Intranasal administration of the chemotherapeutic perillyl alcohol results in selective delivery to the cerebrospinal fluid in rats

Perillyl alcohol (POH) has been extensively studied for the treatment of peripheral and primary brain tumors. The intranasal route of administration has been preferred for dosing POH in early-stage clinical trials associated with promising outcomes in primary brain cancer. However, it is unclear how intranasal POH targets brain tumors in these patients. Multiple studies indicate that intranasally applied large molecules may enter the brain and cerebrospinal fluid (CSF) through direct olfactory and trigeminal nerve-associated pathways originating in the nasal mucosa that bypass the blood–brain barrier. It is unknown whether POH, a small molecule subject to extensive nasal metabolism and systemic absorption, may also undergo direct transport to brain or CSF from the nasal mucosa. Here, we compared CSF and plasma concentrations of POH and its metabolite, perillic acid (PA), following intranasal or intravascular POH application. Samples were collected over 70 min and assayed by high-performance liquid chromatography. Intranasal administration resulted in tenfold higher CSF-to-plasma ratios for POH and tenfold higher CSF levels for PA compared to equal dose intravascular administration. Our preclinical results demonstrate POH undergoes direct transport from the nasal mucosa to the CSF, a finding with potential significance for its efficacy as an intranasal chemotherapeutic for brain cancer.

The intranasal (IN) route of administration has long been appreciated as an option for local and systemic delivery for certain small molecules, peptides, and protein drugs 1,2 . In particular, the nasal route possesses distinct advantages when local effects are needed (e.g. as with decongestants, antibiotics, and mucolytics) or when non-invasive access to the systemic circulation is required for fast drug onset and/or to avoid extensive hepatic first-pass elimination (e.g. as with the application of the opioid antagonist naloxone following opioid overdose). Multiple clinical studies show that intranasally applied small molecules (e.g. zolmitriptan, sumatriptan, butorphanol tartrate, fentanyl, nicotine, and estradiol) and low molecular weight peptide drugs (e.g. calcitonin, desmopressin, buserelin, oxytocin), can achieve improved systemic exposure and therapeutically relevant concentrations in peripheral tissues 3 . Improved systemic exposure for these molecules is likely due to greater paracellular permeability across the nasal epithelia and more efficient absorption into the bloodstream through the extensive nasal vasculature present in the underlying lamina propria 1,4 . In contrast, intranasal delivery results in improved central exposure to the brain or CSF for only a subset of small molecules 1,[5][6][7][8][9][10][11][12] . Mechanistic studies indicate that intranasally applied large molecules can directly access the central nervous system (CNS) through olfactory or trigeminal nerve-associated pathways that originate within the nasal mucosa [13][14][15][16][17] . Collectively, these findings have increased clinical interest in using the intranasal route for enhanced central delivery of small molecules. One such small molecule is perillyl alcohol (POH), a 152 Da plant-derived monocyclic terpene and chemotherapeutic

Results
The experimental setup and study design are illustrated in Fig. 1. Adult female Sprague-Dawley rats with an average weight of 210 g received an equal dose of POH via intranasal (n = 5) or intravascular (n = 4) administration. The fixed dose administered to animals in both groups (4.745 mg) was chosen to represent: (i) the lowest POH dose described in previous preclinical studies and (ii) the allometrically-scaled dose corresponding to the intranasal POH formulation from an ongoing clinical trial examining its safety and efficacy in glioma patients (NCT02704858) 29,[35][36][37][38] (see "Methods" for further description). The intranasal POH formulation was directed to a region near the olfactory mucosa to increase the likelihood of CNS delivery since the olfactory mucosa has relatively lower vascular permeability and mucociliary clearance compared to the nasal respiratory mucosa 4,13 . CSF and plasma samples were processed to precipitate residual proteins (e.g. albumin) and to allow for total  www.nature.com/scientificreports/ (bound and unbound) POH in each compartment to be analyzed. CSF concentration of POH was below the lower limit of detection (LLOD) for the assay in 1 out of 5 animals in the intravascular POH cohort; these values were therefore excluded from data analysis. POH and PA concentrations in the plasma and CSF were above LLOD for all other cases (Figs. 2, 3). Table 1 indicates AUC 70 values for this study, i.e. total analyte concentrations over 70 min of POH administration. Tables 2 and 3 summarize mean concentrations and CSF-to-plasma ratios. These results are also discussed in the following sections.
Plasma and CSF exposure for perillyl alcohol and perillic acid following intravascular application. Intravascular application of POH to female Sprague-Dawley rats (n = 4) resulted in a mean POH plasma concentration of 33.11 ± 15.78 µM ( Fig. 2b; Table 2) and a mean POH plasma AUC 70 of 2317.7 ± 1104.6 µM.min (Table 1). CSF exposure for POH in these animals was approximately 200-fold-lower than plasma POH exposure, with a mean POH CSF concentration of 0.16 ± 0.03 µM ( Fig. 2a; Table 2) and a mean POH CSF AUC 70 value of 11.2 ± 2.1 µM.min (Table 1). The mean CSF-to-plasma ratio for POH was 0.02 ± 0.01 following intravascular application ( Fig. 4b; Table 2). Apart from POH, perillic acid (PA), the major POH metabolite, was also assessed in both CSF and plasma taken from the same animals. Intravascularly applied POH resulted in a mean PA plasma concentration of 2.37 ± 0.27 µM ( Fig. 3b; Table 2) and mean PA plasma AUC70 value of 165.9 ± 18.9 µM.min (Table 1). CSF PA exposure was approximately fourfold lower than plasma PA exposure, with a mean PA CSF concentration of 0.51 ± 0.10 µM ( Fig. 3a; Table 2) and a mean PA CSF AUC 70 value of 35.7 ± 7.0 µM.min (Table 1). Mean CSF-toplasma ratio for PA was 0.21 ± 0.02 in this group ( Fig. 5b; Table 3).
Enhanced CSF exposure for perillyl alcohol and metabolite following intranasal application. Intranasal application of POH to female Sprague-Dawley rats (n = 5) resulted in a mean POH plasma concentration of 7.57 ± 1.77 µM ( Fig. 2b; Table 2) and a mean POH plasma AUC70 of 529.9 ± 123.9 µM.min (Table 1). CSF exposure in these animals was remarkably only about fivefold lower than plasma, with a mean POH CSF concentration of 1.49 ± 0.34 µM ( Fig. 2a; Table 2) and a mean POH CSF AUC 70 value of 104.3 ± 23.8 µM min (Table 1). These values resulted in a CSF-to-plasma ratio of 0.20 ± 0.01 ( Fig. 4a; Table 2). In other words, intranasal administration of POH resulted in a tenfold higher CSF-to-plasma ratio compared to intravascularly applied POH, consistent with a significant degree of direct targeting of POH to the CSF from the nasal passage.
Animals that received POH intranasally also had tenfold higher PA metabolite concentrations in the CSF and plasma compared to animals that received intravascular POH. Specifically, intranasal application of POH resulted www.nature.com/scientificreports/  www.nature.com/scientificreports/    Table 2) and a mean PA plasma AUC 70 value of 1674.4 ± 276.5 µM min (Table 1). These values were approximately threefold higher than mean POH plasma values in the same animals, consistent with significant nasal metabolism and subsequent PA systemic absorption from the nasal mucosa. CSF PA exposure in animals that received POH intranasally was approximately fourfold lower than plasma PA exposure, with a mean PA CSF concentration of 5.88 ± 1.14 µM ( Fig. 3a; Table 2) and a mean PA CSF AUC 70 value of 411.6 ± 79.8 µM min (Table 1). Collectively, these values suggest that intranasally applied POH is rapidly metabolized to PA, resulting in elevated PA levels in both plasma and CSF within 70 min of intranasal application. The mean CSF-to-plasma ratio for PA was 0.26 ± 0.02 in these animals ( Fig. 5a; Table 3), which was similar to the mean CSF-to-plasma ratio for PA in the intravascular group ( Fig. 5b; Table 3).

Discussion
This study has yielded several significant new findings: (i) intranasally applied POH results in a tenfold higher CSF-to-plasma ratio for POH compared to intravascular controls, (ii) intranasally applied POH results in tenfold higher PA concentrations in the CSF and plasma compared to intravascular delivery, and (iii) intranasal and intravascular application of POH resulted in similar CSF-to-plasma ratios for PA. Taken together, our results suggest intranasal POH application is more suited to obtain greater CSF exposure for POH and PA than intravascular POH application. Most importantly, our results provide new evidence supporting significant direct transport from the nasal mucosa to the CSF for POH, suggesting a strong advantage for the intranasal route in the central delivery of POH. Our group has previously described how intranasally applied macromolecules access the CNS via potential pathways originating in the nasal mucosa 2,4,13,16,39,40 . Briefly, intranasally applied molecules may transit through the olfactory and/or respiratory nasal epithelia via intracellular or extracellular pathways to reach the underlying connective tissue 2,4,13,16,39,40 . Once within the lamina propria, molecules may experience several different fates: (i) systemic absorption into the nasal vasculature to enter the systemic circulation, (ii) lymphatic absorption into draining nasal lymphatic vessels and downstream cervical lymph nodes, and/or (iii) direct transport along pathways associated with the olfactory and trigeminal nerves (perivascular or perineural) to enter the CSF and brain. More detailed mechanistic studies by our group have shown that macromolecules enter the CNS by direct transport pathways at the level of the olfactory bulbs (via olfactory pathways) or the brainstem (via trigeminal pathways) 2,4,13,16,39,40 . Subsequent widespread distribution from these regions may then occur within the perivascular spaces around cerebral blood vessels via convective transport or dispersion potentially down to the capillary level 14,39,41 .
Based on its molecular weight, POH (152 Da) belongs to a subset of non-polar small molecules that in theory may passively diffuse through the phospholipid bilayers of intact brain capillary endothelial cells to enter the CNS 42 . POH also has a high binding affinity for albumin (K d = 0.19 M) that results in a relatively low unbound drug fraction within albumin-rich fluid compartments such as plasma or CSF (unbound fraction, fu ~ 0.2-0.4) 43 . Albumin-POH interactions may therefore affect passive diffusion of POH across the blood-brain and blood-CSF barriers. Interestingly, the distribution of POH and its metabolites between the blood and CSF compartments has not been carefully examined.
In the present study, intranasally applied POH resulted in approximately fivefold lower POH concentrations in the CSF than that in plasma sampled from the same animals, indicating that POH is more likely to enter the systemic circulation following intranasal delivery. However, intravascularly applied POH at the same dose resulted in ~ 200-fold lower POH concentrations in the CSF than that in the plasma, indicating that a markedly higher relative POH CSF exposure was obtained with the intranasal route. These findings suggest that the intranasal route provides direct and elevated central exposure to POH.
Wang et al. recently reported a blood-brain barrier permeability increase following intracarotid POH application at doses down to ~ 800 mg/kg in 3-week-old C57BL/6 tumor-bearing mice, resulting in elevated CNS exposure for bloodborne molecules 32 . In contrast, an increase in blood-brain barrier permeability was not observed following POH application to intravascular sites other than to the carotid artery 32 . In our study, we administered a ~ 34-fold smaller dose (23 mg/kg) via the abdominal aorta and observed that POH concentrations in the CSF were 200-fold lower than POH concentrations in the plasma. Based on these findings, we do not believe that POH administration at the dose used in our study resulted in any significant effect on blood-brain barrier permeability.
We further observed that measured PA concentrations were greater than POH concentrations in all physiological compartments assessed (CSF, plasma) following intranasal dosing. This pattern of higher PA than POH levels was also observed for brain lysate following intranasal dosing of POH in a recent study 33 . Collectively, these findings from our study and Wang et al. suggest significant and rapid metabolism of POH to PA occurs following intranasal POH administration. Significant POH metabolism following intranasal application may be due to rapid POH-to-PA conversion at the nasal mucosa as well as along olfactory and trigeminal nerve-associated pathways into the CNS. POH is sequentially metabolized to perillaldehyde by alcohol dehydrogenases and then to perillic acid by aldehyde dehydrogenases. Prior studies have established that POH can also be sequentially altered to PA by cytochrome P450 and aldehyde dehydrogenases [44][45][46] . Interestingly, multiple reports indicate that these transmembrane enzymes are highly expressed in the olfactory mucosa at levels that exceed those within the respiratory mucosa [47][48][49][50][51] . Furthermore, significant alcohol dehydrogenase immunoreactivity appears to be associated with the perivascular compartment of systemic blood vessels, particularly smooth muscle cells of the tunica media 52 . High mRNA expression levels of alcohol and aldehyde dehydrogenases have also been detected at brain-CSF interfaces such as the leptomeninges, choroid plexus, and ependyma, as well as within the brain parenchyma [53][54][55] . The prominent presence of these biotransformative enzymes at the site of nasal administration, particularly the nasal region associated with brain delivery pathways (e.g. the olfactory epithelium), as well as at downstream sites that have been implicated in direct nose-to-brain biodistribution (e.g. the perivascular www.nature.com/scientificreports/ compartment of leptomeningeal blood vessels and brain tissue itself) is consistent with the significant POH-to-PA conversion we and others have observed following intranasal dosing. Despite marked differences in POH CSF exposure, we observed that both intranasal and intravascular POH application resulted in similar CSF-to-plasma ratios for PA. One explanation for this result would be if PA were a substrate for specific transporters at the blood-CSF and blood-brain interfaces. This might lead to some degree of PA equilibration between blood and CSF irrespective of the delivery route and differences in metabolism. Indeed, immunohistochemical staining in the mouse has shown that brain microvessels, choroid plexus, and the glia limitans are highly immunoreactive for monocarboxylate transporters (MCT-1 and MCT-2) 56,57 , for which PA may well be a substrate. Further work will be needed to validate this hypothesis. Clarification of the regional biodistribution of both POH and PA in the brain following intranasal and intravascular dosing may reveal important sites for POH metabolism and PA transport; however, this was beyond the scope of the present study.
Our study has some limitations. Effects on POH biodistribution related to gender, species-differences, and pathology (e.g. tumor) will require further study. CSF and plasma samples were pooled over 70 min to obtain sufficiently large injection volumes for our HPLC setup. Investigating POH concentrations with more sophisticated detection systems may be able to yield more detailed kinetic information not provided in the present study. We also observed relatively high variability in measured POH plasma concentrations after intravascular dosing; the reasons for this variability were not entirely clear. Finally, the use of tracheotomized rats was necessary to ensure more consistent tolerance for the highly viscous POH formulation and to better allow for targeting to the olfactory mucosal area with tracer dosing. While unlikely, this experimental preparation may have had some effect on the resulting transport and central biodistribution observed following intranasal POH. Nevertheless, our results provide among the first evidence supporting intranasal enhancement of central POH exposure by a direct path that bypasses the blood-brain barrier. Our findings support further use of the intranasal route for the delivery of POH, POH-derived analogs, and possibly other similar chemotherapeutic agents to the CSF and brain for neuro-oncology applications. Sample preparation for chromatography. HPLC assay design was based on a previously described pharmacokinetic study by Hua and colleagues 35 with slight modifications. Briefly, a 50 µL volume of sample/ standard/quality control was supplemented with 100 µL acetonitrile and vortexed vigorously for 3 min (POH) or 1 min (PA). Samples were then centrifuged (14,000 g, 4 °C, 10 min for POH; 14,000 g, room temperature, 5 min for PA). Following centrifugation, 20 µL of supernatant was injected into the HPLC system. For PA analysis, 120 µL of supernatant was carefully transferred to a clean tube and dehydrated under a stream of nitrogen. Pellets were then resuspended in 100 µL diluted mobile phase (50% mobile phase, 50% acetonitrile, 10 mM NaHCO 3 ), and 80 µL of this resuspension was injected into the HPLC system. Chromatography setup. The chromatography setup comprised of Shimadzu Prominence HPLC system modules, a degasser, an auto-sampler, a UV detector (wavelengths set to 210 nm and 217 nm for POH and PA detection respectively), and a column oven (Shimadzu Scientific Instruments Inc., Columbia, MD, USA). The separation was performed using a Zorbax Eclipse XDB-C18 column (4.6 × 150 mm, 5 µm; Agilent Technologies, Santa Clara, CA, USA) and a Zorbax C18 guard (8 × 4 mm, 5 µm; Agilent Technologies, Santa Clara, CA, USA). Mobile phase composition for both analytes was isocratic, comprising of water-acetonitrile mixture (60:40, v/v, pre-mixed, 1.0 mL/min) for POH analysis and 0.05 M ammonium acetate (pH 5)-acetonitrile mixture (64:36, v/v, pre-mixed, 2.0 mL/min) for PA analysis. Both phases were passed through 0.45 µm filters (Millipore-Sigma, St. Louis, MO, USA) before use.

Materials and methods
POH formulation. POH formulation used for this study was adapted from the nasal formulation recipe used in an ongoing clinical trial that examines the safety and efficacy of perillyl alcohol in recurrent grade IV glioma patients (NCT02704858). Briefly, intranasal POH formulation consisted of 10% v/v (9.4% w/w) POH, 55% v/v glycerin USP and 35% v/v dehydrated alcohol. Dose for in vivo experiments was determined by allometric scaling based on body weight 37 to determine doses for rats based on the smallest daily dose received by a human subject in the clinical trial i.e. 4.745 mg (i.e. 23.75 mg/kg). Formulations were refrigerated as one-timeuse aliquots (54 µL per aliquot) at 4 °C. The volume (54 µL) was considered safe for intranasal administration based on experiences with similar volumes for intranasal study paradigms in rats in our lab 4 www.nature.com/scientificreports/ Intracisternal sampling of cerebrospinal fluid (CSF). A dedicated CSF withdrawal system was used for POH experiments (Fig. 1). This system comprised of a 20 cm long polyethylene tubing (PE-10; inner diameter: 0.61 mm, outer diameter: 0.28 mm; Plastics One Inc., Roanoke, VA, USA); a 100 µL Hamilton syringe (inner diameter: 1.475 mm; Harvard Apparatus, Holliston, MA, USA), and a two-port shut-off valve assembly (PEEK, inner diameter: 0.5 mm; Idex Health and Science, Oak Harbor, WA, USA). Tubing was firmly secured at both ends of the valve assembly using NanoTight sleeves (FEP; inner diameter: 0.69 mm; Idex Health and Science, Oak Harbor, WA, USA); these sleeves flanked on either side of valve fittings. The valve fitting side that faced the cisternal space was connected to a different set of tubing assembly i.e. 1 cm PE-10 tubing sealed to a 1.27 mm long polyether ether ketone tubing (PEEK; 33 GA; Plastics One Inc., Roanoke, VA, USA). PE-10/PEEK tubing connections were sealed using molten wax and cyanoacrylate on the day of the experiment.

Tracheotomy (prior to intranasal POH administration). Preliminary experiments indicated that
glycerin was increasing the viscosity of POH formulation (ethylene glycol: 0.0162 N s/m 2 , glycerin: 0.95 N s/m 2 ) that resulted in breathing issues and low survival rate of animals at the experimental endpoint (70 min). Hence, animals subjected to intranasal POH administration underwent a brief tracheotomy procedure before cisternal puncture. Briefly, an incision was made parallel to rings of hyaline cartilage in the anterior tracheal region; a short piece (5 cm) of PE-205 tubing (outer diameter: 2 mm) was then inserted into this opening. Tubing was secured by 2-3 surgical knots with a general-purpose thread, followed by sealing the skin with cyanoacrylate. In our experience, tracheotomy prior to intranasal administration increased the survival rate and time for rats. Tracheotomy was not performed for intravascular POH experiments.
Intracisternal cannulation. Intracisternal cannulation was the first procedure conducted on animals undergoing intravascular POH dosing. Cannulation was performed after the tracheotomy procedure for intranasal POH administration experiments. Cisternal cannulation procedure was adapted from a recently reported setup for intrathecal delivery in rats at our laboratory 37 . Briefly, lidocaine hydrochloride (2%; 0.5 mL, s.c.) was applied as a local anesthetic at the scalp before positioning the animals in a stereotaxic frame. A 5 cm midline incision was then made from the base of the neck 2 cm anterior to the bregma. The skin was laterally secured with hemostats and muscle layers were sequentially retracted using vannas scissors and cotton tip applicators. Once the atlanto-occipital layer was exposed, gentle incisions were made by alternating between a goniotomy knife and forceps to avoid dural puncture. When the cisterna magna was visible beneath the intact dura, a small dural hole was made using a dental needle at a 60° angle from the vertical. PEEK tubing was then inserted with the help of a microelectrode holder and stereotaxic frame's manipulator arm. Excess CSF flowing out from the hole was absorbed using cotton tip applicators and tubing was sealed to the cranial opening with cyanoacrylate; sealed setup was then dried until cured.
Abdominal aortic cannulation. Intracisternal cannulation preceded abdominal aortic cannulation for both treatment groups. Briefly, animals were carefully flipped over to the ventral side without displacing the tubing from its position. An incision was made below the xiphoid process along the abdominal midline. Skin and visceral organs were sequentially retracted to obtain a clear view of the aorta. Aorta was separated from the vena cava by cotton tip applicators and curved sharp forceps. The fascia between the two abdominal vessels was broken the opening was widened. Next, a 6-inch long piece of general-purpose thread was tied to the bottom of this opening to cut off blood supply to the lower body. www.nature.com/scientificreports/ using a 3-way stop-cork. Another round of 500 µL saline was injected to flush the entire formulation from the cannula into the artery.
Blood and CSF withdrawal. CSF was withdrawn at 0-12 min, 14-26 min, 28-40 min, 40-54 min, and 56-70 min at 5 µL/min. After each withdrawal, CSF samples were ejected from the tubing into clean microcentrifuge tubes over a 2-min interval. Tubes were pre-cooled on ice. Blood sampling (300 µL) was performed every 14 min. Blood was collected in 1.5 mL microcentrifuge tubes containing 3 µL sodium heparin (anticoagulant; 1000 USP units). Tubes were centrifuged (3000 rpm, 4 °C, 5 min) to separate red blood cells from the CSF and plasma at the end of the experiment. Samples were further spun (12,000 rpm, 4 °C, 5 min) to remove additional contamination. Samples were stored (− 80 °C) until ready for analysis. Before HPLC analysis, samples from each animal were thawed and pooled across all time points to run HPLC assays in triplicates.
Euthanasia. Animals were euthanized by injecting 1 mL KCl (1 M) into the pericardium.

Statistical analysis.
For each HPLC analysis, LLOD were determined before sample analysis. Pure samples used for standard curves were dissolved in artificial CSF (aCSF) and commercially available rat plasma. POH concentrations in the CSF and plasma were each estimated by linear regression analysis. The CSF concentration of POH was below the LLOD for the assay in 1 out of 5 animals in the intravascular POH cohort; these values were therefore excluded from statistical analysis. POH and PA concentrations in the plasma and CSF were above LLOD for all other cases. Treatment groups were compared using the Student's t-test when the two groups exhibited normal distribution and equal variance. In the absence of a normal distribution, groups were compared using the Mann-Whitney rank-sum test. Both analyses were performed using SigmaPlot (Systat Software, San Jose, CA) to identify statistical significance. www.nature.com/scientificreports/