Smart 3D bioprinter minds its own business

There are shortages of life-saving organ donors. To address this, researchers are working on 3D-bioprinting (parts of) organs that can take over certain functions in the body. The greatest challenge here is keeping the cells in the printed tissue alive and endowing them with the correct functionality. 3D-bioprinters use a bioink for this purpose — typically a hydrogel containing cells with substances that the body accepts. By curing the gel into the desired structure, cells are fixed in a set position. But determining that position and ensuring that subsequent layers of tissue align correctly requires a great deal of complex, bespoke work.

Now, researchers at Utrecht University have developed a new method that allows 3D-bioprinters to look along step by step during printing. PhD candidate Sammy Florczak developed the technique, called Generative, Adaptive, Context-Aware 3D Printing (GRACE). GRACE uses lasers and artificial intelligence (AI) to determine which cells are present in the printed ink and where they are located. Biomedical engineer Riccardo Levato of Utrecht University led the research: ‘GRACE takes some of the work off our hands by generating designs that are better than anything we could come up with ourselves.’

Optimal blood vessels

Whereas conventional 3D printers often print layer by layer, Levato and his team use volumetric bioprinting. With this technique, the researchers can build entire structures in a matter of seconds. They do this by projecting a series of 2D light projections onto a rotating vial containing cells in a light-sensitive gel. Wherever the light converges, the gel hardens and fixes the cells in place. The result: 3D mini-tissues.

One of the greatest challenges in 3D bioprinting is printing blood vessels accurately. These are needed to deliver oxygen and nutrients to the cells. But because current printing methods rely on a pre-determined 3D design, placing blood vessels correctly is difficult. ‘It remains unclear exactly where the cells will ultimately end up within the gel,’ says Levato. ‘As a result, some cells end up far from the blood vessels and others close by. GRACE can see where the cells are and then designs blood vessels in such a way that all cells are situated no more than 200 micrometres from a blood vessel, thereby receiving sufficient oxygen and nutrients.’

The technology can also seamlessly connect multiple printing steps. For instance, GRACE can automatically scan the surface of a printed piece of bone tissue and print a perfectly fitting layer of cartilage on top of it. It also recognises opaque structures that do not allow light to pass through, and subsequently generates a design that accounts for this. Levato: ‘In this way, we used GRACE to print cells through a stent — a porous, opaque tube used in cardiac procedures.’

Virtual map

The bioprinter ‘looks’ using light sheet imaging. Levato: ‘We scan the rotating vial of bioink by shining a thin sheet of light through it. This allows the printer to create a virtual map of everything inside the vial.’ By making the cells fluorescent, for example, GRACE can use the light sheet to detect precisely where the cells are located.

The researchers demonstrated the considerable advantages of this virtual map by printing pancreatic cells with surrounding blood vessels. They had GRACE design and print an optimised network of blood vessels, then compared this with a randomly printed network of comparable surface area and diffusion properties. The result: the pancreatic cells in the optimised network produced more insulin. Levato: ‘That is precisely what such a printed structure is intended for.’

Clever baby

As a next step, the researchers aim to improve the resolution in order to make individual cells visible. ‘At present, our technique has a resolution of 14 micrometres, which means we can only identify clusters of cells.’

The researchers also hope to make GRACE more autonomous. Although the technique already employs several AI algorithms, machine learning has not yet been incorporated. ‘In this research, we had to provide GRACE with rules ourselves, such as the required distance between cells and blood vessels and the number of endpoints we wanted for the blood vessels. Now we want the technique to learn from various networks, but that will require a great deal of data and training. In that respect, GRACE is still a clever baby — and we want to turn it into a clever adult.’

Florczak, S. et al. (2025) Nature 645, DOI: 10.1038/s41586-025-09436-7