Key role of UBQLN2 in pathogenesis of amyotrophic lateral sclerosis and frontotemporal dementia

The UBQLN2 gene, also called Chap1/Dsk2 or PLIC, is located on the Xp11.21 chromosome. UBQLN2 is a single-exon coded, 66 kDa protein member of the ubiquilin family [12, 29]. UBQLN2 has a ubiquitin-like domain (UBL) in its N-terminal that can interact with the proteasome, and a ubiquitin-associated domain (UBA) on the C-terminal which recognizes the polyubiquitin chains on marked proteins [63]. UBQLN2 contains four stress-induced protein 1-like domains (STI-1 like), located between the residues 178–247 and 379–462, which are involved in the interaction with heat-shock proteins and autophagy mediators. UBQLN2 has also a unique PXX domain containing 12 tandem repeats implicated in protein-protein interactions [1, 12] (Fig. 1). This domain differentiates UBQLN2 from other members of the ubiquilin family. Interestingly, most of the mutations identified in the UBQLN2 gene are located in the PXX domain. Yet, it is unclear how the mutations impact the secondary structure and functions of the PXX domain. The 3-dimensional structure of UBQLN2 is formed by five β-strands, an α-helix of 3.5 turns and an additional 3₁₀-helix [63].

UBQLN2 mutations were first identified in five unrelated families, suffering of ALS/FTD [12]. From this analysis, five mutations were found in the PXX domain and all consisted of substituted proline residues (P497H, P497S, P506T, P509S and P525S). Subsequently, more than 10 other mutations have been reported [11, 13, 16, 18, 20, 26, 41, 46, 59, 60, 64] (Table 1). Albeit most of mutations have been detected in the PXX domain, some of them were also found in the STI1 domains or between domains [62]. The authors examined 47 post-mortem spinal cord samples of a wide variety of ALS cases, such as sALS (n = 23), fALS without mutations in SOD1, TDP-43 and FUS (n = 5), ALS with dementia (n = 5) and fALS with G298S mutation in TDP-43 (n = 21) [12]. Also, brain autopsy samples of cases harboring UBQLN2^P506T mutation were analysed. All fALS cases exhibited co-localization of UBQLN2 and TDP-43 cytoplasmic inclusions in the brain and the spinal cord neurons [12]. Interestingly, UBQLN2 was also found to be colocalizing with TDP-43 in the spinal cord of sALS patients, making it a component of the neuronal inclusions in patients affected with the sporadic form of the disease, suggesting a central role for this protein in ALS pathology [12, 17].

The degradation of misfolded or redundant proteins is critical for the maintenance of cellular health, specifically in neurons which are not proliferating cells. There is evidence that UBQLN2 mutations can provoke proteasome as well as autophagy impairments. First, an accumulation of the proteasome efficacy reporter Ubiquitin^G76V_GFP was detected in Neuro2a cells expressing mutant UBQLN2^P497H [12]. They further analyzed the dynamic of this reporter after blocking new protein synthesis with cycloheximide for 0, 2, 4 and 6 h in cells. The rates of Ubiquitin^G76V_GFP degradation were significantly slower in UBQLN2 mutations (P497H and P506T) when compared to the WT, suggesting the impairment of the protein degradation pathway by UBQLN2 mutations. Another group took advantage of the endogenous Myc as a proteasome efficiency reporter, as it is quickly degraded by the proteasome when its synthesis is blocked by cycloheximide [6]. In HeLa cells expressing mutant UBQLN2 (P497H, P497S, P506T, P509S or P525S) and treated with cycloheximide, a delay in the degradation of Myc was observed as compared to untransfected cells. However, when a chymotrypsin-like test was used to measure efficiency of the proteasome, no dysfunction of the proteasome was detected, which is contradictory with the previous study. Other studies suggested that mutant IVm-UBQLN2 could impede the UPS by reducing the delivery of ubiquitinated proteins to the proteasome [6, 44]. Indeed, ALS-linked mutation in UBQLN2 led to accumulation of ubiquitinated high-molecular-weight complexes (HMWCs). Deletion of the UBL domain of mutant UBQLN2 sequestered ubiquitinated substrates from both the UPS and autophagy (mATG9 and ATG16L1), resulting in an enhanced HMWCs accumulation [44]. Those results suggest that mutant UBQLN2 potentially interferes with the degradation of the polyubiquitinated substrates. Likewise, deletion of UBQLN2 UBL domain impaired the clearance of unfolded nascent polypeptide chains while UBQLN2 levels remained unaffected [24]. Recently, our group highlighted that overexpression of mutant UBQLN2^P497H in Neuro2A cells sequestered lys^48_ bound ubiquitin and resulted in a reduction in the efficiency of the proteasome [50]. This also supports the evidences that mutant UBQLN2 is actively impairing the UPS pathway.

Another study reported, with a knocked-in hUBQLN2^P506T mouse, that UBQLN2 works with the HSP70 system for proteasomal degradation of insoluble ubiquitylated protein aggregates [24]. Under non-stressed conditions nor mutations, UBQLN2 is inactive in homo- or heterodimers. In presence of HSP70 clients, UBQLN2 binds itself to HSP70 and associated ubiquitinated aggregated/misfolded proteins. HSP70-HSP110-dependent disaggregase activities can pull aggregated proteins apart, allowing UBQLN2 to form a HSP70-client-UBQLN2-proteasome degradation complex. UBQLN2 can then act as a proteasomal shuttle by connecting ubiquitylated proteins to the proteasome for degradation, leading to the proteolysis of the client. However, mutant UBQLN2 no longer recognized client-bound HSP70 and remained in its inactive phase, leading to accumulation of misfolded/aggregated proteins [24]. Defect in HSP70 binding was later confirmed in ALS patients’ lymphoblasts with P494L, P497H or P506A mutations in UBQLN2 [60]. They also confirmed the role of UBQLN2 as a shuttle factor for HSP70-bound substrates which was previously shown [24]. Thus, activation of UBQLN2 in order to bind to HSP70 seems to represent an important step to accomplish its normal degradation function.

Lastly, optineurin (OPTN) is a ubiquitin-binding multifunctional adaptor protein and is mainly implicated in the autophagy processes. It has also been proposed to be an important factor implicated in regulation of inflammation [38]. UBQLN2 and OPTN can be localized to the same cellular compartment, the Rab11-positive endosomal vesicles [45]. This observation led to propose that these proteins are involved in the same pathological processes in motor neurons diseases. Indeed, OPTN+/UBQLN2+ vesicles showed features of recycling endosomes and were positives for the key autophagy-regulating molecules p62, LC3, mAPG9L1, ATG16, and the autophagy initiator ULK1, suggesting that these vesicles are involved in protein homeostasis [45]. Thus, OPTN and UBQLN2 are implicated in protein homeostasis of the endosomal system via autophagy. Disrupting the function of these vesicles may contribute to development of ALS.

It has been reported that cytoplasmic inclusions of mutant C-terminal TDP-43 colocalized with either WT or mutant UBQLN2 in co-transfected Neuro2a cells [12]. UBQLN2 directly interacts with the C-terminal fragments of TDP-43 [4]. Moreover, results from our laboratory suggested that the overexpression of either hUBQLN2^WT or mutated hUBQLN2^P497H in Neuro2A cells can promote the formation of TDP-43 cytoplasmic inclusions [49]. Yet, the mutated form of UBQLN2 induced a stronger cytoplasmic mis-localization of TDP-43 as compared to the WT form. Those results were consistent with other studies that reported the high affinity of TDP-43 and UBQLN2 to form inclusions in the cytosol [4, 12, 16, 27, 34]. The formation of these inclusions has also been reported to be dose dependent. Cells with low expression of WT or mutant UBQLN2 did not exhibit cytoplasmic inclusions of UBQLN2 in TDP-43 aggregates [12, 49].

Strategies to increase ubiquitination have been proposed to boost UPS function [8]. In this regard, our group hypothesised that mutated UBQLN2 may sequester ubiquitin proteins, thereby reducing the degradation of mutated TDP-43 by the UPS and exacerbating the formation of cytoplasmic aggregates [50]. The co-expression, in Neuro2A cells, of hUBQLN2^P497H and hTDP-43^G348C vectors led to the accumulation of cytoplasmic TDP-43. However, ubiquitin overexpression reduced the cytoplasmic aggregation of TDP-43 [50]. A chymotrypsin-like assay was completed to measure the proteasome efficacy in hUBQLN2^P497H and hTDP-43^G348C transfected cells. This assay revealed a significant reduction in proteasome efficacy in all transfected cells as compared to control cells. The presence of up-regulated ubiquitin levels corrected the proteasome dysfunction in cells transfected with hUBQLN2^P497H and in cells co-transfected with both hUBQLN2^P497H and hTDP-43^G348C. These results suggested that the proteasome impairment caused by hUBQLN2^P497H up-regulation may participate to cytoplasmic TDP-43 accumulation. In summary, UBQLN2 overexpression or mutation exacerbated TDP-43 pathology in vitro (Fig. 3). A reduction of UBQLN2 expression or an up-regulation of ubiquitin may constitute potential therapeutic avenues for ALS pathology.

The implication of UBQLN2 in inflammation has emerged in the last two to 3 years. The overexpression of hUBQLN2^WT and hUBQLN2^P497H in Neuro2A cells was found to increase the phosphorylation of p38 MAPK and the nuclear levels of the phosphorylated form of the Nuclear Factor kappa-B (NF-κB), a potent inflammatory mediator [49]. However, it remains unclear if the activation of NF-κB is directly caused by the accumulation of UBQLN2 or comes from the UBQLN2-drived mislocalization of TDP-43 or other mechanisms [58]. Nevertheless, the down-regulation of UBQLN2 with siRNA reversed the increased NF-κB activation. NF-κB is a transcription factor involved in inflammation and probably has a noxious role in ALS and other neurological diseases. Indeed, the RNA levels of NF-κB as well as the levels of TDP-43 were abnormally up-regulated in the spinal cord of ALS patients [58]. TDP-43 was found to interact with p65 NF-κB in spinal cord of ALS but not in control. NF-κB inhibition by withaferin A treatment attenuated disease phenotype in mice expressing mutant TDP-43 and it increased survival in mutant SOD1 mice [14, 15, 48, 58]. Mice harboring UBQLN2^P497H;TDP-43^G348C mutations also exhibited increased levels of NF-κB in spinal cord as compared to TDP-43^G348C mice [50]. The double transgenic mice also expressed higher levels of pro-inflammatory cytokines CCL1, CCL5 and CXCL1 and higher stages of activated microglia in the spinal cord as compared to single TDP-43 transgenic mice. Noteworthy, the UBQLN2^P497H transgene was only expressed in neurons. The increased inflammation markers could suggest that microglia were hyperactivated in reaction to UBQLN2/TDP-43 neuronal pathology. These results also suggested that UBQLN2 pathology may favorize neuroinflammation and the reduction of UBQLN2 cytoplasmic accumulation may be a potential treatment approach to reduce glial cells activation observed in ALS.

Stress granules (SGs) are cytoplasmic inclusions of proteins and maintenance mRNAs with the aim of promoting stress response proteins translation in stressful situations, such as oxidative stress, osmotic stress, infection, ER stress and inhibition of the proteasome [3]. Under stress conditions, rather than engaging with the protein degradation system via its UBA and UBL domains, UBQLN2 associates with SGs components through its STI1-like region. UBQLN2 influences the early processes of molecular complex dynamics in phase separation that drives SGs formation. A group recently suggested that UBQLN2 modulates the state of the components to be recruited to SGs rather than regulating the levels of core SGs components [2]. One of the main components identified as UBQLN2-interacting SGs were heterogeneous nuclear ribonucleoproteins A (hnRNPAs) [2, 19]. UBQLN2 has been shown to colocalize with SGs in vitro under different cellular stressor. Indeed, UBQLN2 undergoes liquid-liquid phase separation (LLPS) which correlates with UBQLN2 ability to form stress-induced puncta inside cells [9]. LLPS consist of phase-separation of proteins composed of low-complexity or prion-like domains into membrane-less and spherical compartments. Persistent stress, mutations and aging are suspected to induce liquid-solid phase separation (LSPS) and might become irreversible aggregation, such as amyloid-like aggregates [52, 57]. Also, UBQLN2 puncta formation and stress granules (SGs) colocalization are highly sensitive to levels of UBQLN2 expression.

Stress granules contain various proteins, including fALS related proteins, TDP-43, FUS, hnRNPA1 and 2. A disturbance in stress granules homeostasis can lead to disruption of RNA processing [37] (Fig. 3). It has been recently reported that overexpression of mutated UBQLN2 (P497H and P506T) in HeLa cells directly impaired the binding of UBQLN2 to FUS which resulted in a loss of the ability of UBQLN2 to regulate FUS-RNA complexes and SGs formation [2]. Mutations in FUS and hnRNPA1 have been reported to increase their propensity to accumulate in SGs, disrupting SGs primary function to act as a stress response modulator [30, 36, 47]. Other studies reported that UBQLN2 upregulation can cause a cellular stress leading to MAPK activation, which have been recently shown to have a critical role in TDP-43 accumulation in stress granules following a chronic stress [40, 49].

Ubiquitin is also a common component of SGs [12, 33]. The specific binding between ubiquitin and UBQNL2 could prevent the self-assembly and oligomerization of UBQLN2 [9]. Spectrophotometric assays of ubiquitin and UBQLN2 at different molar ratios, revealed that LLPS formation decreased with the enhancement of ubiquitin. Those results might provide an insight by which UBQLN2 could traffic ubiquitinated substrates from SGs to the proteasome (Fig. 3) [9]. Another group reported that the UBA domain of UBQLN2 drove amyloid-like aggregation of UBQLN2 in vitro and without the help of ubiquitin or chaperones [56]. The UBA domain responsible for binding ubiquitinated substrates was also implicated in the promotion of aggregation because of the propensity of UBQLN2 to self-assemble and then aggregate. That process was modulated by ubiquitin binding to UBQLN2. Indeed, the behavior of the ubiquitin-binding-deficient form UBQLN2^L619A in neurons could prevent UBQLN2 self-assembly and could inhibit LLPS of UBQLN2 [56]. Also, expression of UBQLN2^L619A induced a 50% increase in neuronal death compared to WT. Collectively, those results suggest that the UBA domain promotes self-assembly of the protein leading to an increase in UBQLN2 aggregation with toxic consequences for neurons.

The LLPS capacity of a protein is based on its multivalent interactions in “sticker” regions which are separated by “spacers” [22]. Recently, a group exposed that ALS-linked mutations, depending of amino acid and sequence position, differently affect the implication of UBQLN2 in LLPS [10]. PXX domain mutations favorize UBQLN2 oligomerization and LLPS formation, creating reversible aggregates which can be cleared by binding to ubiquitin. Mutations to hydrophobic amino acid (P497L, P506A, T487I) increased the interaction between stickers and increased oligomerization of UBQLN2. Inversely, mutations to hydrophile residues (A488T, P500S, P506S, P509S, P525S) have small impact on oligomerization [10]. Increasing the hydrophobicity within “stickers”, but not “spacers”, favorized the oligomerization tendency of UBQLN2 [67]. The sequence position also greatly impacts UBQLN2 capacity for oligomerization. Oligomerization was high with mutation in 497 position and appeared to be little with mutation 500, 509 or 525 positions [10].