Deletion of UBQLN1 shortens telomeres in hESCs
Single cell RNA-seq analysis revealed that UBQLN1 expression level is positively correlated with the telomere length in hESCs (Fig. S1A) [20]. To determine whether telomere length regulates UBQLN1 expression, we generated telomerase-deficient TERC–/– hESCs and compared the telomere length and UBQLN1 expression levels with those of WT hESCs. Telomeres were shorter in TERC–/– hESCs than in WT hESCs (Fig. S1B). By Western blot analysis, UBQLN1 protein levels did not differ between TERC–/– hESCs that had short telomeres and WT hESCs (Fig. S1B), suggesting that telomerase deficiency and short telomeres do not affect UBQLN1 expression. Hence, we generated UBQLN1 knockout human ESCs using CRISPR/Cas9 method [48] and showed that deletion of UBQLN1 consistently resulted in telomere shortening in repeated experiments by Southern blot (Fig. S1C;Fig. 1A,B) and the telomeres continuously shorten following passage without UBQLN1 (Fig. 1B). Shorter telomeres found in both UBQLN1-knockout (UBQLN1–/–) hESC lines than in WT cells, were validated by flow-FISH [38](Fig. 1C), T/S ratio (Fig. 1D), or Q-FISH method (Fig. 1E). These data indicate that UBQLN1 is required for telomere maintenance in hESCs.
hESCs typically express pluripotent transcriptional factor OCT4. By immunofluorescence, OCT4 protein expression was not altered by UBQLN1 deficiency (Fig. S1D). The levels of shelterin complex components TRF1/2 and telomerase activity detected by TRAP assay [34], also remained similar between UBQLN1–/– and WT hESCs (Fig. 1F,G). Analysis of cell cycle revealed that the percentage of S-phase was slightly reduced in hESCs without UBQLN1, compared with that of WT hESCs (Fig. S1E). Nevertheless, DNA damage was notably elevated in UBQLN1–/– hESCs at late passage, as measured by immunofluorescence of γ-H2AX and p-ATM, compared with those of WT hESCs (Fig. 1H and Fig. S1F.G). Moreover, the frequency of micronuclei was increased in UBQLN1-deficient population (Fig. S1G). By simultaneous immunofluorescence of γ-H2AX and hybridization in situ (IF-FISH) analysis of telomeres [49], DNA damage at telomeres was increased in UBQLN1–/– cells as evidenced by co-localization of γ-H2AX and telomere foci (Fig. 1H). In summary, UBQLN1 deficiency leads to telomere shortening without affecting telomerase activity. Instead, UBQLN1 likely maintains telomeres by preventing telomere damage.
Transcriptome and proteome analysis of UBQLN1–/– hESCs
To understand potential signaling by which the loss of UBQLN1 leads to telomere damage and shortening, we performed RNA-sequencing to reveal the molecular changes. Transcriptome of UBQLN1–/– hESCs separated well from that of UBQLN1+/+ hESCs by Principal Components Analysis (PCA). Compared with WT hESCs, 81 genes were upregulated and 128 genes downregulated in UBQLN1–/– hESCs (Fig. S2A,B). By GO and GSEA analysis, the results indicated that upregulated genes were related to biological processes including calcium-dependent cell-cell adhering, regulation of hemopoiesis and cell differentiation in spinal cord (Fig. S2C), while downregulated genes were enriched in mitochondria function, oxidation-reduction process as well as metabolic process (Fig. S2D-F). Integrative Genomics Viewer showed four representative oxidative-reduction related genes were downregulated in UBQLN1–/– hESCs (Fig. S2G). The correlation analysis revealed that a positive correlation between the expression of UBQLN1 mRNA and these 4 genes (Fig. S2H), further indicating that UBQLN1 affects oxidation-reduction process. Reduction-oxidation reactions have an essential role in the protein structure maintenance through providing disulphide bonds and maintaining a proper redox environment for oxidative protein folding [50]. These RNA-seq data indicate that loss of UBQLN1 leads to altered transcriptome compromising mitochondria function and redox.
UBQLN1 is known to involve in ubiquitin proteasome system-mediated protein degradation [10, 51]. To further understand the protein changes in UBQLN1-deficient hESCs, we performed quantitative proteomic-MASS analysis. The proteome of UBQLN1+/+ hESCs can clearly distinguish from that of UBQLN1–/– hESCs by PCA analysis (Fig. 2A). 823 proteins increased and 824 proteins decreased in UBQLN1-deficient hESCs compared with WT hESCs (Fig. 2B). Most of decreased proteins are located in mitochondria and nucleus (about 24% and 24% respectively) (Fig. S3A). GO analysis showed that a large number of down-regulated proteins were enriched in mitochondria and mainly inner mitochondrial membrane protein complex (Fig. 2C). Correspondingly, these decreased proteins in UBQLN1–/– cells were mainly related to the biological process of aerobic respiration and electron transport on respiration chain (Fig. 2C-D). The main decreased electron transport chain proteins revealed by mass spectrometry were validated by western blot (Fig. S3B). Together, the quantitative proteomic-MASS analysis further suggests potential roles of UBQLN1 in mitochondria biogenesis and function.
Protein-protein interaction analysis demonstrated that the down-regulated proteins constituted a network of mitochondrial function, while the up-regulated proteins concentrated on the ubiquitin-proteasome system (Fig. S3C), consistent with the basic function of UBQLN1 in ubiquitylation mediated protein degradation. These data provide evidence that UBQLN1 can influence the mitochondrial functions by direct or indirect pathways such as proteostasis. The reduced electron transport chain (ETC) component also may lead to mitochondrial dysfunction.
UBQLN1 deficiency compromises mitochondria function
It is known that 13 proteins are encoded by mitochondrial genome [52]. Seven mitochondria-encoded protein were identified and all except for MT-ND5 were expressed at lower levels in UBQLN1–/– hESCs than in WT hESCs, based on our MASS data (Fig. S3D). This may also be evidence of organelle defect [53]. The total mtDNA copy number was also decreased after loss of UBQLN1 (Fig. S3E), which may explain for the decreased mitochondria-encoded proteins. However, the ATP level was up-regulated despite impaired mitochondria function accompanied by UBQLN1 deletion (Fig. S3F). It is to note that hESCs possess active glycolytic metabolism which can generate considerable ATP when compared to somatic cells [54], and presumably the increased ATP could be produced by anaerobic respiration independent of mitochondria function.
We compared the morphology of mitochondria by Mitotracker immunofluorescence and found nearly similar mitochondria morphology but slightly reduced quantity in UBQLN1–/– hESCs compared with UBQLN1+/+ cells (Fig. 2E). Moreover, the mitochondrial membrane potential (MMP) measured by JC-1 [55–57] was also reduced in UBQLN1–/– hESCs, like CCCP-induced reduction of MMP in WT hESCs (Fig. 2F), corresponding to decreased componence of ETC in inner membrane of mitochondria and dysfunctional organelle. The reactive oxygen species (ROS) level was notably increased in UBQLN1-deficient cells when compared with WT cells by either flow cytometric analysis or immunofluorescence (Fig. 2G,H). ROS produced by mitochondria has been frequently shown to damage DNA and telomere [58, 59]. Telomeric DNA is thought to be particularly susceptible to ROS-mediated cleavage and base modifications [60, 61]. ROS induced telomere damage is mainly mediated by oxidized guanine (8-oxoG), which can either prevent telomere elongation or even leading to telomere cleavage [62, 63]. Indeed, we observed remarkably increased level of 8-oxoG in the nucleus of UBQLN1–/– cells compared to WT cells (Fig. 2I). Together, these data indicate that UBQLN1–/– deficiency compromises mitochondria biogenesis and function and leads to oxidative damage.
N-acetyl-L-cysteine (NAC) or hypoxia mitigates telomere shortening in UBQLN1–/– hESCs
To test whether the ROS burst contributes to telomere damage and shortens telomeres in UBQLN1–/– cells, we designed two experiments by employing permeable antioxidant NAC or by culture of the cells under hypoxia (5% O2) condition to reduce potential oxidative damage, followed by additional culture for 10 passages and measurement of telomere length (Fig. 3A). NAC is an effective ROS scavenger, which can decrease cellular ROS level with appropriate concentration [59, 64]. NAC can efficiently reduce ROS level in UBQLN1–/– cells approximating that of UBQLN1+/+ hESCs and maintain normal ES cell clone morphology, like WT ESCs (Fig. 3B,C; Fig. S4A). The MMP also was recovered in UBQLN1–/– cells after treatment with NAC (Fig. S4B). The 8-oxoG which may directly damage telomere DNA, was significantly reduced, accompanied by the decreased ROS level (Fig. S4C). Consistently, NAC also decreased γH2AX foci at telomeres in UBQLN1–/– cells (Fig. S4D). Moreover, RNA-seq analysis revealed differential transcriptome between UBQLN1–/– cells treated with and without NAC. Interestingly, many of reduced ETC components of inner mitochondria membrane in UBQLN1–/– cells by MASS analysis as well as reduced expression of genes associated with oxidation-reduction process were restored at transcriptional levels following treatment with NAC (Fig. S4E,F). Also, genes in ETC of mitochondria were upregulated in UBQLN1–/– cells following NAC treatment (Fig. 3D). These results suggest that in addition to its function as a ROS scavenger, NAC also can recover the mitochondria function by intervening the transcriptome.
Furthermore, we examined the telomere length after treatment with NAC for additional 10 passages, compared with DMSO treatment served as vehicle control. NAC alleviated telomere shortening of UBQLN1–/– cells (Fig. 3E). Hence, NAC recovers mitochondria functions as demonstrated by RNA-seq, ATP and MMP, reduces ROS and partly prevent telomere shortening of UBQLN1–/– cells following continuous cultivation.
Additionally, we cultured ESCs under hypoxia (5% O2) and compared with conventional 20% O2 culture conditions. The hESC clones did not change much in morphology between the cultures under 20% and 5% O2 after 10 passages (Fig. 3F). Analysis of the cell cultures for 24 h by RNA-seq showed that the transcriptome differed under the two culture conditions. Notably, the up-regulated genes in UBQLN1–/– cells cultured under low versus normal O2 concentration were enriched in “glycolytic process” as well as “response to hypoxia” (Fig. 3G). Low O2 also improved the oxidation-reduction process weakened by the loss of UBQLN1 (Fig. 3H). Compared to those under 20% O2, UBQLN1–/– cells cultured under 5% O2 exhibited elevated MMP and declined levels of ROS and 8-oxoG (Fig. 3I; Fig. S5A-C). Elevated number of γH2AX co-localized with telomeres in UBQLN1–/– cells under 20% O2 was noticeably reduced by cultures under hypoxia (Fig. S5D). Reduced telomere damage corroborates the lower 8-oxoG levels by 5% O2 shown above.
Furthermore, we compared the telomere length by TRF of UBQLN1–/– cells cultured under low O2 condition for 10 passages with that of the control cultures under 20% O2. Low O2 culture alleviated telomere shortening in UBQLN1–/– cells (Fig. 3J). Transcriptome of UBQLN1–/– cells cultured under 5% O2 was closer to that of UBQLN1+/+ in both first and second principal component (Fig. S5E). By integrated analysis, genes up-regulated in both NAC and 5% O2 conditions were enriched in the pathway of hypoxia response (Fig. S5F), indicating the similar switch of metabolism pattern.
Together, 5% O2 cultures also reduce ROS and recover mitochondria functions and attenuate telomere shortening induced by UBQLN1 deficiency.
UBQLN1 prevents the ubiquitinated proteins from overloading mitochondria
To functionally investigate how UBQLN1 regulates mitochondria function, we searched for the UBQLN1-interacting proteins. We constructed the 3flag-UBQLN1 cell line (Fig. 4A). Unfortunately, we failed to see direct interactions of UBQLN1 with the mitochondria proteins such as SDHB and UQCRC2 that were reduced in UBQLN1–/– cells shown above (Fig. 4B;Fig. 2C,D). It is possible that UBQLN1 regulates shuttling of many ubiquitinated proteins without forming a stable interaction enough with them to be recovered by co-IP. Then, we carried out Co-IP/MASS analysis (Fig. 4C), and found many UBQLN1-interacting proteins such as ARF4 and signal pathways that were upregulated, notably ribosome biogenesis and metabolisms implicated in translation and protein synthesis (Fig. 4D,E). That many of the upregulated protein following UBQLN1 deficiency overlapped with UBQLN1-interacting proteins, such as RPS75b, MYH9, ARF4, PSMA6, EIF4A1 and RPL10, may suggest that UBQLN1 could be implicated in regulation of protein degradation for proteostasis.
We have identified ARF4 as an UBQLN1-interacting protein as well as its increase in UBQLN1–/– cells. By pull-down experiments combined ubiquitylation analysis, we observed ubiquitin modified ARF4 accumulated in UBQLN1–/– cells (Fig. 4F,4G). We also could see abundant protein aggresome accumulated in the cytoplasm of UBQLN1–/– cells, in contrast to WT cells (Fig. 4H). Proteosome inhibitor MG132 effectively promoted the aggresome, which may serve as a positive control, further supporting proteasome degradation deficiency in UBQLN1–/– cells. The aggresome appears to be the product of the most extreme form of protein aggregation that is observed in cells whose proteasome function is chronically blocked [65]. When the “Garbage proteins” failed to be degraded by proteasome and reside largely in cytoplasm, many of them can be carried to mitochondria and lead to mitophagy as well as elevated ROS [66]. Furthermore, increased mitophagy as shown by maker LC3-II (reflecting autophagic activity) and decreased P62 protein was detected in UBQLN1–/– cells, compared to WT cells (Fig. 4I), coincident with more emerging lysosomes (Fig. 4J). These data show that UBQLN1 is required to clear the ubiquitinated proteins and maintain functional mitochondria.
By further analysis of the proteome data, the upregulated proteins in UBQLN1–/– cells showed the enrichment in autophagy and ubiquitin-mediated proteolysis by GO and KEGG analysis (Fig. S6A,B). It is likely that UBQLN1 deficiency compromises protein degradation in proteasome by abrogating ubiquitin-proteasome system, consistent with UBQLN1’s function as a shuttle to carry ubiquitin modified protein to proteasome followed by degradation. In addition to the ubiquitin associated activity, the upregulated proteins were also enriched in multiple metabolisms, immune process, kinase activity as well as cell cycle (Fig. S6A,B). The aberrant up-regulation of these proteins also forebodes accumulation of mass proteins with destiny to degradation in UBQLN1–/– cells. Lysine-48 (K48) linked polyubiquitin chains are well established as the canonical signal for proteasomal degradation [67]. Quantitative mass spectrometry analyses of intracellular ubiquitin linkages support this notion, as K48-polyubiquitin linkage rapidly accumulates when cells are treated with the proteasome inhibitor MG132 [68]. To validate our hypothesis, we detected the ubiquitin (k48) level of total protein which was increased in UBQLN1–/– cells (Fig. 5A). UBQLN1 located mainly at cytoplasm and also with some in nucleus as tagged by flag (Fig. 5B), coincided with the distribution of ubiquitin modified proteins reported in previous research [69].
To examine whether the ubiquitinated proteins are accumulated around the mitochondria, we infected Ub-R-GPF (a GFP- based reporter for ubiquitinated proteins) into UBQLN1+/+ and UBQLN1–/– hESCs, based on the method described previously [70]. Notably, the accumulated GFP signals representative of ubiquitinated proteins supposedly with destiny to degradation were recruited near or even into mitochondria in UBQLN1–/– cells, distinguishable from UBQLN1+/+ cells (Fig. 5C). Mitochondrial-derived vesicles (MDVs), which are enriched for the outer mitochondrial membrane (OMM) import receptor TOMM20 and cannot be stained by mitochondrial probes and lack most of respiratory chain component proteins, are implicated in diverse physiological processes—for example, mitochondrial quality control—and are linked to various neurodegenerative diseases [71]. In early studies, MDVs carrying matrix cargo were stimulated by mild oxidative stress and were mapped to deliver mitochondrial proteins within 1–3 h to multivesicular bodies/lysosomes for degradation, whereas mitophagy would follow at later times (after approximately 24h) [72, 73]. We also observed increased number of large TOMM20+ MDV in UBQLN1-deficient hESCs, compared with those of WT cells (Fig. 5D). Together with increased LC3-Ⅱ and decreased p62, UBQLN1 loss may lead to increased mitophagy. Consequently, aberrantly elevated mitophagy may cause increased ROS levels and stress in UBQLN1–/– hESCs. TOMM20+ MDVs facilitate mitophagy in response to functional impairments, thus building a mitochondrial stress response [71].
Treatment of UBQLN1–/– cells with ML-SA1, which can activate lysosomal pathway by promoting lysosomal acidification and activity of the lysosomal enzymes [74], partly reduced accumulation of protein with ubiquitination modify (Fig. 5E) and also decreased ROS production (Fig. 6F). Notably, activation of lysosomal pathways by ML-SA1 treatment for 10 passages (from passage 10 to 20) reduced telomere shortening in UBQLN1–/– cells during continuously passages (Fig. 5G).
These results support the notion that UBQLN1 maintains mitochondrial function and telomeres by regulating proteostasis.
UBQLN1 mutation leads to neural differentiation defect.
We performed teratoma forming experiment to test if there is differentiation defect after injection of WT or UBQLN1–/– hESCs into immunodeficient nude mice. UBQLN1–/– hESCs can form teratoma at similar size without statistical differences compared with WT hESCs (Fig. 6A), even if UBQLN1–/– hESCs possess shortened telomere, defective protein homeostasis and dysfunctional mitochondria. However, the teratoma formed from UBQLN1–/– hESCs showed deficient neural differentiation as indicated by H&E staining (Fig. 6B) and reduced specific neural markers, such as NESTIN and SOX2, but not β-ⅢTUBULIN (Fig. 6C), while the development of mesoderm and endoderm was normal (Fig. 6B). Moreover, we also compared the embryoid body (EB) formation from WT or UBQLN1–/– hESCs. Deficient EB differentiation ability was found in UBQLN1–/– hESCs as to both the size and number (Fig. S7A). Furthermore, we also found reduced expression of neuronal marker SOX2 (Fig. S7C) but not β-ⅢTUBULIN (Fig. S7B) in EB induced differentiation process of UBQLN1–/– hESCs compared with WT hESCs. Hence, UBQLN1 plays an important role in neuronal differentiation of hESCs. This result may also explain the consequences of UBQLN1 mutation in the pathogenesis of neuron degenerative Alzheimer’s and Parkinson’s diseases.