3D Printed Drug-Loaded Implantable Devices for Intraoperative Treatment of Cancer

Abstract

Surgery is an important treatment for cancer; however, local recurrence following macroscopically-complete resection is common and a significant cause of morbidity and mortality. Systemic chemotherapy is often employed as an adjuvant therapy to prevent recurrence of residual disease, but has limited efficacy due to poor penetration and dose-limiting off-target toxicities. Selective delivery of chemotherapeutics to the surgical bed may eliminate residual tumor cells while avoiding systemic toxicity. While this is challenging for traditional drug delivery technologies, we utilized advances in 3D printing and drug delivery science to engineer a drug-loaded arrowhead array device (AAD) to overcome these challenges. We demonstrated that such a device can be designed, fabricated, and implanted intraoperatively and provide extended release of chemotherapeutics directly to the resection area. Using paclitaxel and cisplatin as model drugs and murine models of cancer, we showed AADs significantly decreased local recurrence post-surgery and improved survival. We further demonstrated the potential for fabricating personalized AADs for intraoperative application in the clinical setting.

1. Introduction

Surgery is an integral component of cancer care, and complete removal of the tumor is associated with long-term survival and improved quality of life for many cancers [1–3]. Unfortunately, local recurrence can occur even after macroscopically-complete resection, and is a common cause of morbidity and mortality [3,4]. Moreover, the perioperative inflammatory process may enhance cancer progression in some settings [5–7]. Thus, there is a strong need for treatments and strategies to reduce local recurrence post-surgical resection. Current strategies to reduce local recurrence rely on adjuvant therapies, including systemic chemotherapy and radiotherapy. While both modalities can improve outcomes, these treatments are limited by their toxicity profiles [8]. In systemic chemotherapy, delivery of drug to residual malignant cells remains a major challenge as these cells are generally poorly vascularized and less accessible by agents delivered intravenously. This lack of vascularization also diminishes radiotherapy efficacy [9]. One approach to address this challenge is to deliver chemotherapeutics locally [10,11]. This strategy can be effective if the surgical cavity is well-confined. However, most surgical margins are not confined to a cavity and are often near normal tissue. To enable local drug delivery, the device would need to be able to attach at the surgical margin and provide local sustained drug delivery. The device should be personalized to fit the unique anatomies of each surgical margin. While this is challenging with traditional drug delivery technologies, recent advances in 3D printing and drug delivery technologies offer unique opportunities. Drug containing devices can be engineered and rapidly fabricated in a personalized fashion to fit unique anatomies for intraoperative applications. We hypothesized that we can engineer a drug-loaded device utilizing 3D printing that can deliver chemotherapeutics directly to the resection bed, thus reducing local recurrence while minimizing off-target toxicity. Additionally, such devices can be tailored to individual patients with tunable, localized drug release. Herein, we report the development of a polymer-based, 3D printed device utilizing Continuous Liquid Interface Production (CLIP) [12–14]. Using paclitaxel (PTX) and cisplatin (CP) as model drugs, we examined this device using mouse models of cancer (A431 and 344SQ).

2. Materials and methods

2.1. Materials

HEMA, PEG₅₅₀-DMA, BLS, and LTPO were purchased from Sigma Aldrich. PTX, CP, and Luciferin were purchased from Fisher Scientific. Microcentrifuge filter units were purchased from Corning. H-cell glassware was obtained from Adams and Chittenden Glass. Gibco brand cell culture reagents were purchased from Fisher Scientific. Calcein AM was purchased from Fisher Scientific, #C1430. Propidium iodide was purchased from Fisher Scientific, #ICN19545825. All other chemicals were acquired from Sigma Aldrich.

2.2. Cell culture

A431 human squamous cell carcinoma cells were obtained from the University of North Carolina Lineberger Cancer Center Tissue Culture Facility, which had previously obtained the cells from the American Type Culture Collection (ATCC® CRL-1555^™). 344SQ-GFP/luc murine NSCLC cells were graciously provided by Dr. Shaomin Tian at the University of North Carolina. Cells were cultured in a complete growth medium made of DMEM (A431) or RPMI (344SQ) supplemented with 10% (v v⁻¹) fetal bovine serum and 1% (v v⁻¹) penicillin/streptomycin.

2.3. Animal maintenance

6–8 week old female athymic nude mice weighing 20–30 g were supplied by the University of North Carolina animal facility and maintained under pathogen-free conditions in the Center for Experimental Animals (an AAALAC accredited experimental animal facility). The animal use protocol (20–187.0) was approved by the University of North Carolina Institutional Animal Care and Use Committee and conformed to the Guide for the Care and Use of Laboratory Animals [15].

2.4. Cisplatin Prodrug synthesis

To form the cisplatin prodrug (CPP) (Fig. S1), dry CP (cis-[PtCl₂(NH₃)₂), 200 mg, 6.7×10⁻⁴ mol) was added to a 20 mL scintillation vial surrounded with foil to minimize light degradation. 5 mL DI H₂0 was then added along with a 100-fold excess of H₂O₂ (30% w/v, 7.6 mL, 6.7×10⁻² mol). The mixture was stirred for 1 hour at 600 rpm at 50 °C, resulting in a pale-yellow powder. The scintillation vial was then placed in a 4 °C refrigerator overnight to recrystallize the cis,trans,cis-[PtCl₂(OH)₂(NH₃)₂] in situ. The following day, the solution was transferred to a 15 mL centrifuge tube, the vial ​was rinsed with cold DI H₂O to recover all solids, and ​the liquid was added to the tube. Wash steps were then performed by centrifuging at 200 g for 5 min, aspirating off the resultant supernatant, then resuspending the pellet with 5 mL of a cold wash solution as follows; twice with DI H₂O, once with ethanol, and once with diethyl ether. After the diethyl ether was aspirated, the tube was placed in a foil covered vacuum desiccator overnight. The following day, the dried powder was weighed (165 mg, 4.94×10⁻⁴ mol, 74% recovery) then dissolved in a 20 mL scintillation vial in 10 mL of dimethyl sulfoxide (DMSO). 300 μl of octanoic anhydride (1.0×10⁻³ mol) was then added to give a 2x excess, and the solution was stirred for 48 hours at 600 rpm at room temperature while covered in foil. The resulting yellow liquid was transferred to a 50 mL centrifuge tube, 20 mL of DI H₂O was added, precipitating a light-yellow solid, and was placed in a 4 °C refrigerator for 1 hour. The tube was then centrifuged at 500 g for 5 min, the supernatant was removed, and 25 mL of acetonitrile was added to dissolve the remaining solid. The solution was transferred to a 200 mL round bottom flask and placed in a rotary evaporator at 80 °C rotating at 60 rpm for 1 hour. After cooling, cold ether was added to the round bottom flask and used to transfer all remaining solid to a 15 mL centrifuge tube. This was washed three times per above at 200 g using cold ether. After the final aspiration, the tube was again placed in a foil covered vacuum desiccator overnight. The following day, the dried powder was weighed (153 mg, 2.60×10⁻⁴ mol, 53% recovery) then verified with ¹H NMR and mass spectrometry using positively charged electrospray with a measured m/z of 608.1587 and a 1.3 ppm error with a [M+Na]⁺ adduct ionization with data previously published [16].

2.5. Device fabrication

Devices were designed using Fusion 360 (Autodesk, San Rafael, CA) and made from a resin consisting of 48.5 wt% HEMA and 48.5 wt% PEG₅₅₀-DMA, with 0.5 wt% BLS added as a UV absorber and 2.5 wt% LTPO as a photoinitiator/crosslinker [12]. PTX and/or CPP were added to the resin to result in 5 wt% and 4 wt% resin concentrations, respectively. Model devices were printed with a 6 × 12 mm base in two steps on a Carbon3D S1 prototype printer at a speed of 10 mm hour⁻¹ with a light intensity of 10 mW cm⁻¹. The inert base layer was first printed with a 300 μm thickness using non-drug-loaded resin. Without detaching the inert base layer, the resin was exchanged for the drug-loaded resin and printing resumed over top of the inert layer. The drug loaded portion of the device was printed with a 400 μm thick base and 1.5 mm arrowheads. After printing, devices were first rinsed in deionized water for 5 s, followed by a brief 2 s rinse with acetone, and finally dried under a gentle stream of air until all acetone had evaporated. Devices were then placed in an LED UV oven (365 nm, 90 mW cm⁻²) for post-curing set at 100% for 2 min. This fully cured and crosslinked all remaining polymer throughout the printed network, and ensured complete bonding between the drug-loaded and inert layers.

2.6. Characterization of resin and printed polymer

Thermal analysis differential scanning calorimetry (DSC) measurements were conducted on a Discovery series DSC (TA Instruments, New Castle, DE). Printed materials (~3 mg) were massed into an aluminum pan and sealed with a T-Zero hermetic lid. Samples underwent a heat-cool-heat cycle by heating from 30 °C to 120 °C, cooling to −85 °C, and subsequently heating to 120 °C at a rate of 10 °C/min, under nitrogen atmosphere. Trios software (TA Instruments, New Castle, DE) was used for analysis of glass transition temperature (T_g) at the midpoint of transition.

Photopolymerization kinetics of non-loaded and drug-loaded resins were characterized with photocalorimetry using a Discovery DSC equipped with an AccuCure LED light source with emission at 385 nm (Digital Light Lab, Knoxville, TN). 9–10 mg of resin was added to an aluminum DSC sample pan without a lid and placed in the DSC cell, which was held at a constant temperature of 25 °C under continuous nitrogen flow. After a 5-min isothermal step, sampes were exposed to UV light for 15 min at 6 mW cm⁻². The light was then turned off for 5 min followed by another 5 min UV exposure to ensure the raction was complete. Heat was normalized with a horizontal baseline from the heat flow value at the end of the UV exposure step. Rate of polymberization (R_p) was calculated by $R_p=\frac{|Q|}{\Delta H_{max}}$ where Q is heat flow in mW and ΔH_max is the theoretical heat evolved if all methacrylate groups in the sample are reacted. For these experiments, ΔH_max was calucated as the product of moles of methacrylate in the sample and ΔH for conversion of a methacrylate (60 kJ mol⁻¹) [17]. Bond conversion was calculated by integrating the R_p equation.

To determine the difference in extractable content of non-loaded and dual-drug-loaded AADs, each type were weighed and then placed into a 10 mL solution of 1:1 acetonitrile:2-propanol (ACN:IPA) overnight in a shaker at 150 rpm. They were then removed, dried, and then a final weight was obtained.

To obtain FTIR data, non-drug-loaded and dual-drug-loaded flat devices were printed without barbs to allow insertion into spectrometer. FTIR data was acquired on a Bruker Tensor 27 using a Pike ATR Max equipped with a ZnSe ATR crystal. Data was acquired at a resolution of 4 cm-1, and 16 scans were averaged for both the background and sample scans.

2.7. In vitro release

Device drug release studies were performed with discs fabricated with equivalent surface area and volume as the AAD to enable sink conditions in a 15 L bath. Each disc was weighed and then placed into a microcentrifuge filter unit with the filter removed, which was then placed in a floating rack in 15 L of pH 7.4 PBS and kept under constant stirring at 37 °C. Devices (n = 5) were removed at each time point with the drug extracted in a 10 mL solution of 1:1 acetonitrile:2-propanol (ACN:IPA) for 3 days in a shaker at 150 rpm. The concentration of drug in extracted solutions was determined using an SPD-M20A (Shimadzu Scientific Instruments, Columbia, MD) high performance liquid chromatograph (HPLC) equipped with a diode array detector and a Chromolith Fast Gradient RP-18e 50 × 2 mm column (EMD Millipore, Burlington, MA). Samples (10 μL) were injected into the HPLC and eluted using a binary solvent system (phase A and B, A for Water, B for ACN) at a flow rate of 0.25 mL min⁻¹. The linear gradient program was set as follows: 0 to 20 min, 0% to 100% B; 20 to 25 min, 100% B; 25 to 30 min, 50% B; 30 to 35 min, 0% B. The column and sample temperature were maintained at 30 °C and 4 °C, respectively. PTX (retention time 11.4 min) and CPP (retention time 12.5 min) were monitored at 200 nm. Concentrations of the extraction solution were calculated by comparing peak integrations to standard curves made with known concentrations for each run.

2.8. In vitro H-cell directional release

To evaluate the ability of a non-drug-loaded inert base layer to prevent drug release, AADs with inert layers of 300 μm were printed as previously described. These AADs were printed with oversized flanges and were clamped in place between the two sides of an H-cell, sealing with O-rings on either side (Fig. S2). Parafilm was also wrapped around the AAD edges sticking beyond the O-ring to prevent fluid loss during the experiment as the polymer AADs allowed some fluid flow through the polymer network. Each side of the H-cell was filled with 30 mL of DI H₂0 and placed over a stir plate with a stir bar to increase drug extraction. Aliquots (1 mL) were removed over 48 hours and concentrated 10x through removal of all water via freeze drying and subsequent dissolution with 100 μL of 1:1 ACN:IPA. Samples were analyzed via HPLC as previously described to determine the amount of drug released towards and away from the AAD arrowheads.

2.9. In vitro cytotoxicity

3D tumor models were created and imaged as previously described [18]. A431 cells were cultured as previously described and suspended in fresh medium at a concentration of 5×10⁵ cells mL⁻¹. Cell suspensions were mixed 1:1 with PureCol Ez Col collagen solution (Advanced Biomatrix, Carlsbad, CA) to achieve a final concentration of 2.5×10⁵ cells mL⁻¹. 3 mL of solution was added to a 35 mm dish with a 14 mm coverslip optical glass bottom (MatTek, Ashland, MA). An AAD was placed into the solution with arrowheads down towards the coverslip glass bottom for optimum imaging. Dishes were then placed in an incubator at 37 °C overnight. Cells were stained 12 hours later by adding 6 μL of calcein AM (4 μM, Invitrogen, Carlsbad, CA) and 300 μL of propidium iodide (1 mg mL⁻¹, MP Biomedicals, Santa Ana, CA) then incubated at 37 °C for 30 min. Gels were then washed twice with 1 mL of PBS and imaged on an LSM 700 confocal microscope (Zeiss, White Plains, NY). Tiled z-stack images were obtained from arrowhead tip to base with tiles stitched in ZEN imaging software before exporting. Z-stacks were trimmed in ImageJ (NIH, Bethesda, MA) to remove slices above or below the relevant arrowhead heights, then maximum intensity projections were obtained. Particle analysis was run using 175 × 500 pixel windows moving away from the AAD edge to obtain live versus dead cell counts in both non-drug-loaded and dual-drug-loaded AADs to determine cell death in 0.4 mm steps with a half size window used for the closest 0.2 mm.

2.10. In vivo release

When mice with drug-loaded AADs reached any endpoint, their AADs were excised after they were sacrificed. AADs were also obtained from mice to prevent progressing skin lesions if present. Complete AADs were not always obtained, so portions of AADs that were obtained were weighed to normalize subsequent release measurements per mg of AAD. Extraction and HPLC were performed as previously described.

2.11. In vivo efficacy

Mice were injected subcutaneously with 1×10⁶ (344SQ) or 5×10⁴ (A431) cells in 100 μL of 1:1 Matrigel:plain medium on the left flank. Tumors were allowed to grow until they reached 100–150 mm³ (7 days). All tumors were then surgically resected while mice were under anesthesia. Incisions were made posterior to the tumor, such that a surgical pocket could be blunt dissected between the skin and muscle to fully visualize and resect the tumor. Mice were then placed into either an IV treatment group and had their incision stapled closed with no implant, or had an AAD inserted into the subcutaneous resection pocket with arrowheads facing towards the body and device inert layer towards the skin. AADs were pressed down, engaging the arrowheads to seat the device and then incisions were stapled closed. To reduce surgical pain all mice were given a 0.1 mg kg⁻¹ subcutaneous injection of buprenorphine perioperatively and a 5 mg kg⁻¹ subcutaneous injection of meloxicam once a day for 3 days following resection. IV groups consisted of PBS, PTX only, CP only, or dual PTX+CPP. Dosing of PTX and CP was based on human clinical equivalent doses of 135 and 75 mg m⁻², respectively [19],and using a normal man equivalency of 70 kg and 1.8 m² [20]. This resulted in 3.5 and 2 mg/kg doses in mice given via tail vein injections of 100 μL drug solutions in PBS. Mice with AADs were implanted with an AAD that was loaded with no drugs, PTX, CPP, or dual PTX+CPP; or a dual PTX+CPP flat device without arrowheads. Mice were monitored every two or three days, to detect and measure any tumor recurrence, and body weight was monitored weekly. Tumor volume was measured using calipers in both studies and with an in vivo imaging system (IVIS) using luciferase bioluminescence in the 344SQ luciferase expressing study. Physical manipulation of the loose murine skin permitted the detection of recurrence and measurement of tumors with calipers, despite the presence of AADs. Correlation of physical measurements with luciferase imaging verified measurement accuracy. Recurring tumors had their length (l) and width (w) measured to calculate the tumor volume (v) as $v=\frac{l \times w^2}{2}$.

2.12. Hematological toxicity, hepatotoxicity, and nephrotoxicity

In vivo toxicity was determined using blood collected via submandibular bleeds. Blood was collected from all groups 2 days before treatment, and 4 and 27 days after treatment (n = 7, n = 7, and n = 5 per group on each day respectively). For hematological toxicity, whole-blood (> 35 μL) was stored in a heparin pre-treated stopper covered tube at 4 °C and analyzed for white and red blood cell counts. For hepatotoxicity and nephrotoxicity, whole-blood (> 50 μL) was centrifuged at 2,500 g for 10 min with plasma then pipetted off the top into an Eppendorf tube. The whole-blood and the isolated plasma were analyzed by the Department of Pathology & Laboratory Medicine, Animal Histopathology & Laboratory Medicine Core, University of North Carolina, for blood cell counts, AST, BUN, and Crea levels.

2.13. IVIS imaging

Bioluminescent imaging was performed on luciferase expressing 344SQ-GFP/luc murine models using luciferin and an IVIS Lumina imaging system (PerkinElmer, Waltham, MA). 100 μL of 30 mg mL⁻¹ luciferin was injected intraperitoneally and then mice were placed into a warmed isoflurane anesthesia induction chamber. Mice were then moved into the warmed IVIS imaging chamber with isoflurane nose cones and imaged 8 min after initial luciferin injection using an F stop of 1.2, binning of medium, and a 30 s exposure time.

2.14. Personalized intraoperative devices

To create personalized intraoperative devices, we first used an HTC U11, 12 megapixel, f/1.7 aperture camera (HTC Corporation, Taoyuan City, Taiwan) to photograph a tumor excision area of a mouse. 46 pictures of the target area were taken from different angles and were then imported into Recap Photo (Autodesk, San Rafael, CA). This software created a 3D model based on those images, which then had the target area cropped and was exported to Fusion 360 (Autodesk, San Rafael, CA) where this surface was thickened and had arrowheads added. This custom formed AAD was then fabricated on a CLIP printer and implanted into the target area to demonstrate the accurate results obtained from this process. We then repeated the same procedure after obtaining 37 images intraoperatively during a human abdominal tumor resection. A 3D model of the surgical cavity was made, the target area was again isolated (a convex area in this case) and a custom device was rendered and printed without subsequent implant.

2.15. Statistical analysis

All experiments were performed at least three times and expressed as mean±SD for in vitro or mean±SEM for in vivo studies. Statistical significance for in vitro drug release was determined using two-way ANOVA, in vivo growth comparisons were determined using area under the curve and unpaired t-tests [21], and Kaplan-Meier survival curves using log-rank tests. All tests were performed in Graphpad Prism 8 (GraphPad Software, San Diego, CA). Release studies’ fit lines were modeled as non-linear regressions using one phase exponential association and the formula y = y_max × (1 - e^−kt). AAD cytotoxicity fit lines were modeled as non-linear regressions using a sigmoidal four parameter logistic of the formula $y=y_{min}+\frac{\left(y_{max}-y_{min}\right)}{\left(1+{\left(I{C}_{50}÷x\right)}^{\mathit{\text{Hillslope}}}\right)}$. Differences were considered significant when p < .05 and significance levels in all figures were represented with * p < .05, ** p < .01, *** p < .001, and **** p < .0001.

3. Results and discussion

To generate the design of our proposed drug delivery devices, we identified the following desired capabilities: (1) the device can adhere to a surgical margin without significant migration or detachment, (2) it can encapsulate chemotherapeutics and release them in a controlled fashion, (3) the device can be formulated in real-time to fit surgical margins with different sizes and shapes. To obtain these capabilities, we developed a device using both physical design elements and specific resins. To enable adherence to surgical margins, we employed a barbed needle arrowhead design in our device as other barbed needle designs have been shown to require low penetration forces while offering increased pull-out forces by up to two orders of magnitude, ranging from 25 to 107 mN per microneedle [22,23]. The device has an array of these arrowhead shapes that can be inserted into tissue to secure the device in place and prevent migration and detachment (Fig. 1). The array of arrowheads extends from a contoured base that can be custom designed to fit in the surgical bed. The device also contains a non-drug-loaded inert layer as an integral part of the base, opposite from the arrowheads, that we used to limit the diffusion of chemotherapeutics towards normal tissue. We termed this device an arrowhead array device or AAD. These physical design choices gave us the ability to adhere to tissue and release them in a controlled direction. The next development step was to choose a suitable resin. To integrate chemotherapeutics, we chose a resin we knew to have good drug loading properties that we had previously reported on [12]. This enabled devices to encapsulate chemotherapeutics during fabrication. For the final step in our device development, we needed to formulate devices in real-time to fit surgical margins. We found that Recap Photo offered a way to rapidly image and create 3D computer renderings to quickly obtain models of tumor resections beds, followed by rapid 3D printing available from continuous liquid interface production (CLIP) printers to fabricate devices capable of providing local drug delivery to precise surgical margins. CLIP was chosen over alternative 3D printing methods due to its ability to more rapidly produce precise 3D printed objects at print speeds up to 500 mm/hr, while maintaining high precision and a lack of vertical anisotropy [13,24].

Fig. 1.

Arrowhead array devices. (a) 3D computer rendering of AAD. (b) Photograph of CLIP printed drug-loaded AAD. (c) SEM image (20×) of arrowhead array on AAD. (d) SEM image (100×) of arrowhead needle tip.

To integrate chemotherapeutics within the AAD, we incorporated drugs directly into a liquid resin during the 3D printing process. We chose to use a resin made from hydroxyethylmethacrylate (HEMA) and polyethylene glycol dimethacrylate M_n 550 (PEG₅₅₀-DMA) as both components are commonly used, biocompatible, and crosslinkable; and we have previously performed a material precursor study of this polymer combination, which was shown to offer good drug loading coupled with high resolution CLIP 3D printing [12]. Resins and monomers of PEG-DMA and HEMA are biocompatible over at least 175 days in multiple in vitro cell models including HeLa and HUVEC, with minimal leachable and degradation product cytotoxicity at 37 °C [12]. The resin was supplemented with a UV absorber 2-tert-Butyl-6-(5-chloro-2H-benzotriazol-2-yl)-4-methylphenol (BLS) and photoinitiator ethyl (2,4,6-trimethylbenzoyl) phenylphosphinate (LTPO) to encourage cross linking during printing. We chose PTX and CP as model chemotherapeutics since this combination is commonly utilized and is effective against a wide range of cancers [25–28]. While PTX readily dissolved in the resin, CP formed large aggregates and was insoluble. To overcome this challenge, we synthesized a hydrophobic CPP via conjugation of the fatty acid octanoic anhydride. CPP can be reduced back to cisplatin through intracellular mercaptans which are often overexpressed in tumors, such as glutathione or ascorbate [29,30]. CPP formed a well dispersed suspension and allowed homogeneous device printing. PTX had a maximum solubility of 8 wt% in the resin, but a final loading of 5 wt% was chosen to avoid saturation concerns. CPP was loaded at 4 wt% to offer a similar molar ratio of PTX and CP as commonly used in combination chemotherapy [19].

Using the described design and drug-loaded resin, we designed and fabricated a model AAD with a 6 × 12 mm base and 1.5 mm arrowheads (Fig. 1a). Initial designs included single versus dual layer topologies, varying resin formulations, and varying sizes and thicknesses to obtain a durable model appropriate for murine models. Physical characteristics such as base flexibility and arrowhead fracture resistance were considered, along with drug solubility. SEM images revealed evenly spaced arrowheads and well-formed barbs (Fig. 1c and 1d). These geometries with arrays of undercut barb structures are not readily made by micromolding techniques, necessitating the need to 3D print these structures.

3.1. Analysis of resin and printed material

To further our previous studies of this resin [12], we performed additional material analysis of the liquid resin and its 3D printed crosslinked solid. Thermal DSC performed on printed parts of the unmodified resin showed a T_g of 7.3 °C, in agreement with previous studies [12]. Photo DSC showed similar photokinetics between both plain and drug-loaded resins, with the drug-loaded resin having a slightly higher maximum double bond conversion rate than the plain resin (R_p max (s⁻¹) of .0048 vs .0039 respectively), and similarly reaching a slightly higher total double bond conversion (88% vs 84% respectively) (Fig. S3).

To determine the difference in extractable content of non-loaded and dual-drug-loaded AADs, an overnight extraction was performed which showed that the non-loaded AADs had a minimal amount of extractable content with only a 0.3% weight change, indicating good crosslinking and curing of the resin (Table S1). The dual-drug-loaded AADs conversely had a much higher extractable content, as would be expected with the extractable drug, decreasing in weight by 2.4%.

FTIR analysis of the 3D printed solid materials noted a difference between samples at 1635 cm⁻¹, correlating to C=C bond stretching (Fig. S4). In the plain, non-loaded material, it can be seen there is a minimal peak at this wavenumber, indicating the consumption of the majority of the methacrylate groups during printing. As similar polymerization kinetics and methacrylate conversion was observed by photo DSC of non-loaded and dual-drug-loaded resins, the larger peak in the dual-drug-loaded material can likely be attributed to the C=C alkene group in paclitaxel.

3.2. AADs encapsulating CPP and PTX can directionally release therapeutics in a controlled fashion

PTX and CPP release from AADs was characterized in vitro using scaled, 3D printed discs with an equivalent surface area to volume ratio as the model AADs designed for use in murine models (Fig. S5). These models were used to reliably maintain sink conditions not readily achievable with larger AADs. We found both drugs have first-order release kinetics in PBS at 37 °C, with t_1/2 = 21 days for PTX and t_1/2 = 23 days for CPP. Drug elution time variability on HPLC was less than 1% using pure drugs or drugs extracted from 3D printed devices, indicating no significant degredation of the drugs due to the CLIP printing process.

As mentioned above, AADs are designed to have directional drug release: towards the arrowheads/surgical margin tissue but away from the base/normal tissue. To accomplish this, AADs included a thin inactive layer of non-drug-loaded polymer as part of the base, opposite the arrowheads, to prevent drug release from the base towards normal tissue. To confirm directional release, we utilized an H-cell setup to characterize the drug release in both directions (Fig. S2a). AADs were printed with surrounding edges to seal with O-rings of an H-cell (Fig. S2b). Using an AAD with a 300 μm inactive layer, we found that over 48 hours significantly more drug was released towards the arrowheads (50±3 μg) than from the base (8±1 μg, p < .0001) (Fig. S2c). These results showed that we were able to design a device with directional release in order to protect normal tissue.

To determine the penetration/diffusion of chemotherapeutics from drug-loaded AADs, we examined the cytotoxicity of drug-loaded AADs on tumor cells using a bioengineered tumor [18,31]. This 3D tumor mimic acted as a modified version of a disk diffusion test to enable examining a 3D AAD with cells distributed throughout a hydrogel in which the AAD was completely submerged. We assessed the viability of A431 cells in this tumor mimic upon exposure to non-drug-loaded or dual-drug-loaded AADs. Non-drug-loaded AADs resulted in minimal cell death after 12 hours, while dual-drug-loaded AADs resulted in a region surrounding the AAD with significant cell death (Fig. S6c & S6d). Live and dead cells at varying distances were counted with ImageJ to yield cell death percentages, which were plotted against distance from AAD, revealing a distance response curve analogous to a dose response curve (Fig. S6a). Dual-drug-loaded AADs exhibited 97±4% cell death within 0.4 mm of the AAD, falling off to 58±6% at 2 mm. This was significantly more than non-drug-loaded AADs (p < .0001) which exhibited a maximum of 6±1% cell death over the same distances. This observed pattern of cell death is consistent with cytotoxicity from the released drug, as opposed to the physical presence of the AAD, showing drug-loaded AADs to be capable of drug delivery with significant cytotoxicity within a localized radius while the non-drug-loaded AAD alone does not demonstrate significant cytotoxicity.

3.3. In vivo drug release from AADs varies with geometry

Drug-loaded AADs in the A431 mouse model that were removed for any reason or at study endpoints up to 82 days post-implant were evaluated to determine in vivo release. Residual drug was extracted and compared to baseline non-implanted devices to quantify in vivo release (Fig. S7). One group had been implanted with flat, dual-drug-loaded devices which lacked an arrowhead array to examine variability of drug release in vivo with a lower surface to area volume device. These flat devices were thicker to have an equivalent volume and total drug loading, but with a smaller surface area due to the absence of an arrowhead array. This led to a surface area to volume ratio of 3.0 mm⁻¹ for flat devices and 3.6 mm⁻¹ for AADs, yielding a 22% increase with the arrowhead array. All AADs loaded with either a single drug or both drugs showed similar release profiles, however the flat devices lacking an arrowhead array and with a smaller surface area trended to release both drugs more slowly. Dual-drug-loaded flat devices showed significantly slower release over 60 days in the case of PTX, with dual-drug-loaded AADs achieving 81% release, PTX single-drug-loaded AADs achieving 85% release, and dual-drug-loaded flat devices only releasing 61% in this time. CPP showed a similar trend in AADs, with dual-drug-loaded AADs achieving 81% release, CPP single-drug-loaded AADs achieving 89% release, and dual-drug-loaded flat devices only releasing 74% over 60 days in vivo. This demonstrates that drug release of each drug from AADs occurs over a prolonged period of time and is influenced by the geometry of the device with lower surface area flat devices releasing both PTX and CPP more slowly than their AAD counterparts.

3.4. Drug-loaded AADs have minimal systemic toxicity

Although the AAD is a local drug delivery device, we recognize that some of the therapeutics may enter systemic circulation. To characterize the systemic toxicity of AADs, we examined hepatotoxicity, nephrotoxicity, and hematological toxicity, some of the main dose-limiting toxicities of chemotherapeutics. This was done using mice implanted with drug-loaded AADs and measuring serum enzyme levels of aspartate aminotransferase (AST), blood urea nitrogen (BUN), and creatinine (Crea) along with CBC counts (Table S2). While Crea levels increased in all CP containing IV arms, all levels for IV and AAD treatment arms remained within normal limits.

3.5. Drug-loaded AADs can inhibit tumor recurrence post-surgery

To investigate whether drug-loaded AADs can prevent tumor recurrence after surgery, we employed two mouse models of cancer: a lung cancer 344SQ-GFP/luc allograft and a squamous cell carcinoma A431 xenograft model. To simulate human surgical resections, mice were implanted with a 344SQ or A431 tumor on the left flank. Tumors were resected when the tumor volume reached 100–150 mm³ (Fig. 2a). Surgical resections were performed by veterinary staff who were blinded to the experimental groups, with an equal number of mice randomly placed into each arm each day, across three days as tumor size was reached. Experimental groups received AADs loaded with both PTX and CPP or AADs loaded with a single chemotherapeutic (PTX or CPP). Control groups include AADs with no chemotherapeutics and systemic (tail vein injection) PTX, CP, or both; and PBS.

Fig. 2.

Treatment scheme and A431 tumor mouse model results. (a) Treatment scheme for the evaluation of drug-loaded AADs. Day 0 – Tumor inoculation through subcutaneous injection of 5×10⁴ A431 or 1×10⁶ 344SQ cells in 1:1 matrigel:medium. Day 7 – Tumors of 100–150 mm³ are surgically resected with half of these mice receiving AAD implants in their resection pockets and the other half receiving IV treatments. > Day 7 – Drug release from AADs. (made in ©BioRender – biorender.com) (b & c) A431 tumor mouse model. (b) Individual mouse tumor growth curves showing all drug-loaded AADs significantly reduce tumor recurrence. (c) Survival curve. All drug-loaded AADs showed significant survival benefits over its corresponding IV treatments. (* p < .05, *** p < .001)

As seen in Fig. 2b, all drug-loaded AADs were effective in preventing tumor recurrence in the A431 tumor model. Median survival time of the PBS and single drug IV treatments groups, and the non-drug-loaded AAD group ranged from 36 to 48 days, while the median survival of the dual IV treatment group was 122 days. Median survival of all AAD groups was not reached after 136 days with 6, 7, and 8 mice (n = 8 for each) remaining alive in the CPP, dual, and PTX AAD groups, respectively. PTX-loaded AADs showed greater survival benefits that PTX IV (p = .0005), as did CPP loaded AADs versus CP IV (p = .0344) and dual-loaded AADs versus dual IV treatments (p = .0448). At least 4 or more mice experienced some tumor recurrence in all groups without drug-loaded AADs, while mice with CPP loaded AADs experienced only two tumor recurrences, and the PTX and dual-drug-loaded AADs had no tumor recurrence over 136 days.

The data was further validated in the 344SQ mouse model. We again saw significant delays in tumor recurrence and growth in mice implanted with any drug-loaded AAD as measured by both tumor volume and luminescence (Fig. 3). We did not obtain the same total tumor inhibition with this more aggressive cell line where many mice in the drug-loaded AAD groups suffered from skin ulcerations while still at smaller tumor volumes. Correlation of physical measurements with luciferase imaging confirmed measurement accuracy. These results show that drug-loaded AADs can delay or inhibit tumor recurrence post-surgical resection.

Fig. 3.

344SQ tumor mouse model results. (a) IVIS luciferase fluorescence imaging 12 days post surgical resection showing decreased luminescence in drug-loaded AAD groups. (b) Individual mouse tumor growth curves showing all drug-loaded AADs significantly reduce tumor recurrence.

3.6. AADs can be formulated in real-time for intraoperative application

To demonstrate that AADs can be formulated in real-time for intraoperative application, we printed custom AADs for implantation using real world resection beds in a mouse and a human patient (Fig. 4). We first experimented using a mouse model. A tumor-bearing mouse underwent surgical resection and the surgical bed was imaged using a digital camera from 46 different angles (Fig. 4a). The images were processed using Recap Photo to create a 3D model of the resection bed (Fig. 4b). The implant target area was selected such that it would encompass the entire resection bed, with the rest of the model removed, leaving a 3D surface of only the target area. A custom fit AAD was designed from the selected area of the model using Fusion 360 (Fig. 4c) by adding both a desired base thickness and an array of arrowheads to the target area, resulting in the modeled device as shown in Fig. 4c. Following fabrication (Fig. 4d), the customized AAD was implanted into the resection area (Fig. 4e) demonstrating that this process can indeed create a personalized AAD device to accurately fit a selected surgical margin for intraoperative application. To examine potential clinical translation, we applied the same technique to a case of abdominal tumor resection. Images were taken from 37 locations during a surgical resection of a tumor in the deep pelvis, and used to generate a 3D model of the resection cavity (Fig. 4b). The resection margin perimeter was selected from the rendering, with the resulting target area leading to a convex surface which again had thickness and arrowheads added in Fusion 360 as seen in the human AAD model (Fig. 4c). This AAD was then fabricated (Fig. 4d) but not implanted. This served to demonstrate the capability of CLIP technology to model and print a personalized device intraoperatively.

Fig. 4.

Real-time AADs for intraoperative application. Image sequence from initial photographs to device implantation in mouse and human models. Approximate tumor margins targeted for device implant are circled in yellow. (a) Photographs of mouse tumor resection area and human abdominal tumor resection cavity taken intraoperatively (sets of 37–46 images). Tumor margin is in circled area. (b) 3D rendered tissue model. Previously selected tumor margin is used as a model surface for a 3D drug-loaded AAD. (c) 3D computer model of implant device with needle array towards tissue. Shape was extracted from previously selected tumor margins. (d) 3D printed CLIP fabricated device ready for implant. (e) Device applied within tumor resection area (mouse only at this time).

4. Conclusions

In summary, we have developed an innovative intraoperative device that can deliver local chemotherapy to prevent tumor recurrence after surgery. Moreover, the device is biocompatible and can be engineered in real-time for personalized cancer treatment. We hope to continue this work by optimizing resin chemical and physical properties, developing automated design software, and maximizing print speeds to enable rapid start to finish device production sufficient for intraoperative utilization. Given intraoperative radiation is an established treatment and has shown benefit in preventing cancer recurrence, other intraoperative treatments such as implantable AADs hold high potential for clinical translation and clinical impact. Importantly, our work takes full advantage of the unique properties afforded by 3D printing/CLIP: rapid synthesis, synthesis of complex shapes, and the ability to incorporate therapeutics into printed objects. Our work can facilitate further work in utilizing 3D printing to improve personalized treatment.

Abbreviations

AAD

arrowhead array device

AST

aminotransferase

BLS

2-tert-Butyl-6-(5-chloro-2H-benzotriazol-2-yl)-4-methylphenol

BUN

blood urea nitrogen

CLIP

continuous liquid interface production

CP

cisplatin

CPP

cisplatin prodrug

Crea

creatinine

DSC

differential scanning calorimetry

HEMA

hydroxyethylmethacrylate

IVIS

in vivo imaging system

LTPO

ethyl (2,4,6-trimethylbenzoyl) phenylphosphinate

PEG₅₅₀-DMA

polyethylene glycol dimethacrylate M_n 550

PTX

paclitaxel