Many drugs fail to make the transition from preclinical to clinical studies and do not prove to be effective in phase 2 studies despite promising reports in preclinical models. Likewise, many drugs are discarded based on lack of activity in preclinical models but may have activity in vivo. This contributes to the inefficiency of drug development and requires better models that more accurately recapitulate the host disease state including models that reflect drug resistance or particular mutations as seen in patients [1
]. Failure to respond to treatment as well as drug resistance to anti-cancer drugs are critical challenges in clinical oncology that can be ascribed to factors like genetic mutation [3
], oncogenic amplification [6
], and other changes in the tumor cell machinery that result in changes in the uptake capability, metabolism of drugs, and removal of drugs and drug metabolites from the cells. Because of such dynamic evolving factors in cancer cells, resistance to drugs can arise quickly and has been documented for most commonly prescribed anti-cancer treatments, often with strikingly high rates, but varying widely between patients. Such variable response to therapy is currently addressed through precision medicine (PM) by relating genetic mutations and expression profiles to therapeutic options. However, in practice, even after the identification of probable targets, modification of a predetermined traditional treatment strategy is still rare given the lack of empirical data demonstrating the efficacy of a targeted therapy. Moreover, despite its potential, the clinical efficacy of PM-driven treatment is typically poor [8
]. Although the statistics vary with disease, in general, only 11% of patients entered into PM programs experience a change to an alternative therapeutic; of those patients who do receive an altered treatment, only 3% experience what could be considered a significant improvement in outcome compared to traditional therapy [9
]. This means that current PM tools used to predict and drive treatment decisions are not much better, if any, than the standard approach and thus needs to be improved.
Two dimensional (2D) cell culture techniques have provided biological models that have resulted in numerous scientific advances and discoveries. However, these culture techniques fail to accurately represent the three dimensional (3D) biology of the in vivo tumor microenvironment [10
]. Cells placed into 2D environments experience completely foreign topographies, substrate mechanical properties, cell–cell and cell–matrix interactions, and diffusion kinetics of nutrients and oxygen in comparison to bioengineered 3D models. Instead, 2D cell culture conditions, such as plastic surfaces, can induce significant changes in cells at the genetic and phenotypic level, resulting in experimental results that may not be truly representative of in vivo biology [11
]. In earlier studies, members of our team demonstrated that when metastatic colorectal cancer cells were maintained in 2D culture, they took on a clear epithelial phenotype, but when incorporated a 3D organoid environment they quickly adopted a phenotype that appeared both mesenchymal and metastatic, which was accurately reflective of their in vivo tumor of origin [13
]. Numerous cell line-based tumor organoids and tumor-on-a-chip models have been created by our group and others that show the importance of the 3D microenvironment.
This discovery has led to the creation of 3D organoid models created from patient tumors, filling a critical scientific gap, to facilitate drug screening studies that can provide patient-specific empirical data to better predict a patient’s drug response. Typically, we combine organ micro-engineering [15
], often with microfluidics [13
], with tissue/tumor-inspired extracellular matrix (ECM) biomaterials, to produce more biologically relevant organoid models. This 3D organoid biofabrication approach combines multiple facets that support improved recapitulation of certain in vivo conditions, including architectures in 3D rather than 2D, cell–cell and cell–ECM interactions, circulatory systems if relevant, and even integration of a subset of immune cells should the application demand it. Our lab has succeeded in creating such patient-derived tumor organoids (PTOs) and tumor-on-a-chip models from lung, mesothelioma, melanoma, colorectal, appendiceal, glioblastoma, sarcoma, and several rare tumors, and have deployed these models in chemotherapy and immunotherapy drug screens [20
]. While certainly useful for disease modeling, mechanistic study, and drug development, we have also used such models in a diagnostic sense to influence therapy, which is perhaps the ultimate goal.
There has been a challenge in deploying PTOs in high-throughput formats and maintaining consistency. Most current PTO systems are inconsistent in terms of size, geometry, and cell density; have poor take rates, fail to form spheroids in hanging drop scenarios, are not amenable to high-throughput formats, and rely on Matrigel-induced self-organization/differentiation. ECM bioinks and bioprinting offer a potential solution.
Bioprinting is a versatile technology with potential for a wide variety of applications in regenerative medicine and tissue engineering. Bioprinting can be described as computer-controlled additive biofabrication with the potential to build or pattern viable organ-like or tissue structures in 3 dimensions (3D) using cells and biomaterials. To date, complete human-sized organs have not been printed, but this remains the ultimate goal. Currently, bioprinted constructs have been implanted in animals [24
], and small-scale bioprinted tissue constructs, or “organoids”, are being implemented in a number of applications. These include disease modeling, drug and toxicology screening, and, recently, personalized medicine in cancer [17
]. One of the major problems that the field of bioprinting, is that few advances have occurred in regard to approaches to the printing process itself, or generation of novel, more user-friendly bioinks. Unfortunately, many bioprinting studies are somewhat repetitive—falling back on traditional biomaterials and their crosslinking approaches, which were never developed to be bioprinted or to accurately represent the complexities of the native ECM. Our laboratory has over a decade of experience in developing ECM-derived bioinks with inherent characteristics to improve printability [29
]. These have ranged from multi-step crosslinking reaction bioinks to, recently, thixotropic bioinks that allow for the simple introduction of cells within the bioink precursors and maintenance of the mechanical properties necessary to ensure extrusion, while still supporting deposition of free-standing 3D structures [17
]. We have recently been working to create modular, defined bioinks with key ECM components that boost cell viability, phenotype, and function of healthy primary cells and patient-derived tumor cells [25
]. Additionally, in recent years we have seen various laboratories pushing the bioprinting field by developing novel methodologies of bioprinting that are not simply inkjet and extrusion bioprinting. This includes laser-induced forward transfer bioprinting [35
] and freeform reversible embedding of suspended hydrogels (FRESH) bioprinting [36
] (from which we certainly have taken inspiration), among others. Advances in methodologies such as these, paired with improved bioinks developed specifically for bioprinting, are opening up new opportunities for bioprinting-based applications.
Here, we describe a 3D bioprinting approach—“immersion bioprinting”—which aims to mitigate the limitations that have plagued tumor organoid systems as we described above. Realization of this technology that can fabricate PTOs in a consistent and high-throughput fashion will provide a valuable ex vivo/in vitro tool that can be deployed for many subsequent studies, including target discovery, mechanistic investigation of tumor biology, drug development, and personalized drug screens to aid in treatment selection in the clinic.
2. Materials and Methods
2.1. Hydrogel Bioink Formulations and Preparation
The collagen–hyaluronic acid (HA) bioink was made using methacrylated collagen and thiolated HA and was prepared using the following steps. Methacrylated collagen type I (Coll-MA, Advanced BioMatrix, Carlsbad, CA, USA) was reconstituted with 20 mM acetic acid according to the manufacturer’s protocol to produce a concentration of 6 mg/mL. Immediately prior to use, 1 mL Coll-MA was neutralized with 85 µL of neutralization buffer (Advanced BioMatrix) and mixed with thiolated and heparinized HA (Heprasil®, ESI-BIO, Alameda, CA, USA). The HA component was first dissolved in sterile water containing 0.05% w/v of the photoinitiator 2-Hydroxy-4′-(2-hydroxyethoxy)-2-methylpropiophenone (Sigma, St. Louis, MO, USA) to make a solution 1% w/v HA. The final hydrogel was comprised of Coll-MA and HA at a ratio of 1:3, in which the thiol groups of the HA and the methacrylate groups of the Coll-MA were crosslinked. As a bioink, typically this formulation is allowed to partially crosslink through thiol-methacrylate crosslinking, resulting in a soft hydrogel that can be further crosslinked by methacrylate-methacrylate photopolymerization at a later point after printing.
As a comparison, the HyStem–HP hydrogel product was employed and prepared as previously described [38
]. Briefly, a thiolated HA component (Heprasil®
), a thiolated gelatin component Gelin-S®
), and polyethylene glycol diacrylate crosslinker (PEGDA, Extralink®
) were dissolved separately in sterile water containing 0.1% w/v of the photoinitiator 2-Hydroxy-4′-(2-hydroxyethoxy)-3-methylpropiophenone (Sigma) to make solutions containing 1% w/v. These solutions were then mixed together in a 2:2:1 ratio by volume, respectively, for immediate testing.
2.2. Immersion Bioprinting Evaluation
Two commercially available bioprinters were employed to evaluate the compatibility of the collagen–HA hydrogel and the HyStem–HP hydrogel—the Cellink Incredible bioprinter (Cellink, Boston, MA, USA) and the Allevi Allevi2 bioprinter (Allevi, Philadelphia, PA, USA). First, simple prints were performed to assess the gcode that was written, and the printers’ capability to correctly deposit material in the center of each well of 96-well plates. This was performed simply by using printed water droplets to assess print accuracy. Once gcode–printer compatibility was established, hydrogel-only constructs were bioprinted using custom-written gcode to drive the bioprinters (Supplementary File S1
). The 96-well plates were prepared by filling each well with 150 μL of a 10 mg/mL gelatin solution. Each hydrogel was then deposited by the bioprinter in approximately 20 μL volumes in each well of the 96-well plates. Printing procedures were evaluated largely visually and quantitatively, and the total time of printing was recorded.
Additionally, the rheological properties of the collagen–HA hydrogel and the gelatin immersion baths were assessed to determine whether we were employing optimal printing conditions. The bioink as well as 3 gelatin bath concentrations (5, 10, and 20 mg/mL, stored at 4 °C until printing) were analyzed as previously described [32
] using a TA instruments DHR-2 rheometer (TA Instruments, New Castle, DE, USA) with a 25 mm plate and 25 mm 2° cone system with 120 grit sandpaper intimately bonded to the surfaces. This addition mimics a roughened geometry for better adhesion to the bioink, preventing slippage during testing. The rheological test was a simple strain sweep from 1% to 1000% shear strain (γ), during which the storage modulus (G
′) and the loss modulus (G
″) of each material were recorded.
2.3. Cell Line Culture
For the initial analysis of hydrogel biocompatibility, two common cell lines were employed. Human liver cancer cell line (HepG2, HB-8065; American Type Culture Collection (ATCC), Manassas, VA, USA) and human colorectal cancer epithelial cell line (Caco2, HTB-37; ATCC) were cultured in Dulbecco’s Modified Eagle Medium—High Glucose (4.5 g/L) (DMEM-HG; Lonza, Benicia, CA, USA) supplemented with 10% fetal bovine serum (FBS; Hyclone, Logan, UT, USA), 1% L-glutamine (Hyclone) and 1% penicillin/streptomycin (P/S; Hyclone) at 37 °C with 5% CO2. Cells were used at 90% confluence.
2.4. Cell Line Organoid Immersion Bioprinting
The cells were trypsinized with 0.05% of trypsin (Hyclone) and counted. Then, collagen–HA hydrogel precursors at a concentration of 10 million cells pTissue constructs of each model cell line (HepG2 and Caco2) were created by suspending ter mL, after which the cell-containing suspensions were drawn into printhead syringes. Prior to cell preparation, 96-well plates were prepared by filling each well with 150 μL of the 10 mg/mL gelatin solution. Well plates were placed at 4 °C until needed.
Hydrogel–cell organoids were bioprinted using custom-written gcode to drive the bioprinters (Supplementary File S1
). The printhead needle traveled into the gelatin bath of each well and deposited approximately 20 μL of cell-laden bioink before moving to another well and repeating. A UV lamp (365 nm, 18 W/cm2
, Dymax BlueWave 75, Dymax Corporation, Torrington, CT, USA) was then employed to irradiate each well for about 1 s each, initiating the crosslinking reaction between remaining thiol groups and methacrylate groups and between methacrylates groups themselves. Following crosslinking, well plates were transferred to 37 °C briefly to decrease the viscosity of the gelatin, after which it was aspirated from the wells and replaced with fresh cell culture media.
2.5. Viability and Proliferation Analysis
Viability of HepG2 and Caco2 organoids following immersion bioprinting in both the Inkredible and Allevi2 bioprinters was assessed on days 1, 3, 5, and 7 (n = 3) using a standard live/dead cell viability kit (Thermo Fisher, Waltham, MA, USA) according to the manufacturer’s recommendations. Stained organoids were imaged using macro-confocal microscopy (Leica TCS LSI, Leica Microsystems, Buffalo Grove, IL, USA) to visualize viable cells (green) and dead cells (red). This study was also applied to A549 lung epithelial cell-based organoids, of which the cells were cultured in the same manner as the other cell lines.
Proliferation of cells within the different organoids was evaluated on days 1, 3, 5, and 7 (n = 3) by quantification of mitochondrial metabolism with a CellTiter 96® Aqueous One Solution Cell Proliferation Assay kit (Promega, Madison, WI, USA) according to the manufacturer’s protocol. Absorbance was quantified on a Varioskan Lux multimode plate reader (Thermo Fisher) at 490 nm. Cell number was proportional to the absorbance signal.
2.6. Patient-Derived Tumor Biospecimen Processing
Two glioblastoma (GBM) biospecimens and one sarcoma biospecimen were obtained from 3 surgically treated patients in adherence to the guidelines of the Wake Forest Baptist Medical Center IRB protocols. All biospecimens were completely de-identified prior to use. The specimens were placed in RPMI and transferred fresh to the laboratory by a dedicated tissue procurement manager. Clinical information was not shared with the lab with the exception of the type of tumor and type of prior treatments (if any). Once received, biospecimens were washed in phosphate-buffered saline (PBS) with 2% penicillin-streptomycin for three 5 min cycles. Tissues were individually minced and placed into serum-free low glucose DMEM with 2% penicillin-streptomycin and collagenase (Vitacyte, Indianapolis, IN) and protease (Vitacyte) for 2 h on a shaker plate in 37 °C. Media for glioblastoma biospecimens was also supplemented with 3.3 mg/mL hyaluronidase (STEMCELL Technologies, Seattle, WA, USA). Digested tissues were neutralized with cold serum-supplemented high glucose DMEM and then filtered through a 100 μm cell filter and centrifuged to create a cell pellet. Plasma and non-cellular material were removed, and the pellet was re-suspended in 1 mL BD PharmLyse (BD, San Diego, CA, USA) with 9 mL deionized water for 15 min protected from light. The conical was then centrifuged at 200 rpm for 5 min and the pellet was suspended in media and counted. The cell suspension was then centrifuged to a pellet and then suspended in PBS along with dead cell labeling solution from a dead cell removal kit (Miltenyi Biotec, Germany). The cell suspension was then sorted using a magnetic column to remove the dead cells labeled with microbeads. The effluent cell suspension was centrifuged at 1200 rpm for 5 min and the pellet was suspended in media and counted for use. It should be noted that cell suspensions are not subjected to any cell sorting or isolation protocols so that tumor heterogeneity can be best preserved.
2.7. Patient-Derived Tumor Organoid Chemotherapy Screening
GBM 1, GBM 2, and sarcoma PTO sets were prepared in 96-well plates by immersion bioprinting in the same manner as described above for the cell line organoid studies. Bioprinted PTOs were maintained in DMEM-HG (Lonza) supplemented with 10% FBS (Hyclone), 1% L-glutamine (Hyclone) and 1% penicillin/streptomycin (P/S; Hyclone) at 37 °C with 5% CO2.
Proof-of-principle drug screening experiments were initiated on day 7 following PTO immersion bioprinting and involved administering the drug compounds to the organoids and incubating under these conditions for 72 h. For GBM PTOs, drug response was assessed for dacomitinib (Selleckchem, Houston, TX, USA) at 6, 600 M, and 60 μM concentrations and an experimental p53 pathway activator (NSC59984, Selleckchem) at 1 μM, 10 μM, and 100 μM concentrations. For sarcoma PTOs, drug response was assessed for imatinib (Selleckchem) at 5 μM, 50 μM, and 500 μM concentrations and doxorubicin (Sigma Aldrich) at 10, 1, and 100 μM concentrations.
Following the 72-h incubation of PTOs with drug compounds, relative viability was assessed by quantifying adenosine triphosphate (ATP) activity using Celltiter Glo 3D assays and measuring relative luminescence using a Varioskan Lux multimode plate reader (Thermo Fisher). Cell number was proportional to the luminescence signal.
Organoid and tissue chip technology has advanced incredibly quickly in the last decade to the point where clinical deployment is likely inevitable. However, to be considered realistic for clinical diagnostics and personalized medicine, these technologies need to be incredibly consistent, reliable, scalable, and user-friendly. For organoids with more complexity than simple hanging drop-style spheroids, this has been a challenge. Here, we have posited that bioprinting, and more specifically, immersion bioprinting, can be used to overcome this challenge. With advances in bioprinting hardware, software, functional ECM-derived bioinks, and modifications to printing protocols, bioprinting can be harnessed not only to print larger tissue constructs, but also large numbers of micro-scaled tissue and tumor models for applications such as drug development, diagnostics, and personalized medicine.