Recombinant rabies viruses (RVs) can be used to infect neurons at axonal projections, at which point they are transported to the soma. The infection can either spread to neurons presynaptic to the starter neurons or simply lead to marker gene expression in the targeted projection neurons, both of which are important applications in neural circuit mapping. RVs known as DG viruses are attenuated owing to the deletion of the viral envelope glycoprotein, but even they cause cytotoxicity that can impair the structure and function of infected neurons.

Attempts to reduce RV-associated toxicity have included the use of a less toxic RV strain or a self-inactivation strategy, but these approaches do not entirely overcome the problem. “We just wanted to make a completely benign rabies virus,” says Ian Wickersham from the Massachusetts Institute of Technology. In collaboration with scientists from the Allen Institute for Brain Science, Wickersham and his team pursued a double-deletion strategy. “The idea was to remove almost entirely [the virus's] ability to express any genes from its genome,” he explains. The commonly used DG virus is impaired in the final stages of the life cycle, but it can still replicate, which dooms the infected cells. To avoid viral replication, the researchers generated ΔGL viruses by deleting the viral polymerase gene, “which is absolutely required for expression of genes from the viral genome,” says Wickersham.

Neurons remain healthy after long-term infection with ΔGL RV. Adapted with permission from Chatterjee et al. (2018).

Abolishing essentially all expression from the viral genome means that the expression of engineered cargo, such as fluorescent proteins, also will be affected, which can be a problem. However, each viral particle “comes with a few polymerase proteins packaged in the virion,” says Wickersham. These polymerases lead to a “tiny amount” of transcription, which, although insufficient for reporter gene expression, can generate enough Cre or Flp recombinase to induce Cre- or Flp-dependent reporter gene expression.

The researchers verified that the ΔGL virus is benign by monitoring the health of infected neurons over four months. They found that neither the structure nor the response properties of infected neurons changed over time.

In addition to the complete absence of cytotoxicity, the tropism of the ΔGL virus, meaning its preference for infecting particular cell types, distinguishes this RV from alternative retrograde tracing tools such as CAV2 and rAAV-retro. For corticocortical projections, “the rabies virus had broader tropism than either of those two species,” says Wickersham. This property makes the ΔGL RV more broadly applicable.

The researchers demonstrated the utility of the nontoxic RV for labeling projection neurons in different brain regions in mice and rats. However, a major application for RVs is monosynaptic tracing. For this, the DG RV must pick up G-glycoproteins that are supplied in starter neurons, so that the virus can complete its life cycle and infect neurons presynaptic to the starter cells. Wickersham and his team are currently working to establish monosynaptic tracing for their ΔGL virus by providing both G and L proteins in the starter cells. “It will lead to toxicity in the starting cells,” concedes Wickersham, but they are planning to express the proteins at low levels and to use the previously described less toxic CVS N2c strain. Such a system would enable long-term functional studies with, for example, optogenetic tools.