Polymer-assisted intratumoral delivery of ethanol: Preclinical investigation of safety and efficacy in a murine breast cancer model

Focal tumor ablation with ethanol could provide benefits in low-resource settings because of its low overall cost, minimal imaging technology requirements, and acceptable clinical outcomes. Unfortunately, ethanol ablation is not commonly utilized because of a lack of predictability of the ablation zone, caused by inefficient retention of ethanol at the injection site. To create a predictable zone of ablation, we have developed a polymer-assisted ablation method using ethyl cellulose (EC) mixed with ethanol. EC is ethanol-soluble and water-insoluble, allowing for EC-ethanol to be injected as a liquid and precipitate into a solid, occluding the leakage of ethanol upon contact with tissue. The aims of this study were to compare the 1) safety, 2) release kinetics, 3) spatial distribution, 4) necrotic volume, and 5) overall survival of EC-ethanol to conventional ethanol ablation in a murine breast tumor model. Non-target tissue damage was monitored through localized adverse events recording, ethanol release kinetics with Raman spectroscopy, injectate distribution with in vivo imaging, target-tissue necrosis with NADH-diaphorase staining, and overall survival by proxy of tumor growth. EC-ethanol exhibited decreased localized adverse events, a slowing of the release rate of ethanol, more compact injection zones, 5-fold increase in target-tissue necrosis, and longer overall survival rates compared to the same volume of pure ethanol. A single 150 μL dose of 6% EC-ethanol achieved a similar survival probability rates to six daily 50 μL doses of pure ethanol used to simulate a slow-release of ethanol over 6 days. Taken together, these results demonstrate that EC-ethanol is safer and more effective than ethanol alone for ablating tumors.


Introduction
Ablation is the focal destruction of tissue using a small instrument delivered under the skin, typically performed when surgical excision of the tissue is impractical, inaccessible, or dangerous for the patient.
Ethanol ablation, or percutaneous ethanol injection (PEI), kills tumors by causing coagulative necrosis upon contact with tissue (1-3). Ablation of small focal tumors with ethanol is a well-accepted technique because it is cost-effective, can be visualized with ultrasound, and has acceptable clinical outcomes for treatment of hepatocellular carcinoma (4). Ethanol ablation has been used as an alternative to radiofrequency ablation for hepatocellular carcinomas in cirrhotic patients because it is less expensive and less time-consuming with a similar 5-year survival rate for small lesions (3,5). More recently, ethanol ablation has been used successfully for neurolysis (6,7), cardiomyopathy septal ablation (8,9), treatment of cystic thyroid nodules (1, 2, 10,11), palliation of osteolytic bone metastases (1, 2,12,13) and ablation of neuroendocrine tumors (14,15). Notably, the precision with which any ablation modality can inflict necrosis on a target tissue without damaging nontarget tissue is the key metric for success of an ablation technique. Unfortunately, pure ethanol leaks freely out of the tumor, along the path of least resistance resulting in high rates of post-treatment tumor progression (16,17). In fact, ethanol ablation often requires multiple sessions to achieve the same efficacy as radiofrequency ablation (18) as it provides incomplete coverage of large lesions in a single session (19). For these reasons, microwave ablation and cryoablation have gained favor in high-resource settings, as tissue destruction can be directed to a radially symmetric volume of tissue more reliably. However, use of thermal ablation requires hard to access resources including CO 2 cylinders for cryotherapy, expensive specialized machinery for microwave ablation, and a consistent power supply for all thermal ablation limiting use of thermal ablation to high-resource settings. Therefore, an unmet clinical need remains for local treatment, especially in low-resource settings, where surgery remains widely inaccessible (20).
Here a simple method for percutaneous tumor ablation using polymer-assisted delivery of ethanol with ethyl cellulose (EC) is demonstrated in murine tumors. EC is an inert, ethanol-soluble polysaccharide used as a coating for medical pills and regarded as safe by the Food and Drug Administration. Once the EC-ethanol mixture encounters an aqueous environment, a fibrous gel is formed via non-solvent induced phase separation when water contacts the EC-ethanol mixture (21). EC was incorporated into the ethanol solution in order to increase the viscosity of the ethanol and limit the spread of ethanol out of the tumor. We hypothesize that EC will increase the time that ethanol is held at the injection site in an intratumoral depot, and as the ethanol remains in contact with the target-tissue longer it will increase local necrosis. In previous studies we characterized the rheological properties of the gel produced by EC-ethanol (22) and its delivery into ex vivo tissue (23). We previously demonstrated that 3% EC-ethanol forms a fibrous gel upon contact with water and in superficial hamster cheek pouch tumors decreased tumor volume observed over a 7 day period compared to pure ethanol (22). EC-ethanol's unique physical properties have been safely utilized clinically to treat venous malformations (24,25) and herniated discs (26), but has yet to be assessed for the ablation of subcutaneous tumors.
In this study we utilized murine breast cancer tumors to test the ability of EC-ethanol to precisely ablate tumors percutaneously. Cell death due to ethanol is dependent on two factors: the concentration of ethanol and the exposure time. Thus, our goal was to slow the release of ethanol from the injection site to allow for maximum ethanol exposure to the tissue near the needle while limiting ethanol exposure to non-target tissue.
Ethanol's cytotoxic ability is due to the fact ethanol molecules are very-small and polar; thus, tagging ethanol with a radio-opaque or fluorescent probe to monitor its location was deemed impractical for this study as it would affect the transport of ethanol through tissues. Ethanol does, however, produce a distinct signal with Raman spectroscopy (27,28) which was utilized to monitor the release kinetics of ethanol through tumor tissue. To monitor the spatial distribution of EC-ethanol a fluorescent powder, fluorescein, was mixed with ECethanol to monitor the approximate location of the injectate in vivo and ex vivo. Fluorescein acts as a representative small molecule, to demonstrate how chemotherapeutic agents could easily be incorporated into the EC-ethanol solution to improve intratumoral delivery. Ablation efficacy was then validated with viability staining of tumors and monitoring the tumor volume after EC-ethanol ablation.
Large 1 cm 3 mouse tumors were used to model clinically relevant tumor sizes for assessment of fluid distribution (fluorescein distribution) and tumor necrosis (viability staining). To assess safety (adverse events) and efficacy (tumor growth) small 50 mm 3 tumors were used to allow for sufficient time between the treatment and the ethical tumor burden limit of 2000 mm 3 in mice. 67NR cells were selected for this study because they are non-metastatic and grow rapidly without producing large necrotic cores, unlike their sister line 4T1, which could artificially contain injection fluids. 67NR cells are triple-negative breast cancer with a luminal phenotype and BALB/c background. This study characterizes how the inclusion of EC alters ethanol delivery and the resulting safety and efficacy of percutaneous EC-ethanol injections in subcutaneous breast tumors.
All experiments used 2.5% w/w of fluorescein mixed in ethanol. The stability of EC polymer is well documented (29), however the rate of EC degradation in pure ethanol has yet to be quantified. To limit possibility of chemical modification of EC polymers, 6% EC-ethanol solution was mixed without heat no more than 24 h before injection. All mixtures were mixed with an ethanol safe stir-bar in ethanol-safe containers. All injections were performed using 27-gauge needles with manual needle placement in the center of the tumor. Injections were performed at a fixed rate of 1 mL/h with a syringe pump unless otherwise specified. One mouse died hours after a 4 mL/kg dose; however, the mouse was also under isoflurane anesthesia for a prolonged time (1 hour, 3%). Thus, considerations were made to limit isoflurane exposure after this adverse event. No other changes in systemic adverse events was noted for doses below 6 mL/kg. Mice were monitored for adverse events for a minimum of 14 days and up to 35 days (dictated by tumor growth) to observe any delayed effects after treatment. Greater than 200 breaths per minute qualified as respiratory distress. Limping or dragging of the tumor-bearing leg at any point indicated mobility impairment. Swelling and redness at any time after injection was noted as inflammation/edema. Bruising or bleeding of the tumor-bearing leg was noted as bleeding. Ulceration was defined as a scab on the skin over the tumor persisting for more than 3 days. A drop of 15% or greater of the baseline weight at any time after treatment was recorded as a severe loss in body weight. Because the tumors in this study were on the flank, mobility impairment, inflammation, edema, and ulceration of the tumor-bearing foot were considered localized adverse events. Lethality, respiratory distress, and loss in body weight were considered systemic adverse events.

Quantification of Ethanol diffusion coefficient using Raman spectroscopy
The diffusion coefficient of ethanol through tumor tissue was measured using a Raman spectroscopy assay coupled with a deterministic mathematical model of ethanol diffusion. Tumor specimens (n=6) were cut longitudinally and trimmed to fit 12 mm Snapwell inserts containing porous polycarbonate membranes (Corning-Costar®, New York, NY). Each snap well was placed in a custom-built chamber filled with 600 µL of ethanol to enable ethanol diffusion through the membrane and upwards into the tumor (Fig 2). A custom-built confocal Raman spectroscope with a 785 nm excitation laser diode (LD785-SH300, Thorlabs Inc., Newton, NJ) was used to capture Raman spectra at the surface of the tumor (30). The least squares fit was performed in MATLAB (MathWorks, Natick, MA) to extract the relative contribution of ethanol and tumor tissue to the measured Raman spectra at the tumor surface, using a previously validated method (30)(31)(32).
To compute the diffusion coefficient of ethanol, a deterministic two-compartment transport model was used that recapitulates the Raman assay, using Fick's law of diffusion to characterize ethanol diffusion from an ethanol compartment at the bottom, to the tumor at the top (Supplemental Fig 1). An analogous model was outlined previously in (33) and was applied to estimate the diffusion coefficients of drugs in biological tissues (32). Here, we assumed that the partition coefficient between the ethanol and tumor compartments (ET) was 1 due to the high porosity of the polycarbonate membrane, which is fully permeable to small molecules like ethanol. The diffusion coefficient of pure ethanol is 10 -5 cm 2 /s (34), and the height of the ethanol compartment, hE, was 1.5 cm. The height of each tumor sample, hT, was measured using a micrometer. The two unknowns, the diffusion coefficient of ethanol in the tumor (DT) and the first-order loss term that accounts for ethanol evaporation from the tumor (KEV) were computed and optimized by fitting the predicted concentration of ethanol at the surface of the tumor to the measurements obtained using Raman spectroscopy.

Whole-body Fluorescent Imaging
Mice were imaged using a whole-body in vivo IVIS Lumina imaging system (Perkin Elmer) in the prone position, 30 minutes after injection with 6% EC-ethanol or pure ethanol (n=5) to allow for injectate diffusion.
Images were acquired with excitation at 430 nm, emission at 520 nm, and an exposure time of 1 second. Mice with 5 mm 67NR flank tumors (n=5) were given an injection of 6 mL/kg of either 6% EC-ethanol or pure ethanol. Pixels with a radiant efficiency of 0.5x10 10 or higher we selected to represent the injection area. Area calculations and compactness calculations were performed using standard image processing tools in MATLAB (MathWorks).

Imaging with a hand-held fluorescence Microscope
For ex vivo imaging, 6% EC-ethanol was injected into ~1000 mm 3 tumors. Tumors were injected with 150 μL of either pure ethanol (n=10), 3% EC-ethanol (n=10), or 6% EC-ethanol (n=20) at 1 mL/h. A control image of a tumor without any injection was acquired to account for tumor autofluorescence. Tumors were flash frozen and cross-sectioned centrally for imaging at a 10 mm working distance. Tumors were imaged using a fluorescence microscope with excitation at 480 nm and an emission bandpass filter at 520±15 nm. Pixels with >80% saturation in the green channel were counted as the area of fluorescence. Pixel dimensions were converted to an area using a digital ruler to scale each set of photos, and the radius r 1 was computed. The fluorescent volume V 1 (calculated using r 1 ) was calculated assuming the fluorescent image was taken at the center of a spherically distributed injection.

Histology and Immunohistochemistry
For viability staining, a 5 µm section was taken serially every 2 mm throughout the tumor and stained with NADH-diaphorase to quantify necrosis. NADH-diaphorase staining was performed using Nitrotetrazolium Blue Chloride (Sigma-Aldrich) and B-Nicotinamide Adenine Dinucleotide (Sigma-Aldrich). The region of interest (ROI) was selected manually from a digital scan at 5x magnification to include only non-viable cellular regions. Adjacent H&E slides were used to confirm tumor regions. Necrotic volume was calculated by the Reimann Sum: , where volume (V) is equal to the summation of area (A) from 1 to n times = ∑ = 1 ( )∆ the distance between each measurement (x). In this case, x = 2 mm.

Statistical Analysis
Rates of adverse events were assessed with a Chi Squared test with a confidence level of 95% with a degree of freedom of 1. Two-tailed ANOVA testing with unequal variance was performed on all groups against each other and controls with a confidence level of 95%. Post-hoc multiple comparisons were performed using Tukey's HSD test. Survival curves were quantified using Kaplan-Meier analysis, and a logrank test was performed to determine the significance of a P-value less than 0.05 with a confidence level of 95%.

EC-Ethanol Significantly Reduces Adverse Events compared to Pure Ethanol
We investigated whether using EC-ethanol would impact the rates of adverse effects compared to pure ethanol injections (recorded in Supplemental Table 1). Mice with 67NR flank tumors received a single intra-tumoral injection of different volumes of 6% EC-ethanol (2-10 mL/kg), 4 days after tumor inoculation when tumors were approximately 50 μL in volume. A dose of 6 mL/kg of EC-ethanol was identified as the maximum tolerable dose (MTD) because it did not cause significantly more systemic adverse than untreated tumor bearing controls (Supplementary Table 1). These mice were compared to both mice that received a single intra-tumoral injection of 6 mL/kg (150 μL for average 25 g mouse) of ethanol or a repeat dose of 2 mL/kg (50 μL for average 25 g mouse) of ethanol over 6 days. All injections were delivered at a rate of 1 mL/hr.
An untreated tumor bearing group was also included. A mouse was considered to have a localized adverse event if it displayed signs of mobility impairment (limping), subdermal bleeding, inflammation, or ulceration (scabbing) at any time during the two-week monitoring period (Fig 1A-B). Mobility impairment in the form of dragging the tumor-bearing leg or limping was most likely due to muscle or nerve damage in the leg from either ethanol or tumor invasion. EC-ethanol mixtures reduced limping, ulceration, and bleeding, but not inflammation, when compared to the same dose of pure ethanol (Fig 1C-F). A 6 mL/kg dose of EC-ethanol resulted in less local bleeding, inflammation, and limping than the same dose of pure ethanol, indicating that EC helps limit off target effects of intratumoral ethanol injections. Chi squared test significance of P<0.05 is indicated with (*) asterisk compared to untreated group, ( †) dagger compared to 6 mL/kg ethanol group, ( ‡) double dagger compared to repeated 6 x 2 mL/kg/day ethanol group.

EC-Ethanol Slows Ethanol Diffusion
In order to investigate the mechanism for improved safety of EC-ethanol over ethanol, the release kinetics of ethanol through tissue were investigated using Raman Spectroscopy. The tissue-depot interface was modeled as shown in Supplemental Fig 1. EC was hypothesized to reduce off-target leakage by slowing the release of ethanol from the injected depot site. The relative concentration of ethanol over time a sample was measured with Confocal Raman Spectroscopy (30-32) as ethanol diffused through thick (5 -10 mm) sections of 67NR tumors. The change in ethanol concentration over time was used to fit a model of molecule transport in a two-compartment model. The effective diffusion coefficient (D eff ) of a molecule represents the effective transport rate at which molecules move from one compartment to the other. D eff can be approximated from the concentration over time recordings (Fig 2A-B) as in previously validated algorithms (30)(31)(32). The D eff of ethanol of 6%EC-ethanol was approximately half that of pure ethanol (3.3 ± 1.5 x 10 -6 cm 2 /s vs. 5.9 ± 1.2 x 10 -6 cm 2 /s ; P<0.05; n=7,6) (Fig 2C). Taken together, these results demonstrate that addition of EC to ethanol significantly slows the release of ethanol through tumor tissue.

EC-Ethanol Delivery is More Uniformly Distributed Compared to Pure Ethanol
To demonstrate that the slower diffusion of EC-ethanol influences the local distribution of ethanol, the localization of EC-ethanol and ethanol-only injections were quantified. Both EC-ethanol and pure ethanol were mixed with the fluorescent dye, fluorescein (2.5% w/w) to enable in vivo tracking of the injectate. Small, 50 mm 3 67NR tumors were injected with 150 µL at 1 mL/hr to force overflow from the tumor and monitor the event of tumor leakage. Nude mice were used to limit autofluorescence from fur. Whole-body imaging was performed 30 min after injection (Fig 3A, B) using an in vivo imaging system (IVIS Lumina, Perkin Elmer. Compactnessdefined as 4πA/P 2 , where A and P are the area and perimeter of the fluorescent region, respectively-was calculated using MATLAB. A perfect circle has a compactness of one, and any deviations from a circle will have a compactness that deviates from one. The compactness of the EC-ethanol injectate was significantly greater than that of pure ethanol (0.96 ± 0.01 vs. 0.80 ± 0.13; P<0.05; n=8,9) (Fig 3C), indicating reduced injectate leakage when using the EC-ethanol mixture over pure ethanol.

EC-ethanol Increases Intra-tumoral Ethanol Retention
The effect of relative concentration of EC and injection rate on injectate retention in the tumor were evaluated. A large, 1 cm 3 tumor volume was selected to replicate clinically relevant tumor sizes while remaining under the ethical tumor burden limit for mice of 2 cm 3 . Nude mice with 1 cm 3 67NR flank tumors received a single 150 μL intra-tumoral injection of either pure ethanol, 3% EC-ethanol, or 6% EC-ethanol.
Injections of pure ethanol (n=10), 3% EC-ethanol (n=10), and 6% EC-ethanol (n=20) were performed at 1 mL/hr Injections and also at 5 mL/hr (n=5) and 10 mL/hr (n=5) for 6% EC-ethanol injections. The contrast agent fluorescein (2.5% w/w) was added to image intra-tumoral injectate retention using fluorescence microscopy. The area of fluorescein was quantified in frozen tumor cross sections immediately after injections as outlined in the methods section. The intra-tumoral fluorescein retention was greater following injection of 6% EC-ethanol than following injection of 3% EC-ethanol or pure ethanol (P<0.01; all others n.s.) (Fig 4A,B).
The intra-tumoral fluorescein retention was also greater when using slower injection flow rates: injection of 6% EC-ethanol at 1 mL/hr yielded a greater fluorescent volume than at 5 or 10 mL/hr (P<0.05) (Fig 4A,B). The ratio of fluorescent volume to injected volume was 1.85 ± 1.00 for 6% EC-ethanol, which was significantly greater than that for 3% EC-ethanol (1.02 ± 0.75, P<0.05) or pure ethanol (0.74 ± 0.58. P<0.05) .  (Fig 5A) and NADH-diaphorase to distinguish viable cells (blue) from necrotic cells (white) (Fig 5B). Untreated tumors were included as a control.
The necrotic volume following injection of 6% EC-ethanol was 5-fold greater than that of pure ethanol. The average ratio of necrotic volume to injected volume for 6% EC-ethanol was 2.00  0.42, which was significantly greater than 3% EC-ethanol (0.80  0.2) or pure ethanol (0.40  0.09) (Fig 5C). As inducing necrosis is the primary objective of any ablation technique, 6% EC-ethanol is more effective at inducing target tissue necrosis compared to 3% EC-ethanol or pure ethanol.   Figure 6A. Survival probability was quantified with Kaplan-Meier curves in Figure 6B. Mice treated with a single ethanol injection did not experience a growth delay compared to untreated controls. Only mice treated with EC-ethanol experienced significantly increased survival rates compared to untreated controls and the same dose of ethanol (Fig 6B). Repeated ethanol injections did not produce significantly greater survival times compared to any group (Fig 6B).

Discussion
In this study it was demonstrated that 6% EC-ethanol delivered at a rate of 1 mL/hr is safer and more with chemotherapy could potentially create a more robust response compared to either one alone (35).
Additionally, immunomodulatory agents may enhance EC-ethanol therapy. As EC-ethanol creates a large necrotic region like other ablation modalities, there is a potential for adaptive immune system activation through the release of tumor antigens as seen with cryotherapy and thermal ablation (36). Moreover, the response to immunotherapies is known to be bolstered with local tumor ablation (36). Thus, future studies should investigate the effect that EC-ethanol ablation has on drug delivery and antitumor immunity.
In conclusion, it was demonstrated that augmenting percutaneous ethanol injections with the ECethanol treatment improves therapeutic efficacy and safety compared to injection of pure ethanol. The practical advantage of EC-ethanol ablation is in the reduction of number of treatment sessions and the reduced dose of ethanol needed to achieve the same efficacy. Further, EC-ethanol ablations are ultra-low-cost, allowing their use in low-and high-resource settings alike in the absence of bulky, expensive machinery. Thus, ECethanol may provide an additional line of treatment enabling more patients to access appropriate treatments.

Supporting Information Captions:
Supplemental Table 1