TECHNOLOGY FEATURE NEURAL CIRCUITS NATURE|Vol 461|22 October 2009 present 20 × 20 × 25-nanometre resolution desirable, and scientists since Ramón y Cajal limit of SBF-SEM. However, SBF-SEM can have pursued chemical and biological meth- image much greater volumes — on the scale of ods for exclusively targeting neurons that two to three orders of magnitude more — mak- are functionally linked via active synapses.

ing focused ion beam and SBF complementary One promising method, being pioneered by FMI/FEI C. GENOUD, rather than competitive tools. scientists such as Lynn Enquist at in New Jersey and Ed Callaway at A beautiful mind the Salk Institute for Biological Studies in La There are alternatives to electron microscopy Jolla, California, exploits natural infection pat- — particularly for researchers interested in terns of neurotropic viruses for the fluorescent more than a static snapshot. “All the things labelling of individual neural circuits4. I’ve studied up until now have been dynamic Callaway works with modified rabies virus, questions,” says Lichtman. “And you just can’t a pathogen that spreads so efficiently across do that with electron microscopy — you’ve got mouse neurons that a single particle injected to kill it to look at it!” into the brain can prove lethal. His viruses are Lichtman’s solution was the Brainbow constrained via deletion of a key glycoprotein transgenic mouse3, which uses a site-specific Focused-ion-beam microscopy, as performed gene. “We preserved the ability to replicate and DNA recombination system to randomize with instruments such as FEI’s Helios NanoLab amplify, but provided a means to control the expression of multiple fluorescent protein DualBeam, allows more energy to be used for spread,” he says. “Deleting the glycoprotein gene genes in neurons, yielding intermediate col- imaging, improving the resolution. also allows us to control the initial infection and our combinations that distinguish each cell target specific cell types.” Some investigators are from its neighbours. With a broad portfolio distinct labels. “All of the colours of the rainbow applying viral tracing to trace entire networks of commercially available fluorescent proteins that we see are interpreted from three pigments of interconnected cells, but Callaway is mostly from which to choose — including the Living in our ,” he explains. “So we just inverted interested in targeting smaller ‘neighbourhoods’. Colors proteins made by Clontech in Mountain that, thinking that if we could just mix different “When we get to the point where we can go View, California, and the TurboColors proteins amounts of three colours in different cells, we into a live animal and target one cell and label from Evrogen in Moscow — Lichtman’s group should be able to get all the visible colours of every single input to that cell, that will be a huge had many options. However, just a handful of the rainbow.” advance,” says Callaway. “But it’s clear we’re far colours proved sufficient to generate nearly 100 In other cases, more selective labelling is from labelling all of them. We’re now labelling

WHOSE MAP IS IT ANYWAY?

Even as ‘’ makes its solving that problem, all the other brain, and Mitra thinks that linking for those of us who are pursuing way into the mainstream scientific problems become trivial.” these will prove challenging but — to do not one, but vocabulary, there is already On the other hand, Partha Mitra not insurmountable. “Larry has many. Neural plasticity is just one debate over what — if anything of Cold Spring Harbor Laboratory deep knowledge of the relevant of the many interesting questions — it actually means. “It’s sort of in New York thinks that the tools literature, and estimates that only that will be open for new attack.” analogous to how ‘’ used to are already at hand for creating a around a third of these possible Scientists in both camps hold up mean the set of genes, but now it sparser ‘mesoscopic’ map of the mesoscopic connections have work done by researchers at the means the whole DNA sequence,” projections that link functionally ever been studied,” Mitra says. Allen Institute for Brain Science in explains Sebastian Seung of discrete brain regions, which some “But when I sat down and thought Seattle, Washington, in mapping the Massachusetts Institute of call a projectome — although you about the cost to map out those gene expression in the brain as Technology in Cambridge. won’t catch Mitra using that term. connections, I was shocked to an example of how good science, There is fairly broad agreement “Everything has an ‘-ome’ added to find that it actually shouldn’t take careful planning and efficient that mapping the wiring in it, and that’s ok if you’re in a yoga that much time, money or effort.” workflows can yield tremendous mammalian brains is a worthwhile class,” he jokes. “But I prefer ‘brain With the recent awarding of a pay-offs. Smith and others think endeavour. The issue is one of scale architecture’ because it conveys Transformative R01 grant from the that maturation of high-resolution — should these be comprehensive structure and function; architects US National Institutes of Health, circuit-mapping techniques will reconstructions of neuronal shape space for human use, and Mitra’s team is now taking first ultimately bring high-throughput circuitry, or more macro-scale evolution shaped our nervous steps towards making their Mouse ‘dense’ reconstruction within representations of long-range system for appropriate behavioural Brain Architecture Project into a reach. connections between regions of the repertoires and so on.” reality. Accordingly, Mitra emphasizes brain? This is the neuroanatomical Mitra and dozens of colleagues At the same time, by deliberately that his group’s data — which it equivalent of choosing between recently published a plan for overlooking the highest orders of intends to make freely available creating a road atlas or Google integrating existing tools — neuroanatomical complexity, this via an open-access model — Earth. including chemical labels and approach leaves open numerous should provide a framework for Arguments can readily be engineered viral tracers — into questions that will probably be future reconstructions. “This marshalled for and against either a concerted effort to chart the answered only by dense mapping. is just supposed to be the first approach; most boil down to how connections between functionally “These things are just so incredibly generation,” he says. “I have no best to invest time, money and homogeneous clusters of cells via tangled and complicated, it’s doubt that if this succeeds — this technology. “Dense reconstruction light microscopy9. Larry Swanson inconceivable that you’ll come whole-brain approach to brain of a cubic millimetre of the cortex at the University of Southern across two identical brains,” architecture and neuroanatomy is kind of a ‘going to the Moon’ California, Los Angeles, has says Stephen Smith of Stanford — then we will see successive goal, where we think it’s possible proposed that 500–1,000 such University in California. “And I waves of technology hitting the but difficult,” says Seung, “but in anatomical units exist within the think it is a wonderful opportunity problem.” M.E.

1150 © 2009 Macmillan Publishers Limited. All rights reserved