home | lab | find me | science | publications | software | toolbox | site map

Reading Notes

What I read, when, some quotes and some thoughts.

In hindsight, I am a bit shocked that I have managed to meet nearly one author per paper over the last two years (including math/engineering people like Alon and Chklovskii).

Index


Geinisman, Y., Gundersen, H.J.G., van der Zee, E. and West, M.J. 1996. Unbiased stereological estimation of the total number of synapses in a brain region. Journal of Neurocytology 25: 805-19.

20080205

"Identification of synapses and their counting with dissectors."

"Each dissector consisted of two adjacent ultrathin sections, a reference section and a look-up section [...] The postsynaptic density (PSD) was used as a counting unit."

"[...] Two or more profiles belonging to the same PSD [post-synaptic density] of a perforated synapse (as determined by following them on consecutive sections) were connected by an imaginary straight line and treated as a single entity. [...] only those labeled synapses that had a PSD in the reference, but not in the look-up, section were counted."

"Unbiased estimation of total synapse number"

"Estimation of ultrathin section thickness."

v(dis) = a(fr) * h(dis)

where:

"Calculation of synaptic numerical density per unit volume"

Nv = Q / v(dis)

where:


"Mob Software: The Erotic Life of Code." An Essay in First Person by Richard P. Gabriel & Ron Goldman.

URL: http://www.dreamsongs.com/MobSoftware.html

20080202

"The swarm"

* Each individual shall steer toward the average position of its neighbors.
* Each individual shall adjust its speed to match its neighbors.
* Each individual shall endeavor to not bump into anything.

"The general principle for complexity design is this: Think locally, act locally."


Hanesch, U., Fischbach, K.-F. and Heisenberg, M. 1989. Neuronal architecture of the central complex in Drosophila melanogaster. Cell Tissue Res. 257:343-66.

"The central complex consists of four substructures. These are from rostral to caudal: (1) the ellipsoid body, (2) the fan-shaped body, ad below it (3) the paired noduli and (4) the protocerebral bridge. Closely associated with this arrangement of fibers are two further areas: (5) the paired ventral bodies, and (6) the paired lateral triangles."

"The central complex lies in the middle of the brain between the pedunculi of the mushroom bodies and is bounded laterally by the two antenno-glomerular tracts. Dorsally, it reaches the pars intercerebralis; below, the esophagus and the great commissure; and frontally the median bundle and the beta-lobes of the mushroom bodies."

"Strausfeld (1976) uses the term "central body" to refer to the close association of three substructures: the ellipsoid body, the fan-shaped body, and the noduli."

---> The term 'central body' is ambiguously used for different parts of the central body throughout the literature and should be avoided, in favor of 'central complex'.

"The largest of the parts is the fan-shaped body (fb). It is a regular structure of horizontal layers and vertical segments."

"Just anterior to the fan-shaped body, half embedded in its concavity, lies the ellipsoid body (eb). It is a perfectly round doughnut with the hole, the ellipsoid body canal, pointing anteriorly and slightly dorsally. (Since it is tilted, it appears as an ellipse in frontal sections.)"

"The ellipsoid body seems to be a specialty of dipterans. In other insects [...] its homologue is a half-circular structure, which is part of the central body."

"Ventral to the fan-shaped body and just caudal to the ellipsoid body lie the two mirror-symmetrically arranged noduli [no, a.k.a. 'ventral tubercles'] [...] These are roughly spherical glomeruli each segmented into two subunits along the antero-posterior axis."

"The fourth structure, the protocerebral bridge (pb), lies slightly removed at the dorso-posterior margin of the central neuropil embedded in the cellular cortex between the two calyces of the mushroom bodies. [...] The bridge is continuous across the midline [...]"

"The two ventral bodies [a.k.a. 'lateral accessory lobes'] are roughly spherical lobes just lateral, and partially rostro-ventral to, the ellipsoid body [...] They are bounded rostro-caudally and laterally by the beta-lobes and peduncles of the mushroom bodies and rostro-ventrally by the antennal lobes. The two ventral bodies are connected across the midline by the ventral body commissure, which runs ventral to the ellipsoid body and dorsal to the esophagus."

"The two lateral triangles abut the fan-shaped body rostro-laterally at the level of its largest lateral extent. They are also lateral to the upper half of the ellipsoid body and dorso-caudal to the ventral bodies. [they express GABA]. The lateral triangles are connected to the ellipsoid body by a prominent fiber tract, the later ellipsoid body tract."


Strauss, R. 2002. The central complex and the genetic dissection of locomotor behaviour. Current Opinion in Neurobiology 12:633-8.

20080118

(after recommendation from Michael Knust)

"The local neuronal circuits are found in the insect thoracic ganglia, which carry out tasks remotely comparable to those performed by the vertebrate spinal cord during locomotion."

"[...] in decapitated fruit flies the trunk is able to coordinate its legs if the cut cervical connective is non-rhythmically stimulated with biogenic amines in place of the missing 'go' signal from the brain."

"Walking analysis in Drosophila strains with mutations affecting brain structure soon associated the central complex [...] with functions related to higher locomotor control. [...] flies [...] with altered structure of the CX walk more slowly [...], react less quickly to changing stimuli during flight and show altered orientation behaviour toward landmarks. They are less active or quickly lose activity, or fail to start walking or flying [...]"

"The CX [..] comprises four neuropilar regions, the fan-shaped body, the ellipsoid body, the protocerebral bridge and the paired noduli, which are all interconnected by sets of columnar interneurons that form many regular patters of projection [...]"

"[The CX] receives input from most parts of the brain [...], however, no obvious prominent tract either from sensory areas or to motor areas exists that might prompt an easy guess at its functions."

[The paper describes leg locomotion details]

"The mirror-symmetrical architecture of the CX seems suited to exchange and to adjust neuronal information from both the brain and body halves. Consequently, 'right-left bargaining' has been among the suggested general functions of [CX]."

"[mosaic] experiments show that an intact brain can apparently compensate for existing side asymmetries in the generation of steps, whereas a C31-defective brain is incapable of either recognizing or correcting such asymmetries. [...] CX accomplishes the respective balancing functions."

Genes that generate mutations in the CX:

"The corresponding behavioural and morphological deficits in mutants of five different genes virtually exclude a mere coincidence. The specific regulation of the step length predominantly through the swing speed of legs does require the intact protocerebral bridge."

"In conclusion, an intact ellipsoid body is required to establish the fractal structure in the temporal pattern of the walking bouts. An intact protocerebral bridge and fan-shaped body along with connecting circuitry are required to up-regulate the walking motivation to a normal level, whereas the intact mushroom bodies are required to down-regulate it to the wild-type level. Regulation is achieved by either extending or limiting the bout length, respectively."

"Flies from a few of the lines with structural defects in the ellipsoid body and/or the fan-shaped body display altered orientation behaviour when walking or flying toward permanently visible attractive landmarks [...]. But flies from almost all of these lines quickly loose their bearings as soon as the chosen target landmark becomes invisible."

[i.e. the CX is necessary for the internal representation of the physical world, however simple that is in flies (perhaps just the current course)]

"[...] part of [navigational] functions are located in the ellipsoid body."

"[in locust] identified interneurons that are sensitive to polarised light [...] function in sky compass-mediated spatial navigation. [...] most of the neurons were found in the lower division of the central body, which corresponds to the ellipsoid body of Diptera."

"The precursor of the CX in larvae of holometabolous insects -except for that of legged maggots- is poorly developed compared with the CX of nymph stages of hemimetabolous insects."

[i.e. having legs and a proper CX seem to go together for insect larvae.]

"[...] the bar-bell shaped CX of late third instar Drosophila larvae has only about 1970 fibers within the interhemispheric commissure and hardly any synapses."

"But would the less demanding tasks of a Drosophila larva really already require the larval CX precursor? This seems to be the case because locomotor defects were found in early third instar larvae of six strains that were originally isolated for their adult CX abnormalities."

Summary:

"Genetic dissection of locomotion in Drosophila has identified the CX as a principal site of locomotor control in larvae and adults. One of its functions is to regulate locomotor activity in interplay with the mushroom bodies. It establishes the across-body symmetry of locomotion through right-left bargaining, facilitates landmark orientation and enhances walking speed through and adaptive increase of the swing speed of legs. Thus, comparatively easy-to-conceive technical functions such as the balancing of step length on both sides of the body, along with seemingly complex decision functions such as whether to initiate walking voluntarily, are located in the CX."

Speculation:

A possible unifying concept for CX-related mutants and observed phenomena may be that of a 'comparison centre', with winner-take-all events (and plasticity of the weights, etc) driving decision-making over competing sensory inputs.


Küppers, B., Sánchez-Soriano, N., Letzkus, J., Technau, G.M. and Prokop, A. 2003. In developing Drosophila neurones the production of gamma-amino butyric acid is tightly regulated downstream of glutamate decarboxylase translation and can be influenced by calcium. J. Neurochemistry 84: 939-51.

20080116 Flying high!

"Surprisingly, and in contrast to vertebrates, detectable levels of GABA occur late during Drosophila neurogenesis, after essential neuronal proliferation and growth have taken place and synaptogenesis has been initiated."

---> Perhaps the strong hard-wiring of Drosophila renders the early activatory role of GABA unnecessary. Carsten Duch said that larvae develop normally in panneuronal TTX expression; i.e. synapse activity is not necessary for the general wiring.

"Cell counts reveled on average 27 strongly and 25 weakly labeled GABA-positive neurones on either side of each segment. Thus, about 20% of all neurones show GABA immunoreactivity (in total approximately 250 neurones/hemisegment; Schmidt et al. 1997)."

"Double-labelling with anti-Elav antibodies suggest that GABA is restricted to neurones [...]"

"CNS development in Drosophila embryos can be roughly subdivided into three partially overlapping phases: a phase of birth and specification of neurones (ending at about 14 h), a phase of major axonal outgrowth (9-13 h), and a stage of synapse formation and neurite refinement (13-21 h)."

"We could not detect GABA before 16 h of development [...]. At this time, synaptogenesis in the CNS has started as suggested [...] by Synapsin and reports of first synaptic events and co-ordinated movements [...]"

"GABA staining appears in gradual fashion, some hours after first excitatory synaptic transmission occurs [...]"

Elav-Gal4 has pan-neuronal expression.


Zito, K., Parnas, D., Fetter, R.D., Isacoff. E.Y. and Goodman, C.S. 1991. Watching a synapse grow: noninvasive confocal imaging of synaptic growth in Drosophila. Neuron 22:719-29.

20080116 flying at 10000 m above Canada

"In contrast to the fairly static NMJ [neuromuscular junction] in mammals, the Drosophila NMJ is dynamic, growing new boutons and sprouting new branches during larval development."

"The Drosophila NMJ shares several important features with central excitatory synapses in the vertebrate brain: it is glutamatergic, with homologous ionotropic glutamate receptors [...]; it does not have a conspicuous basal lamina separating the pre- and postsynaptic sides (as does the vertebrate NMJ) but rather appears by ultrastructural analysis to consist of close membrane apposition; and it is organized into a series of boutons that can be added or eliminated during development and plasticity."

"[during development] there can be [...] 100-fold increase in the surface area of the postsynaptic muscle [... with] a concomitant growth of the presynaptic nerve terminal [... of] up to 10-fold increase in the number of boutons and a 10-fold increase in the number of active zones per bouton."

---> But the above numbers are maxima. A 2-fold increase is also normal, depending upon the specific muscle.

"The Drosophila NMJ exhibits activity-dependent plasticity in the extent of this synaptic growth."

"After synapse formation, Fas II becomes localized both pre- and postsynaptically, where it controls synapse stabilization [...]; in Fas II null mutants, synapse formation is normal but boutons then retract during larval development."

"Each abdominal hemisegment of the Drosophila larva contains a stereotyped pattern of 30 muscles that can be identified by their size, shape, and body wall insertion sites."

"These 30 muscles are innervated by ~45 motor neurons that make specific connections with particular muscles."

"We observed that, as the animal grows, preexisting boutons spread apart."

"New boutons are distinct from old boutons in both size and shape. Older boutons are larger and rounder in shape [...]"

"[...] the generation of a mature bouton, on average, spanned over [...] 48 hr at room temperature."

"[Figure 2] New boutons intercalate between existing boutons or add at the end of strings of existing boutons. [and boutons may undergo division]"

"We observed that new boutons often arose from existing boutons, either asymmetrically 'budding' in a manner akin to yeast division [...] or undergoing and apparently symmetric division with one old bouton dividing into two new boutons ['division'...]. New boutons also added directly from the axonal membrane ['de novo']. In summary, we observed no special growth zone. Rather, boutons added almost anywhere along the [NMJ] synapse."

"[...] the basic shape of the synapse was determined early on, and most of the synapse expansion occurred through an increase in the number of boutons, not through an increase in branching [only 27% add a new branch]."

"The formation of new branches always stemmed from a single preexisting bouton, rather than from the axon."

[If a bouton buds from an old bouton, then it directly has the characteristic ultrastructure of a mature bud despite its short age]

"[fig. 6] Some budding boutons have lower levels of [...] Fas II [when compared to mature boutons; and have the same levels of synaptotagmin]. [...] This suggests that the levels of [Fas II] may be locally modulated within a growing synapse."

[Summary]

  1. "Nascent boutons contain the same components as mature boutons"
  2. "We observed no special growth zone, and no special growth structure."
  3. "New boutons, which are relatively small, develop over time into larger boutons, with a concomitant increase in the number of active zones."
  4. "[...] at least some new boutons at the ends of strings of preexisting boutons exhibit a lower level of Fas II, suggesting [local modulation] during synaptic growth."

Landgraf and Thor. 2006. Development and structure of motoneurons. International Review of Neurobiology Vol. 75.

20071223

[i.e. Drosophila motoneurons]

"The motor system is composed of four principal components, which are segmentally repeated and amenable to targeted genetic manipulations. These are:
(1) a cuticular exoskeleton with hooklike protrusions (denticles) [...] moved by
(2) arrays of body wall muscles (30 per abdominal half segment [...] innervated by
(3) complements of motoneurons located in the ventral nerve cord [...] receiving input from
(4) sets of interneurons that are predominantly cholinergic [...]"

---> The mania in inserting literature references as author and multiple author surnames completely disrupts reading. Just like in conference talks: I don't attend them to be blasted with details that I am not interested in, and which prevent me from understanding what is being said. Ideally, references would be cited as numbers within the text, or not at all.

"[motoneurons] constitute the output node of the motor system [and] their functions are as much defined by their peripheral projections and postsynaptic targets [...] as they are by their central presynaptic connections."

"[...] segmentally repeated sets of motoneurons are generated during the first half of embryonic development [from neuroblasts]."

"[Each neuroblast] expresses a unique [...] combination of regulatory genes rather than a common motoneuron progenitor-specific code. This situation in insects differs from that found in vertebrates where most motoneurons are generated through a common genetic program occurring in a restricted set of precursors from a discrete region of the ventral neural tube."

"[...] motoneurons are born early in neuroblast lineages, each deriving from a separate ganglion mother cell which typically also produces a sibling interneuron."

"[In A1-A6], about 15 of the 30 existing neuroblasts give rise to [...] approximately 36 efferent neurons per half segment. Of these, 31 can be considered bona fide motoneurons since they express the main excitatory neurotransmitter glutamate and are capable of eliciting muscle contractions."

"Variations of this set of efferent neurons are apparent in [A7-A9 and T1-T3]."

"[...] neuroblasts are regulated by homeotic genes to produce segment-specific progenies [like in vertebrates]"

"The large diversity among motoneurons [... is] generated [...] through a series of hierarchical partitions [...]. The first subdivision [...] between ventrally projecting motoneurons (vMN) and dorsally projecting motoneurons (dMN)."

"vMN identity is controled by the combined action of the homeobox genes Nkx6 and hb9 while dMN identity is indicated and partially controlled by expression of the even-skipped homeobox gene."

"[On the specification of unique motoneuron identities] First, unique identities could emerge as the product of each cell's birth history [related to neuroblast specification genes][...] On the other hand, when similar motoneurons arise from a common progenitor [...], the temporal 'NB-clock' succession of NB determinants: Kr -> Hb -> Pdm -> Cas -> Gh that molecularly modify the progenitor cell, might participate [...] in the specification of unique motoneurons. [plus other speculative sources of uniqueness]"

---> It amazes me that, despite knowing lots of genes involved and knowing the developmental geometric choices of motoneurons, the authors did not create a computer model that could suggest missing factors, or the actual mechanics of the process, for testing then in the bench.


Alexandre V. Borovik. Mathematics under the microscope. v.0.919 20070905 12:39

20071222

Tropical mathematics: sum() and max() only. Also know as (max, +)-algebra

"Captures the essential properties of "distance" between species in the phylogenetic tree."

"A third one is computer science and the theory of time-dependent systems, like queuing networks [...] Indeed, we do not normally multiply time by time; instead, we either add two intervals of time [...] or compare the lengths of two intervals [...]"

"[...] tropical mathematics is the mathematics of time -which also explains its applications to genomics: phylogenetic trees grow in time,and the geometry of phylogenetic trees reflects the geometry of time."

---> The same could be said about neuronal trees and development / developmental biology.

"Still, why is Sudoku popular? I believe the answer is in a rhythm of repeated cycles of operations each of which engages our brain just up to a right and most pleasurable level of intensity."

---> Same about computer programming: a continuous stream of isolated pleasant moments. The best example that comes to mind is bug fixing (I recall Paul Graham already mentioned something along these lines regarding bug fixing.)

"We will not understand the psychological and neurophysiological roots of an important aspect of mathematical practice until we figure out why bubble wrap popping is such an addictive and pleasant activity."

"We have to do mathematics using the brain which evolved 30,000 years ago for survival in the African savanna." [Stanislas Dehaene, "The Number Sense"]

"In particular, should we be surprised it if were confirmed indeed that the most comfortable pace of execution of a recursive algorithm is set by a gene responsible for grooming behaviour?"

(To be continued)


T.S. Ray. 1990. Evolution, Ecology and Optimization of Digital Organisms.

200712XX

"I will consider a system to be living if it is self-replicating, and capable of open-ended evolution."

"The work presented here aims to parallel the second major event in the history of life, the origin of diversity [the first being the origin of life itself]."

"From a single rudimentary ancestral creature containing only the code for self-replication, interactions such as parasitism, immunity, hyper-parasitism, sociality and cheating have emerged spontaneously."

[I have to make mutation my food, diversity my flag, and selection my sword.]

"[Emergence:] In light of the nature of the physical environment, the implicit fitness function would presumably favor the evolution of creatures which are able to replicate with less CPU time, and this does in fact occur. However, much of the evolution in the system consists of the creatures discovering ways to exploit one-another."

"Because the fitness landscape includes an ever increasing realm of adaptations to other creatures which are themselves evolving, it can facilitate an auto-catalytic increase in complexity and diversity of organisms."

"Evolutionary theory suggests that adaptation to the biotic environment (other organisms) rather than to the physical environment is the primary force driving the auto-catalytic diversification of organisms ([34])."

---> Not only for organisms: also for individuals. The group of humans surrounding a human has a very acute effect on the expectations and performance of the individual.

"It seems that what we see [in the digital creatures] is what we know. It is likely to take longer before we appreciate the unique properties of these new life forms."

"Digital organisms program themselves, using evolution."


A. Fröhlich and I.A. Meinertzhagen. 1983. Quantitative features of synapse formation in the fly's visual system. I. The presynaptic photoreceptor terminal. J. Neurosci. 3(11)2336-49.

20071130

"The formation of precisely constituted synaptic populations in the developing nervous system involves two types of phenomena. The first requires assembly of individual synaptic contacts between physiologically appropriate neurons, by mechanisms which presumably depend ultimately upon neuronal recognition [...], and the second requires he regulation of the quantity or number of such contacts within any one functional class of synapse."

---> The Cadherin code in self-avoidance and cell class-avoidance is precisely at play for the first. What regulates the second other than activity patterns?

"In certain types of specialized synapses [...] where one neuron is presynaptic to clusters (diads, triads, tetrads, etc.) [referred to as 'multiple-contact synapses'] of postsynaptic elements at a single synaptic junction [...] more than one neuron must explore and recognize each synaptic target."

---> The authors keep trying to inject a "qualitative" versus "quantitative" classification of aspects of synaptogenesis, which is as unuseful as "morphalactic" and "epimorphic" terms are for studying regeneration in planarians.

" [multiple contact synapses] are perhaps the norm throughout invertebrate nervous systems [...] as well as in specific areas of the vertebrate brain [...]"


Ian Meinertzhagen. 1993. The synaptic populations of the fly's optic neuropil and their dynamic regulation: parallels with the vertebrate retina. Progress in Retinal Research 12:13-39.

20071130

The paper claims in the introduction to compare vertebrate and invertebrate visual systems, but it's a collection of very detailed information about Musca's visual system-related synapses.

Interesting to me are time sequence data plots, revealing that at early phases of development there are more but smaller synapses, which get reduced in number while increasing in size at later stages.


Big Ball of Mud. By Brian Foote and Joseph Yoder

20071129

URL: http://www.laputan.org/mud/

"Maintenance needs have accumulated, but an overhaul is unwise, since you might break the system."

---> This is the problem addressed by version control systems: one can branch out and perform a thorough overhaul of the system while it never stopped being a working system. 'git' excels at the task because, by allowing local branches and local commits [i.e. plenty of flexibility], it fosters experimentation and innovation by people who's only hope of authority is the presentation of a completely renewed, but working and history-traceable system to the powers-that-be.

"[...] if you can't easily make a mess go away at least cordon it off. This restricts the disorder to a fixed area, keeps it out of sight, and can set the stage for additional refactoring."


Melvin E. Conway. 1968. How do committees invent? Datamation magazine, April 1968.

20071129

URL: http://www.melconway.com/research/committees.html

"[...] the choice of design organization influences the process of selection of a system design [..] To the extent that an organization is not completely flexible in its communication structure, that organization will stamp out an image of itself in every design it produces."

---> Hence the failure of governments in combating the internet, for example. Or the failure of a single, but experienced and budget-constrained developer in creating any innovative design.

"There is a tendency to overpopulate a design effort [which contributes a lot to the disintegration of the organization] [... because ] A manager knows that he will be vulnerable to the charge of mismanagement if he misses is schedule without having applied all his resources."

"Assuming that two men and one hundred men cannot work in the same organizational structure (this is intuitively evident and will be discussed below) our homomorphism [between the structure of the human organization and the structure of the system they are designing] says that they will not design similar systems: therefore the value of their efforts may not even be comparable."

---> As a corollary, startup companies have a huge advantage over established companies in flexibility. In the information age, where access to and retrieval of knowledge are at the tips of nearly anyone, there is no longer the need to host in-house experts, which was the advantage of large organizations until now.

"Assumptions which may be adequate for peeling potatoes and erecting brick walls fail for designing systems."

---> I.e. not all tasks are parallelizable.

"As long as the manager's prestige and power are tied to the size of his budget, he will be motivated to expand his organization."

[Which leads to large, purposeless organizations, and then:]

"Probably the greatest single common factor behind many poorly designed systems now in existence has been the availability of a design organization in need of work."

The sentence would have been more powerfully written like this:

"Probably the [...] common factor behind [...] poorly designed systems [...] has been the availability of a design organization in need of work."

"Common management practice places certain numerical constraints on the complexity of the linear graph which represents the administrative structure of a military-style organization. Specifically, each individual must have at most one superior and at most approximately seven subordinates."

---> Same applies to a Principal Investigator and his lab: don't get too many postdocs, PhD students and technicians. If you do, ensure that some of the lower levels are controlled directly from intermediate levels, not from you as a PI!

"[...] organizations which design systems [...] are constrained to produce designs which are copies of the communication structures of these organizations."

"[...] a criterion for the structuring of design organizations: a design effort should be organized according to the need for communication."

"There is need for a philosophy of system design management which is not based on the assumption that adding manpower simply adds to productivity."

---> Besides reminding me so much of the book "The Mythical Man-Month", reminds me also of all or most of the mismanaged research labs I've seen so far, and also of the current hiring style of many companies: hire the very best people, even if very few, and assign them later; and do not hire anyone just because a project needs another team member. Enron also followed the "hire the genius" philosophy, but they failed to put an iron fist on their heads and left them switch projects continuously following individual wishes. Volker Hartenstein does this: infuses research project direction into anyone willing to join his group, but only accepts able people -or has been lucky that the system weeds out those not sufficiently able before candidates reach him.


A.H.D. Watson and F-W. Schürmann. 2002. Synaptic structure, distribution, and circuitry in the central nervous system of the locust and related insects. Microscopy Research and Technique 56:210-26.

20071128

"Electron microscopy reveals two major types of synaptic contacts between nerve fibers: chemical synapses (which predominate) and electrotonic gap junctions. The chemical synapses are characterized by a structural asymmetry between the pre- and postsynaptic electron dense paramembranous structures. [...] Synaptic bars are the most prominent presynaptic element at both monadic and dyadic (divergent) synapses."

"[Synaptic bars] are associated with small electron lucent synaptic vesicles in neurons that are cholinergic or glutamatergic (round vesicles) or GABAergic (pleomorphic vesicles). Dense core vesicles of different sizes are indicative of the presence of peptide or amine transmitters."

"The contacts between neurons that are interpreted as being the sites of chemicals synapses typically share a number of structural features: [1] synaptic vesicles associated with presynaptic electron dense membrane appositions, [2] a widened synaptic cleft, and [3] a sub-membranous elecron-dense coating in the opposing postsynaptic elements [...]"

---> I can't see [2] in first instar of Drosophila larvae. Perhaps I am not fixing properly.

"The sub-membranous densities and the material within the synaptic cleft are best visualized in aldehyde fixed, non-osmificated tissue (Schürmann, 1980) in which membranes themselves are left unstained (Fig. 1C)."

"While the postsynaptic element at central chemical synapses is almost universally another neurone, occasionally this can be demonstrated as a glial cell process (Fig. 1F)(Watson and Burrows, 1982)."

"In locusts, the bar, which is by far the most abundant presynaptic structure [...] and depending on fixation, a narrow plate may be seen running above it [...] This differs significantly from the presynaptic bars seen at the synapses in the [adult] fly brain, at which the bar has a prominent broad plate giving it a 'T'-shaped profile in cross-section (Burkhardt and Braitenberg, 1976; Fröhlich and Meinertzhagen, 1983)."

"At some diadic synapses, two or even all three of the neurites involved may contain a presynaptic bar (Burrows et al. 1989; Watson and Pflüger 1984). [common at the MB]"

--> What does this imply, that diadic synapses are reciprocal or what? I can't match my own understanding of what a diadic synapse is with the sentence above.

"[...] a single neuron may form both dyadic and monadic synapses with or without presynaptic structures."

"Pleomorphic vesicles [that is, variable in size and contents instead of homogeneous] are found almost exclusively in terminals that show evidence for the presence of GABA, while round agranular vesicles are found in terminals that are immunoreactive for glutamate [...] Neurons that appear to be cholinergic [...] also contain round agranular vesicles [...] Synaptic terminals containing round agranular vesicles are generally more numerous than those containing pleomorphic vesicles and in ultrastructural immunocytochemical studies of identified interneurons or efferent neurones, about 20-49 % of presynaptic terminals that contain electron-lucent vesicles exhibit GABA-like immunoreactivity. [...]"

"Large dense core vesicles with a diameter of 60-120 nm occur in all synaptic neuropiles of Orthopteran insects (Fig.3C-G). These may be scattered among electron lucent vesicles but more often lie in distinct clusters in the vicinity of the presynaptic membrane."

"Dense core vesicles are particularly abundant in certain neuropile compartments within the brain (e.g. lateral protocerebrum, central body complex, mushroom bodies, and deutocerebral neuropiles) that may be richly endowed with neuropeptides and biogenic amines [...] Such regions have been tentatively proposed as centres of neuromodulation (Schürmann et al., 1991)."

---> Centres of neuromodulation? What the hell is that? Doesn't fit at all with reward-maximization.

"An abundance of dense core vesicles is also typical of neurosecretory efferent neurons containing these classes of neurotransmitters [i.e. GABA, glutamate, acetyl-choline]"

"In the Orthoptera, biogenic amines such as 5-HT, octopamine, and histamine are found in terminals that make conventional synapses and that contain both dense core vesicles and electron lucent vesicles (Fig. 3C-E)."

"[...] serotonin is associated with large dense core vesicles in neurosecretory terminals of Rhodnius [...]"

"[The tannic acid -induced massive release of dense core vesicles into the extracellular space] suggests a paracrine form of release of neuroactive compounds (volume transmission) in insects, in addition to focal synaptic transmission associated with small agranular vesicles [...]"

"Though less common, electrotonic connections [electrical synapses] are present between neurones and between glial cells in the Orthoptera and other insect groups. [...] At these sites [...] about 0.1 nm in length, the membranes [...] lie only 0.2 nm apart [...]"

NOTE: only the mushroom bodies, the central body complex, and the antennal and optic lobes are organized into columns, strata and glomeruli with obvious geometrical order. The rest is "foam" like the VNC neuropile. In such foamy neuropiles ("diffuse") synapsin 1 is uniformly distributed.

"The concept of microcircuitry within the insect nervous system is well established (Pearson, 1979) and has received support by the finding of co-distribution of transmitters and neuromodulators in discrete neuropile compartments (Homberg ?)"

---> Need to read Pearson 1979. I sense that they are using the word 'microcircuitry' to mean something else other than 'geometry of the fine branching levels and their synaptic contacts'.

"Within discrete regions of neuropile containing repeated neuronal units, unusually structured synaptic complexes that comprise elements of microcircuitry (divergent, convergent, serial, reciprocal synaptic coupling) are found in both the vertebrate and the invertebrate nervous systems."

---> In the pipeline paper I must clearly specify what I mean by microcircuitry. Perhaps my usage of the term is totally incorrect and I should change it to "fine branching" or "tips of the arbors", "local wiring", etc.

"[regarding microcircuitry] A recent paper describes a special convergent wiring of cholinergic visual interneurons in the locust brain (Rind and Leitinger, 2000; [also Rind in Part II]). In comparison, by far the most intensely investigated insect brain neuropiles are the optic lobes of flies (see Meinertzhagen, 1993)."

"The total number of synapses of single identified neurones has so far only been estimated from selected samples of serial sections or from serial ultrathin sections of restricted parts of their central arborizations."

---> As the authors note: because the task is very labor intensive.

"[...] estimates of the numbers of synaptic contacts that comprise a single physiological connection between a pair of neurones also vary considerably depending on the neurones concerned. For some neurons, the number of contacts constituting a synapse may be as low as a few tens (Meinertzhagen, 1989) but more often this lies in the range of several hundreds if not thousands of synaptic contacts (Burrows et al., 1989; Simmons and Littlewood, 1989; Watson and Burrows, 1983)."

---> This is the number of synapses connecting any two neurons.

NOTE: labeling dendrites with horse-radish peroxidase (which is visualized as black precipitate) is ok for dendrites, because one can still see the pre-synaptic components (which are in different neurons, and thus not obfuscated in black). But for insect neurons, where dendrites are known and expected to have both pre- and postsynaptic components, such labeling would completely obliterate any chance of identifying presynaptic sites on the dendrite.

"The FETi [fast extensor tibiae motorneuron] is unique among the arthropod neurones [...] in the complexity of the local circuitry in which it participates [...]"

---> Its dendrites are small and curly and sport both pre- and postsynaptic sites all over, making a lot of recurrent connections with its partners.

"Local interneurones have all their neurites confined within a single ganglion or neurome or even a particular region within a ganglion. Functionally, they fall into two broad categories, those that produce action potentials and those that do not."

"[...] local spiking interneurones [...] have one set of branches in a sensory neuropile where they receive direct input from sensory afferent neurones [...] A short unbranching, axon-like neurite may link this field with a second field of branches that lie in a different region of the neuropile and that are predominantly sites of synaptic output. The branches of the input field are often of small diameter and relatively smooth, while those of the output field may be more varicose [...] (Watson and Burrows, 1985)"

"It has been suggested that on the basis of morphology alone, the polarity of the neurone and the distribution of input and output synapses can be inferred (Römer and Marquart, 1984; Strausfeld, 1976; Tyrer and Altman, 1974)."

---> Then they proceed to say that no, they have data to the contrary, but their data is not generalizable.

"Many populations of non-spiking local interneurones appear to have only a single arborisation within the neuropile, and on these branches input and output synapses may be intermingled though inputs are more abundant on smaller diameter branches and outputs on larger ones (Kondoh and Hisada, 1986b)."

---> Interestingly, the small size of insect brains and the average closeness of any two potentially interacting neurons may contribute to the the lack of need for digital transmission (i.e. spiking, which is a robust form of long-range information relay, is not needed, neurons can be analog, particularly interneurons).

BEWARE though: the authors describe a variety of possible interneuron branching I/O organization:
- fine branches for input, thick and varicose branches for output
- fine branches for input, thick branches for intermingled input and output
- thick branches for input, fine branches for output

Each one may correspond to a different neuronal type with different information integration and transmission needs and bottlenecks.

"In [non-spiking inter-] neurones where input and output synapses are intermingled, however, an output synapse will be most strongly affected by the inputs nearest to it."

"If input synapses from single or functionally equivalent presynaptic and postsynaptic neurones are clustered on particular branches, the interneurone might be functionally compartmentalized. [because of the analog operational mode] compartments would not be exclusive, rather inputs from different sources would constitute a series of overlapping spheres of influence having strong effects on nearby output synapses and progressively weaker influences on more distant output sites."

"Intersegmental [inter]neurones are spiking neurones whose axons link one ganglion or neuromere to another. They may ascend or descend between two adjacent ganglia or run for considerable distance along the nerve cord arborizing in several ganglia and within the brain Killman et al., 1999; Heinrich 2002)."

"[...] though as the authors point out [Peters et al., 1986], varicosities may be more reliable indicators of clustered mitochondria than synapse sites [at least for auditory ascending interneurones AN1 and AN2 of cricket prothoracic ganglion]."

"Further investigations of the functional morphology and role of mature synapses of identified locust neurons should be directed towards three important problems: (1) Understanding the subcellular topography of the synaptic complexes, (2) Investigating the principles of synaptic microcircuitry, (3) Unraveling the pharmacology of identified central synaptic interactions [...]."

"Neuropeptides outnumber classical transmitters by far (Homberg, 2002) and their distribution is concentrated in some brain areas, among them the central body complex."


B. Gerber and R.F. Stocker. 2007. The Drosophila larva as a model for studying chemosensation and chemosensory learning: a review. Chem. Senses 32:65-89.

20071128

"Adults and larvae are anatomically and behaviorally much different, reflecting their different lifestyles. Foe example, adult Drosophila flies need to find food (as well as mates, egg-laying sites, etc.), which requires sophisticated odor-driven behaviour. Fly larvae, in contrast, live on their food source and hence do not need long-range odor detection to find food."

Then the paper becomes totally unreadable. What a "review"! It's a listing of genes and organs.

The figures contain projection maps which are useful.

The authors become interested in their own script again at page 71 with "Larval versus adult olfactory circuits" and beyond.

"The salient feature of [larvae on agar plate] experiments is that larvae are attracted by odors' however, closer inspection reveals that, similar to adult flies, this is a concentration-dependent effect [that is, concentration of the odor in the air, not of larvae attracted to it]. At very low concentrations, larvae behave indifferently; at low-to-medium concentrations they are attracted, but as concentration further increases, they are eventually repelled by the odor."

"In cases where only attraction or repulsion is observed, this may be due to testing a too restricted range of concentrations; this is understandable as generating very high or very low odor concentration is technically challenging."

"Effectiveness of reward, but not of punishment"


Strausfeld NJ and Bacon JP. 1982. Multimodal convergence in the central nervous system of dipterous insects. In: Multimodal Convergences in Sensory Systems. Ed. E. Horn. Fortschritte der Zoologie. Band 28. Verlag.

20071123

"Sensory convergence in the lateral deutocerebrum."

"Axons from three sensory systems, visual, olfactory and tactile, converge at specific regions of the lateral deutocerebrum where they end amongst clusters of descending neurons."

---> the BPM and BPL neuropile compartments! In first instar, we have observed plenty of tracts starting off or converging into the lateral deutocerebrum.

"[the mushroom bodies] occupy a prominent location in the protocerebrum and are possibly involved in complex multimodal integration even though their most direct connection is with the olfactory lobes [..]"

---> With Wayne Pereanu's lineage connectome data we can disprobe this: they potentially receive connections from other sensory systems, but relayed through the central complex -not directly.

Note: the reason why males react to females so abruptly and irrationality may lay on the direct connections of the olfactory lobe (think pheromones) to the output regions of the brain, not the sensory integration regions, thus bypassing higher-order regions. Considering that evolutionary processes incur in the development of layered structures, the newer ones being functionally laid on top of the newer ones, it is no surprise that the easiest sensory modality, i.e. chemical, and its related variant olfaction, are at the core of the oldest animal behaviours such as reproduction.

"General comparisons between [olfaction] and thoracic mechanosensory neuropiles can be made: the antennal lobes, being ventral neuromeres of the deutocerebrum, are homologous to thoracic leg neuromeres."

"... the mushroom bodies merit intensive investigation with respect to their role as centers for long-term multimodal association, timing, learning and the generation of complex patterns of behaviour [...] The present anatomical evidence is that mushroom body outputs eventually interact with descending pathways. Possibily they influence the input selectivity and control the gain of certain descending neurons."

"In the entire cortex of a mammal it is difficult to distinguish more than five or six basic morphological types of neurons despite the multiplicity of cortical functions [...] and molecular heterogeneity [...]. In contrast there is an enormous variety of cell shapes in insects."

---> In insects functions may be encoded genetically, and thus neurons are diverse in both their genetic profile and their morphological appearance. In mammals, the cortex may be doing a single function: prediction, and thus the genetic profile of a neuron may have to do with the encoding of its long term state a lot more than with its specific function.

"Intracellular recordings show [the giant fiber] to be multimodal [...]. Seven out of nine endings on to it have been demonstrated [...] These are inputs 1 to 4 (visual), 5 and 6 (mechanosensory, from the antennae) and 8 (ascending from the thoracic ganglia)."


Douglas RJ and Martin KAC. 2007. Mapping the Matrix: The Ways of Neocortex. Neuron 56.

20071108

Peter's rule: neurons interconnect in proportion to the contribution to the neuropil of their dendrites and axonal synaptic boutons. From Baitenberg and Schuz, 1991.

"It is perhaps important to point out that these computational primitives [linear operations such as summation, division, and sign inversion, and non-linear operations such as winner-take-all, invariance and multistability] arise through the collective action of the whole circuit and are not carried out within the dendritic tree of single neurons."

---> My (silly) dreaming hypothesis for dendro-dendritic interactions between fly brain lineages included dendritic trees replacing whole cells ... which must be wrong, then.


Douglas RJ, Martin KAC, Whitteridge D. 1989. A canonical microcircuit for neocortex. Neural Computation 1:480-488.

20071102 at Salk Institute with Ping Wang

Rodney build a computer simulation of a simplified, modular neocortex that fits rather well into lineage-resolution fly brain maps.

What he put in the model:

Cortical model:

Unfortunately this paper does not exist in electronic form.


Geoffrey Miller. 2007. "Sexual Selection for Moral Virtues. The Quarterly Review of Biology, 82(2):97-125

200710XX

" ... viewed many ... prosocial behaviours as hard-to-fake indicators of animal fitness ... "

"... many prosocial behaviors that were assumed to arise through kinship or reciprocity are now thought to have emerged as costly signals of individual fitness favored by social and sexual selection."

"... most successful hunters, who provide the prosocial good of hunted meat [for the entire clan], also tend to attract more high quality female mates. [...] altruistic meat-provisioning was favored, at least in part, by sexual selection."

Hilarious translation:

"SHF, 26, seeks kind, generous, romantic, honest man."

as:

"Singe Hispanic female, 26, seeks a healthy male of breeding age with a minimal number of personality disorders that would impair efficient coordination and parenting in a sustained sexual relationship, and a minimal number of deleterious mutations on the thousands of genes that influence the development of brain systems for costly, conspicuous, altruistic displays of moral virtue."

"... whenever there are incentives to act like a better partner during courtship than after reproduction, the problem of trait stability arises."

---> If there is no incentive for keeping a smooth paired relationship for the rest of the pair's life, then it may be that males or females can act properly only during courtship, and not anymore afterwards. There are many instances of this in the literature and public spoken knowledge, fables, etc.

" ... a sexual selection account of moral virtues does not imply that males evolve all the conspicuous virtues and females play the passive role of virtue-assessment. Given mutual choice, both human sexes should show conspicuous, sexually attractive moral virtues during mate attraction and retention."

What costly signals do politicians show? None. Perhaps that explains their generally bad public perception.

"Moral culpability is a slippery idea, since everyone must be a joint product of their genes, environment, and random developmental events."

From Nietzsche's pagan virtues (leadership, bravery, strength, skill, health, fertility, beauty, tolerance, joy, humor, and grace) I feel a starting decline in two: health and beauty. My skin is no longer shiny and turgent, but shows small wrinkles; and I get sick easily if I overwork myself despite my very good physical output and overall athletic condition.

"For courtship to be reliable, valid, and discriminating as a moral test, it must lead to a perceivable range of moral failures (e.g. broken promises, revealed prejudices, irritabilities, infidelities, impatient sexual pressures) that reflect an underlying population distribution of moral traits."

"If neither individual in a sexual relationship cares about projecting moral virtues (as in relations between prostitutes and clients or masters and slaves), then the relationship is considered superficial and unloving."

"Borderline personality disorder (the tendency to view intimate partners in unstable, dichotomized ways, as alternately extremely good or extremely evil) is just an exaggerated form of the normal human tendency to alternately overvalue and undervalue our lovers' virtues."

"Courtship generosity may even include much of the paternal effort that is usually assumed to arise through kin selection (where 'kin' includes 'offspring'), since most divorced fathers reduce their paternal investment as soon as they are denied sexual access to mothers. Thus, what looks like simple paternal investment in one's offspring may turn out to be better described as ongoing courtship generosity by males to maintain sexual access to the mothers of those offspring."

---> Hey! It's women who are always enticing men sexually, for the purpose of gaining their purse! Well, the previous sentence is a half truth (and a half lie).


Roth, F., Siegelmann, H., Douglas, R.J. 2007. "The Self-Construction and -Repair of a Foraging Organisms by Explicitly Specified Development from a Single Cell." Artificial Life 13:347-68.

200710XX

"... nature does not encode the entire target system explicitly in the description. Instead, nature encodes modular rules that provide direct mapping between genotype and phenotype. [...] Modularity also promotes the automatic emergence of complexity through iterative application of simple rules, and the ability to adapt to environmental signals."

"The final organism is encoded in the description of the stem cell as a kind of a state machine in which the states, once activated, become persistent [cell and tissue differentiation]. Each state corresponds to a population of cells that has a particular functionality by virtue of the particular set of elementary intracellular mechanisms that it expresses. Transitions between the states (and so the development of later cell populations) are triggered by local environmental conditions, which are themselves a function of which states have been previously activated."

"... a simple multicellular organism that expresses attractive or aversive foraging behaviour analogous to a Braitenberg vehicle."


Douglas, Mahowald, Martin and Stratford. 1996. The role of synapses in cortical computation. J Neurocytol 25: 893-911.

200711XX

"Gray subsequently established the morphological criteria for distinguishing these two types [excitatory and inhibitory synapses] in the electron microscope (EM) in the neocortex that are still used today (Gray, 1959)."

"Parallel paths do converge in the brain. Sherrington conceived this in synaptic terms as a convergence towards a final common path: the means of the convergence of the inhibitory and excitatory pathways was through their synapses formed with their target neuron ..."

"Such linearity [of some cat visual cortex receptive field neurons] is a feature only of neurons with simple receptive fields, since the other varieties of receptive fields - complex and end-inhibited, are not linear. However, the very fact that there is a class of neurons, that exhibits quasi-linear behaviour is a puzzle, because in addition to the action potential threshold there are so many sources of non-linearities in neurons."

"If the inhibitory conductances were located proximally on the dendritic tree and were large, then the interaction between the excitatory and inhibitory synapses would be non-linear. The inhibitory conductances would 'shunt' the excitatory currents significantly and thus would appear to act 'divisively' [actually performing a division operation]. Small inhibitory conductances located on the proximal regions of the neuron, or inhibitory conductances located on distal portions of the dendritic tree would interact linearly with the excitatory synapses and produce the appearance that the inhibition was subtracting the excitatory current arriving at the neuron [...]"

"The fundamental formulation of our model of cortical microcircuits [...] is that large number of excitatory and inhibitory neurons are connected in recurrent circuits."

"[about recurrent networks] Thus, the major output of the neuron is to its nearest neighbours. However, the neuron's neighbours have similar morphology and thus the neuron may receive excitatory connections back from neighbours to which it has given connections. Such 'first-order' reciprocal connections between pairs of neurons have been reported between pyramidal cells in rat visual cortex [...]"

---> Is it possible that in Drosophila, considering that the axon is always very much detached from the dendritic tree, such recurrent networks are generated by dendrodendritic connections?

" [since both excitatory and inhibitory neurons are plugged into a cortical spiny stellate] the overall 'gain' of the system is to some degree embedded in the physical connections."

"Each [spiny stellate cell] makes about 5000 synapses, of which about 1200 are made with other spiny stellate cells. We have assumed that about one third of the boutons occur in the primary cluster and that they are homogeneously distributed in the three-dimensional space. [...] we make the simplifying assumption that local connections are made between spiny stellates on a random basis [... ; ...] we have calculated that for axonal arbors whose primary clusters have standard deviations of 100 microns, the spiny stellate neuron would participate in 117 first-order recurrent connections; if the standard deviation of the primary cluster increases to 150 microns, which effectively reduces the synaptic bouton density, then a given spiny stellate would participate in only 34 such pairs [...]. While it is obvious that the more first-order connections a given neuron is involved in, the stronger will be the synaptic current it receives from those neurons, this is a very interesting case for consideration, because the effects of the feedback may differ significantly between the two cases. It might be expected intuitively that if the excitatory current is sufficient to drive the interconnected neuron to threshold, then, in the absence of any inhibition, the neurons in the network will be driven to their maximum discharge rates. Inhibition, it seems, is essential. Surprisingly, however, analysis of these first-order excitatory networks indicates that in some configurations they can be stable, in the sense that they remain bounded without the restraint of saturation, without the addition of inhibition. [...] The essential problem is how to control the gain, or amplification, that is inherent in these recurrent excitatory circuits."

"Although the synapses themselves are linear, the change in slope of the FI curve [spike frequency versus current] during recurrent inhibition is proportional to the output of the neuron. Interpreted physiologically, feedback inhibition that acts through approximately linear 'subtractive' synapses nevertheless generates a network conductance that appears as a shunting-like inhibition of the output of the neuron. This inhibition changes he gain of the cortical response to a given input current. This crucial insight provides the solution to the problem posed by the experimental data. It explains how shunting, or divisive inhibition can be revealed in the spike discharge of the neuron [...], while intracellular recordings indicate the shunts associated with inhibitory events are extremely modest [...] and largely induce small amplitude hyperpolarizations."

"[...] orientation tuning of cortical neurons remains largely unchanged over considerable changes in the contrast of the stimulus [...] and is surprisingly robust in the face of potentially confounding stimuli [...]. Such invariance in the face of noise is in fact a cardinal property of cortical processing and visual perception."

[about the simulation, explaining how recurrent connections (both activatory and inhibitory) result in enhancement of orientation selectivity]

[[ NOTE: the paragraph below explains how competitive networks can compute]]

"In this circuit, the inhibitory neuron acts as a summing device and provides the same inhibitory current to all 40 spiny neurons, in proportion to the total activity in the circuit. Virtually all 40 spiny neurons are active initially, because geniculate relay neurons will response to a stimulus of any orientation [...]. As the total activity in the network increases due to the positive feedback, the excitatory drive to the inhibitory neuron increases. Neurons that are just above threshold will thus be silenced by inhibition and they will no longer contribute to the total excitation [...]. However, the pattern of connectivity ensures that they will continue to receive an inhibitory input, and so their threshold for activation will continue to be raised as the total activity in the circuit increases through the recurrent excitation between remaining active members. The inhibited neurons will always be these who were most weakly driven by the stimulus and thus received the least amount of feedforward synaptic current from the geniculate relay neurons. The neurons who were most biased for the stimulus orientation will receive the most feedforward current [why, no need! what they receive is the most recurrent excitatory input!]. They will produce the most activity and re-excite each other most strongly and thus remain above threshold despite the recurrent inhibition. In this way the initially small input from the lateral geniculate relay neurons will be amplified selectively by a subgroup of the total 40 excitatory neurons to provide a robust orientation signal."

"Once the stimulus is presented, the tuning of the inhibition rapidly changes from broad band to narrow band."

---> The fact that activating a neuron is hard, i.e. several incoming spikes have to be integrated for a neuron to fire in response, is already an element of the system, for the activation of specific orientation selective neurons. Other neurons need temporal integration to get enough stimuli to spike, but it never happens because of the recurrently activated, generally connected inhibitory neurons.

"In the recurrent configuration [...] it is interesting to note that the role of inhibition changes over time. Initially, it is used as a means of setting a threshold to extract the best estimate of the signal arising from the visual input. Later, inhibition assists in stabilizing the recurrent excitation of the active population: as the excitation grows, so the inhibition grows proportionately and acts as a divisor. This proportionality is important in helping to maintain a balance between inhibition and excitation."

"To reduce the firing of the neuron to zero, the inhibition has only to be the magnitude of the smaller feedforward component of the excitation."

"There are a number of feature that become apparent in this simulation. [1] First, the system is robust in the face of noise. The pattern of intracortical connectivity ensures that even if the stimulus is embedded in noise, there will be a subpopulation of 'winners'. The selective, noise resistant properties arise out of the actual pattern and weighting of the synaptic connections between the cortical neurons, here modelled as a Gaussian weighting. This property does not arise out of feedforward circuits. [2] Secondly, the circuit performs a version of gain control or 'normalization'. [When Rodney says "divisive" he means normalizing:] "The average activity of the excitatory network determines the strength of the recurrent inhibition and thus the threshold for spike activity. When the network has converged, then the inhibition is essentially divisive: strong inhibition is seen for strong excitation, weak inhibition for weak excitation. This extraction of invariance is one cardinal feature of cortical operations that feedforward models lack. [3] Thirdly, it is obvious that the current threshold required to produce a minimal discharge for a given neuron will vary accordingly to the amount of recurrent inhibition it receives at any moment. In this respect, the circuit responds dynamically to the incoming stimulus."

"Finally, this form of circuit [the recurrent circuit with high number of local recurrent connections, both inhibitory and excitatory neurons included] provides a reconciliation of the disparity we detected in the experimental data: the observation of divisive or shunting inhibition [...], with the failure to detect this shunt biophysically [...]. The division arises out of the network conductance generated by the recurrent inhibition and not, as was previously thought, out of the individual inhibitory synaptic conductances of the neuron."

"The recurrent circuit also permits the excitation to be controlled by small levels of inhibition, provided that the inhibition is provided to all members of the excitatory circuit."

---> Same thing for transcription factors and gene expression in general!

"What chiefly distinguishes cerebral cortex from other parts of the central nervous system is the great diversity of its cell types and interconnexions. It would be astonishing if such a structure did not profoundly modify the response patterns of the fibers coming into it. [Lorente de Nó, 1949]"

---> This paper scores among the top five papers I've ever read. Well written. concise but properly detailed, without a single superfluous paragraph. It has 3 typographic mistakes in the form of misplaced words in half-completed sentences and a duplicated draft-like sentence, but beyond these ignorable stains, it's fabulous. I finally grasped the kind of computation that the brain cortex may be performing -at least, as the authors imagine it.


Douglas RJ and Martin KAC. 2007. "The butterfly and the loom". Brain Research Reviews.

200710XX

" ... the dimensions that astrophysicists and astronomers talk about have their parallel in the numbers that anatomists deal with. In the neocortex we have 10 thousand million nerve cells, which is the same order of number as the starts in our galaxy. The number of connections they make are the same order as a thousand galaxies."

[The neocortex has some sort of compartments:]

"Sharp transitions [in electrophysiological recordings] were observed from a region with one set of properties, to the adjacent region with different properties."

[Heraclitus implicit:]

"The conundrum pointed to by Sherrington was how two successive touches ever felt alike since the peripheral activity was entering a cortical network whose state was continually varying. Since subjectively successive touches can feel alike, his conclusion was that there is a mystery still to be solved in sense perception. Despite the extraordinary technical advances in recordings from awake humans and non-human primates, this stability of perception remains a mystery."

"Only 1/1000 fibers in the white matter itself connects to a structure other than neocortex."

"Thus there was no way to assign to the layers of the cortex the separate tasks of reception, association, and projection."

"[Fig. 8 (A)] The circuit proposes that the cortex is composed of three dominant populations that interact with one another. One population is inhibitory (smooth cells, filled synapses), and two are excitatory (open synapses). The latter represent superficial (P2 + 3) and deep (P5 + 6) layer pyramidal neurons, respectively."

"After 2 years of intense study [of human fetuses], his [Cajal's] conclusion was that the superiority of the human brain arouse from, 'the prodigious abundance and unaccustomed wealth of forms of the so-called neurons with short axons' [...] We would today call these the inhibitory neurons."

---> Rodney notes, though, that other primates have a similar average: about 15-20% of the cortical neurons are inhibitory neurons.

"Both Cajal and Sherrington were artists, imaginative, widely read, charismatic, hugely energetic, enormously fluent with prose and poetry. In their work they ever sought the bigger picture, the integration of small facts into larger concepts. They had formidable powers of observation."

"' [...] The intense anthropomorphism of his [Cajal's] descriptions of what the preparations showed was at first startling to accept ... We must, if we would enter adequately into Cajal's thought in this field, suppose his entrance, through his microscope, into a world populated by tiny beings actuated by motives and strivings and satisfactions not very remotely different from our own... Listening to him I asked myself how far this capacity for anthropomorphizing might not contribute to his success as an investigator. I never met anyone else in whom it was so marked.' (Sherrington ..)]"

---> I don't like to anthropomorphisize, but: to what extent I do so, unaware? I have no idea.

"Anyone browsing through the early scientific literature that is the foundation of modern neuroscience will soon discover that in the formulaic, bland, and often semi-literate writing of contemporary science, we have lost an undeniably human presence. The passionately engaged, literate author, in energetic pursuit of nature's secrets, is now an endangered species."


Erich Staudacher. 1998. Distribution and morphology of descending brain neurons in the cricket Gryllus bimaculatus. Cell Tissue Res (1998) 294:187±202

200709XX

"Various groups of cells and single neurons have been identified, and the morphology of more than 40 cells is described. Nearly 200 descending brain neurons can be stained via one cervical connective."

"The main arborizations of the cells from the prominent ventral i5 group are found in the same part of the protocerebrum. In contrast, various cells arborize in the ventral posterior deuterocerebrum, but their somata are not located in different clusters. Thus, neurons from the same cluster may, but need not necessarily, arborize in the same brain area."

"Hedwig (1994) has identified a command system [...] consisting of at least two pairs of descending cells, for stridulation [i.e. singing by friction of body parts] in grasshopers."

[He already describes contralateral cell bodies in descending projections.]

"The remaining, more posterior clusters were not arranged in any obvious pattern (Fig. 2)."

--> clear reference to the posterior crescent group.

"... there are more clusters descending ipsilaterally than contralaterally."

[On ipsilateral descending neurons:]

"... irrespective of the location of their perikarya, these cells arborize in dorsal and ventral areas of the brain but never enter any of the glomerular neuropils of the brain (mushroom bod, central complex, antennal lobe)."

"Most ipsilateral descending neurons with a soma in one of the protocerebral clusters arborize in the diffuse neuropil of the protocerebrum and in the deutocerebrum."

---> he notes, though, that deuterocebral cells may differ in this respect.

The 'i' and 'c' cluster naming convention originates in "ipsilateral" and "contralateral".

Contralaterally-projecting neurons ('c') seem to belong to clusters containing also neurons projecting ipsilaterally.

If anything:

Identity [as in table 1]:

Functionality:

In crickets, it looks like the axes are differently situated, for the calix is dorsal, but the alpha lobe (dorsal lobe) of the mushroom body doesn't point dorsal.

Reads to me that they consider the neuraxis for axial spatial cues, i.e. fly brain "posterior" (which is dorsal to the neuraxis) is considered dorsal here.

"Four of the six contralateral clusters are grouped around the mushroom body calyx, [..] one cluster is situated in the anterior ventral deutocerebrum. [..] one cluster [..] in the tritocerebrum [there are no ipsilateral clusters in the tritocerebrum]."

Clearly some contralaterally-projecting neurones have their bodies in the posterior crown group, near medial, near the pars intercerebralis. I have DiI of them, plus they are mentioned as well in Boyan and Williams 1981 (for cricket). Such neurons respond to auditory stimulus.

"Like ipsilateral neurons, the contralateral cells are never found to arborize within any glomerular neuropil and, independent of soma location, the branches of these cells are found in most parts of the brain, except the optic lobes. [...] in contrast to ipsilateral descending neurons, many of them arborize in both hemispheres [...]"

The author describes examples of single cells that project contralaterally and then, in the contralateral neuropil, the axon describes a loop before projecting downwards.

"[...] there seems to be no obvious rule governing the distribution of somata in relation to the (dorso-ventral-lateral) location of their descending axons in a connective."

"only 0.03 - 0.15 % of all cells descending to the thoracic ganglia"

"... descending cells only represent about 6.5% of the intersegmental neurons running through this connective. This suggest that they are greatly outnumbered by cells descending from the suboesophageal ganglion and intersegmental neurons with axons ascending to both head and ganglia."

"The relatively small number of descending neurons suggests that they may be more important for the initiation of behavioral responses and less important for determining details of their performance."

--> seems to correspond very well with the situation in mammalian cortex, where 97 % of all synapses into a layer 4 neuron are from the cortex itself; only a tiny 3% are thalamic.

"For example, with respect to a simplified model of cricket singing and grasshoper stridulation, command neurons in the brain determine whether the animal sings or not, the central pattern generator in the pterothoracic ganglion organizes the basic pattern, whereas the suboesophageal ganglion determines quantitative aspects of the song."

---> the author notes that the situation may not be so simple for other sensory modalities and motor systems, such as walking, were i5 plays a direct role.

"In general, the arborizations of descending cells are found in all parts of the brain except the optic lobes."

Clusters definition: two criteria:

  1. the primary neurites of all the cells of the same cluster should enter the ganglion at the same location.
  2. somata of the same cluster should be surrounded by a common glial sheath.

"A cell-lineage analysis of the neurons of the Drosophila brain has revealed that somata of s single cluster are not packed tightly at the end of metamorphosis (Ito et al. 1998)."

"... neurons of the same cluster innervate the same brain neuropils, although they may show different branching patterns within these neuropils (Ito et al. 1998)."

---> In crickets they find some differences, but not as big as the author claims -from his own data- or, rather, the structure of dendritic arborization is not so simple as stated in Drosophila.

"Outside the Orthoptera and Blattaria, no homologous descending cells have been found [..]"

---> ??? There are obvious correlations, at least for cluster 3D location, and also regarding generic descending neuron axonal projection pattern.

Clusters and function:

"... neurons with somata in different clusters (e.g. i5, c5) may respond to similar stimuli, as found for cricket auditory neurons [...]"

---> The author concludes that related (and thus named similarly) ipsi- and contralateral clusters are not related. Uh?


Ryuichi Okada, Midori Sakura, and Makoto Mizunami. 2003. Distribution of Dendrites of Descending Neurons and Its Implications for the Basic Organization of the Cockroach Brain. The Journal of Comparative Neurology 458:158–174.

"We propose, based on our results and documented findings, that many parallel processing streams [direct connections] function in various forms of reflexive and relatively stereotyped behaviors, whereas indirect pathways [mediated by interneurons] govern some forms of experience-dependent modification of behaviour."

"In general, central release of relatively stereotyped behavior seems to be controlled mainly by labeled lines, and more plastic behavior seems to be activated by populations of DNs [Descending Neurons]".

"Staudacher (1998) showed, by retrograde dye fillings from the cervical connective, that there are nearly 200 pairs of somata of DNs in the brain of the cricket, and the classified these neurones into 17 clusters based on soma locations."


Pflüger H.J., Bräunig P. and Hustert R. 1988. The organization of mechanosensory neuropiles in Locust thoracic ganglia. Phil. Trans. R. Soc. Lond. B 321:1-26.

Sensory organs projecting/receiving to/from the thorax:

Tyrer & Gregory made an anatomical guide for the locust thoracic nervous system (based on anatomical studies of the cockroach nervous system by Pipa el al (1959) and Gregory (1974).

"[The guide allows for the description of neuronal arborizations] in relation to commissures and prominent longitudinal tracts which coalesce to for the connectives."

"A further conspicuous structural feature in sections of ganglia is the unstructured, 'amorphous' neuropiles formed by areas of densely packed small diameter fibers. The density of fibers in these neuropiles is greater than anywhere else in the ganglia ..."

"The ventral neuropiles are thought to be predominantly sensory [...] whereas the dorsal neuropiles contain a mixture of mechanosensory, motor, interneuronal and neurosecretory fibres [...]"

"This paper attempts to show the extent to which different sensory modalities project to specific areas of neuropile."

Abbreviations of described sensory modalities:

I love the wording of this paper:

"In representative cross sections the neuropilar areas in which their afferents terminate are shown and their branching patterns are described in relation to ganglionic landmarks."

"Binet (1894) described the ventral regions of a ganglion of a better as 'sensory association centres'."

--> It's been known since forever! How come there isn't any insect nervous system book that could educate me with the basics?

Figure 1. (a) Ventral view of a mesothoracic ganglion cut into 16 micron transverse sections starting with section 1 (first cell bodies visible) ending with section 53 (last cell bodies visible). Prominent neuropiles, commissures and other structures in the sections are indicated as black bars. Thickness of bars gives a relative measure of the size of the structures in the sections. (b-m) Camera lucida drawings (left) and photographs (right [destroyed by a careless photocopier]) of selected sections of a mesothoracic ganglion stained with osmium and ethyl-gallate. Numbers in brackets refer to section numbers in (a). Calibration 100 microns; see List of abbreviations used.


Sprecher SG, Reichert H, Hartenstein V 2007. "Gene expression patterns in primary neuronal clusters of the Drosophila embryonic brain" Gene Expression Patterns 7: 584-95.

"We searched for [] markers among the large number of patterning genes defined for the embryonic neuroblast map [], and we describe here 18 genes whose expression persists in the late embryonic brain."

---> These 18 markers would be great for my neurite-BLAST program.

---> Elav-Gal4 is expressed in all lineages: good candidate for MARCM.

---> They did it all IGNORING the neurite projection pattern!

"Most molecular markers are expressed in NBs and their progeny in a highly dynamic pattern. For instance _engrailed_, even though roughly expressed in all cells of the described projection clusters, shows a temporally dynamic expression pattern, in that it declines as neurons maturate. Other markers, such as _orthodenticle (otd) or _eyes absent_ (eya), remain expressed in in all neurons of some clusters, whereas they are confined to only a few cells in other clusters. Only a few, if any, of the molecular markers described in this study show a 'simple' continuous expression pattern which includes all neurons of a fixed set of primary clusters."

" [...] our analysis revealed that clusters that by late stage 11 [the stage represented in the neuroblast map of Urbach et al. (2003)] express a given gene, generally maintain some level of expression of that gene in the late embryonic brain."

---> Table 1 has a nice listing of lineages and the 18 markers. There's plenty of overlap, but each lineage has its own pattern of expression. Now if only we could identify them easily, using the neurite projection BLAST!


Peter Norvig: Artificial Intelligence - A Modern Approach.

"Goal formulation [...] is the first step in problem solving. [...] help organize behaviour by limiting the objectives that the agent is trying to achieve."

"Problem formulation is the process of deciding what actions and states to consider, and follows foal formulation."

"A search algorithm takes a problem as input and returns a solution in the form of an an action sequence. [...] Thus we have a simple 'formulate, search, execute' design for the agent [...]'

"[...] there are four essentially different types of problems--single-state problems, multiple-state problems, contingency problems, and exploration problems."

"The output of a search algorithm is a solution, that is, a path from the initial state to a state that satisfies the goal test."

"The process of removing detail from a representation is called abstraction."

"The choice of a good abstraction thus involves removing as much detail as possible while retaining validity and ensuring that the abstract actions are easy to carry out."

"The right formulation makes a big difference to the size of the search space."


Douglas RJ & Martin KAC. 1991. A functional microcircuit for cat visual cortex. J. Physiol. 440:735-769.

"Most putative inhibitory synapses are not on spines, but on proximal dendritic shafts [...], which are probably electrotonically close to the soma and thus should be visible in intracellular recording."

--> in insects, close to the spike-generating mechanism, I bet.


Maye, A., Hsieh, C., Sugihara, G. and Brembs, B. 2007. Order in spontaneous behavior. PLoS One, May 16.

"... that one of the evolutionary benefits of this superimposed spontaneity is to be able to discern which of the incoming complex stream of sensory stimuli comes as a result of our own actions and which doesn't. This is actually a very old hypothesis, ... "

URL: http://brembs.net/spontaneous/

Using the so-called "S-Map Procedure" the researchers detected a non-linear signature in the fly behavior. Such a signature can only be found in systems whose indeterminate behavior is not due to noise but originates in their design. "This signature indicates that there is a function in the fly brain which evolved to generate spontaneous variations in the behavior" Sugihara said. "This function appears to be common to many other animals and could form the biological foundation for what we experience as free will".

Björn Brembs adds: "Our subjective notion of 'Free Will' is essentially an oxymoron: we would not consider it 'will' if it were completely random and we would not consider it 'free' if it were entirely determined."

Nobody would attribute any responsibility to our action if it had happened entirely coincidental. On the other hand, if our action was completely determined by external factors such that there was no alternative, again the person would not be held responsible. So if there is anything remotely close to free will, it must exist somewhere between chance and necessity - which is exactly where fly behavior comes to lie.

"The question of whether or not we have free will appears to be posed the wrong way," says Brembs. "Instead, if we ask 'where between chance and necessity are we located?' one finds that this is precisely where humans and animals differ".

With this small reformulation, the topic of free will becomes the new biological research area of studying spontaneous behavior and can thus be discerned from the philosophical question.

Brembs emphasizes: "Regardless of our speculations on free will, the most important scientific aspect of our work is the evidence we found for a brain function which appears evolutionarily designed to always spontaneously vary ongoing behavior. There is tentative evidence that such a function may be very widespread in the animal kingdom, including humans. Why would all brains have this function?" If this indeed turned out to be the case, the scientists may have discovered the first evidence for something truly fundamental to the understanding of brains.


NJ Strausfeld, U Bassemir, RN Singh, JP Bacon. 1984. Organizational principles of outputs from dipteran brains. J. Insect Physiol. 30(1):71-93.

20070504

20070122

"Most supraoesophageal descending neurones (DNs) arise in the deuterocerebrum. Their dendritic fields extend outwards from the midline towards the optic lobes. Typically, dendrites of DNs are subdivided into several domains and receive several types of sensory inputs."

"DN axons project dorsally in the ventral nerve cord to one or more target neuropils in one or more thoracic ganglia."

---> No wonder most DiI backfills must be done on thoracic segments to catch any DN in the brain.

---> The term "Command" interneurones is also present in this paper. Same with 'neck connectives' (target of the most recent paper).

"In Drosophila, the GDN [giant descending neurones] is known to mediate the jump response via the tergotrochanter motor neurone (TTM) [...] In Musca, as in Drosophila, the TTM neurone and the GDN are cobalt-coupled [i.e. connected by gap junctions along with normal chemical synapses]."

It is clear from figure 4 that neuron's thicker neurite is not necessarily stemming from the cell body. Not at all! This corresponds to my observations as well from DiI injections and genetic GFP labeling.

In addition, the above correlates with the idea that neuronal computation occurs at the dendrite tree, and the cell body has little to do in signal transmission, at least in insects. Given that insect neurons can be made to be bipolar as well (Sánchez-Soriano's paper), one may attempt the conclusion that the situation is the same in vertebrates; that the cell body position at the interface between the dendritic tree and the axonal tree is accidental or a mere convenience obeying volumetric constraints and minimization.

Fig. 13 A shows the 3 tracts (medial, intermediate and lateral) of each half of the VNC, although such tracts are never described in the text (so far).


Jeff Dawkins on Hirarchical Temporal Memory (HTM)

URL: http://www.spectrum.ieee.org/apr07/4982/3

"... the only way to train an HTM [hierachical temporal memory] is with input that changes over time. How that is done is the most challenging part of HTM theory and practice."

"Strange though it may seem, we cannot learn to recognize pictures without first training on moving images."

---> Yet another explanation on why we perceive the 3D structure when moving through a stack at high speed.

---> A structure is classified as noise (such as a lead citrate precipitate drop) if, through time and space, it exists briefly.

"HTMs are hierarchical, dynamic memory systems."

Hierachical Hidden Markov Models are similar to HTM, but simpler, and have trouble with spatial variation, as if "you could learn melodies but not be able to recognize them when played in a different key."

".. they work [...] handling distortion and variances in visual images."

---> HTMs smell like Lisp: programs and data are one and the same. Yet numenta has implemented them in C++! They must have reinvented Lisp in the process.


Chklovskii's website

URL: http://research.janelia.org/Chklovskii/past.html

"We suggest that the sizes of axonal and dendritic arbors are chosen to minimize wiring volume. Here is a simple example to illustrate this idea. Suppose you need to wire up a topographic projection between two layers of neurons - circles and squares in (A) below. Two arrangements that implement this wiring diagram with realistic neurons are shown below (B, C). These two arrangements have the same electrical properties (barring non-linear interactions in dendrites) but differ in the placement of synapses. Arrangement (B) is preferred because it has shorter total length of wires (i.e. axons + dendrites) than arrangement (C).Therefore, topographic projections from many neurons to few should have wide dendritic and narrow axonal arbors. In projections from few neurons to many the opposite should be true: wide axonal arbors should synapse onto narrow dendritic arbors."

---> What this means: the large presynaptic components and small postsynaptic components of fly brains indicate a system where every neuron projects to a lot of target neurons.

This reminds me of the Raemaekers paper on transcription factor networks: a network worked if the majority where activating in a default inactive state of the genes, or viceversa, but not in the middle ground.

The relative sizes of pre/post synaptic partners in the compartments will tell me, according to the above, the nature of the connections (many to few or few to many). Since the sensorial input is known, then perhaps a pattern emerges on how is sensorial input integrated (received by interneurons) in the brain.


Zlatic M, Landgraf M, and Bate M. 2003. Genetic specification of axonal arbors: atonal regulates robo3 to position terminal branches in the Drosophila nervous system. Neuron 37(1):41-51.

20070317

About robo, robo3, slit and atonal.

"... in one axis at least, terminating axons detect and are positioned by their response to a cue that is produced not by their targets but by cells on the midline. This might suggest that sensory terminals are located at appropriate positions within the neuropil by factors that are quite independent of the neurons with which they will ultimately form connections."

"We cannot necessarily conclude from our experiments that sensory terminals in Drosophila are similarly [to zebrafish Mauthner cell] positioned by factors entirely independent of the target. In the periphery, motoneuron axons are guided to their proper muscles by a hierarchy of cues, starting with a transcription factor code that delivers them to particular regions of the muscle field within which they then seek out and synapse with their targets. In the case of the embryonic sensory neurons, we could also envisage a hierarchy of cues, including target-derived signals, that would contribute to the final projection pattern and certainly to synaptogenesis."

When robo is mutated, sensorial neurites project to the proper FasII bundle in the VNC neuropile, but also cross the midline and project "properly" to the contralateral, but correct, FasII bundle.

Line PO163-Gal4 labels sensorial cells that project to the ventral nerve cord neuropile FasII bundles.


Prospects and Problems of Cortical Theory - video by Jeff Hawkins

(Inventor of Palm Pilot, from talk at Berkeley)

[He mentions 'Kevan' (Martin?), whose balding head shows 13 min from the end at the first question. He, he.]

Use the morphology as the constrains of your theoretical model. Extracting all from the anatomy is almost insane.

"my goal is to get as many people as possible working on cortical theory"

"you can get more people to work on something if they sniff successful business"

"Hierarchical Temporal Memory (HTM)": technology with which the brain works, according to him

All functional regions have common architecture.

"Vision is not a sense, but a million senses together, all providing points of sensory data. Same with audition."

University and PhD: no place to park my car, no place to make a copy, no place to actually work ... hum!

HTM:

  1. Discovers causes in the world (causes == a human on which light is reflected and you visualize it; the object leading to a sensory representation).
  2. Infers causes of novel input
  3. Predicts future causes & input
  4. Creates motor behaviour

visual pattern recognition
language understanding
machine language

Each node (such as cortical columns) discover causes, passes beliefs up and passes predictions down. 'Up' and 'down' in the hierarchy of nodes.

The hierarchy makes a difference: having noise low levels makes no difference because the high levels can discard all that doesn't match.

"Belief propagation-like techniques can be used to quickly have entire system reach best overall consensus"

"Shared representations lead to generalizations and efficiency"

Modular learning: you know what eyes and hair are, when you see them in a novel animal you don't need to re-learn them.

Mechanisms of attention: afforded by the hierarchy.
Matches hierarchies in the world: structures and substructures. One can find local correlations and then larger and larger ones.

Each node in the hierarchy stores sequences of patterns:

How does each node discover causes?

He keeps mentioning a section of V1 cortex.

"All I can say is I see these things happening together often"

Insect brains could be MORE complex than mammalian brains, just because they are less flexible.

Feedforward innervation: proximal to soma "drivers": spatial coincidence detection (like moving pattern)

Feedback innervation: distal channels "modulators": temporal pooling

Thalamus ---> L4 -> L3 -> L5 ---> Specific thalamus relay: attention ... plus lots of collateral input

[HTMs] do spatial coincidence followed by temporal pooling

L4 temporal pooling:
- adjacent in space
- contiguous in time
"look at my neighbors"

L3 temporal pooling:
- non-adjacent in space
- delayed in time
"look at far away in the cortex"

L5 temporal pooling:
- non-adjacent in space
- specific time intervals (bursts)

"It's essential to have the motion to discover the causes in the world"

If it can recognize the mug, it should be able to recognize profiles in TEM data!

Numenta: HTM technology, GO USE IT [did that: urgh. Unusable.]

Inherently asynchronous system: good for distributed systems (processing in parallel)

Of course HTM can interface exotic senses (non-human)

"Just systems that discover causes and make inferences"

3D world projected into a 2D cortex! Actually the HTM works better with more dimensions than humans have.

URLs:

why companies go out of business: lack of planning, bad timing, loose focus

Can you learn without interacting with the environment? No ho crec! Although he doesn't think so.


Kurz Weil on the Technological Singularity

20070217

URL: Technological singularity (printer a.k.a. non-adds version)

"The massive parallelism of the human brain is the key to its pattern recognition abilities, which reflects the strength of human thinking. As I discussed above, mammalian neurons engage in a chaotic dance, and if the neural network has learned its lessons well, then a stable pattern will emerge reflecting the network's decision."

--> The pattern of activation in a network is its answer! I suspected it all along.

[on other civilizations]

"Perhaps they are among us, but have decided to remain invisible to us. Incidentally, I have always considered the science fiction notion of large space ships with large squishy creatures similar to us to be very unlikely. Any civilization sophisticated enough to make the trip here would have long since passed the point of merging with their technology and would not need to send such physically bulky organisms and equipment. Such civilization would not have any unmet material needs that require it to steal physical resources from us. They would be here for observation only, to gather knowledge, which is the only resource of value to such civilization."

--> What the approach of the singularity means is the approach of the closing of the loop: the creation of a non-human device from currently existing and developing parts, which will then support itself. The stage is equivalent to that of the origin of the first cell from molecular machines in the prebiotic soup of catalytic RNA and proteins. We can use this opportunity to understand how can many disassembled parts become a self-replicating autonomous unity; to extract general principles on such state transition, which would help in understanding how biological cells arouse in the first place.


Grueber et al. 2007. Projection of Drosophila multidendritic neurons in the central nervous system: links with peripheral dendrite morphology. Development 134:55-64.

20070124

Figure 2: fasII labeling of VNC. They cite Langraff 2003 as the paper establishing the VNC tracts nomenclature.

first_instar_1200 7of12 microtubule density distribution corresponds nicely to panel A in Fig. 2. Visible are: DL, DI (not labeled in panel A), DM, VMd, VMv, VI, CI1-3 and part of VL. One is not really visible: CL.

first_instar_1200 6of12 microtubule distribution corresponds to some extend as well, although the tracts visible are different. DL and DI are very visible (isolated). VMd and VMv are somewhat visible. DM is visible, but only if one focus on the core, high-microtubule count area. VI has two candidates. CL and VL are visible, although rather more dorsal.

---> Note: the physical entity of VNC tracts may have to do with the simple fact that elsewhere in the VNC, transversal fibers or local varicosities take up the space. the disposition, though, is remarkably constant across individuals, and also remarkably constant when comparing different antero-posterior sections of the VNC across individuals.


Fusi S. and Senn W. Eluding oblivion with smart stochastic selection of synaptic updates. Chaos 16:026112.

20061209

Awesome paper.

"In this simple example we showed that forgetting is directly related to hitting the upper or the lower [potentiation] bound of the synapse."

"In the slow learning scenario [that optimizes the amount of memories not forgotten] the whole learning process develops in the neighborhood of the equilibrium distribution [that is, all synapses show an activation pattern that is not dependent on any particular learning event]. In order to be able to read the perturbations provoked by each stimulus and to retrieve the information about its activity pattern, it is important to make the neuronal dynamics sensitive to small fluctuations. For instance, if two levels of activity should be discriminated depending on the pattern to be recalled, the threshold separating the two postsynaptic current distributions should be finely tuned. [...] To make the neuron sensitive to the fluctuations of the post-synaptic current around the equilibrium distribution, we subtract a negative quantity [inhibitory synapses as information retrieval-enabling systems!] [...]"

"This subtraction has a twofold value: on the one hand it allows the neuron to read small perturbations and allows [...] which linearize the learning process. On the other hand it permits to change only a small fraction of synapses to discriminate between patterns which should produce different responses."

Holy shit: "If the activity of the inhibitory cells is correlated with the activity of the output neurons, presentation by presentation, then g^*_I [inhibitory] will be correlated with G^* [activatory]. In particular, if the excitatory connections to both the excitatory output neurons and the inhibitory neurons are updated with a Hebbian rule [wikipedia: the increase in synaptic efficacy arises from the presynaptic cell's repeated and persistent stimulation of the postsynaptic cell], then g^*_I will tend to cancel the signal produced by G^*. To preserve the memory, the signal of g*_I should be smaller than the one produced by G^*."

---> What the above is saying is that the modulations of synaptic plasticity or potentiation state (LTP or LTI) is *scaled down* using inhibitions, so that more memories can be saved overlapping it without reaching the upper boundary of the synapse (that would lead to oblivion of old learned memories).

---> What I can extract for Drosophila brain: a lot!


D.B. Chklovskii. 2004. Synaptic Connectivity and Neuronal Morphology. Two sides of the same coin. Neuron 43(5):609-17.

"One hundred years later [from Cajal], the role of axons and dendrites in communication is not in doubt, yet a quantitative theory of neuronal shape is still missing. Such a theory would establish how much of neuronal shape can be explained by communication requirements and how much it reflects signal processing, or computation, requirements. In particular, dendrites may perform nonlinear operations, axons may serve as delay lines or frequency-dependent filters, and spines may be filtering compartments."

"[...] brain design is viewed as a solution to an optimization problem, where the wiring cost is minimized for a given network functionality."

"Wiring optimization has been invoked to explain many aspects of brain design: why the brain is located in the head, why neocortex folds in a characteristic species-specific pattern, why gray and white matter segregate in the cerebral cortex, why there are separate visual cortical areas, why the number of areas and neuron density scale with brain size, why cortical areas in mammals and ganglia in C. elegans are arranged as they are, why topographic maps exist, why ocular dominance patterns and orientation preference maps are present in the visual cortex, why axonal dendritic arbors have particular dimensions and branching angles, and why axons and dendrites occupy a certain fraction of the gray matter."

"Axons and dendrites take up valuable space, introduce delays and attenuation, require material and metabolic energy, and rely on genetic information for guidance and development. [...] the wiring cost can be approximated by the wiring volume. [...] pattern formation in the cortex, [...] angles of dendritic branching, differential axon diameters in rod and cone pathways, and equipartition of volume between axons and dendrites in the cortical neuropil are best explained by the minimization of the wiring volume. [...] the assumption that evolution minimized the wiring cost, while maximizing the network functionality, leads to [...] for a fixed functionality of the network, as specified by the synaptic connectivity, find the wiring design that minimizes the wiring volume."

"[...] neuronal network [...] all to all [...] the network volume depends on the wiring design. [...] the simplest possible wiring design, non-branching [...] axons [...] occupy a prohibitively large volume. Adding features of neuronal morphology, such as branching axons, branching dendrites, and dendritic spines, derives the size of the network [...]. Therefore, if one assumes that cortical function requires high (potential) interconnectivity in a small volume, one needs not look further to find a reason for the existence and total length of axons and spiny dendrites."

1 mm^3 of mouse neocortex contains 10000 neurons, with an average intracortical axonal diameter of 0.3 microns (very small!)

"The axonal length per neuron is given approximately by the number of neurons, N, times the typical interneuron distance."

--> How can I port this to the "all-to-all" of the fly neuropil?

Dendrites are a means to shorten axons! To reduce wiring volume! But also, to process multiple incoming signals.

"Because a single dendrite takes up less volume than the many converging axons, this solution is more efficient."

"... I show that adding dendrites to the all-to-all connected network, which possesses both high convergence and divergence, also improves the wiring efficiency."

"[...] adding dendritic spines to the wiring design III [...] reduces [...] the network volume."

"[...] an effect similar to adding dendritic spine might be achieved by positioning synaptic boutons on short axonal branches, i.e., terminaux boutons."

---> The latter is similar to the varicosities observed in Drosophila first instar larva neuropil.

"Why could not spined be much longer? They could be, but then their volume should be counted toward wiring cost, just as dendritic branches were. So far the spine volume has been excluded from the wiring cost because it depends weakly on the spine length. Indeed, the spine volume is dominated by its head, which does not scale with the spine length."

"If evolution attempts to minimize the wiring volume, why not make axons and dendrites thinner? The answer is that thinner wires impair brain functionality by adding to signal delay in axons and to attenuation in dendrites and by reducing information transmission capacity in synapses."

"Then, the trade-off between signal delay, attenuation, and information rate on the one hand and wiring volume on the other determines the wire diameter. This argument explains the observed difference in axonal diameters between different pathways across branch points and explains the fraction of neuropil taken up by wiring."

Network volume considering different wire diameters for axons (d_a) and dendrites (d_d):

R^3  ~  N^2 * (d^2_a * d^2_d) / s         [14]

Where:

"In addition, Equation 14 shows that axons and dendrites occupy approximately equal volume, a result consistent with anatomical data."

---> Which makes sense considering that all information from axons has to be conveyed effectively and without loss into dendrites, and thus an approximate volume of wires is necessary (both to reach the axons in the first place, and to transmit the information without attenuation).

"[...] comparison of the all-to-all network with the cortical column may seem artificial because connectivity in the cortical column is sparse. This is not a problem, however, if the brain functionality is specified by the potential synaptic connectivity."

"Potential synapse means a location in the neuropil where an axon and a dendrite come within a spine length of each other. The potential synapse is a necessary although not sufficient condition for the actual one. Its significance derives from the observation of the structural plasticity in adult neocortex: longitudinal in vivo imaging shows that dendritic spines constantly extend and retract, forming and eliminating actual synapses. At the same time, axonal and dendritic branches do not change, meaning that the potential synapses remain stable. Therefore, it may be more appropriate to characterize the cortical column by its potential connectivity."

---> YES, exactly, the potential connectivity is what I should measure in the fly brain neuropil. I'm on my way.

"The ration between the numbers of actual and potential synapses is called the filling fraction. Its value is typically much smaller than one, as would be expected from the sparseness of local cortical connectivity."

---> By design, the first instar larval neuropil has potential all-to-all connectivity, since all lineages are projecting towards a common center.

---> reminds me of that paper where synapses and physical contacts where described in 3D in the potentially all-to-all connectivity at the local level in the neocortex.

"If the cortical column includes ~ 10^5 neurons, then only the final design [with everything] fits into the volume allowed for the cortical column, meaning that all the morphological features are necessary to implement observed high interconnectivity. It is possible that the number of neurons in the cortical column is smaller. Then, depending on how many neurons there are, some of the features may not be needed. Similar considerations may explain why smaller neuronal networks can be implemented without some morphological features, such as dendrites or spines."

---> We may be seeing one such variation in insects, although I recall smaller Diptera had smaller cells per brain, not smaller number of cells (meaning, the complexity of the connectivity was in the same order).

"According to existing data, there is little variation in the average diameter of local axons and dendrites, spine length, or filling fraction between mammalian species. Then Equation 15 predicts that the density of synapses does not vary significantly either. This prediction is consistent with available data."

---> As an overall derivation: if Diptera's nervous system relies on a bigger axon fraction relative to dendrites, no problem, as long as there are means to enhance the potential synaptic contacts without a penalty in excessive wiring. The presence of multiple postsynaptic terminals plugged to the same presynaptic terminal, and the latter existing as thick varicosities, could perfectly make up for the above requirements to minimize wiring volume while maximizing potential connectivity.

"[...] the present work strongly suggests that neuronal morphology is largely a reflection of synaptic connectivity."

"Because convergence and divergence must be the same (averaged over all cortical neurons), the volume of the axonal arbor approximately equals that of the dendritic arbor. There observations coming from various brain structures, when taken together, point to a general relationship between morphology and connectivity."

---> Convergence means that axons from many neurons reach the same target neuron, and divergence that dendrites from a single neuron reach axons from many neurons.


D.B. Chklovskii, T. Schikorski and C.F. Stevens. Wiring Optimization in Cortical Circuits. 2002. 2002 Neuron 34(3):341-347

"[...] so the maximum possible number of synapses (give the actual brain architecture_ occurs when the wire fraction is 3/5 and, in this sense, the cortical circuits are optimal when the wire fraction is actually 3/5; in order for the number of synapses to exceed this maximum, some other characteristic of the brain's structure or function must be altered, such as the conduction delay from one location to another or the size of synapses."

"Our observation of an actual wire fraction close to the predicted 3/5 argues that conduction delay and cable attenuation are close to their minimal values, and that the 'layout' parameter and number of synapses (given the actual cortical architecture and the properties of other elements) are close to the maximum. Therefore, we suggest that these parameters play a key role in determining cortical architecture." (but they note, there might me other parameters for which optimization has happened)

"Our observations have an interesting consequence for notions about memory storage in the brain. Long-term memory is often considered to involve the formation of new synapses [...] but, according to our analysis, an increased number of synapses could not be accommodated without degrading performance in some way because the cortex is already optimally wired in the sense that the number of synapses is already maximal. To fit in additional synapses without compromising performance, some synapses would have also to be eliminated. If memories were stored by increasing synapse size, compensatory decreases in synapse size would have to accompany to maintain optimal wiring."

"The idea that neural circuit design is under pressure to minimize signal delay and attenuation dates back to Cajal. Our observations suggest that the layout cannot be less efficient than in the real brain without compromising brain function, thus supporting the importance of wire lengths minimization in brain organization. Such wire length minimization arguments have recently been used to explain why cortical regions are separated, why ocular dominance and orientation preference patterns are present in primary visual cortex, why white and gray matter is partitioned as it is, why axonal and dendritic arbors have particular size and branching angles, and why cortical areas and ganglia in C. elegans are arranged as they are." (with many references inserted)


Kayoko Ishii. 2006. Cognitive robots to understand human beings. Life Science Unit. Quarterly Review no. 20.

"The enlargement of human brains obeys, according to [14], to the acquisition of social behaviour."

--> This matches what I've read on mushroom bodies in hymenoptera! Does it spring from the same source?

[14] R.I.M. Dunbar, "Origin of language," Kagaku, Vol. 67, pp. 289-196, 1997. (Japanese)


Heisenberg, M. 1998. What do the mushroom bodies do for the insect brain? An introduction. Learn. Mem. 5(1)1-10.

20060304

Nice overview. The entire issue is devoted to insect MB.

Quite pitiful: recordings have been done on single neurons of large insects. Not much data has been extracted from such experiments, other than the activity profile of a single neuron here and there which is, in my opinion, wishfully presented as an example of MB activity.

MB receive both optical and chemosensory input.

Optical input reaches the calyx.

Kenyon cells are two synapses away from the olfactory receptors on the antenna.

Function of the MB in temporal integration of sensory input: the length of its parallel fibers is important. Thus in small brains the MB lobes extend to the limits of the neuropil, not occurring in larger brains (the logic is, the lobes are long enough).

I postulate that large cells that contact broad areas of the brain (and thus hundreds of cells) have a neuromodulatory function. No other interpretation is consistent with complex systems theory, where the number of connections is key to network dynamics, and in particular only a low number of connections enables ordered dynamics.

Speculation: elements responsible for the control of neuromodulatory elements may be implementers of emotions (the drum beat players in the roamer benches of a Roman warship). I've read already several times (readership and researcher's easy topic to work on bias warning!) that olfactory-related neurons have a large number of targets, suggesting that olfactory senses affect emotions directly (think about courtship and sex!).

Everything is expressed in Kenyon cells. Suspect background noise enhanced by anatomical particularities (i.e. packed small long parallel fibers).

Researchers are naive in studying only structures whose anatomy makes them stand out on the background of other tissues. The MB stands out because of its three long thick bundles of orthogonal fibers. The rest of the neuropil is not unstructured, rather, the structure is not recognizable as simple geometric figures. No matter how geometrically complex, the neuropil gross structure looks simple and homogeneous at the macro scale, and is thus labeled uninteresting. I suspect that the lack of macro structures provide the best excuse not to dive into the barely approachable extreme complexity of the neuropil.

It amazes me to note that research has been conducted on adult and pupae, when nothing is known on earlier stages of development that would explain how the brain structure came to be!


K. Yasuyama et al. 2003. Synaptic organization of the mushroom body calyx in Drosophila melanogaster. JCN, 445(3):211-26.

20060304

GABAergic cells can be labeled for TEM with a UAS(membrane-bound)HRP, and then color developing the HRP

There are synapses, even if in small numbers, in axon bundles.

Nice reference pictures of Drosophila brain TEM.

Many references to "boutons" when the axonic/dendritic profile is large. Perhaps these are varicosities.


K. Yasuyama et al. Synaptic connections of cholinergic antennal lobe relay neurons innervating the lateral horn neuropile in the brain of D. melanogaster. 2002. JCN, 466:299-315.

20060304

Technique: ChAT immunolabeling in 1-micron or ultrathin plastic sections of Vibratome slices labeled by the ABC-DAB procedure.

The "large diameter axons" within the iACT in Figure 4C may just be varicosities.


Libersat, F. 2004. Maturation of dendritic architecture: lessons from insect identified neurons.

20060304

"Some insect neurons are individually identifiable, n that they show a characteristic dendritic architecture, and this sets the ground for quantitative analysis of dendritic architecture at different developmental stages because a given identified neuron can be sampled in a large number of animals [...]"

---> Nice reason to work with insects.

"The constant three-dimensional architecture of identified neurons is used as a template for examining the effect of various experimental manipulations of their geometry [..]"

---> The author is assessing the constancy in terms of "fuzzy signal in the same volumetric areas. There is no quantification of this constancy anywhere in the literature.

Mutations in dendritic architecture, in Drosophila:

Questions the paper attempts to answer:

  1. "To what extent is dendritic architecture regulated by intrinsic factors?"
  2. "Regarding extrinsic factors, what is the relative contribution of activity-dependent versus activity-independent mechanisms to specific features of dendritic architecture?"

"In any event, metric measures commonly used for dendrites include the total dendritic length, total dendritic surface area, dendritic segment length and diameter, and number of branches."

"Topological measurements, such as Sholl analysis (Sholl, 1953) and branch order analysis (see Mizrahi and Libersat, 2002), are implemented to evaluate the pattern of branching. Sholl analysis consists of counting the number of occurrences of branch points in the dendritic tree falling between concentric spheres that are separated by a fixed number of microns. Branch order analysis consists of plotting the frequency distribution of the number of dendritic branches as a function of branch order. Branch order analysis is useful for evaluating growth and pruning of trees."

"[...] Sholl analysis has been criticized for not discriminating between topological and metric aspects of tree structure [...]"

---> Hence the need of the branch order analysis for the study of the tree's topology.

"Among the best studied examples of identified neurons are the large interneurons of insects, such as the abdominal giant interneurons (GIs) of crickets and cockroaches (Mendenhall and Murphey, 1974; Mizrahi et al., 2000), the giant interneurons of the vertical system (VS) in the Drosophila lobula plate an optic lobe (Scott et al., 2002), and motoneurons in the holometabolous insects, Manduca sexta (Weeks and Levine, 1990; Consoulas et al., 2000; Duch and Levine, 2000, 2002), and Drosophila melanogaster (Consoulas et al., 2002)."

On branching variability:

"[...] when examining pairs of homologous neurons, a given identified neuron is more similar to its contralateral mirror image in the same animal than it is to its homologue from different animals, as show for the visual neurons of grasshoppers (Goodman, 1978)."

---> There is a "Hausdorff" method to perform quantitative measurements of similarity between two branching trees.

"Although the locations of the primary dendrites occur with little variability, the fine pattern of higher order distal dendritic branching is variable [...]"

"[...] [in] thoracic spiking local interneurons in the locust [..] dendritic growth can be divided into at least two phases: a first phase of dendritic branching during which the basic skeleton of major neurites is formed, and a second phase during which the basic skeleton is elaborated by the addition and retraction of many side branches (Shepherd and Laurent, 1992)."

---> When comparing the branching extension of a neuron, the author measures the area! What a bad measure!

"[...] cercal sensory axons enter the CNS at the 45% stage of cockroach embryogenesis and start branching at 50%, whereas the giant interneurons send out a primary dendrite followed by the formation of low order branches at 45%. However, pre- and postsynaptic processes start interacting via filopodial contacts only at 55% of development [...]. Thus, during the first half of embryogenesis, lower branch order is established for both pre- and postsynaptic elements of this circuit in the absence of any physical contacts."

The number of synapses can increase even when the number of contacts doesn't:

"Synapse formation between the main sensory axons and GI dendrites is initiated at 55% to 65% of [cockroach] embryogenesis and, while the number of contacts does not significantly increase after about 70%, the number of synapses doubles between 65% and 75% of embryogenesis."

"[...] the formation of distal terminal branches in insect interneurons and motoneurons requires incoming afferent input."

"After targeting the TeTxLC Drosophila central nervous system with a pan-neuronal GAL4 driver, which abolishes synaptic activity, motoneurons involved in the peristaltic locomotory behavior of Drosophila develop with an apparently normal morphology and retain heir capacity to form synapses [...]"

---> So the formation of the high order branches is perhaps way more dependant on the presynaptic higher order branches than the activity itself transmitted through such synapses.

"Furthermore, vertical system (VS) neurons in the Drosophila lobula plate of the optic lobe develop in the absence of synaptic activity [...]"

"[...] the development of the sophisticated receptive field organization of VS neurons does not depend on sensory experience [...]"

---> Which suggests to me the first instar larval brain could be developing completely switched off.

"Taken together, it seems likely that the dendritic morphogenesis of VS neurons (and presumably other insect neurons) is largely determined by internal genetic programs, combined, perhaps, with the local environment and diffusible cues released by afferent neurons. One possible explanation [...] is that dendrite maturation does not depend on synaptic activity, but maintenance of the dendritic tree does. [...] it remains unclear whether normal synaptic activity might be crucially important for the maturation of dendritic shape during later stages. A reasonable explanation might be that early phases of dendritic growth follow an activity-independent program, but that synaptic activity becomes increasingly important for dendritic maturation during refinement of the circuitry."

"[...] postsynaptic elements play a role in the maintenance of the synaptic circuitry by means of mechanisms that do no involve modifications of neuronal architecture." (suggested from single neuron ablation experiments, where the synaptic strength was altered)

"[...] the presence of a postsynaptic target [with retrograde signaling] is required for synaptic reorganization."

In the conclusions the author overemphasizes the role of afferent inputs in morphing the higher-order dendritic branches. There is no support for such a conclusion in the data presented in the paper .

UAS-Shibire: its dynamin, there's a dominant negative and a temperature sensitive mutants. Gets in the way of releasing vesicles, stops synaptic transmission.


Alon Kauffman et al. 2005. Multi-perturbation experiments.

20060225

This paper will be looked upon, in the future, as the formal declaration of the inflexion point in the default research focus in biology, from a single gene to a network of genes. It's about time the concept of "industrial gene" is disregarded; such concept arose only from the very first research results that suggested a "one gene, one function" relationship, and which the bio industry and university and high school lecturers alike adopted (because of its inviting simplicity?).

"However, the vast majority of these studies have perturbed only one element at a time, often resulting in little phenotypic effect. Hence, in complex biological systems, multiple concomitant perturbations should be employed to reveal the contributions of the different elements to the system's functioning."

---> We have to free ourselves from the idea that one gene is involved in one process, and thus dive into the idea that when a gene is malfunctioning or absent, others may take over its role in each of the different functional modules in which the malfunctioning gene participates.

[this can be linked nicely with the Libersat 2004 review on dendrite architecture, where the ablation of one neuron results in compensated, strengthened synapses with remaining neurons]

"The goal of MPA is to define and calculate the contribution (importance) of system elements to a certain function, from a dataset of a series of multi-perturbation experiments."

"The dataset included 21 multi-knockout experiments (see Table S1). Prediction of the full multi-knockout set (i.e., 32 multi-perturbations) was obtained using projection pursuit regression, and explains 79.6% of the data variance via leave-one-out cross validation."

---> From the set of MPA experiments, the function and importance of each element is deduced. We could randomly inactivate and label (UASmcd8GFP and UAStetrotoxin background over which flip-out clones are generated) and then evaluate a particular behavior of the fly or larva, and thus deduce the paper that each neuroblast lineage plays in that behavior. The scale problem (100x100 lineages is 10^4 pair cases, and 10^6 triplet cases), is nicely taken care of by mathematical inference methods described in this paper.

---> A nice experiment to investigate the role of different brain areas could be performed in Drosophila under the MPA logic, and also under a Bayesian logic.

Extracting information from a multiperturbation experiment:

"The multi-knockout data can be utilized to construct a unique weighted multilinear performance prediction function F, which, given any configuration of knocked-out and intact genes, can accurately predict the PRR performance level. However, this function contains 32 (or, more generally, 2^n) terms corresponding to all possible knockout configurations and hence is unintelligible and uninformative to the biologist. To extract the relevant information in the data and make it explicit, F is further processed via the FIN (Functional inference network) algorithm to construct a compact and yet fairly accurate functional prediction function, F'. Each of the terms in F' can be viewed as a serial functional pathway, whose contribution depends on the intactness of all its component genes. (Obviously, membership in a functional pathway does not necessarily imply that there are direct physical interactions between the elements of the pathway.)"


Sanchez-Soriano 2005 "Are dendrites in Drosophila homologous to vertebrate dendrites?" Dev. Biol. 288:126-38.

20060205 Zurich

"Neurons in vertebrates and arthropods are functionally similar and are believed to have common evolutionary roots. However, nerve cords of vertebrates and arthropods are very differently organized."

---> Really that different? I don't think so. Both structures are constrained by similar spatial and functional requirements.

"Here, [...] motorneurons in the Drosophila nerve cord, and characterize the compartmentalisation of these cells. Various features have been described for these motorneurons: they send usually one primary neurite into the neuropile from where they enter the dorsal motor nerves. All motorneurons form side branches in the dorsal neuropile (Landgraf et al 2003b). These side branches are commonly referred to as dendrites. They display tree-like shapes which could be classified as partial spherical radiation type when using nomenclature for vertebrate dendrites (Fiala and Harris, 1999). They are arranged into somatotopic maps which are roughly comparable to motor columns in the ventral horn of the vertebrate spinal cord (Landgraf et al., 2003a; Tsuchida et al., 1994). Thus, dendrites of Drosophila motorneurons and of vertebrates show similarities at the gross morphological level."


Prokop A and Meinertzhagen IA. 2006. Development and structure of synaptic contacts in Drosophila. Sem. Cell Dev. Biol. 17:20-30.

20060205 Zurich

[idea: the organization of the nerve cord of vertebrates and that of higher invertebrates is similar, even though the common ancestor didn't have any identifiable and/or similarly complex ventral nerve cord. Its similarities thus stem from the sharing of the elements that compose it, i.e. neurons and their neuritic arborizations, glial cells and vessels, plus the need of a similar function, that demands capacity of computation paired with speed of computation. These constraints are in the line of the paper on the "Segregation of the brain into gray and white matter: a design minimizing conduction delays", by Wen Q and Chklovskii DB. BUT, if the common ancestor truly didn't have a ventral nerve cord organization, how come other parts of the nervous system are so similar, such as the corpora cardiaca / glandula pineal? The organization of the ventral nerve cord must predate the segregation of the deuterostome and protostome trees. ACTUALLY, this sort of organization can be seen in planarians and macrostomum as well, indicating a very old invention date.]

The unsheathing of the neuropil by glial cells is thought to provide a more favorable environment in terms of ions than the haemolimph.

[Experiment: generate clones on a background of both GFP and electron-dense compound, so that 1) clones can be identified under the scope, and scanned with the confocal while alive, and 2) immediately fixed for electron microscopy, where one can go and see the fine structure of the branches. The EM part can also be achieved by photoconverting solvent into a precipitate wherever there is GFP emitting green light.] -- if you, reader, do it, I'd love to hear about.

"NMJs and giant fiber terminals form monadic synapses, in which one presynaptic site displays a continuous junction with just a single postsynaptic site. In contrast, most central synapses have two, or sometimes three, even four or more postsynaptic elements opposite a single presynaptic release site, in constellations commonly referred to as diads, triads, tetrads, etc."

--> What sort of advantage or functionality does such a network provide? In vertebrate brains monadic is the rule, except for the retina. [perhaps synchronization?]

""In contrast to NMJs, synapse assembly at photoreceptor contacts occurs mainly at one stage of development, i.e. during pupal life, although certain plastic properties persist into adult life."

--> Sounds reasonable that the eyes are not connected until they are needed! But then, what is the pupal brain processing, if anything?

"... developmental studies have taught us that each synapse seems to assemble on-site in a progressive manner."

"... we do not yet know whether the structural and functional proteins of fly synapses assemble on-site, or whether they are already pre-assembled and delivered via specialised vesicles, as is the case in vertebrate release sites."

According to de Ruyter and Laughlin in Nature 1996, the amount of information encoded and transmitted per synapse is higher in non-spiking synapses, and are most used in sensorial systems.


Quan Wen, Dmitri B. Chklovskii. 2005. Segregation of the Brain into Gray and White Matter: A Design Minimizing Conduction Delays. PLoS Comput Biol. 1(7):e78.

"Local cortical circuits may be viewed as a network of n neurons with all-to-all potential synaptic connectivity, meaning that the axons and dendrites of most neurons come close enough to form a synapse [23–25]."

"As a result, the optimal volume of the network is of the same order as the non-wire volume. Assuming that non-wire consists mostly of synaptic components, such as axonal boutons and spine heads, the optimal network volume is of the same order as the total synaptic volume. Therefore, the local network volume is given by:

     l^3 ~ n^2 * vs

where vs is the average synapse volume and n is the total number of neurons in the local network. (In a network with all-to-all connectivity, n is also the number of local connections made by a neuron via potential synapses.) For the sake of clarity, we ignore the fact that only a fraction (0.1– 0.3) of potential synapses are converted into actual synapses [23]."

"... the maximum number of neurons in the all- to-all connected network is on the order of 10^4. This corresponds to roughly the size of a cortical column, which is then the largest network that can combine all-to-all potential synaptic connectivity with tolerable conduction delay."

"Anatomical evidence suggests that the brain maintains short conduction delays by implementing sparse global interconnectivity while preserving high local interconnectivity [31]. Such design resembles the small-world network [38], as has been noticed by several authors [39–42]."


Hamasaka Y, Nassel DR. 2006. Mapping of serotonin, dopamine, and histamine in relation to different clock neurons in the brain of Drosophila. J Comp Neurol 494(2):314-30.

Clock neurons in the brain are related to light input from the Bolwig's organ. Since all motor output occurs in the VNC, I suspect clock neurons sitting in the brain may connect directly to the VNC.

A line to highlight synaptic terminals: tim-GAL4 x n-synaptobrevin-GFP (nsyb-egfp).

Bloomington number 9263, w[*]; P{GawB}D42, P{UAS-n-syb-GFP.E}3/TM3, Sb[1] - 3rd chromosome.

Volker said: sineoculis neurons are known to exist in the tritocerebrum, at the interface between the brain and the suboesophageal ganglion. Such cells could be responsible for the long sineoculis-GFP projections into the ventral nerve cord. Further, the sineoculis group that seems to project to the ventral nerve cord is not any of the cells that overlap with timeless (tim) or pigment dispersing (pdf) expressing cells.

On the Bolwig organ: its neurons project to a dozen of cells in the lateral side of the brain, which constitute the access medulla and in turn project to the CP2, a neuropil compartment that develops into the lateral horn of the adult, and which is known for receiving input from olfactory and visual input.

Helfrich-Förster have papers on the above.


Trimble WS et al. 1991. Cellular and molecular biology of the presynaptic nerve terminal. Annu. Rev. Neurosci. 14:93-122.

Dense-cored vesicles are assembled in the neuron's soma and transported through neurites. Vesicles mature during transport, including proteolysis (low pH? Yes, protons are pumped in, and the proton differential is used to translocate acethylcoline inside the vesicle from the cytoplasm).

The membranes of dense-cored vesicles contain many of the components of small, clear vesicles (meaning small vesicles may be derivatives, by means of vesicle membrane recycling, of dense-core vesicles.)

"Mitochondria and endoplasmic reticulum-like structures often found in terminals are excluded from the direct vicinity of the active zones."


Qin thesis.

20051115:

Lead me to:

Koh YH, Gramates LS, Budnik V. 2000. Drosophila larval neuromuscular junction: molecular components and mechanisms underlying synaptic plasticity. Microsc Res Tech 49(1):14-25.

PSD: postsynaptic density

MAGUK Discs-Large (DLG): Drosophila gene for the MAGUK protein.


Crick F, Koch C, Kreiman G, Fried I. 2004. Consciousness and neurosurgery. Neurosurgery. 2004 Aug;55(2):273-281; discussion 281-2.

20051114:

"Within the cortex, approximately 80 % of the neurons are excitatory. Their connections are primarily within the cortex, many of them with neighboring neurons, so there is much rather specific mutual excitation."

--> I would say this depends on whether the inhibitory neurons are much larger and harbor way more synapses with excitatory neurons than the latter with themselves.

"Each neuron makes and receives some thousands of connections."

"The main job of a neuron in any cortical area is to respond to the significant correlations in its inputs and, gradually, to learn them, so that it can build the appropriate feature detectors (such as those for faces). It probably does this primarily by modulated Hebbian (or Pseudo-Hebbian) learning mechanisms (19). Early cortical areas discover simple, local correlations, whereas later areas look for correlations of these correlations, and so on, in a hierarchical manner."

--> What is this Hebbian learning mechanism? Oh, has to do with belief propagation networks.

I like very much the straightforward style in writing. Without being vulgar, it is very clear and suggestive, powerful!


Kalisman N, Silberberg G, Markram H. 2005. The neocortical microcircuit as a tabula rasa. Proc Natl Acad Sci U S A. 2005 Jan 18;102(3):880-5

20051114

Interesting method for labeling neurons:

Neurons can be labeled in vivo by microinjecting biocytin, and then fixing and staining with fluorescent avidin.

So, they do the recordings by patch-clamp with a pipette that contains biocytin in the electrolyte solution, and later on stain with fluorescent avidin to see which neurons they recorded on.

They created their own software for 3D reconstruction.

Two different types of axodendritic connections:

  1. Geometrical connectivity: close axodendritic appositions and touch sites, regardless of the existence of a functional synapse.
  2. Functional connectivity: when a button is observed at the interface of axon and dendrite.

---> I would add: 3) Potential connectivity: by physical proximity of branches. In a way, a milder case of geometrical connectivity.

"As revealed by paired recordings, only a small fraction of the geometrical contacts is used to form functional connections. How, then, are geometrical contacts important? We propose that the importance of all-to-all geometrical connectivity is primarily to enable the dynamic rewiring of the cortical microcircuit without the need for major axonal and dendritic remodeling."

"Any learning process that might require two specific neurons in the column to be functionally connected could therefore occur into a synaptic contact between a bouton and spine."

"[Touches Versus Synapses] (geometrical contacts, as opposed to functional contacts, are axodendritic connections that do not consist of a synapse) [...]"

"As revealed by paired recordings, only a small fraction of the geometrical contacts is used to form functional connections. How, then, are geometrical contacts important? We propose that the importance of all-to-all geometrical connectivity is primarily to enable the dynamic rewiring of the cortical microcircuit without the need for major axonal and dendritic remodeling. All-to-all geometrical connectivity therefore allows for any two pyramidal neurons in the cortical column to form a reliable synaptic connection without arbor remodeling by simply using the existing fiber infrastructure. Any learning process that might require two specific neurons in the column to be functionally connected could therefore occur by the transformation of existing geometrical touches into a synaptic contact between a bouton and spine. Such transformations are probably faster (8, 9) and metabolically cheaper than any process requiring axonal or dendritic growth. Secondly, although the mechanisms triggering and enhancing formation and modification of synapses are only partially understood, it is likely that some form of "molecular handshaking" between the neurons, reflecting their relative activity, is required (46-49). Because most pairs of pyramidal neurons in the column are geometrically connected, they can relay molecular signaling between them in a truly parallel manner. The fact that multiple contacts are observed between synaptically connected neuronal pairs may be explained by the hypothesis that molecular interactions at axodendritic touch sites are driven by the electrical activity of the pre- and postsynaptic neurons. If the molecular signaling is considered to be a computational process ending in a decision to create, modify, or remove a functional synapse, it significantly enhances parallel information processing in the neocortex. One may also speculate that an intermediate state of such transformations is the silent synapse (30-32). Finally, the fact that only a small fraction of potential touch sites are used to form functional connections endows the microcircuit with a vast number of possible connectivity configurations (50, 51). The "filling fraction" (51), which denotes the ratio between the actual and potential connectivity, is used to assess the information content of the cortical tissue in terms of information bits per synapse, as a function of the possible network configurations. According to this formalism, the low filling fraction (0.1) implied in our study suggests a high information content of 4-5 bits per synapse (51)."

"Any learning process that might require two specific neurons in the column to be functionally connected could therefore occur into a synaptic contact between a bouton and spine."

"Spines and filopodia could therefore be considered to be devices that increase the geometrical connectivity between neurons. This possibility is further suggested by high spine and filopodia motility in comparison to negligible axonal and dendritic arbor remodeling (54-56)."

Reminds me of the sine oculis screen spines on the primary neurons (from the mutant screen done by Wayne, Shana and Siau Min.)

Wayne Pereanu & I (wildly speculating): the distinction between axons and dendrites may be absolutely irrelevant to information flow between neurons. The only morphological difference between axons and dendrites is that axons are usually much longer, and thus an optimal configuration of its membrane and cytoskeleton demand a thin profile, an appropriate set of channels for depolarization transmission, and a forward-only arrangement of microtubules.

How does a neuron know that is not connecting to itself? How can a "molecular handshaking" mechanism aid in this? Well, turns out there are plenty of papers explaining self-avoidance in dendritic trees, both within a neuron and between neurons of the same class and/or parent progenitor cell. Has to do with membrane cadherins having a combinatorial, antibody-like genomic recombination and diversity generation mechanisms.

---> The combinatorial code of Cadherins explains it, enabling self-repulsion and also cell type repulsion.


Borst A and Haag J. 2002. Neural networks in the cockpit of the fly. J. Comp. Physiol. 188:419-37.

20051020:

"Based on the anatomical overlap between the various columnar elements, the retinal images are thought to be processed by at least three different retinotopic pathways in parallel [...]."

Only lamina neurons have been studied using electrophysiology. The rest are too small (in a fly larger than Drosophila! But there are some studies of the rest of the cells in other insects.)

"On each hemisphere, there exist about 60 different LPTCs [lobula plate tangential cells] each of which is individually identifiable based on its characteristic anatomy and response properties. In general, LPTCs are sensitive to visual motion in a directionally selective way."

---> Like cat brain cells in V1 (Nuno & Elisha, the Institute of Neuroinformatics)

"... on the observation that certain varicosities (blobs), which have often been found to be presynaptic specializations in invertebrates, are seen in the lobula plate branches of CH-cells, at the light microscopy level. This presumption was indeed confirmed by subsequent ultrastructural investigations (Gauck et al. 1997). The protocerebral branches were found to house exclusively postsynaptic densities, and presynaptic sites were clearly visible in branches of CH-cells in the lobula plate." (this sentence was badly and misleadingly phrased)

"However, as a further complication, the lobula plate arborization of CH-cells were also found to house presynaptic sites that makes the lobula plate arbor of CH-cells a dendrite and an axon terminal at the same time."

Motion-detection computing is done by a bi-fan network motif, in which one of the signals is delayed, "waiting" for the other, and then one signal is subtracted from the other, so that it either fires high or gets totally canceled depending upon the direction.

Interesting references:

Sodium channels can be blocked with TTX and QX314.

"... neural nicotinic receptors in insects [...] found to also be permeable to calcium, in addition to Na and K-ions (Goldberg et al. 1999; Oertner et al. 1999, 2001). Together with the voltage-activated calcium current described above, nicotinic receptors constitute a second gate for calcium entry in these [LPTC] neurons."


Patrick Forterre. 2005. The two ages of the RNA world, and the transition to the DNA world: a story of viruses and cells. Biochimie 87.

20051019

This paper is the first convincing and consistent explanation of the origin of bacterial, archeal and eucaryotic cells I've ever read.

Typographic errors:

"Many orphan proteins [i.e. without known homologues] found in completely sequenced genomes thus could be of viral origin."

"Briefly, transfer of DNA from a simple virus could have led to a prokaryotic cell, whereas transfer of DNA from a complex virus could have led to a eukaryotic cell, if the virus used to recruit intracellular membranes of the RNA-cell for the formation of its envelope."


U. Homberg and J.G. Hildebrand. 1994. Postembryonic development of the gamma-amminobutyric acid-like immunoreactivity in the brain of the sphinx moth Manduca sexta. J. Comp. Neurol. 339:132-149.

20051014

Summary: The authors describe the pattern of GABA neurons in the brain throughout larval and pupal development.

Flaw: fig. 1c describes "one major fiber" which most likely corresponds to a bundle of fibers. Low resolution study! Or, they consider "fiber" to mean a nerve, a bundled group of neurites.

Ecdisone-related hormones are responsible for "changing" the GABA expression of a large group of neurons. The authors speculate this has to do with rewiring/reprogramming large chunks of the brain in the impasse from larva to adult.

Apparently primary neurons have way many more GABAergic representatives than secondary neurons during larval development.

"The tritocerebrum receives the primary neurites of approximately 50 neurons with somata lateral to the neuropil." [referring to GABAergic neurons]

"During metamorphosis, two phases in the development of the GLI [GABA-like immunoreactivity] can be distinguished. The first phase [...] is characterized by the disappearance of GLI in most brain areas and new expression of GLI in the optic lobes, in the lower division of the central body, and in some neurosecretory cells. During the second phase [...] the adult pattern of GLI in areas other than the optic lobe and central body emerges [...]."

"GLI [...] decline peaks at P3-P5 [...]."

"[in the LAC=larval antennal center] During [late] larval development, immunoreactive soma move to the periphery of clusters of [secondary neurons]."

There are bilateral GABAergic connections (between hemispheres).

Flaw: the groups of neurons defined in this study are useless because they can't be identified in space; other than a general idea of primary and secondary neurons, there is no positional map to refer to.

page 138 is misplaced after 139.

The mushroom body contains way more GLI cells in the adult than in the larva (I need to learn about what the heck is the mushroom body, because a "center of olfactory sensory input" is just too generic, doesn't tell me much!).

---> it's also a memory and high-integration center.

[page 142 bottom left] "In contrast to those in the midbrain, immunoreactive neurons in the optic lobe are probably generated from neuroblasts that were not previously active in the embryo."

---> how reliable is this information? No clue.

Where do GABAergic neurons come from? Perhaps a few from each lineage?

"The axonal projections of certain larval GABA-immunoreactive neurons seem to change during metamorphosis."

---> Corresponds nicely with Libersat's 2004 review on dendritic architecture.

"Hormone titles rise steadily until about stage P9 and then drop abruptly. This large and long increase might trigger transient synthesis of GABA even in neurons not committed to the GABA phenotype."

"Neural circuits in the ventrolateral protocerebrum might therefore continue to operate and remain relatively unchanged during the synaptic reorganization of the brain, and [...] the ventrolateral protocerebrum communicates via several intensely GABA-immunoreactive fibers with the ventral nerve cord."


Vladislav Volman, Itay Baruchi and Eshel Ben-Jacob. Manifestation of function-follow-form in cultured neuronal networks. Phys. Biol. 2 (2005) 98-110.

20051012

"The cultured networks spontaneously form from a dissociated mixture of cortical neurons and glia cells drawn from 1-day old Charles River rats. The cells are spread homogeneously over a lithographically specified area of Poly-D-Lysine for attachment to the recording electrodes. Subsequently, the neurons send dendrites and axons to form a wired network. Although this process is self-executed with no externally provided guiding stimulation or chemical cues, a relatively intense dynamical activity is spontaneously generated within several days. As we show in the next section, the spontaneous activity is marked by the formation of synchronized bursting events (SBEs)-short time windows during which most of the recorded neurons participate in relatively rapid firing. These SBEs are separated by long (seconds to minutes) intervals of sporadic neuronal firing."

"In contrast, the activity of small and medium-sized neuronal networks does not exhibit any such organizational motifs. This observation supports the notion of unitary, or elementary, networks about 1mm3 in size, based on the following reasoning. The elementary time scale of neuronal activity is 10^3 s, while the propagation speed of action potentials along the axon is 1ms1, yielding a characteristic length scale of 1 mm. In the cortex, 1 mm3 of tissue contains about 10^5 neurons, which corresponds to a cultured network of 1 mm2 and 10^4 neurons. Hence, we define a unitary (or elementary) network as having O (10^4) neurons in an area of 1 mm2, i.e. corresponding to the medium-sized cultured networks. Higher density cultures grown in an area of 1mm2 form a web of neuronal clusters linked by axon bundles [7]. This is a hint that the cells are programmed to form networks with special characteristics."

"By the same token, when 10^6 neurons are spread over 10^2 mm2, they are free to form a homogeneous fabric of 10^2 coupled elementary networks. The measurements presented here were performed from electrode arrays of 10^2 mm2. It has been proposed that the existence of sub-groups of SBEs in these large networks reflects the co-existence of functionally distinguishable coupled elementary networks [8]."

"Put together, these observations suggest that large cultured networks are able to sustain different spatio-temporal patterns of neurons collective activity. In the next section, we investigate the possible origin of this kind of dynamics."

"These overlapping networks are then allowed to evolve according to equations (B.1)(B.3), generating synchronized bursting events, as explained earlier."

---> means: the activation level of each neuron fluctuates according to the equations defining their spike-firing potential.

"Indeed, there is evidence that neurons and glia maintain intricate dialogue, exchanging information on the molecular level [19-21]. To provide an instructive example, it has recently been shown that astrocytes synchronize neuronal activity by generating glutamate-related currents. These currents, shown in figure 9(a), are non-synaptic, as their decay time (700 ms) is much larger than that of the synaptic ones."

"These considerations lead us to speculate that real neuronal networks might utilize mechanisms other than those of synaptic transmission in order to regulate the expression of different spatio-temporal activity patterns."

[from Appendix B] "According to the theory of neuronal group selection, the size of the brains basic functional assembly varies between 50 and 10^4 cells. Motivated by this, and by the notion of unitary networks (as explained in the main text), we study the dynamics of networks composed of 20-60 cells. To follow physiological data [31], 20% of the cells are usually set to be inhibitory."

Experiment: affect the ability of glial cells in the Drosophila neuropile to secrete glutamate or similar things which may be acting in coordinating and synchronizing whole sets of neurons.

According to Volman, some authors have described that glial cells generate currents which are non-synaptic and have a long decay time (~700ms), and are related to glutamate.

Could it be that fly brain somas are generating this sort of currents as well to affect neighbour somas in the cortex?

Segregating all the somas from the neurites allows for them to be affected by glial signals, such as the observed synchronization by glutamate.

Connectivity:


N. Kashtan, S. Itzkovitz, R. Milo, and U. Alon. 2004. Topological generalizations of network motifs. Physical Review E 70, 031909.

Network motifs:

1 - FFL or Feedforward Loop

	x --> y -->
	|          z 
	---------->
	

In the case of positive regulation and AND-logic (on integrating input in each node), the FFL acts as a persistence detector: a circuit that rejects transient activation signals, responds only to persistent signals, and yet allows rapid response to inactivation signals.

With an OR-gate the FFL filters OFF pulses and responds rapidly to ON pulses.

With other sign combinations, the three-node FFL can function as a pulse generator or response accelerator.

The FFL in neurons can act as a synchronizing pipe, where asynchronous X1,X2 stimuli result in a single Z stimuli by the first Xi to reach Y increasing Y's sensitivity, as the shortly-followed, second Xj makes Y fire before its sensitivity decays too much. It has the effect also of discarding stimuli too separated in time (because of Y sensitivity decay), which could be noise and/or meaningless.

Also, this very same mechanism acts as a sensitization: for multi-input (multi-X) FFLs, sensitivity to any given input is conditioned by the previously received input.

When comparing transcription networks to neuronal networks, the alpha factor in transcription networks is the protein decay, whereas in neuronal networks is the sensitivity decay.

A 2X,Y,2Z FFL can implement an XOR functionality, responding to one stimulus only when the other is not present (in the case of 2X).

A FFL also detects a long stimulus, while ignoring short stimuli if not closely related in time (fig. 5) (coincidence detector of two independent weak stimulus).

When considering FFL in transcription networks, the X role is usually a global transcription factor, the Y role a local (more specific) transcription factor, and the Z nodes are the coregulated genes.

2 - single-input module:

Generates a temporal order of gene expression which correlates with the functional order of the genes in the pathway.

3bi-fan:

Performs hard-wired combinatorial decisions governed by the input functions of the output genes.

	A --> C & D ; B --> C & D

4- three-loop motif:

A circle of connections:

	A --> B --> C --> A

These guys are way too theoretical and are selling thin air. For example, they claim the different nodes in a network can have different roles; a role defined by the connectivity properties of the node. But in the real world such three-node graphs would have other connections on each node and different weights that restrict their swapping and thus provide every node with a unique role. IOW, the real world is a lot more complicated.

Topological generalizations of subgraphs defined by the role of the nodes looks promising nonetheless:

Severe flaw of the study: when discussing, they disregard the importance of a certain type of motif on the basis of its low abundance, when such low abundance could actually be very significant: for example, in controlling in a certain way a very specific pathway. For example, there is a low abundance of multi-input FFL in transcription networks, but then unique, central proteins like p53 may be regulated that way and are capital to the system.

I have no data to discard the application of such an argument to neuronal networks as well.


Peter Bräunig and Malcolm Burrows. 2004. Projection Patterns of Posterior Dorsal Unpaired Median Neurons of the Locust Subesophageal Ganglion. J.Comp. Neurol. 478:164175.

20051007

Check Bräunig, 1991, for descriptions of the connectivity of the neurons.




Last updated: 2012-05-08 11:10 Zurich time. Copyright Albert Cardona.