Article | Published:

Toll-6 and Toll-7 function as neurotrophin receptors in the Drosophila melanogaster CNS

Nature Neuroscience volume 16, pages 12481256 (2013) | Download Citation


Neurotrophin receptors corresponding to vertebrate Trk, p75NTR or Sortilin have not been identified in Drosophila, thus it is unknown how neurotrophism may be implemented in insects. Two Drosophila neurotrophins, DNT1 and DNT2, have nervous system functions, but their receptors are unknown. The Toll receptor superfamily has ancient evolutionary origins and a universal function in innate immunity. Here we show that Toll paralogs unrelated to the mammalian neurotrophin receptors function as neurotrophin receptors in fruit flies. Toll-6 and Toll-7 are expressed in the CNS throughout development and regulate locomotion, motor axon targeting and neuronal survival. DNT1 (also known as NT1 and spz2) and DNT2 (also known as NT2 and spz5) interact genetically with Toll-6 and Toll-7, and DNT1 and DNT2 bind to Toll-6 and Toll-7 promiscuously and are distributed in vivo in domains complementary to or overlapping with those of Toll-6 and Toll-7. We conclude that in fruit flies, Tolls are not only involved in development and immunity but also in neurotrophism, revealing an unforeseen relationship between the neurotrophin and Toll protein families.


The Toll receptor superfamily, comprising Toll and Toll-like receptors (TLRs), has ancient evolutionary origins, arising over 700 million years ago, and is present throughout metazoans1. Toll and TLRs have a universal function in innate immunity, and they initiate adaptive responses in vertebrates1,2. In humans the ten TLRs are pattern recognition receptors that directly bind to microbial antigens and activate proinflammatory and co-stimulatory responses. Mammalian TLRs were identified by homology to Drosophila Toll (Toll-1). The Drosophila genome contains nine Toll receptor genes (Toll-1 to Toll-9), which, except for Toll-9, are phylogenetically distinct from the vertebrate TLRs1. Thus, Drosophila Toll-1 to Toll-8 form one clade and Toll-9 together with vertebrate TLRs form another3. Toll-1 functions in developmental processes, including the establishment of the embryonic dorso-ventral axis, in axon targeting and degeneration, and in innate immunity1,4, but the roles of the remaining Tolls are largely unresolved. Reports have indicated that Toll-7 to Toll-9 have developmental functions but no antibacterial immunity functions, although Toll-7 is involved in antiviral responses5,6,7,8,9 and Toll-6 and Toll-7 are expressed in the CNS10. Unlike the TLRs, Toll-1 does not bind microbial products directly. Instead, detection of bacterial molecules by the soluble recognition proteins PGRP and GNBP triggers a serine protease cascade11. This leads to the cleavage and activation of Spätzle (Spz), an endogenous protein ligand for Toll-1 (ref. 12).

Spz belongs to the neurotrophin family of growth factors, which in vertebrates comprises nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT3) and neurotrophin-4 (NT4) (refs. 13,14). Spz comprises a signal peptide, an unstructured pro-domain and an active cystine knot domain of 13 kDa (also known C-106), which dimerizes, binding Toll with 2:2 stoichiometry13,14,15. Spz is secreted as a pro-protein and is cleaved extracellularly by the serine proteases Easter, acting in development, and Spätzle Processing Enzyme (SPE), acting in immunity, to release the active cystine knot13. This mechanism resembles the extracellular cleavage of BDNF at the synaptic cleft by the serine protease plasmin (which is also involved in the blood-clotting cascade) and which is activated by the presynaptic release of plasminogen activating factor (tPA) upon high frequency stimulation16. The characteristic neurotrophin cystine knot, formed by antiparallel β-sheets held together by three intersecting disulfide bonds, can be precisely aligned between the crystal structures of Spz and NGF17,18,19.

DNT1 was identified independently as related to BDNF, using vertebrate neurotrophin sequences as query to search the Drosophila sequenced genome with bioinformatics tools20. DNT1 was found to be spz2, a paralog of spz (refs. 20,21). Structural prediction analysis showed that, of the spz paralogs, DNT1 and DNT2 (spz5)21 are closest to the neurotrophin superfamily, followed by spz (refs. 20,21).

There is also functional conservation between DNT1, DNT2 and Spz and the mammalian neurotrophins in the nervous system20. The vertebrate neurotrophins have essential functions during development in neuronal survival, axon targeting and connectivity and during adult life in learning, memory and cognition16. During development, DNT1, DNT2 and spz are expressed in target cells for CNS neurons, such as the embryonic en-passant midline target of interneurons and the muscles, the target of motor neurons20. DNT1 and DNT2 are required for neuronal survival, as neuronal apoptosis decreases upon their overexpression in the CNS and increases in the loss-of-function mutants, leading to neuronal loss, with apoptotic neurons comprising those identified as bearing the Even-skipped (Eve) or Homeobox 9 (HB9) neuronal markers20. DNTs are required for motor-axon targeting, as interfering with the function of DNT1, DNT2 and Spz causes misrouting, mistargeting and sprouting defects in motor axon terminals20. Thus, DNT1 and DNT2, as well as Spz, are Drosophila neurotrophins on the basis of sequence, structural and functional homology to the vertebrate neurotrophins20.

There is further cellular and molecular evidence that neurotrophism operates in the Drosophila nervous system. During normal Drosophila development, many neurons and glial cells die22,23,24, and ablation or mutation in glial cells results in neuronal death in several contexts25. Identified Drosophila neurotrophic factors include the homolog of mesencephalic astrocyte-derived neurotrophic factor (MANF), which promotes dopaminergic neuron survival in fruit flies using a noncanonical pathway26, and Netrin, which promotes interneuron survival from the en-passant midline target27. Gliotrophic factors of the transforming growth factor (TGF)-α, neuregulin and PVF/platelet-derived growth factor (PGDF) protein families have also been shown to maintain glial survival in Drosophila22,28,29,30.

The mammalian neurotrophins signal through three distinct receptor types—p75NTR, Trk and Sortilin—and share a downstream target, the activation of NF-κB31,32,33. In Drosophila there are no canonical homologs of these receptors. The receptors for DNT1 and DNT2 are unknown, although one hypothesis is that orphan Tolls fulfill this function in insects. Toll receptors are generally thought to function by activating NF-κB signaling, which regulates the production of antimicrobial peptides in immunity1. Neurotrophins also function in immunity34, but these roles have been largely unexplored. TLRs are also present in the CNS, primarily in microglia, where they have immunity-related functions35. Thus, potential relationships between the Toll and neurotrophin families may have been overlooked. Here we ask whether Toll-6 and Toll-7 can function as receptors for DNT1 and DNT2 during CNS development.


Toll-6 and Toll-7 are expressed and required in the CNS

Toll-6 and Toll-7 mRNA are expressed in the embryonic and larval CNS and adult central brain (Fig. 1). To visualize protein distribution, we used a GFP exon trap insertion into Toll-6 (hereafter named Toll-6MIMICGFP), and we raised antibodies to Toll-7. The Toll-6MIMICGFP insertion is likely to result in a truncated Toll-6 protein and thus a mutant allele that could conceivably affect expression from the locus. However, we have no evidence that Toll-6 regulates its own expression. The Toll-7 antibodies were found to be specific, as they did not detect signal in Toll-7 null mutant embryos or in deficiency embryos lacking the Toll-7 locus (Supplementary Fig. 1a). The distribution of GFP in Toll-6MIMICGFP and of Toll-7 in wild-type matched the expression patterns of Toll-6 and Toll-7 transcripts, respectively (compare Fig. 1a,b,e,g,i,k with Fig. 1c,d,f,h,j,l). This indicates that the protein patterns most likely represent the endogenous distribution of the receptors. In Toll-6MIMICGFP, GFP is mostly cytoplasmic, whereas Toll-6 is localized to the membrane; therefore, the GFP does not reveal the subcellular distribution of Toll-6. Toll-6MIMICGFP and Toll-7 proteins are distributed in the CNS (Fig. 1c,d,f,h,j,l), including both interneurons and motor neurons (Fig. 1c,d), but we cannot rule out expression also in glia. Toll-6MIMICGFP is present in ventral embryonic HB9+ neurons (Fig. 2a) in all Eve+ motor neurons except raw prawn 2 (RP2) (Fig. 1c, Fig. 2 and Supplementary Fig. 2c) and in longitudinal interneuron axons (Fig. 2e). Toll-7 protein is distributed in ventral HB9+ and LIM-homeodomain-3 (Lim3)+ RP motor neurons (Fig. 1d, Fig. 2b,d and Supplementary Fig. 2d), in interneuron axons that cross the midline (Fig. 1d) and project along the three fasciclin II (FasII)+ longitudinal fascicles (Figs. 1d and 2f), and possibly in motor neuron dendrites or perhaps glia (Fig. 2d). Ventral HB9+ and Lim3+ motor neurons project along the intersegmental nerve b/d (ISNb/d), which targets muscles 6,7,12,13. The above distributions overlap with those of GAL4 reporters for Toll-6 (ref. 36) and Toll-7 (Supplementary Fig. 2a,b,e–i), which include expression in the ISNb/d axonal terminals (Fig. 2g). In the larva, immunoreactivity for GFP in the Toll-6MIMICGFP and immunoreactivity for Toll-7 were distributed along the ventral nerve cord (VNC) neuropil (Figs. 1f,h and 2h), and Toll-6MIMICGFP is detectable in anterior corner cell (aCC) motor neurons (Fig. 2h and Supplementary Fig. 2i). In the adult brain, Toll-6MIMICGFP is distributed in dopaminergic neurons (Fig. 2i). Toll-6MIMICGFP and anti–Toll-7 immunoreactivity are distributed in complementary layers of the fan-shaped body and in complementary rings of the ellipsoid body (Fig. 2j), the sites for the central control of locomotion. Thus, Toll-6 and Toll-7 are distributed in the locomotor circuit, including motor neurons, interneurons of the central pattern generator, and locomotion centers and dopaminergic neurons in the brain.

Figure 1: Toll-6 and Toll-7 are expressed in the CNS through all stages.
Figure 1

In situ hybridizations showing transcripts for Toll-6 and Toll-7 in (a,b) stage 13, 15 and 17 embryos; (e,g) larval optic lobes, central brain and ventral nerve cords; and (i,k) adult brain in central complex (arrows). (c,d,f,h,j,l) Distribution of GFP immunoreactivity in Toll-6MIMICGFP and Toll-7 immunoreactivity. (c) Arrows, GFP signal in distinct neuronal types. (d) White arrows in first and second images from left, motor axons exiting the CNS; yellow arrowheads, motor neuron cell bodies; white arrows in third image, axons crossing the midline; white arrows in fifth image, axons along three interneuron fascicles and in thickenings (yellow arrowhead) that might correspond to dendrites or glia. (i,k) Arrows, signal in and around fan-shaped body. ELs, U/CQ: Eve+ neurons. Anterior is up. Scale bars: a,c, 10 μm; f,h,j,l, 50 μm.

Figure 2: Identification of Toll-6 and Toll-7 cells in the locomotor circuit.
Figure 2

(a,c,e) Anti-GFP in Toll-6MIMICGFP embryos is distributed in ventral lateral nerve cord HB9+ neurons (a), Eve+ EL interneurons and all Eve+ motor neurons (MN) except RP2 (c,e, arrows; for RP2 neurons, see Supplementary Fig. 2c). IN, interneurons; arrowheads point to longitudinal connectives. (b,d,f) Toll-7 protein is localized to ventral lateral and medial HB9+ neurons (b, arrows), Lim3GAL4UASmyrRFP (Lim3>RFP) RFP+ RP motor neurons (d, arrows) and possibly dendrites (pink arrow), and to FasII+ interneuron fascicles (f, arrows). (g) tdTomato in D42GAL4UAS10xmyr-tdTomato (Toll-6>Tom) and GFP in w;Toll7GAL4/+;UASGAPGFP/+ (Toll-7>GFP) reveal ISNb/d terminals. (h) Toll-6MIMICGFP and Toll-7 are present in the larval VNC neuropil (arrows point to axons) and at least Toll-6MIMICGFP in motor neurons. (i) Toll-6MIMICGFP colocalizes with the dopamine precursor tyrosine hydroxylase (TH) in most dopaminergic neurons. (j) Toll-6MIMICGFP and Toll-7 are distributed in distinct layers of the fan-shaped body (fsb; arrows) and rings of the ellipsoid body (eb; arrows). Scale bars: ag, 10 μm, hj, 50 μm.

To investigate the functions of Toll-6 and Toll-7 in the CNS, we generated null mutant alleles (Supplementary Fig. 3). The embryonic motor neurons are preserved in the larva; thus, we tracked crawling mutant larvae. Most particularly, Toll-7P8/Toll-7P114;Toll-626/Toll-631 double mutants crawled more slowly than controls (Fig. 3a,b, P < 0.0001, corrected P < 0.001). To test whether Toll-6 and Toll-7 function in motor-axon targeting, we visualized the projections of the FasII+ ISNb/d. Toll-631/Df(3L)XG4, Toll-7P8/Toll-7P114 and Toll-7114/Df(2R)BSC22 single mutants and Toll-7P8/Toll-7P114;Toll-626/Toll-631 double mutants showed deficient targeting and axonal misrouting (Fig. 3, χ2(7) = 136.247, P < 0.001, corrected P < 0.001). Overexpression of constitutively active forms of the receptors, Toll-6CY and Toll-7CY, in neurons also caused targeting defects (Fig. 3d,e, χ2(7) = 136.247, P < 0.001; corrected P = 0.032 and P < 0.001, respectively). To test whether Toll-6 and Toll-7 can regulate cell survival, we visualized cell death with anti–cleaved-Caspase-3 antibodies and quantified the number of apoptotic cells using DeadEasy Caspase software37. Apoptosis increased in Toll-626/Toll-631, Toll-626/Df(3L)XG4, Toll-7P8/Toll-7P114 and Toll-7P8/Df(2R)BSC22 mutant embryos compared to that in wild-type controls, showing that Toll-6 and Toll-7 are required for cell survival in the CNS (Fig. 4a–c) (Fig. 4b: F(2,70) = 5.782, P = 0.005; corrected P = 0.006 and P = 0.015, respectively; Fig. 4c: F(2,71) = 7.010, P = 0.002; corrected, P = 0.001 and P = 0.032, respectively). Overexpression of Toll-6CY and Toll-7CY in neurons rescued naturally occurring cell death (Fig. 4d, ANOVA, F(2,68) = 4.811, P = 0.011; corrected P = 0.021, P = 0.012, respectively). Thus, Toll-6 and Toll-7 can promote cell survival. The increase in dying cells in the double mutants compared to wild-type controls affected HB9+Caspase+ neurons (Fig. 4e–h, Student t(26) = −2.230, P = 0.035) and, albeit not significantly, Eve+Caspase+ EL interneurons (Fig. 4i,j, χ2(1) = 1.992, P = 0.158), which normally express Toll-7 and Toll-6 (Fig. 2a–f). Not all HB9+ neurons normally express Toll-6 or Toll-7, and thus we could not confirm that all dying HB9+ neurons necessarily corresponded to Toll-6+ or Toll-7+ neurons. However, there was a good correlation between the ventral and central locations of the dying HB9+ neurons in the double mutants and the equivalent location of HB9+ Toll-6MIMICGFP-positive and HB9+Toll-7+ neurons in normal embryos (compare Fig. 4e with Fig. 2a, and Fig. 4f with Fig. 2b), indicating that, in the mutants, the HB9+ dying neurons most likely included Toll-6+ and Toll-7+ neurons. We were able to confirm that because all Eve+ neurons except RP2 were also Toll-6MIMICGFP positive, the apoptosis of Eve+ neurons in the mutants corresponded to cell death of at least Toll-6+ neurons. Apoptosis resulted in neuronal loss, as in Toll-7P8/Toll-7P114;Toll-626/Toll-631 doubles mutants there was a reduction in the number of Eve+ EL interneurons (Fig. 4k,l, χ2(1) = 9.645, P = 0.002). Altogether, our data show that Toll-6 and Toll-7 are required for locomotion, motor-axon targeting and neuronal survival.

Figure 3: Toll-6 and Toll-7 are required for larval locomotion and motor-axon targeting.
Figure 3

(a) Trajectories of larvae crawling for 400 frames per larva, n = 50 larvae per genotype. (b) Toll-7;Toll-6 double mutant larvae crawled more slowly. Kruskal-Wallis = 814, P < 0.0001, and Dunn's test for pair-wise comparisons; asterisks refer to double mutants versus yw controls (Dunn = 4,474), n = 50 larvae and 19,950 frames per genotype. (ce) The incidences of FasII+ motor axon misrouting in one or more projections and of loss of two or more projections per hemisegment increased in stage 17 mutant embryos (c,e) and embryos overexpressing activated forms of Toll-6 or Toll-7 in all neurons (elavGAL4>Toll-6CY;Toll-6CY and Toll-7GAL4;elavGAL4>Toll-7CY) (d,e). Scale bar, 10 μm. For e, chi-squared χ2(7) = 136.247, P < 0.001, pair-wise comparisons to yw chi-squared with Bonferroni correction, n = 169–465 hemisegments per genotype. **P < 0.01; ***P < 0.001. For details, see Supplementary Table 1.

Figure 4: Toll-6 and Toll-7 maintain neuronal survival.
Figure 4

(a) Embryonic VNCs labeled with anti–cleaved-Caspase-3. (b,c) Apoptosis increased in Toll-7 and Toll-6 mutant embryos, as quantified with DeadEasy software: one-way ANOVA F(2,70) = 5.782, P = 0.005; post hoc Dunnett P = 0.006, P = 0.015, respectively; n = 19–28 embryos per genotype (b); one-way ANOVA F(2,71) = 7.010, P = 0.002; post hoc Dunnett P = 0.001, P = 0.032, respectively; n = 21–31 embryos (c). (d) Pan-neuronal overexpression of activated Toll-6 and Toll-7 rescued naturally occurring cell death in the CNS; one-way ANOVA F(2,68) = 4.811, P = 0.011; post hoc Dunnett P = 0.021, P = 0.012, respectively; n = 22–27 embryos. In bd, asterisks refer to pair-wise comparisons to yw, post hoc Dunnett tests. (eh) Apoptotic Caspase+HB9+ cells in Toll-7P8/Toll-7P114; Toll-626/Toll-631 double mutant embryos in locations corresponding to neurons that normally express Toll-6 (e) or Toll-7 (f), high magnification view (g) and quantification (h); unpaired Student t-test t(1) = −2.230, P = 0.035, n = 9–19 embryos. (i,j) In Toll-7P8/Toll-7P114; Toll-626/Toll-631 double mutant embryos, more (albeit not significantly more) EL clusters had Eve+Caspase+ apoptotic interneurons compared to wild-type (yw) controls (j, χ2(1) = 1.992, P = 0.158, n = 109–138 EL clusters). (k,l) Apoptosis leads to loss of Eve+ EL interneurons in the double mutants. More EL clusters had fewer neurons than the normal 8–10 per cluster (arrowheads in k); chi-squared χ2(1) = 9.645, P = 0.002, n = 22–260 EL clusters. *P < 0.05; **P < 0.01; ***P < 0.001. All embryos stage 17. Scale bars: a, 20 μm; e,f,k, 10 μm; g,i, 5 μm. For details, see Supplementary Table 1.

Toll-6 and Toll-7 interact genetically with DNT2 and DNT1

The observed CNS phenotypes resemble those caused by DNT1 and DNT2 (ref. 20), which is consistent with DNTs and Tolls being involved in common developmental processes. We next asked whether DNT, Toll-6 and Toll-7 mutants might interact genetically. Single DNT1, DNT2, Toll-6 or Toll-7 mutants are viable. However, DNT1 DNT2 double mutation is semi-lethal in progeny of a heterozygous stock maintained over the TM6B chromosome at 18 °C (Fig. 5a, χ2(11) = 360.277, P &lt; 0.001, corrected P < 0.001). Similarly, Toll-7;Toll-6 double mutation is also embryonic semi-lethal at 18 °C in progeny of a heterozygous stock over SM6aTM6B (Fig. 5a, corrected P < 0.001). This semi-lethality can be rescued with the expression of constitutively active forms of the receptors in cholinergic interneurons, Toll-6ΔLRR, Toll-7ΔLRR and Toll-7CY (see Online Methods) (Fig. 5b, χ2(6) = 85.028, P < 0.001; corrected P < 0.001, P = 0.003, P < 0.001, respectively). Exploiting this cold-sensitive semi-lethality, we tested genetic interactions between the DNTs and the Tolls. DNT141Toll-626 and Toll-7P114;DNT2e03444 double mutation is semi-lethal under the above conditions, whereas DNT2e03444Toll-626 double mutant embryos are viable (Fig. 5c). The semi-lethality of DNT141Toll-626 and Toll-7P114;DNT2e03444 double mutation is consistent with lack of the receptor from one signaling pathway and the ligand from the other being equivalent to losing both ligands or both receptors. The viability of DNT2e03444Toll-626 double mutants is consistent with lack of both the receptor and the ligand from the same pathway being equivalent to but not worse than losing only one of them. The semi-lethality of DNT141Toll-626 double mutation can be rescued by the overexpression of Toll-6CY or Toll-7CY with Toll-7GAL4 (Fig. 5c, χ2(5) = 653.525, P < 0.001; corrected P < 0.001, P < 0.001 respectively). These data suggest that Toll-7 may function downstream of DNT1 and Toll-6 downstream of DNT2. DNT141Toll-7P114 double mutants also have reduced viability, suggesting that Toll-7 may also act downstream of DNT2.

Figure 5: Toll-6 and Toll-7 interact genetically with DNT2 and DNT1.
Figure 5

Survival index for homozygous yw;;+/+ controls bred from an outcross to TM6B at 18 °C is 1. (a) Single homozygous mutants lacking one DNT or Toll-6 or Toll-7 are viable, whereas homozygous double losses of DNT1 and DNT2 or Toll-6 and Toll-7 are semi-lethal if bred at 18 °C as progeny of a stock maintained over a TM6B or SM6aTM6B balancer. Chi-squared χ2(11) = 360.277, P < 0.001, n = 126–872 pupae per genotype. (b) The semi-lethality of Toll-7P8;Toll-626 double mutation can be rescued by overexpressing the activated receptors in cholinergic neurons. χ2(6) = 85.028, P < 0.001, n = 102–467 pupae. (c) Homozygous double losses of one DNT and one Toll recapitulate the semi-lethality of DNT141DNT2e03444 and Toll-7P8;Toll-626 double mutations, and the lethality DNT1 Toll-6 double mutations can be rescued by expressing the activated receptors with Toll-7GAL4. χ2(5) = 653.525, P < 0.001, n = 72–991. (d) The semi-lethality of DNT141DNT2e03444 double mutation can be rescued by expressing activated Toll-6, Toll-7 or Toll-1 receptors in neurons. χ2(10) = 401.419, P < 0.001, n = 83–1,461 pupae. (e,f) Quantification of anti–cleaved-Caspase-3 labeling in embryonic VNCs: apoptosis increase in DNT55 and DNT2e03444/Df6092 mutant embryos compared to wild-type (yw) controls (e, one-way ANOVA F(2,69) = 10.479, P < 0.001; post hoc Dunnett P < 0.01, P = 0.051, respectively) is rescued with the overexpression of activated Toll-7CY and Toll-6CY in all neurons (f, Welch ANOVA F(2,63) = 5.143, P = 0.009; post hoc Dunnett P = 0.011, P = 0.017, respectively). In e,f, asterisks refer to pair-wise comparisons to yw, post hoc Dunnett tests. (g) Pan-neuronal overexpression of activated Toll10b, Toll-6CY and Toll-7CY rescues the semi-lethality of the spz2 mutation; χ2(7) = 99.272, P < 0.001. ***P < 0.001; **P < 0.01; *P < 0.05. (ad,g) Asterisks refer to chi-squared comparisons to fixed controls with Bonferroni corrections. For detailed genotypes and statistics details, see Supplementary Tables 1 and 2.

We thus asked whether activated forms of Toll-6 and Toll-7 could rescue DNT mutant CNS phenotypes. Pan-neuronal expression of Toll-6ΔLRR, Toll-7ΔLRR, Toll-6CY or Toll-7CY rescued the semi-lethality of DNT141DNT2e03444 double mutation (Fig. 5d, χ2(10) = 401.419, P < 0.001, corrected P < 0.001 for all). Overexpression of Toll-7CY in all neurons rescued the apoptosis caused by loss of DNT1 function, and overexpression of Toll-6CY in all neurons rescued the apoptosis caused by loss of DNT2 function (Fig. 5e, F(2,69) = 10.479, P < 0.001, and Fig. 5f, F(2,63) = 5.143, P = 0.009; corrected P < 0.01, P = 0.051, P = 0.011, P = 0.017, respectively). Altogether, these data indicate that Toll-7 and Toll-6 most likely function as receptors for DNT1 and DNT2, respectively, although these interactions may be promiscuous.

Toll-6 and 7 function upstream of NF-κB and bind DNT1 and 2

We next asked whether potential interactions between the DNT1 and DNT2 ligands and Toll-6 and Toll-7 receptors could induce NF-κB signaling. Pan-neuronal overexpression constitutively active Toll10b, which activates NF-κB homologs Dorsal and Dorsal-related immunity factor (Dif)4, rescued the semi-lethality in spz2 mutants, and this rescue was replicated with the overexpression of activated Toll-6CY and Toll-7CY in neurons (Fig. 5g, χ2(7) = 99.272, P < 0.001; corrected P < 0.001, P = 0.018, P < 0.001, respectively). Transfection of the S2 cell line with activated Toll-6CY and Toll-7CY resulted in the activation, upon induction, of snail-luciferase, a reporter for Dorsal, and drosomycin-luciferase, a reporter for Dif (Supplementary Fig. 4a, F(5,24) = 27.165, P < 0.001; corrected P = 0.054, P = 0.000084, respectively; Supplementary Fig. 4b, F(5,24) = 6.574, P = 0.001; corrected P = 0.01, P = 0.018, respectively). Furthermore, when S2 cells transfected with hemagglutinin (HA)-tagged Toll-6HA or Toll-7HA were stimulated with purified DNT2, this triggered a drosomycin-luciferase readout indicative of Dif signaling (Supplementary Fig. 4c,d, F(5,27) = 16.788, P < 0.001, corrected P = 0.034, P = 0.09, respectively; and Supplementary Fig. 5a–d). However, in vivo, overexpression of Toll-6CY and Toll-7CY did not induce drosomycin-GFP expression (Supplementary Fig. 4e), as is consistent with previous reports7,8. Nevertheless, activated Toll-6CY and Toll-7CY induced increases in Dorsal, Dif and Cactus proteins (Supplementary Fig. 4f; Dorsal: F(2,9) = 10.382, P = 0.005; corrected P = 0.003, P = 0.085, respectively; Dif: F(2,9) = 12.898, P = 0.005; corrected P = 0.002, P = 0.006, respectively; Cactus: F(2,4.14) = 28.233, P = 0.004; corrected P = 0.038, P = 0.011, respectively). Although we do not provide mechanistic evidence for whether or not Toll-6 or Toll-7 signaling involves the canonical Toll pathway, our data indicate that Toll-6 and Toll-7 function upstream of NF-κB.

In the light of the genetic evidence that Toll-6 and Toll-7 are receptors for Drosophila neurotrophins, we next asked whether DNT1 and DNT2 could bind Toll-7 and Toll-6. To test whether they can interact in vitro, we purified His-tagged secreted forms of the receptors comprising only the extracellular domain, Toll-6-ECDHis and Toll-7-ECDHis (Fig. 6a,b and Supplementary Fig. 5e); and a baculovirus-expressed cleaved DNT2 cystine knot domain, DNT2-CK-His (Fig. 6a–c and Supplementary Fig. 5a–d). We mixed these to allow formation of complexes, which were subjected to native gel electrophoresis (Fig. 6d,e). In native gels, protein mobility does not depend on molecular weight but on conformation and charge relative to the pH of the buffer. At the pH of our buffers, all three proteins were negatively charged, but DNT2CK just barely so, as its pI was close to the buffer pH. Thus the mobility of the latter was limited, whereas the Toll-6ECD and Toll-7ECD migrated further (Fig. 6d), as their pI values differed from the buffer pH. Adding DNT2-CK shifted the mobility of Toll-6ECD and Toll-7ECD, and new bands appeared at the top of gel, relative to those in controls (Fig. 6d,e). This indicates that DNT2 interacts with both Toll-6 and Toll-7.

Figure 6: In vitro, cell culture and in vivo evidence that Toll-7 and Toll-6 bind DNT1 and DNT2.
Figure 6

(a) The constructs encoding tagged proteins. (b) Coomassie stainings showing secreted Toll-6ECD and Toll-7ECD purified from S2-cell conditioned medium (left panel; for mass spectrometry, see Supplementary Fig. 5e); DNT2 purified from a baculovirus expression system as a secreted cleaved cystine knot (CK) dimer (middle panel); and DNT1 purified from S2 conditioned medium as cleaved 45-kDa cystine knot plus terminal extension (CK, CTD; left arrow), plus full-length form and cleavage products, and DNT2 purified from S2 conditioned medium as cleaved CK (right arrow) only (right panel). (c) The mass of DNT2 purified by reverse phase chromatography, as determined by MALDI TOF mass spectrometry. The N-terminal residue and the seven cysteines that form three intra- and one inter-molecular disulfide bond are highlighted. Gln1 + −17.03 indicates that approximately 17 Da are lost due to cyclization of the N-terminal glutamine to pyroglutamate. Together with 7 Da eliminated in the formation of seven disulfide bonds, this gives an expected mass for the DNT2 monomer of 13,112.65 Da, or a dimer mass of 26,225.3 Da. The observed mass of peak A is 26,225.5 Da. (d) Native gel showing complexes of purified DNT2CK with purified Toll-6ECD, Toll-7ECD or both Toll-6ECD plus Toll-7ECD. WB, western blot. (e) Predicted mobility of native Toll-6ECDHisFLAG, Toll-7ECDHisFLAG and DNT2CKHis at pH 8.8. (f) Cotransfection S2 cell lysate controls for ELISA and coimmunoprecipitation (co-IP) experiments showing proteins expressed in each experiment in gi. (g) ELISAs using cotransfected S2 cells revealed a significant difference in absorbance comparing singly transfected and cotransfected S2 cell lysates. Unpaired t-tests: top, condition 6 versus 2: t(4) = −10.485, P < 0.001; 7 versus 3: t(4) = −7.619, P = 0.002; bottom, 7 versus 4: t(4) = −5.574, P = 0.005; 6 versus 5: t(4) = −13.504, P < 0.001; n = 3 repeats. (h,i) Co-IP of full-length Toll-7HA and DNT1V5, and full-length Toll-6HA and DNT2V5, from cotransfected S2 cells. (h) Precipitation of receptors with anti-HA coprecipitates bound ligands detected with anti-V5. (i) Precipitation of ligands with anti-V5 coprecipitates bound receptors detected with anti-HA. (j) In vivo co-IP from transgenic flies overexpressing full-length Toll-7HA and DNT1-FLAG in the retina with GMRGAL4. Two examples are shown, using rabbit (left) or mouse (right) anti-Flag antibodies to precipitate DNT1, coprecipitating bound receptor detected with anti-HA. ***P < 0.001, **P < 0.01, *P < 0.05. (hj) Full-length blots are shown in Supplementary Figures 6–9.

In S2 cell culture, cotransfection with full-length DNT1 tagged with the simian virus 5 (V5) epitope (DNT1V5) and full-length Toll-7HA, and with DNT2-V5 and Toll-6HA (Fig. 6f), revealed interactions between DNT1 and Toll-7 and between DNT2 and Toll-6 in ELISAs (Fig. 6g). This interaction was also demonstrated by coimmunoprecipitation of the ligands and receptors expressed in cotransfected S2 cells. Anti-hemagglutinin precipitated full-length Toll-6HA and Toll-7HA receptors, and did so only in receptor-transfected cells (Fig. 6h). Whereas we detected no ligands after immunoprecipitation from controls that had not been transfected with receptors, antibodies to the hemagglutinin-tagged forms of Toll-6 and Toll-7 copurified DNT2-V5 and DNT1-V5, respectively, from cotransfected cells (Fig. 6h). There was some nonspecific binding of DNT1 in the control lacking receptor, but at lower levels than in the cotransfected cells (Fig. 6h). In the reverse coimmunoprecipitation, anti-V5 precipitated full-length and cleaved V5-tagged DNT2 and DNT1, and did so only in ligand-transfected cells (Fig. 6i). By contrast, we detected no receptors after immunoprecipitation from control S2 cells that had not been transfected with ligands, whereas antibodies specific for the V5-tagged forms of DNT1 and DNT2 copurified Toll-7HA and Toll-6HA, respectively, from cotransfected cells (Fig. 6i). Together, these data show that DNT1 or DNT2 and full-length transmembrane Toll-6 or Toll-7 can be coimmunoprecipitated from S2 cells. In vivo, we immunoprecipitated DNT1 bound to Toll-7 from transgenic flies overexpressing (using a GMRGAL4 driver) both Flag epitope–tagged DNT1 cystine knot domain (DNT1-CK-Flag) and full-length Toll-7HA in the retina (Fig. 6j). Together, these data demonstrate that DNT1 binds Toll-7 and DNT2 binds promiscuously Toll-6 and Toll-7.

Consistent with their functions as ligands for Toll-7 and Toll-6 in interneurons and motor neurons, DNT1 and DNT2 are expressed in embryonic CNS midline and muscle20. We found anti-DNT1 antibodies to be specific, as they did not reveal signal in DNT141 null mutant embryos (Supplementary Fig. 1b). In normal embryos, DNT1 protein was detectable at the midline (Fig. 7a), a target of interneurons, and in high levels in muscles 13,12 and lower levels in muscles 6,7, the targets of ISNb/d axons (Fig. 7b). We were not able to generate a DNT2 null allele; thus, we could not confirm that anti-DNT2 signal was absent in mutants. However, anti-DNT2 detected ectopic DNT2 distribution in embryos, larval and adult brains (Supplementary Fig. 1c–e), suggesting that it detects endogenous DNT2 protein in vivo. In normal larvae, anti-DNT2 revealed puncta along FasII+ interneuron axons (Fig. 7c), Toll-6GAL4>myrRFP RFP+ (Fig. 7d) and Toll-6MIMICGFP GFP+ (Fig. 7e) axons. In adult brains, DNT1 overlay Toll-7 in fan-shaped-body layers (Fig. 7f), whereas DNT2 and Toll-6MIMICGFP were distributed in complementary layers (Fig. 7f), compatible with the nonautonomous function of DNT2. Thus, the distribution of DNT1 and DNT2 supports their functioning as Toll-7 and Toll-6 ligands in vivo.

Figure 7: The relative distributions of DNT1, 2 and Toll-7, 6, respectively, in vivo are consistent with their functions are ligand-receptor pairs.
Figure 7

(a) Anti-DNT1 reveals DNT1 protein distributed in the embryonic CNS midline (stage 15) and (b) at high levels in muscle 13,12, in lower levels in muscles 6,7 and possibly others too (stage 17). (c,d) Anti-DNT2 reveals punctate signal along larval CNS axons revealed with (c) FasII+, (d) DsRed+ in Toll-6GAL4(D42)>myrRFP and (e) Toll-6MIMICGFP. (f) Anti-DNT1 and anti–Toll-7 colocalize in fan-shaped-body layers. Anti-DNT2 and anti-GFP in Toll-6MIMICGFP are distributed in complementary fan-shaped-body layers. Anterior is up. Scale bars: a,d,e, 10 μm; c,f, 50 μm.


We have found that neurotrophic functions in the fruit fly are carried out by Toll-7 and Toll-6 binding DNT1 and DNT2, respectively. Toll-6 and Toll-7 are expressed in the locomotor circuit, including motor neurons and interneurons of the embryonic central pattern generator and locomotion centers of the adult central brain. By removing Toll-6 and Toll-7 function in mutants or adding them in excess, we have shown that Toll-6 and Toll-7 are required for normal locomotion and motor axon targeting, and to maintain neuronal survival. In the absence of Toll-6 and Toll-7 function, at least some of the dying cells are HB9+ and Eve+ EL interneurons that normally express the receptors. Using genetic interaction analysis, we have shown that Toll-6 and Toll7 function together with DNT1 and DNT2 in vivo. Using biochemical approaches in vitro, in cell culture and in vivo, we have shown that Toll-6 and Toll-7 directly bind DNT2 and DNT1, respectively. Finally, the relative in vivo protein distribution patterns of the ligands and the receptors are consistent with their shared functions. Most importantly, we have shown that Toll receptors underlie neurotrophism in fruit flies, which is therefore implemented using a different molecular mechanism from the canonical vertebrate mechanism involving p75NTR, Trks and Sortilin.

Our data show that Toll-6 and Toll-7 have neurotrophic functions in the Drosophila CNS matching those of DNT1 and DNT2 (ref. 20). As in the mammalian neurotrophin system, these functions are pleiotropic. Mammalian neurotrophin ligands and receptors have functions ranging from maintaining neuronal survival to axon targeting, dendritic arborization and synaptic transmission, which vary with context, cell type and time16,38,39. For instance, whereas vertebrate neurotrophins and Trk receptors maintain neuronal survival in the peripheral nervous system, they do not have a prominent role in maintaining motor neuron survival, instead functioning at the neuromuscular junction in synaptogenesis and synaptic plasticity40. Our data show that Toll-6 and Toll-7 also have pleiotropic functions, maintaining predominantly interneuron survival and regulating motor-axon targeting.

Our data indicate that Toll-7/DNT1 and Toll-6/DNT2 are the most likely ligand-receptor pairs, but there appears to be promiscuity in ligand binding, as at least DNT2 can bind both receptors. This may also be the case for DNT1, but pure mature DNT1 protein could not be obtained using the baculovirus system, restricting the tests that we were able to perform. Such promiscuity may account for the redundancy between Toll-6 and Toll-7 observed in genetic and functional tests (for example, compromised locomotion and viability in the double mutants only). It may indicate that in vivo the binding partners might be determined by the relative temporal and spatial distribution patterns of the proteins. Alternatively, it is also conceivable that DNT1 and DNT2 have distinct functions and may bind each receptor according to functional requirements. DNT1 and DNT2 have distinct biochemical properties: whereas DNT2 is consistently secreted from S2 cells as a mature, cleaved form consisting of the cystine knot domain, DNT1 is secreted both as full-length and mature forms, as well as products of cleavage in the disordered pro-domain. The protease that might cleave DNT1 in vivo is unknown, but these properties are akin for DNT2 to the intracellular cleavage of NGF and for DNT1 the extracellular cleavage of BDNF16. In either case, the observed promiscuity is reminiscent of the binding of all mammalian neurotrophins to a common p75NTR receptor.

Although vertebrate neurotrophin receptors are structurally and functionally distinct from the Tolls, both regulate NF-κB1,13,32,33,41. NF-κB is also one of the transcription factors that activates the innate immune response downstream of the TLRs, and it also has extensive and highly conserved functions in neurons. Neuronal NF-κB controls gene expression as a potent prosurvival factor; it controls neurite extension; it also has non-nuclear synaptic functions, including the clustering of glutamate receptors; and it underlies synaptic plasticity during learning and memory, from crustaceans to mammals31,32,41,42,43. In humans, alterations in NF-κB function lead to psychiatric disorders41. Previous reports have shown that Toll-6 and Toll-7 do not activate Drosomycin upon immune challenge, indicating that Toll-6 and Toll-7 do not have innate immunity functions and do not activate NF-κB–Dif in cell types involved in immunity7,8. In future work we plan to elucidate the signaling mechanism downstream of Toll-6 and Toll-7 in the CNS and, in particular, to determine whether it uses downstream signal transducers such as MyD88 that are required for the immune and developmental functions of Toll-1. The mammalian TLR-8 is required for neurite extension in the neonatal brain, but this activity is not MyD88 dependent44. Thus, although our data do not confirm or refute whether Toll-6 and Toll-7 can signal through the canonical Toll signaling pathway, they do show that Toll-6 and Toll-7 function upstream of NF-κB.

This conclusion is supported by several observations reported here. First, in cell culture, activated forms of Toll-6 and Toll-7 and stimulation with DNT ligands were able to induce NF-κB signaling via Dorsal and Dif. Second, in vivo, overexpression of activated Toll-6CY and Toll-7CY in retinal photoreceptor neurons resulted in the elevation of Dorsal, Dif and Cactus proteins, as was previously reported for Toll-145. Third, in vivo, overexpression in neurons of activated Toll-6CY and Toll-7CY, like activated Toll10b, rescued the semi-lethality of the spz2 mutation; and conversely, overexpression of activated Toll10b in neurons rescued the semi-lethality of the DNT1 DNT2 double mutation. Our data also show that signaling by Toll-6 and Toll-7 differs in at least some respects from that mediated by Spz–Toll-1. For example, in cell culture the activation of NF-κB signaling by Toll-6 and Toll-7 was not as strong as that reported by others to be induced by Toll-17,8; and in vivo genetic rescues revealed a specific and stronger relationship between Toll-6 and Toll-7 and DNT1 and DNT2, compared to Toll-1. Understanding the molecular mechanisms of Toll-6 and Toll-7 signaling that underlie the developmental programs that they promote is a key objective of future research.

Notably, NF-κB, p75NTR and Toll receptors are all evolutionarily very ancient molecules, present in cnidarians (for example, Nematostella); thus, they evolved long before the common ancestor of flies and humans and since the origin of the nervous and immune systems1,46. Of note, the Toll homolog in the worm Caenorhabditis elegans is expressed in neurons and can implement an immune function by means of a behavioral response of pathogen avoidance47. p75NTR is a member of the tumor necrosis factor receptor superfamily, which is closer to the Tolls than to the Trks48. Toll receptors resemble p75NTR intracellularly, through their ability to activate a downstream signaling pathway resulting in the activation of NF-κB, and Trk receptors in the extracellular ligand-binding module, with a combination of leucine-rich repeats and cysteine repeats48. Trk receptors, with an intracellular tyrosine kinase domain, emerged later in evolution49. Although Toll receptors are evolutionarily conserved, they are not, at least in the innate immunity context, activated by the same ligands in flies and humans1. This raises questions: if in Drosophila the Trk receptors were lost and Tolls are the only neurotrophic receptors, is this a key difference that underlies the distinct brain types and behaviors in flies and humans? In the course of evolution, did the Tolls become specialized for immunity functions in vertebrates? Or is the relationship uncovered here between the neurotrophin-ligand and Toll-receptor superfamilies an ancient mechanism of nervous system formation? In mammals TLRs also have nervous system functions, including ones in neurogenesis, neurite growth, plasticity and behavior, but the endogenous ligands in the mammalian CNS are unknown50. A key objective of future research will be to investigate whether the neurotrophin and TLR protein families interact in the mammalian brain, particularly in the context of learning, memory, and neurodegenerative and neuroinflammatory diseases.



Mutants and reporters. Toll-6MiMICMI02127 encodes a GFP-bearing insertion into the coding region of Toll-6 ( Toll-626, Toll-631, Toll-7P8 and Toll-7P114 are null mutant alleles generated by imprecise excision of P-element insertions (gift of J.L. Imler). Deficiencies Df(3L)DXG4 and Df(2R)BSC22 uncover Toll-6 and Toll-7, respectively. DNT141, DNT2e03444 and spz2 have been described20,51. All stocks were balanced using lacZ-marked balancers and/or TM6B Tb to identify mutants. Drosomycin-GFP (gift of J.M. Reihhart) is a reporter for Dif signaling.

Overexpression in vivo. Overexpression in vivo used the following GAL4 drivers: (1) w;;elavGAL4 for all neurons, (2) w;chaGAL4 (gift from R. Baines) for cholinergic neurons, (3) line D42 (gift from S. Sanyal) for Toll-6-GAL4 (ref. 36), (4) w;Toll-7GAL4, (5) w;GMRGAL4 (gift from M. Freeman) for retina, (6) w;;HB9GAL4 for HB9-neurons, and (7) w;engrailedGAL4 (Bloomington). These were crossed to (1) membrane-tethered reporters: (i) w;;UASGAP-GFP (gift from A. Chiba), (ii) w;UASmCD8-GFP, (iii) w;;10xUAS-myr-td-Tomato (gift from B. Pfeiffer), (iv) UASDsRed (gift from K. Ito), (v) w;Lim3GAL4UASmyrRFP/CyOactYFP (gift from M. Landgraf), (2) activated forms of the receptors: (i) w;UASToll-6ΔLRR and w;UASToll-7ΔLRR , (ii) w;;UASToll-6CY and w;;UASToll-7CY, (iii) w;UASToll10b (gift from T. Ip). Other lines were generated by conventional genetics.

Survival assays. Flies were bred at 18 °C, as stocks or crossed from heterozygous mutants over balancer chromosomes. The survival index (SI) is given by SI = 2 × TM6B+/TM6B. A SI of 1 is the Mendelian expectation when viability is unaffected. See Supplementary Table 2 for genotypes of parental flies and sample sizes.

Generation of fusion constructs.

Toll-7GAL4. To generate Toll-7GAL4, a 5-kb fragment immediately upstream of the Toll-7 start ATG was amplified by PCR using primers 1 and 2 (Supplementary Table 3) and cloned into 5′ NotI and 3′ BamHI restriction sites of the pPTGAL vector.

Cloning of full-length Toll-6 and Toll-7. Toll-6 and Toll-7 are intronless genes. Full-length open reading frames, equivalent to cDNAs, were PCR-amplified from genomic DNA using primers 3–6 (Supplementary Table 3), and Gateway cloning was used to generate pAct-Toll-6-HA and pAct-Toll-7-HA fusion constructs.

Cloning of Toll-6ECD-His-FLAG and Toll-7ECD-His-FLAG. Sequences encoding the extracellular domains (ECD) of Toll-6 and Toll-7 were cloned to produce secreted forms of the ligand-binding domains for binding assays in vitro. Sequences encoding Toll-6 and Toll-7 ECDs tagged with 6His were PCR-amplified using primers 7–10 (Supplementary Table 3). Gateway cloning was used to generate pAct-Toll-6-ECD-6His-3xFLAG and pAct-7-ECD-6His-3xFLAG fusion constructs.

Generation of activated forms of Toll-6 and Toll-7 for cell culture and transgenesis. The activated Toll-6 and Toll-7 receptors UASToll-6ΔLRR-HA and UASToll-7ΔLRR-HA comprise the signal peptide, transmembrane and intracellular domains of Toll-6 and Toll-7, respectively, but lack the entire extracellular domain. PCR from cDNA was used to amplify signal peptide sequences (primers 11, 12, 15, 16; Supplementary Table 3) separately from transmembrane-intracellular domain sequences (primers 13, 14, 17, 18; Supplementary Table 3), and MluI sites were added to each, 3′ and 5′ respectively. MluI sites were used to ligate fragments, which were then cloned into the pUAS-Gateway-HA-attB destination vector (gift of C. Basler) to produce pUAS-Toll-7ΔLRR-HA-attB and pUAS-Toll-6ΔLRR-HA-attB fusion constructs.

In UAS-Toll-6CY and UAS-Toll-7CY the conserved cysteines at position 1020 of Toll-6 and position 993 of Toll-7 are substituted by tyrosine, mimicking the constitutively active allele of Toll, Toll10b, and the functional UASToll10b constructs. Overlap extension PCR following standard procedures was used to make the cysteine-to-tyrosine mutations in each receptor. The primers used were as follows: 5′ primers 19, 20, 23 and 24 and 3′ primers 21, 22, 25 and 26 (Supplementary Table 3). The resulting products were cloned into the UAS-Gateway-FLAG destination vector to produce pUAS-Toll-7CY-FLAG and pUAS-Toll-6CY-FLAG fusion constructs.

Sequencing confirmed that in UAS-Toll-7CYFLAG and UAS-Toll-6CYFLAG the targeted cysteines had been mutated to tyrosines. There is an additional mutation in UAS-Toll-7CYFLAG of Pro58 to leucine and one in UAS-Toll-6CYFLAG of Ser862 to threonine, both in the extracellular domain.

Cloning for DNT protein production. DNT1 and DNT2 proteins were produced by S2 cell expression, from pAct5C-Pro-TEV6HisV5-DNT1-CK-CTD and pAct5C-Pro-TEV6HisV5-DNT2-CK fusion constructs encoding full-length DNT1 and DNT2. TEV protease sites designed to aid protein cleavage were not used, as they cleaved spontaneously. The DNT1 and DNT2 signal peptide and pro-domain sequences were PCR-amplified from DNT1 and DNT2 cDNAs, with primers 27, 28, 33 and 34 (Supplementary Table 3). The DNT1 and DNT2 cystine knots (plus the C-terminal extension CK-CTD for DNT1) were amplified using primers 29, 30, 35 and 36 (Supplementary Table 3). Using overlapping PCR, tagged full-length DNT1 and DNT2 sequences were obtained and cloned using Gateway.

To produce DNT2 protein using baculovirus-infected Sf9 insect cells, full-length DNT2 was PCR amplified from clone LD26258 (Berkeley Drosophila Genome Project) using primers 31 and 32 (Supplementary Table 3). The insert was cloned into the pFastBac1 vector (Invitrogen) using EcoRI and NotI sites.

Protein purification.

Purification of Toll-6ECD and Toll-7ECDs produced from S2 cells. S2 cells were transfected with pAct-Toll-6-ECD-6His-3xFLAG for Toll-6ECD or pAct-7-ECD-6His-3xFLAG for Toll-7ECD and incubated for 72 h. Protein purification was performed as described above for DNT1 and DNT2. Toll-6 and Toll-7 ECDs were identified by mass spectrometry at the Proteomics Facility, University of Birmingham. Ni-NTA–purified Toll-6 and Toll-7 ECD proteins were further purified with anti-FLAG magnetic beads (Sigma-Aldrich) by standard procedures. The purified proteins were used for native gel electrophoresis.

Purification of DNT2 produced from baculovirus expression system. A secreted DNT2 protein was produced in Sf9 insect cells by baculovirus infection with the DNT2-Pro-CK-TEV-6xHis sequence. The protein was purified as described previously17. Edman (N-terminal) sequencing was carried out at the Protein and Nucleic Acid Chemistry (PNAC) facility at the Department of Biochemistry, University of Cambridge. The cleaved, mature cystine knot domain of DNT2 was purified by reverse phase chromatography, as previously described for Spz (ref. 15).

Mass spectrometry.

Toll-6ECD and Toll-7ECD verification. Coomassie bands were excised from a gel, destained and subjected to in-gel digestion with trypsin for overnight at 37 °C using standard procedures. Peptides were extracted from gel pieces with acetonitrile and formic acid and dried in an evaporator. The samples were resuspended in 0.1% formic acid/water and subjected to liquid chromatography–tandem mass spectrometry that was performed using an Ultimate 3000 HPLC series (Dionex) coupled to a LTQ Orbitrap Velos ETD mass spectrometer (ThermoFisher Scientific) via a Triversa Nanomate nanospray source (Advion Biosciences). Peptide separation, mass spectrometric analysis and database search were carried out as specified at the University of Birmingham Proteomics Facility.

DNT2 verification. DNT2CK was analyzed by MALDI-TOF mass spectrometry, following procedures previously described for Spz15.

Western blotting.

Western blotting was carried out following standard procedures. Primary antibodies used were mouse anti-6-His (1:4,000, BD Pharmingen, #552565) or mouse anti-6-His (1:1,000, Thermo Scientific, #MA1-21315), mouse anti-V5 (1:5,000, Invitrogen, #R960-25), mouse anti-Dorsal (1:500, Hybridoma Bank, 7A4), mouse anti-Cactus (1:500, Hybridoma Bank, 3H12), rabbit anti-Dif (1:500, ref. 52, gift from D. Ferrandon), chicken anti-HA (1:2,000 for S2 cell co-IP western blot; 1:5,000 for fly in vivo co-IP, #ET-HA100) and mouse anti-HA (12CA5) (1:2,000, Roche, #11 583 816 001). Secondary antibodies used were HRP–anti-mouse (1:5,000, Vector Labs, #PI-2000), HRP–anti-rabbit (1:5,000, Vector Labs, #PI-1000) and HRP–anti-chicken (1:10,000, Jackson ImmunoResearch, #703-035-155). For quantitative analysis of dorsal, cactus and Dif expression, five adult fly heads were pooled per sample (n = 4, total 20 flies per genotype), then lysed in NP-40 lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% NP-40). Western blot images were analyzed by GeneTools software (Syngene). Band intensities were normalized to Ponceau red staining of the same membrane.

S2 cell culture signaling assays.

Toll-6 and Toll-7 transfections. DNT1 was produced in transfected S2 cells (Invitrogen), and DNT2 was produced by both transfected S2 cells and Sf9 cells infected with baculovirus. Purified ligands were added to S2 cells transfected with Toll-6 and Toll-7 receptors to test activation of the snail-luciferase reporter by Dorsal or nuclear translocation of Dorsal; the cell line 648-1B6 stably transfected with drosomycin-luciferase was used to test the activation of the drosomycin-luciferase reporter by Dif. To test downstream signaling by the activated receptors, S2 cells were transfected with 1 μg pMTGAL4 and 1 μg pUAS-Toll-7CY-FLAG or with 1 μg pUAS-Toll-6CY-FLAG. To test signaling by Toll-6 and Toll-7 upon DNT binding, full-length Toll-6 and Toll-7 were expressed in S2 cells by transfecting with 1 μg of the pAct-Toll-7-HA or pAct-Toll-6-HA constructs. S2 cells and the S2 cell line 648-1B6 1B6 stably transfected with drosomycin-luciferase were maintained by standard procedures, and the TranIT-2020 (Mirus) transfection reagent was used. In all cases, pAct-Renilla was cotransfected as a control. To determine luciferase activity, we used the Dual-Glo Luciferase Assay System (Promega). Cell suspension (50 μl) was transferred in triplicate to an opaque 96-well plate. The supplied firefly luciferase substrate (40 μl) was added, samples were incubated for 10 min at room temperature, and luminescence was measured using a SPECTRAFluro Plus (Tecan). Stop & Glo substrate (40 μl) was added, samples were incubated at room temperature for 10 min and luminescence was measured. The relative luciferase activity was determined by normalizing the firefly over the Renilla readout for each well.

Coimmunoprecipitation (co-IP).

In vivo co-IP from transgenic flies. Transgenic flies of genotype w;GMRGAL4/UASDNT13′+FLAG; UASToll-7HA/+ and wild-type Oregon R control flies were used for in vivo co-IP. Heads were homogenized in NP40 lysis buffer and spun at 1,200g for 5 min at 4°. Supernatant was precleared with 20 μl of protein-A/G magnetic beads (Pierce) in a total volume of 500 μl for 1 h at 4° and then incubated with 1 μg mouse anti-Flag antibody (cat. no. F3165, clone M2, Sigma-Aldrich) overnight at 4 °C. Protein-A/G magnetic beads (25 μl) were added to the lysate and incubated at room temperature for 1 h. The beads containing the immune complex were washed twice with lysis buffer and once in PBS. The immune complex was eluted in 2× Laemmli buffer by incubating for 10 min at room temperature and was analyzed by SDS-PAGE.

Co-IP from S2 cells. S2 cells were transfected with 1 μg pAct-Toll-6-HA, pAct-Toll-7-HA, pAct5C-Pro-TEV6HisV5-DNT1-CK-CTD or pAct5C-Pro-TEV6HisV5-DNT2-CK, or cotransfected with 1 μg pAct5C-Pro-TEV6HisV5-DNT1-CK-CTD plus 1 μg pAct-Toll-7-HA per well or with 1 μg pAct5C-Pro-TEV6HisV5-DNT2-CK plus 1 μg pAct-Toll-6-HA per well. After 48 h cells were collected and centrifuged at 1,000g for 5 min at 4 °C. After washing in PBS, cells were lysed in 600 μl NP-40 lysis buffer with protease inhibitor cocktail (Pierce Biotechnology) and left on ice for 30 min. After lysis, samples were centrifuged at 14,000g for 10 min at 4 °C. Five hundred microliters from each lysate was precleared with 20 μl of protein-A/G magnetic beads (Pierce) for 1 h at 4 °C. Precleared lysates were incubated with 1.5 μg anti-V5 or 1 μg anti-HA overnight at 4 °C. Cell lysate–antibody mixtures were incubated with 25 μl protein-A/G magnetic beads for 1 h at room temperature. Beads were washed two to four times in NP-40 lysis buffer and once in PBS. Proteins were eluted in 40 μl 2× Laemmli buffer and analyzed by SDS-PAGE.


S2 cells were seeded, transfected and processed as described above for co-IP. ELISA was carried out following standard procedures. High-binding-capacity ELISA plate (Greiner Bio-One) wells were coated with anti-V5 (1:1,000) or anti-HA (1:1,000). Each well was incubated with 100 μl of cell lysate and then with anti-HA (1:1,000) or anti-V5 (1:1,000), respectively. ELISA signal was developed by TMB substrate (Pierce) and the reaction was stopped with 1 M H3PO4. Absorbance was measured at 450 nm.

Native gel electrophoresis.

To test whether the secreted Toll-6ECD and Toll-7ECD could interact with mature DNT2-CK, proteins were mixed and putative complexes were run on a native gel. In native gels, protein mobility does not depend on molecular weight, but on conformation and charge, relative to the pH of the buffer (as the difference between the buffer pH and the pI of the proteins increases, the proteins migrate further). We used sample and gel buffers at pH = 8.8 and running buffer pH = 8.6. The pI of DNT2CK-TEV-6His is 7.94, very close to the buffer pH. The pI of Toll-6ECD-6His-3xFLAG is 6.06 and that of Toll-7ECD-6His-3xFLAG is 6.0, quite different from the buffer pH. For pI and net charge calculation, we used Protein Calculator v3.3 ( At the pH of our buffers, both proteins are negatively charged.

Toll-6ECD and Toll-7ECD were purified from S2 cell conditioned medium, and DNT2-CK was produced using baculovirus infection and purified by reverse phase chromatography. 12.5 μl Ni-NTA- and FLAG-purified Toll-6 and/or Toll-7 ECD were mixed with 2 μl of reverse-phase-purified DNT2-CK and left on ice for 30 min to form complexes. Protein mixtures or proteins alone were supplemented with 4× native gel-loading buffer (62.5 mM Tris-HCl, pH 8.8, 20% glycerol and 0.005% bromophenol blue). Samples were separated on a 6% polyacrylamide gel for 1 h at 100 V followed by 1 h at 150 V in the absence of SDS. Proteins were then analyzed by western blotting.

Generation of anti–Toll-7, anti-DNT1 and anti-DNT2 antibodies.

Antibodies were raised to Toll-7 using peptide AAQRAQTWRPKREQLHLQQA injected into guinea pigs, and antisera were affinity purified (Davids Biotechnologie). Antibodies were raised to DNT1 using peptide VRYARPQKAKSASGEWKY and to DNT2 using peptide KRLIALQGNGQN; peptides were injected into rabbits and antisera were affinity purified (Davids Biotechnologie).

In situ hybridizations and immunohistochemistry.

In situ hybridizations followed standard procedures using mRNA probes, from pDONR-Toll-7 linearized with HindIII and pDONR-Toll-6 linearized with SmaI, both transcribed with T7 RNA polymerase.

Immunolabeling was carried out following standard procedures. Primary antibodies used were rabbit anti-GFP (1:1,000 for embryos and larvae, 1:250 for adult brains, Invitrogen #A11122), mouse anti-GFP (1:1,000, Invitrogen #A11120), rabbit anti-DsRed (1:100, Clontech #632496), mouse anti-FasII (1:4 for embryos, 1:250 for larvae, DSHB ID4), mouse anti–tyrosine hydroxylase (1:50, Immunostar #22941), rabbit anti–cleaved-Caspase-3 (1:50, Cell Signaling #9661; 1:250, AbCam #Ab13847), mouse anti–β-gal (1:750, Sigma #G4644), guinea pig anti-HB9 (1:1,000, ref. 53 gift from H. Broihier), rabbit anti-HB9 (1:1,000, ref. 53, gift from H. Broihier), mouse anti-Eve (1:5–1:10, DSHB 2B8), mouse anti-Dorsal (1:10, DSHB 7A4), guinea pig anti–Toll-7 (see above; 1:10), rabbit anti-DNT1 (see above; 1:50 for adult brains, 1:100 for embryos) and rabbit anti-DNT2 (see above; 1:100 for embryos and larvae, 1:50 for adult brains). Secondary antibodies were directly conjugated Alexa 488, 546 and 647 (1:250, Molecular Probes) or biotinylated mouse or rabbit (1:300) followed by avidin amplification using the Vectastain ABC Elite kit (Vector Labs) or Streptavidin–Alexa 488 (1:250, Molecular Probes). Stainings were carried out in populations of hundreds of embryos, at least three adult brains and three larval VNCs per experiment, and experiments were repeated at least twice and most often more.


Laser-scanning confocal microscopy was carried out using a Leica SP2-AOBS and a 40× or 63× lens at 512 × 512 or 1,024 × 1,024 pixels resolution, with 0.5- or 1-μm steps. Caspase-positive apoptotic cells in vivo were counted automatically using DeadEasy Caspase software37.

Automatic tracking of larval locomotion.

Larvae were collected from crosses kept in vials with standard yeast-rich food at 25 °C in a 12 h light:12h dark regimen. All larvae analyzed were also homozygous white (w) mutant. For our w+;Toll-626 stock, to obtain w;Toll-631/Toll-626 larvae, F1 larvae from a cross to w;Toll-631 females were sexed and only w males were used. Larvae were collected, rinsed in water and placed on a large petri dish containing agar, Vogel-Bonner salts and 40% glucose at room temperature and on a light box. Larvae were allowed to recover for 10–20 s before filming began and were then filmed for 1 min. All filming was carried out in the morning between 1 and 5 h after zeitgeber 'lights on' time, and larvae from all genotypes were filmed in each session. Larvae were filmed using a Motic MC Camera 1.1 or a Canon video camera.

Films were analyzed using FlyTracker software (our modification of the ImageJ MTrack2 plug-in). Canon films were first converted to .avi format using Any Video Converter Professional with codec .mjpeg, and Motic films were decompressed using VirtualDub software ( and saved in .avi format. Films were checked in VirtualDub for sequence continuity and images not containing larvae edited out to ensure a continuous sequence of filmed crawling larvae. The first 400 frames were opened blindly in ImageJ and converted to grayscale, the resolution was set to 800 × 600 pixels, and the frames were saved as a stack of .tiff images. These were next analyzed using FlyTracker in ImageJ. FlyTracker tracks the crawling larvae and produces three outputs per film: a stack of 400 images with the location of the larva as identified by the software in each corresponding raw image; a path or trajectory of each larva in each film as a single JPEG image; and an Excel spreadsheet with quantitative locomotion parameters. The speed in mm/s was converted from the output in pixels using a photographed reference of known size in mm. All trajectories for each genotype were plotted onto a single image using Adobe Photoshop layers.

Statistical analysis.

For continuous data, kurtosis, skewness, histograms and Kolmogorov-Smirnov normality tests were carried out to test data distribution, and Levene tests for the homogeneity of variance. For a few samples that were not distributed normally within a data set that showed normal distribution, normality was assumed. Normally distributed data were analyzed with unpaired t-tests for two-sample comparisons and one-way ANOVA (or Welch versions for significant Levene tests) for the whole data set followed by Dunnett (fixed control group) or Games-Howell (significant Levene tests) post hoc tests or Bonferroni corrections for multiple t-tests. Data that were not distributed normally were analyzed with Kruskal-Wallis for the whole data set, followed by Dunn's test for multiple comparisons. Categorical data were analyzed with chi-squared tests, followed by a z-test and Bonferroni post hoc to the whole data set or Bonferroni corrections for pairwise comparisons to controls. No blinding was done. For genetic in vivo experiments, the reproducibility of the experiment was verified by the overall large population sizes; cell culture experiments were carried out in general at least three times in triplicate. P values, tests and sample sizes are provided in the Results text and figure legends, and further details are provided in Supplementary Table 1.


  1. 1.

    & Toll-like receptors—taking an evolutionary approach. Nat. Rev. Genet. 9, 165–178 (2008).

  2. 2.

    & Innate immune recognition. Annu. Rev. Immunol. 20, 197–216 (2002).

  3. 3.

    & Biology of Toll receptors: lessons from insects and mammals. J. Leukoc. Biol. 75, 18–26 (2004).

  4. 4.

    , , , & The dorsoventral regulatory gene cassette spatzle/Toll/cactus controls the potent antifungal response in Drosophila adults. Cell 86, 973–983 (1996).

  5. 5.

    , & Functional analysis of Toll-related genes in Drosophila. Dev. Growth Differ. 52, 771–783 (2010).

  6. 6.

    et al. Virus recognition by Toll-7 activates antiviral autophagy in Drosophila. Immunity 36, 658–667 (2012).

  7. 7.

    , , & Toll-related receptors and the control of antimicrobial peptide expression in Drosophila. Proc. Natl. Acad. Sci. USA 97, 10520–10525 (2000).

  8. 8.

    , , & The Drosophila Toll-9 activates a constitutive antimicrobial defense. EMBO Rep. 3, 82–87 (2002).

  9. 9.

    , , & Induction of neuron-specific glycosylation by Tollo/Toll-8, a Drosophila Toll-like receptor expressed in non-neural cells. Development 130, 1439–1448 (2003).

  10. 10.

    , , & Tissue and stage-specific expression of the Tolls in Drosophila embryos. Gene Expr. Patterns 2, 311–317 (2002).

  11. 11.

    The immune response of Drosophila. Nature 426, 33–38 (2003).

  12. 12.

    et al. Binding of Drosophila cytokine Spatzle to Toll is direct and establishes signaling. Nat. Immunol. 4, 794–800 (2003).

  13. 13.

    & Structure and function of Toll receptors and their ligands. Annu. Rev. Biochem. 76, 141–165 (2007).

  14. 14.

    & Proteolytic processing of the Drosophila Spatzle protein by easter generates a dimeric NGF-like molecule with ventralising activity. Mech. Dev. 72, 141–148 (1998).

  15. 15.

    et al. Structural insight into the mechanism of activation of the Toll receptor by the dimeric ligand Spätzle. J. Biol. Chem. 283, 14629–14635 (2008).

  16. 16.

    , & The yin and yang of neurotrophin action. Nat. Rev. Neurosci. 6, 603–614 (2005).

  17. 17.

    , & Molecular mechanism that induces activation of Spatzle, the ligand for the Drosophila Toll receptor. J. Biol. Chem. 285, 19502–19509 (2010).

  18. 18.

    et al. Biophysical characterization of refolded Drosophila Spatzle, a cystine knot protein, reveals distinct properties of three isoforms. J. Biol. Chem. 283, 32598–32609 (2008).

  19. 19.

    , , & Crystallization of Spatzle, a cystine-knot protein involved in embryonic development and innate immunity in Drosophila melanogaster. Acta Crystallogr. Sect. F Struct. Biol. Cryst. Commun. 64, 707–710 (2008).

  20. 20.

    et al. Drosophila neurotrophins reveal a common mechanism for nervous system formation. PLoS Biol. 6, e284 (2008).

  21. 21.

    , & A family of proteins related to Spatzle, the toll receptor ligand, are encoded in the Drosophila genome. Proteins 45, 71–80 (2001).

  22. 22.

    The midline glia of Drosophila: a molecular genetic model for the developmental functions of glia. Prog. Neurobiol. 62, 475–508 (2000).

  23. 23.

    , , , & Programmed cell death in the embryonic central nervous system of Drosophila melanogaster. Development 134, 105–116 (2007).

  24. 24.

    et al. Genetic control of programmed cell death in Drosophila. Science 264, 677–683 (1994).

  25. 25.

    et al. Trophic neuron-glia interactions and cell number adjustments in the fruit fly. Glia 59, 1296–1303 (2011).

  26. 26.

    et al. Evidence that DmMANF is an invertebrate neurotrophic factor supporting dopaminergic neurons. Proc. Natl. Acad. Sci. USA 106, 2429–2434 (2009).

  27. 27.

    et al. Blocking apoptosis signaling rescues axon guidance in Netrin mutants. Cell Rep. 3, 595–606 (2013).

  28. 28.

    , , & Regulation of cell number by MAPK-dependent control of apoptosis: a mechanism for trophic survival signaling. Dev. Cell 2, 159–170 (2002).

  29. 29.

    , & The Drosophila neuregulin vein maintains glial survival during axon guidance in the CNS. Dev. Cell 1, 679–690 (2001).

  30. 30.

    , & Gliatrophic and gliatropic roles of PVF/PVR signaling during axon guidance. Glia 56, 164–176 (2008).

  31. 31.

    & Regulation of neural process growth, elaboration and structural plasticity by NF-κB. Trends Neurosci. 34, 316–325 (2011).

  32. 32.

    et al. NF-κB signaling promotes both cell survival and neurite process formation in nerve growth factor-stimulated PC12 cells. J. Neurosci. 20, 7556–7563 (2000).

  33. 33.

    et al. Selective activation of NF-κB by nerve growth factor through the neurotrophin receptor p75. Science 272, 542–545 (1996).

  34. 34.

    , & A role for Nerve Growth Factor in nervous, endocrine and immune systems. Prog. Neuroendocrinimmunol. 3, 1–10 (1990).

  35. 35.

    Regulation of innate immune responses in the brain. Nat. Rev. Immunol. 9, 429–439 (2009).

  36. 36.

    Genomic mapping and expression patterns of C380, OK6 and D42 enhancer trap lines in the larval nervous system of Drosophila. Gene Expr. Patterns 9, 371–380 (2009).

  37. 37.

    , , & DeadEasy caspase: automatic counting of apoptotic cells in Drosophila. PLoS ONE 4, e5441 (2009).

  38. 38.

    & Trk receptors: roles in neuronal signal transduction. Annu. Rev. Biochem. 72, 609–642 (2003).

  39. 39.

    & Neurotrophin-mediated rapid signaling in the central nervous system: mechanisms and functions. Physiology (Bethesda) 20, 70–78 (2005).

  40. 40.

    & Severe peripheral sensory neuron loss and modest motor neuron reduction in mice with combined deficiency of brain-derived neurotrophic factor, neurotrophin 3 and neurotrophin 4/5. Dev. Dyn. 218, 94–101 (2000).

  41. 41.

    & Roles for NF-κB in nerve cell survival, plasticity, and disease. Cell Death Differ. 13, 852–860 (2006).

  42. 42.

    & Participation of Rel/NF-κB transcription factors in long-term memory in the crab Chasmagnathus. Brain Res. 855, 274–281 (2000).

  43. 43.

    , , , & NF-κB, IκB, and IRAK control glutamate receptor density at the Drosophila NMJ. Neuron 55, 859–873 (2007).

  44. 44.

    et al. Toll-like receptor 8 functions as a negative regulator of neurite outgrowth and inducer of neuronal apoptosis. J. Cell Biol. 175, 209–215 (2006).

  45. 45.

    et al. Expression and nuclear translocation of the rel/NF-κB-related morphogen dorsal during the immune response of Drosophila. C.R. Acad. Sci. III 316, 1218–1224 (1993).

  46. 46.

    et al. Sea anemone genome reveals ancestral eumetazoan gene repertoire and genomic organization. Science 317, 86–94 (2007).

  47. 47.

    et al. A reverse genetic analysis of components of the Toll signaling pathway in Caenorhabditis elegans. Curr. Biol. 11, 809–821 (2001).

  48. 48.

    Evolution of the neurotrophin signaling system in invertebrates. Brain Behav. Evol. 68, 124–132 (2006).

  49. 49.

    Tracing the evolution and function of the Trk superfamily of receptor tyrosine kinases. Brain Behav. Evol. 68, 145–156 (2006).

  50. 50.

    , & Toll-like receptor signaling in neural plasticity and disease. Trends Neurosci. 34, 269–281 (2011).

  51. 51.

    et al. Role of the Spatzle pro-domain in the generation of an active toll receptor ligand. J. Biol. Chem. 282, 13522–13531 (2007).

  52. 52.

    The Rel protein DIF mediates the antifungal but not the antibacterial host defence in Drosophila. Immunity 12, 569–580 (2000).

  53. 53.

    & Drosophila homeodomain protein dHb9 directs neuronal fate via crossrepressive and cell-nonautonomous mechanisms. Neuron 35, 39–50 (2002).

Download references


We thank C. Arnot, J. Wen and M. Wheatley for advice; S. Jondhale, J. Ng and S. Quayle for technical help; S. Bishop and K. Kato for comments on the manuscript; A.J. Courey (University of California, Los Angeles), J.L. Imler (Institut de Biologie Moléculaire et Cellulaire, CNRS, Strasbourg), T. Ip (University of Massachusetts), J.M. Reichhart (Institut de Biologie Moléculaire et Cellulaire, CNRS, Strasbourg), S. Sanyal (Emory University), R. Baines (University of Manchester), M. Freeman (University of Oxford), K. Ito (University of Tokyo), A. Chiba (University of Miami), B. Pfeiffer (Janelia Farm), M. Landgraf (University of Cambridge), C. Basler (University of Zurich), H. Broihier (Case Western Reserve University), D. Ferrandon (CNRS, Strasbourg), the Bloomington Stock Center and Iowa Hybridoma Bank for reagents; the Birmingham Mass Spectrometry Facility (Birmingham Science City, Advantage West Midlands); and Len Packman for mass spectrometry and Edman sequencing in Cambridge. The LTQ Orbitrap Velos ETD mass spectrometer used in this research was obtained through the Birmingham Science City Translational Medicine: Experimental Medicine Network of Excellence project, with support from Advantage West Midlands (AWM). This work was funded by a UK Medical Research Council Career Establishment Grant (MRCG0200140) to A.H., Wellcome Trust project grant (WT094175/Z/10/Z) to A.H. and N.J.G., Wellcome Trust equipment grant (WT073228/Z/03/Z) to A.H., Wellcome Trust programme grant (WT081744MA) to N.J.G., European Union Marie Curie International Incoming Fellowship (PIIF-GA-2010-274193-NPN) to J.S.W., UK Medical Research Council studentship to G.M. and Brunei government studentship to M.A.L.

Author information

Author notes

    • Graham McIlroy
    •  & Istvan Foldi

    These authors contributed equally to this work.


  1. School of Biosciences, University of Birmingham, Edgbaston, Birmingham, UK.

    • Graham McIlroy
    • , Istvan Foldi
    • , Jill S Wentzell
    • , Mei Ann Lim
    • , Janine C Fenton
    •  & Alicia Hidalgo
  2. Department of Biochemistry, University of Cambridge, Cambridge, UK.

    • Jukka Aurikko
    •  & Nicholas J Gay


  1. Search for Graham McIlroy in:

  2. Search for Istvan Foldi in:

  3. Search for Jukka Aurikko in:

  4. Search for Jill S Wentzell in:

  5. Search for Mei Ann Lim in:

  6. Search for Janine C Fenton in:

  7. Search for Nicholas J Gay in:

  8. Search for Alicia Hidalgo in:


G.M., I.F., J.A., J.S.W., M.A.L., J.C.F. and A.H. performed experiments; A.H. and N.J.G. conceived and directed the project; A.H., N.J.G. and G.M. wrote the paper; all authors contributed to planning experiments and analyzing data and to discussions and improvements to the manuscript.

Competing interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to Alicia Hidalgo.

Supplementary information

PDF files

  1. 1.

    Supplementary Figures and Text

    Supplementary Figures 1–9 and Supplementary Tables 2 and 3

Excel files

  1. 1.

    Supplementary Table 1

    Supplementary Table 1

About this article

Publication history





Further reading