Concentration-dependent Early Antivascular and Antitumor Effects of Itraconazole in Non–Small Cell Lung Cancer

Repurposing of clinically available pharmaceuticals provides an efficient and cost-effective means of bringing new treatments to patients. By bypassing animal toxicology and human dose-finding studies, this approach allows researchers and clinicians to advance their findings directly from proof-of-principle in vitro and in vivo studies to efficacy evaluations in patients. Successful examples of repurposed pharmaceutical agents include aspirin (employed as an analgesic for decades before its use to prevent and treat cardiovascular disease) and hydroxychloroquine (an antimalarial found to have immune-modulating properties for the treatment of systemic lupus erythematosis and other autoimmune conditions; ref. 1).

In oncology, several medical therapies in use for other conditions have been explored and in some cases established as anticancer agents. Among others, examples include cyclooxygenase inhibitors, such as celecoxib, for treatment of gastrointestinal and thoracic malignancies, aspirin for the prevention of colorectal cancer (PIK3CA-mutant tumors in particular), the antidiabetic agent, metformin, for multiple solid tumors, and statin therapies in gynecologic and breast cancers (2–5). In addition, the lymphocytotoxic properties of glucocorticoids have resulted in their widespread use in combination regimens for non-Hodgkin lymphoma.

Itraconazole is an orally bioavailable and well-tolerated azole antifungal in clinical use for decades for the treatment of invasive fungal infections (6). In a broad drug screen, itraconazole emerged as a potent inhibitor of the Hedgehog signaling pathway (7, 8), which serves as a critical axis in numerous cancers (9). Unlike vismodegib, a Hedgehog pathway inhibitor approved for the treatment of advanced basal cell carcinoma (BCC), itraconazole maintains efficacy against many secondary drug-resistant smoothened (SMO) mutations (8). Separately, itraconazole has demonstrated potent inhibition of endothelial cell proliferation and angiogenesis (10–12), which may account for its apparent activity in non-Hedgehog–dependent malignancies, such as prostate and lung cancers.

Clinical studies of itraconazole as an antineoplastic agent have been performed in prostate cancer, BCC, and non–small cell lung cancer (NSCLC). A study using itraconazole 600 mg/day for castrate-resistant prostate cancer resulted in decreased Hedgehog pathway activity that correlated with decreased PSA levels (13). In a preliminary phase II clinical trial, the addition of itraconazole to pemetrexed chemotherapy for second-line treatment of advanced nonsquamous NSCLC resulted in a doubling of progression-free survival and a 4-fold increase in overall survival (14). The survival difference reached statistical significance, even though the trial was closed to enrollment prematurely due to practice changes hindering accrual. In an open-label, proof-of-concept study in advanced BCC, itraconazole 400 mg daily for 1 month produced substantial decreases in Hedgehog pathway activation, proliferation, and tumor size (15).

Because the biologic effects of itraconazole in human lung cancer cannot be entirely extrapolated from preclinical models or from other human malignancies, we conducted a window-of-opportunity, preoperative study of itraconazole in patients with early-stage NSCLC to determine the mechanism of anticancer effects and identify predictors of efficacy.

This study was approved by the University of Texas (UT) Southwestern Medical Center Institutional Review Board (Dallas, TX, STU122014-038) and was registered on clinicaltrials.gov (NCT02357836). Following Department of Defense guidance, a data and safety monitoring committee comprising individuals not directly involved with the study was established. The full protocol was reviewed and approved by the U.S. Army Medical Research and Material Command, Office of Research Protections. All patients provided written informed consent and were recruited and enrolled in the Harold C. Simmons Comprehensive Cancer Center at UT Southwestern Medical Center (Dallas, TX). This study was conducted in accordance with “Declaration of Helsinki.”

Up to 15 eligible patients were planned for enrollment, with a target of 10 cases with completion of treatment and all biomarker studies. Key eligibility criteria included NSCLC planned for surgical resection, Eastern Cooperative Oncology Group performance status 0–2, adequate hepatic and renal function, ability to undergo MRI studies, and no concurrent use of medications significantly affecting the metabolism of itraconazole. Although some antiangiogenic agents (e.g., bevacizumab) have demonstrated safety concerns (hemoptysis) in squamous histology (16), we enrolled all histologic subtypes because itraconazole had an established safety profile for the treatment of lung abscesses in the setting of lung squamous cell cancer (17, 18).

Supplementary Figure S1 shows the study schema. Briefly, prior to initiation of itraconazole treatment, patients underwent dynamic contrast-enhanced () MRI, 4-mm skin punch biopsy, and collection of peripheral blood. Archival tissue from the diagnostic biopsy was obtained. Subjects then received itraconazole 600 mg orally daily (administered as 300 mg orally twice daily) with food for 7–10 days. This dose, higher than the antiangiogenic dose of itraconazole (14), has been shown to inhibit the Hedgehog pathway in human studies (13). The duration of itraconazole therapy was selected to provide sufficient drug exposure to reach near steady-state pharmacokinetics in plasma and tissues without significantly delaying surgery. It was also expected to be sufficient for target pharmacodynamics effects, as tumor Gli1 mRNA has been shown to decrease after as little as 3 days of itraconazole treatment in medulloblastoma xenografts (7). Blood samples were collected for pharmacokinetics analysis. After 7–10 days of itraconazole therapy, subjects underwent repeat DCE MRI, skin punch biopsy, and blood collection. They then resumed itraconazole for 2–3 days, followed by surgical resection. Any standard postoperative therapy and clinical surveillance were determined by patients’ treatment teams.

Blood samples with EDTA were centrifuged at 1,100 × g (relative centrifugal force) for 15 minutes at 4°C for separation of plasma and mononuclear cell layers. Plasma was stored at −70°C. Before analysis, samples were thawed overnight at 4°C and centrifuged at 1,500 × g to remove debris. Skin punch biopsy specimens were flash-frozen for RNA procurement (PCR). Research specimens from surgically resected tumors were divided, with half flash-frozen in liquid nitrogen (for pharmacokinetics analysis) and half fixed in 4% paraformaldehyde (for angiogenic marker IHC). Individuals performing and overseeing correlative studies were blinded to patient identity and clinical effects of itraconazole.

DCE MRI is an established technology to assess microvessel density (MVD) and tumor capillary permeability (19). Each consented participant was screened for renal disease either by completing a Choyke Questionnaire (20) or undergoing point-of-care creatinine testing on the day of the examination. Scans were performed on a 3.0 Tesla Phillips whole-body scanner using a three-dimensional spoiled gradient echo sequence combined with parallel imaging, that is, SENSE technique. Anatomic imaging of the lungs was obtained with breath-held axial and coronal T2-weighted half-Fourier single-shot turbo spin-echo images acquired with in-plane resolution of 1.5 × 2 mm² and slice thickness of 5 mm. These images were used to determine tumor length (i.e., maximum dimension). Because calculated tumor volume can be unreliable due to irregular tumor shape, we calculated the whole tumor volume after segmenting the tumor on each slice where the tumor was visible, using the volume tool on the open-source dicom viewer Horos v2.1.1 (21, 22). Diffusion weighted imaging and arterial spin–labeled acquisitions were obtained, but provided no valuable information (data not shown). Then, the DCE MRI acquisition plane was set in the oblique sagittal plane so that the target tumor moved along the imaging plane with patient's respiration. Images were acquired before, during, and after intravenous administration of a standard dose of 0.1 mmol/kg of gadobutrol (Bayer Healthcare) with subjects in the supine position. Patients were instructed to perform intermittent breath-holds with free breathing in between. Imaging parameters were as follows: repetition time/echo time, 2.4/1.1 milliseconds; flip angle, 10 degrees; receiver bandwidth, 1,857 Hz/pixel; field of view, 320 × 260 mm²; acquisition voxel size, 2 × 2 × 4 mm³; and dynamic temporal resolution, 2.6 seconds. A total of 90 dynamic frames were acquired. To estimate tumor vascular measures, the extended Tofts model was used on a center slice for each tumor to compute the volume transfer constant (K^trans), which reflects the efflux rate of gadolinium contrast from blood plasma into the tissue extravascular extracellular space, using PMI.

On days 5, 6, 7, or 8, plasma samples were obtained at approximately 0, 2, 3, and 4 hours following a 300 mg oral dose of itraconazole. We determined plasma and tissue levels of itraconazole and its primary metabolite, hydroxyitraconazole, using a validated ultraperformance liquid chromatography/mass spectroscopy-mass spectroscopy method (23). We considered the duration of itraconazole therapy to be sufficient to achieve steady state, based on a half-life (t_1/2) of 24–30 hours (24). Pharmacokinetics estimates [area under the time–concentration curve calculated from 0 to 4 hours (AUC_{0–4 h}) and maximum concentration (C_max)] and pharmacokinetics parameters (such as clearance, volume, and t_1/2) for itraconazole were calculated using the pharmacokinetics modeling program WinNonlin 6.1 (Pharsight Corporation). Data were expressed as the means of the relevant measure and/or parameter and interindividual variances. Additional detail regarding pharmacokinetics methodology is provided in Supplementary Materials and Methods.

As performed in earlier studies (25), we employed a commercially available kit to measure plasma cytokine and angiogenic factor levels. Kits (Human Group I, 27-plex Cytokine Panel and Human Group II, 23-plex Cytokine Panel, Bio-Rad) were used according to the manufacturer's instructions using a BioPlex 200 Machine (Bio-Rad). Percent change of cytokine levels of post-itraconazole- compared with pre-itraconazole–treated samples was plotted as a heatmap in GraphPad prism.

Tissues were fixed in 4% formalin, washed in PBS, and embedded in paraffin. Sections (5-μm thick) were cut and stained with routine hematoxylin and eosin. Tissues were deparaffinized and rehydrated. After antigen retrieval (using Dako Target Retrieval Solution, catalog no., S1699) and blocking of endogenous peroxidase activity, biotin, and nonspecific proteins with 20% Aquablock (East Coast Biologicals), tissues were incubated in primary antibody (5–10 μg/mL) overnight at 4°C.

Briefly, sections were first deparaffinized and then subjected to antigen retrieval as detailed by Sorrelle and colleagues (26). The sections were incubated with rabbit anti-human CD31 (Abcam, ab81289, dilution 1:40,000), washed, and developed with TSA detection system using OPAL 570. The sections were then incubated with mouse anti-human CD31 (Abcam, ab9498, dilution 1:250) and subsequently developed with OPAL 520. The sections were counterstained with DAPI and imaged. Images of DAPI, CD31, and CD34 were taken from the same field of view and then merged. Colocalization of CD31 and CD34 was determined using Fiji Software as reported by Sorrelle and colleagues (26).

Three fragments per tissue were mechanically homogenized in 80% MeOH. Fragments were centrifuged at 14,000 × g for 15 minutes (4°C). The supernatant was transferred to a new tube and evaporated to dryness using a SpeedVac Concentrator (Thermo Savant). Metabolites were reconstituted in 100 μL of 0.03% formic acid in LCMS-grade water, vortex-mixed, and centrifuged to remove debris. LC-MS/MS and data acquisition were performed using an AB QTRAP 5500 Liquid Chromatography/Triple Quadrupole Mass Spectrometer (Applied Biosystems SCIEX) with an injection volume of 20 μL, as described previously (27). Chromatogram review and peak area integration were performed using MultiQuant Software version 2.1 (Applied Biosystems SCIEX). The peak area for each detected metabolite was normalized against the total ion count of that sample to correct for any variations introduced by sample handling through instrument analysis. The normalized areas were used as variables for the multivariate and univariate statistical data analysis. All multivariate analyses and modeling on the normalized data were carried out using Metaboanalyst 4.0 (http://www.metaboanalyst.ca).

We analyzed frozen-treated tumor tissue and serial skin biopsies for GLI1 and PTCH1 mRNA by qPCR. Originally, we had planned to analyze serial tumor specimens as well, but the requirement for a fresh, study-specific baseline biopsy before itraconazole treatment initiation proved impractical. Total RNA was extracted using TRizol (Invitrogen) and purified with PureLink RNA Mini Kit (Invitrogen). cDNA was generated using Superscript III First Strand Synthesis System (Invitrogen). qPCR was performed using Bio-Rad CFX Real-Time Cycler and SYBR Green Master Mix (Bio-Rad). Data are presented as fold change relative to control samples using the ΔΔC_t (2^−ΔΔC_t) method, with HPRT1 as an internal control gene. Primer pairs for qPCR are: (i) GLI1: 5′-AGCCGTGCTAAAGCTCCAGT-3′ and 5′-CCCACTTTGAGAGGCCCATAG-3′; (ii) PTCH1: 5′-GGACACTCTCATCTTTTGCTG-3′ and 5′-GGTAGTCTGCTTTCTGGGT-3′; and (iii) HPRT1: 5′-GGTCAGGCAGTATAATCCAAAG-3′ and 5′-GGACTCCAGATGTTTCCAAAC-3′. For skin biopsy specimens, we compared pre- and posttreatment specimens. For tumor specimens, we compared resected itraconazole-treated lung tumor specimens with untreated NSCLC tumor specimens and normal lung samples that were obtained from the Harold C. Simmons Comprehensive Cancer Center Tissue Resource.

To determine effects of itraconazole on tumor angiogenesis, we used paired t tests or Wilcoxon signed rank tests to investigate whether there was a significant change in the values of tissue and peripheral blood samples, and imaging from pretreatment to posttreatment. To determine effects of itraconazole on the Hedgehog pathway, we used paired t tests on pre- and posttreatment skin biopsies for GLI1 and PTCH1 mRNA levels. Spearman rank correlation was computed to investigate the association between pharmacokinetic parameters and changes in tumor volume, cytokine, GM-CSF, and tumor MVD. We employed the nonlinear mixed effects model to determine the effect of itraconazole pharmacokinetics on the pharmacodynamics profile of itraconazole. As with similar biomarker studies (25, 28), due to the exploratory nature of this study, we did not adjust for multiple comparisons.

We planned to enroll up to 15 patients. A sample size of 10 patients provides 90% power to detect a 30% reduction in MVD between the pre- and post-itraconazole tissue samples, assuming a SD of 25% using a paired t test with a two-sided significance level of 0.05. We considered this as a relatively conservative estimate of effect size, as itraconazole reduces MVD by 50%–75% in preclinical lung cancer models (10). Up to 5 additional patients were planned for accrual as needed to ensure 10 matched pairs of imaging, blood, and tissue data for analysis. The planned total sample size of 15 patients was also consistent with the FDA recommendations for human pharmacokinetics studies. Specifically, the number of subjects was projected to provide (i) 20% precision (SEM) within pharmacokinetics parameters and (ii) a reasonable understanding of intraindividual variability in pharmacokinetics parameters (29).

Thirteen patients were enrolled to the study. Median age was 64 years, and 9 were female. Additional baseline case characteristics are shown in Table 1. All patients initiated itraconazole therapy. Median total duration of therapy was 12 days (range, 8–14 days). With one exception, all patients were entirely adherent to the dosing regimen of 300 mg orally twice daily. The remaining patient, misunderstanding therapy instructions, took only half of the intended dose (300 mg orally once daily). Itraconazole was well-tolerated. Reported toxicities included fatigue, nausea, and elevated transaminase levels, all of which were low grade, reversible, and manageable with standard supportive measures. No complications were reported from study-related skin biopsies, DCE MRI scans, or blood draws. There were no bleeding or thrombotic complications related to surgery.

The method validation parameters for determination of itraconazole and hydroxyitraconazole in plasma and tissue met all acceptance criteria for sensitivity, selectivity, matrix effect, linearity, accuracy, precision, recovery, dilution integrity, and stability as set forth in the FDA's May 2018 Guidance for Industry on “Bioanalytical Method Validation” (https://tinyurl.com/rvd9m9f). A representative set of chromatograms is presented in Supplementary Fig. S2.

Plasma samples were available for 11 patients and tissue samples were available for 9 patients. The determined pharmacokinetics parameters for itraconazole are presented in Supplementary Table S1A and for hydroxyitraconazole in Supplementary Table S1B. In general, the plasma concentrations of hydroxyitraconazole were higher than that of itraconazole, consistent with previous studies of orally administered itraconazole (30). The steady-state concentrations demonstrated relatively low intrapatient variability. However, we did observe relatively high interpatient variability, with pharmacokinetics parameters varying more than 6-fold across patients. It is noteworthy, a single subject (patient #5) self-administered 300 mg/day versus the intended 600 mg/day, therefore, this patient's exposure was significantly less than that of other patients. Other clinical parameters, including body mass index, body surface area, renal function, and hepatic function, did not account for the interpatient variability in itraconazole concentrations. Concentrations of itraconazole and hydroxyitraconazole for individual patients are shown in Supplementary Fig. S2. Table 2 displays significant associations between plasma and tissue concentrations of itraconazole, hydroxyitraconazole, and correlative studies. In general, tissue concentrations of itraconazole and hydroxyitraconazole trended with the plasma concentrations.

Nine patients completed pre- and post-itraconazole MRI studies. Reasons that enrolled patients did not complete paired research MRI scans included presence of metal implants, inadequate renal function, and patient preference. Supplementary Table S2 shows MRI measurements, including tumor length, volume, and K^trans. Representative examples of imaging studies are displayed in Fig. 1. We observed a wide range of changes in tumor volume (−26% to +13%) and perfusion (−29% to +59%) across cases. With the exception of one case, tumor volume and perfusion changed in the same direction (overall Spearman correlation, 0.77; P = 0.03). There was a significant association between itraconazole C_max, AUC_{0–4 h}, and tissue concentration and hydroxy-itraconazole average plasma concentration (C_ave) and tumor volume reduction (Spearman correlation, −0.71; P = 0.05; Table 2). Tissue concentration of itraconazole also significantly associated with change in tumor perfusion (ΔK^trans; Spearman correlation, −0.71; P = 0.01; Table 2).

We used CD31 and CD34 colocalization (Supplementary Fig. S3) to quantify tumor MVD because CD31 staining alone identified both endothelial cells and nonendothelial cells (data not shown). Figure 2A shows pre- and post-itraconazole MVD by individual patient. Samples for paired pre-and post-itraconazole tumor MVD were available for 10 patients. As with tumor volume and perfusion, we observed a wide range of changes in MVD, with three cases demonstrating at least 2-fold increase and six cases demonstrating at least 2-fold decrease after itraconazole treatment. Similar to the imaging parameters, MVD changes were also strongly and significantly correlated with pharmacokinetics parameters (Table 2). Specifically, the greater the plasma or tissue itraconazole levels, the greater the decrease in MVD. Changes in MVD were highly correlated with reduction in tumor volume (Spearman correlation, 0.77; P = 0.03). In addition, changes in MVD correlated positively with changes in the proangiogenic cytokines IL1β (ref. 31; Spearman correlation, 0.57; P = 0.05) and GM-CSF (ref. 32; Spearman correlation, 1.00; P < 0.001).

Cytokine changes over the course of therapy are shown in Fig. 2B. Changes in concentrations of the proangiogenic cytokines IL1β (ref. 31; Spearman correlation, 0.7; P = 0.03), G-CSF (ref. 33; Spearman correlation, 0.67; P = 0.05), and GM-CSF (ref. 32; Spearman correlation, 1.0; P < 0.001) were significantly correlated with tumor volume reduction. In addition, changes in IL1β, GM-CSF, and the antiangiogenic cytokine IFNγ (34) were significantly associated with pharmacokinetics parameters in such a way that greater itraconazole concentrations were associated with reduced proangiogenic and increased antiangiogenic factors (Table 2).

Twelve patients underwent paired skin biopsies and were analyzed for Hedgehog pathway activity. In general, itraconazole therapy did not appear to impact Hedgehog pathway activity as measured by transcription of pathway target genes, GLI1 and PTCH1. In pre- and posttreatment skin biopsies, there was no change in transcription of GLI1 (P = 0.20) and PTCH1 (P = 0.34) mRNA (Fig. 3A and B; Supplementary Fig. S4). Comparison of itraconazole-treated tumors, normal lung, and untreated NSCLC tumors from separate, unrelated sample sets demonstrated no significant decrease in GLI1 and PTCH1 mRNA expression with itraconazole treatment (Fig. 3B and C).

Resected tumors from 6 patients (11 tumor regions) underwent metabolic analysis. These were compared with seven resected untreated NSCLC tumors (eight tumor regions). Both principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) demonstrated distinct metabolic profiles between treated and untreated tumors (Fig. 4A and B). It was interesting that itraconazole impacted some of the same metabolites in both the tumor and the lung, suggesting that the drug has metabolic effects in nonmalignant tissue. Several tricarboxylic acid (TCA) cycle metabolites (citrate, succinate, glutamate, and malate) were relatively depleted in itraconazole-treated conditions. Furthermore, we observed an accumulation of some metabolites related to ketone metabolism, including l-acetyl carnitine and β-hydroxybutyrate.

Itraconazole, an antifungal agent in clinical use for more than 25 years, has demonstrated promising antitumor efficacy in patients with prostate, lung, and basal cell skin cancer (13–15). Preclinical studies have identified effects on angiogenesis and the Hedgehog pathway as potential mediators of these anticancer effects (7, 8, 10–12). To elucidate the therapeutic mechanism of itraconazole in NSCLC and to identify potential correlates of efficacy, we performed this preoperative “window-of-opportunity” trial focused on intensive biospecimen analysis. At higher than standard doses, itraconazole was well-tolerated. Although we did not discern any effects on the Hedgehog pathway, we did detect early signals of antiangiogenic and antitumor effects, evidenced by changes in tumor MVD, angiogenic cytokines, perfusion, and volume.

Most notably, we observed a strong and significant association between drug exposure and these potential benefits. Itraconazole concentrations, which varied more than 6-fold across cases, were significantly correlated with multiple imaging and tissue and cytokine parameters, with correlation coefficients ranging 0.6–1.0. The patient with the lowest drug concentrations experienced the greatest increase in tumor volume over the course of the study (patient #5; Supplementary Tables S1 and S2), whereas the patient with the highest drug concentrations achieved the greatest decrease in tumor volume (patient #10; Supplementary Tables S1 and S2). The natural history of malignancy and longstanding dose–response principles may account for these observations. In preclinical tumor models, clear proliferation of blood vessels may occur over as little as 4 days (35). Reviews and meta-analyses of oncology phase I clinical trials have shown that the clear majority of responses occur at doses ≥75% of the MTD (36, 37). We also observed that intratumoral itraconazole concentrations exceeded those in plasma in all cases. As expected, levels of the active metabolite, hydroxyitraconazole, exceeded that of the parent drug.

The plasma itraconazole levels on days 6 and 11 of our study represent approximately 80% and 95% of steady-state levels (38), and thus, the large variance in plasma and tumor concentrations is unlikely to be explained by the inability to reach steady-state levels of itraconazole by the end of the patients’ study period. However, despite the low number of patients and large variance of itraconazole levels, our study identified several significant biological correlates with itraconazole levels. Future trials with larger cohorts of patients and longer duration of therapy may identify greater numbers of significant biological correlates and further elucidate itraconazole biology in NSCLC.

Reasons for the wide range in pharmacokinetics parameters are not clear. Apart from a single case, in which a patient misunderstood dosing directions and, therefore, ingested one-half the intended daily dose over the course of the study, dosing diaries and pill counts suggested 100% adherence. Furthermore, patient size, renal function, and hepatic function were not associated with itraconazole concentrations. To our knowledge, patients were not taking concurrent medications known to interact with itraconazole. Nevertheless, unpredictable and heterogeneous itraconazole pharmacokinetics parameters may not be entirely unexpected. In contrast to fluconazole, echinocandins, and polyene antifungals, there is strong evidence supporting the recommendations in favor of monitoring itraconazole levels in both prophylactic and therapeutic contexts. These recommendations for itraconazole reflect clinically meaningful drug exposure–response relationships (39, 40) and substantial inherent pharmacokinetics variability, mediated, in part, by food and gastric pH effects on oral bioavailability of itraconazole (38), as well as potent drug–drug interactions via effects on CYP3A4 (41) PGP1 (41, 42), and protein glycosylation (43, 44).

The wide range of itraconazole concentrations may also account for the lack of Hedgehog pathway inhibition in skin biopsies and tumors (Fig. 3; Supplementary Fig. S4). Only patients 7 and 10 reached plasma itraconazole levels to those reported on 400 mg daily after 15 days (https://tinyurl.com/ya64msm4) despite taking 600 mg daily. Higher itraconazole levels are required to inhibit the pathway in in vivo preclinical studies (7). Furthermore, our study treatment period was also significantly shorter than BCC (15) and prostate cancer (13) trials that treated patients for 4–24 weeks and reported Hedgehog pathway inhibition. Biologic effects on target pathways may lag substantially behind pharmacokinetics steady state as has been shown in studies of EGFR inhibitors (45). Thus, the longer duration of therapy of the latter two studies may have allowed a significant biological effect of the Hedgehog pathway in contrast to our study. Finally, GLI1 activation by other pathways that are independent of the Hedgehog pathway in NSCLC (46, 47) may account for lack of tumor activity by itraconazole.

In contrast to bevacizumab and ramucirumab, antibodies approved for NSCLC that target the VEGF ligand and receptor, respectively, itraconazole appears to affect multiple antiangiogenic pathways. Preclinical studies have demonstrated effects on sterol biosynthesis, disruption of cholesterol trafficking, induction of HIF1α, and inhibition of VEGFR2 and mTOR pathways (10, 11, 43). This clinical trial also suggests broad effects of itraconazole, as we observed exposure-dependent changes in multiple cytokines relevant to vascular biology, including IL1β, IFNγ, G-CSF, and GM-CSF (31–34), echoing preclinical effects of itraconazole (48, 49). Recently, antiangiogenic agents have received renewed attention in thoracic oncology. In numerous studies, the addition of anti-VEGF or -VEGFR2 therapies to EGFR inhibitors in EGFR-mutant NSCLC significantly prolongs disease control (50, 51). Furthermore, tumor angiogenesis may have relevance to cancer immunotherapy through effects on T-cell infiltration into the tumor microenvironment and immune regulatory cell function (52).

Our metabolomic assessment of a subset of these patients was hypothesis neutral because the metabolic effects of itraconazole in lung cancer are unknown. Despite the small cohort, we observed several metabolic differences in itraconazole-treated patients that may indicate a change in the mechanisms of energy formation. The TCA cycle is a central metabolic pathway that allows cells to oxidize metabolic fuels to produce biosynthetic intermediates and energy. Although metabolic activity cannot be inferred directly from metabolite abundance, the relative depletion of several TCA cycle intermediates in itraconazole-treated patients suggests that this pathway functions differently in the presence of the drug. Because NSCLCs use abundant plasma-borne fuels like glucose and lactate to supply the TCA cycle (53, 54), the depletion of TCA cycle intermediates could reflect a reduction in the availability of such fuels. The increased abundance of metabolites related to ketone utilization, a pathway that provides alternate fuels when carbohydrate metabolism is impaired, is also consistent with a change in central metabolism in the presence of itraconazole. Interestingly, several of itraconazole's metabolic effects, including its effects on the TCA cycle and ketone-related metabolites, were observed in both the tumor and the adjacent lung, suggesting that they resulted from systemic effects rather than interruption in oncogenic signaling in the tumor. Determining whether these metabolic changes are related to therapeutic responses will require further study in preclinical models and a larger placebo-controlled clinical study.

Preoperative, phase 0 biomarker studies provide an optimal setting to elucidate pharmacodynamic effects of and predictive biomarkers for emerging cancer therapies. Such studies are particularly essential in lung cancer, where real and perceived challenges to tissue acquisition may hinder implementation of essential biocorrelatives (55). We designed this trial to provide adequate exposure to a novel therapeutic agent, without subjecting patients to a prolonged delay before definitive surgery. The 10–14 days of itraconazole, selected to achieve near steady-state pharmacokinetics, is considerably less than the duration of preoperative gefitinib and pazopanib (up to 42 days) in previously published preoperative lung cancer studies, both of which were administered to molecularly unselected populations where clinical effects are unknown (25, 56). Furthermore, the entire study duration fit well within standard diagnostic and treatment intervals for early-stage lung cancer (57).

Despite these favorable characteristics, we faced challenges performing the intended biomarker studies. For various reasons, including ineligibility for particular studies (e.g., renal dysfunction or potential metal implants for research MRI scans) and patient preference, only 9 of the 13 enrolled patients completed all planned biomarker studies. Because of logistical considerations, we eliminated baseline study-specific fresh tumor biopsies. Accordingly, we were not able to perform serial assessment of Hedgehog pathway components in tumor tissue. Whether the Hedgehog analysis in skin biopsies truly represented intratumoral Hedgehog pathway status in this study cannot be determined. However, lack of a fresh pretreatment biopsy did not affect evaluation of pharmacokinetics or angiogenic parameters. In addition, in this preoperative clinical trial, patients did not receive any intervening therapy in between their diagnostic biopsy and the initiation of itraconazole. We would, therefore, expect the diagnostic biopsy to represent accurately the tumor biology at the time study therapy was started. This study design also precluded analysis of tumor metabolomics in paired pre- and post-itraconazole samples. Although we did analyze tumor MVD on archival tumor biopsies, permitting comparison with resected specimens, but these small specimens may not be representative of the larger tumor. While the short duration of study treatment was logistically appealing to patients and providers, it may also pose limitations. Longer therapy may have resulted in greater effects on biomarker endpoints, or may have unearthed other toxicities of high-dose itraconazole. Finally, we recognize the exploratory nature of our findings, which require confirmation in larger clinical trials and biomarker studies.

In conclusion, short-course, high-dose itraconazole exhibited evidence of early antivascular and antitumor effects that were highly correlated with drug exposure. While reasons for the marked interpatient variation in itraconazole concentrations observed in this study remain unclear, this observation is consistent with experience in infectious disease settings, where pharmacokinetics variability and clinically meaningful drug exposure–response relationships have led to recommendations for routine monitoring of itraconazole levels. Given the relatively low cost of and accessibility to itraconazole (less than $1,000/month, compared with more than $8,000/month for approved antiangiogenic agents and more than $7,000/month for approved Hedgehog antagonists; refs. 58–60) and growing interest in antiangiogenic therapies across multiple cancer types, further studies incorporating real-time drug monitoring and dose titration are warranted. More broadly, as the number of fixed dose cancer therapies increases, attention to interpatient pharmacokinetics and pharmacodynamics differences may represent important considerations.

D.E. Gerber reports grants from Karyopharm (funding for investigator-initiated trial), AstraZeneca (funding for investigator-initiated trial), and BerGenBio (funding for investigator-initiated trial), personal fees from Catalyst (consulting), and other from Gilead (stock ownership) outside the submitted work. F.J. Fattah reports grants from Department of Defense Lung Cancer Research Program (W81XWH-14-1-0540) during the conduct of the study. I. Pedrosa reports personal fees from Bayer Healthcare (honorarium, scientific advisory board) and other from Philips Healthcare (institutional research agreement), Health Tech International (stock options, scientific advisor), and Philips Healthcare (coinventor of patents, no royalties) outside the submitted work. J.M. Saltarski reports grants from Department of Defense during the conduct of the study. C. Padmanabhan reports grants from Department of Defense Lung Cancer Research Program (W81XWH-14-1-0540) during the conduct of the study. V. Chemburkar reports grants from Department of Defense Lung Cancer Research Program (W81XWH-14-1-0540) during the conduct of the study. B. Faubert reports grants from Canadian Institutes of Health Research (MFE 140911) during the conduct of the study. R.J. DeBerardinis reports personal fees from Agios outside the submitted work. J. Kim reports grants from Department of Defense, NCI, and American Cancer Society during the conduct of the study and outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

The views expressed in this article should not be construed to reflect the views or policies of the FDA.

D.E. Gerber: Conceptualization, resources, supervision, funding acquisition, validation, investigation, writing-original draft, project administration, writing-review and editing. W.C. Putnam: Data curation, formal analysis, investigation, methodology, writing-review and editing. F.J. Fattah: Data curation, software, project administration. K.H. Kernstine: Supervision. R.A. Brekken: Supervision. I. Pedrosa: Software, formal analysis, supervision. R. Skelton: Project administration. J.M. Saltarski: Project administration. R.E. Lenkinski: Software, formal analysis, supervision. R.D. Leff: Supervision, validation, methodology. C. Ahn: Data curation, software, formal analysis. C. Padmanabhan: Project administration. V. Chemburkar: Project administration. S. Kasiri: Formal analysis, investigation, project administration. R.R. Kallem: Formal analysis, validation. I. Subramaniyan: Formal analysis, validation. Q. Yuan: Software, formal analysis, validation, investigation. Q.N. Do: Formal analysis, validation, investigation. Y. Xi: Investigation. S.I. Reznik: Supervision. L. Pelosof: Supervision, funding acquisition. B. Faubert: Data curation, software, investigation. R.J. DeBerardinis: Supervision. J. Kim: Resources, data curation, supervision, funding acquisition, writing-original draft, writing-review and editing.

The authors thank Ms. Dru Gray for assistance with article preparation and Steven Wright for assistance with IHC. This study was funded by the Department of Defense Lung Cancer Research Program grants (W81XWH-14-1-0540 to D.E. Gerber and W81XWH-14-1-0338 to J. Kim), NCI grants (K24CA201543-01 to D.E. Gerber and 1R01CA196851 to J. Kim), American Cancer Society grant (RSG-16-090-01-TBG to J. Kim), Cancer Prevention Research Institute of Texas Core Facilities Support Award (RP170003 to W.C. Putnam), Canadian Institutes of Health Research (MFE 140911 to B. Faubert), NCI grant (R35CA22044901 to R.J. DeBerardinis), Cancer Prevention Research Institute of Texas (RP160089 to R.J. DeBerardinis), and the National Heart, Lung, and Blood Institute (5T32HL098040 to S. Kasiri). Additional support was provided by the Biomarker Research Core, Tissue Resource, and Biostatistics Shared Resource in the Harold C. Simmons Comprehensive Cancer Center (P30 CA142543, to principal investigator Carlos Arteaga, MD).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.