Written by Xinran Feng

Revised by Lin Mei

In the brain, billions of neurons are connected by trillions of synapses with specificity, however, the molecular mechanism remains largely unknown. One plausible explanation for this biased connectivity is that processes of certain cell types express cell recognition proteins that selectively bind to receptors on partner cells' surfaces. This recognition event could trigger adhesive interactions that promote synapse formation between appropriate partners. [1] In 2019, the team of Thomas C. Südhof and Shuai Wang of Stanford University published a Science article [2] and reported that synapse formation in the mouse hippocampal CA1 region requires latrophilin 2 and 3 (Lphn2 and Lphn3). Interestingly, Lphn2 mediates synapse formation projected from the entorhinal cortex, while Lphn3 mediates synapse formation projected from CA3 region. This function required the binding of both teneurins and FLRTs.

Lphn3 is a multi-domain protein encoded by a gene with multiple exons. The RNA molecules transcribed from the genome are called pre-mRNAs. In a process called “splicing”, the pre-mRNAs should be further processed by splicing factors to excise introns out before they are transported to the cytoplasm for protein translation. However, in most cases, one gene can undergo different splicing patterns, so-called alternative splicing (AS). Alternative splicing can lead to mRNA decay or produce diverse protein isoforms. By analyzing the PacBio RNA-seq data and neuron subtypes specific RNA-seq data, the authors found that Lphn3 has 32 exons with at least 10 exons of them undergoing alternative splicing. Many of the alternative exons showed unique splicing patterns in excitatory and inhibitory neurons, especially exon 31 (E31) and 32 (E32). First, E31 is more included in inhibitory neurons. E32 is more included in excitatory neurons. To be specific, E31 and E32 are actually mutually exclusive. Second, the splicing of E30, E31, and E32 determines the cytoplasmic region of Lphn3. Third, the splicing patterns of Lphn3 exons also differ in brain regions and postnatal developmental stages. Among them, the E31-included variant is the most abundant Lphn3 C terminal splice variant in mouse hippocampus.

Given that the C terminal of a GPCR is important to G protein coupling, the authors then asked whether Lphn3 C terminal splicing events alter its Gα protein coupling preferences. They used TRUPATH, an open-source biosensor platform for interrogating the GPCR transducerome, to measure the G protein coupling ability of Lphn3 isoforms by BRET2 (bioluminescence resonance energy transfer) assay. They found that the E31 variant preferentially couples to Gαs, while the E32 variant preferentially couples to Gα12/13. The expression of E31 variant dramatically increased cAMP level in cells. Interestingly, they also noticed that, the splicing patterns of the homologous genes, Lphn1 and Lphn2, may also regulate their G-protein coupling ability.

Which Lphn3 C terminal splice variants promote synapse formation? To answer this question, the authors used CRISPR to knockout E31 containing Lphn3 transcripts. E31 KO did not change Lphn3 protein levels, but indeed increased E32-containing transcript levels. They found that E31-specific KO decreased neuronal firing rates as observed for the global loss of Lphn3 proteins. Meanwhile, E31-specific KO, like global KO, decreased in excitatory synapse density.

As illustrated above, hippocampal CA1 neurons receive major inputs from entorhinal cortex (EC) and the CA3 region.[3] Südhof and colleagues previously showed that Lphn3 in hippocampal CA1 region mediates synapse formation from CA3 region but not from the entorhinal cortex. To study whether E31-specific KO impairs this function, the authors used the monosynaptic retrograde tracing technique to trace neural connections to the CA1 region. They found that E31-specific KO decreased in synaptic inputs to CA1 pyramidal neurons from the ipsilateral and contralateral CA3 region. Thus, E31-containing isoform coupled to Gαs is essential for Lphn3-mediated synaptic connectivity.

Next, the authors asked whether the Lphn3 E31 splice variant perform additional functions other than Gα protein coupling? E31-containing Lphn3 isoforms have a PDZ binding motif at the C terminus, but the E32-containing Lphn3 isoform does not. This motif interacts with Shank in postsynaptic protein assemblies. The authors, therefore, did the in vitro sedimentation assay with GKAP, Homer, PSD95, Shank (GHPS proteins), as well as the indicated Lphn3. As expected, only Lphn3 containing E31 robustly co-sedimented with the GHPS complex, whereas Lphn3 containing E32 or E31(ΔPBM) did not. Adding the presynaptic ligands, Lphn3, TENM2, and FLRT3 to the mixture promotes the clustering of phase-transitioned droplets. Again, the effect was not observed in E32-containing Lphn3 or E31 without PBM. Notably, Lphn3 exhibited faster recovery kinetics than most scaffold proteins, suggesting that Lphn3 E31 variant forms a fluidic shell on the surface of the postsynaptic scaffold protein condensates. Thus, LPHN3 E31 but not LPHN3 E32 is located on the surface of phase-transitioned droplets formed by postsynaptic scaffold proteins, LPHN3 ligands TENM2 and FLRT3 can further cluster the droplets.

So, is the PDZ binding motif encoded by Lphn3 E31 required for synapse formation? To answer this question, the authors, again, used CRISPR to selectively delete the short PBM from the E31-containing Lphn3. This manipulation did not decrease the protein or mRNA levels, or change the splicing patterns of Lphn3. However, as expected, PBM deletion significantly decreased the excitatory synapse density in hippocampal neurons. Thus, the PBM encoded by Lphn3 E31 is important for excitatory synapse formation.

Ultimately, the authors were curious whether the splicing of Lphn3 E31 and E32 is regulated by neuronal activities since it showed distinct patterns in different types of neurons and synapses. The authors analyzed the published RNA-seq datasets and found that either using potassium chloride, PTX, or KA to activate the neurons, Lphn3 splicing pattern shifted from the inclusion of E32 to E31. Also, when checking the single-cell RNA-seq results, they found that the neurons with higher immediate early gene expression also had higher E31 inclusion ratio.

Over all, this paper was well organized and the experiments were well designed. However, I still have some concerns and questions to the authors:

Original publication:

Wang, S., DeLeon, C., Sun, W., Quake, S.R., Roth, B.L., and Sudhof, T.C. (2024). Alternative splicing of latrophilin-3 controls synapse formation. Nature 626, 128-135.

Reference:

[1] Sanes, J.R., and Zipursky, S.L. (2020). Synaptic Specificity, Recognition Molecules, and Assembly of Neural Circuits. Cell 181, 536-556.

[2] Sando, R., Jiang, X., and Sudhof, T.C. (2019). Latrophilin GPCRs direct synapse specificity by coincident binding of FLRTs and teneurins. Science 363.

[3] Petrantonakis, P.C., and Poirazi, P. (2014). A compressed sensing perspective of hippocampal function. Front Syst Neurosci 8, 141.

[4] Wallis, D., Hill, D.S., Mendez, I.A., Abbott, L.C., Finnell, R.H., Wellman, P.J., and Setlow, B. (2012). Initial characterization of mice null for Lphn3, a gene implicated in ADHD and addiction. Brain Res 1463, 85-92.