AEP binds to Tmod3 and specifically cleaves Tmod3 at N157.
To screen substrates of AEP, we conducted proteomic analysis with immunoprecipitated AEP complex. Silver staining and mass spectrometry were performed, and Tmod3 was thus identified as a candidate substrate (Fig. 1A-B, Supplementary Table. 2). We found that plenty of cytoskeleton related proteins identified in our data, such as Vimentin, Cortactin, and IQ Motif Containing GTPase Activating Protein 1 (IQGAP1), also have been revealed in gastric cancer [43]. More, endogenic coimmunoprecipitation (co-IP) assays revealed that AEP interacted with Tmod3 in GBM cells (Fig. 1C). Immunofluorescent (IF) staining indicated that Tmod3 colocalized with AEP around the nucleus (Fig. 1D). AEP has been reported to act as an important lysosomal protease that regulates protein substrate homeostasis [44]. Detection of lysosomal status with a lysosomal probe in AEP-knockdown GBM cells showed that AEP knockdown significantly activated lysosomes (Fig. S1A). To further investigate whether the interaction of AEP with Tmod3 occurs in the lysosomes or not, triple immunofluorescence was used to detect co-localization of AEP/Tmod3 with the lysosomal biomarker LAMP2. The result showed that AEP was highly co-localized with Tmod3, but both were only marginally co-localized with LAMP2 (Fig. S1B), suggesting that the interaction of AEP with Tmod3 mainly occurs outside lysosomes, which is consistent with our previous finding [36].
Next, we detected the cleavage of Tmod3 in GBM cells with or without high AEP expression. Two fragments with molecular weight between 15 and 25 kDa were found in cells with high AEP expression (Fig. 1E). To further explore whether the cleavage of Tmod3 was dependent on AEP enzymatic activity, wild-type (WT)-AEP or enzymatically inactive AEP (C189S mutant) was transiently co-expressed with carbon terminal Flag-tagged Tmod3 (Tmod3-Flag) in HEK293 cells. Only cells co-transfected with WT-AEP could be detected with cleaved Tmod3, while the cells transfected with AEP-C189S failed to exhibit the cleavage of Tmod3. These results suggested that AEP was responsible for eliciting Tmod3 proteolytic cleavage (Fig. 1F).
AEP is known to be strictly selective in cleaving asparagine [30]. To identify the precise cleavage site in Tmod3, we generated a panel of carbon (C)-terminal Flag-tagged Tmod3 mutants with asparagine (N) to alanine (A) substitution mutations. The results indicated that only the cells co-transfected with the Tmod3-N157A mutant and AEP failed to produce the cleavage product, definitively indicating that N157 was the specific cleavage site at which AEP cleaved Tmod3 (Fig. 1G and Fig. S2A). Through multispecies sequence alignment, N157 was found to be species conserved and precisely fit into the AEP enzymatic center (Fig. 1H-I).
As AEP cleaves Tmod3 at site N157 between the TMBS2 and ABS2 functional domains of Tmod3, Flag-tagged C-terminal truncated Tmod3 (amino acids, 1–157, referred to as tTmod3-N) and N-terminal truncated Tmod3 (amino acids, 158–352; referred to as tTmod3-C) were constructed (Fig. S2B). The interactions between AEP and two truncations were detected by co-IP, and it was found that tTmod3-C played a critical role in interacting with AEP (Fig. S2B).
Considering that specific mutations in a protein may affect its biological function, we searched Cancer Cell Line Encyclopedia (https://portals.broadinstitute.org) and collected sixteen natural missense mutations of Tmod3 in cancer. The corresponding mutant Tmod3 was co-transfected with AEP into HEK293 cells. Most of them and wild-type Tmod3 differed negligibly, except for the Tmod3 T256S, V262M and L303M mutant, which significantly promoted the cleavage of Tmod3 (Fig. S2C-D). In addition, Tmod3 was reported to be phosphorylated at Ser71 upon insulin-stimulated Akt2 activation, with its function significantly facilitated in actin remodeling [23]. However, phosphorylation of Ser71 did not affect the cleavage of Tmod3 (Fig. S2E). And that, we also examined the AEP-induced cleavage of other Tmod family proteins and found similar cleavage of Tmod1 or Tmod2, but not of Tmod4 (Fig. S2F). Taken together, these results indicate that AEP interacts with Tmod3 and cleaves at N157, generating two Tmod3 truncations with complete functional domains in cancer cells.
Tmod3 is highly expressed in many solid cancers and is associated with poor patient prognosis.
To investigate the expression of Tmod3 in diverse tumors and its correlation with prognosis. We searched the Gene Expression Profiling Interactive Analysis (GEPIA). Tmod3 was found to be upregulated in diverse types of tumors such as cholangio carcinoma (CHOL), pancreatic adenocarcinoma (PAAD) and stomach adenocarcinoma (STAD), compared to the corresponding normal tissues (Fig. S3A). We also searched the Gene Expression Omnibus (GEO) database, and analysis of the GSE 4290 dataset confirmed that Tmod3 was upregulated in HGG compared with LGG and normal brain tissues (Fig. S3B). High Tmod3 expression is associated with poor patient prognosis in many types of tumors, including GBM and LGG, cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), lung adenocarcinoma and adenosquamous carcinama (LUAD and LUSC), PAAD and bladder urothelial carcinoma (BLCA) (Fig. S3C-G).
Since malignant tumors often exhibit marked heterogeneity, in order to investigate the expression of Tmod3 in different cell subtypes within the tumor, we measured the expression of Tmod3 by performing single-nuclear RNA-seq (snRNA-seq) with 6 cases of GBM and single-cell RNA-seq (scRNA-seq) with 8 cases of cervical cancer tissues. After single-nucleus/cell suspension preparation and single cDNA library sequencing, the t-distributed stochastic neighbor embedding (t-SNE) algorithm distinguished 23 clusters and 28 clusters of cells in GBM and cervical cancer, respectively. The results indicated that Tmod3 was highly and ubiquitously expressed in most clusters (Fig. 2A-B). We further detected the expression of Tmod3 by immunohistochemical staining in a tissue array of 123 human glioma specimens. Tmod3 was found to be highly upregulated in HGG compared with LGG (Fig. 2C-D). Patients whose Tmod3-stained samples had a high H-score presented with poorer overall survival (OS) (60 of 123 patients with definitive survival follow-up data, P = 0.0044, HR = 1.963, 95% CI, 1.118–3.448. Figure 2E, Supplementary Table 3). Western blotting analysis of glioma tissues consistently showed that Tmod3 was significantly upregulated in HGG compared to LGG and normal brain tissues (Fig. S3H).
To identify the functions of Tmod3 in cancer cells, we constructed Tmod3-knockdown (KD) U87-MG and A172 cells using lentivirus. The effects of Tmod3 KD were confirmed by western blotting (Fig. 2F). Lysosomal probe detection indicated that Tmod3 KD had a limited effect on lysosomes (Fig. S4A). The CCK-8 assay indicated that Tmod3 KD significantly inhibited the proliferation of GBM cells (Fig. S4B). Flow cytometry results suggested that Tmod3 KD led to increased apoptosis rates in both cells (Fig. S4C-D). The colony formation ability was significantly impaired in the cells with Tmod3 KD (Fig. 2G-H). In addition, the invasive ability of U87-MG and A172 cells was also inhibited by Tmod3 KD, as detected by Transwell assay and scratch wound-healing assay (Fig. 2I-J and Fig. S4E-F). Simultaneously, in vivo animal models suggested that Tmod3 KD reduced the tumorigenesis of GBM (Fig. 2K-L and Fig. S4G). Tmod3 KD reduced weight loss and prolonged the OS of tumor-bearing mice (Fig. S4H and Fig. 2M). Immumohistochemical staining (IHC) indicated that biomarkers of proliferation (Ki-67) and invasion (Vinculin) were significantly reduced by Tmod3 KD (Fig. S4I-J). Taken together, the results suggested that Tmod3 may exert deleterious effects in different types of solid tumors, and that suppression of Tmod3 significantly inhibits the malignancy-related functions of GBM cancer cells in vitro and in vivo.
AEP-produced tTmod3 is found in many types of solid tumors and is highly associated with poor prognosis in patients with GBM.
In biochemical experiments, Tmod3 was found to be cleaved by AEP in cancer cells. Does this event truly occur in clinical tumor tissues, and if it does, then what does this cleavage indicate about clinical diagnosis and treatment? We collected 72 human freshly frozen glioma tissues (LGG, 24; HGG, 48), 8 normal brain tissues, 25 hepatocellular carcinoma and 30 cervical cancer tissues to detect the cleavage of Tmod3 by western blotting. Obviously, bands indicating specific Tmod3 cleavage were definitively detected in the HGG tissues (25.0%, 12 of 48 cases, Fig. 3A and Fig. S5A-B), and there was a positive correlation between AEP activation and cleavage of Tmod3 (Fig. 3B). However, the cleavage of Tmod3 was nearly undetectable in LGG and normal tissues (Fig. 3C-D and Fig. S5C). The cleavage of Tmod3 was also detected in the hepatocellular carcinoma tissues (32.0%, 8 of 25 cases, Fig. 3E and Fig. S5D) and the cervical cancer tissues (26.7%, 8 of 30 cases, Fig. 3F and Fig. S5E). We analyzed the correlation between the high cleavage of Tmod3 and the clinical characteristics of HGG patients, and the results were shown in Supplementary Table 4. We found that age, sex, cancer recurrence, WHO tumor grade and isocitrate dehydrogenase 1 (IDH1) mutation were without difference in patients with or without Tmod3 cleavage. Tumors with high Tmod3 cleavage tended to be invasive in multiple lobes (Fig. 3G). Moreover, compared to patients with lower cleavage of Tmod3, patients with obvious cleavage of Tmod3 presented with shortened OS (n = 48, P = 0.0357, HR = 2.243, 95% CI, 0.7491–14.04, Fig. 3H, Supplementary Table 4). Thus, these results strongly indicated that AEP cleavage of Tmod3 is a common event in many types of solid tumors and that the cleavage of Tmod3 is a promising biomarker for indicating poor prognosis in patients with GBM.
Tmod3 truncations produced by AEP cleavage promote GBM proliferation and invasion.
Since truncations of Tmod3 are present in tumor tissues and are prognostically relevant, what is their function? To answer this question, U87-MG and A172 cells with tTmod3-N or tTmod3-C overexpression (OE) were constructed (Fig. 4A). The promoted proliferative ability of cells with tTmod3-C OE was validated by colony formation assays (Fig. 4B-C) and CCK-8 assays (Fig. S6A). In contrast, in the invasion assays, compared with negative control (NC) cells, cells with tTmod3-N OE were found to have significant elevated motility (Fig. 4D-E). In the in vivo experiments, the group of mice with tTmod3-C OE exhibited significantly enhanced tumor progression (Fig. 4F-I), and the OS of the tumor-burden mice was reduced compared to that of the NC group of mice (Fig. 4J). Next, we measured the expression of Tmod3, Ki-67 and Vinculin in related groups of mice. Consistently, Ki-67 was significantly upregulated in the tTmod3-C OE group, while Vinculin was significantly upregulated in the tTmod3-N OE group (Fig. S6B-C). In addition, in an unexpeted finding, tTmod3-C was found to have markedly increased nuclear accumulation (Fig. S6B). Together, these results suggested that two truncations of Tmod3 may promote GBM cancer cell malignancy in different ways: tTmod3-C mainly promotes tumor proliferation, while tTmod3-N primarily promotes invasion.
AEP promotes GBM progression by cleaving Tmod3 in vitro and in vivo
To investigate whether the elevated proliferation and motility mediated by AEP is dependent on AEP cleavage of Tmod3, we constructed cell lines for use in functional assays (NC, AEP KD, AEP KD/tTmod3-C res, AEP KD/tTmod3-N res and AEP KD/tTmod3-N res/tTmod3-C res). Cells with AEP KD presented impaired proliferation and motility, which was consistent with our previous findings [36]. However, the rescue of tTmod3-C strikingly enhanced the proliferation and colony formation ability of U87-MG and A172 cells (Fig. 5A-B and Fig. S7A). The rescue of tTmod3-N significantly promoted the invasion of U87-MG and A172 cells in a Transwell assay (Fig. 5C-D).
To fully address the effects of AEP cleavage of Tmod3 on GBM progression in vivo, nude mice orthotopic glioma model using the constructed U87-MG cells (NC, AEP KD, AEP KD/tTmod3-C res, AEP KD/tTmod3-N res and AEP KD/tTmod3-N res/tTmod3-C res) were established. The MRI scan showed that AEP KD significantly inhibited tumorigenic ability of U87-MG cells, which was consistent with our previous article on AEP in GBM [36]. However, tTmod3-N and tTmod3-C, especially the latter one rescued the tumor size reduction caused by AEP KD. Notably, tTmod3-N together with tTmod3-C greatly promoted tumor progression in vivo (Fig. 5E-G). Mice injected with AEP-KD/tTmod3-C res or AEP-KD/tTmod3-N res/tTmod3-C res U87-MG cells presented with greater significant weight loss on the day of MRI detection (Fig. 5H). We calculated the OS of the remaining mice. Consistently, AEP KD significantly prolonged the survival of the mice; however, tTmod3-C res alone or in combination with tTmod3-N res dramatically shortened the OS of the related mice (Fig. 5I). Further measurement of the corresponding proliferation and invasion indicators by immunohistochemical analysis showed that Ki-67 was upregulated in tissues with tTmod3-C res, while Vinculin was upregulated in tissues with tTmod3-N res (Fig. S7B-D).
In order to exclude the excessive rescue of the two truncations of Tmod3 from producing side effects on cellular functions, we constructed GBM cells that were individually or collectively rescued with two truncations of Tmod3 at the physiological level (Fig. S8A). And we also detected and confirmed that the rescue of truncated Tmod3 does not have a significant effect on the known substrate protein p53 of AEP (Fig. S8A). Results of cellular functional assays consistently showed that truncated Tmod3 mediates the AEP-enhanced aggressive phenotype of cancer cells (Fig. S8B-D). In addition, the rescue of truncated Tmod3 had a limited effect on lysosomes (Fig. S9).
Taken together, the in vitro and in vivo experiments revealed that truncated Tmod3 produced by AEP promoted GBM progression. In particular, the proliferation-promoting ability of tTmod3-C led to a rapid increase in tumors in a short period of time, which directly affected the OS of model mice, while tTmod3-N significantly enhanced cancer cell migration and invasion.
tTmod3-N promotes cancer cell invasion through actin remodeling.
Considering that Tmod3 is important for the regulation of the cytoskeleton, how do two truncations of Tmod3 contribute to cytoskeletal regulation? We used phalloidin staining to detect cytoskeletal changes in A172 cells with tTmod3-C or tTmod3-N overexpression. The cells overexpressing tTmod3-N tended to be disordered, with the cellular edges changing from sharp to rough and the disordered actin fibers were co-located with tTmod3-N at cell edges (Fig. 6A, upper panel). tTmod3-C was not closely related to actin, but was found translocated to the nucleus (Fig. 6A, bottom panel). Moreover, co-IP assay confirmed that both Tmod3 and tTmod3-N could bind to actin (Fig. 6B). Thus, these findings hinted that tTmod3-N may be involved in regulating the formation of actin fibers.
To investigate the effects of truncated Tmod3 on F-actin polymerization, we purified his-tagged full-length Tmod3 (Tmod3-FL), tTmod3-N and tTmod3-C (Fig. 6C). The rate of actin polymerization was detected after adding 200 mM Tmod3 or its truncations to the reaction system using an Actin Polymerization Biochem Kit. Indeed, Tmod3 markedly inhibited the polymerization of actin as previously reported [18], however, the two truncations alone failed to or mildly damage actin polymerization (Fig. 6D). When tTmod3-N combind with Tmod3-FL was added to the reaction system, the polymerization of actin was significantly rescued, which suggested that tTmod3-N may competitively inhibit the functions of Tmod3-FL and promote F-actin polymerization (Fig. 6D).
To further detect the effects of AEP cleavage of Tmod3 on actin remodeling. U87-MG and A172 cells with AEP KD or with truncated Tmod3 rescue and the related controls were analyzed (NC, AEP KD, AEP KD/tTmod3-N res, AEP KD/tTmod3-C res, AEP KD/tTmod3-N res/tTmod3-C res). Under normal conditions, globular-actin (G-actin) and F-actin are in dynamic equilibrium, and the F-actin to G-actin (F/G-actin) ratio can reflect the invasion ability of cells to some extent [45]. To test the changes in the F/G-actin ratio, IF staining of F-actin and G-actin with fluorescence-conjugated phalloidine and deoxyribonuclease I (Fig. 6E) and western blotting (Fig. 6G) were performed. The results indicated that AEP KD induced a decrease in the ratio of F/G-actin, while the rescue of tTmod3-N reversed this effect. The cells with tTmod3-N rescue appeared as extended pseudopodia and a disordered actin cytoskeleton. The rescue of tTmod3-C alone did not change the ratio of F/G-actin (Fig. 6F and Fig. 6H). Additionally, the cellular pseudopodia-related marker Cortactin was detected in related cells. In cells with the tTmod3-N res or tTmod3-N res/tTmod3-C res, Cortactin and F-actin were significantly colocalized at the edge of the cell, which was considered an invasive pseudopod [46] (Fig. 6I-J).
Hence, these findings demonstrated that tTmod3-N remodeled the actin cytoskeleton by interacting with actin and competitively suppressing full-length Tmod3 functions, which significantly promoted the invasion of cancer cells.
tTmod3-C is accumulated in the nucleus and enhances SND1-mediated cancer cell proliferation.
tTmod3-C was found to aggregate in the nucleus of A172 and U87-MG cells by confocal laser scanning microscopy (Fig. 7A), which was also confirmed by western blotting (Fig. 7B). Tmod1, previously known as erythrocyte tropomodulin (E-Tmod), was transported to the nucleus depending on its functional nuclear import and export domains, and in the nucleus, it played a novel role in the proliferation of cells [47]. Inspired by this, we next set out to determine whether the nuclear export sequence (NES) (speculated to be located at aa_130-139) or nuclear localization sequence (NLS) (speculated to be located at aa_343-351) of Tmod3 is critical for the nuclear translocation of Tmod3. We transiently expressed Tmod3 or its truncations with NES or NLS mutations or deficiency in HeLa cells (Fig. 7C-D). Consistent with the results in the A172 and U87-MG cells, Tmod3 was mainly distributed in the cytoplasm, tTmod3-N was negligible in the nucleus, and tTmod3-C was obviously enriched in the nucleus (Fig. 7E). Mutations in the core amino acids of the NES sequence (I137D/L138E) led to nuclear accumulation of Tmod3, while mutations in the core amino acids of the NLS sequence (K344A/R345C) resulted in total cytoplasmic localization. These results indicated that both the NES and NLS contributed to the nuclear-cytoplasmic shuttling of Tmod3 (Fig. 7F). To determine whether the shuttling of tTmod3-C into the nucleus is dominantly mediated by the NLS or caused by NES deletion. We constructed tTmod3-C with an NLS mutation (tTmod3-C mutNLS) or deletion (tTmod3-C ΔNLS) and tTmod3-C with an NES addition (NES-tTmod3-C) (Fig. 7D). HeLa cells transiently expressing tTmod3-C mutNLS or tTmod3-C ΔNLS exhibited partial nuclear localization of the mutants, while most of the green fluorescence in cells that transiently expressed NES-tTmod3-C was in the cytoplasm (Fig. 7G). By the way, we treated HeLa cells transfected with the abovementioned truncations with leptomycin B (LMB, 100 nM, 3 h), a nuclear export inhibitor that inhibits the transport of proteins and RNAs carrying the NES. Indeed, truncations containing NES clearly responsed to LMB treatment (Fig. 7H). Defective NLS does not completely inhibit the nuclear aggregation of tTmod3-C. Functionally, the NES sequence addition can inhibit cell proliferation caused by tTmod3-C, while NLS mutated tTmod3-C still has a promotive effect on the proliferation of GBM cells (Fig. 7I-J). These results suggested that NES deficiency dominantly mediated the nuclear translocation of tTmod3-C, and AEP cleavage of Tmod3 at N157 was a natural regulatory mechanism of Tmod3 nuclear translocation because it cleaved the NES in the N-terminus.
To further address the mechanism by which tTmod3-C promotes cancer cell proliferation, mass spectrometry following co-IP using nuclear fractions of U87-MG cells with AEP KD and tTmod3-C rescue was conducted. The results revealed that Staphylococcal Nuclease And Tudor Domain Containing 1 (SND1) was a candidate for interacting with tTmod3-C (Fig. 8A-B, Supplementary Table 5). The interaction of tTmod3-C with SND1 was confirmed by co-IP with transiently expressed Flag-tagged SND1 and ZsGreen1-tTmod3-C in HEK293 cells followed by immunoblotting with anti-ZsGreen1 antibody (Fig. 8C).
SND1 is a novel transcriptional coactivator that activates downstream Ras Homolog Family Member A (RhoA) transcription, sequentially regulates expression of Cyclin D1 (CCND1), Cyclin E1 (CCNE1) and Cyclin Dependent Kinase 4 (CDK4), and accelerates cell proliferation in GBM [48, 49]. Together with LGMN and TMOD3, members of SNA1/RhoA signaling such as SND1, RHOA and CDK1 were upregulated in GBM (Fig. S10A). We reasonably assumed that tTmod3-C promoted GBM cell proliferation by interacting with SND1. To confirm this supposition, A172 and U87-MG cells with indicated treatment: NC, AEP KD, AEP KD/SND1 OE, AEP KD/tTmod3-C res, AEP KD/tTmod3-C res/siSND1, were used. The effect of SND1 interference was confirmed by western blotting (Fig. 8D). The reported SND1 downstream targets such as RhoA, CDK1, CDK2, CDK4 and Cyclin D1, were significantly downregulated by AEP KD, but rescued by SND1 OE (Fig. 8E). tTmod3-C rescued the downregulation of the related markers caused by AEP KD, but SND1 interference reduced the effect of tTmod3-C in A172 and U87-MG cells (Fig. 8F). Compared to cells with AEP KD, GBM cells with SND1 OE showed reestablished proliferation, as detected by CCK-8, similar to the proliferation of cells with tTmod3-C rescue (Fig. 8G). However, when SND1 was interfered by siRNA, the promoted proliferation driven by tTmod3-C rescue was significantly inhibited, thus indicating that tTmod3-C promoted GBM proliferation partially depending on SND1 (Fig. 8H). In addition, correlation analysis of TMOD3 and SND1 or RHOA expression by GEPIA indicated that Tmod3 is correlated with SND1/RhoA signaling (Fig. S10B). These findings indicated that tTmod3-C was translocated to the nucleus mainly due to the absence of the NES in the N-terminus. Nuclear tTmod3-C interacting with SND1 promoted the transcription of downstream genes, facilitating cell proliferation.