On being the right size: scaling effects in designing a human-on-a-chip

Abstract

Developing a human-on-a-chip by connecting multiple model organ systems would provide an intermediate screen for therapeutic efficacy and toxic side effects of drugs prior to conducting expensive clinical trials. However, correctly designing individual organs and scaling them relative to each other to make a functional microscale human analog is challenging, and a generalized approach has yet to be identified. In this work, we demonstrate the importance of rational design of both the individual organ and its relationship with other organs, using a simple two-compartment system simulating insulin-dependent glucose uptake in adipose tissues. We demonstrate that inter-organ scaling laws depend on both the number of cells and the spatial arrangement of those cells within the microfabricated construct. We then propose a simple and novel inter-organ 'metabolically supported functional scaling' approach predicated on maintaining in vivo cellular basal metabolic rates by limiting resources available to cells on the chip. This approach leverages findings from allometric scaling models in mammals that limited resources in vivo prompt cells to behave differently than in resource-rich in vitro cultures. Although applying scaling laws directly to tissues can result in systems that would be quite challenging to implement, engineering workarounds may be used to circumvent these scaling issues. Specific workarounds discussed include the limited oxygen carrying capacity of cell culture media when used as a blood substitute and the ability to engineer non-physiological structures to augment organ function, to create the transport-accessible, yet resource-limited environment necessary for cells to mimic in vivo functionality. Furthermore, designing the structure of individual tissues in each organ compartment may be a useful strategy to bypass scaling concerns at the inter-organ level.

Figure 1

Microtechnologies to generate multi-compartment organ systems have been demonstrated to effectively conduct hypothesis-driven research into suspected mechanisms of toxicity arising from actions by a few key interacting organs. By extending these approaches to include many organ systems, a “human-on-a-chip” may be developed. Such a system could be used to conduct generalized drug discovery screening, to identify potential toxicities prior to human clinical trials. In both cases, physiologically-based pharmacokinetic and pharmacodynamic (PBPK/PD) models may be applied to determine in vivo relevance, but directly observing toxicities that may occur through unknown and unsuspected mechanisms between yet to be determined inter-organ interactions will require that many organ compartments be scaled appropriately, to accurately mimic physiological interactions. There is currently a lack of understanding of how to perform this “appropriate scaling”. This article reviews scaling approaches that have been used or discussed to date and to introduce some new ideas for appropriate scaling.

Figure 2

Allometric scaling of basal metabolic rate with organism size. (A) Kleiber's law demonstrates that when plotted on a log-log scale, a 3/4 power scaling dependence can be observed for basal metabolic rate of animals across a wide range of sizes. (graphs based on data in Kleiber). (B) Distinct differences in metabolic rate per cell are found between cells within organisms of different size, and cultured in vitro. While cells in vivo scale with a -1/4 power scaling exponent, cell cultured in vitro significantly raise basal metabolic rate and are no longer dependent on the organism mass from which they came. These findings demonstrate that the allometric network is responsible for controlling cell function, and that metabolic rate is not an intrinsic property of the cell (graphs based on data in West et al. )

Figure 3

Two-compartment organ-on-a-chip devices. (A, B) Schematics demonstrating device structure and operation. Cell- or spheroid-laden collagen is polymerized in the collagen chamber. The posts around the chamber prevent the collagen from travelling into the channels. A syringe pump forces media through the system at a volumetric flow rate Q of 1 μL/s, and the pooled perfusate is pipetted away at the end of the experiment. Tissue structure is manipulated by (C) mechanically dispersed spheroids or (D) intact spheroids encapsulated in the collagen matrix.

Figure 4

Characterizing adipose function in two-compartment devices. (A) Sample bright field image of an intact adipocyte spheroid, prior to dispersion. (B) Dispersed cells assayed for viability using a LIVE/DEAD fluorescent kit (Green = calcein AM stained cells, live; Red = ethidium homodimer stained cells, dead. Scale bar = 300 μm). (C) Dispersed spheroids were loaded into the device at varying densities (0× = no cells; 1× = cells dispersed from 0.3 spheroids; 10× = cells dispersed from 3 spheroids), and perfused with either control media, or with insulin-containing media. Insulin prompts significant uptake of glucose by the adipocytes in the 10× condition. (* p < 0.001 by ANOVA as compared to all other conditions; data plotted as means +/− SEM, n = 7-8). (D) Adipose tissue architecture plays a significant role in uptake of glucose in this system. Dispersed cells from three spheroids consume significantly greater amounts of glucose (* p < 0.001, n = 7-8) as compared to three intact spheroids, likely due to transport limitations into intact spheroids.

Figure 5

(A) Schematic of a simplified human-on-a-chip flow distribution. Blood substitute is pumped through an oxygenator and a lung-on-a-chip, before being distributed to multiple organs comprising the system. The pooled liquid is collected and cycled through the system. (B) Distribution of blood flow through the organ compartments. 100% of the volume is passed through the lungs, and then split into different volumetric flow rates, reflecting blood flow distribution in a human at rest (calculated from ref ).