Axial Discovery - Image-based cell sorting

Call for talented individuals and teams

Build with Axial: https://axial22.axialvc.com/

Axial partners with great founders and inventors. We invest in early-stage life sciences companies such as Appia Bio, Seranova Bio, Delix Therapeutics, Simcha Therapeutics, among others often when they are no more than an idea. We are fanatical about helping the rare inventor who is compelled to build their own enduring business. If you or someone you know has a great idea or company in life sciences, Axial would be excited to get to know you and possibly invest in your vision and company . We are excited to be in business with you - email us at   info@axialvc.com

Image-based cell sorting

A picture is worth a thousand words and for biology, better images have powered breakthroughs in structural biology, drug development, and more. Over the last 2 decades with leadership recently from Keisuke Goda and Daniel Schraivogel, the field of image-based cell sorting has made significant progress from 10s-100s of cells per second to tens of thousands and maybe millions.

Amnis was one of the first movers in this field founded in 1999 and acquired in 2011 by EMD Millipore. Their imaging flow cytometry devices are still on the market and can sort ~5000 cells/s. However, they are limited in their ability to sort in real-time and have a multi-second lag due to image processing done in parallel to sorting. As a result, the device never really scaled in its ability to sort 10Ks of cells per event and beyond. Advances in neural networks (CNN) over the last 7 years has enabled the ability to sort cells (i.e. real-time actuation) by image more precisely and with a higher throughput. The entire workflow - data acquisition -> image reconstruction -> analysis -> sorting - is integrated and happens in microseconds. The potential is to have a sorter that is only limited by software and not hardware.

There are various terms for image-based sorting with neural networks: image-activated sorting (IAC), image-enabled cell sorting (ICS), but the key advance is merging flow cytometry and microscopy with machine learning. Fluorescence-activated cell sorting (FACS) has been a workhorse for immunology over the last 5 decades since its invention in the 1960s. The tool relies on extracellular markers and corresponding antibodies conjugated with fluorophores. Then the FACS machine gates on a particular wavelength to sort different cell types. However, FACS can only capture 10s of fluorescence signals and can capture very little spatial information. As a result, using imaging to sort cells could increase the resolution of sorting - a ~1,000x improvement in content/signals measured and capturing spatial relationships.

So what could this enable? So far image-based cell sorting can massively increase the efficiency of isolating cells across different development cycles. Then IACS has been combined with genetic screening to study combinatorially the effects of knockouts/downs on subcellular processes and morphology. The interesting applications are around cell therapies. Mainly around figuring out active ingredients. When Amnis came to market the only real use-case was in tools. Diagnostics might also make some sense particularly in cancer and genetic screening at birth. But over the last few years, the success of CAR-T cell therapies has opened up a new market for image-based cell sorting. There is potential to build new platforms that rely on images to formulate and discover new cell therapies.

FACS is still pretty useful for cell therapy research and about 10x lower cost than a IACS machine. Then the scale of magnetic sorting (1Bs of cells/s) is needed for clinical applications. Image-based approaches are reaching parity to FACS (10Ks of cells/s) and have a shot, driven by software, to get to billions of cells. The key part here is matching the right business model to IACS. A tools approach probably won’t work - sales cycles are too long, not that many use cases known, device cost. As a result, image-based cell sorting can only support a drug development company right now.

The next question is then for what ends? Where is multi-dimensional data valuable in cell therapies? There might be opportunities in studying neutrophils given their short half-life. But they probably won’t make good medicines to engineer. The killer application might be in systematically screening immune cells in a large set of cancer patient biopsies and discovering new cell types that FACS and scRNA-seq methods would miss out on. The sorting would need to be gated on some sort of functional proxy. Then engineering immune cells isolated from the tumor microenvironment (i.e. T-cells, NK cells) into drug candidates. The functional readout is the key here to make a fishing expedition worthwhile.