Epigenetic marks do not act alone. DNA and histone modifications can be grouped into epigenetic 'states' that affect gene expression and cell identity in both normal development and disease. The hunt for these states led the engineering group of Harold Craighead and the epigenetic team of Paul Soloway at Cornell University into a creative collaboration. Their brainchild: a nanofluidic device that can identify and sort single chromatin fragments based on the combination of marks that they carry.

Robust methods to locate various marks in the genome already exist. Chromatin immunoprecipitation detects histone modifications, and bisulfite sequencing identifies DNA methylation. So why the need for a new tool? Epigenetic states are derived by overlaying maps of these marks, but comparing data from different cell populations can obscure how marks are coordinated within each cell. “One thing that you can never resolve from superimposition,” notes Soloway, “is whether those epigenetic marks that are identified on a given gene are really present at the very same time, on the same molecule.”

Although chromatin immunoprecipitation can be performed in successive rounds or combined with bisulfite sequencing on a single sample, either option requires a lot of starting material (typically more than 1,000 cells per round), whereas nanofluidics can characterize multiple marks directly from a single cell.

Soloway likens nanofluidics to fluorescence-activated cell sorting (FACS), except that nanofluidics uses voltage gradients to propel molecules rather than using pressure to propel cells. Chromatin stained with affinity fluorescent probes is loaded and threads like a noodle through a 250-nm-deep channel. Lasers excite the passing fragments prior to a branch point, and sensitive photodiodes detect the emission of single fluorophores thanks to the tiny interrogation volume of the channel. Signal is processed in real time by on-board circuitry, which can be programmed to score fluorophore combinations and actuate a voltage switch that sends DNA to the positive output channel.

The engineering was tricky, but the groups achieved a sorting rate of more than 500 molecules per minute, with a low (1–2%) incidence of false positives. They improved accuracy by observing output channels as well as the input and tweaking the sorting parameters. Traditional FACS operates at higher speeds, but the tiny nanodevice can be multiplexed to greatly increase throughput.

The researchers demonstrated the ability to sort plasmids, either naked or bearing methyl groups, using a DNA dye and labeled MBD1 protein that binds methylated DNA. They enriched at levels similar to FACS, but missed up to 25% of positive molecules. To lower the false-negative rate, they have now optimized MBD1 binding conditions and are exploring brighter probes.

The technology is still at the proof-of-principle stage, but it has intriguing potential. One attraction of chip-based approaches is their easy integration. Any version of a Lilliputian factory can be dreamed up, conceivably including DNA extraction, probe binding, sorting and library preparation for sequencing, all on the same factory floor. Profiling single cells may also address the provocative question of how epigenetic states are propagated through cell division, or during differentiation or reprogramming. Soloway wants to profile rare and hard-to-culture cells such as mouse primordial germ cells.

The groups are working on directly sorting chromatin, adjusting optics to add a third laser, optimizing probes and attempting unbiased amplification of the mere femtograms of sorted output to allow sequencing. Soloway hopes that commercialization will offer a route for technical improvements like mutliplexing to ramp up throughput.

Ultimately, the device could be used to profile the effect of epigenome-modifying drugs. “This tool,” says Soloway, “could provide a very effective means to quite rapidly screen for efficacy of different compounds and query the effective compounds on multiplicities of epigenetic states.”