The prototype rectal heating device was designed and constructed using a hollow, heat-conductive copper cylinder (15 cm length, 2.23 cm outer diameter (OD), 2.0 cm inner diameter (ID)) (Figure 1A). One end of the cylinder was fitted with a 3D-printed endcap and flow channel assembly (Figure 1B). The flow channel assembly included discrete inlet and outlet water channels, as well as a groove containing a rubber O-ring to ensure a watertight seal with the copper tube. The inlet channel was designed to eject water from the cross-sectional center toward the closed end of the cylinder. The endcap had a hollow, rounded region protruding 1 cm from the end of the cylinder.

A 50 L/min submersible pump (AE-172, AquaEuroUSA, Los Angeles, United States) was connected to the inlet channel with rubber tubing (2 m length, 1.7 cm OD, 0.95 cm ID). The outlet channel was connected to an external, variable-temperature water bath and the flow channel assembly was affixed to the copper tube using epoxy adhesive.

To assess heat distribution, we performed a benchtop test using a tissue-mimicking thermochromic gel phantom that changes color from tan to magenta when heated from 40 °C–70 °C. Synthesis of this polyacrylamide-based gel phantom was performed as previously described (Negussie et al., 2016; Mikhail et al., 2016; Eranki et al., 2019). The rectal heating device was wrapped in polyolefin film and fixed at the center of a 1 L glass beaker filled with the gel phantom precursor solution. After solidifying overnight, the phantom was equilibrated in a 37 °C insulated water bath to match normal physiologic temperature. A separate water bath was heated to 65 °C and the water pump was set to its maximum flow rate (50 L/min) to circulate water through the rectal heating device. This temperature was selected to ensure visible color change in the phantom, enabling assessment of spatial heating homogeneity. After 1 h of water circulation, the rectal heating device was removed, and the phantoms were cut longitudinally or axially and photographed.

Rectal tissue heating and doxorubicin delivery were simulated using computer models as previously described, with modifications (Mikhail et al., 2017; Gasselhuber et al., 2012). The rectum was modeled as cylinder of 25 mm inner diameter with 3 mm thickness. The model included surrounding tissue, assuming axisymmetric geometry. In addition, the model included systemic plasma and systemic tissue compartments. Briefly, the heat transfer model assumed surface tissue heating from circulating water at 44 °C, heat conduction into the wall of the rectum, and cooling via blood perfusion. The drug delivery model considered temperature-dependent intravascular release of doxorubicin, extravasation of free doxorubicin, and cellular doxorubicin uptake (Mikhail et al., 2017; Gasselhuber et al., 2012). Plasma pharmacokinetics of LTLD were considered based on clinical data from a Phase I trial (Wood et al., 2012). Since no cellular uptake parameters for colon cells was available in the literature, we used prior parameters derived from in vitro studies in hepatocytes (Rossmann et al., 2017). IV and IA infusion of drug were simulated by either administering LTLD to the systemic plasma compartment (IV), or to the capillary bed of the rectum (IA) for 30 min. Heating was assumed to continue for a total of 60 min after the start of the infusion, and the heating device was assumed to have been pre-heated as in the animal studies.

All animal procedures were approved by the Animal Care and Use Committee of the NIH Clinical Center and conducted in accordance with applicable federal regulations. Eight female Yorkshire swine (16–18 weeks old, 53–63 kg; Oak Hill Genetics, Ewing, IL) were studied. Due to a known anaphylactoid reaction to liposomal formulations, all swine received premedication as previously described (Mikhail et al., 2017). Dexamethasone (0.12 mg/kg, IM) was administered twice daily for 48 h prior to the day of the study. On the day of study, dexamethasone (0.12 mg/kg, IM), famotidine (0.5 mg/kg, IM) and diphenhydramine (2 mg/kg, IM) were given 1.5–3 h before infusion of study formulations. Finally, meloxicam (0.3 mg/kg, IV) was given 10 min before infusion of study formulations. Anesthetic induction was achieved with IV propofol (1 mg/kg) following sedation with IM ketamine (25 mg/kg), midazolam (0.5 mg/kg), and glycopyrrolate (0.01 mg/kg). Endotracheal intubation was performed, and a surgical plane of anesthesia was maintained with 100% oxygen and isoflurane (1%–3%). An introducer sheath was surgically inserted in the jugular vein for administration of IV fluids and blood collection. For those cohorts where the study drug was administered IV, a second introducer sheath was placed in the contralateral jugular vein. An enema (100–150 mL) was administered under anesthesia, followed by irrigation of the rectum.

Two T-type thermocouples (accuracy: ±0.5 °C) were affixed to the heating tube at 6.5 cm and 12.5 cm positions using silicon tape and connected to a LogMaster 4-Channel Logger (ThermoWorks, Utah, United States), which measured temperature every second (Figure 1C). For analysis, a rolling average of the current and prior 9 measurements was used. The assembled rectal heating device was equilibrated to 44 °C by recirculating warm water prior to insertion into the rectum. Hyperthermia was maintained for 1 h from the start of drug infusion. For the IV LTLD cohort where the rectum was not heated, the rectal heating device was inserted but maintained at 37 °C.

For IA drug delivery, introducer sheaths were surgically placed in the carotid arteries bilaterally. Catheters were advanced through the sheaths and diagnostic pelvic angiography was performed and the arteries supplying the rectum were visualized. Then, catheters were advanced into the internal iliac arteries bilaterally. The catheter tips were positioned in the distal internal iliac arteries at the level of the caudal aspect of the obturator foramina in the posterior-anterior imaging projection.

The swine were divided into four treatment groups (n = 2/group): (1) IV LTLD (Thermodox®, Celsion/Immunon, New Jersey, United States), (2) IA doxorubicin with hyperthermia, (3) IV LTLD with hyperthermia, and (4) IA LTLD with hyperthermia. Drug formulations were infused over 30-min at a dose of 0.7 mg/kg. For IA drug delivery, the drug dose was divided equally between the two catheters in the internal iliac arteries. Catheters were connected to an automatic dual syringe injector to deliver the entire dose in 30 min. Drug infusion began following temperature equilibration at 44 °C inside the rectum.

Venous blood samples (3 mL) were collected at baseline and sequentially for 1 h from the start of infusion. Plasma was separated by centrifugation, and doxorubicin was quantified by LC/MS. A non-compartmental analysis was employed to determine pharmacokinetic parameters. Maximum plasma concentration (C_max) and time to C_max (T_max) were observed values. The elimination rate constant and terminal half-life were calculated from the log-linear regression of terminal exponential phase data points. The area under the concentration-time curve (AUC_0–60min) was calculated from the beginning of the drug infusion to the time of the last measurable plasma concentration using the trapezoid method and extrapolating to infinity (AUC_0-   ∞ ) by dividing the last measurable plasma concentration by the terminal exponential rate constant. Clearance (CL) was calculated as dose/AUC_0-   ∞ .

One hour after the start of drug infusion, swine were euthanized under general anesthesia by administration of Beuthanasia-D or equivalent (Pentobarbital Sodium 390 mg/mL and Phenytoin Sodium 50 mg/mL; 1 mL/10 lbs IV), consistent with current AVMA guidelines for euthanasia. Immediately following euthanasia, the rectum was exposed via a midline abdominal incision, marked at cranial and caudal aspects using sutures, and dissected en bloc. The rectum was opened with a midline incision and divided into 12–16 pieces, approximately 2 × 2 cm. For each animal, the mean concentration of doxorubicin in eight pieces taken from ventral, dorsal, cranial, and caudal aspects was determined. Samples of perirectal fat and the heart were also acquired. Tissues were rapidly frozen in liquid nitrogen. Frozen tissues were mounted with Tissue-Tek cryoadhesive (Sakura Finetek, Torrance, CA) and cut into 10 μm cross-sections for fluorescence imaging. Doxorubicin was extracted from homogenized rectum, heart, and perirectal fat samples and quantified by LC/MS.

Doxorubicin distribution in the rectal wall was assessed by fluorescence microscopy (Olympus VS200-6 FL, Center Valley, Pennsylvania, United States; excitation 468 nm, emission 635 nm) of transverse rectal wall sections. Images of intrinsic doxorubicin fluorescence were pseudo-colored yellow. All images were captured with identical exposure and window/level settings.

Comparisons of mean tissue doxorubicin concentrations and pharmacokinetic parameters between groups were performed using one-way ANOVA with a single pooled variance (GraphPad Prism v9.0.0). Post hoc analyses were performed using Tukey’s multiple comparison test. T_max is presented as median and range.

After 1 hour of circulating 65 °C water through the rectal heating device embedded in the thermochromic phantom, a uniform color change was observed along both the length and circumference of the device (Figure 2). This temperature was selected to ensure visible color change in the phantom.

Computational simulations were performed to estimate tissue temperature and doxorubicin distribution in the rectal wall (Figure 3). The model geometry reflected the experimental setup with full-length heating for IV and IA LTLD simulations. For simulation of IA drug infusion, perfusion was included in only half of the model to mimic selective arterial catheterization. After 60 min of heating, tissue temperatures were predicted to reach the threshold for doxorubicin release from liposomes up to a few millimeters away from the rectal lumen. Simulations also suggested doxorubicin accumulation in the heated rectum following both IA and IV LTLD infusion, with differing spatial distribution patterns. The former appeared only in regions perfused by catheterized arteries and was more concentrated near the lumen, whereas the latter appeared at more moderate concentrations but with greater transverse and longitudinal distribution.

Bilateral catheter placement in the internal iliac arteries was confirmed by angiography (Figure 4A). Catheter localization and rectal perfusion were confirmed using contrast-enhanced cone-beam CT (Figure 4B). The position of the rectal heating device within the rectum and the catheters was confirmed by x-ray (Figure 4C).

During hyperthermia, the two thermocouples attached to the proximal and distal ends of the rectal heating device recorded mean temperatures of 43.5 °C and 44.3 °C, respectively (Figure 5A), indicating maintenance of the target hyperthermic range throughout the procedure.

All animals tolerated the procedures without adverse events. No macroscopic signs of thermal injury to the rectum were observed. In animals receiving LTLD and hyperthermia, a sharply demarcated red discoloration was evident in the lower rectum and anus, most profoundly following IA LTLD delivery, corresponding to the region supplied by the catheterized vessels and exposed to the rectal heating device. This discoloration was consistent with doxorubicin accumulation (Figure 5B).

Fluorescence imaging of rectal wall cross sections revealed intrinsic doxorubicin fluorescence, with the most pronounced deposition observed in swine that received IV or IA LTLD combined with hyperthermia. In these treatment groups, doxorubicin was distributed throughout the rectal wall including the muscularis (Figure 6). The punctate fluorescence pattern, especially evident in LTLD-treated swine in combination with hyperthermia, is consistent with nuclear uptake of doxorubicin.

Plasma doxorubicin concentrations during treatment are shown in Figure 7A, and the corresponding pharmacokinetic parameters are summarized in Table 1. A complete list of p-values for comparisons between treatment groups is provided in the Supplementary Material. Compared to IA doxorubicin + HT, the values of AUC_0–60min and C_max were greater, and clearance (CL) was lower for IV LTLD, IV LTLD + HT, and IA LTLD + HT. The elimination half-life (t_{1/2 elimination}) was longest for IV LTLD compared to IA doxorubicin + HT, IV LTLD + HT, and IA LTLD + HT.

Doxorubicin concentrations in the rectal wall were highest in swine that received either IV LTLD + HT or IA LTLD + HT compared to IV LTLD (p = 0.0651 and 0.0405, respectively) or IA doxorubicin + HT (p = 0.0707 and 0.0438, respectively) (Figure 7B). Mean doxorubicin concentrations in full thickness samples of the rectal wall were 0.49 ± 0.16 (IV LTLD); 0.67 ± 0.46 (IA doxorubicin + HT); 7.45 ± 6.18 (IV LTLD + HT); and 8.41 ± 5.15 μg/g (IA LTLD + HT) (Figure 7C). Low doxorubicin levels were detected in the heart across all groups (Figure 7C). One swine treated with IA LTLD + HT had high doxorubicin concentration in the perirectal fat.