Excitatory axons, but not basket cell axons, selected excitatory targets with similar orientation beyond axon–dendrite overlap and 10 µm soma distance in MICrONS mouse visual cortex
Question: In mouse visual cortex, do axons select excitatory (E) targets by orientation preference beyond axon–dendrite overlap and soma distance (Peters' rule), and do excitatory and inhibitory axons differ? Approach: We used the public MICrONS minnie65 v1507 data from one mouse: 2,141 proofread axons, all 337,312,429 synapses, and digital-twin orientation tuning for 21,381 co-registered E neurons. Overlap was the number of shared 5 µm cubes between the output synapses of an axon and the input synapses of a neuron. For each real partner, we drew surrogates from unconnected candidates in the same overlap bin and 10 µm soma-distance bin. Results: Connected E→E pairs of tuned neurons had a mean cos 2Δθ 0.050 higher than matched unconnected pairs (95% CI 0.041 to 0.059; 41,058 connections). The excess was 0.044, 0.111 and 0.148 for 1, 2 and 3 or more synapses. The summed input vector of a neuron predicted its own orientation better than surrogate inputs (margin 0.041, CI 0.030 to 0.053; 18,956 neurons). The excess pointed to the neuron's own orientation, not to that of its neighbours. It was clear for L4 and L5 targets but not for L2/3 targets. Surrogate inputs gave 80% of the prediction score at 20 inputs, because orientation preference is spatially clustered. At 50 targets per axon, the targets of E axons had higher pair similarity than surrogates (ΔS 0.020, CI 0.016 to 0.024; 1,198 axons). Basket cell axons did not (ΔS −0.000, CI −0.009 to 0.008; 200 axons). The E minus basket difference was 0.020 (CI 0.010 to 0.030) and stayed positive (0.012 to 0.020) with manual co-registration, a stricter model fit, and 10 µm cubes. Martinotti, bipolar and neurogliaform results were not clear. The targets of 284 inhibitory neurons did not align with the orientation of their own excitatory inputs. Pitfall: basket and Martinotti axons contact more than 80% of high-overlap candidates. Surrogates drawn without replacement then lost the distance match for 42% of basket targets. With 25–50 µm bins, this gave a false basket effect equal to the E effect. Limitations: one mouse; model-derived tuning of E neurons only; proofread axons are not a random sample; synapse positions approximate axon and dendrite paths; absolute effects are small. Code, result tables, derived tables, figures and instructions are included.
connectomics · MICrONS · mouse visual cortex · orientation tuning · Peters' rule · like-to-like connectivity · basket cells · inhibitory interneurons · structure-function · surrogate design · axon-dendrite overlap · digital twin
References
15- usesdoi:10.1038/s41586-025-08790-w
Source of the MICrONS minnie65 v1507 connectome, co-registration and digital-twin tuning tables that we analyse.
- supportsdoi:10.1038/s41586-025-08840-3
Our overlap- and 10 µm distance-matched E→E result agrees with their like-to-like rule; we add a test of inhibitory axon outputs.
- supportsIn MICrONS mouse visual cortex, connected excitatory neurons had more similar preferred orientations mainly when their synapses were large (14,412 neurons, digital-twin tuning) sci:sha256:07b8a9b83eea2b7443dc99d29805440aefc0d23414bb9c8959decc534824bd92
With a different overlap measure (shared 5 µm cubes) we get a similar tuned E→E excess (0.050 vs 0.046) and a rise with synapse count.
- Requiring valid axon review marks reduced prediction gains for estimated orientation from 2.23% to 0.18% in a fixed kernel test of 3,461 neurons from one MICrONS mouse sci:sha256:614767952cadc655012f6ebed4f66c161f7cfbadb527ef544ca8d35afb13ffc7
Related test of orientation prediction from inputs; we use only proofread axons and Peters'-rule surrogates and find a small, neuron-specific margin.
- With relay and partner controls, input orientation profiles gave no clear output prediction gain in a ridge test of 1,314 neurons from one MICrONS mouse sci:sha256:e143b4493109cbb616b349ebcd59caca419d6ffab0908410ea6ff05cae709a4e
Related input-to-output orientation test for pyramidal relays; we test the E→I→E case and also find no clear alignment.
- Separate penalties reduced the added inhibitory input gain from 1.41% to 0.18% in a model-derived orientation test of 5,193 MICrONS neurons sci:sha256:66588b94767d58b8bc4d385a08080d9abd1f0d7d4943566a8a2f4ead93f1620d
Related study of inhibitory inputs and orientation; we study inhibitory outputs.
- Paired recordings in mouse V1 found most excitatory but less than half of the inhibitory connections that MICrONS electron microscopy predicted for 17 matched cell classes sci:sha256:ce046109af65d9a43f99d098f26cd85017dc112137362e3deed08cf3846aea5f
Shows differences between EM-predicted and physiologically detected inhibitory connections, a limitation for EM-based inhibitory analyses.
- doi:10.1038/s41586-024-07780-8
Connectomic census of inhibitory target specificity in the same volume.
- doi:10.1016/j.neuron.2023.12.013
Functional specificity of PV inhibition measured as strength; our test concerns connection choice and synapse count.
- doi:10.1038/nature09802
Earlier EM evidence that inhibitory neurons receive input from excitatory neurons with diverse orientation preferences.
- doi:10.1016/j.neuron.2010.08.002
Inhibitory neurons in mouse V1 are broadly tuned.
- doi:10.1038/nn.2876
Dense, unselective excitatory input to PV interneurons in mouse V1.
- doi:10.1038/nature09880
Like-to-like connectivity between L2/3 pyramidal neurons from paired recordings.
- doi:10.1038/nature17192
EM evidence for orientation-related connectivity among L2/3 pyramidal neurons.
- supportsdoi:10.1038/ncomms12270
We also find spatial clustering of orientation preference, which makes distance matching necessary.
Cited by
6 publicationsPublic catalog entries on Sciscape.
- Recorded movie similarity gave no clear loss decrease after model similarity in a fixed test of 234,393 MICrONS pairs from one mouse
This earlier SCI study tested model orientation specificity after anatomical overlap controls. Our test uses recorded and model movie similarities together.
- In one mouse visual cortex volume, like-to-like synapses formed on basal and apical dendrites in proportion to axon–dendrite contact, for within-area, feedforward and feedback axons
Related result on axon target selection by orientation beyond axon-dendrite overlap in the same volume.
- Reverse excitation reduced ridge prediction error for inhibition by 36.9 percent in 66 reciprocal mouse V1 pairs with visual controls
Related SCI work tests anatomical target choice and orientation in MICrONS. The present test concerns physiological weights and full visual response similarity.
- In MICrONS mouse V1 layer 2/3, inhibitory connections spread 85–113 µm, not the 40 µm a published holographic-stimulation model used; the stimulation data fit wide inhibition with narrow excitation
Found that MICrONS inhibitory targets do not match the orientation of their excitatory inputs; we did not refit orientation rules.
- supportsIn mouse V1 EM, E→I→E wiring was like-to-like with the sign that holographic far suppression needs, but EM orientation specificity was 4–7 times too weak in a truncated linear model
Our cell-specific part of E→I→E orientation specificity (real minus block shuffle, about 0.01; proofread axons 0.014) agrees with its null result for interneuron input-output alignment; the total specificity at fixed distance is larger because of the spatial clustering of orientation.
- In the V1 Deep Dive mouse, connected excitatory neurons had only slightly more similar movie responses than nearby unconnected neurons; the same test gave a 2 to 7 times larger effect in MICrONS
Earlier SCI study of excitatory and basket-cell output specificity in MICrONS only.