Our ability to sense and navigate the world requires the precise assembly and function of neural circuits in the brain. During development, neuronal-cell projections called axons are guided by molecular cues to extend away from non-target regions of the brain and towards their target regions1, where axons make synaptic connections with partner neurons. Over the past few decades, several candidate molecular cues have been identified2; however, questions remain as to whether distinct sets of cell-surface molecules mediate attraction to targets and avoidance of non-target regions. Writing in Science, Pederick et al.3 show in mice that axon attraction and repulsion are guided by the same cell-surface molecule during circuit assembly in the hippocampus, a brain region involved in spatial memory and navigation4.
The hippocampus contains the subfields CA1, CA2 and CA3, and neurons in the CA1 subfield project to a target region in an adjacent brain structure called the subiculum. CA1 projections to the subiculum are organized along a medial-to-lateral anatomical axis. In this way, in the medial part of the network, neurons in the proximal CA1 (located near the border with the CA2 region) project to the distal subiculum (the part farthest from the CA1 border), whereas, in the lateral part of the network, distal CA1 neurons project to the proximal subiculum (Fig. 1).
A previous study by Pederick and colleagues’ laboratory showed5 that a cell-surface molecule called teneurin-3 (Ten3) is expressed both by proximal CA1 neurons and by neurons in their target region, the distal subiculum. The study showed that molecules of Ten-3 adhere to each other, and that this binding leads to attraction between neurons expressing this protein. Through this interaction, Ten3-expressing projections are attracted to target regions that express Ten3. Pederick et al. hypothesized that a similar mechanism — in which protein binding causes projections expressing that protein to be attracted to target regions expressing the same protein — might be involved in directing the formation of the lateral hippocampal network.
Using a technique called single-cell RNA sequencing to profile gene expression in individual cells from the developing mouse hippocampus, Pederick et al. found that the cell-surface protein latrophilin 2 (Lphn2) was expressed both in distal CA1 projections and in proximal-subiculum target neurons of the lateral hippocampal network. They initially investigated whether Lphn2–Lphn2 adhesion and attraction might, in a similar way to Ten3–Ten3 adhesion and attraction, direct the formation of hippocampal circuits. However, this was not the case: when the authors overexpressed Lphn2 in a non-adhesive cell line, the cells did not adhere to each other. By contrast, Lphn2-expressing cells readily formed aggregates with Ten3-expressing cells, consistent with previous reports of Lphn2–Ten3 binding6.
The binding of Lphn2 to Ten3 could potentially trigger the activation of signalling pathways inside an axon, resulting in it moving towards or away from the region in which the interaction takes place. Because the areas targeted by Ten3-expressing axons and Lphn2-expressing axons do not overlap, Pederick and colleagues reasoned that Ten3–Lphn2 interactions might result in repulsion. To test this, the authors used a clever approach that involved injecting engineered viruses into the hippocampus to manipulate the expression of Lphn2 and Ten3 by CA1 neurons, and by neurons in their subiculum target regions.
The authors injected a virus expressing Lphn2 into the distal part of the developing subiculum, where Lphn2 levels are normally low, to increase Lphn2 levels there. Once the hippocampus had developed fully, the authors injected a virus expressing a fluorescent protein into the proximal CA1, where Ten3 expression is high, to visualize the axons from that region that had innervated the subiculum. These axons avoided the regions where Lphn2 was artificially expressed. The authors then performed the converse experiment: reducing Lphn2 expression in the developing proximal subiculum, where expression of this protein is usually high. In this case, Ten3-expressing axons from the proximal CA1 region invaded the regions where Lphn2 expression was reduced. Together, these results suggest that Lphn2–Ten3 interactions are necessary and sufficient for repulsive guidance of Ten3-expressing axons.
Crucially, the authors assessed the relative contributions of Ten3-mediated attraction and Lphn2-mediated repulsion of Ten3-expressing proximal CA1 axons, by reducing the expression of both Ten3 and Lphn2 across the entire subiculum. This manipulation led to an increase in axon projections in non-target regions and reduced innervation of the target region. Therefore, the precise targeting of Ten3-expressing proximal CA1 axons seems to require both Lphn2-mediated repulsion away from non-target regions and Ten3-mediated attraction.
How do Lphn2-expressing CA1 axons respond to target regions containing high levels of Ten3? Deletion of Ten3 expression from cells in the distal subiculum led to greater innervation of Lphn2-expressing axons from neurons in more-distal parts of CA1 into more-distal subiculum than was observed in hippocampi from control mice; this indicated that target-derived Ten3 repels Lphn2 axons.
Cooperation between attraction and repulsion of axon projections is a familiar theme in the development of neural circuits2. This study demonstrates that the connectivity between CA1 axons and subiculum neurons in the hippocampal network tightly follows a ‘Ten3 axon to Ten3 target, Lphn2 axon to Lphn2 target’ rule, instructed by reciprocal repulsions between Ten3-expressing and Lphn2-expressing cells.
Pederick et al. beautifully demonstrate how the binding interactions between cell-surface molecules depend on cellular context: on CA1 projections, Ten3 acts as a receptor for both attractive (Ten3) and repulsive (Lphn2) target-derived cues, whereas in the subiculum, it serves to repel Lphn2-expressing axons (Fig. 1). The importance of developmental context for the mechanisms that guide circuit assembly is further indicated by another study7 showing that coincident binding of Lphn2, Ten3 and another cell-surface molecule, Flrt2, is required for the formation of neuronal synaptic connections between CA1 neurons and their partners upstream in the hippocampal circuit, rather than for axon guidance.
Further research is needed to identify the signalling cascades that are triggered by cell-surface molecules such as Ten3 and Lphn2, and that determine whether and how an axon is attracted to or repulsed by a given molecular cue. Also, if attraction and repulsion are both necessary for precise circuit assembly, what is the identity of the cell-surface molecule that mediates the attraction of Lphn2-expressing axons?
Given that the number of cell-surface molecules encoded by the genome is limited but the circuitry of the mammalian brain is highly complex, each cell-surface molecule that is involved in guiding axons to their appropriate targets probably serves multiple such functions in different circuits, depending on the cellular and developmental context. It will be crucial to account for each molecule’s context-dependent roles during the assembly of diverse neuronal circuits.
C.H. is in the process of transitioning from his appointment at Harvard Medical School to UCSF where he will also be a Chan Zuckerberg Biohub investigator. Stephen Quake is a co-author of the study referred to in this News & Views article and is currently co-president of the Chan Zuckerberg Biohub.