BioMicroSystems

Cécile Bosmans¹, Marcel Karperien¹, Liliana Moreira Teixeira¹, Jeroen Leijten¹

¹Department of BioEngineering Technologies, University of Twente, Enschede, The Netherlands

In vivo, the human body is organized in three dimensions, where tissues are intertwined and arranged in a defined manner. Tissue architecture is inherently linked to its function, therefore the replication of such organization in vitro is essential in academic and pharmaceutical settings. New approach methodologies as alternatives to conventional 2D in vitro systems and animal models are increasingly being recognized by regulatory agencies in a context of high attrition rates in drug development pipelines and a need for robust platforms. Organ-on-chips (OoC) provide defined physiological conditions and microfluidics that allow compartmentalization and environmental control of human tissue features. However, the biofabrication of tissue compartments remains mostly random, and materials of interest are commonly introduced through, e.g., pipetting. The combination of OoC and bioprinting technologies could thus strengthen efforts directed toward faithful tissue biofabrication and more reproducible microarchitectures, thereby yielding more robust platforms. Within this ecosystem, laser-induced forward transfer (LIFT) has been identified as a promising technology to deterministically bioprint biological features within OoC. Here, we describe the mechanisms of LIFT and its ability to deposit single or multiple cells in a spatially controlled manner. Further, we demonstrate the fabrication of standardized OoC platforms compatible with direct LIFT printing.

Sensors are traditionally designed to isolate a specific physical or chemical quantity, while unwanted dependencies and cross-sensitivities are treated as limitations. Machine learning offers a different perspective: these dependencies may contain additional information that can be extracted rather than eliminated.

This presentation explores this concept using a compact 1-D CMOS-MEMS thermal anemometer as an example. Although its outputs depend simultaneously on the speed and direction of the wind, machine learning can disentangle these effects and extract subtle features that are not accessible through conventional calibration. Even small asymmetries in the sensor, normally regarded as imperfections, can become useful information channels.

This approach points towards a broader opportunity for microsensor systems. In lab-on-a-chip and biological sensing, where multiple physical, chemical, and biological phenomena can influence a small number of sensor outputs, deliberately embracing cross-sensitivity could enable simpler and more compact sensor architectures. Rather than requiring a dedicated sensing element for every measurand, a single device could generate a rich, multidimensional signal from which multiple quantities are inferred.

The central question is therefore not only what a sensor was designed to measure, but what additional information is already hidden in its signals—and how machine learning can reveal it.

Osteoarthritis (OA) is a multifactorial degenerative joint disease that involves aberrant mechanical loading, inflammation, and immune interactions resulting in a vicious cycle of inflammation and joint degradation. Organ-on-chip (OoC) models have shown great potential for studying aspects of OA in vitro and have been progressing towards more complicated multi-OoC models of the knee joint: the Joint-on-Chip (JoC). However, studying OA in vitro remains challenging, as most JoC models do not combine inter-tissue communication with relevant mechanical stimulation. Therefore, we aimed to create a modular JoC platform integrating two key tissues: articular cartilage and the synovial membrane. In this presentation, I will discuss how standardization and modular platform technology enabled the development of this JoC platform, integrating tissue-specific OoC models with fluidic circulation and pneumatic mechanical stimulation. I will furthermore present a multiplexed OoC platform for cartilage mechanobiology, made possible by the interoperability enabled through standardization. Together, these platforms illustrate how increasing system complexity can be addressed through standardized interfaces, semi-automated fluidic circuit board design, and microfluidics design automation. More broadly, these developments illustrate how standardized interfaces and design automation methodologies can facilitate the transition from individually developed OoC devices towards modular, integrated, and interoperable platforms for studying (joint) biology in vitro.