Connectivity between L3 and Tm9

Asked at the VFB table, NeuroFly 2026 (Cologne), 9 September 2026.

The question

“Find me connectivity data between L3 and Tm9, then plot synapse count against frequency, coloured by dataset.”

L3 is a lamina monopolar neuron, Tm9 a transmedullary neuron — a canonical, heavily studied connection in the optic lobe’s motion/ON pathway. The asker didn’t want a single headline number but the shape of the connection: how many synapses each individual L3→Tm9 pair makes, and whether that shape is reproducible across the connectomes VFB now holds.

How we answered it

1. search_terms to canonicalise both types — L3 = lamina monopolar neuron L3 (FBbt:00003721), Tm9 = transmedullary neuron Tm9 (FBbt:00003797). Worth noting the immature-neuron terms and the Tm9a/Tm9b subtypes that also match the text search.
2. list_connectome_datasets to see what’s currently loaded.
3. query_connectivity with both ends fixed (L3 upstream, Tm9 downstream) and group_by_class=false, so we get individual neuron-to-neuron pairs rather than a class-level aggregate, at synapse threshold ≥1. This is a live cross-dataset query; the both-ends L3→Tm9 form returned 5,356 pairs across four datasets — we set BANC’s two aside (far too few for a distribution), leaving 5,354 pairs across three optic-lobe connectomes.
4. Binned each pair by its synapse count and drew one frequency polygon per dataset.

Direction matters: we queried L3→Tm9 (Tm9 postsynaptic), the biologically dominant direction — the reverse is negligible.

What we found

The three well-sampled datasets agree almost exactly — median synaptic weight of 21–23 per L3→Tm9 pair — a reassuring cross-connectome check on a textbook connection. The distribution is bimodal: a low-count shoulder (2–4 synapses, weak or incidental touches) and a broad main peak around 22–30 synapses, the genuine columnar retinotopic connection.

Worth being honest about:

• Dataset coverage. Only optic-lobe connectomes reconstruct these cells. Hemibrain, MANC, FAFB-CATMAID and the larval L1 CNS were excluded — they don’t contain L3/Tm9 — and BANC was set aside because its current VFB load holds only 2 L3→Tm9 pairs, far too few to say anything about a distribution.
• Not normalised. These are raw pair counts. Datasets differ in how many optic-lobe columns were reconstructed, so compare the shapes of the distributions, not the absolute peak heights.

Sources

• FBbt:00003721 — lamina monopolar neuron L3
• FBbt:00003797 — transmedullary neuron Tm9
• Fischbach and Dittrich, 1989, Cell Tissue Res. 258(3):441–475; connectome sources: FlyWire (Dorkenwald, Schlegel et al., 2024), Optic Lobe (Nern et al., 2024), male-CNS (Berg et al., 2025).