Defective protein folding and proteostatic stress are common pathogenic mechanisms linking genetically and anatomically diverse neurodegenerative diseases (Ling et al., 2013). The steady state levels—and ultimately neurotoxicity—of aggregation-prone proteins are determined through a balance of protein production and protein clearance (Prahlad and Morimoto, 2009). The convergence of aging-dependent declines in protein degradation and nuclear import with environmental stresses may push this equation toward irreversible protein aggregation that sets the stage for neurodegenerative processes (Morimoto, 2008). Thus, enhancing protein degradation capacity—or reducing the aggregation potential of aggregation-prone proteins—represents a promising therapeutic avenue for amyotrophic lateral sclerosis (ALS) and other neurodegenerative proteinopathies.

The highly conserved ubiquilin (UBQLN) gene family fulfills diverse roles in protein folding, shuttling, and degradation (Zheng et al., 2020; Rothenberg and Monteiro, 2010). All eukaryotic UBQLNs feature an amino-terminal ubiquitin-like (UBL) domain and a carboxyl-terminal ubiquitin-associated (UBA) domain separated by a low complexity, methionine-rich central region harboring variable numbers of STI1-like repeats first identified in the yeast stress-inducible 1 (Sti1) protein and later described in several distinct classes of protein chaperones (Fry et al., 2021; Zientara-Rytter and Subramani, 2019; Howe et al., 2020; Schmid et al., 2012; Li et al., 2013). Mammals encode four ubiquilin proteins of which three (UBQLN1, UBQLN2, and UBQLN4) are widely expressed. UBQLN1 and UBQLN2 share >70% amino acid identity and are presumed to function in semi-redundant fashion, whereas UBQLN4 is a more distantly related paralog.

Although specific functions of individual ubiquilins in mammals are largely unknown, a generic model for UBQLN function holds that the UBA domain engages ubiquitylated substrate while the UBL domain engages the proteasome, leading to substrate degradation (Itakura et al., 2016). STI1-like repeats form a hydrophobic groove that is thought to engage hydrophobic regions of client proteins (Fry et al., 2021). It has been reported that UBQLN deficiency (Df) leads to defects in autophagy and ER-associated protein degradation (Rothenberg et al., 2010; Lim et al., 2009; N’Diaye et al., 2009; Lee et al., 2013). The central methionine-rich domain has been shown to bind transmembrane domains of mitochondrial proteins, which appears central to their UBQLN-dependent triage and degradation (Itakura et al., 2016).

Interest in ubiquilin function was greatly stimulated by the discovery that dominant mutations in UBQLN2 cause X-linked ALS/frontotemporal dementia (FTD) (Deng et al., 2011; Gellera et al., 2013). Most ALS-associated mutations in UBQLN2 are clustered within 42-amino acid proline-rich repeat (PRR) that is unique to UBQLN2 (Deng et al., 2011) however, disease-linked mutations outside this region have also been described (Synofzik et al., 2012; Daoud et al., 2012). In addition, UBQLN2-mutant patients exhibit a range of phenotypes that includes FTD, ALS, and spastic paraplegia (Gkazi et al., 2019). Interestingly, ubiquilin-positive aggregates are a near universal occurrence in TDP-43-positive ALS, as well as ALS linked to C9ORF72 expansions (C9-ALS) (Brettschneider et al., 2012). These correlative findings suggest that ubiquilin pathology may contribute to the molecular pathogenesis of ALS even in the absence of UBQLN2 gene mutations.

Transgenic or virus-directed expression of UBQLN2^ALS mutants in rodents recapitulates protein UBQLN2 inclusions seen in ALS/FTD patients and elicits variable phenotypic abnormalities ranging to mild gait and memory defects to neuronal loss, paralysis, and early death (Gorrie et al., 2014; Wu et al., 2015; Huang et al., 2016; Le et al., 2016; Sharkey et al., 2020). Coexpression of UBQLN2^P497H and an ALS-associated TDP-43 allele under control of the neurofilament heavy gene promoter caused severe motor neuron loss and muscle wasting (Picher-Martel et al., 2019). Brain-directed expression of wild-type UBQLN2 also caused toxicity phenotypes (Huang et al., 2016; Le et al., 2016; Sharkey et al., 2020), potentially due to disruptions in Ub homeostasis. Finally, like many proteins implicated in ALS/FTD (Pakravan et al., 2021), UBQLN2 harbors low complexity regions and undergoes liquid-liquid phase separation (Sharkey et al., 2018; Dao et al., 2018). While ALS-associated mutations in the PRR may interfere with UBQLN2 liquid demixing, the relevance to disease pathogenesis is presently unclear.

In previous work, we reported that expression of ALS-associated UBQLN2 mutants caused mutation-dependent neurotoxicity in Drosophila (Kim et al., 2018). Here, we carried out genetic screens for UBQLN2 toxicity modifiers using flies expressing UBQLN2^ALS alleles with differing aggregation potential. Suppressor genes emerging from this screen were then tested for impacts on the toxicity of endogenous UBQLN2^ALS mutants in iMNs. Our findings suggest that endolysosomal dysfunction and axon guidance defects are phenotypic drivers of neurodegeneration in UBQLN2-associated ALS/FTD.

We previously reported that homozygous expression of single-copy human UBQLN2^ALS alleles caused mild eye toxicity when expressed under control of an eye-specific GMR driver at 22°C (Kim et al., 2018). Reasoning that UBQLN2-associated phenotypes may be worsened by heat stress (HS), we compared eye morphologies between flies expressing UBQLN2^WT, a clinical UBQLN2^P497H allele, and a highly aggregation-prone UBQLN2^4XALS mutant that harbors four different clinical mutations (P497H, P506T, P509S, and P525S, Figure 1A; Kim et al., 2018). None of the UBQLN2 transgenes caused an overt external eye phenotype when expressed from a hemizygous GMR>UBQLN2 Chr2 locus at 22°C (Figure 1B). By contrast, GMR>UBQLN2^4XALS flies—but not GMR>UBQLN2^WT or GMR>UBQLN2^P497H flies—exhibited a moderately severe rough eye (RE) phenotype that was characterized by eye depigmentation and loss of ommatidial facets in both male and female flies (Figure 1B). As expected, UBQLN2^4XALS was significantly less soluble than UBQLN2^WT and UBQLN2^P497H at both 22°C and 29°C; however, the relative proportion of insoluble UBQLN2^4XALS was comparable at both temperatures, suggesting that increased aggregation of UBQLN2^4XALS cannot wholly account for its enhanced toxicity at 29°C (Figure 1—figure supplement 1A, B).

We also performed quantitative mass spectrometry (MS) to assess relative abundance of UBQLN2^WT, UBQLN2^4XALS, and endogenous Drosophila Ubqln (dUbqln) in whole-head extracts. The average number of peptide spectral matches for hUBQLN2 was comparable between GMR>UBQLN2^WT and GMR>UBQLN2^4XALS flies reared at 22°C and 29°C (Figure 1—figure supplement 1C). hUBQLN2-unique peptides were ~6-fold more abundant than dUbqln peptides in both GMR>UBQLN2^WT and GMR>UBQLN2^4XALS heads, providing a lower limit of hUBQLN2 overexpression (Figure 1—figure supplement 1C). As expected, given the heat inducibility of UAS, absolute numbers of hUBQLN2 peptides were higher in GMR>UBQLN2^WT and GMR>UBQLN2^4XALS flies reared at 29°C versus 22°C; however, the difference was not statistically significant across three replicates. Altogether these findings established HS-dependent toxicity of UBQLN2^4XALS that is not solely due to increased protein expression.

We next performed RNA-Seq analysis of whole heads from GMR>UBQLN2^WT, GMR>UBQLN2^P497H, GMR>UBQLN2^4XALS. Each of the three GMR>UBQLN2 lines exhibited hundreds of gene expression changes relative to GMR-Gal4 controls, suggesting that UBQLN2 overexpression has a substantial impact on cellular regulation (Figure 1—figure supplement 2A). Consistent with their more severe eye phenotype, GMR>UBQLN2^4XALS flies exhibited a distinct RNA-Seq gene expression signature relative to GMR>UBQLN2^WT and GMR>UBQLN2^P497H flies, which clustered together in principle component analysis (Figure 1—figure supplement 2B). Using an FDR of <0.05, 629 genes were uniquely changed in GMR>UBQLN2^4XALS flies versus GMR>UBQLN2^WT, GMR>UBQLN2^P497H, and GMR-Gal4 controls at 29°C (Figure 1—figure supplement 2C, Figure 1—figure supplement 2—source data 1). Of these, 402 genes were upregulated and 227 genes were downregulated in GMR>UBQLN2^4XALS flies (Figure 1—figure supplement 2D, E, Supplementary Dataset 1 and 2, respectively). Gene ontology (GO) analysis revealed that mRNAs involved in eye development and phototransduction, including inaC, ninaE, and Rh3, were broadly downregulated in GMR>UBQLN2^4XALS flies, likely reflecting degenerative cell loss (Figure 1—figure supplement 2D, E, Supplementary Dataset 2). Upregulated GO terms included response to biotic stimulus, which includes Drosophila innate immunity genes such as Dro, AttB, AttC, and Listericin, and response to UV light, which includes the small heat-shock protein-encoding genes Hsp23, Hsp26, and Hsp27 (Figure 1—figure supplement 2D, E, Supplementary Dataset 1). Upregulation of HSPs may be driven by UBQLN2^4XALS misfolding. In contrast to the profound differences between GMR>UBQLN2^4XALS and GMR>UBQLN2^WT flies, only 64 genes were differentially expressed between GMR>UBQLN2^WT and GMR>UBQLN2^P497H flies, which is consistent with their qualitatively similar eye phenotypes. Finally, 28 differentially expressed genes were common to GMR>UBQLN2^4XALS and GMR>UBQLN2^P497H flies relative to GMR>UBQLN2^WT flies (Figure 1—figure supplement 2F). Overall, these findings reveal that the strong eye phenotype of GMR>UBQLN2^4XALS is accompanied by robust changes in gene expression.

The RE phenotype of GMR>UBQLN2^4XALS flies allowed us to perform genetic modifier screens. To this end, we screened a Bloomington Df library of 194 lines spanning 85% of chromosome 2 (Figure 2A). The screen was carried out at 29°C and eye phenotypes were scored on a scale of 1–5, with a score of ‘1’ representing a morphologically normal eye; a score of ‘3’ representing the unmodified UBQLN2^4XALS eye phenotype; and a score of ‘5’ representing eyes harboring >50% necrotic patches (Figure 2B). Lethal enhancers were also noted. While the mild eye phenotype of GMR>UBQLN2^P497H flies largely precluded identification of suppressors, the side-by-side screening of this line allowed us to identify shared and/or mutation-specific enhancers. Candidate modifier Dfs were subjected to secondary screens against UBQLN2^WT, UBQLN2^P497H, and UBQLN2^4XALS.

The UBQLN2^4XALS screen identified seven suppressors, three of which overlapped a common genomic interval, and 23 enhancers, including 15 lethal enhancers (Figure 2C). The UBQLN2^P497H screen identified six enhancers, two of which were also identified in the UBQLN2^4XALS screen (Figure 2D). Notably, all Dfs that enhanced the GMR>UBQLN2^P497H eye phenotype also caused enhanced eye phenotypes in GMR>UBQLN2^WT flies.

A combination of iterative Df screening and RNAi screening was then used to map causal genes within three UBQLN2^4XALS suppressor loci (BSC180, ED2426, and Exel6038) and one enhancer locus common to UBQLN2^4XALS and UBQLN2^P497H (BSC37) (Figure 2C and D). Among 17 annotated genes within BSC37 (Figure 2—figure supplement 1A), we prioritized Rab5, which encodes an early endosomal protein that interacts with the ALS2 gene product, Alsin (Hsu et al., 2018; Kunita et al., 2007). In support of a role for Rab5 as a UBQLN2 modifier gene, Rab5 knockdown phenocopied the hyperpigmented phenotype seen in GMR>UBQLN2^4XALS flies crossed to BSC37 at 29°C (Figure 2—figure supplement 1B, C). Rab5 knockdown also caused a hyperpigmented eye phenotype in GMR>UBQLN2^WT and GMR>UBQLN2^P497H flies (Figure 2—figure supplement 1C), indicating Rab5 is a mutation-independent UBQLN2 enhancer. Interestingly, while Rab5 knockdown also caused hyperpigmented eye patches in GMR>UBQLN2^WT, and GMR>UBQLN2^P497H flies reared at 22°C, GMR>UBQLN2^4XALS/shRab5 flies reared at 22°C failed to exhibit eye patches (Figure 2—figure supplement 1C). This finding supports the idea that UBQLN2^4XALS phenotypes are HS dependent and is consistent with an earlier study showing stronger eye phenotypes in GMR>UBQLN2^WT and GMR>UBQLN2^P497H flies relative to GMR>UBQLN2^4XALS flies at 22°C (Kim et al., 2018). Finally, we showed that overexpression of GFP-Rab5 partially rescued the eye phenotype of GMR>UBQLN2^WT, GMR>UBQLN2^P497H, and GMR>UBQLN2^4XALS flies reared at 29°C (Figure 2—figure supplement 1F). Rab5 knockdown or overexpression did not affect UBQLN2 expression (Figure 2—figure supplement 1D, E, G, H). The combined findings suggest that Rab5 is a general modulator of UBQLN2 toxicity and that UBQLN2 overexpression interferes with endolysosomal function in flies.

A similar approach was used to map the causal UBQLN2^4XALS suppressor gene in BSC180. Two different overlapping Dfs (C144 and ED4651) rescued UBQLN2^4XALS eye toxicity to a similar extent as BSC180 (Figure 2—figure supplement 2A, B). Within this overlap we identified lilliputian (lilli) as a gene of interest. lilli encodes a transcriptional elongation factor that is orthologous to mammalian AF4/FMR2 (fragile X mental retardation 2) (Wittwer et al., 2001). Interestingly, lilli mutations were recently shown to suppress toxicity of TDP-43 and C9ORF72-derived dipeptide repeat proteins (DPRs) in Drosophila (Chung et al., 2018; Yuva-Aydemir et al., 2019). In support of lilli as a UBQLN2^4XALS suppressor, a lilli^A17-2 LOF allele diminished UBQLN2^4XALS-mediated eye toxicity to a similar extent as lilli-spanning Dfs (Figure 2—figure supplement 2B, C). Thus, lilli may be of interest as a general suppressor of toxicity arising from misexpression of neurodegeneration-associated genes in Drosophila.

Next, we mapped a UBQLN2^4XALS suppressor in the genetic interval spanned by ED2426 and BSC346, which suppressed GMR>UBQLN2^4XALS eye phenotypes to a similar extent (Figure 3A and B). The region of overlap between ED2426 and BSC346 contains Unc-5, which encodes a transmembrane dependence receptor that mediates axonal repulsion and apoptosis suppression in response to secreted netrin ligands (Boyer and Gupton, 2018; Labrador et al., 2005; Keleman and Dickson, 2001; Llambi et al., 2005; Wang et al., 2009; Ahn et al., 2020).

In support of a genetic interaction between Unc-5 and UBQLN2^4XALS, two Unc-5 LOF alleles (Labrador et al., 2005), and three different RNAi lines diminished the RE phenotype of GMR>UBQLN2^4XALS flies without affecting UBQLN2 expression (Figure 3C, D and E). By contrast, Unc-5 knockdown had no effect on the RE phenotype caused by the expression of ALS-associated FUS alleles that cause severe degenerative phenotypes in Drosophila (Figure 3—figure supplement 1; Lanson et al., 2011; Kwiatkowski et al., 2009; Vance et al., 2009). This suggests that genetic interaction with Unc-5 may be specific to UBQLN2 versus other ectopically expressed ALS-associated proteins in Drosophila.

The repulsive activity of Unc-5 on axon guidance is antagonized by frazzled (fra), which encodes the fly ortholog of mammalian deleted in colon carcinoma (DCC) (Chan et al., 1996; Keino-Masu et al., 1996; Kolodziej et al., 1996). fra silencing by two different RNAi lines rescued the UBQLN2^4XALS eye phenotype to a similar extent as Unc-5 knockdown (Figure 3F). The combined findings suggest that signaling through Unc-5/Frazzled potentiates UBQLN2^4XALS toxicity in the Drosophila compound eye.

To evaluate the impact of Unc-5 silencing on motor function, we measured climbing behavior of recombinant flies expressing UBQLN2^4XALS under control of a D42 motor neuron driver in the presence of shUnc-5 or control shRNAs at 29°C. The moderate climbing defect of D42>UBQLN2^4XALS flies relative to D42-Gal4 flies was partially reversed by an shUnc-5 (TRiP) allele in both male and female flies, while a second Unc-5 shRNA line (KK) rescued climbing in male but not female flies (Figure 4A).

Consistent with their climbing defects, D42>UBQLN2^4XALS flies exhibited neuromuscular junction (NMJ) abnormalities, including increased numbers of satellite boutons and reduced numbers of mature boutons. Both phenotypes were corrected by Unc-5 silencing (Figure 4B and C). Finally, we also measured the effect of Unc-5 silencing on lifespans of flies expressing UBQLN2^4XALS under control of a pan-neuronal Elav driver. Elav>UBQLN2^4XALS/shUnc-5 flies showed a significant increase in lifespan relative to Elav>UBQLN2^4XALS flies crossed to shLuci, with the effect being most pronounced in female flies (Figure 4D). Unc-5 knockdown also modestly increased the lifespan of Elav-Gal4 male flies relative to Elav-Gal4/shLuci flies while having no impact on lifespan of Elav-Gal4 female flies (Figure 4D). Altogether, these experiments suggest that aberrant Unc-5 signaling contributes to neuronal phenotypes in UBQLN2^4XALS flies.

The identification of Unc-5 as a UBQLN2^4XALS suppressor raised the possibility that axonal guidance defects are particularly relevant to the UBQLN2 toxicity mechanism. Interestingly, the overlapping genetic interval spanned by the UBQLN2^4XALS suppressors Exel6038 and r10 contains beat-1b and beat-1c, two members of the beaten path (beat) family of axon guidance genes (Figure 5A–C; Vactor et al., 1993). Neuronally expressed Beat proteins regulate motor axon guidance and defasciculation in response to Sidestep (Side) ligands expressed on target substrates (Fambrough and Goodman, 1996; Sink et al., 2001; Siebert et al., 2009; de Jong et al., 2005). UBQLN2^4XALS flies crossed to a Beat-1b shRNA line exhibited less severe eye degeneration than UBQLN2^4XALS flies expressing shRNAs targeting luciferase, mCherry, or beat-1c (Figure 5D, E and F). GMR>UBQLN2^4XALS flies heterozygous for a p-element insertion in the beat-1b ORF also exhibited improved eye phenotype relative to control GMR>UBQLN2^4XALS controls (Figure 5D). These findings support the idea that Beat-1b signaling contributes to UBQLN2 toxicity and further implicate axonal guidance defects as a contributing pathomechanism to UBQLN2-associated neurodegeneration.

We employed CRISPR/CAS9 to introduce ALS-associated mutations into the X-linked UBQLN2 gene in neonatal male induced pluripotent stem cells (iPSCs) (WC031i-5907-6, see Materials and methods, and Du et al., 2015). During the course of this work, we serendipitously derived an UBQLN2 allele harboring clinical P497H and P525S mutations (termed UBQLN2^2XALS) as well as a UBQLN2^I498X allele that truncated the UBQLN2 ORF at codon 498 within the PRR. The expression and RIPA solubility of UBQLN2^P497H, UBQLN2^2XALS, and UBQLN2^4XALS were comparable to UBQLN2^WT in undifferentiated iPSCs, while UBQLN2^I498X could not be detected by Western blotting suggesting it is a null allele (Figure 6A). We also failed to detect cytologic UBQLN2^P497H, or UBQLN2^4XALS aggregates in immunostaining experiments, indicating that endogenous ALS mutations are insufficient to promote UBQLN2 aggregation in undifferentiated iPSCs (Figure 6B). As expected, UBQLN2^I498X iPSCs exhibited very weak immunoreactivity with UBQLN2 antibodies.

Previous work demonstrated that the 4XALS mutation increased the chymotrypsin sensitivity of purified UBQLN2, likely due to defective folding of the PRR (Kim et al., 2018). To determine whether the 4XALS mutation altered the folding of endogenous UBQLN2, we incubated detergent extracts from UBQLN2^WT, UBQLN2^P497H, UBQLN2^4XALS iPSCs with increasing concentrations of chymotrypsin. As shown in Figure 6C, UBQLN2^4XALS exhibited a unique chymotryptic fragmentation pattern relative to UBQLN2^WT and UBQLN2^P497H, suggesting that endogenous UBQLN2^4XALS is misfolded.

Reasoning that transient inhibition of protein degradation may potentiate UBQLN2 aggregation, we prepared soluble and insoluble fractions of UBQLN2^WT, UBQLN2^P497H, and UBQLN2^4XALS iPSCs exposed to the proteasome inhibitor MG132 or the autophagy inhibitor bafilomycin A1 (BafA1). While MG132 had no effect (not shown), BafA1 decreased the solubility of both wild-type and mutant UBQLN2 proteins, with the effect being most pronounced for UBQLN2^4XALS (Figure 6D). In addition, while the solubility of UBQLN2^WT and UBQLN2^P497H recovered following BafA1 washout, the fraction of insoluble UBQLN2^4XALS remained elevated (Figure 6D and E). Finally, the relative abundance of lipidated and non-lipidated forms of LC3 were comparable between UBQLN2^WT, UBQLN2^P497H, and UBQLN2^4XALS, and UBQLN2^I498X iPSCs, suggesting similar levels of autophagic flux (Figure 6D).

We next evaluated the localization of wild-type and ALS-mutant UBQLN2 proteins to endolysosomal structures. UBQLN2^WT, UBQLN2^P497H, and UBQLN2^4XALS partially localized with the lysosomal marker LAMP1 in BafA1-treated iPSCs (Figure 6—figure supplement 1A). By contrast, UBQLN2 punctae did not significantly overlap with Rab5-positive endosomes (Figure 6—figure supplement 1B). Given colocalization of UBQLN2 with LAMP1, we evaluated impacts of P497H and 4XALS mutations on lysosomal number and size in live cells using the lysosome-tropic fluorescent probe, LysoTracker. Both the abundance and average size of lysosomes were significantly elevated in UBQLN2^P497H and UBQLN2^4XALS iPSCs, relative to UBQLN2^WT or UBQLN2^I498X iPSCs, which exhibited qualitatively similar LysoTracker staining patterns (Figure 6—figure supplement 1C). In particular, the frequency of lysosomes greater than 5 μm and 10 μm was significantly elevated in UBQLN2^4XALS iPSCs versus UBQLN2^P497H iPSCs (Figure 6—figure supplement 1D). These findings support the existence of endolysosomal defects in UBQLN2^ALS iPSCs.

A failure to detect UBQLN2^ALS aggregates in untreated iPSCs could be due to continuous cytosolic dilution of UBQLN2 during mitotic cell division or may reflect the absence of neuron-specific stimuli that promote UBQLN2 aggregation. To explore these ideas, we carried out immunostaining experiments using UBQLN2^WT, UBQLN2^P497H, or UBQLN2^4XALS iMNs (Figure 7A). UBQLN2^WT, UBQLN2^P497H showed largely diffuse localization patterns, whereas UBQLN2^4XALS formed discrete aggregates distributed throughout iMN cell bodies and axons (Figure 7A). UBQLN2 aggregation was also observed in iMNs differentiated from an independently generated UBQLN2^4XALS iPSC clone (Clone 2, Figure 7—figure supplement 1A).

F-actin labeling with phalloidin revealed that UBQLN2^WT and UBQLN2^P497H were evenly distributed throughout growth cone lamellipodia and filopodia. Lamellipodial UBQLN2^4XALS staining was generally weaker, except for occasional brightly staining aggregates that were observed in ~50% of growth cones examined (Figure 7B). Notably, the UBQLN2^4XALS mutation did not significantly impact the localization of DCC, which was highly enriched in filopodial spikes in iMNs of all UBQLN2 genotypes (Figure 7—figure supplement 2).

We evaluated neurite length and complexity in UBQLN2^WT, UBQLN2^P497H, and UBQLN2^4XALS iMNs using Sholl analyses according to the diagram shown in Equation 1 and Figure 7C, respectively. UBQLN2^4XALS iMNs exhibited a significant reduction in average total neurite length, primary neurite length, and neurite branch length relative to UBQLN2^WT iMNs, which manifested as an increased rate of Sholl decay (Figure 7D). By contrast, neurite length and complexity were comparable between UBQLN2^P497H and UBQLN2^WT iMNs, indicating the clinical P497H mutation is insufficient to disrupt neurite growth dynamics (Figure 7D).

Given that autophagy inhibition with BafA1 induced UBQLN2^4XALS aggregation in iPSCs, we carried out similar studies in UBQLN2^WT, UBQLN2^P497H, and UBQLN2^4XALS iMNs. BafA1 strongly increased the size and number of UBQLN2^4XALS aggregates while having minimal effects on UBQLN2^WT or UBQLN2^P497H localization. BafA1-induced UBQLN2^4XALS aggregates were observed in two different iMN clones and were strongly colocalized with the autophagy receptor, p62 (Figure 8A, B and D). In addition, UBQLN2^4XALS aggregates often colocalized with, or were adjacent to, LAMP1-positive lysosomes (Figure 8C and D). By contrast, UBQLN2^4XALS only weakly colocalized with endosomal Rab5 or the autophagosome marker, LC3 (Figure 8C and D). These findings suggest that the autophagy pathway suppresses endogenous UBQLN2^4XALS aggregation in iMNs.

Mammals harbor a single fra ortholog, DCC, and four closely related paralogs with homology to Unc-5: UNC5A, UNC5B, UNC5C, and UNC5D. Among these, we focused on UNC5B, which is widely expressed in nervous tissue and has well-described roles in axon guidance and apoptosis regulation (Wang et al., 2009; Ahn et al., 2020; Pradella et al., 2021; Tang et al., 2008). To assess contributions of DCC and UNC5B signaling to UBQLN2-associated toxicity, we transduced UBQLN2^4XALS iPSCs with lentiviral shRNA vectors targeting DCC or UNC5B and differentiated the cells into iMNs for neurite analysis. qPCR confirmed that expression of UNC5B and DCC was reduced ~40–60% in their respective shRNA-transduced iPSCs relative to iPSCs transduced with an NT shRNA vector (Figure 9—figure supplement 1A, B). UBQLN2^4XALS iMNs expressing DCC shRNA also showed reduced DCC immunoreactivity at filopodial spikes (Figure 9—figure supplement 1C). Both UNC5B and DCC knockdown significantly increased average total neurite length, primary neurite length, and branch length in UBQLN2^4XALS iMNs relative to UBQLN2^4XALS iMNs expressing a non-targeting (NT) control shRNA (Figure 9A). Neither UNC5B nor DCC silencing affected the size or number of UBQLN2^4XALS aggregates (Figure 9—figure supplement 2), suggesting they influence toxicity pathway(s) downstream of UBQLN2 aggregation.

We next evaluated growth cone morphology in wild-type and UBQLN2^ALS iMNs stained with phalloidin. Compared to UBQLN2^WT or UBQLN2^P497H iMNs, UBQLN2^4XALS iMNs showed a high proportion of blunt-end termini versus growth cone termini, suggesting a defect in growth cone elaboration (Figure 9B and D). A preponderance of blunt-end termini was also observed in UBQLN2^4XALS (Clone 2) iMNs, suggesting they are a specific consequence of the 4XALS mutation (Figure 7—figure supplement 1B). By contrast, UBQLN2^4XALS iMNs expressing UNC5B or DCC shRNAs exhibited supernumerary growth cones along primary and secondary neurites and an increase in growth cone size and abundance relative to blunt-end termini (Figure 9C and D). These findings suggest that aberrant DCC-UNC5 signaling suppresses growth cone elaboration in UBQLN2^4XALS iMNs and that axon guidance defects contribute to toxicity phenotypes in fly and iMN models for UBQLN2-associated ALS.

Flies were maintained with the standard cornmeal-yeast medium (Nutri-Fly BF #66-112, Genesee Scientific) supplemented with propionic acid and all crosses were performed at 22°C. For HS experiments, all crosses were performed at indicated temperatures (27°C or 29°C). Note: the UAS-Gal4 promoter contains heat-shock elements that increase transgene expression at 27°C and 29°C. Generation of isogenic UAS-UBQLN2 stocks using PhiC31 integration was previously described.

Total RNA was isolated from 100 male fly heads using the TRIzol reagent (Invitrogen, 15596018) following the manufacturer’s protocol and treated with TURBO Dnase (Invitrogen, AM2239). RNA samples were prepared with three biological replicates for each genotype and each temperature. Then RNA samples were sent to Novogene (Novogene Co., Ltd, Sacramento, CA, USA) for non-stranded cDNA library building and sequencing at PE150 with NovoSeq 6000. Raw reads adapters were trimmed by fastp (Chen et al., 2018) and then were mapped to D. melanogaster genome (dmel-all-chromosome-r6.27, FlyBase) by STAR with the setting suggested by ENCODE project (https://github.com/ENCODE-DCC/rna-seq-pipeline, Dobin et al., 2013 STAR 2.7.1a). The number of RNA-Seq reads mapped to each transcript was summarized with featureCounts (Liao et al., 2014) and differential expression was called using DESeq2 (Love et al., 2014). The GO analysis was performed on MetaScape website (Zhou et al., 2019).

The protocol for fly brain immunohistochemistry was adapted from a previously published protocol (Zhan et al., 2013). Adult fly brains were dissected using a pair of fine forceps in PBS (or 0.3% TX-100 in PBS), fixed, blocked with normal goat serum (NGS), and stained with primary antibodies at 1:500 dilution for two overnights at 4°C. After subsequent secondary antibody staining, DAPI was added for the nuclear staining. Images were acquired using a Nikon A1 confocal microscope using a 60× oil lens.

Fly lysates were prepared by homogenizing 100 male fly heads in 200 μl of lysis buffer containing 20 mM Tris-HCl (pH 8.0), 138 mM NaCl, 10 mM KCl, 1 mM MgCl₂, 1 mM EDTA, 0.5% sodium deoxycholate w/v, and 0.1% SDS w/v (RIPA buffer w/o NP-40). Samples were centrifuged at 20,000 × g for 10 min and soluble fractions subjected to tryptic digestion and orbitrap MS using the filter aided sample preparation method (Wiśniewski et al., 2009). We performed two technical replicates for each of the three biological replicates. The tryptic digest solution was desalted/concentrated using an Omix 100 μl (80 μg capacity) C18 tip and the peptides were analyzed by HPLC-ESI-MS/MS using a system consisting of a high-performance liquid chromatograph (nanoAcquity, Waters) connected to an electrospray ionization Orbitrap mass spectrometer (QE HF, Thermo Fisher Scientific). HPLC separation employed a 100×365 μm fused silica capillary micro-column packed with 20 cm of 1.7 μm diameter, 130 Å pore size, C18 beads (Waters BEH), with an emitter tip pulled to approximately 1 μm using a laser puller (Sutter Instrument). Peptides were loaded on-column at a flow rate of 400 nl/min for 30 min and then eluted over 120 min at a flow rate of 300 nl/min with a gradient of 5–35% acetonitrile, in 0.2% formic acid. Full-mass profile scans were performed in the FT orbitrap between 375 and 1500 m/z at a resolution of 120,000, followed by MS/MS HCD scans of the 10 highest intensity parent ions at 30% relative collision energy and 15,000 resolution, with a mass range starting at 100 m/z. Dynamic exclusion was enabled with a repeat count of one over a duration of 30 s. The MetaMorpheus software program was used to identify peptides and proteins in the samples (Shortreed et al., 2015; Solntsev et al., 2018). Protein fold changes were quantified by FlashLFQ (Millikin et al., 2020; Yin et al., 2019; Millikin et al., 2018).

We employed the Bloomington Deficiency Kit for chromosome 2 (DK2L and DK2R) comprised of 194 different lines. All 194 lines were crossed to GMR-Gal4/CyO or homozygous GMR>UBQLN2^P497H or GMR>UBQLN2^4XALS flies at 29°C. A minimum of 30 F1 progeny containing GMR>UBQLN2 either the Df chromosome or balancer chromosome were analyzed for eye morphology 1–3 days post eclosion using a blinded, 1–5 grading system, with a score of 1 representing a control (GMR-Gal4) eye; 3 corresponding to the unmodified UBQLN2^4XALS phenotype at 29°C; and 5 representing severe eye degeneration featuring more than 50% necrotic tissue. We were unable to derive UBQLN2 progeny for a handful Df lines crossed to UBQLN2^4XALS, suggesting lethal genetic interactions. All putative modifier Dfs were retested in secondary screens that included side-by-side crosses to GMR-Gal4, GMR>UBQLN2^WT, GMR>UBQLN2^P497H, and GMR>UBQLN2^4XALS. Those Dfs that were confirmed to modify GMR > UBQLN2 eye phenotypes in both screens were deemed bona fide modifier Dfs. Sexually dimorphic phenotypes were also scored. Drosophila eye pictures were acquired using Leica S9 i Stereomicroscope.

Climbing assay was modified from methods described previously (Kim et al., 2018). Climbing ability was measured by tapping ~10 flies to the bottom of a graduated testing vial (15 cm) and taking videos over of fly movement over the course of 10 s. More than 100 flies for each genotype and each gender were used for climbing ability. Video frames at the 5 s time point were used to record the position of each fly using the multi-point plugin in ImageJ. Using the final positions of every fly and respective starting points also marked with ImageJ’s multi-point, the vertical displacement and velocity of every fly was calculated. Data showing velocity of each individual fly were graphed as scattered plots with mean climbing distance and SEM. Unpaired t-test with Welch’s correction were used for statistical analysis for different groups of flies.

NMJ assay was modified from methods described previously (Kim et al., 2018). Third-instar, wandering larvae from the F1 generation were rinsed in ice-cold PBS (Lonza, 17512F) and dissected along the dorsal midline. All tissues except the brain and nerves were removed to expose the muscles and NMJs. The dissected larval pelt was fixed in 4% paraformaldehyde for 20 min at room temperature. The larval pelts were given a wash with PBS followed by blocking with 5% NGS in 0.1% PBST (0.1% TX-100 in PBS). Following blocking, the larval pelts were probed with primary antibodies overnight at 4°C. They were then washed several times with 0.1% PBST followed by incubation with secondary antibodies for 2 hr at room temperature, subsequently followed by washes with 0.1% PBST. Larvae were then mounted onto slides using Prolong Gold mounting media. Confocal images were acquired using Zeiss LSM 710 confocal microscope and a 60× oil objective was used to image the NMJs. Both primary and secondary antibody solutions were prepared in 5% NGS in 0.1% PBST. For primary antibodies, the following dilutions were used: 1:100 Cy3-conjugated goat anti-HRP (Jackson ImmunoResearch, 123-165-021); 1:100 mouse anti-DLG 4F3 (DSHB). For secondary antibodies, the following antibody dilutions were used: 1:250 Alexa Fluor 647-conjugated phalloidin (Invitrogen, A22287); 1:500 goat anti-mouse Alexa Fluor 488 (Invitrogen, A-11029). For the analyses, NMJs innervating muscle 4 on segments A2-A3 were imaged and analyzed for synaptic bouton quantification. Mature boutons are defined as boutons that are included in a chain of two or more boutons. Satellite boutons are defined as a single bouton that is not included in a chain of boutons, and instead, sprout off of a mature bouton or branch. The groups were compared using unpaired Student’s t-test on GraphPad Prism software. p-Value less than 0.05 was considered statistically significant.

Longevity assay was modified from methods described previously (Zhan et al., 2013). For survival analysis, flies were aged at 27°C with no more than 15 flies per vial. Total more than 100 flies were used for each genotype. Vials were changed on a 2- to 3-day cycle. Death events were scored on a daily basis. Rescue in longevity was defined as greater than 5% increase in median lifespan in addition to the statistical threshold according to the Log-rank (Mantel-Cox) test, p<0.05. In the survival graphs shown, each set of experiments was done in the same time period with the corresponding control subjects in order to control longevity variation caused by environmental factors. Both genders were used in the survival assay unless otherwise specified.

A normal iPSC line (WC031i-5907-6, fibroblasts from neonatal male) was obtained from WiCell Research Institute (Yin et al., 2019). Into this line we introduced the following mutations using CRISPR/CAS9: P497H, 2XALS (P497H, P525S), 4XALS (P497H, P506T, P509S, P525S), and I498X, which harbors a 1 nt deletion in codon 497 that leads to frameshift and translation termination at codon 498. UBQLN2^P497H, UBQLN2^2XALS, and UBQLN2^4XALS lines were sequenced for the top five ranking off-target cleavages (none were found) and confirmed for expression of pluripotency markers. All iPSCs have normal karyotypes, express stem cell markers, exhibit pluripotency (as assessed by capacity to differentiate into three germ layers), and were mycoplasma negative (Yin et al., 2019). STR analysis defines profile for each line, confirms clonality and purity to 95–98% confidence. Fifteen loci were tested, all matched appropriate source fibroblasts. iPSCs were cultured with mTeSR1 (Stemcell Technologies) on Matrigel (Corning). iPSCs on Matrigel were passaged with 0.5 mM EDTA. iPSC colonies were passaged every 4–7 days at a 1:3 to 1:6 split ratio.

Differentiation of iPSCs into iMNs was carried out as previously described (Du et al., 2015). In brief, iPSCs were dissociated and placed in Matrigel-coated plates. On the following day, the iPSC medium was replaced with a chemically defined neural differentiation medium, including DMEM/F12, Neurobasal medium at 1:1, 0.5×N2, 0.5×B27, and 1×Glutamax (all are from Invitrogen). CHIR99021 (3 μM, Torcris), 2 μM DMH-1 (Torcris), and 2 μM SB431542 (Stemgent) were added in the medium. The culture medium was changed every other day. Human iPSCs maintained under this condition for 7 days were induced into neuroepithelial progenitors (NEP). The NEP cells were then dissociated with dispase (1 mg/ml) and split at 1:6 with neural differentiation medium described above. Retinoic acid (RA, 0.1 μM, Stemgent) and 0.5 μM purmorphamine (Stemgent) were added in combination with 1 μM CHIR99021, 2 μM DMH-1, and 2 μM SB431542. The medium was changed every other day. NEP cells maintained under this condition for 7 days differentiated into OLIG2+ motor neuron progenitors (MNPs). To induce motor neuron differentiation, OLIG2+ MNPs were dissociated with EDTA (0.5 mM) and cultured in suspension in the above neural differentiation medium with 0.1 μM RA and 0.1 μM purmorphamine. The medium was changed every other day. OLIG2+ MNPs under this condition for 6 days differentiated into HB9+ NMPs. The HB9+ NMPs were then dissociated with Accutase (Invitrogen) into single cells and plated on Matrigel-coated plates. The HB9+ NMPs were cultured with 0.1 μM RA, 0.1 μM purmorphamine, and 0.1 μM Compound E (Millipore) for 6 days to mature into CHAT+ iMNs.

For immunostaining, iPSCs and iMNs were fixed with 4% paraformaldehyde in PBS, permeabilized with 0.2% PBST, blocked with 2% BSA, and stained with primary antibodies for overnight at 4°C and then stained with α-rabbit-Alex-488 and α-mouse-Alexa-594-conjugated secondary antibodies. Images were acquired using a Nikon A1 confocal microscope using either a 20× lens or a 60× oil lens. For Lysotracker assays, iPSCs were incubated with 70 nM LysoTracker Red DND-99 (L7528, Invitrogen) for 1 hr at 37°C in culture medium. Live microscopy of lysosomes was performed using a Nikon A1 confocal microscope with a heated chamber and an objective to maintain the cells at 37°C using a 60× oil lens.

Fly lysates were prepared by homogenizing 10 fly heads in 50 μl of RIPA lysis buffer. The soluble fraction was taken after centrifugation at 21,000 × g in a microcentrifuge for 10 min iPSC extract were prepared using in lysis buffer containing 20 mM Tris-HCl (pH 8.0), 138 mM NaCl, 10 mM KCl, 1 mM MgCl₂, 1 mM EDTA, and 1% Triton-X 100 vol/vol (TX buffer). All lysis buffers were supplemented with protease inhibitor cocktail (Sigma, P8340), 10 mM NaF, and 1 mM DTT. Following centrifugation at 21,000 × g for 15 min, the insoluble pellet was washed twice with PBS, then suspended and boiled in Laemmli buffer. For immunoblotting, samples were separated by SDS-PAGE and transferred to PVDF membranes and immunoblotted with primary antibodies and LI-COR IRDye secondary antibodies (IRDye 800CW goat anti-rabbit and IRDye 680RD goat anti-mouse) as described. Signals were acquired using Odyssey bio-systems (LI-COR Biosciences). Immunoblotting results were analyzed and organized with ImageStudio Lite software (LI-COR).

iPSCs were lysed in buffer containing 20 mM Tris-HCl (pH 8.0), 138 mM NaCl, 10 mM KCl, 1 mM MgCl₂, 1 mM EDTA, and 0.2% NP-40 vol/vol. A total of 20 μg of each protein lysate in 10 μl was digested with increasing amounts of chymotrypsin (0.05–0. 2 μg/ml, final volume, Sigma) for 5 min at room temperature. Digestion was terminated by addition of Laemmli loading dye and boiling at 95°C for 5 min. The digested proteins were analyzed by Western blotting with anti-UBQLN2 antibodies.

Sholl analysis was done by determining the number of neurite branches at various radial distances from the cell body. The rate at which branching decreases as a function of distance is the Sholl regression coefficient, or Sholl decay. Digitizing individual neurite branching patterns, or tracing, was performed using the Simple Neurite Tracer (SNT) plugin in ImageJ. Semi-automatic tracing of β-tubulin staining in iMNs was carried out blinded on individual neurons. SNT’s Python application programming interface was implemented to measure the number of branch points and their respective radial distances. Other morphological descriptors including neurite length, branching number, and primary neurite number were measured in a similar way and graphed as scatter plots in GraphPad. Using the Python programming language, linear regression was fitted to the semi-log plots of the total branch points for every 1 μm radial distances.

There are a variety of ways to perform curve fitting for Sholl analysis, including linear mixed models. However, mixed models are necessary for samples with extensive clustering or heterogenity (Wilson et al., 2017). Since each sample was differentiated, immunostained, and imaged simultaneously and in the same way, a simple linear regression model was appropriate. To ensure each linear fit was an accurate approximation of branch point distribution, only linear fits with a high Pearson’s correlation (R²>0.8) were used to calculate the Sholl coefficient (Equation 1). Individual neuron Sholl decay was graphed as scatter plots using GraphPad, where each neuron was one data point. Since neurons with a low Pearson’s correlation were excluded in this analysis, all neurons of a given genotype were combined into one Sholl plot, in an additional, more inclusive analysis. This was accomplished by plotting the average number of branch points at 1 μm intervals among all neurons of a given genotype. Due to high variation of branching within each sample, only branch points that fell within the 10–90 percentile were linearly fitted to yield the overall Sholl decay for each genotype.(1)log⁡Npr2= -kr+mN=# branch points (intersections)r= radial distance from somak= Sholl regression coefficientm= y intercept of linear fitted line

Neurite terminals to the cell body were classified as growth cones (filopodial and lamellipodial ends, Figure 9B, arrows in UBQLN2^WT and UBQLN2^P497H) or blunt ends (Figure 9B UBQLN2^4XALS). The percent of neurite terminals classified as growth cones was measured for individual neurons.

For the fluorescence quantification of colocalization images, neuronal cell body regions were extracted from each image using ImageJ/Fiji’s selection tool. Thresholding of every region removed background signals. Pearson’s colocalization coefficients of each cell body region were determined using ImageJ/Fiji’s colocalization package. Pearson’s coefficients were plotted as a bar graph to compare the colocalization (rho = 1) or exclusion (rho = –1) of UBQLN2 with LC3, Rab5, LAMP1, or P62. Ten sight fields for each group were analyzed.

Statistical analysis information including individual replicates and biological replicates number, mean or median, and error bars are explained in the figure legends. The statistical tests and resulting p-values are shown in the figure legends and/or figure panels.