyperlip-idemia-triggered neutrophilia promotes early atherosclerosis. Circulation 122(18):1837–1845.

28. Ren S, Ning Y (2014) Sulfation of 25-hydroxycholesterol regulates lipid metabolism,inflammatory responses, and cell proliferation. Am J Physiol Endocrinol Metab 306(2):E123–E130.

29. Hanahan D, Folkman J (1996) Patterns and emerging mechanisms of the angiogenicswitch during tumorigenesis. Cell 86(3):353–364.

30. Görlach A (2009) Regulation of HIF-1alpha at the transcriptional level. Curr Pharm Des15(33):3844–3852.

31. Wenger RH, Rolfs A, Spielmann P, Zimmermann DR, Gassmann M (1998) Mousehypoxia-inducible factor-1alpha is encoded by two different mRNA isoforms: Ex-pression from a tissue-specific and a housekeeping-type promoter. Blood 91(9):3471–3480.

32. Luo W, et al. (2011) Pyruvate kinase M2 is a PHD3-stimulated coactivator for hypoxia-inducible factor 1. Cell 145(5):732–744.

33. Semenza GL (2010) HIF-1: Upstream and downstream of cancer metabolism. CurrOpin Genet Dev 20(1):51–56.

34. Guarnerio J, et al. (2014) Bone marrow endosteal mesenchymal progenitors dependon HIF factors for maintenance and regulation of hematopoiesis. Stem Cell Rep 2(6):794–809.

35. Cyster JG, Dang EV, Reboldi A, Yi T (2014) 25-Hydroxycholesterols in innate andadaptive immunity. Nat Rev Immunol 14(11):731–743.

36. Lu H, Forbes RA, Verma A (2002) Hypoxia-inducible factor 1 activation by aerobicglycolysis implicates the Warburg effect in carcinogenesis. J Biol Chem 277(26):23111–23115.

37. Murphy RC, Johnson KM (2008) Cholesterol, reactive oxygen species, and the for-mation of biologically active mediators. J Biol Chem 283(23):15521–15525.

38. Kim ND, Luster AD (2015) The role of tissue resident cells in neutrophil recruitment.Trends Immunol 36(9):547–555.

39. Yi T, et al. (2012) Oxysterol gradient generation by lymphoid stromal cells guidesactivated B cell movement during humoral responses. Immunity 37(3):535–548.

40. Sadanandam A, et al. (2015) A Cross-Species Analysis in Pancreatic Neuroendocrine Tu-mors Reveals Molecular Subtypes with Distinctive Clinical, Metastatic, Developmental, andMetabolic Characteristics. Cancer Discov 5(12):1296–1313.

41. Nielsen SF, Nordestgaard BG, Bojesen SE (2012) Statin use and reduced cancer-relatedmortality. N Engl J Med 367(19):1792–1802.

42. Villablanca EJ, et al. (2010) Tumor-mediated liver X receptor-alpha activation inhibitsCC chemokine receptor-7 expression on dendritic cells and dampens antitumor re-sponses. Nat Med 16(1):98–105.

43. Parangi S, Dietrich W, Christofori G, Lander ES, Hanahan D (1995) Tumor suppressorloci on mouse chromosomes 9 and 16 are lost at distinct stages of tumorigenesis in atransgenic model of islet cell carcinoma. Cancer Res 55(24):6071–6076.

44. Griffiths WJ, et al. (2013) Analytical strategies for characterization of oxysterol lip-idomes: Liver X receptor ligands in plasma. Free Radic Biol Med 59:69–84.

45. Ramirez DM, Andersson S, Russell DW (2008) Neuronal expression and subcellularlocalization of cholesterol 24-hydroxylase in the mouse brain. J Comp Neurol 507(5):1676–1693.

Soncini et al. PNAS | Published online September 26, 2016 | E6227

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Aug

ust 1

8, 2

020