A null model of the mouse whole-neocortexmicro-connectomeMichael W. Reimann, Michael Gevaert, Ying Shi, Huanxiang Lu, Henry Markram & Eilif Muller
Blue Brain Project
Laye
r 23
Laye
r 4
Laye
r 5I
T
Laye
r 5P
T
Laye
r 6
PF AL
SoM
Vis
Med
Tem
p
PF
AL
SoM
Vis
Med
Temp
-3 -1-2-2.5 -1.5
ipsi-lateral contra-lateral ipsi-lateral contra-lateral
log 10 synapses / um3
PF AL
SoM
Vis
Med
Tem
p
PF
AL
SoM
Vis
Med
Temp
PF AL
SoM
Vis
Med
Tem
pPF
AL
SoM
Vis
Med
Temp
PF AL
SoM
Vis
Med
Tem
p
PF
AL
SoM
Vis
Med
Temp
PF AL
SoM
Vis
Med
Tem
p
PF
AL
SoM
Vis
Med
Temp
PF AL
SoM
Vis
Med
Tem
p
PF AL
SoM
Vis
Med
Tem
p
PF AL
SoM
Vis
Med
Tem
p
PF AL
SoM
Vis
Med
Tem
p
PF AL
SoM
Vis
Med
Tem
p
sourc
e re
gio
n target region
VISa VISrl VISal
VISli
VISpl
VISpor
VISpm
VISam
VISl
Source region Raw projection strengths
Not everyprojectioncompletely coversboth source ortarget region.This is capturedand quantified bythe model.
FRP
MO
sACAd
ACAv PL ILA
ORBl
ORBm
ORBvl
AId
AIv
AIp GU
VIS
CSSs
SSp-b
fdSSp-t
rSSp-ll
SSp-u
lSSp-u
nSSp-n
SSp-m
MO
pVIS
al
VIS
lVIS
pVIS
pl
VIS
liVIS
por
VIS
rlVIS
aVIS
am
VIS
pm
RSPa
gl
RSPd
RSPv
AU
Dd
AU
Dp
AU
Dpo
AU
Dv
TEa
PERI
ECT
FRPMOs
ACAdACAv
PLILA
ORBlORBmORBvl
AIdAIvAIpGU
VISCSSs
SSp-bfdSSp-trSSp-ll
SSp-ulSSp-unSSp-nSSp-m
MOpVISalVISl
VISpVISplVISli
VISporVISrlVISa
VISamVISpmRSPagl
RSPdRSPvAUDdAUDp
AUDpoAUDv
TEaPERIECT 0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.00 0.25 0.50 0.75Predicted innervation prob.
0.0
0.2
0.4
0.6
Obse
rved in
nerv
ation p
rob.
MOs - L5MOs - L2/3MOs - L6a
2
4
6
8
10
12
14
IPSI
IPS
I
CONTRA
CO
NT
RA
n.s
no data
10
20
30
40
50
Proj. density
IPSI CONTRA
Axo
n nu
mbe
r
FRP
MO
sACAd
ACAv PL
ILA
ORBl
ORBm
ORBvl
AId
AIv
AIp GU
VIS
CSSs
SSp-b
fdSSp-t
rSSp-ll
SSp-u
lSSp-u
nSSp-n
SSp-m
MO
pVIS
al
VIS
lVIS
pVIS
pl
VIS
liVIS
por
VIS
rlVIS
aVIS
am
VIS
pm
RSPagl
RSPd
RSPv
AU
Dd
AU
Dp
AU
Dpo
AU
Dv
TEa
PERI
ECT
FRP
MO
sACAd
ACAv PL
ILA
ORBl
ORBm
ORBvl
AId
AIv
AIp GU
VIS
CSSs
SSp-b
fdSSp-t
rSSp-ll
SSp-u
lSSp-u
nSSp-n
SSp-m
MO
pVIS
al
VIS
lVIS
pVIS
pl
VIS
liVIS
por
VIS
rlVIS
aVIS
am
VIS
pm
RSPagl
RSPd
RSPv
AU
Dd
AU
Dp
AU
Dpo
AU
Dv
TEa
PERI
ECT
First order innervation prob.predicted from projectiondensity
Pairs of regionstend to be innervated moreoften together
A tree-based model createsstatistical predictions of individualregion innervation profiles
Validation of the region innervation model
1 target 74%
2 targets 24%
3 targets 2%
Model-based neurons(6 visual areas)
Han et al., 2018
1 target 44% 2 targets 38%
3 targets 10%4 targets 6%5 targets 2%
1 target 53% 2 targets 28%
3 targets 14%
4 targets 5%5 targets <1%
1 target 35%
2 targets 33%
3 targets 20%
4 targets 9%5 targets+ 3%
Fluorescence-based MAPseq-based
1 target 56% 2 targets 36%
3 targets 8%4 targets <1%5 targets 0%
1 target 50% 2 targets 38%
3 targets 4%4 targets 8%
5 targets 0%
L2/3 L4
L5 L60.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
VIS
li
VIS
l
VIS
al
VIS
pm
VIS
am
VIS
rl
VISli
VISl
VISal
VISpm
VISam
VISrl
Area A
Are
a B
P(B
|A)
Model
P(B
|A)
VIS
li
VIS
l
VIS
al
VIS
pm
VIS
am
VIS
rl
Area A
VISli
VISl
VISal
VISpm
VISam
VISrl
Are
a B
Han et al., 2018
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
L2/3 L2/3
within
-reg
ion m
odule
sVIS
a
within-region modulesVISam
within-region modulesVISa
within
-reg
ion m
odule
sVIS
am
0 100 200 300 400 500
0
2
4
6
8
10
Unid
irec
tion
al c
on. pro
b. (%
)
0.0
0.2
0.4
0.6
0.8
Bid
irec
tion
al c
on. pro
b.
(%)
Sampling offset (um)200 400 600
Region radius (um)
1.00
1.25
1.50
1.75
2.00
2.25
2.50
2.75
Reci
pro
cal o
vere
xpre
ssio
n
Model instancesMean
0 100 200 300 400 5001
2
3
4
5
6
7
Reci
pro
cal o
vere
xpre
ssio
n
Model instancesMean
Sampling offset (um)200 400 600
Region radius (um)
2
4
6
8
10
12
14
16
Unid
irec
tion
al c
on. pro
b. (%
)
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Bid
irec
tion
al c
on. pro
b.
(%)
VISam->VISa
VISa->VISam
(VISa->VISam * VISam->VISa)
VISa<->VISam
10-6 10-5 10-4 10-3 10-110-2 100
0
50
100
150
200
250
300
350
Count
(model)
log edge density
0
500
1000
1500
2000
2500
3000
3500
Count
(contr
ol)
25
20
15
10
5
0
Edge
den
sity
(%
)
source: prefrontalsource: medialsource: somatomotorsource: temporalsource: anterolateralsource: visual
intra-moduleinter-module
log c
ontr
ol h
alf-
wid
th
1.2
1.0
0.8
0.6
0.4
log model half-width1 2 3 4 5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
[...] [...]
Model
Expected
Log-c
ount
MOs MOp
FRP
Neuron triplets
FRP
MOpMOs
Sampling radius 150 um
The modelpredicts over-expression ofcertain long-range motifs
modeldata
Rel
ativ
e fr
equen
cy
profile number
Per projection class
Per module: intra-module
Per module: inter-module
1 2 3 4 5 6
5pt
1 2 3 4 5 6
5it
1 2 3 4 5 6
4
1 2 3 4 5 6
23
1 2 3 4 5 6
6
1 2 3 4 5 6
visual
1 2 3 4 5 6
somatomotor
1 2 3 4 5 6
anterolateral
1 2 3 4 5 6
temporal
1 2 3 4 5 6
medial
1 2 3 4 5 6
prefrontal
1 2 3 4 5 6
visual
1 2 3 4 5 6
somatomotor
1 2 3 4 5 6
anterolateral
1 2 3 4 5 6
prefrontal
1 2 3 4 5 6
temporal
1 2 3 4 5 6
medial
Source
Target
source
target
Wang andBurkhalter (2007)
VISpl
VISpor
VISl
VISli
VISal
VISrl
VISa
VISam
VISpm
Yes
90 deg
None
180 deg
Yes
None
None
None
None
180 deg
Yes
None
None
90 deg
None
90 deg
Yes
None
Yes
90 deg
None
180 deg
Yes
None
None
None
None
180 deg
Yes
None
None
90 deg
None
90 deg
Yes
None
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
reflection?
rotation?
Yes
-
None
-
Yes
-
None
-
None
-
Yes
-
-
-
None
-
Yes
-
---
Juavinettet al. (2017)
ourmapping
Contact: [email protected] // bluebrain.epfl.ch
Strengths of individual projections Layer profiles of synapse densities
Validation of layer profiles
Validation of the mapping
FRP
MOs
ACAd
ACAv
PL
ILA
ORBl
ORBm
ORBvl
AIdAIv
AIpGU
VISC
SSsSSp-bfd
SSp-tr
SSp-ll
SSp-ul
SSp-unSSp-n
SSp-m
MOp
VISal
VISl
VISp
VISpl
VISliVISpor
VISrl
VISaVISamVISpm
RSPagl
RSPd
RSPv
AUDd
AUDp
AUDpoAUDv
TEa
PERI
ECT
FRP
MOs
ACAd
ACAv
PLILA
ORBl
ORBm
ORBvl
AIdAIv
AIp
GU
VISC
SSs
SSp-bfd
SSp-tr
SSp-ll
SSp-ul
SSp-unSSp-n
SSp-m
MOp
VISal
VISl
VISp
VISpl
VISli
VISporVISrl
VISa
VISamVISpm
RSPagl RSPd
RSPv
AUDd
AUDp
AUDpo
AUDv
TEa
PERI
ECT
1 5 10 15
Statistical interaction for innervations from MOs
P(MOs->A|MOs->B) / P(MOs->A)
Predicted innervation
Which combination of regions is innervated byindividual axons?Analyze innervation patterns of reconstructed axonsfrom Janelia Mouselight
Topographical mapping of projectionsRegion innervation profiles of individual axons
Incomplete region coverage
We model a topographical mappingbetween brain regions by defininglocal coordinate systems in thesource and target regions.Fit to projection data from the voxelized model of Knox et al., 2018
We first validate the model againstthe input data: How much is lost by
modelling only linear mapping?
Next, we validate against the literature.This is functional, not anatomical data!
Predictedmotifs
and micro-connectivity
We predict the fullmicro-connectivity
for over 10 million neurons
Reciprocal connectivity between regions
Multi-region connectivity motif counts
Long-range connectivity shaped by local clusters
We predict a connectivity principledescribed by Taylor et al., 2017 ex-
tends to single-neuron scale.
Julie A Harris, Stefan Mihalas, et al. The organization of intracortical connections by layer and cell class in the mouse brain. April 2018. doi: 10.1101/292961. URL http://biorxiv.org/lookup/doi/10.1101/292961.
Almut Schüz and Günther Palm. Density of neurons and synapses in the cerebral cortex of the mouse. Journal of Comparative Neurology, 286(4):442–455, August 1989. Joseph E. Knox, et al. High resolution data-driven model of the mouse connectome. April 2018. doi: 10.1101/293019. URL http://biorxiv.org/lookup/doi/10.1101/293019.
Charles R. Gerfen, Michael N. Economo, and Jayaram Chandrashekar. Long distance projections of cortical pyramidal neurons. Journal of Neuroscience Research, 96(9):1467–1475, September 2018. doi: 10.1002/jnr.23978.
Ashley L Juavinett, Ian Nauhaus, Marina E Garrett, Jun Zhuang, and Edward M Callaway. Automated identification of mouse visual areas with intrinsic signal imaging. Nature Protocols, 12(1):32–43,
Quanxin Wang and Andreas Burkhalter. Area map of mouse visual cortex. The Journal of Comparative Neurology, 502(3):339–357, May 2007. doi: 10.1002/cne.21286.
Yunyun Han, et al. The logic of single-cell projections from visual cortex. Nature, 556(7699):51–56, March 2018. doi: 10.1038/nature26159.
Peter N. Taylor, Yujiang Wang, and Marcus Kaiser. Within brain area tractography suggests local modularity using high resolution connectomics. Scientific Reports, 7(1):39859, December 2017. doi: 10.1038/srep39859.
References
Availability
Both the code, parameterizedconnectome constraints (recipe)and connectome instances areavailable at:
https://portal.bluebrain.epfl.ch/resources/models/mouse-projections/
Motivation: Connectomics takes place in two separate worlds: Micro-connectomics, studying connections between individual neurons, but at comparatively small scales; and macro-connectomics, studying connections at large scales, but only between regions of hundreds of neurons. We take a first step towards consolidating the two approaches. We start by considering the micro-connectivity implied by the macro-connectivity models, that of unstructured region-to-region connections, then apply a number of additional biological constraints:
We then find that these simple constraints lead to highly nonrandom micro-connectivity between regions.
Projections have a certain layer profile of synapse density in the target region
Projections implement a topographical mapping between regions
Individual neurons innervate only a subset of the regions targeted by their source region.
We start with macro- or meso-scale connectomics. Harris et al., 2018 report the strengths of projections between regions in the Allen Common Coordinate Framework. They provide strengths for five different projection classes. We convert the reported strengths to average synapse densities in terms of synapses um-3 using a predicted total number of synapses (88 billion, Schuz & Palm, 1989).
We predict and apply layer profiles of synapse density in the target region based on data from Harris et al., 2018
We validate the prevalence of profiles and theresulting densities against raw data of thevolumetric connectome model of Knox et al.,2018