1.Hypoxia promotes the mitochondria transfer from the higher invasive HCC cells to the lower invasive HCC cells
First, the interaction between co-cultured HCC cells was examined by field emission scanning electron microscope (FESEM). We found that the physical connection between HCC cells could be made through nanoscale tubular structure (Fig. 1a). MitoTracker-Red was used to label mitochondria, and phalloidin green was used to stain F-actin in the nanotubes. We observed the co-localization of MitoTracker labeled mitochondria in the nanotubes through immunofluorescence (IF)(Fig. 1b). Then, as shown in Fig. 1c, we used MitoTracker-Red dye to label the mitochondria in MHCC-97H cells, fully washed and removed the unbound dye, and then co-cultured them with green fluorescence protein (GFP) labeled Hep3B cells for 24 hours, and then performed DAPI nuclear staining and phalloidin green F-actin staining. We found that as shown by the yellow arrow in Fig. 1d, there are mitochondria labeled with MitoTracker-Red dye from MHCC-97H cells in GFP labeled Hep3B cells. The same phenomenon was also observed in the co-culture of PLC cells and Hep3B cells (Fig. 1c-d). This suggests that mitochondria can transfer between HCC cells through TNTs.
In order to analyze the role of mitochondrial transfer between different HCC cells, we compared the invasion and metastasis abilities of three commonly used HCC cells, MHCC-97H, PLC, and Hep3B, using transwell's migration and invasion experiments (Fig. 1e). We found that Hep3B cells were relatively the lower invasive HCC cells. We used Mito-PA-GFP to label the mitochondria in MHCC-97H and PLC cells, and the labeling results are shown in Fig. S1a. Then, as shown in Fig. 1f, Mito-PA-GFP labeled MHCC-97H and PLC cells were co-cultured with red fluorescence protein (RFP) labeled Hep3B cells for 24 hours respectively, and fluorescence activated cell sorting (FACS) was used to screen Hep3B cells containing MHCC-97H mitochondria or PLC mitochondria as the experimental group24. After sorting, Hep3B cells were subjected to transwell migration and invasion experiments (Fig. 1g), and we found that Hep3B cells containing MHCC-97H or PLC mitochondria had improved migration and invasion abilities. However, there was no statistically significant change in the migration and invasion ability of MHCC-97H cells or PLC cells containing Hep3B mitochondria (Figure S1b-c). This suggests that the transfer of mitochondria from the the higher invasive HCC cells to the lower invasive HCC cells can enhance the metastasis and invasion ability of the lower invasive cells.
To quantify the rate of mitochondrial transfer, we used a co-culture system of MHCC-97H and PLC cells labeled with Mito-PA-GFP and Hep3B cells labeled with RFP, respectively, to analyze the proportion of FITC+ ECD+ Hep3B cells to ECD+ Hep3B cells. Cytochalasin D can inhibit the formation of TNT25-28. As shown in the results in Fig. 1h, we added different concentrations of cytochalasin D in co-culture. Under the treatment of higher concentrations of cytochalasin D, the transfer rate of mitochondria decreased. HCC, as a solid tumor, has always been characterized by hypoxia in its microenvironment29. So, we treated the co-cultured systems under normal and hypoxic conditions for 24 hours and found that the rate of mitochondria transfer increased under hypoxic conditions (Fig. 1i). This suggests that hypoxia can promote the mitochondria transfer from the higher invasive HCC cells to the lower invasive HCC cells.
2. HMGB1 promotes mitochondrial transfer and promotes the development of HCC cells
In order to verify the role of HMGB1 in mitochondrial transfer and the migration and invasion of HCC, the following studies were carried out. According to data from TCGA and ICGC (Fig. 2a-b), the mRNA expression level of HMGB1 in HCC tissue was significantly higher than that in adjacent non tumor tissues. HCC patients with high expression of HMGB1 had a shorter overall survival period. Subsequently, in the clinical samples we collected, it was also detected through WB analysis that the protein level of HMGB1 in HCC was higher than that in adjacent non tumor tissues (Fig. 2c). Through immunohistochemistry staining, we found that HMGB1 was more localized in the cell nucleus in adjacent non tumor tissues, and the cytoplasm was also stained deeper in HCC (Fig. 2d).
In order to clarify the role of HMGB1 in mitochondrial transfer and the invasion and metastasis of HCC cells, we constructed cell lines with overexpression and knockdown of HMGB1 using lentivirus vector transfection and antibiotic screening. The related protein expression was shown in Fig. 2e. Firstly, we tested the effect of changes in HMGB1 expression on mitochondrial transfer rate through the co-culture model in Fig. 1h for 24 hours. From the results shown in Fig. 2f, we could see that overexpression of HMGB1 could promote mitochondrial transfer, while knocking down HMGB1 could reduce the transfer. Through the experimental results of Fig. 2g, it could be seen that overexpression of HMGB1 enhanced the migration and invasion ability of HCC cells, while knockdown resulted in a decrease in their ability. Similar to the results, we observed from the tail vein lung metastasis experiment in nude mice (Fig. 2h-j) that the HMGB1 overexpression group had stronger biological signals in the lung metastasis lesions on in vivo imaging, a relatively higher incidence of lung metastasis, and the larger area of lung metastasis lesions.
3.RHOT1 promotes mitochondrial transfer and promotes the metastasis and invasion of HCC cells
The spatial location and transportation of mitochondria are mainly regulated by RHOT1 (also called Miro1), which belongs to Miro protein of small GTPase subfamily and is located in the outer membrane of outer mitochondrial membrane30. As seen in published article, RHOT1 is up-regulated in pancreatic cancer, cholangiocarcinoma, esophageal cancer, glioblastoma, acute myeloid leukemia, thymoma and other cancers31. Due to HIF1A being a crucial molecule closely related to hypoxia32, so we first analyzed the relationship between RHOT1 and HIF1A using bioinformatics, to explore the expression of RHOT1 under hypoxia in advance. It was found that in the data of TCGA and ICGC (Fig. 3a), the correlations between the two were 0.40 and 0.36, respectively, with P<0.001. So we used western blot (WB) assay to analyze the protein expression of RHOT1 in PLC cells and MHCC-97H cells before and after hypoxia, and found that the expression of RHOT1 increased after hypoxia (Fig. 3b). According to data from TCGA and ICGC (Fig. 3c-d), the mRNA expression level of RHOT1 in HCC tissue was significantly higher than that in adjacent non tumor tissues. Meanwhile, HCC patients with high expression of RHOT1 had a shorter overall survival period. Subsequently, in the clinical samples we collected, it was also detected through WB assay that the protein level of RHOT1 in HCC was higher than that in adjacent non tumor tissues (Fig. 3e). Through immunohistochemistry staining, we found that RHOT1 was more localized in the cytoplasm of cells, and HCC staining was deeper compared to matched adjacent non tumor tissues (Fig. 3f).
Similarly, we constructed RHOT1 overexpression and knockdown cell lines using lentivirus vector transfection and antibiotic screening, with the associated protein expression shown in Fig.3g. From the results shown in Fig. 3h, we could see that overexpression of RHOT1 could promote mitochondrial transfer, while knocking down RHOT1 could reduce the transfer. Through the experimental results of Fig. 3i, it could be seen that overexpression of RHOT1 enhanced the migration and invasion ability of HCC cells, while knockdown resulted in a decrease in their ability. Similar to the results, we observed from the tail vein lung metastasis experiment in nude mice (Fig. 3j-l) that the RHOT1 overexpression group had stronger biological signals in the lung metastasis lesions on in vivo imaging, a relatively higher incidence of lung metastasis, and the larger area of lung metastasis lesions.
To further analyze the relationship between HMGB1 and RHOT1 in clinical samples, we used tissue microarrays for immunohistochemistry staining, which included 87 paired HCC patient tissues. Fig. 3m-o show that, similar to the results of bioinformatics, the expression of HMGB1 and RHOT1 in HCC patients was higher than that in adjacent non tumor tissues. HCC patients with high expression of HMGB1 and RHOT1 had the shorter overall survival period. In addition, we conducted a correlation analysis on the expression of HMGB1 and RHOT1, and found a positive correlation (n=87, r=0.3414, P=0.0012) between HMGB1 and RHOT1 expression in HCC patients in the tissue microarray (Fig. 3p). Similarly, in the data of TCGA (n=374, r=0.58, P<0.001) and ICGC (n=243, r=0.56, P<0.001), the two were also positively correlated (Fig. 3q) .
4.HMGB1 enhances RHOT1 expression through NFYA and NFYC subunits in the NF-Y complex under hypoxia
Previously, our research team found that hypoxia can promote the release of HMGB1 from HCC cells, which interacts with mitochondria to promote tumor progression15. Consistent with our previous research results, HMGB1 translocated to the cytoplasm and extracellular space of cells under hypoxic conditions (Figure 4a, Fig S1d). The biological behavior of mitochondria is closely related to the regulation of endoplasmic reticulum, and endoplasmic reticulum stress plays a key role in the communication between endoplasmic reticulum and mitochondria33. In our previous research, we found that HMGB1 can activate endoplasmic reticulum stress16. So we wondered whether HMGB1 could regulate RHOT1 expression through endoplasmic reticulum stress. So we detected the changes of RHOT1 and endoplasmic reticulum stress signal pathway through WB analysis after the PLC cells were stimulated by recombinant human HMGB1 protein and the high expression of HMGB1. As shown in Fig.4b, after stimulation of recombinant human HMGB1 protein and overexpression of HMGB1, RHOT1 protein expression increased, and endoplasmic reticulum stress signal pathway was activated. Subsequently, we placed PLC cells under hypoxic conditions and subjected them to various combinations of si-HMGB1 treatments (Fig. 4c). We found that the expression of RHOT1 and IRE1, PERK, and GRP78 increased under hypoxic conditions, and their overexpression was inhibited after si-HMGB1 treatment.
To confirm the specific effect of endoplasmic reticulum stress on RHOT1, we used the activator tunicamycin (TM) and the inhibitor 4-PBA to treat PLC cells and HMGB1 overexpressed cells respectively (Fig. 4d). We found that the expression of RHOT1 was increased under the stimulation of TM, while inhibited under the treatment of 4-PBA. The perception and response of ERS is coordinated by unfolded protein response (UPR), which has three endoplasmic reticulum membrane embedded sensors, namely double-stranded RNA-dependent protein kinase-like ER kinase (PERK), activated transcription factor 6 (ATF6) and inositol required enzyme 1 (IRE1)34. In order to explore the specific regulatory mechanism of endoplasmic reticulum stress on RHOT1, we used siRNA to inhibit the key molecules XBP1, PERK and ATF6 of the three pathways. Fig. 4e shows that the specific inhibitory effects were effective. Fig. 4f shows the changes of RHOT1 mRNA level under the treatment of three pathways, and shows that under the treatment of si-ATF6, the changes of RHOT1 were statistically significant. Studies have shown that when cell UPR is activated, ATF6α and/or ATF6β can combine with NF-Y complex to form heterologous protein complexes35. After analyzing various bioinformatics websites such as JASPAR (https://jaspar.genereg.net), it was found that NF-Y complex might be the transcription factor of RHOT1. So we conducted co-immunoprecipitation (CoIP) experiments using ATF6 and discovered the NF-Y transcription factor complex. Subsequently, we detected specific protein and mRNA changes in the three components of NF-Y in PLC and MHCC-97H cells under hypoxic conditions, and found that the expression of NFYA and NFYC increased (Fig. 4h). The results of Fig. 4j-k show that the expression of NFYA and NFYC increased under the effects of hypoxia and HMGB1. So we constructed HCC cells overexpressed with NFYA and NFYC using lentiviruses (Fig. 4l), which could enhance the expression of RHOT1. According to the luciferase reporter assay (Fig. 4m) and chromatin immunoprecipitation (ChIP) experiment (Fig. 4n-o), NFYA and NFYC regulated the expression of RHOT1 by binding to the first binding site of the RHOT1 promoter. The results of Fig. 4p-q show that after inhibiting the expression of NFYA and NFYC under hypoxia and HMGB1 stimulation, the expression of RHOT1 could be inhibited. In conclusion, we found that HMGB1 increases RHOT1 expression through NFYA and NFYC subunits in the NF-Y complex under hypoxia.
5 RAC1 binds to RHOT1 and aggregates to the cell membrane under the regulation of HMGB1 under hypoxia
The mechanism of TNTs formation is largely related to the interactions between protein complexes36,to explore the specific mechanism of nanotunnels in mitochondrial transfer, we conducted CoIP experiments using RHOT1 and conducted protein liquid chromatography tandem mass spectroscopy. We found that RAC1, ARP2/3 and β-actin could interact with RHOT1 (Fig. 5a). RAC1 belongs to the Ras superfamily and Rho family of small GTPases37, is the most famous member, mainly involves in the reorganization of actin cytoskeleton, which can promote the formation of TNTs through ARP2/3 complex38. The immunofluorescence experiments of RAC1 and RHOT1 were performed on HCC cells before and after hypoxia in Fig. 5b. It was found that there was fluorescence co-localization between the two. At the same time, it was found that RAC1 was relatively uniform in the cytoplasm of the cells under normoxia, but the fluorescence signal would aggregate on the cell membrane during hypoxia. Extracting proteins from the cell membrane of cancer cells before and after hypoxia, it was found that the RAC1 content of the cell membrane increased after hypoxia (Fig. 5c). The WB and IF results in Fig. 5d-e show similar phenomena in HCC tissue. Due to the phenomenon of HMGB1 translocation after hypoxia mentioned in Fig. 4a and Fig S1d, we are very curious whether the translocation of RAC1 is related to HMGB1. So we used recombinant human HMGB1 protein to stimulate HCC cells, and treated HCC cells with HMGB1 translocation inhibitor (ethyl pyruvate, EP) under hypoxia39. The results from Fig. 5f-g and Fig.S1e show that HMGB1 could regulate the aggregation of RAC1 onto the cell membrane under hypoxic conditions. At the same time, it was found through Fig. 5h results that HMGB1 could also promote the expression of RAC1 under hypoxic conditions.
To further analyze the relationship between HMGB1 and RAC1 in clinical samples, we also performed immunohistochemistry staining of tissue microarrays. Fig. 5i-k show that, similar to the results of bioinformatics, the expression of RAC1 in HCC patients was higher than that in adjacent non tumor tissues. HCC patients with high expression of RAC1 had a shorter overall survival period. In the correlation analysis of HMGB1 and RAC1 expression, it was found that there was a positive correlation (n=87, r=0.3769, P=0.0003) between the two in HCC patients with tissue microarray (Fig. 5l). Similarly, in the data of TCGA (n=374, r=0.43, P<0.001) and ICGC (n=243, r=0.43, P<0.001), the two were also positively correlated (Fig. 5m)
6.HMGB1 promotes mitochondrial transfer through RHOT1 and RAC1 in HCC cells
In order to verify the role of RAC1 in mitochondrial transfer and the migration and invasion of HCC, the following studies were carried out. According to data from TCGA and ICGC (Fig. 6a-b), the mRNA expression level of RAC1 in HCC tissue was significantly higher than that in adjacent non tumor tissues. HCC patients with high expression of RAC1 had a shorter overall survival period. Subsequently, in the clinical samples we collected, it was also detected through WB analysis that the protein level of RAC1 in HCC was higher than that in adjacent non tumor tissues (Fig. 6c). Through immunohistochemistry staining, we found that RAC1 was more localized in the cytoplasm of adjacent non tumor tissues, while the staining was deeper in HCC (Fig. 6d).
Similarly, we constructed RAC1 overexpressed and knockdown cell lines using lentivirus vector transfection and antibiotic screening, with associated protein expression shown in Fig. 6e. From the results shown in Fig. 6f, we could see that overexpression of RAC1 could promote mitochondrial transfer, while knocking down RAC1 could reduce the transfer. Through the experimental results of Fig. 6g, it could be seen that overexpressing RAC1 enhanced the migration and invasion ability of HCC cells, while knocking down showed a decrease in their ability. Similar to the results, we observed through the tail vein lung metastasis experiment in nude mice (Fig. 6h-j) that the RAC1 overexpression group had stronger biological signals in the lung metastasis lesions on in vivo imaging, a relatively higher incidence of lung metastasis, and the larger area of lung metastasis lesions.
In order to further clarify the specific effect relationship between RHOT1, RAC1, and HMGB1, we conducted sequential regulation on the basis of HMGB1 overexpression. From the results in Fig. 6k-o, it was shown that knocking down the expression of RHOT1 and RAC1 on the basis of Lv-HMGB1 PLC cells could reduce the mitochondria transfer, the migration and invasion ability of HCC cells, and the in vivo tail vein lung metastasis ability. At the same time, we analyzed the correlation between the expression of HMGB1, RHOT1, and RAC1 in tissue microarrays and clinical pathological data. According to the results in Table 1, the high expression of RHOT1 was closely related to HBsAg positivity, while the high expression of RAC1 was closely related to AFP>20ng/ml.