ABSTRACT
Movement is the main output of the nervous system. It emerges during development to become a highly coordinated physiological process essential to the survival and adaptation of the organism to the environment. Similar movements can be observed in morphologically-distinct developmental stages of an organism, but it is currently unclear whether these movements have a common or diverse molecular basis. Here we explore this problem in Drosophila focusing on the roles played by the microRNA (miRNA) locus miR-iab4/8 which was previously shown to be essential for the fruit fly larva to correct its orientation if turned upside down (self-righting) (Picao-Osorio et al., 2015). Our study shows that miR-iab4 is required for normal self-righting across all three Drosophila larval stages. Unexpectedly, we also discover that this miRNA is essential for normal self-righting behaviour in the Drosophila adult, an organism with radically different morphological and neural constitution. Through the combination of gene-expression, optical imaging and quantitative behavioural approaches we provide evidence that miR-iab4 exerts its effects on adult self-righting behaviour through repression of the Hox gene Ultrabithorax (Ubx) (Morgan, 1923; Sánchez-Herrero et al., 1985) in a specific set of motor neurons that innervate the adult Drosophila leg. Our results show that this miRNA-Hox module affects the function, rather than the morphology of motor neurons and indicate that post-developmental changes in Hox gene expression can modulate behavioural outputs in the adult. Altogether our work reveals that a common miRNA-Hox genetic module can control complex movement in morphologically-distinct organisms and describes a novel post-developmental role of the Hox genes in adult neural function.
INTRODUCTION
Movement, the main output of the nervous system, first emerges during embryonic development. Although in its initial embryonic manifestation movement typically appears highly uncoordinated, as development proceeds, movement turns into an exceedingly coordinated physiological process (Landmesser and O’Donovan, 1984; Suzue, 1996; Saint-Amant and Drapeau, 1998; Crisp, 2008) that ultimately enables the fully formed animal to feed, escape from predators or find a suitable partner to mate. As such, adequate movement control is key to the animal’s adaptation to the environment and represents an essential intrinsic attribute fundamental to organismal survival and evolution. A point worth noting is that animals as distinct as insects and mammals all have a preferred orientation in respect for the surface so that their grip is maximised to facilitate motion.
To what extent does the genetic make-up of the organism influence the control of its movements? In principle, genetic mutations could affect the control of movement in two fundamentally distinct ways: they could impair the developmental formation of the networks underlying movement control, or, instead, interfere with the function of the cellular components involved in the physiological regulation of movement. A priori, these two levels of action of the genetic system need not be mutually exclusive.
A key experimental system to study the effects of genes on movement control is the fruit fly Drosophila melanogaster. Here, following the behavioural genetics approach pioneered by Seymour Benzer and his colleagues (Benzer, 1967; Hotta and Benzer, 1972), it became possible to isolate several genes with associated roles in movement control, including the Zinc-finger transcriptional co-repressor gene scribbler (Yang et al., 2000), the cGMP-dependent protein kinase gene foraging (de Belle et al., 1989; Osborne et al., 1997), the Ig superfamily gene turtle (Bodily et al., 2001), the phosphatidic acid transporter gene slowmo (Carhan et al., 2004) and other genes, such as pokey (Shaver et al., 2000) whose molecular functions have not yet been established. Of note is the case of the Hox genes, which encode a family of transcription factors key for the correct development of body structures along the main body axis (Lewis, 1978; McGinnis and Krumlauf, 1992; Alonso, 2002; Mallo and Alonso, 2013), and whose function has been shown to be required for the correct development of the neuromuscular networks underlying larval crawling (Dixit et al., 2008).
Nonetheless, much of the genetic dissection of movement control has so far focused on so-called protein-coding genes. Recent work in our laboratory showed that a single non-coding RNA, the microRNA miR-iab4, can affect the complex motor sequence that allows the young fruit fly larva to rectify its orientation if turned upside down (self-righting, SR) (Picao-Osorio et al., 2015). SR is an adaptive innate response that ensures an adequate position of the organism in respect for the surface, and it is evolutionarily conserved all the way from insects to mammals (Ashe, 1970; Penn, and Brockmann, 1995; Faisal. and Matheson, 2001; Jusufi, et al., 2011), including humans.
microRNAs (miRNAs) repress gene expression by blocking protein translation or promoting target mRNA degradation of their targets (Bartel, 2018). Our previous work showed that miR-iab4 affects larval movement through regulation of one of its molecular targets, the Hox gene Ultrabithorax (Ubx), whose level of expression in a set of metameric motor neurons is critical for normal SR behaviour (Picao-Osorio et al., 2015). To explore the generality of the effects of miRNA regulation on larval SR movement we recently conducted a genetic screen that revealed that at least 40% of all miRNAs expressed in young Drosophila larva can affect SR demonstrating an unprecedented and widespread role of miRNA regulation in the control of postural adjustments and locomotor behaviour (Picao-Osorio et al., 2017).
Despite this progress, it is currently unclear whether similar (adaptive) movements performed by morphologically distinct organisms rely on common or different genetic operators. Here we investigate this problem looking at the effects of the miR-iab4/Ubx system on a series of distinct developmental stages of the fruit fly including the larvae and adults: organisms with substantially different somatic and neural anatomies, behavioural repertoire and life style.
Through the combination of gene expression, optical imaging and behavioural analyses we show that a single genetic module composed of the miRNA miR-iab4 and the Hox gene Ubx contributes to the SR response in both Drosophila larvae and adults. Our study also reveals a novel neural role of the Hox genes in the fully formed adult, suggesting that these key developmental genes also perform biological functions once development has ceased.
RESULTS AND DISCUSSION
Our previous work in the young, first instar Drosophila larvae showed that ablation of the miR-iab4/8 locus (Bender, 2008) leads to significant defects in the SR response (Picao-Osorio et al., 2015). To investigate whether miR-iab4/8-dependent effects were confined to the L1 stage or had impact throughout larval development we conducted a series of SR tests in first, second and third instar larvae (L1, L2 and L3 larvae, respectively) (Figure 1A-B). SR was assayed by briefly putting individuals upside-down and monitoring the time they took to come back to a normal right side up position (see Materials and Methods). miRNA mutants show an increased time for the completion of the SR sequence (Figure 1C), indicating both that activities derived from the miR-iab4/8 locus affect SR, and that this miRNA system is important for the normal timing of the SR response across all three larval stages.
Like in all holometabolous insects, the Drosophila life cycle involves the transformation of the larva into the adult through the process of metamorphosis (Truman and Riddiford, 1999). Given the substantial anatomical and functional remodelling that metamorphosis imposes on the Drosophila soma and nervous system, genetically-induced behavioural defects observed in the larvae may simply disappear in the adult stage. Remarkably, a modification of the SR test performed in the Drosophila adult (see Materials and Methods), reveals that integrity of the miR-iab4/8 locus is important for normal SR response also in the adult fly (Figure 1C and Figure S1). That is, a common miRNA system controls the same adaptive behaviour in two stages of the life of a fly, with radically different morphological and neural constitution.
Importantly, analysis of free-walking behaviour in adult flies (see Methods) show that the mutation of the miR-iab4/8 locus does not impair locomotion in adult flies (Figure S2). This indicates that the absence of the miR-iab4/8 system does not lead to a generalised motor deficient phenotype. However, miRNA mutant flies tend to walk less than controls, and previous work (Bender, 2008) showed that the miR-iab4/8 mutants displayed posture control defects during mating, suggesting that miR-iab4/8 may function in other posture control systems in addition to SR. Alternatively, these different posture control systems may share some of the same neural substrate upon which miR-iab4/8 exerts its biological role.
The miR-iab4/8 locus encodes two distinct miRNA molecules: miR-iab4 (Ronshaugen et al., 2005) and miR-iab8 (Bender, 2008; Tyler et al., 2008; Stark et al., 2008), each produced from pri-miRNA transcription of opposite DNA strands. To tease apart the individual contributions of miR-iab4 and miR-iab8 towards adult SR we performed a series of genetic tests in adults using a collection of chromosomal variants that specifically disabled miR-iab4 or miR-iab8 placing them in combination with the miR-iab4/8 mutation (DmiR) and determined that miR-iab4 (and not miR-iab8) is responsible for the effects on the adult SR response (Figure S3).
Behavioural observation of the SR routine in the adult shows that legs perform a key role during the SR response (Figure 2A), allowing the animal to swiftly grab the substrate and use this point of contact to flip its body into the right side up position. The halteres, important mechanosensory organs that control body manoeuvres in flight (Nalbach, 1993), may also contribute to the control of body manoeuvres underlying the SR response. However, we found that flies with ablated halteres showed no effect in the time to complete the SR response as compare to controls (Figure S1C). Next, we asked which pair of legs derived from the pro-(T1), meso- (T2), or meta- (T3) thoracic segments contributed to the control of SR. We performed a series of ablation experiments in which we surgically removed T1-, T2-or T3-legs from wild type individuals and assessed their performance in SR tests (Figure 2B). These experiments showed that the activities of T1- and T3-legs contribute to normal SR, whereas removal of T2-legs had no detectable effects.
We then sought to establish whether leg movement showed any anomalies in miR-iab4/8 mutant flies when compared with wild type specimens. Quantification of leg movement in immobilised upside-down flies showed that in DmiR mutant flies legs displayed a reduction in activity levels compared to those observed in WT flies (Figure 2D-E,). This observation suggests that the impact of miR-iab4 on the SR response is mediated –at least in part– through effects in the levels of activity of the legs (Figure 2F-G).
To explore the molecular basis underlying miR-iab4 effects on adult SR we considered the hypothesis that miR-iab4 exerts its effects on adult SR via the same molecular target established in the larva, the Hox gene Ubx (Picao-Osorio et al., 2015) (Figure 3A). In addition, Ubx plays a key developmental role in allocating morphological specificity to one of the thoracic ganglia (T3) (Lewis, 1978; Mallo and Alonso, 2013), including effects on detailed patterning of the T3-leg (Rozowski and Akam, 2002). To test the model that miR-iab4 represses Ubx to allow for normal adult SR response, we increased the expression levels of Ubx within its natural transcriptional domain in normal flies seeking to emulate the de-repression effects caused by miRNA removal. The results of this experiment (Figure 3B) show that an increase of Ubx levels phenocopies the effects of the miR-iab4/8 mutation on adult SR response, suggesting that the expression levels of Ubx are important for normal behavior.
Taking into consideration: (i) that mutation of the miR-iab4/8 locus disrupts the SR response; (ii) that legs play a key role in the SR sequence, and (iii) that modulation in the levels of the miR-iab4 target Ubx within its transcriptional domain had significant impact on SR, we decided to explore the cellular basis underlying SR control by testing the model that modulation of Ubx in leg motor neurons – the direct effectors of leg activity – may play a role in the adult SR response. For this, we artificially upregulated Ubx in different neuronal assemblies known to innervate the Drosophila leg (Bierley, 2012; Baek and Mann, 2009; Lacin and Truman, 2016) using the available lineage specific leg motor neuron GAL4 drivers VT006878-GAL4 (NB2-3/lin15) and R24C10-Gal4 (NB5-7/lin20). These experiments showed that upregulation of Ubx within the domain demarked by the VT006878-Gal4 (Lacin and Truman, 2016) was sufficient to cause an increase in the time that individual flies take to complete the SR response (Figure 3C and Figure S5), whereas the other motor neuronal drivers had no effect. In addition, inactivation of VT006878 neurons (Figure 3D-E) through expression of a temperature-sensitive allele of shibire (Kitamoto, 2001) has pervasive effects on the timing of the SR response (Figure 3F), highlighting the contribution of these motor neurons to the normal SR response.
We noted that the VT006878 line seems to be transcriptionally active in wing and haltere sensory axons (Figure 3D), making it plausible that wings and/or halteres may play a role in adult SR. However, different manipulations indicated this might not be the case. First, all adult flies had surgically removed wings in our SR behavioural paradigm, making it unlikely that these appendages are key contributors to this behaviour. Second, ablation of halteres resulted in no apparent effect in the time wild type flies took to complete a SR response (Figure S1C). Altogether, these results suggest that the effect observed by the overexpression of Ubx in the VT006878-Gal4 line cannot be accounted by an effect derived from the haltere/wing sensory neurons.
Upregulation of Ubx using VT006878-Gal4 is expected to increase Ubx levels in all three thoracic segments (T1-T3) and scattered neurons in the brain (Lacin and Truman, 2016), making it unclear whether ectopic Ubx expression in the brain might be the cause of the SR defects observed in treated adults. To constrain the expression pattern of VT006878>Ubx to the brain only, we used the ventral nerve cord (VNC) specific tool teashirt-Gal80ts. Upregulation of Ubx within circuits in the brain has no effect on the timing of SR in adults (Figure S4) indicating that: (i) an increase of Ubx within the VT006878 domain in the brain is insufficient to cause an adult SR phenotype; and (ii) Ubx upregulation within the thoracic VT006878 domain is indeed responsible for the triggering of SR defects in the adult.
Ubx protein is detected in subsets of neurons within the T1-T3 ganglia, with a larger population observed within the T3 segment of the VNC (Figure 4A-B). The RNA in situ hybridisations show that miR-iab4 is highly expressed in the T3 ganglion of the VNC (Figure 4A and C, E-F) and that both Ubx and miR-iab4 are expressed within the VT006878 domain (Figure 4D and G). In miRNA mutants, Ubx expression is significantly increased in the T3 segment of the VNC, but not in T2 (Figure 4H-I and Figure S6), in agreement with the idea that increase of Ubx expression (de-repression) within the VT006878 domain in T3 leads to SR defects in the adult. A prediction that emerges from this idea is that artificial reduction of Ubx in DmiR mutants, specifically confined to the VT006878 domain should ameliorate (or even rescue) the SR phenotype observed in adult mutants. In line with this prediction, RNAi-mediated reduction of Ubx driven by VT006878-Gal4 rescues the SR phenotype in adult flies (Figure 3G). This experiment also indicates that levels of Ubx in the T3 segment are key for normal timing of adult SR.
Detailed anatomical examination of T3 VT006878 motor neurons (also known as ventral lineage 15 motor neurons (Lacin and Truman, 2016)) in wild type and DmiR specimens showed no detectable differences in axonal projections or morphologies (Figure 5A-C) suggesting that – as observed in the larva (Picao-Osorio et al., 2015) – the microRNA under study might have effects on neuronal function rather than on neuronal morphology. Indeed, quantification of varicosities (a known indicator of neuronal activity (Cox et al., 2000; Petreanu et al., 2012) at the junction of VT006878 neurons with the muscle system of the third leg reveals a statistically-significant reduction in varicosities in the DmiR samples (Figure 5D-E) in line with the model that absence of the miRNA leads to diminished levels of neural activity. Remarkably, in DmiR mutants, RNAi-mediated reduction of Ubx expression in the VT006878 neurons rescues the normal number of varicosities strongly indicating a role of Ubx in the formation of active contact points between the neuronal and muscle systems. Furthermore, multiphoton microscopy analysis of genetically-encoded calcium reporters (GCaMP6m) specifically expressed in the VT006878 motor neurons shows an overall reduction of spontaneous neural activity in DmiR samples in T3 when compared to wild type (Figure 5F-G). Altogether, our data suggests that miR-iab4 represses Ubx within the VT006878 motorneuron domain in T3 ensuring the normal neural functions that underlies the adult SR response.
Lastly, we sought to determine whether the effects of the miRNA on adult SR behaviour emerge from a progressive developmental function of the miRNA, or rather, are the consequence of the activity of the miRNA on the physiology of the VT006878 motor neurons in the adult. For this we performed a conditional expression experiment in which we maintained normal expression of Ubx in the VT006878 domain during the full developmental process that spans from embryo to adulthood, increasing Ubx expression only after adult eclosion (Figure 6A). Our data shows that an increase in Ubx expression, exclusively delivered in the adult, is sufficient to induce SR defects (Figure 6B and Figure S7) revealing a novel post-developmental role of Hox genes in the control of neural function in the fully formed organism.
Our work reveals that complex adaptive movements performed by organisms with distinct morphologies and neural anatomies can rely on a simple genetic module involving a miRNA and a Hox gene. We also describe what is –to our best knowledge– the first case of post-developmental roles of the Hox genes with impact on adult behaviour. Based on the pervasive evolutionary conservation on the Hox gene system and the key roles played by these genes in the nervous systems of animals as different as insects and mammals, we propose that similarly simple genetic modules including miRNAs and Hox targets may conform part of the molecular circuitry underlying movement control in other species, including humans.
COMPETING INTERESTS
The authors declare no competing interests.
MATERIALS AND METHODS
Drosophila Culture and Strains
Fly stocks and crosses were raised at 25°C on standard corn meal/yeast/agar medium, under a 12 h/12 hr light/dark cycle. The following fly strains were used: VT006878-Gal4 (Vienna Drosophila Resource Center, ID200694); R54F03-Gal4 (BDSC, #39078); R24C10-Gal4 (BDSC, #49075) (Lacin and Truman, 2016); ΔmiR-iab4/iab8 (Bender, 2008); iab-3277; iab-5105 and iab-7MX2 (Karch et al., 1985); UbxM3-Gal4 (De Navas et al., 2006); UAS-UbxIa (Reed et al., 2010); UAS-UbxRNAi (BDSC, #31913); UAS-Myr::GFP (Pfeiffer et al., 2010); UAS-Nls::GFP(BDSC, #4775); UAS-GCaMP6m (Chen et al., 2013); UAS-shits (Kitamoto, 2001).
Self-Righting tests
Larval SR behaviour was assayed as previously described (Picao-Osorio et al., 2015; Picao-Osorio et al., 2017). For adult SR behaviour tests, flies were grown in non-crowed conditions at 25°C. The day before the SR test, the wings of cold-anesthetised 2-to-4-day old flies were surgically removed (clipped). Flies recovered for one day at 25°C. Flies were assayed for SR behaviour by being introduced individually into an arena and rolled over with a brush to an “upside-down” position (“legs up”) and the time taken by the fly to return to its normal position (“right-side up”) was recorded. A maximum of ten minutes was given to the fly to SR. All experiments were done with flies 4-6 days after eclosion and tested at 25°C. For silencing of VT006878 neurons, 3-to 4-day-old flies expressing shits1, were incubated for 10 min at 31°C, or at 18°C for controls, just before the SR test. SR behaviour was assessed within seconds (50±10) after incubation. Similar results to those observed using this procedure were obtained when measuring the time to SR of adult flies with intact wings after recovery from CO2-induced or cold-induced anaesthesia (Figure S1). The absence of halteres showed no effects on SR (Figure S1).
Walking behaviour
Locomotion was assessed by placing single flies (males or females) on a circular arena with sloped edges (Simon and Dickinson, 2010), and their spontaneous walking behaviour was recorded from the top at 60 Hz with a Flea FL3-U3-32S2M Point Grey camera with a M1214-MP2 lens (Computar) for 15 minutes. To automatically track the position of flies in the arena, we used Ctrax (Branson et al., 2009). Walking bouts were defined as segments in time when the body moved through space with speed of at least 5 mm/s for (at least) 500ms. As a measurement of locomotion performance, we calculated the straightness of a walking bout. Straightness was defined as the mean angular deviation from a line defined between the start and end points of the segment. A straightness of 0 indicated walking along a perfect straight line. Straightness greater than 0 indicated curvilinear trajectories, and the greater the value, the more prominent the deviations were from a straight course.
Quantification of leg movements
Flies 3-5 days post eclosion, were cold-anesthetised and its thorax tethered to a glass microscope slides with UV-activated glue (BONDIC®). We next removed the tibia of mid leg to avoid interference of this leg in the quantification of first and third leg movements (this leg had no apparent effect on the timing of the SR response, Figure 2B). The leg movement videos were captured at 200 Hz with a monochrome digital camera (Bonito CL-400B, Allied Vision, with a M1214-MP2 lens and EX2C extender from Computar). We used a custom-made MATLAB script to quantify leg activity levels. Regions of interests (ROI) for analysis were automatically drawn based on the centre of mass (CM) of the thorax of the fly. Video images were converted into binary values using a threshold, and a time averaged image was calculated. Because the thorax of the fly was glued to a cover slip it was the only part of the fly that remained stationary throughout the video. To isolate pixels corresponding to the fly thorax, we identified those that did not change in intensity for more than 90% of the video. Pixels that did were converted to a background-related pixel. Next the thorax of the fly and its CM, were extracted using the connected components method. From the CM, we automatically defined two regions of interest (ROIs), one on each side of the fly that were separated by the width of the fly thorax, and with 459×100 pixel size. For each pixel inside of these ROIs, we extracted the pixel intensity (in A.U) and calculated the change in pixel intensity as a function of time. Leg activity per pixel was classified as 1 if the instantaneous change in pixel intensity was at least 10 pixels per time step, which corresponded to approximated 20%of the total change in pixel intensity. From this, we averaged the change in pixel intensity over the course of the experiment and generated a mean heat map (over all flies) for WT and miRiab4/8 flies (Figure 2D). To quantify the average leg activity for each fly (Figure 2E), we calculated the mean response of both the Left and Right ROIs, normalized by the area of each ROI.
Adult leg preparation and mounting
Tissue dissection and mounting were performed as described (Enriquez et al., 2016). Fly legs were dissected with forceps in 0.3% triton in 1x phosphate buffered saline (PBS). Adult legs attached to thoracic segments were fixed in 4% formaldehyde in PBS overnight at 4°C followed by five washes in PBTx for 20 minutes at room temperature. legs were mounted onto glass slides using 70% glycerol medium.
Immunohistochemistry and RNA in situ hybridisation
Adult brains were dissected in 1X PBS. Tissues were then fixed for 1h in 4% formaldehyde in 1X PBS at room temperature. After fixation, brains were washed 3 times (30 min per washing) in PBS with 0.3% Triton-X-100 (PBTx) and incubated at 4°C overnight in primary antibodies. The following primary antibodies were used: mouse monoclonal anti-Ubx (FP3.38 (White and Wilcox, 1985) 1:500 from the Developmental Studies Hybridoma Bank) and chicken anti-GFP (Abacam Probes, 1:3000). The secondary antibodies were anti-mouse Alexa Fluor 555 (Invitrogen Molecular Probes, 1:1000) and anti-chicken Alexa Fluor 488 (Invitrogen Molecular Probes, 1:1000). RNA in situ hybridisation in adult ventral nerve cords for the precursor RNA transcripts of miR-iab-4 was performed by designing 48 unique 20nt-probes labelled with Quasar 570 in the Stellaris platform from Biosearch Technologies, and using an adapted version of the protocol by Raj A. and Tayagi S., 2010. Images were acquired with a Leica SP8 confocal microscope, processed and analysed using FIJI Image J. The VT006878 nerve-ending varicosities were quantified by measuring the puncta they covered in VT006878-labelled by myrGFP in leg or VNC.
Two-photon calcium imaging
To prepare flies for in vivo imaging in the ventral nerve cord (VNC) (Figure 5F) we adapted existing methods (Chen et al., 2018; Seeholzer et al., 2018). In brief, a single fly (3–5 days old) was cold-anaesthetised after eclosion and tethered using UV-curable glue to a piece of aluminium foil that covered a hole in the bottom of a modified polystyrene weighing dish. The fly’s body was positioned such that the dorsal side of the thorax covered the small hole made in the centre of the aluminium foil. The dish was then held by blue-tack on a glass microscope slide with ventral side and legs facing the slide. Next, the dish was filled with saline solution and a small hole in the thorax was opened by removing the cuticle covering the T3 ganglion using sharp forceps to avoid damaging nerves. The preparation was positioned under the two-photon microscope (see details below) and spontaneous GCaMP6m activity within VT006878 neurons was recorded. Composition of saline solution was as used previously (Seeholzer et al., 2018): 108 mM NaCl, 5 mM KCl, 2 mM CaCL2, 8.2 mM MgCL2, 4 mM NaHCO3, 1 mM NaH2PO4, 5 mM trehalose, 10 mM sucrose, 5 mM HEPES pH 7.5. All imaging experiments were performed on a MOM-type two-photon microscope (designed by W. Denk, MPI, Martinsried; purchased from Sutter Instruments/Science Products) equipped with a mode-locked Ti:Sapphire, Chameleon Vision-S laser set at 927nm. Emitted fluorescence was detected with F48×573, AHF/Chroma, and a water immersion objective 20x/1,0 DIC M27 Zeiss was used for image acquisition. For image collection we used custom-made software running under IGOR pro 6.3 for Windows (Wavemetrics) (Zimmermann et al., 2018), at 64 × 32 pixel resolution with 15.625 frames per second image sequences for activity scans or 512 × 512 pixel images for high-resolution morphology scans.
Statistical analysis
Statistical analyses were performed with GraphPad Software Prism using Mann-Whitney U test or one-way ANOVA with the post hoc Tukey-Kramer test. Error bars in figures represent SEM. Significant values in all figures: *p < 0.05, **p < 0.01, ***p < 0.001.
ACKNOWLEDGEMENTS
We thank members of the Alonso lab for helpful discussions and comments. We wish to thank Filip Janiak and Tom Baden (University of Sussex) for their assistance with multiphoton microscopy experiments and Clare Hancock for preliminary observations in this project. This research was funded by a Wellcome Trust Investigator Award made to C.R.A. (098410/Z/12/Z).