CHAPTER IV
DISCUSSION
The C. elegans genome contains a single homolog of vertebrate neuroligins. The nlg-1 gene encodes a set of proteins quite similar in overall structure to vertebrate neuroligins: a type I transmembrane protein with a signal sequence followed by a cholinesterase-like extracellular domain, extracellular proline-rich and O-linked sugar motifs, and a relatively small intracellular domain terminating in a PDZ-binding motif (Figure 7 and Figure 9).
In contrast to vertebrate neuroligins, where alternative splicing affects primarily extracellular parts of the protein and thus affects binding interactions with neurexins (Ichtchenko et al., 1995, Boucard et al., 2005, Chih et al., 2006, Comoletti et al., 2006, Ichtchenko et al., 1996), alternative splicing of C. elegans nlg-1 transcripts leads almost exclusively to diversity in the intracellular domain of the protein (Figure 8 and Figure 9). Based on sequence analysis of a limited number of independent cDNAs, it appears that exon 13 is more likely to be included than skipped (it is present in six of six cDNAs), and exon 14 is more likely to be skipped than included (it is present in only one of six cDNAs). We have not yet explored the significance and regulation of the extensive alternative splicing of nlg-1 transcripts. However, it is possible that the isoform diversity generated by alternative splicing of nematode neuroligin corresponds to the diversity provided by multiple neuroligin genes in mammals. For example, we note that the intracellular domain of mammalian Neuroligin-2 has a proline-rich region not found in the other mammalian neuroligins. Similarly, exon 14 of C. elegans nlg-1 encodes a proline-rich region; perhaps nematode isoforms containing exon 14 correspond to mammalian Neuroligin-2. If this is true, we would predict that animals lacking exon 14 would have synapse maintainence and maturation issues for only GABAergic neuromuscular junctions.
Our reporter studies have positively identified 33 of the approximately 45 neuroligin expressing neurons in C. elegans. These are the 4 URA, 2 URB, 2 AIY, 2 PVD, 2HSN, 9 DA and 12 VA neurons. We have also excluded a number of neurons in the head, including AFD, AWA and AWC (Jessica Heatherly, personal communication), IL2, RME, AIZ, AIB, AVG, RIA, RIF, RIV and ASG. In addition, we have excluded the AS, DB, VB, DD and VD neurons in the ventral nerve cord. Current neurons of interest for the remaining unidentified neurons include the motor neuron/interneurons: RMH and RMF; the interneurons: AVB, AVF, AVK, RIB, RID and RIG; the sensory neuron: ASE; and neurons of unknown function: AVH and ALA. (for a summary, see Figure 12)
Other than the nlg-1-expressing cells in the ventral nerve cord, which are all cholinergic motor neurons, the nlg-1-expressing cells elsewhere in the animal do not fall neatly into a single neuron class or neurotransmitter type. The AIY and URB cells are interneurons, the PVD cells are sensory neurons and the URA and HSN cells are motor neurons (Figure 10, Figure 11 and Figure 12). Furthermore, the AIY interneurons are cholinergic (Altun-Gultekin et al., 2001), the PVD neurons are glutamatergic (Lee et al., 1999a), and the HSN cells release both serotonin and (weakly) acetylcholine (Desai et al., 1988, Duerr et al., 2001). Neurotransmitter assignments have not been reported for the remaining nlg‑1-expressing neurons; however, they do not express GABAergic, dopaminergic, serotonergic or (eat-4-positive) glutamatergic reporters (see Methods).
In addition to neurons, we observed nlg-1 expression in the body wall muscles; muscle expression has also been reported for human neuroligin-4 (Bolliger et al., 2001). We also note that, unlike vertebrate muscles, nematode body muscles send neuron-like processes ("muscle arms") to the nerve cords to receive input (White et al., 1976). These nematode neuromuscular junctions therefore correspond more closely anatomically and structurally to vertebrate central cholinergic synapses than they do to vertebrate neuromuscular junctions.
We have also found that, in at least some neurons, neuroligin is present presynaptically, where it colocalizes with the synaptic vesicle marker synaptotagmin (Data not shown). The presynaptic localization requires the function of the UNC-104 and UNC‑116 motor proteins. The presence of neuroligin at presynaptic sites in C. elegans has been previously reported (Feinberg et al., 2008).
Although nlg-1 mutants display superficially normal growth, locomotion and nervous system structure, we identified a number of specific behavioral and biochemical nlg-1 mutant phenotypes. For each of these phenotypes, we have documented phenotypic rescue by transgenic expression of an NLG-1 cDNA and/or an NLG-1::YFP fusion protein. In addition, measurements of thermal response and spontaneous reversal using the nlg-1 deletion allele tm474 are the same as measurements using the ok259 deletion allele. Together, these data argue strongly that the phenotypes we report are due solely to loss of nlg-1 gene function.
As mentioned in Results, the most obvious behavioral output of C. elegans is its locomotion. C. elegans explores its environment using a biased random locomotion. The animals move by generating sinusoidal waves along the length of their bodies. When these waves are propagated from head to tail the animal moves forward, and when propagated from tail to head the animal moves backward. The changes of direction are called "reversals." These reversals can occur spontaneously, or can be induced as a direct response to a sensory stimulus. The animals move forward for some amount of time and then spontaneously reverse direction, backing up for approximately three body-lengths. When the animal resumes forward movement, it is in a random direction, and so this behavior is also termed a “pirouette” (Pierce-Shimomura et al., 1999). C. elegans can also execute "omega" turns, during forward or reverse locomotion. The body of the animal bends such that the head touches the tail (causing the animal to trace a path shaped like the Greek letter Ω) (Wallace, 1969, Croll, 1975). The animal then proceeds in a new direction. Animals may bias their forward locomotion ventrally or dorsally, causing them to trace an arc or spiral. Such biased locomotion is referred to as “veering” or “shallow turns.” In general, navigation is achieved by 33% “reversals”, 10% “omega turns” and 57% “shallow turns.” Presumably, these mechanisms for changing the bearing of the animal allow it to navigate in a complex soil environment (Zhao et al., 2003). When the worm is in an environment with plentiful food, it slows considerably, and spontaneous reversals are suppressed (Zhao et al., 2003). This state has been termed dwelling, and is mediated by serotonin and dopamine (Hills, 2004). When the worm is removed from food, it switches to a state called local search, area restricted search, or pivoting (Gray et al., 2005). This state is characterized by rapid movement in short runs with frequent reversals and deep omega turns. The animals gradually, over the next thirty minutes, shift to another state called dispersal or traveling (Wakabayashi et al., 2004).
When C. elegans is removed from food, the area restricted search portion of the behavior is mediated by the AWC olfactory neurons, the ASK gustatory neurons and the AIB interneurons (Wakabayashi et al., 2004). The animals gradually, over the next thirty minutes, shift to a dispersal or traveling state. The ASI gustatory neurons and the AIY interneurons suppress omega turns and reversals (Wakabayashi et al., 2004, Hills, 2004, Gray et al., 2005). However, it is not obvious how or if this is related to the reduced frequency of spontaneous reversal observed in nlg-1 mutants. We note that nlg-1 mutants exhibit apparently normal touch-induced reversal behavior; in addition, once initiated, backwards locomotion (whether induced or spontaneous) seems to be completely normal. It therefore seems that the mutant phenotype is associated with the "decision" to back up, rather than the execution of the backwards locomotion motor program per se. Although the circuitry involved in touch-induced reversal has been well-characterized (Chalfie et al., 1985b), the circuitry associated with spontaneous reversals has not yet been fully determined; however, control of spontaneous reversal has been shown to involve multiple transmitters (e.g., glutamate, dopamine, and acetylcholine) (Hills, 2004, Gray et al., 2005, Mullen et al., 2007). Our data suggest that spontaneous reversal circuitry contains some element(s) not found in the induced-reversal circuit.
In the ventral nerve cord, expression of nlg-1 is limited to the DA and VA neurons (Figure 10 and Figure 12). However, although these are the motor neurons which mediate backwards locomotion (Chalfie et al., 1985b), we are not sure how or if this is related to the reduced frequency of spontaneous reversal observed in nlg-1 mutants, since the mutants display apparently normal touch-induced reversal behavior and normal backwards locomotion once it is initiated. The neuroligin-expressing AIY interneurons (Figure 11) also play a role in the regulation of spontaneous reversal. In AIY ablated animals, or ttx-3 mutants (in which AIY interneurons do not differentiate), there is an increased frequency of spontaneous reversal (Altun-Gultekin et al., 2001). However, as already described in the Results section (Figure 16), nlg-1 mutants have a decreased frequency of spontaneous reversals, whereas ttx-3 mutants have an increased frequency of spontaneous reversals. Thus, loss of NLG-1 from AIY cells does not mimic loss of AIY function.
Some sensory behaviors of C. elegans have been assigned to specific, relatively simple neuronal circuits. Thermotaxis is one of the most interesting behaviors in C. elegans. This behavior, reported first by Hedgecock and Russell (Hedgecock and Russell, 1975), requires a mechanism for sensing temperature, the association of recent feeding state with that temperature and persistence of the association memory for at least four hours. One pathway has been proposed by Mori et al. (Mori and Ohshima, 1995). Thermal sensation is mediated by the pair of AFD sensory neurons (Mori and Ohshima, 1995); the synaptic output of AFD is primarily to the pair of AIY interneurons, whose output, in turn, is primarily to the AIZ and RIA interneurons (White et al., 1986). The only neuroligin expressing neurons in this circuit are the pair of AIY interneurons. It is noteworthy that laser ablation of each of these cell types individually does not mimic the nlg-1 mutant phenotype. When AFD neurons are ablated, ~50% of the worms are cryophilic and ~50% are athermotactic. When AIY neurons are ablated, the worms are cryophilic. When AIZ neurons are ablated, the worms are thermophilic. When AFD and AIY neurons are ablated, or when AIY and AIZ neurons are ablated, the animals barely move. However, ablation of the AFD and AIZ cells together provides a close resemblance to the completely athermotactic behavior of nlg-1 mutants (Mori and Ohshima, 1995). We do not yet know if the insensitivity to temperature associated with the loss of neuroligin function (as well as each of the other phenotypes discussed below) results from failure of a single synaptic connection, malfunction of an individual neuron, or disruption of one or several neuronal circuits.
Aversion to 1-octanol by wild-type animals is mediated by the ADL and ASH sensory neurons (Troemel et al., 1995); both of these cell types have synaptic output to the AIA and AIB interneurons (White et al., 1986). However, none of these four cell types express neuroligin, so we have no clear model of this nlg-1 mutant phenotype. The most plausible explanation is that NLG-1 facilitates or inhibits the secretion of a soluble factor acting at a distance.
A particularly intriguing phenotype of nlg-1 mutants involves the integration of conflicting sensory inputs in an "Approach-Avoidance" paradigm. The AWA sensory neurons mediate sensation of the volatile attractant diacetyl (Bargmann et al., 1993), and the ADL, ASE and ASH sensory neurons collectively mediate sensation of the repellent cupric acetate (Sambongi et al., 1999). When presented individually to nlg-1 mutants, each of these two compounds elicits completely normal sensory responses, suggesting that these sensory neurons (and at least some of their associated interneurons) are able to function appropriately. However, the nlg-1 response to simultaneous presentation of these two compounds is clearly not normal (Figure 19). Under the assay conditions we use, the repellency of the cupric acetate to wild-type animals appears to be stronger than the relative attractiveness of the diacetyl (80% of the animals "decide" not to cross the repellent barrier). In nlg-1 mutants, however, the attractiveness of the diacetyl "wins out" over the repellency of the cupric acetate (and 60% of the mutants are clearly "willing" to cross the repellent barrier). The mutant phenotype, therefore, should not be viewed as a failure to integrate the two signals; rather, the integration appears to use an alternative algorithm, which may reflect functional deficits in one or more of the "processing" interneurons.
We note that using a similar "Approach-Avoidance" (diacetyl/cupric acetate) paradigm, Ishihara et al. (2002) characterized altered sensory integration behavior associated with hen-1 mutants (Ishihara et al., 2002). However, although both nlg-1 and hen-1 mutants are similar to wild type in their attraction to diacetyl by itself and their avoidance of cupric acetate by itself, there is a dramatic difference between the two mutants when the two compounds are presented simultaneously. nlg-1 mutants are significantly more likely than wild-type to cross the cupric acetate barrier under these conditions (Figure 19), whereas hen-1 mutants are significantly less likely than wild-type to cross the barrier (Ishihara et al., 2002). hen-1 mutants also display deficits in behavioral plasticity mediated by paired stimuli (Ishihara et al., 2002), while nlg-1 mutants have wild-type responses using such paradigms (not shown); hen-1 mutants would thus seem to have a more general type of sensory integration deficit than nlg-1 mutants.
We have utilized an established model of oxidative stress in C. elegans (Yamamoto et al., 1996, Yanase et al., 2002) to demonstrate that neuroligin deficient mutants are hypersensitive to paraquat-induced oxidative stress (Figure 20). Although it was convenient to assess paraquat sensitivity by measuring survival (Figure 20), we also obtained similar results measuring swimming rate (Figure 40). We believe that this paraquat sensitivity results from an increased basal level of oxidative stress in nlg-1 mutants, and we have shown that nlg-1 mutants contain a higher level of protein carbonylation, a biochemical protein modifications characteristic of oxidative stress (Figure 20). A related nlg-1 mutant phenotype is hypersensitivity to the toxic effects of both organic and inorganic mercury compounds (Figure 20). We do not yet understand how lack of neuroligin in a subset of neurons is able to sensitize the entire organism to toxins such as paraquat and mercury. There may be a circuit that is responsible for being the sentinel for oxidative stress, or the neuroligin ectodomain could be acting as a soluble factor confiring resistance to oxidative stress.
Some studies have reported the presence of biomarkers associated with oxidative in individuals with autism (Reviewed in (Chauhan and Chauhan, 2006, Kern and Jones, 2006, Deth et al., 2008)). James et al. (James et al., 2006, James et al., 2004) analyzed plasma collected from ASD-affected individuals. They reported a metabolic profile consistent with impaired capacity for methylation (significantly lower ratio of S-adenosylmethionine to S-adenosylhomocysteine) and increased oxidative stress (significantly lower redox ratio of reduced glutathione to oxidized glutathione) in children with autism. Geier (Geier et al., 2009) also reported a lower ratio of reduced glutathione to oxidized glutathione in plasma of ASD affected individuals. Other studies have found an increase in 3-nitrotyrosine in cerebellar extracts (Sajdel-Sulkowska et al., 2008), increased urinary excretion of 8-isoprostane-F2α (a non-enzymatic oxidation product of arachidonic acid) (Ming et al., 2005) and increased plasma levels of malonyldialdehyde (an end product of peroxidation of polyunsaturated fatty acids and related esters) (Chauhan et al., 2004a, Chauhan et al., 2004b). Together these studies sugest a general increase in oxidative stress in people with ASDs
Human twin studies by Bailey (Bailey et al., 1995) and Muhle (Muhle et al., 2004) have shown that the concordance among monozygotic twins for a strict diagnosis of autism is 60%, which is about 12-fold higher than the concordance among dizygotic twins (Bailey et al., 1995, Muhle et al., 2004, Kern and Jones, 2006). This clearly provides strong evidence for a hereditary basis for autism, but also highlights the importance of non-hereditary (environmental) factors Our data on the sensitivity of neuroligin deficient mutants to oxidative stress (e.g., paraquat) and mercury compounds demonstrate a clear connection between the nematode equivalent of an autism associated synaptic mutation and hypersensitivity to environmental toxins. We believe that this provides an important model for understanding how both genetic and environmental contributions to a neurological disorder can have a single underlying basis.
There are many molecules with different yet overlapping functions involved in synapse formation. Although neuroligin clearly plays an important role, it is certainly not the only factor. The human families in which neuroligin defects are associated with autism presumably have a number of additional contributing genetic and environmental factors that collectively result in autism. It is now widely accepted that mutations affecting structural components of synapses (e.g., scaffolding molecules and adhesion proteins) provide a significant risk factor for the development of autism spectrum disorders; however, a credible mechanistic explanation for this is still lacking.
We undertook a structure-function study of NLG-1 to identify regions that are essential for neuroligin localization and/or function. We used three approaches to identify critical regions or motifs. First, we generated deletions within the intracellular domain of NLG-1. Second, we introduced neuroligin mutations associated with ASDs in humans into NLG-1. Third, we mutagenized putative phosphorylation sites or a kinase binding motif within the intracellular domain of NLG-1. As described in the Methods, these mutations were introduced into the rescuing NLG-1::YFP clone.
Through a systematic deletion analysis, we identified a portion of the intracellular domain of NLG-1 required for neuroligin function but not localization in C. elegans. Expression of a Δ 15-16 NLG-1::YFP fusion protein failed to rescue any of the nlg-1 mutant phenotypes tested. However, although fluorescence from the Δ 13-16 NLG-1::YFP fusion protein was faint, we observed puncta in the dorsal and sublateral nerve cords, suggesting that the protein was localized correctly. Within the intracellular domain, the constitutively expressed exons 15 and 16 (see Figure 8 and Figure 9) are required for neuroligin function, whereas the alternatively spliced exons 13 and 14 (see Figure 8 and Figure 9) appear to be dispensable. Finally, the PDZ-binding motif does not appear to be required for either the localization or function of neuroligin in C. elegans.
Structure-function studies suggest that there are several noteworthy differences between the mammalian and C. elegans neuroligin proteins. The C-terminal PDZ-binding motif appears to be required for postsynaptic assembly in mammals; Nam (Nam and Chen, 2005) noted reduced recruitment of PSD95 and AMPA receptors to nascent postsynaptic sites in neurons expressing neuroligin lacking the PDZ-binding motif. We note, however, that the in vivo behavioral consequences of deleting the PDZ-binding motif have not yet been examined in mammals. In the present study, we show that the PDZ-binding motif is dispensable for neuroligin function in C. elegans. The mutant neuroligin in the Nam study was expressed in neurons which contained endogenous neuroligins, whereas our transgenic animals only express the mutant ΔPDZ NLG-1. Furthermore, Dresbach et al. (Dresbach et al., 2004) identified an intracellular 11 amino acid region near the transmembrane domain that is required for rNLGN1 localization. In contrast, we find that the entire intracellular domain of C. elegans NLG-1 is dispensable for synaptic localization. However, since the level of NLG-1 fluorescence is reduced in Δ 13-16 NLG‑1 transgenic animals relative to controls, we believe that the intracellular region may contribute to protein expression or stability. We conclude that there is either no localization information in the intracellular domain of NLG‑1, or there is redundant localization information elsewhere in the protein or in the complex of multiple protein-protein interactions in which NLG-1 participates.
Although we do not currently have localization information for the mammalian neuroligins when heterologously expressed in C. elegans, the ability of hNLGN4 and rNLGN1 to rescue the nlg-1 mutant phenotypes argues strongly that they are localized correctly in worms. In the future, it would be useful to generate transgenic worms carrying functional mammalian neuroligin::GFP fusion proteins. For example, it would be informative to delete the 11 amino acid region identified in the Dresbach (Dresbach et al., 2004) study and determine the extent to which this mutant neuroligin localizes and functions in C. elegans (rescues the nlg-1 mutant phenotypes). This system could be very powerful for studying the in vivo consequences of amino acid substitutions or deletions on neuroligin localization and function.
We also introduced mutations into C. elegans NLG-1 that are homologous to the ASD-associated mutations in hNLGN3 and hNLGN4. Two of our mutant constructs, V397M and R714C, gave complete rescue of nlg-1 mutant behavioral phenotypes, whereas the R430C and R62W constructs only gave partial rescue (Figure 29, Error! Reference source not found., Figure 31 and Figure 32). Overall, this approach was not especially informative, since mutating random amino acids may have given a similar outcome. Another limitation is that we made the homologous mutation in the C. elegans nlg-1 cDNA, and the mutation may affect the protein in a species specific manner. Although NLG-1 is approximately 26-28% identical (45-47% similar) to the four human neuroligins, there are enough sequence differences that a given ASD-associated mutation may not have the same effect on the worm protein. Finally, C. elegans live at 15-25ºC whereas the human core body temperature is 37ºC. Mutations in human neuroligin presumably disrupt function at 37ºC. The mutations may have less effect at 20ºC. It is notable, however, that all four mutated NLG-1 proteins localized normally in C. elegans, suggesting that the mutant phenotypes arise from a functional problem with the protein, rather than a deficit in protein synthesis, trafficking or stability. Consequently, the R430C and R62W mutations may be worthy of additional study.
We also analyzed putative phosphorylation sites within the C-terminal domain. For this analysis, we generated two mutant constructs. In the first construct, we simultaneously eliminated two potential PKC phosphorylation sites. In the second construct, we mutated an LAL motif, a potential MEK/JNK binding site. In both cases, the resultant fusion proteins were localized to synaptic regions, and we observed discrete puncta in the dorsal and sublateral nerve cords. Furthermore, the fusion proteins rescued the entire battery of nlg-1 mutant phenotypes (data not shown). It is possible that these sites are not phosphorylated, or that phosphorylation does not play a role in the intracellular signaling, localization or functioning of neuroligin. Similarly, the LAL motif may not be a Mek/Jnk kinase-binding site, or that binding plays no role in the intracellular signaling, localization or functioning of neuroligin. It is also possible that the role of these motifs is not revealed by our limited number of behavioral phenotypes.
A significant limitation of our approach is that we cannot precisely control transgenic expression level. It is likely that the presence of multiple copies of the transgene results in overexpression, and this may mask the effect of a given mutation on NLG-1 function. In the context of alternative splicing, for example, we show that expression of a single isoform rescues all nlg-1 mutant phenotypes. However, it is possible that at physiological levels, the different NLG-1 isoforms might actually differentially mediate specific cellular processes or behaviors. Similarly, the ASD-associated mutations may have reduced function that is compensated for by overexpression. Any modulation of function due to phosphorylation or Mek/Jnk kinase binding potentially may have been masked by overexpression. In the future, some of these limitations could be addressed by using the new Mos-based single-copy insertion method (Robert et al., 2009). Rather than predicting what is important based on amino acid sequence, it is often easiest to have the organism tell us what is important. With the nlg-1 mutant phenotypes that I have identified and characterized in this study, such traditional genetic approaches may be now possible.