Doe Lab Research
Temporal transcription factors: timers, integration, and function
We previously showed that embryonic neuroblasts sequentially express five temporal transcription factors (TTFs) – Hunchback (Hb), Kruppel, Pdm, Castor, Grainy head – that specify temporal identity during the rapid (45 minute cell cycle) embryonic neuroblast lineages. The need for speed probably explains why the timer for this temporal cascade is intrinsically programmed into the neuroblast lineages: the TTF cascade occurs in relatively normal neuroblasts cultured in vitro and even in G2-arrested neuroblasts (Grosskortenhaus et al., 2005).
More recently, we have extended our analysis of TTFs to larval brain neuroblasts, which go through over 100 divisions over five days (Figure, above), and shown that they require extrinsic hormonal signaling to generate temporal identity. My postdoc Mubarak Syed, and other labs, identified candidate TTFs in larval neuroblast lineages, including the 'type II' neuroblasts that uniquely make Intermediate Neural Progenitors (INPs) which have an extended lineage and contribute neurons to the adult brain central complex. We have shown that each INP undergoes a temporal transcription factor cascade (D, Grh, Ey) that is required for generating neuronal diversity (Figure, below). More recently, grad student Luis Sullivan, when in my lab, tested the role of the Eyeless late INP temporal factor in specifying central complex neuronal identity. Luis found two subtypes derived from young Dichaete+ INPs and two subtypes from old Eyeless+ INPs. Importantly, he found that neurons in a common temporal cohort (e.g. P-FN and P-EN) had similar neuronal targeting and connectivity, matching our findings in larval motor circuits (below). Luis went on to show that the Eyeless TTF was required for proper neuronal fate specification, connectivity, and adult navigation behavior.
There are many interesting open questions. Can the larval neuroblasts undergo a normal early or late TTF cascade when cultured in vitro? Do the TTF transitions require cell cycle progression? Importantly, do the candidate TTFs actually specify neuronal temporal identity? Do the TTFs act in the neuroblast (or INP) to specify permanent neuronal identity? What are the targets of the TTFs, and do they have "terminal selector" properties (persistent expression and coordinate regulation of multiple neuronal features)? To begin to address these latter questions, we will map the developmental origin of the ~30 new adult central complex neuron subtype markers generated at Janelia, and to use these markers to test candidate neuroblast TTFs for a role in generating neuronal diversity.
Developmental origin of larval motor circuits
Three mechanisms are used to generate neuronal diversity in the Drosophila CNS. First, spatial patterning cues specify individual neuroblast identity; second, a cascade of temporal transcription factors (TTFs) specify neuroblast progeny temporal identity based on their birth-order; and third, each progeny (GMC) divides asymmetrically into NotchON/NotchOFF sibling neurons, resulting in hemilineage identity. Ultimately, these three mechanisms generate 300 unique neurons in each hemisegment (Figure).
To understand how each mechanism contributes to neural circuit formation, we needed to identify neurons in a circuit and map their developmental origin – e.g. do neurons in a circuit come from a single neuroblast, single temporal cohort, or single hemilineage? Graduate student Brandon Mark took on this challenge, ultimately mapping the progeny of seven neuroblast lineages (78 neurons) within the serial section TEM reconstruction of segment A1. The combination of complete developmental history and connectomic information allowed Brandon to determine the relationship beween developmental mechanism and circuit assembly. He made four key observations.
(1) Neurons within a lineage, hemilineage, or temporal cohort were not preferentially interconnected.
(2) Neuroblasts always generated one hemilineage projecting to a dorsal (motor) neuropil domain, and the other hemilineage projecting to a ventral (sensory) neuropil domain; thus each neuroblast contributes neurons to both sensory and motor processing.
(3) Within a hemilineage, neurons with different temporal identity target their pre- and post-synapses to different neuropil sub-domain; similar to observations in mammals.
(4) Within a hemilineage, neurons with different temporal identity have distinct connectivity (see below).
These observations led to a simple model for how developmental mechanisms generate neuronal connectivity. Each neuroblast contributes a motor hemilineage and a sensory hemilineage to the CNS; neurons in a hemilineage target a specific neuropil domain based on a shared response to global pathfinding cues. Within a hemilineage, neurons in different temporal cohorts target their synapses to distinct sub-domains. Ultimately, synapse formation and circuit assembly occur within each “hemilineage-temporal” neuropil subdomain (Figure).
Our model makes many predictions, including some we have already tested. The model predicts altered temporal identity would affect connectivity but not regional targeting (which is conferred by hemilineage identity). This is precisely the effect of misexpressing the Hb TTF in the NB7-1 lineage: axons project normally to the correct spatial muscle group, but synaptic connections to individual muscles are abnormal (Seroka et al., 2019).
There are many interesting open questions remaining. First, the model predicts that dorsal and ventral hemilineages respond differentially to the Sema2 dorsoventral gradient (Figure, above), but how this occurs is unknown; identifying receptors differentially expressed in dorsal and ventral hemilineages would help identify the relevant ligand. Second, it remains unknown whether the temporal identity sub-domains arise by self-avoidance, by spacing cues, or by a precise response to global patterning cues. Lineage-specific ablation of a single temporal cohort will help distinguish among these possibilities. We will continue to test this model at every level by manipulating neuroblast identity, hemilineage identity, and temporal identity. Most importantly, we will investigate the molecular mechanisms linking developmental origin to neuropil targeting and circuit assembly. To determine how hemilineage identity leads to regional neuropil targeting, we can overexpress Notch in a single neuroblast lineage to create a common hemilineage identity, TU-tag or FACS purify the neurons at the time of synapse choice (late embryo), examine their transcriptome for cell surface receptors, and test receptor function for hemilineage targeting. Similarly, we can overexpress both Notch and a temporal factor to create a single hemilineage/temporal identity, purify the neurons, and mine the transcriptome for cell surface molecules that can be tested for a role in subregional targeting and synaptic partner choice.
Developmental origin of neural circuits – the adult brain central complex
An important feature of the type II neuroblast lineages is that they contain two axes of temporal identity (neuroblast and INP), and they make substantial contributions to the adult central complex, a brain region required for celestial navigation. We identified a dozen candidate TTFs sequentially expressed in neuroblasts (Syed et al., 2017), and three TTFs in INPs (Bayraktar and Doe, 2013).
Graduate student Luis Sullivan tested the role of the late INP temporal factor Eyeless in specifying central complex neuronal identity. He found that all four types were made from young neuroblasts, with two derived from the young INP temporal factor Dichaete+ window and two from the old INP temporal factor Eyeless+ window (Figure). He showed that loss of Eyeless eliminated late-born neuron types (e.g. E-PGs), while increasing early-born neuronal types (e.g. P-ENs). Importantly, the ectopic early-born neurons showed morphology and connectivity identical to the endogenous early-born neurons. Furthermore, transient Eyeless knock-down at the time these neurons are generated resulted in a morphologically normal adult central complex (as expected due to the precision of the manipulation), but a defect in flight navigation behavior (Sullivan et al., 2019). Thus, the Eyeless TTF regulates the identity and behavioral function of central complex neurons, similar to the role of embryonic neuroblast TTFs in larval motor circuit formation.
Future directions. Our goal is to link developmental origin within type II lineages to neural circuit formation in the adult central complex (Figure). We can perform highly specific targeted developmental manipulations (e.g. increasing or decreasing the number of one class of central complex neurons) and assay the consequences anatomically or behaviorally. We can use optogenetics and two photon holographic targeting to precisely activate/silence single adult central complex columns of a single neural subtype (e.g. E-PG), and a second laser to image putative downstream neuronal responses (e.g. PF-R). We have also established a behavioral arena where tethered adult flies can be presented with visual stimuli (e.g. fictive sun, polarized light), and measure flight behavior (Figure). In particular, we are interested in testing a recently proposed navigation circuit model which includes three of the neuronal subtypes that we have mapped to their developmental origin. As we analyze the role of neuroblast and INP TTFs in specifying neuronal identity and circuit formation, we will run parallel behavior analyses to identify defects in flight navigation.
Developmental origin of neural circuits – the MDN circuit drives backward locomotion in larvae and adults
We used an optogenetic behavioral screen to identify neurons that drive specific larval behaviors (Clark et al., 2016), with the goal of mapping the relevant circuits and determining the developmental origins of the component neurons. Here we identified two cholinergic (excitatory) brain descending neurons that, when activated, induced backward larval locomotion (Movie). We named them Mooncrawler Descending Neurons (MDNs) as a tip of our hat to the Moonwalker Descending Neurons that induce backward walking in adult flies. Arnaldo traced the MDN downstream neurons and found three independent motor circuits: one that inhibits forward locomotion via Pair1-mediated inhibition of the forward-specific A27h premotor neuron, one that induces backward locomotion via activation of the backward-specific A18b premotor neuron, and one that we are currently characterizing (Figure). We concluded that the MDNs induce a coordinated switch between antagonistic locomotor behaviors (halt forward; initiate backward). We were curious if the larval MDNs might survive metamorphosis to become the adult MDN neurons, so we genetically immortalized the larval MDNs and, surprisingly, found they could trigger backward walking in adults, despite the profound differences in target motor neurons and motor rhythms. Our model is that MDN activity coordinately halts forward locomotion and drives backward locomotion.
We plan to determine the developmental origins of the neurons in this circuit to ask whether they share a spatial identity (common neuroblast), temporal identity (common birthdate or TTF expression), or common hemilineage (NotchON/NotchOFF). We also plan to use genetics to permanently mark each of the neurons in the larval circuit to see if they all, like MDN itself, persist into adult. We expect the adult circuit to have novel sensory input, and new leg motor neuron output, but the interneuronal core of the larval circuit may be maintained and remodeled to drive both limbless locomotion (larvae) and limbed locomotion (adult). Ultimately, we want to identify the cell surface molecules that promote larval MDN circuit formation, and ask if the same cell surface molecules are subsequently used to promote adult MDN circuit formation.
Integration of spatial and temporal cues to generate neuronal diversity
An important question in neuroscience is how spatial and temporal cues are integrated by progenitors to generate unique cell lineages. For example, each embryonic temporal transcription factor specifies a different neuronal identity in each neuroblast lineage (see below). Thus, spatial information (neuroblast identity) alters the output of each temporal factor. This could be due to spatial factors altering chromatin and thus changing temporal factor binding, or spatial and temporal factors binding genomic loci independently but acting combinatorially. Distinguishing these models requires mapping chromatin landscapes and TTF genomic binding sites in single neuroblast lineages – posing significant technical hurdles.
My postdoc Sonia Sen made impressive progress on this long-standing question (Sen et al., eLife 2019). Sonia used the DamID method to compare open chromatin landscapes in two adjacent neuroblast lineages, and Dam:Hb to map Hb genomic binding sites in each neuroblast lineage. Sonia found that NB5-6 and NB7-4 have different chromatin landscapes; that the spatial factor Gsb had binding correlated with open chromatin in NB5-6 but not NB7-4; and that Dam:Hb binding occurred within open chromatin in NB5-6, but in NB7-4 those sites were “closed” and unbound by Dam:Hb. Our model is that spatial factors generate neuroblast-specific chromatin landscapes, which lead to neuroblast-specific binding of temporal factors, resulting in the production of neuroblast-specific progeny. This does not exclude the possibility that temporal factors like Hb also modify chromatin, resulting in neuron-specific chromatin states following integration of spatial and temporal factors.
We plan to extend these studies to determine whether Gsb is necessary and sufficient to open chromatin domains, as well as testing the spatial factor Engrailed for a similar role. We will use our Dam:Hb data to identify lineage-specific Hb target genes, which will help reveal the mechanism by which the Hb TTF specifies lineage-specific neuronal identity. We also plan to characterize the role of Hb and other temporal factors in creating distinct chromatin landscapes.