FTD/ALS-associated poly(GR) protein impairs the Notch pathway and is recruited by poly(GA) into cytoplasmic inclusions

Artificial DNA sequences were designed to express 80 repeats of GA, GR, and PR following a Flag tag at the N-terminus. As a control, a stop codon TAA was introduced instead of the ATG codon at the beginning of the open reading frame of GA. The DNA sequences were synthesized (Genewiz) and subcloned into pUASTattB vector between the BglII and XhoI sites. The UAS-(GR)₈₀ control construct was generated by site-directed mutagenesis (QuikChange II Site-Directed Mutagenesis Kits, Agilent Technologies), in which the first two nucleotides of the start codon ATG were mutated to TA to form the stop codon TAG (the a5263t_t5264a mutations primer: TCGTTAACAGATCTCCAC CTAGGATTACAAGGACGACGAC). Transgenic flies were made with transgenes UAS-(GA) ₈₀ control, UAS-(GA) ₈₀ , UAS-(GR) ₈₀ control, UAS-(GR) ₈₀ , and UAS-(PR) ₈₀ present on the second or third chromosome. The presence of transgenes in the fly was confirmed by sequencing the genomic region amplified by RT-PCR. Primers used were CTGCAACTACTGAAATCTGCCA (forward) and TGTCACACCACAGAAGTAAGGT (reverse). CTGCAACTACTGAAATCTGCCA (forward) was used for sequencing. All flies were raised at 25 °C on a standard diet. GMR-Gal4, OK371-Gal4, UAS-GFP/CyO, Vg-Gal4, and w ¹¹¹⁸ ; P{NRE-EGFP.S}5A flies were from the Bloomington Drosophila Stock Center. Notch ⁵⁴¹⁹ mutant and UAS-Notch ^FL7 flies were kindly provided by Dr. S. Artavanis-Tsakonas.

Wings were arbitrarily classified into four groups according to the strength of the phenotype (absent, weak, medium, and strong), as judged from the number of notches and the size of wing area lost.

UAS-(GA) ₈₀ control, UAS-(GA) ₈₀ , UAS-(GR) ₈₀ , and UAS-(PR) ₈₀ flies were crossed with OK371-Gal4, UAS-GFP/CyO flies at 18 °C to obtain flies expressing dipeptides in motor neurons. For the climbing assay, individual 3-day-old adult flies were placed into a 15-ml polypropylene centrifuge tube (CellTreat Scientific Products). After 1 min, each fly was lightly taped to the bottom of the tube and allowed to climb for 10 s. The climbing distance was scored as the average of five tests for each fly. For quantification of dendritic branching, ddaE sensory neurons were labeled with mCD8-GFP driven by 221-Gal4. The number of dendritic ends was counted at the third-instar larval stage even though the mCD8-GFP signal was reduced by (GR)₈₀ and (PR)₈₀.

Third-instar larval wing imaginal discs, brains or salivary glands were dissected in PBS and fixed in 4 % paraformaldehyde for 20 min at room temperature. After three washes in PBS, samples were permeabilized in PBS containing 0.5 % Triton X-100 (PBT) for 30 min at room temperature and then blocked in 0.5 % goat serum in PBT for 1 h at room temperature. Samples were then incubated with the primary antibody overnight at 4 °C, washed three times with PBT, and incubated with secondary antibody for 2 h at room temperature. The primary antibodies were mouse anti-Flag (Sigma; 1:500) and mouse anti-Wingless (Wg) (Developmental Studies Hybridoma Bank; 1:50). The secondary antibodies were goat anti-mouse Alexa 488 and Alexa 594 (Invitrogen; 1:200). DNA staining was carried out by incubating salivary glands with Quant-iT OliGreen ssDNA reagent (Life Technologies; 1:1000) at room temperature for 5 min. Alternatively, samples were loaded on slides with Vectashield containing 4′,6-diamidino-2-phenylindole (DAPI, Vector laboratories).

Published iPSC lines from two control subjects and three C9ORF72 carriers [1, 2, 34] were differentiated into cortical neurons as described [2]. Neuronal cultures were aged for 8 weeks before RNA extraction. For (GA)₈₀ and (GR)₈₀ expression experiments, neuronal cultures from one control line were aged for 4 weeks before transfection. Middle frontal gyrus brain samples from three healthy control subjects and eight C9ORF72 repeat expansion carriers were obtained from the UCSF Memory and Aging Center. Another three control subjects were from the Mayo Clinic Jacksonville and were used in a recent study [14]. The mean age at death was 78 ± 8 years in six control subjects and 64 ± 7 years in eight C9ORF72 repeat expansion carriers.

For transfection of 4-week-old iPSC-derived neurons and HeLa cells, we used 1.6 µg of plasmid DNA containing HA-(GA)₈₀, FLAG-(GR)₈₀, or empty vector (pcDNA 3.1, Life Technologies) and Lipofectamine 2000 (Life Technologies). Forty-eight hours after transfection, cells were fixed with 4 % paraformaldehyde and permeabilized with 0.2 % Triton X-100, blocked with 3 % bovine serum albumin, and incubated overnight at 4 °C with primary antibodies against HA (Roche; 1:600) or FLAG (Sigma; 1:1000). After three washes with PBS, the cells were incubated with Alexa Fluor-conjugated secondary antibodies (Invitrogen; 1:300) for 1 h at room temperature and mounted on glass slides with Vectashield HardSet Mounting Medium with DAPI. Immunostained cells were examined by fluorescence microscopy.

The images of adult eyes and wings were acquired with a Nikon DS-Fi1 camera and a Nikon SMZ1500 stereomicroscope. The image of the dorsoventral boundary was acquired with a Nikon D-Eclipse C1. All other images were acquired with a Leica TCS SP5 II laser scanning confocal microscope and Leica LAS AF software. To quantify Wg fluorescence intensity in wing imaginal discs after staining, 30 Z-stack images (step size 0.5 μM) were acquired for each sample. Image-J was used for Z projection, and the integrated intensity of the same region at the dorsoventral boundary of six discs was measured per genotype.

To determine which DPR is toxic in vivo and identify the mechanisms, we generated DNA constructs containing 80 copies of GGXGCX, GGXCGX, or CCXCGX (with X being randomly any one of the four nucleotides) that can be transcribed under the control of UAS elements and translated into (GA)₈₀, (GR)₈₀, or (PR)₈₀, respectively. All constructs contained the CCACC consensus Kozak sequence adjacent to the ATG start codon and a DNA sequence encoding the Flag tag at the N-terminus of DPRs (Fig. 1a; Tab. S1). To generate transgenic fly lines, we used the PhiC31 integrase-mediated site-specific integration system [5] to ensure equal transcription of different DPR constructs in a cell-type specific manner when a unique Gal4 driver is used [8].

To dissect the toxicity of DPRs, we first used GMR-Gal4, which drives ectopic gene expression mostly in the eye, a widely used model for studying neurodegeneration [6, 24]. Expression of (PR)₈₀ and (GR)₈₀, but not (GA)₈₀, driven by GMR-Gal4 resulted in a mostly adult lethal phenotype and caused a drastic deformation of eyes in all flies at the pupal stage; neither the GMR-Gal4 nor any of the UAS transgenic lines alone had any eye phenotype (Fig. 1b). More than 200 flies were examined for each genotype. To confirm that different DPRs were expressed in the eye tissue, we performed RT-PCR analysis using primers specific to the 3′UTR sequence common to all DPR constructs. Indeed, all DPR mRNAs were expressed, although at different levels (Fig. S1), which is probably due to differences in mRNA stability. The stability, degradation, and biophysical properties of (GA)₈₀ proteins may also differ from those of (PR)₈₀ and (GR)₈₀. Indeed, immunostaining analysis with Flag tag-specific antibody showed all three DPRs were expressed when OK371-Gal4 was used (Fig. S2a–c), which drives target gene expression in cholinergic neurons including motor neurons. However, unlike (GR)₈₀ (Fig. S2b) and (PR)₈₀ (Fig. S2c), (GA)₈₀ forms distinct inclusions, mostly one in each neuron (Fig. S2a). Thus, although the extent of toxicity between different DPRs cannot be compared directly, lack of (GA)₈₀ toxicity is correlated with its inclusion formation, while more diffuse (GR)₈₀ and (PR)₈₀ cause drastic phenotypes under the same experimental condition.

To further exclude the possibility that (GR)₈₀ toxicity might be caused by transcription of GC-rich RNAs in vivo, we also generated transgenic fly lines that expressed the same (GR) ₈₀ mRNA except that the AUG start codon was replaced with the UAG stop codon (Fig. 1a). Expression of this control construct from the same genomic locus under the same Gal4 driver was not toxic (Fig. 1b). Thus, (GR)₈₀ and (PR)₈₀ proteins are highly toxic when overexpressed in the Drosophila eye by the GMR-Gal4.

Since GMR-Gal4 drives target gene expression at a high level, the drastic toxicity of (GR)₈₀ and (PR)₈₀ in the eye should be considered in the context of their overexpression. Therefore, we searched for Gal4 drivers that would give rise to a more moderate phenotype, enabling us to study the mechanism. Expression of the (GA)₈₀ construct by various Gal4s did not result in any obvious phenotypes (Table S2). However, expression of (GR)₈₀ and (PR)₈₀ in all cells by tubulin-Gal4, in all neurons by elav-Gal4, or in motor neurons by D42 or OK371-Gal4 at 25 °C resulted in a mostly lethal phenotype (Table S2). The strength of Gal4 drivers is dependent on temperature. Lower expression of (GR)₈₀ and (PR)₈₀ by OK371-Gal4 at 18 °C caused a semi-lethal phenotype (Table S2), and surviving adult flies had greatly reduced locomotor activity (Fig. 2a). Moreover, expression of (GR)₈₀ and (PR)₈₀ in ddaE sensory neurons decreased dendritic branching (Fig. 2b). (GR)₈₀-control was not used in these experiments because it did not produce any eye phenotype (Fig. 1b). Expression of (GR)₈₀, but not the (GR)₈₀ control construct, by Vg-Gal4, which drives gene expression in the dorsoventral boundary and some other cells of the wing imaginal discs [11], gave rise to wing margin defects in 90 % of flies of both sexes at 25 °C (Fig. 2c, d). Thus, (GR)₈₀ is toxic in multiple neuronal and non-neuronal cell types in vivo.

Misregulation of several signaling pathways results in distinct wing defects [4]. However, the mild notching defects at or near the tip of the wing caused by (GR)₈₀ are remarkably similar to the defect due to partial loss of Notch activity, as in flies heterozygous for the N ⁵⁴¹⁹ allele [20]. Therefore, we investigated the genetic interaction between Notch and (GR)₈₀. We grouped wing notching phenotypes by their severity: absent, weak, medium, and strong (Fig. 3a). Since the Notch gene is located on the X chromosome, we examined only female flies in this genetic interaction experiment. About 90 % of female flies heterozygous for the N ⁵⁴¹⁹ allele had a weak wing margin defect (Fig. 3b). To facilitate the genetic interaction study, Vg-Gal4 and UAS-(GR) ₈₀ were recombined onto the same second chromosome. The wing notching phenotype in the resulting fly line (Fig. 3b) was stronger than that in flies transheterozygous for Vg-Gal4 and UAS-(GR) ₈₀ (Fig. 2c). When this line was examined in the N ⁵⁴¹⁹ /+ background, the wing notching phenotype was substantially stronger than would be expected from an additive effect of (GR)₈₀ and N ⁵⁴¹⁹ /+: 18 % of (GR)₈₀ flies versus 47 % of (GR)₈₀ flies in the N ⁵⁴¹⁹ /+ background showed a strong wing margin phenotype (p < 0.05, n = 3), raising the possibility that the two genes genetically interact (Fig. 3b). These results suggest that (GR)₈₀ compromises the Notch signaling pathway. Indeed, (GR)₈₀ toxicity was largely suppressed by ectopic expression of full-length Notch; a wing notching phenotype was absent in 67 % of (GR) ₈₀ flies when Notch was co-expressed but in only 12 % of (GR) ₈₀ flies when GFP was co-expressed (p < 0.001, n = 3) (Fig. 3c–e). (PR)₈₀ toxicity was not suppressed by Notch expression, and PR₈₀ expression in N ⁵⁴¹⁹ /+ background did not significantly enhance the PR₈₀ wing defect (not shown), suggesting divergent pathogenic mechanisms of these two DPRs. To ensure that the suppression by Notch was not due to the dilution of Gal4 caused by a second copy of the UAS transgene, flies expressing both (GR)₈₀ and GFP by Vg-Gal4 were used as controls in this experiment (Fig. 3c). These genetic analyses suggest that the Notch pathway is a major target of (GR)₈₀ toxicity in vivo.

To confirm (GR)₈₀ indeed suppresses Notch signaling, we expressed a Notch activity reporter, UAS-Notch Response Element (NRE)-EGFP [33], in the wing disc cells by Vg-Gal4. Notch signaling was generally lower in all cells at the dorsoventral boundary that express UAS-(GR) ₈₀ (Fig. 4b) than in control flies expressing UAS-(GR) ₈₀ Control (Fig. 4a). The dorsoventral boundary remained intact, and no obvious cell loss was observed, as judged from caspase-3 immunostaining. To further confirm this finding, we examined the expression level of endogenous Wingless (Wg) at the dorsoventral boundary of the wing disc, which is a direct target of Notch signaling [12]. Indeed, (GR)₈₀ significantly suppressed endogenous Wg expression (Fig. 4c, d). Although we cannot rule out the possibility that (GR)₈₀ affects the Wg level through other mechanisms, this result is consistent with the finding that (GR)₈₀ decreases endogenous Notch signaling.

Poly(GR) proteins have been detected mostly in the cytoplasmic inclusions in brain tissues of patients with C9ORF72 repeat expansion [3, 30, 42], thus it is possible that Notch signaling in patient neurons may also be compromised. In recent years, patient-specific iPSCs have been used as a powerful model for many neurodegenerative diseases [18, 38, 39]. We differentiated iPSC lines from two control and three subjects with C9ORF72 repeat expansions [1, 2, 34] into 8-week-old postmitotic neurons of cortical lineage and found reduced expression of the Notch target genes HES1 (Fig. 4e) and HEY1 (Fig. 4f). HEY1, one of the target genes more sensitive to Notch level [35], showed reduced expression in brain tissues of C9ORF72 patients (Fig. 4f). The expression of Notch target NFKB1 was also lower in both iPSC-derived neurons and brain tissues of C9ORF72 patients (not shown). Although these results are correlative, Notch signaling seems to be compromised in patient cells as well.

To further examine the effects of (GR)₈₀ on Notch signaling, we examined the subcellular localization of (GR)₈₀ expressed in a subset of wing disc cells by Vg-Gal4 (Fig. S2d). Immunostaining with either anti-Flag antibody or anti-GR antibody revealed that this DPR was localized in the cytoplasm (Fig. S2e and g), suggesting that (GR)₈₀ interferes with the Notch signaling pathway in the cytoplasm. In both wing disc cells (Fig. S2f) and salivary gland cells (Fig. S3d), (PR)₈₀ also seemed to be mostly cytoplasmic.

Since both poly(GA) and poly(GR) proteins are concomitantly expressed in patient cells [3, 30, 42], we first co-expressed (GA)₈₀ and (GR)₈₀ in Drosophila salivary gland cells, which are large and facilitate imaging analysis. (GA)₈₀ by itself formed mostly cytoplasmic inclusions (Fig. 5a), while (GR)₈₀ was localized throughout the cytoplasm in salivary gland cells (Fig. 5b), as in wing disc cells (Fig. S2e). This subcellular distribution of (GR)₈₀ in salivary gland cells was confirmed with the anti-GR antibody (Fig. S3e). We also noticed one or two small (GR)₈₀-positive dots on chromatin in each salivary gland cell (Fig. 5b, Fig. S3e), as well as in Drosophila motor neurons (Fig. S3f). Their nature and significance remain to be determined. Shorter GR forms have been reported to accumulate in nucleoli of cultured cells [21], but we found no significant presence of (GR)₈₀ in nucleoli of salivary gland cells, even though their nucleoli were enlarged (Fig. 5b). But unexpectedly, in the presence of HA-(GA)₈₀ (Fig. 5c, left panel), a portion of Flag-(GR)₈₀ formed cytoplasmic inclusions as well (Fig. 5c, middle panel), in contrast to the diffuse cytoplasmic localization when only Flag-(GR)₈₀ was expressed (Fig. 5b). Because (GA)₈₀ alone forms cytoplasmic inclusions (Fig. 5a), co-localization of Flag-(GR)₈₀ and HA-(GA)₈₀ suggested that (GA)₈₀ recruits (GR)₈₀ into these inclusions (Fig. 5c, right panel).

Similarly, in human HeLa cells, (GA)₈₀ alone formed inclusions (Fig. 6a), while (GR)₈₀ alone was localized throughout the cytoplasm without any detectable accumulation in the nucleus (Fig. 6b). (GA)₈₀ inclusions seemed to be aggresomes surrounded by vimentin and were often close to the nucleus (Fig. S4a) and positive for p62 (Fig. S4b). When (GA)₈₀ and (GR)₈₀ were co-expressed, some (GR)₈₀ co-localized with (GA)₈₀ inclusions in all HeLa cells we examined (Fig. 6c). Moreover, in iPSC-derived human neurons, (GA)₈₀ recruited (GR)₈₀ into cytoplasmic inclusions as well (Fig. 6d–f). These findings raise the possibility that (GA)₈₀ protein is protective and can sequester highly toxic poly(GR) protein into inclusions. Indeed, (GA)₈₀ expression partially suppressed the (GR)₈₀-induced cell-loss phenotype at the wing margin (Fig. 6g) and correspondingly increased the Notch signaling, as indicated by the elevated expression of Wg at the dorsoventral boundary of the wing disc (Fig. 6h–j). (GA)₈₀ expression did not suppress the (GR)₈₀-induced eye phenotype (not shown), presumably because GMR-Gal4 is a very strong driver so that the level of non-aggregated (GR)₈₀ remained high. Thus, (GA)₈₀ partially suppresses (GR)₈₀ toxicity in vivo by sequestering (GR)₈₀ into inclusions.