A GABAergic Maf-expressing interneuron subset regulates the speed of locomotion in Drosophila

Interneurons (INs) coordinate motoneuron activity to generate appropriate patterns of muscle contractions, providing animals with the ability to adjust their body posture and to move over a range of speeds. In Drosophila larvae several IN subtypes have been morphologically described and their function well documented. However, the general lack of molecular characterization of those INs prevents the identification of evolutionary counterparts in other animals, limiting our understanding of the principles underlying neuronal circuit organization and function. Here we characterize a restricted subset of neurons in the nerve cord expressing the Maf transcription factor Traffic Jam (TJ). We found that TJ+ neurons are highly diverse and selective activation of these different subtypes disrupts larval body posture and induces specific locomotor behaviors. Finally, we show that a small subset of TJ+ GABAergic INs, singled out by the expression of a unique transcription factors code, controls larval crawling speed.


Introduction
The wiring and functioning of the neuronal circuits that provide animals with the ability to move over a range of speeds have been extensively studied (1). In mammals as in invertebrates the speed of locomotion is regulated by central pattern generator (CPG) neuronal circuits which coordinate, via motoneurons (MNs), the sequential activation of muscles (2)(3)(4)(5). the controlled expression of ion channel proteins that regulate INs activity (7,8). Such an approach used in mouse and in Drosophila has proved instrumental in dissecting the core logic of the CPG circuits that generate the rhythm and pattern of motor output (3,5,9). Still, in light of the remarkable diversity of INs in the vertebrate spinal cord (6,10) and in the Drosophila nerve cord (11,12), a thorough description of the functioning of the CPG regulating locomotion in animals is far from complete.  (5,(13)(14)(15). However, little is known about the combinatorial expression of TFs within these different IN subtypes identified so far and this lack of knowledge impedes cross species comparisons thus limiting [5] our understanding of the common principles of CPG organization in vertebrates and invertebrates.
Here we investigate in the Drosophila larva the role of a small pool of highly diverse INs (23/hemisegment) expressing the evolutionary conserved TF Traffic Jam (TJ), the orthologue of MafA, MafB, c-Maf and NRL in the mouse. Interestingly, MafA, MafB and c-Maf are expressed by restricted subpopulations of ventral premotor INs in the developing mouse spinal cord (6,10,16) but the function of these INs subtypes is to date unknown. To characterize TJ-expressing INs we generated a TJ-Flippase line and used intersectional genetics to activate TJ + subpopulations depending on their neurotransmitter properties. We found that manipulation of these IN subsets modulates larval locomotor behaviours. Our results also showed that activation of a restricted subpopulation of GABAergic/Per + /TJ + (3 INs/segment), belonging to the PMSIs group of INs and known as MNB progeny neurons, significantly impacts the crawling speed of the larvae. [6]

Traffic Jam is expressed in a restricted subpopulation of neurons located in the VNC and involved in larval locomotion
We initiated a detailed analysis of the transcription factor (TF) Traffic Jam (TJ) expression in the embryonic and larval nervous systems, using a previously characterized TJspecific antibody (17) and an enhancer trap for TJ (TJ-Gal4) (18). During embryogenesis, TJ expression is first detected in late stage 12 (st 12) in few cells in the brain and in 12 to 15 cells/hemisegment in the ventral nerve cord (VNC) (Supp. Fig.1A). Co-immunostaining with the glial marker Repo showed that TJ is exclusively expressed in neurons and excluded from glia cells (Supp. Fig.1B). Using TJ-Gal4::UAS-GFP we found that TJ-Gal4 faithfully recapitulates TJ expression in all embryonic (Supp. Fig.1C) and larval stages analysed ( Fig.1C-G). Closer analysis of TJ expression over time showed that TJ is consistently found in a subset of 29 neurons per hemisegment in the VNC abdominal region (A2-A6) from st17 to L3 larval stages (Fig.1A, A', B, B') and excluded from sensory neurons (data no shown). We used TJ-Gal4::UAS-H2AGFP in combination with anti-TJ immunostainings to establish a precise topographic map of TJ + neurons in second instar larvae, a stage representative of the continuous expression pattern of TJ (Fig.1 C-G, C1-G1).
Next, we decided to explore the function of TJ + neurons in larval locomotion using the TJ-Gal4 driver as a tool to either inactivate or activate the entire TJ + neuronal population.
We inactivated neurons by expressing thermosensitive shibire (shi ts ) (19); neuronal activation was achieved by expressing TrpA1 (20). Inactivation of the entire TJ + population led to a slight decrease in the number of larval peristaltic waves ( Fig.1H -second beige bar). This decrease seemed caused by a general disorganization of the peristaltic waves, with the segments of the larvae failing to contract in the coordinated, sequential way ( Fig.1J and 1K, video 1). Activation of the TJ + neurons had more drastic effects, with almost complete abolition of locomotion ( Fig.1I -second red bar) and larvae displaying a complete paralysis which we named "spastic paralysis". This phenotype was characterized by immobility, tonic contraction of all body segments and a drastic shortening of the whole larval body length ( Fig.1L, video 2). When placed back at permissive temperature (23°C), the larvae resumed normal locomotion, proving that this spastic paralysis phenotype is fully reversible. TJ is expressed in a restricted number of neurons in the cerebellar lobes. To assess the role of [7] these neurons in the spastic paralysis phenotype, we restricted the expression of TrpA1 to the TJ + cerebellar lobes neurons using Tsh-Gal80 (21) in combination with TJ-Gal4 and UAS-TrpA1. Under these conditions locomotion appeared normal ( Fig.1I -second salmon pink bar), arguing that TJ + neurons in the cerebellar lobes do not have a function in locomotion in Drosophila larva.
We thus conclude that TJ + neurons within the VNC are part of a neuronal circuit controlling Drosophila larval crawling and that normal function of TJ + neurons is to maintain proper muscle contraction and peristaltic wave propagation during locomotion.

Activation of TJ + neurons in the VNC using an intersectional genetic approach leads to spastic paralysis of the larvae
To further characterize the identity of TJ + neurons regulating crawling behaviour in the larva we developed an intersectional-based genetic approach, using candidate LexA  Fig.2A). We quantified the efficiency of "flip-out" events in TJ + neurons and found from hemisegment to hemisegment and within all specimen analyzed (n=4) that more than half the TJ + neurons (64%) are already recombined in L1 larva ( Fig.2A-C) and 89% in young L2 stage ( Fig.2D-F). These numbers demonstrate the accurate recombination triggered by TJ-Flp. An examination of developing egg chambers of the ovary, a structure in which TJ has been reported to be specifically expressed by somatic cells (17), revealed that all TJ + follicular and border cells are GFP-labelled thus confirming the accuracy of TJ-Flp (supplementary Fig.2B-C). Finally, we used Tsh-LexA to express a LexAop-FRT-STOP-FRT-dTrpA1 transgene in combination with TJ-Flp, anticipating that activation of TJ + neurons in the VNC only should give rise to the drastic spastic paralysis phenotype described above. We found it was indeed the case in all L1 larvae tested (n=12) (Fig.2K, video 3). [8] Collectively, these results show that the TJ-Flp line we generated is an accurate and powerful genetic tool to genetically manipulate TJ + neurons, thus validating our intersectional-based genetic approach.

TJ + motoneurons activation only partially impacts larval crawling
We next asked whether the spastic paralysis phenotype we observe upon activation of the whole TJ + population would actually be caused by TJ-Gal4 driving in MNs. Using pMad and Eve as reliable molecular markers for MNs, we found TJ expression from st13 onward in 3 CQ/Us' MNs namely U1, U2 and U5 (asterisks in Fig.3  is a contingent of 2 TJ + /pMad + /Islet-myc + MNs (ISNd MNs that project on muscles VO3-VO6), 3 TJ + /pMad + /Eve + MNs (ISNdm MNs U1, U2 and U5, that respectively project on muscles DO1, D02 and LL1) and 1 TJ + /pMad + MN that projects on muscle DO5.
To investigate the role of TJ + MNs on locomotion we used CQ2-LexA (13) which allows specific activation of the three TJ + U MNs U1, U2 and U5. Detailed monitoring of CQ2-LexA expression also revealed that this line additionally drives in mid-L1 stage in the TJ + DO5innervating MN (arrowhead in Fig.3D) and in 1 unidentified TJ + IN located dorsally (not shown). Using CQ2-LexA, we therefore activated 4 of the 6 TJ + MNs per hemisegment, and observed a moderate decrease in the number of peristaltic waves and no obvious locomotor phenotype nor body posture defect in the vast majority of the larvae (Fig.3G -second black [9] bar). Importantly, no spastic paralysis was ever observed upon activation of TJ + MNs using the CQ2-LexA driver (Fig.3E, F and video 4).
From this experiment we conclude that activation of more than 50% of TJ + MNs does not trigger the spastic paralysis phenotype observed upon activation of the whole TJ + population but rather causes a slight defect in locomotion probably due to the constant contraction of the dorsal muscles innervated by TJ + MNs. indicating that those cells are unpaired, midline cells (24). When activated, the TJ + - [10] GABAergic IN subpopulation led to seemingly normal locomotion (Fig.4U, video 6). However, counting the number of peristaltic waves actually revealed that locomotion was slowed, with a reduction of 37% of the number of waves compared to control specimen at 31°C ( Taken together this data shows that activation of restricted subpopulations of TJ + neurons defined on the basis of their neurotransmitter identity impact the crawling of the larvae with distinct behavioural hallmarks.

A population of 3 TJ + GABAergic INs per segment regulate the speed of locomotion
While searching for more restricted LexA drivers that would allow us to subdivide more finely the TJ + population implicated in locomotion, we identified the per-LexA driver (5) whose expression co-localizes with 9 of the most ventral TJ + neurons per segment (  28,29) and that do not have counterparts in the adjoining hemisegment. We were intrigued by those 3 unpaired neurons and used a [11] sophisticated triple intersectional genetic approach based on the combinatorial expression of TJ, Per and Gad1 to specifically activate them. A large proportion of the manipulated larvae (~70%) displayed a marked reduction in their speed of locomotion compared to control larvae at 31°C (Fig.5J-L -video 9), while a minority appeared unaffected (~30%). The incomplete or low dTrpA1 expression in every pool of 3 neurons in all segments for a given larva could explain the heterogeneity within the experimental group. In agreement with such possibility we noticed rather weak Per-Gal4 expression in these neurons, which would be predicted to reduce the expression levels of the LexA DBD component in this triple intersectional approach.
Together these results show that activation of a population of 3 GABAergic TJ + /Per + neurons per segment impacts the speed of locomotion in Drosophila larva.

TJ + Per + GABAergic neurons, known as MNB progeny neurons, express a unique combination of TFs
Given the median position of the 3 GABAergic TJ + /Per + neurons and the fact they lack counterparts in the adjoining hemisegment, we hypothesized that those neurons are a subset of the midline cells. Midline cells belong to the sim domain (24) and we confirmed by quadruple immunostaining with TJ, Gad1, Prospero and sim-Gal4 driving a nuclear GFP that TJ + median GABAergic neurons are sim + in late embryonic stage 17 VNC ( Fig.6A-B2, single and double empty arrowheads). We noticed weak sim expression in these neurons at this stage, an observation consistent with low sim expression in late embryonic stages as previously reported (25). It is important to note that the TJ + non GABAergic (glutamatergicpositive) neurons located in the ventral part of the VNC are not sim-Gal4 + (depicted by arrows in Fig.6C-C2). We then found that 2 of the 3 TJ + GABAergic midline cells belong to the Median Neuroblast (MNB) progeny subpopulation identified by nuclear Prospero expression (Prosp-nucl) (24) (Fig.6A, double empty arrowheads). We also found that all 3 TJ + GABAergic cells are fkh + and EN + (Fig.6D-E, full arrowheads), two TFs known to be expressed in a subpopulation of MNB progeny but also iVUMs (24). To further delineate the exact identity of the third TJ + GABAergic midline neuron (GAD + , Per + , fkh + , En + , Prosp -) we examined stage 16 embryo where midline cell identities can be determined accordingly to the highly stereotyped dorso-ventral and anterior-posterior location of the cells. Using these stereotyped positions along with Per, fkh and TJ immunostainings, we showed that TJ is not [12]  Altogether we identify the TJ + Per + GABAergic population regulating the speed of locomotion as 3 MNB progeny neurons derived from a Jumu + progenitor domain and singled out by the combinatorial expression of TJ, En, fkh, Per, Hlh3b and grain. [13]

Discussion
In this study, we characterized from embryogenesis to larval stage L3 TJ-expressing neurons in the VNC and investigated their role in the crawling behaviour of freely moving Drosophila larvae. We generated a TJ-Flp line and developed an intersectional genetic approach based on the use of TrpA1 to specifically activate different TJ + subpopulations depending on their neurotransmitter properties.  (13,29). [14] The identification of such cell type-specific TF and the characterisation of genetic tools such as Gal4 or LexA lines expression patterns of these "markers" have recently proved instrumental for characterizing neuronal circuits in the larval VNC. For example the recently elucidated neuronal circuit that promotes escape behaviour upon noxious stimuli in Drosophila larvae (30) involves the contribution of SNa MNs that were genetically amenable due to the highly specific BarH1-gal4 line (27). Similarly, the very restricted EL-Gal4 driver active in Eve-expressing lateral (EL) INs (31)capturing the has allowed deciphering the implication of EL INs within a sensorimotor circuit that maintains left-right symmetry of muscle contraction amplitude in the Drosophila larva (13). We thus foresee that our extensive mapping of TJ-expressing neurons, together with the reliable TJ-Flp line we generated will facilitate future studies aiming at identifying and investigating neuronal circuit formation and functioning in embryonic or larval VNC.

TJ + cholinergic neurons control body posture in Drosophila larva
Activation of the TJ + cholinergic subpopulation gave rise to a ventral contraction phenotype, with larvae frequently adopting a "crescent shape" position. We noticed that the ventral contraction phenotype was heterogeneous between individuals. The larvae with the most dramatic features were persistently immobile and ventrally curved but peristaltic waves could still be observed emerging from the posterior part of the body (video 5).
Another group of larvae displayed bouts of ventral contraction interrupting otherwise seemingly normal crawling phases that were characterized by regular propagation of peristaltic waves along the body. In light of these observations we currently favour the hypothesis that TJ + cholinergic INs are part of a neuronal circuit controlling the body posture of the larva rather than being intrinsically involved in controlling the speed of locomotion.
Recently, Clark and collaborators (2016), while surveying Janelia Gal4 lines (32) crossed to UAS-TrpA1, identified from their screen several lines that gave rise to similar ventral contraction phenotypes. Interestingly, they also reported a spectrum of severity with some larvae continually blocked with tonically contracted ventral muscles, while others would go through periods of ventral contraction followed by attempts to crawl. From this screen three different lines specifically expressed in subsets of INs were identified and in the future it will be interesting to determine if these subsets include TJ + cholinergic INs.
Alternatively, it is possible that these INs subsets and TJ + cholinergic INs are different [15] components of the same circuit regulating ventral bending of the larva. In such a scenario we can foresee two major alternatives: 1) either these different IN subsets are linearly and sequentially activated, resulting in the contraction of the entire ventral muscle field via the activation of ISNb, ISNd and SNc MNs or 2) each IN subset is selectively and independently used as a premotor excitatory command allowing specific ventral groups of muscles to be activated via only one MN subpopulation (ISNb or ISNd or SNc). Such precise coordination of muscles contraction within a muscle field is exemplified by the recently described activity of an inhibitory IN denoted iIN. iIN specifically innervates "transverse" MNs and not "longitudinal" MNs, thus allowing for the sequential contraction of transverse and longitudinal muscles for an efficient contraction of the larval body segment (15).
Our study on the functional role of TJ + glutamatergic INs is hindered by the fact that when using vGlut-LexA both TJ + MNs and TJ + INs are targeted; to our knowledge no genetic tools exist that would allow us to specifically activate TJ + glutamatergic INs. We nevertheless investigated the implication of TJ + MNs using two non-optimal drivers (CQ2-LexA and RapGAP1-LexA (data not shown)) and found that TJ + MNs activation leads to a defect in locomotion, i.e reduction of the number of peristaltic waves. Similarly, when we used per-LexA to activate the entire TJ + PMSI population (6 TJ + glutamatergic INs and 3 TJ + GABAergic INs per segment) we noticed a reduction of the speed of locomotion and a partial relaxed paralysis of the posterior segments of the larvae. Interestingly, in both cases these genetic manipulations did not give rise to the severe spastic paralysis phenotype we observed while activating the entire TJ + glutamatergic population. It could be argued that activation of the TJ + PMSI GABAergic INs when using per-LexA is in some way "dominant" over the activation of TJ + PMSI glutamatergic INs, and thus, it will be informative to only activate TJ + PMSI glutamatergic INs. Unfortunately a vGlut AD driver, an equivalent of the GAD1 AD line, has not been generated, thus precluding the implementation of this strategy at the present time.
Nevertheless, it would be of interest to solely activate all Per + glutamatergic INs (and thus not the Per + TJ + GABAergic INs); this could theoretically be achieved using the following transgenes: lexAop>>dTrpA1, vGlut-LexA, per-Gal4 and an UAS-Flp (33). Since we reported [16] strong expression of per-Gal4 in the vast majority of Per + glutamatergic neurons, we expect recombination events in the targeted neurons to be efficient and thus assume that this experiment will be conclusive.

TJ + GABAergic neurons regulate the speed of locomotion
Activation of TJ + GABAergic neurons gave rise to an apparently normal, though slowed locomotion, with a number of peristaltic waves accomplished by larvae placed at 31°C similar to the number at 23°C. Since larvae tend to crawl faster when the temperature exceeds their 24°C comfort temperature (in young L3) (34), we wondered if TJ + GABAergic neurons could be modulator component(s) of the temperature sensing system that detects uncomfortable temperatures and induces acceleration as a way to escape. TrpA1 + neurons located in the larval brain have been recently reported to be sensitive to the speed of the temperature increase (35) and we have found that these neurons do not express TJ (data not shown). The possibility that TJ + GABAergic neurons are nonetheless part of this temperature sensing neuronal circuit awaits future experimental analysis.
Further subdivision within the TJ + GABAergic INs pool, using a triple intersectional genetics approach, revealed that 3 Per + /TJ + GABAergic INs located at the midline and known as MNB progeny neurons substantially impact the crawling speed of the larvae. It thus appears that this Per + GABAergic population was overlooked in the previous characterization of PMSIs. This might be due to the fact that of the 20 Per + INs present in each segment, only 3 are indeed GABAergic and these express low levels of period, especially in third instar larvae, a stage in which the characterization PMSIs was originally carried out (5).

Molecular characterization of the Per + /TJ + GABAergic (MNB progeny neurons); searching for equivalents throughout evolution
Cell body position and molecular characterization of the Per + TJ + GABAergic neurons allowed us to identify them as a subset of the MNB progeny among the midline cells.
Although development of the midline cells has been meticulously described (24)(25)(26), the functional implication of these cells in locomotion or other behaviours is comparatively poor (36,37). Here we show for the first time that MNB progeny neurons have a relevant function in the locomotor behaviour of the larva. Moreover, our studies add TJ as a new marker of [17] the MNB progeny, offering the possibility to identify and genetically manipulate these neurons with TJ-Gal4 or TJ-Flp.  Fig.6Q). In the annelid Platynereis dumerilii, a recent study focusing on the molecular characterization of neuron types brought to light a group of neurons specifically co-expressing the TFs Gata1/2/3 and Tal that may be related to CSF-cN (41) indicating that the molecular nature and physiological function of this neuronal type may have been conserved during evolution.

Our detailed molecular characterization of the
The remarkable similarities of combinatorial expression of TFs within this IN class further exemplifies that the molecular mechanisms used during the wiring of the locomotor system are conserved and evolutionarily ancient. [18]   On larva VNC and muscle wall (for the study of MN projections)

Antibody list
For first instar larval (stL1) dissection embryos were preselected during the time when their main dorsal tracheae begin to fill with air, which represents 18 h after egg laying (AEL) and allowed to develop for further 3 h. First instar larvae (21 h AEL) were dissected as described in (53). Third instar larvae were heat-killed at 56°C for 5 sec and dissected as previously described in (54).

On embryo VNC
Immunolabeling of embryos was carried out as previously described (55).
Image acquisition and processing [23] Images were acquired on a Zeiss LSM700 confocal with 40x or 63x objectives, treated and cropped in Photoshop (Adobe) and assembled in Illustrator (Adobe). For the benefit of colour-blind readers, double-labelled images were falsely coloured in Photoshop. 3D projection of whole VNC was implemented using Zen software (Zeiss).

Locomotion assay
Larvae sorting For first instar larvae testing, decorionated late stage 17 embryos of the right genotype were sorted out and placed on an agar/grape juice plate supplemented with some feeding medium (maize, sucrose and yeast). Hatching time was monitored and larvae locomotion assessed 6 hours after hatching.
For third instar larvae testing, eggs were laid for 5 hours on basic maize feeding medium.
Approximately 72 hours later burrowing third instar larvae were picked up and assessed for locomotion.

Assay
Larvae clean of food were gently picked up with tweezers (in the case of the third instar larvae) or the back of tweezers (for first instar larvae) and placed on a 56 mm-agar plate supplemented with grape juice. After a 30-seconds acclimation period, the number of peristaltic waves done by the larvae was manually counted (at that time plate surface temperature was 23°C). The plate was transferred on a hot plate to heat for 2 minutes and 30 seconds or until it reached 31°C. The plate was quickly removed from the heat and the number of peristaltic waves done by the larvae manually assessed for 30 seconds more.
Plate was left to rest for 4 minutes until surface temperature reached 23°C. Number of peristaltic waves in 30 sec was assessed once more.

Statistical analysis
Statistical tests were carried out using Graphpad Prism (Graphpad software, Inc). We used one-way ANOVA with a Tukey post-hoc test to analyse more than two groups of data. When comparing only two groups we used unpaired Student t-test. [24] Figure 1 [32] temperatures. Upon activation of the TJ + within the VNC only (second red bar), we recapitulate the behaviour observed when the entire TJ + population is activated (presented in Fig.1I). Each single point represents a single 12h-old first instar larva. Error bars indicate the SD and n denotes the number of larvae tested. Statistical analysis: One way ANOVA. ***p<0.001.