I earlier wrote two pieces on a microfluidics chip for counting nucleic acids by limiting dilution PCR -- in one application it was used to count mRNAs and the other for counting bacteria. Last week's Science has a nice complement to this: microfluidic chips that count proteins.
A really cool aspect of the chip is its assembly line nature: it doesn't just count proteins, it performs the upstream steps as well. Starting with a sample of cells, it plucks out a single cell. The cell is rinsed and then lysed. The fluorescent antibodies are introduced (if necessary) and the an electrophoretic separation performed. Finally, fluorescent molecules are counted as they pass through a chip region which is illuminated with a sliver of light of the correct excitation wavelength.
They actually demonstrated two variants of the basic scheme: one chip performed an immunoassay on eukaryotic cells as described above, while the other looked at naturally fluorescent proteins in cyanobacteria. The immunoassay chip analyzed a single cell, whereas the cyanobacterial chip ran three in parallel.
This is a really neat development and presuming the costs can be made reasonable, would be very interesting for many applications. But, there are very few naturally fluorescent proteins and so for most applications high quality antibodies (or equivalent specific binders) will be needed -- a nut that has yet to be cracked.