Rapid advances in technology have created the realistic possibility of personalized medicine. In 2000, Time magazine listed tissue engineering as one of the ‘hottest 10 career choices’. However, in the past decade, only a handful of tissue-engineered products were translated to the clinical market and none were financially viable. The reality of complex business planning and the high-investment, high-technology environment was not apparent, and the promise of tissue engineering was overstated. In the meantime, biologists were steadily applying three-dimensional benchtop tissue-culture systems for cellular research, but the systems were gelatinous and thus limited in their ability to facilitate the development of complex tissues. Now, the bioengineering literature has seen an emergence of literature describing biofabrication of tissues and organs. However, if one looks closely, again, the viable products appear distant. ‘Rapid’ prototyping to reproduce the intricate patterns of whole organs using large volumes of cellular components faces daunting challenges. Homogenous forms are being labelled ‘tissues’, but, in fact, do not represent the heterogeneous structure of the native biological system. In 2003, we disclosed the concept of combining rapid prototyping techniques with tissue engineering technologies to facilitate precision development of heterogeneous complex tissue-test systems, i.e. systems to be used for drug discovery and the study of cellular behaviour, biomedical devices and progression of disease. The focus of this paper is on the challenges we have faced since that time, moving this concept towards reality, using the case of breast tissue as an example.
(a) Tissue engineering
Tissue engineering has been traditionally thought of as the development of biologically based replacement tissues and organs for the repair or restoration of tissue/organ function; that is, tissue engineering evolved for surgical end-use. Autologous tissue-engineered devices are formed by deriving normal, healthy cells from a patient, expanding them in culture, combining the cells with an absorbable or degradable material, implanting the material and monitoring the development of tissue as the material either absorbs or is degraded in the body. The material, usually termed a scaffold or matrix, is generally porous or gelatinous in nature, such that the cells may be incorporated within the substrate and not solely on the surface. The scaffold provides form, and sometimes function, for the tissue and may also provide topographical or biochemical signals (either naturally released or designed to release) to the developing cells.
(b) Tissue-engineered test systems
Animal models provide a three-dimensional environment for cell growth, but they are expensive and complex, and it often becomes difficult to isolate a single variable (Yamada & Cukierman 2007) or to ask a very specific question and design a well-controlled, repeatable experiment. Additionally, animal models may not provide a suitable biochemical and/or biomechanical representation of the human organ or tissue in question; even well-constructed human clinical trials are fraught with patient variability, resulting in therapies that are suited to the ‘norm’ rather than the individual. The original concept of engineering cellular scaffolds for tissue repair (Yannas et al. 1982) was subsequently expanded to include the engineering of two-dimensional tissues for use on the laboratory bench (Bell et al. 1988). More recently, the concept of engineering a three-dimensional complex tissue or organ for benchtop use has emerged (Griffith et al. 1997). That is, tissue engineering can be used to provide three-dimensional tissue systems with which one can test new ideas regarding cellular behaviours or the disease process, or with which one can develop new therapies and vaccines. Tissue-test systems are not likely to replace animal or human subjects; however, they have the potential to provide substantive, detailed information regarding very specific conditions—i.e. the test-system environment can be designed and repeatably controlled to answer very specific individualized questions.
One of the first engineered test systems was Test Skin, a two-dimensional engineered full-skin equivalent, designed for toxicity testing (Bell et al. 1988); following the development of Test Skin, however, attention in the engineering domain remained largely focused on tissue engineering for regenerative medicine applications. The cell biology literature contains many examples of three-dimensional environments (Yang et al. 1979) and the enhanced environment that they provide for cell culture; however, these traditional three-dimensional environments are largely gels with homogenous physical consistency and are therefore neither representative of a heterogeneous tissue, nor is the intended purpose to engineer complex tissue.
Synthetic gels have been manufactured, but even those with microfibrillar structures are homogenous throughout. Most anchorage-dependent cells will not perform in a biologically relevant manner when suspended within a gel. Cell attachment to materials is correlated to, among many other factors, the stiffness or modulus of the material as well as attachment area (Aydelotte et al. 1998; Rowley et al. 1999; Ingber 2008). Gels do not always provide sufficient stiffness or surface area for cell attachment and spreading to occur. Likewise, gelatinous cultures are often hindered by transport limitations, which lead to deficiencies in the movement of nutrients, wastes and biochemical molecules. Consequently, we first suggested that the positive attributes of a gel could be enhanced by incorporating smaller, more mobile units of rigid scaffold materials (Burg et al. 2000; Burg 2006) to accommodate a diversity of cells and structures to form a tissue-engineered composite. We further postulate that, through the use of biofabrication, these mobile units may be distributed in a heterogeneous manner throughout a gel to form a biologically relevant engineered tissue (Burg & Boland 2003). We and others have demonstrated the use of mobile units, in the form of beads and granules, to form homogenous cellular composites for a variety of applications (Burg et al. 2000; Eiselt et al. 2000; Halberstadt et al. 2000; Marijnissen et al. 2000; McGlohorn et al. 2004; Brown et al. 2005; Burg 2006); the challenge remains in using select configurations to form heterogeneous, complex tissue structures.
(c) Breast cancer test systems
In vitro breast-tissue models that are structurally relevant with respect to the biological environment have enhanced potential for use in testing regimens of breast cancer drug therapies and vaccines. Furthermore, researchers can use these engineered-tissue models to better understand complex cellular pathways and behaviours, e.g. the metastasis of tumour cells. Using three-dimensional breast-tissue systems, breast cancer cells and normal human mammary cells can be observed in a variety of relevant conditions to deduce why variations in subsequent cellular processes and matrix mechanics lead to cancer progression and, thus, what approaches will be most likely to lead to prevention.
The current ‘gold standard’ for tumour modelling involves the suspension of cells within gel-like matrices, mainly Matrigel (Timmins et al. 2005; Sasser et al. 2007); though these three-dimensional systems add considerable information beyond that gleaned from traditional two-dimensional models, they lack the rigidity necessary to allow for normal functioning of anchorage-dependent constitutive breast-tissue cells, specifically adipocytes (Discher et al. 2005). Matrigel, also generically termed ‘reconstituted basement membrane’, is a commercially available material consisting of a combination of basement membrane-like proteins including type IV collagen and laminin. Because the Matrigel raw material is extracted from murine tumours, the composition is ill defined, varies from batch to batch and does not necessarily contain all of the components found in native basement membrane.
Current tissue models rarely take into account the heterogeneity of tissue mechanics or structure; likewise, they are constructed with the hope that cells (even immortalized cells that are often phenotypically dissimilar to cells in native tissue) will self-assemble into relevant structures, e.g. mammary-gland-like structures. That is, many model systems are dependent on the three-dimensional organization of distinct heterogeneous tissue structures from simple homogeneous cell suspensions that lack spatial cues (Campbell & Weiss 2007). Studies have shown that physical mechanisms have a strong effect on the final tissue structure (Jakab et al. 2004). When developing new model systems, assembled cellular conformations can be difficult to predict because, as described by the differential adhesion hypothesis, spontaneous tissue patterning is an energy-driven process. Cells will arrange such that the total adhesion energy between cells and between cells and extracellular matrix (ECM) will be minimized (Marga et al. 2007); hence, the cells in a three-dimensional model are unlikely, without external direction, to arrange themselves in a biologically relevant manner representative of a complex tissue.
A three-dimensional in vitro breast model should recapitulate the in vivo behaviour of cells and of ECM components in the mammary gland and surrounding stromal tissue. Therefore, the quest to develop a model that can promote the formation of both ducts and acini remains and is of enormous interest in the realm of three-dimensional tumour modelling. Krause et al. (2008) have attempted to recapitulate the formation of ductal structures within mixed type I collagen and Matrigel matrices by introducing fibroblasts into their culture model. They successfully demonstrated the importance of stromal cells in the organization and maintenance of epithelial structures; however, the limitation of this model in providing complex tissue is the spatial organization of the cells within the structure. For example, the methodology includes seeding a specified ratio of stromal and epithelial cells within the three-dimensional gel matrices, causing both cell types to reside in the same ‘compartment’ with no separating basement membrane. The ductal-like structures that form within the gel are multi-layered, unlike the hallmark single-layer lumens apparent in vivo. In order to build a complex model, the collective culture of stromal and parenchymal cells (e.g. fibroblasts, adipocytes, endothelial cells, inflammatory cells, luminal epithelial cells and myoepithelial cells) will be required. Unfortunately, each of these cell types requires its own unique and extremely dynamic and nonlinear set of physical, biochemical and electrical signals to appropriately stimulate cell activity; hence, the task is not trivial.
Previous in vitro studies have demonstrated the importance of spatial alignment of cells in three-dimensional culture in accurately mimicking the in vivo breast-tissue micro-environment, the latter which can include cell–cell and cell–ECM contacts in addition to physical forces and soluble factors (Falconnet et al. 2006; Hebner et al. 2008). The ability to define cell placement, and therefore spatial location, within a three-dimensional tissue can lead to better predictions of cell function and an overall increase in the stability of cell phenotype (Sodunke et al. 2007). Although epithelial cell self-assembly has proven successful in the creation of rudimentary acini and ductal structures, stromal cells (e.g. adipocytes, fibroblasts and endothelial cells) do not have the innate ability to generate relevant architectural arrangements in traditional three-dimensional culture systems. This limitation is important because evidence indicates that interactions among malignant cells with the surrounding stromal tissue have the strong capacity to influence patterns of tumour growth and metastasis (Iyengar et al. 2003). Multiple researchers have noted that the breast micro-environments of both normal and malignant cells are responsible for controlling phenotypic expression (Bissell & Labarge 2005; Booth et al. 2008).
Over 20 years ago, collagen gels were used to stabilize mammary cell boluses so that paracrine signalling could be evaluated and compared with that observed in two-dimensional monolayer culture (Miller et al. 1989). In other studies (Zhang et al. 2005, 2006), alginate–poly-l-lysine–alginate microcapsules were used to encapsulate oestrogen receptor-positive breast cancer MCF-7 cells. The authors demonstrated that the capsule micro-environment (i.e. liquid versus gelled) had an effect on cell viability and activity. Additionally, when compared with MCF-7 monolayers, encapsulated cells were more resistant to drug regimens; the authors suggest that these micro-encapsulated multi-cellular tumour spheroids will have application in high-throughput drug screening processes. To date, the use of chitosan in three-dimensional breast-tissue model applications has been minimal, though researchers have demonstrated chitosan’s utility as a three-dimensional matrix for MCF-7 cells (Dhiman et al. 2005). Chitosan microbead scaffolds with varying degrees of deacetylation were used in in vitro culture, and the growth response of MCF-7 cells cultured on these three-dimensional scaffolds was evaluated in static conditions and was shown to be superior to that of MCF-7 cells in monolayer culture. Three-dimensional chitosan scaffolds have subsequently been used for initial drug-testing protocols. When MCF-7 cells were grown in these three-dimensional systems, they secreted more cathepsin-D in the presence of tamoxifen, but the drug’s subsequent uptake was inhibited. Chitosan scaffolds are porous and can house spheroids without a loss of nutrient and waste transport capabilities; however, the three-dimensional nature of the culture allows MCF-7 cells to resist tamoxifen treatment more readily than MCF-7 cells in two-dimensional monoculture (Dhiman et al. 2005).
In breast-tissue modelling applications, Matrigel has been used traditionally to induce hollow mammary acini formation via polarization of normal luminal epithelial cells. Debnath et al. (2003) discussed the morphogenesis and oncogenesis of MCF-10A human mammary epithelial cells grown in Matrigel, using an overlay method. They aimed to provide a model that could answer questions pertaining to oncogenes and the ultimate signalling pathways that promote epithelial organization; these explanations could probably not be deduced using standard two-dimensional culture. Micropatterning of Matrigel breast-tissue cultures has been reported (Sodunke et al. 2007), but does not address the heterogeneous nature of breast tissue.
Synthetic polymeric mesh and microbeads have been used independently as bases for adipose growth, but again, these models do not provide a heterogeneous tissue structure (Fischbach et al. 2004; Kang et al. 2005; Sahoo et al. 2005). Indeed, past efforts indicate that, in order to construct a complex breast tissue that is representative of the native tissue, efforts will need to focus on the placement of cells and biomaterials into spatially relevant locations with continued focus on the development of the appropriate micro-environment.
We begin this discussion with a working description of biofabrication as the purposeful assembly of cellular and non-cellular biomaterials. Figure 1 overviews select technical considerations in building functional tissues or tissue-test systems, including defining the desired tissue structure (e.g. cellular makeup and distribution), the expected function of the tissue structure, the biomaterial and biomaterial-assembly methodologies. It is also important to plan the integration of sensing systems, for example, to measure oxygen or lactic acid distribution, especially in a tissue-test system. Microfabrication, a subset of biofabrication, can be defined as the application of prototyping techniques to form spatially arranged biological elements. Conventional cell-seeding methods are inadequate in the development of in vitro tissue-test systems because they involve random placement of cells and, therefore, lack the precision necessary for spatial control. As a result, many studies have been conducted to assess current microfabrication tools, and new technologies are being designed and developed, which can be used to produce spatially arranged culture models. Microfabrication tools allow the creation of select biomaterial surface variations as well as the precise placement of cellular components (Khetani & Bhatia 2006). An additional advantage of microfabrication technology is that small cultures can be manufactured with great accuracy and repeatability, thus minimizing expense and use of scarce reagents (Park & Shuler 2003). Previously explored techniques for micropatterning include microcontact printing, laser guidance, photolithography and bioprinting based on inkjet, laser and piezoelectric technology (Ringeisen et al. 2006). Specifically, agarose stamps were used to print human osteoblasts onto porous tissue-engineering substrates in circular patterns of 200, 700 and 1000 μm diameters (Stevens et al. 2005). Lee et al. (2005) described the use of ‘indirect bioprinting’, where they used solid free-form fabrication to print moulds, which were subsequently used to construct scaffolds for cell seeding. The ultimate goal for scientists studying tissue fabrication at the micron scale is to engineer a microenvironment for each particular cell type within a system, where the convergence of these cell types to accurately express a particular breast cancer phenotype would allow the elucidation of events that promote tumour initiation, invasion and metastasis.
Drop-on-demand microfabrication, often referred to as bioprinting, has been explored as a tool for creating spatially arranged biological systems. The earliest literature on bioprinting technology discusses precise deposition of organic molecules, molecular aggregates, cells and single-celled organisms such as bacterium (Blanchard et al. 1996). The basic concept is shown in figure 2 where an extrudant, ink in the case of an office printer, is fed into a chamber with a nozzle. Pressure is created inside the chamber to eject a drop through the nozzle; printers are categorized according to the means of producing this pressure. The most direct method of creating this pressure is a mechanical displacement; for example, the piezoelectric inkjets have a material that changes shape with the application of a voltage waveform. The thermal inkjet family of printers, in contrast, uses an expanding vapour bubble to create this pressure. Other systems, such as continuous inkjets, use mechanical pumps.
The reality is that, no matter the intended application, inkjet printing has many logistical barriers that prevent it from realizing high-throughput precision fabrication of complex, three-dimensional tissues. Cell printing has been described as a powerful ‘bottom-up’ tissue-engineering approach, but with substantial technical limitations, including nozzle clogging and sustainability of printed tissue constructs (Khademhosseini & Langer 2007). The Hewlett-Packard HP 26 thermal inkjet cartridge used in the HP 500 series consumer inkjet printers is capable of printing viable cells and low-viscosity biomaterials (Roth et al. 2004). Indeed, drop-on-demand inkjet printing systems may be employed to create two-dimensional patterns of a variety of cell types (e.g. primary rat hippocampal and cortical neurons, Chinese hamster ovary (CHO) cells and bovine endothelial cells) (Nakamura et al. 2005; Varghese et al. 2005; Xu et al. 2006) and low-viscosity biomaterials, including collagen and alginate (Boland et al. 2006).
Roth et al. (2004) printed type 1 collagen onto glass coverslips in various two-dimensional patterns, including lines, circles, dot arrays and gradients; smooth muscle cells and primary neurons were pipetted onto these substrates and preferentially adhered to the printed collagen patterns. Similarly, CHO cells and embryonic rat motor neurons were successfully seeded onto two-dimensional agar and collagen substrates. Less than 8 per cent of the CHO cells were lysed during the printing process; the maintenance of motor neuron cell morphology post printing was confirmed by observing extension of cell processes following 7 days in culture (Xu et al. 2006). Similarly, cell viability of human fibroblasts was assessed using an alamarBlue assay after cells were printed using piezoelectric controlled inkjet printing. Changing the amplitude and the rise time of the electrical pulse to the piezoelectric printer appeared to have no statistically significant effect on cellular viability (Saunders et al. 2008). Bioprinting has also been employed to create specific two-dimensional patterns of immobilized growth factors on prepared biomaterial substrates (Miller et al. 2006; Ilkhanizadeh et al. 2007; Phillippi et al. 2008). In summary, the potential of inkjet printing is limited to low-viscosity substances.
2. Establishment of the relevant toolbox and challenges in creating composite tissue systems
The literature clearly indicates a high level of excitement with respect to biofabrication and the big picture potential; however, very little focus has been given to addressing the large number of technological problems preventing the reality of high-throughput precision fabrication of three-dimensional, large volume forms. Our long-term objective is to develop composite engineered-tissue systems, capitalizing on technologies developed for our injectable composite systems for regenerative medicine (Burg et al. 2000; McGlohorn et al. 2004; Brown et al. 2005; Burg 2006; Gomillion et al. 2007). We have focused on building a toolbox of techniques and methodologies to allow the development of tissue systems that are specific to a particular goal—i.e. addressing a particular scientific question or a particular drug discovery objective. Our toolbox development began with a simple proof of concept study, simply demonstrating the importance of biomaterial selection in building an engineered tissue—i.e. the biomaterial is a tool that must be selected carefully, in accordance with the desired goal, and there is no ‘one-size-fits-all’ solution. Our fundamental premise is that the many previous studies performed in gelatinous systems allow the elucidation of general principles, but do not allow direct extrapolation to complex native tissues because gel systems do not provide the physiochemical or mechanical heterogeneity found in native tissue. Accordingly, we showed that anchorage-dependent cells behave very differently on substrates of low modulus versus those of higher modulus and that our composite approach, e.g. embedding beads or fibres in a gel, allows a modular system that can be customized to a particular application (Xu & Burg 2007; Xu et al. 2009a,b). By way of example, figure 3a,b shows 3T3 mouse fibroblasts after 3 days of culture in collagen gel compared with cells grown on collagen-coated beads and embedded in collagen gel for the same time frame. Even during this short time frame, one can clearly see that, although the cells are viable in both systems, the cells in the gel are largely isolated and rounded, whereas the cells on the beads have spread and are bridging the beads. Figure 4 similarly shows that MAC-T bovine mammary epithelial cells printed onto fibres, then embedded in a collagen gel, also behave very differently than those embedded into a gel directly. The morphology of the fibre encourages the cells to align; again, the material provides a tool that one could use if interested, for example, in creating zones of alignment within a tissue system (e.g. a collagenous structure). In contrast to homogeneous gel systems, the composite system is modular and can be tuned according to the specific application.
Our toolbox also includes material-processing instrumentation. By way of example, we will focus on the development of a microfabrication tool, specifically the challenges and limitations of a thermal inkjet approach. Although we will focus on the development of a breast-tissue system, the challenges and limitations are applicable to an array of other biofabrication applications.
(a) Thermal inkjet for high-throughput microfabrication
Specific impediments to high-throughput precision biofabrication are now discussed in the context of a custom-designed and built system (PawPrint) that exploits the thermal inkjet microfabrication technology to create drop-on-demand cell placement (Parzel et al. 2009a). The PawPrint system is depicted in figure 5a, and a photograph of the system is shown in figure 5b. An open-architecture machine that incorporates multiple HP 26 cartridges along with other microfabrication and assembly tools, such as placement of prefabricated beads or fibres, was pursued. The HP 26 thermal inkjet cartridge is the heart of the system and operates through a process in which a heating element inside the print chamber forms a bubble and the pressure from the bubble creation forces a liquid drop to be ejected. As the drop is ejected, the bubble rapidly cools and shrinks; liquid from the reservoir fills the void left by the evacuated drop. The heating elements, small thin-film resistors, are heated with a short, precisely timed pulse of electrical current. The application of current to a nozzle’s thin-film resistor to eject a drop is termed ‘firing’ the nozzle. This drop-formation process occurs in less than 3 μs. The HP 26 cartridge is the consumable supply in the HP 500 series office printing system and is replaced when all of the ink is consumed. Surprisingly, this consumable cartridge contains the most important technology in the system, i.e. the nozzles and drop-generation mechanisms.
The potential of the HP 26 cartridge-based system to print a two-dimensional cellular biomaterial has been established (Boland et al. 2006). Before initial application of a new HP 26 to printing, the cartridge is opened, the ink is drained and the cartridge and print head are cleaned and sterilized. The ink is replaced by a biologically relevant biomaterial, such as cell culture medium, or dilute hydrogel, and these biomaterials are often referred to as ‘bio-inks’. The HP 26 cartridge was identified for potential use in high-throughput three-dimensional microfabrication because of the large number (50) of appropriately sized (diameter 50 μm) nozzles per small area, its low cost and wide availability, the ability to fire the nozzle using custom electronics and its demonstrated success in depositing viable cells in two-dimensional arrays.
A significant limitation in adapting an off-the-shelf printer for use as a microfabrication tool is the fact that the print head only has one degree of freedom and the mechanism that feeds paper is used to produce two-dimensional motion of the paper printing media relative to the print head. This creates some difficulty in adapting such a system for use as a flexible, general-purpose microfabricator. Hence, the HP 26 cartridge and cartridge holder are the only components that we salvaged from the original HP 520C printer; the cartridge firing electronics and motion system have been replaced with more flexible components. The system is capable of a printing resolution of 85 μm, but the repeatability of the positions system is 20 μm and variation in drop placement fired from any given nozzle is approximately 30 μm. The biofabrication system pictured in figure 5b can hold four HP 26 cartridges, where each can have a different cell type or biomaterial.
Adapting and using the HP 26 inkjet cartridge in a biofabrication system with the goal of high-throughput precision fabrication produces many challenges that must be addressed. The challenges are grouped into four categories in figure 6: mechanical, biological, chemical and electrical/physical. The nature of these challenges and specific solutions that we have pursued are discussed below.
(i) Mechanical challenges
We chose to move the print surface instead of the print cartridge, a different approach than that of the traditional inkjet systems. This decision was based upon: (i) the ease of designing a sample holder that moved with the two-dimensional positioning system, (ii) the difficulty of designing cabling to move the print cartridge relative to the electronics that control the print cartridge, and (iii) the flexibility of adding other fixed stations, such as a microscope, to analyse or manipulate the samples. The possible negative to this approach is that the acceleration and deceleration of the sample stage while printing may cause movement of the media or deposited material. The decision to move the printing substrate also affected cartridge calibration. An important feature of the original HP printer that used HP 26 cartridges is that the cartridge is disposable and can be easily replaced by the user. The cartridge snaps into a cartridge holder that provides mechanical alignment within the printer and electrical connection between the printer and the cartridge. In single cartridge printing, the cartridge needs only to be aligned ‘close to’ the absolute print area on the print media, for example, the position on a page in the original printer, but it is the accuracy of relative motions between printed drops that determines the quality of the printed image. That is, it is not necessarily important that a drop be placed accurately with respect to the print area, but it must be placed accurately relative to the other drops that constitute a pattern. This requirement means that the positioning system must accurately move the printing substrate relative to the print head and the cartridge must accurately deposit a drop at the location of the nozzle. The first panel in figure 6 illustrates some of the primary mechanical error sources.
An initial source of error is that the drops do not print directly below the nozzle; a standard deviation of 10 μm was observed at a nozzle height of 2 mm above the substrate. This error cannot be directly compensated, and must be included in the error budget for drop-placement accuracy. In the case of printing a cellular load, the goal is to place a single cell under the nozzle. The position of the cell within a drop also adds to the error. Air drafts and the printer motion also affect placement accuracy.
Single-cartridge printing can be accomplished without an absolute calibration of the printing cartridge to any fixed coordinate system; however, a fundamental challenge arises in using multiple cartridges to deposit biomaterials. The cartridges, more specifically the nozzles, must be placed in an exact known location in order to print accurately from one cartridge relative to the other, but this is not a practical approach given that the cartridges must be replaced frequently. An alternate approach is to find the relative location of the cartridges, depicted in figure 7a, where a fixed microscope is used to find the relative nozzle positions. In this approach, the first nozzle of the first cartridge is used to print a drop on the movable stage. The drop is moved into the microscope field of view; note that the approximate position of the nozzle is known a priori. Matlab image analysis software is used to locate the drop relative to the centre of the microscope field of view. In this way, the vector VMA shown in figure 7a is found. The vector VMA should be interpreted as the position of the first nozzle on Cartridge A relative to the centre of the microscope field of view. The process is repeated for the next cartridge and a second vector VMB is found. From the two locations, VMA and VMB, the relative offset between Cartridge A and Cartridge B can be found as VAB. It is now possible to print an image from Cartridge B by using the position commands that would be needed for Cartridge A plus the offset VAB. Figure 7b shows a checkerboard pattern of yellow and black ink that demonstrates successful alignment of two cartridges. This approach can be extended to any number of cartridges, four in the current PawPrint system, and can be used to replace any of the cartridges.
Calibration is crucial in building a tissue-test system in order to align multiple cell types in spatially relevant locations. To replicate the complex parenchymal–stromal interactions in the breast, one must be able to reproduce the cellular collocations. Figure 8 shows D1 mesenchymal stem cells that have been deposited in defined locations to complement the 4T07 non-metastatic, tumourigenic cellular pattern (Pepper et al. 2009). The cells originated in two independent cartridges; hence, the calibration was crucial to the precise alignment shown below. Biologically relevant patterns may now be produced in order to study the cellular interactions and behaviours.
(ii) Biological challenges
The deposition of a defined number of cells in a controlled manner is crucial to building a tissue-test system, in order to produce a replicate of the complex native tissue structure and therefore to capture the appropriate cellular messaging. Figure 9 shows a typical scenario, where the aggregation of cells within the nozzle results in the deposition of multiple cells within one drop. Time in suspension affects this factor, but we have shown that the addition of agents such as ethylene diamine tetraacetic acid (EDTA) in the media can reduce aggregation. Figure 10a–d histograms demonstrate the frequency of drops that contained a specified number of 4T07 cells, using solutions of 50 per cent Dulbecco’s Modified Eagle’s Medium (DMEM) and 50 per cent Hanks’s Balanced Salt Solution (HBSS), with cell concentration of 3×106 cells ml−1, and a final EDTA concentration of 0.27, 0.53 or 1.06 mM. Comparison of histograms indicates that printing time was extended and number of cells per drop diminished, when 0.27 or 0.53 mM EDTA was added to the 4T07 cell suspension. In fact, the greatest proportion of ‘successful’ drops was printed using a 4T07 cell suspension containing 0.53 mM EDTA. 1.06 mM of EDTA was deemed unsuitable for printing, as this condition encouraged cellular aggregation.
If one wanted to study the purported ‘stealth’ behaviour of a cancer stem cell within normal breast tissue, it would be crucial to accurately place one and only one cell. Placement of the biomaterials alone is not sufficient to ensure that patterned tissue is created. Two factors can contribute to the movement of cells, even if the cells are placed accurately. First, the cells must attach to the substrate in order to create a stable pattern. Figure 11a shows a pattern that was correctly deposited but was damaged when the structure was covered in media for incubation. A significant challenge is to produce a healthy culture of cells that will readily attach to the surface, or conversely provide a surface that will adhere the cells, so that the cells can be cultivated. In addition to changing the surface material, a working solution has been provided with a high-humidity environment so that the cells can loosely attach before flooding the surface with culture media.
A second problem occurs during the deposition of either high-density cell population or multiple cell types. Either of these operations requires that multiple drops be printed at the same location. The amount of medium included with the cellular component can cause local flooding, where cells float away from the desired location. Figure 11b shows a printed pattern that required close placement of two cell types, i.e. printing of one type onto the other. The bottom part of the bow is distorted because of local motion of the cells.
(iii) Chemical challenges
The inkjet printer offers an inexpensive tool for microfabrication; limitations of the technology noted in the literature include the ejection of satellite drops (outliers that disrupt pattern resolution), transmission of heat and shear force to extrudants, and difficulties associated with thorough cleaning of cartridge reservoirs (Barbulovic-Nad et al. 2006). Drying of material at the nozzle of the thermal inkjet cartridge is a well-known challenge, even when printing with standard inks. The ink may dry at the nozzle and can partially or fully block the nozzle. The problem appears exacerbated when printing a biomaterial, such as a cell solution, from the cartridge as nozzle clogging occurs but deposits may also occur within the nozzle chamber. We have set a cartridge printing requirement of one million cells in order to build a three-dimensional tissue of clinical relevance; without technical refinement, the clogging and deposits cause the cartridge to fail prematurely, well short of printing one million cells. We hypothesized that the salts in standard cell culture media present a second problem as they can deposit inside the nozzle chamber (Parzel et al. 2009b). Both problems occur as a function of printing time and printed volume and therefore represent a direct impediment to high-throughput printing.
We provide a solution for increasing the restricted number of drops, and corresponding low number of cells, which can be printed prior to cartridge failure due to clogging (Parzel et al. 2009b). In our work, using the HP 26 cartridge to print a serum-free cell culture medium, nozzles failed, i.e. the nozzles did not eject a drop when fired, after only a relatively short amount of printing time. This failure was generally attributed to clogging by adsorbed proteins (for which reason only serum-free medium is now used) and cellular components, in addition to cell aggregates. We assessed the efficacy of EDTA as an anti-scalant and anti-aggregant in two-dimensional bioprinting applications and demonstrated enhancement of the printer’s high-throughput potential following addition of EDTA to a bio-ink (Parzel et al. 2009b). Each of four extrudants was fired from a single nozzle at 1000 Hz in successive test periods for a maximum of 25 min (1.5×106 total drops) or until the nozzle failed. Figure 12 displays the approximate number of drops that could be printed before nozzle failure. Approximately 1.5×106 (standard error of the mean; s.e.m.=±0) drops could be printed with bio-inks containing 0.53 and 1.06 mM EDTA before failure occurred, while only 1.0×106 (s.e.m.=±1×105) drops could be printed with 0.27 mM bio-ink and only 2.0×105 (s.e.m.=±1.7×104) drops with 0 mM bio-ink. A 0.53 mM solution of EDTA was found to be the best solution tested because it prevented nozzle failure, had the greatest proportion of successful drops printed in the cell ejection study (i.e. it reduced aggregation) and had no statistically significant toxic effects on the cells. The challenge to creating a breast-tissue-test system of value is in rapidly creating a large volume system that will allow the cellular behaviours and interactions inherent to the biological system; hence, the test system must be constructed in high-throughput fashion. EDTA substantially extends the printing longevity of the cartridge and enhances quality of the printed drops, a large step towards the realization of a high-throughput system.
(iv) Electrical and physical challenges
The ultimate objective of the inkjet printing system is to reliably produce and deposit drops of the cellular and acellular biomaterial extrudants. The current waveform in the resistive heater plays an important role in droplet formation; the main parameters that define a square pulse of current (voltage) are shown in figure 13a. A study to demonstrate the effect of pulse width was conducted in which the pulse width was varied from 0.5 to 5 μs and droplets were printed using a new HP 26 cartridge that included HP ink. The results are summarized as follows:
— less than 1.75 μs, drops are not reliably formed. The top pictures in figure 13b illustrate deformed drops,
— 2–2.5 μs yields best drop production,
— greater than 2.5 μs for long periods caused nozzle to stop printing (no visible damage), and
— greater than 5 μs, nozzle is visibly damaged.
Although this study defines a fairly limited range over which the pulse width can be varied, it can be conjectured that this parameter should be optimized to match a specific biomaterial.
The current waveform is repeated in figure 14a to highlight the period as a second parameter that affects the droplet generation process. A range of extrudant viscosities will be necessary to create the varied mechanical environment of breast-tissue stroma. In cases of tumours, a higher viscosity extrudant will be required. Early experiences with the systems suggested a link between viscosity and the frequency of the firing waveform that were explored in a short study (Parzel et al. 2009a). The cartridge was loaded with suspended cells at 7.7 million ml−1 in a HBSS–DMEM solution with added EDTA to reduce clogging. The following scale was created to evaluate the printed drops:
— 1 = good,
— 0.5 = printed with satellites around main drop, and
— 0 = not all drops printed.
The printing firing frequency was set to 1000 Hz (period = 1 ms) with a 2 μs pulse width. A series of 30 000 drops was printed in the first 0.5 min and the results evaluated. Printing continued for 30 s intervals and the results were evaluated at the cumulative printing times of 1, 1.5, 2 and 2.5 min. The cartridge stopped producing drops after 2 min and did not recover at 2.5 min. The frequency was reduced to 33 Hz and printing resumed. A series of 990 drops was printed and the results evaluated using the same scale; results were collected at the cumulative printing times 3, 3.5, 4, 4.5 and 5 min, and are plotted in figure 14b.
Additionally, glycerol was printed in varying viscosities. The frequency (1/period) was increased to determine the maximum firing frequency for each viscosity. Note that 1000 drop s−1 was the frequency limit at which the system could generate firing pulses. The results in figure 15 show a dramatic drop in maximum firing frequency at 72 per cent volume ratio of glycerol in water and that it is not possible to print at viscosities higher than 72 per cent volume ratio of glycerol in water.
The interactions of electrical and physical parameters demonstrated here, coupled with the mechanical, chemical and biological parameters discussed above, suggest a complex optimization to effectively print a given biomaterial. That is, in order to use the HP 26 cartridge for high-throughput microfabrication, it is necessary that the best combination of parameters be determined.
3. Conclusion and perspectives
The role of the biomaterial in influencing the micro-environment is often underappreciated. Biomaterial implantologists have long recognized the profound effects of seemingly simple material properties such as topography, chemistry, molecular weight, modulus and surface charge on cellular behaviour in vivo (von Recum 1999). Indeed, these features are equally important to the design of viable test systems for benchtop use; the material is a powerful player in influencing cellular behaviour. As we cycle through the Technology Hype Cycle (Linden & Fenn 2002) from the Peak of Inflated Expectations (i.e. the ‘wow’ concept of printing whole organs) to the Trough of Disillusionment (e.g. the printer nozzle clogs, what now?), we must maintain focus on addressing and wrestling with the seemingly mundane roadblocks to improved biofabrication methodologies in order to reach the Slope of Enlightenment and the Plateau of Productivity.
Portions of this work were funded by a National Science Foundation Presidential Early Career Award for Scientists and Engineers (BES 0093805), a Department of Defense Era of Hope Scholar award (BC 044778) and an NSF Emerging Frontiers in Research and Innovation award (CBE 0736007).
One contribution of 14 to a Theme Issue ‘Advanced processing of biomaterials’.
- © 2010 The Royal Society