The most mature, widespread sorting technology, fluorescence-activated cell sorting (FACS), offers high throughput and sorts predicated on a wide range of phenotypes that can be conveyed through average cellular fluorescence or light scattering levels. However, FACS cannot detect a wide range of phenotypes easily observable through microscopy, such as details of cell morphology, sub-cellular localization, and temporal fluorescence expression with single-cell resolution. Presently, there is no widespread, user-friendly technique to sort cells following microscopy. Commercial solutions offering this capability are prohibitively expensive to individual labs, and are most likely to be found only in well-resourced core facilities.

This thesis presents the design, implementation, and testing of three new technologies for simple, inexpensive, viable, image-predicated cell sorting that can be disseminated to individual labs. For anchorage-dependent cells, we have implemented two technologies. The first is a method termed polymerization-activated cell sorting, or PACS, that permits the selective hydrogel-mediated photo-encapsulation of undesired cells in a culture dish and the retrieval of desired cells using enzyme-mediated release. The second is a method termed radical-activated cell sorting, or RACS, that permits the light-mediated killing of undesired cells in a culture dish through photopatterning of radical toxicity. For non-anchorage-dependent cells, we have developed optofluidicell sorting, or optoflucs. In this approach, cells are injected into a microfluidic device where they are allowed to settle into an array of cell-sized microwells. A laser is then used to optically levitate desired cells out of the array and into a flow stream for collection.

This thesis demonstrates > 100-fold enrichment of target cell populations and up to 89% output purity following a single round of image-predicated, viable cell sorting. In addition, this thesis demonstrates examples of sorts predicated on fluorescence localization, a phenotype that cannot be sorted using FACS. These results, coupled with the minimal expense and simplicity of these technologies make these techniques appealing for widespread adoption by individual biology labs.