The vast majority of the molecules critical to neuronal function are conserved between Drosophila and more complex vertebrate systems, thus Drosophila is a powerful model to study neurodegenerative diseases including ALS. In this study, we showed that the overexpression of human Fus as well as the Drosophila homolog Caz caused neuronal toxicity in vivo in all Gal4 lines that we tested, including ubiquitous expression (act5C-Gal4), wings (MS1096-Gal4), eyes (GMR-Gal4), neurons (Elav-Gal4) and motor neurons (D42- and OK371-Gal4). Since Fus/Caz protein was expressed at late larval or pupal stages and the degenerative phenotypes were progressive with aging, the Fus/Caz overexpressing flies are a reasonably good model to study the relevance of Fus in motor neuron degeneration in ALS. Furthermore, we provide evidence that both loss of Caz and overexpression of Fus/Caz can induce phenotypic defects although the underlying cellular mechanisms are likely to be different in these two models.
Nuclear toxicity of Fus/Caz in the absence of cytoplasmic inclusions
It has been recently shown that the overexpression of Fus in flies leads to neuronal toxicity [
25‐
27]. However, it is unknown whether overexpression of
Drosophila Caz can cause toxicity. In this study, overexpression of fly Caz caused similar or even more severe toxicity in all experiments compared to overexpression of human Fus (Figures
1,
2,
3,
4). This is reminiscent of the studies of another ALS-implicated RNA binding protein TDP-43 that the overexpression of TDP-43 was shown to be toxic in yeast [
33],
Drosophila [
23,
24,
34,
35],
C. elegans [
36], and mouse [
37,
38]. It is noted that protein inclusions of wild-type Fus have also been reported in human ALS patients [
7,
8]. These findings underscore the importance of understanding the underlying molecular mechanisms of Fus in ALS.
Our group and others have shown that the last 32 amino acids of the protein contain an NLS that is both essential and sufficient for nuclear targeting. It has also been shown that the ALS-related mutations in Fus increased the cytoplasmic localization of the protein and promoted the co-localization of Fus with stress granules in the cytosol [
19‐
21]. However, the cytoplasmic accumulation may not be essential or sufficient for the toxicity observed for the protein [
39,
40]. Thus, the link between toxicity and subcellular localization of Fus needs to be characterized
in vivo.
A novel finding of this study is that Fus nuclear localization is required for its toxicity in
Drosophila. The overexpression of FusΔ32 at comparable levels consistently showed little or no toxicity in all tissues. The
in vivo subcellular localization of wild-type Fus, R521G mutant and FusΔ32 was confirmed to be consistent with the findings in cell culture, i.e. FusΔ32 was largely diffused outside of the nucleus. Moreover, the fusion of an NES at the C-terminus of the full-length wild-type Fus dramatically reduced toxicity
in vivo (Figures
5G, I and
6F). FusNES was also diffused in the motor neurons (Additional file
1: Figure S2), thus the data from both FusNES and FusΔ32 suggest that the nuclear localization is critical to the toxicity. Furthermore, the chimeric protein FusΔ32NLS with an exogenous NLS was localized in the nucleus (Additional file
1: Figure S2) and produced similar toxicity as wild-type Fus (Figure
5H, I, and
6G). All these results consistently suggest that the nuclear localization is required for Fus toxicity.
This surprising result differs from the common presumption that the toxicity is caused by the cytoplasmic accumulation of Fus. Our study is the first one characterizing transgenic flies expressing truncated Fus mutant with a primarily cytoplasmic localization. We further manipulated the subcellular localization by using a typical NES and the other different NLS. All experiments consistently support the conclusion. Interestingly, another study showed that mutation in the NLS of TDP-43 relieved the toxicity in
C. elegans [
36], supporting the conclusion of our study. However, the study by Lanson
et al. employed another approach by deleting a postulated NES within the RNA recognition motif of Fus and showed reduced toxicity [
25]. It should be noted that Lanson
et al. did not characterize the NES or the subcellular localization of the NES deletion mutant. Moreover, the postulated NES is within the RNA binding domain in Fus, therefore such deletion may cause a disruption of the RNA binding properties of Fus. Thus, the mechanism of the reduced toxicity of that particular NES deletion mutant needs to be better elucidated. We are confident in our conclusion since the results using three different approaches (FusΔ32, FusNES, and FusΔ32NLS) consistently support the requirement of nuclear localization for Fus toxicity.
Another critical difference between this study and the study by Lanson
et al. is how the transgenic flies were generated. The site-specific transgenic approach using the integrase-mediated insertion at the specific attP locus allowed us to ensure equal expression of the Fus proteins without positional effects. We think that the discrepancy is likely due to the expression level of the protein. This likely explains the discrepancy regarding the toxicity of the Fus transgene in different studies. The R521G mutation produced comparable levels of toxicity as wild-type Fus whereas the other study showed more severe toxicity in flies expressing R521H or R521C [
25]. The experiments in our study are more likely to produce comparable expression levels of the Fus transgene since transgene insertion was random in other studies.
Disruption of muscle 6/7 NMJs was observed in the Fus transgenic flies in this study. It is important to note that NMJ vulnerability occurs selectively in different subgroups of motor neurons or muscles. The diversity of subtypes of motor neurons and muscles and their susceptibility are well documented in ALS [
41]. We examined the muscle 6/7 NMJs in the abdominal segment A2 of the third instar larvae, which is widely used in
Drosophila studies. In addition, we used GFP as a marker to monitor the target protein expression. Our results showed the expression of Fus/Caz in the disrupted NMJs. Studies characterizing NMJs in other muscles may obtain different results, for example another study examined muscle 4, segment A2 and 3 and did not observe reduction of synaptic boutons in Fus overexpressing flies [
25].
Notably, neither wild-type Fus/Caz nor the ALS mutants of Fus showed cytoplasmic inclusions in transgenic flies where severe toxicity was demonstrated in this study. The results suggest that the cytoplasmic inclusions are not essential to Fus toxicity although they have been prominently observed in cell culture systems and in human patient tissues. It has long been proposed that soluble oligomers may be the culprit species causing neurodegenerative disease. A recent study showed that aberrant high molecular weight complex of TDP-43 was toxic, although the nature of the species was still unknown [
42]. A recent study also showed that Fus toxicity could be suppressed without eliminating protein inclusions in yeast [
40], suggesting that protein inclusions might be a non-toxic and non-essential feature of the disease.
Although the results in this study consistently suggest that sufficient steady-state levels of Fus/Caz are required for the toxicity and neurodegenerative phenotypes, it is noted that Fus/Caz undergoes dynamic trafficking in and out of the nucleus in live cells. It remains possible that Fus/Caz with steady-state nuclear localization could produce detrimental effect while it is temporarily outside of the nucleus. Alternatively, the minute amount of Fus/Caz in the cytoplasm that cannot be detected by confocal microscope could also produce cytosolic toxicity. We believe that these two are remote possibilities, however additional studies using more sophisticated approaches are needed to test them in the future. Moreover, the exact nature of Fus toxicity in the nucleus also remains to be determined in future studies.
The phenotypes of caz deletion in Drosophila
Deletion of a specific gene provides insight into understanding the
in vivo function of the protein. By taking advantage of the
Drosophila model, we found that endogenous Caz is required for normal functions of the neurons (Figure
7). The deficiency of
caz in flies induced a strong eye phenotype with disrupted ommatidia structure (Figure
7A-I) and caused defects in the locomotive function in larvae (Figure
7K) and adult flies (Figure
7J). Consistently,
caz1 mutant flies exhibited decreased adult viability and diminished locomotive activity [
27]. These phenotypes were actually similar to those caused by Fus/Caz overexpression. It would be interesting to examine whether some cases of ALS disease may be caused by Fus loss-of-function in patients.
Given the motor deficiencies of Fus overexpression and mutant
caz flies, the presynaptic structure in NMJ was examined. As expected, the number of boutons was significantly reduced in the larvae overexpressing Fus (Figure
4) as well as in the larvae lacking
caz (Figure
7L-M). Thus the disruption of NMJ was evidently the cause of the locomotive defects. Therefore, the Fus/Caz overexpression and loss of
caz obviously caused similar disruption of presynaptic terminals in NMJs. It is likely that Caz is required for neuronal function and therefore the lack of Fus/Caz perturbs the motor neuron distal terminus and causes NMJ disruption.
Different molecular mechanisms for degenerative phenotypes in Fus overexpression and caz deletion Drosophila
By examining the nuclear structure and monitoring apoptosis of cells in the VNC, we found that overexpression of Fus/Caz induced apoptotic cell death (Figure
6). However, neither the
caz deficiency (
BSC759) nor the
caz1 mutation led to apoptosis (Figure
7P and
7S, data not shown). Despite the critical difference, both Fus/Caz overexpression and
caz deletion disrupted the NMJs. This raises the question of where the toxicity originates, either from the death of neuronal cell bodies or disruption of the NMJs.
Studies in the SOD1 mutant-mediated familial ALS have suggested an axon die-back model in which perturbation at the distal NMJs occurs prior to the death of motor neurons [
41]. The Fus/Caz overexpression phenotypes support this model, but the time sequence of the NMJ disruption and motor neuron death has not been distinguished in these flies as yet. Moreover, both the
caz deficiency and
caz1 mutant flies strongly support this model since loss of Fus/Caz caused NMJ disruption and motor function deficiency without the presence of motor neuronal cell death. The findings in our study support two potential mechanisms for the Fus/Caz mediated neurotoxicity: neuronal apoptosis and NMJ perturbation. These two mechanisms are not necessarily mutually exclusive, and in fact may occur simultaneously. Further studies exploring of the in-depth mechanisms by which Fus/Caz causes alterations in RNA processing and nuclear toxicity are currently ongoing. Many laboratories in the field also use various models, including yeast [
39,
40,
43,
44] and
Drosophila [
25‐
27], to investigate the role of Fus in RNA metabolism and the genetic interactions between Fus and other RNA processing proteins. The overexpression and deletion
Drosophila models in this study provide valuable insight into the etiology of ALS and potentially other degenerative diseases in which RNA processing proteins are implicated.