From my personal perspective, this is outstanding work, and it is not a paper you read once. It is a paper you come back to.
What struck me most is not the scale of dSC1 but what SPINE learned. The attribution analysis shows the classifier discriminating subtypes on vesicle morphology: dense-core vesicles, large pale vesicles, the clustered pale vesicles of the type I glomerulus. These are exactly the criteria that Ribeiro-da-Silva, Coimbra, Réthelyi and Todd established decades ago in the darkness steering on the green screen. A network trained without priors recovered the classical diagnostic features and then showed that they carry physiological identity. The descriptive EM literature was not wrong about the dorsal horn. It described objects it had no means to count. This work supplies the denominators, and a collection becomes a circuit.
I want to say something about the honesty of the report, because it is unusual these days. The APEX strategy is elegant, but what I value more is the care taken over the morphological characterization of the primary afferent terminals, and the willingness to formulate intellectual and interpretational strategies based upon the classical work rather than step around it. That choice elevates the intellectual value of the paper.
For me this is not electron microscopy work. It is functional work. A single Aβ-LTMR with more than 17,000 axodendritic synapses against roughly 320 for a C-cold thermoreceptor; the near-absence of axo-axonic input on thermoreceptors and on the itch-associated populations; five inhibitory populations each driven by and feeding back onto one modality, while sending their axodendritic and dendrodendritic output across modalities. Our community has speculated for a long time about how signal processing occurs at the first sensory gate. Now we can begin to reverse engineer it. The Acr1 control, showing that afferents lacking PAD remain competent to generate it, is what makes the anatomy predictive rather than descriptive.
One thought on what the resource now makes possible. The connectivity analyses treat each synapse as a unit, but the segmented clefts carry apposition area, and active zone size is a reasonable first-order proxy for receptor density, release probability and postsynaptic current. A weighted version of these graphs looks to be within reach of the released data. Are the sparse axo-axonic synapses on MrgprD+ afferents also small ones? Do the distal clusters differ in size from the en passant contacts? Do the perforated Mrgpr boutons carry larger appositions? This speaks to density and not to receptor identity, but it would move the connectome from counts toward weights, and these are questions that only become askable once a dataset like this exists.
In my opinion this is the most logical work exemplifying the value of volume electron microscopy, leading to direct functional interpretations.
Congratulations to the authors!