The role of lineage, hemilineage and temporal identity in establishing neuronal connectivity in the Drosophila larval CNS

The mechanisms specifying neuronal diversity are well-characterized, yet it remains unclear how or if these mechanisms regulate neuronal morphology and connectivity. Here we map the developmental origin of 78 bilateral pairs of interneurons from seven identified neural progenitors (neuroblasts) within a complete TEM reconstruction of the Drosophila newly-hatched larval CNS. This allows us to correlate developmental mechanism with neuronal projections, synapse targeting, and connectivity. We find that clonally-related neurons from project widely in the neuropil, without preferential circuit formation. In contrast, the two NotchON/NotchOFF hemilineages from each neuroblast project to either dorsal motor neuropil (NotchON) or ventral sensory neuropil (NotchOFF). Thus, each neuroblast contributes both motor and sensory processing neurons. Lineage-specific constitutive Notch transforms sensory to motor hemilineages, showing hemilineage identity determines neuronal targeting. Within a hemilineage, temporal cohorts target processes and synapses to different sub-domains of the neuropil, effectively “tiling” the hemilineage neuropil, and hemilineage/temporal cohorts are enriched for shared connectivity. Thus, neuroblast lineage, hemilineage, and temporal identity progressively restrict neuropil targeting, synapse localization, and connectivity. We propose that mechanisms generating neural diversity are also determinants of neural circuit formation.


38
Tremendous progress has been made in understanding the molecular mechanisms generating neuronal 39 diversity in both vertebrate and invertebrate model systems. In mammals, spatial cues generate distinct pools 40 of progenitors which generate a diversity of neurons and glia appropriate for each spatial domain (1). The 41 same process occurs in invertebrates like Drosophila, but with a smaller number of cells, and this process is 42 particularly well-understood. Spatial patterning genes act combinatorially to establish single, unique 43 progenitor (neuroblast) identity; these patterning genes include the dorsoventral columnar genes vnd, ind, msh 44 (2-4) and the orthogonally expressed wingless, hedgehog, gooseberry, and engrailed genes (5-8). These factors endow 45 each neuroblast with a unique spatial identity, the first step in generating neuronal diversity ( Figure 1A, left). 46 Here we focus on the left and right sides of abdominal segment 1 (A1L, A1R) and so segment-specific 47 patterning due to Hox gene expression is not relevant. The second step occurs as each neuroblast "buds off" 48 a series of ganglion mother cells (GMCs) which acquire a unique identity based on their birth-order, due to 49 inheritance from the neuroblast of a "temporal transcription factor"-Hunchback (Hb), Krüppel (Kr), Pdm, 50 and Castor (Cas) -which are sequentially expressed by nearly all embryonic neuroblasts (9). The combination 51 of spatial and temporal factors leads to the production of a unique GMC with each neuroblast division 52 ( Figure 1A, middle). The third and final step in generating neuronal diversity is the asymmetric division of 53 each GMC into a pair of post-mitotic neurons; during this division, the Notch inhibitor Numb (Nb) is 54 partitioned into one neuron (Notch OFF neuron) whereas the other sibling neuron receives active Notch 55 signaling (Notch ON neuron), thereby establishing two distinct hemilineages (10-13)( Figure 1A, right). In 56 summary, three developmental mechanisms generate neuronal diversity within the embryonic CNS: 57 neuroblast spatial identity, GMC temporal identity, and neuronal hemilineage identity. 58 A great deal of progress has also been made in understanding neural circuit formation in both vertebrates 59 and invertebrate model systems, revealing a multi-step mechanism. Mammalian neurons initially target their 60 axons to broad regions (e.g. thalamus/cortex), followed by targeting to a neuropil domain (glomeruli/layer), 61 and finally forming highly specific synapses within the targeted domain (reviewed in 14).

62
Despite the progress in understanding the generation of neuronal diversity and the mechanisms 63 governing axon guidance and neuropil targeting, how these two developmental processes are related remains 64 unknown. While it is accepted that the identity of a neuron is tightly linked to its connectivity, the 65 developmental mechanisms involved remain unclear. For example, do clonally-related neurons target similar 66 regions of the neuropil due to the expression of similar guidance cues? Do temporal cohorts born at similar 67 times show preferential connectivity? Are neurons expressing the same transcription factor preferentially 68 interconnected? It may be that lineage, hemilineage, and temporal factors have independent roles in circuit 69 formation; or that some mechanisms are used at different steps in circuit assembly; or that mechanisms used 70 to generate neural diversity could be independent of those regulating circuit formation. Here we map 71 neuronal developmental origin, neuropil targeting, and neuronal connectivity within a whole CNS TEM 72 reconstruction (15). This provides us the unprecedented ability to identify correlations between development 73 and circuit formation -at the level of single neurons/single synapses -and test those relationships to gain 74 insight into how mechanisms known to generate diversity might be coupled to mechanisms of neural circuit 75 formation. We find that lineage, hemilineage, and temporal identity are all strongly correlated with features of 76 neuronal targeting that directly relate to establishing neural circuits. It is not possible to determine the clonal relationship of neurons in the TEM volume based on anatomical 83 features alone; for example, clonally-related neurons are not ensheathed by glia as they are in grasshopper 84 embryos or the Drosophila larval brain (16, 17). We took a multi-step approach to identify clonally-related 85 neurons in the TEM reconstruction. First, we generated sparse neuroblast clones and imaged them by light 86 microscopy. All neuroblasts assayed had a distinctive clonal morphology including the number of fascicles data not shown). The tendency for neuroblast clones to project one or two fascicles into the neuropil has also 89 been noted for larval neuroblast clones (11-13). We assigned each clone to its parental neuroblast by 90 comparing our clonal morphology to that seen following single neuroblast DiI labeling (18)(19)(20), and what has 91 been reported previously for larval lineages (21, 22), as well as the position of the clone in the segment, and in 92 some cases the presence of well-characterized individual neurons (e.g. the "looper" neurons in the NB2-1 93 clone). Note that we purposefully generated clones after the first-born Hb+ neurons, because the Hb+ 94 neurons have cell bodies contacting the neuropil and do not fasciculate with later-born neurons in the clone, 95 making it difficult to assign them to a specific neuroblast clone. We found that neurons in a single neuroblast 96 clone, even without the Hb+ first-born neurons included, project widely throughout the neuropil, often 97 targeting both dorsal motor neuropil and ventral sensory neuropil, as well as widely along the mediolateral 98 axis of the neuropil ( Figure 1B).

99
Next, we used these neuroblast lineage-specific features to identify the same clonally-related neurons in 100 the TEM reconstruction in A1L. We identified neurons that had clustered cell bodies, clone morphology 101 matching that seen by light microscopy ( Figure 1C), and one or two fascicles entering the neuropil ( Figure   102 1D,E). The similarity in overall clone morphology between genetically marked clones and TEM reconstructed 103 clones was striking (compare Figure 1B and 1C). We used two methods to validate the clonal relationship 104 observed in the TEM reconstruction. We used neuroblast-specific Gal4 lines (13, 23) to generate MCFO 105 labeling of single neurons, and found that in each case we could match the morphology of an MCFO-labeled 106 single neuron from a known neuroblast to an identical single neuron in the same neuroblast clone within the 107 TEM reconstruction (data not shown). We also validated the reliability of clone morphology and neuron  whereas the Notch OFF hemilineage shares a different morphology (11-13). We hypothesized that the observed 123 morphological differences may be due to hemilineage identity ( Figure 2). First, we used NBLAST (24) to 124 compare the morphology of clonally related neurons. We observed that five of the seven neuroblast lineages 125 generated two highly distinct candidate hemilineages that each projected to a focused domain in the dorsal or previously been shown to directly generate a single Notch OFF hemilineage due to direct differentiation of the 129 neuroblast progeny as neurons, bypassing the terminal asymmetric cell division (25, 26). We conclude that 130 NBLAST can identify candidate hemilineages, with one projecting to the ventral neuropil, and one projecting 131 to the dorsal neuropil ( Figure 2G). This is a remarkable subdivision within each lineage, because the dorsal  can achieve all three goals by using neuroblast-specific Gal4 lines to drive expression of constitutively active 146 Notch (Notch intra ) to transform Notch OFF hemilineages into Notch ON hemilineages.

147
There are Gal4 lines specifically expressed in NB1-2, NB7-1, and MB7-4 (13, 29) which we used to drive hemilineage that projects to dorsal/motor neuropil (or makes glia), and a Notch OFF hemilineage that projects 159 to ventral/sensory neuropil. In conclusion, we show that NBLAST can be used to accurately identify 160 neuroblast hemilineages; that Notch ON /Notch OFF hemilineages project to motor/sensory neuropil domains, 161 respectively; and most importantly, that hemilineage identity determines neuronal targeting to the motor or 162 sensory neuropil.   Additionally, we generated a new Hb-LexA construct in order to identify additional Hb+ neurons, which 201 we then traced in the EM volume ( Figure 6E,F, cyan neurons). We also used cas-gal4 to drive MCFO in order 202 to identify new late-born neurons ( Figure 6E,F magenta neurons). In total, we identified 18 neurons in the 203 EM volume with known birthdates ( Figure 6E,F; Fig.S4). In order to quantify distance from the neuropil, we 204 measured the neurite length between the cell body and the neuropil entry point. We found that all confirmed 205 Hb+ neurons were located close to the neuropil, whereas late-born neurons were located more distantly 206 ( Figure 6G,H). We also confirmed that left/right neuronal homologs had extremely similar cortex neurite 207 lengths ( Figure 6I). Thus, we confirm that neuronal cortex neurite length is consistent across two 208 hemisegments, and can be used to approximate the temporal identity of any neuron in the TEM 209 reconstruction. cohorts appeared far more overlapping (Fig. S5). To quantify this, we compared the synapse similarity of 217 hemilineage-related neurons and temporal-related neurons and found that neurons related by hemilineage 218 were more similar than those related by birthdate (Fig. S6). We conclude that hemilineages, not temporal 219 cohorts, are more important determinants of neuropil targeting. 220 We next asked whether temporal identity is linked to more precise sub-regional targeting or "tiling" of  We next tested whether other lineages contained hemilineage/temporal cohorts that "tile" neuronal 231 projections and synapse localization. Indeed, examination of the NB5-2 ventral hemilineage showed that 232 early-and late-born neurons targeted their projections to "sub-regional" domains of the full hemilineage  neurons that share a common input or output have a value of 2 synapses apart, with a maximum of seven 263 synapses apart. We found that neurons in a hemilineage had a much lower minimum synapse distance than 264 random, indicating shared connectivity; similarly, neurons in a temporal cohort within a hemilineage also have 265 significantly lower minimum synapse distances, with over 60% of all neurons in the same temporal cohort 266 being separated by two synapses or less ( Figure 9I,J). We conclude that temporal cohorts share common 267 connectivity. Notch OFF neurons projected to the ventral neuropil. It is unlikely that all Notch ON hemilineages target the 289 dorsal neuropil, however, as the NB1-1 interneuron pCC is from a Notch ON hemilineage (10) yet projects 290 ventrally and receives strong sensory input, and its sibling aCC motor neuron is from the Notch OFF 291 hemilineage (10) and projects dendrites in the dorsal motor neuropil. We think it is more likely that the 292 Notch ON /Notch OFF provides a switch to allow each hemilineage to respond differently to dorsoventral 293 guidance cues: in some cases the Notch ON hemilineage projects dorsally, and in some cases it projects 294 ventrally. Nevertheless, our finding that neuroblasts invariably produce both sensory and motor hemilineages 295 reveals the striking finding that the sensory and motor processing components of the neuropil are essentially 296 being built in parallel, with one half of every GMC division contributing to either sensory or motor networks. 297 This has not been observed in larval hemilineages, and may be the result of an evolutionary strategy to 298 efficiently build the larval brain as fast as possible.

299
While we do observe some differences between embryonic and larval hemilineages, the similarities are far 300 more striking. Previous work has shown that larval and embryonic hemilineages have similar morphological 301 features (13), suggesting the possibility that these neurons could be performing analogous functions. Here we 302 show that two components of a proprioceptor circuit, the Jaam and Saaghi neurons (32), are derived from 303 two hemilineages of NB5-2 (also called lineage 6 (21)). Activation of either of these hemilineages in adults 304 results in uncoordinated leg movement, consistent with the idea that these hemilineages could be involved in 305 movement control. Similarly, adult activation of the NB3-3 lineage (also called lineage 8 (21)) caused postural 306 8 effects, again consistent our previous findings that activation of this lineage in larvae cause postural defects 307 (32). In the future, it will be interesting to further explore the functional and organizational similarities of the 308 embryonic and larval nervous systems. 309 Our results suggest that all neurons in a hemilineage respond similarly to the global pathfinding cues that hemilineages (see Figure 3), or expression of Numb will make two Notch OFF hemilineages. In this way it will 317 be possible to obtain RNAseq data on neurons with a common neuropil targeting program.

324
We used the cortex neurite length of neurons as a proxy for birth-order and shared temporal identity. We 325 feel this is a good approximation (see Figure 5 for validation), but it clearly does not precisely identify  In this study we demonstrate how developmental information can be mapped into large scale 351 connectomic datasets. We show that lineage information, hemilineage identity, and temporal identity can all 352 be accurately predicted using morphological features (e.g. number of fascicles entering the neuropil for 353 neuroblast clones, and radial position for temporal cohorts). This both greatly accelerates the ability to 354 identify neurons in a large EM volume as well as sets up a framework in which to study development using 355 datasets typically intended for studying connectivity and function. We have used this framework to relate 356 developmental mechanism to neuronal projections, synapse localization, and connectivity; in the future we