A genome-wide sgRNA library screen identified vulnerabilities in OS treated with doxorubicin
To functionally identify vulnerabilities of doxorubicin resistance in OS cells, we infected 143B cells with an sgRNA lentiviral library targeting 20,914 human genes, covering each gene with at least 6 independent sgRNA sequences. Cells were then cultured for 14 days under selection conditions in the presence of doxorubicin. Then, genomic DNA was extracted, and sgRNA barcodes were amplified for next-generation sequencing to identify sgRNAs and their target genes that were lost after selective culture, indicating that these genes may be critical for the maintenance of doxorubicin resistance (Fig. 1A). By analyzing the sgRNA of the control and doxorubicin treatment groups on day 14, we obtained the key genes promoting doxorubicin resistance in 143B OS cells (Fig. 1B). Among the top 10 genes, ASF1B, NPM1, FZR1, CENPW and SUMO1 were associated with the progression of multiple tumors, drug resistance and a poor prognosis in patients [18,19,20,21,22] (Fig. 1C), suggesting that this screening system is a reliable approach in the search for doxorubicin resistance-related targets for OS.
Through enrichment analysis of the genes obtained from negative screening, we found that the pathways about cell death, cell proliferation, cell cycle and drug transport processes, which play a key role in cell survival and chemotherapy resistance, were significantly enriched (Fig. 1D, E). In addition, gene set enrichment analysis (GSEA) showed that cell apoptosis and multiple DNA damage repair pathways were significantly affected (Fig. 1F and Supplementary Fig. 1A). We reasoned that cells with DNA damage repair defects were selected due to the DNA toxicity of doxorubicin. To further identify doxorubicin resistance-related genes, we made an analysis of data of 143B cells treated with DMSO and those treated with doxorubicin for 7 and 14 days, and revealed a total of 28 significant hits (Fig. 1G), among which genes with more than 3 independent sgRNA sequences were picked for further study (Fig. 1H and Supplementary Fig. 1B).
Rad18 is a critical driver for doxorubicin resistance
To further identify the key genes that lead to chemotherapy resistance, we constructed doxorubicin-resistant cells (143B-Res) from parental 143B cells. The successful development of doxorubicin-resistant cells was evidenced by increased cell viability under doxorubicin treatment conditions (Fig. 2A, B). RNA sequencing of doxorubicin-sensitive parental 143B cells and 143B-Res cells revealed that (Fig. 2C and Supplementary Fig. 2A), in addition to changes in cellular stress pathways, various drug-resistant pathways, including DNA replication related pathways, epoxygenase P450 pathway and ABC transporters, had dynamically shifted in 143B-Res cells (Fig. 2D and Supplementary Fig. 2B). In addition, ABCB1, ABCG2, MYCN and other known drug resistance genes were highly expressed in 143B-Res cells (Fig. 2E). These results verified the successful induction of doxorubicin resistance based on pathway changes and gene expression changes.
By overlapping the differentially expressed genes from the CRISPR/Cas9 knockout library screening and the genes with transcriptomic alterations from the RNA sequencing analysis, we obtained four potential doxorubicin resistance-related genes: Rad18, GPRC5B, KIF1A and FOXN4 (Fig. 2F). We speculated that cells with knockdown of these genes would show therapeutic sensitivity to doxorubicin. Therefore, small interfering RNAs (siRNAs) of these genes were designed for gene knockdown. The interference efficiency of each siRNA was evaluated by the relative expression of mRNA and protein (Supplementary Fig. 2C, D), and the one with the highest interference efficiency was picked for subsequent experiments. Subsequently, we found that among the four genes, Rad18 knockdown led to the most apoptosis of 143B cells induced by doxorubicin (Fig. 2G and Supplementary Fig. 2E). On the other hand, we found that Rad18 was significantly overexpressed in multiple types of cancers (Supplementary Fig. 2F) and associated with a poor prognosis of patients with sarcoma (Supplementary Fig. 2G). We further assessed the relative expression level of Rad18 in hFOB1.19 and OS cell lines and found that the expression level of Rad18 in each OS cell line was higher than that in normal cells (Fig. 2H). We also found that Rad18 was highly expressed in OS tissues compared with normal tissues according to analysis of clinical specimens and data from GSE39058 (Fig. 2I, J). Thus, we speculate that Rad18 is a determinant of doxorubicin sensitivity and the malignant progression of OS.
Subsequently, the half-maximal inhibitory concentration IC50 of doxorubicin in each cell line was measured, and regression analysis showed that the higher the expression level of Rad18 was, the stronger the tolerance of cells to doxorubicin (Fig. 2K). Doxorubicin treatment induced high Rad18 expression in vitro and in vivo (Fig. 2L, M and Supplementary Fig. 2H), which was also observed in the OS tissues of patients receiving chemotherapy (Fig. 2N). In addition, data of GSE87437 indicated the higher the expression level of Rad18 was, the worse the pathological response to chemotherapy among OS patients (Fig. 2J). Taken together, these results demonstrate that Rad18 acts as a critical player in OS cell acquisition of doxorubicin resistance.
Rad18 depletion sensitized OS cells to doxorubicin treatment
To investigate the effects of Rad18 on doxorubicin resistance, we generated stable Rad18 knockout subclones in 143B and Saos-2 cells by CRISPR–Cas9 lentiviral vectors with sgRNA#3 of Rad18 (Fig. 3A). We found that Rad18 knockout increased the sensitivity to doxorubicin in 143B, Saos-2 and 143B-Res cells (Fig. 3B), We further overexpressed Rad18 in 143B and Saos-2 cell lines (Supplementary Fig. 3A). while overexpression of Rad18 led to doxorubicin resistance (Supplementary Fig. 3B). Rad18-knockout subclones showed no significant effects on OS cell proliferation in vitro, whereas knockout of Rad18 significantly induced a reduction in the number of clones after doxorubicin treatment (Fig. 3C, D). In addition, overexpression of Rad18 significantly restored the reduction in the number of clones caused by doxorubicin (Supplementary Fig. 3C, D). Thus, Rad18 expression is closely related to doxorubicin resistance. Furthermore, we found that doxorubicin-induced apoptosis significantly increased in all three Rad18-knockout cell lines (Fig. 3E, F). Subsequently, a caspase-3/7 staining assay revealed that doxorubicin induced increased apoptosis in the Rad18-knockout group (Fig. 3G). In contrast, doxorubicin-induced apoptosis and caspase-3/7 activation were alleviated after overexpression of Rad18 (Supplementary Fig. 3E, F). These suggest that cytotoxicity of doxorubicin was correlated with expression level Rad18.
Next, we sought to confirm the phenomenon in vivo. 143B and Rad18 knockout 143B cells carrying luciferase were injected into the tibia of nude mice, and orthotopic OS tumors were formed after 3 weeks. After treatment with doxorubicin or DMSO for 4 weeks, OS was significantly inhibited by doxorubicin, and the inhibition was significantly increased after Rad18-knockout (Fig. 3H, I). HE staining and immunohistochemical staining indicated that doxorubicin caused more OS necrosis and caspase-3 activation in the Rad18-knockout group than in the control group, suggesting that Rad18 knockout can increase doxorubicin-induced apoptosis (Fig. 3J, K). Together, these data demonstrate that knockout Rad18 induces OS sensitive to doxorubicin treatment both in vivo and in vitro.
Rad18 promotes the HR repair pathway to increase the tolerance of doxorubicin-induced DNA damage
To determine how Rad18 functions, transcriptomic sequencing was performed on 143B and Rad18-knockout 143B cells. DNA replication and DNA repair pathways, such as the HR and Fanconi anemia pathways, were significantly enriched (Fig. 4A, B and Supplementary Fig. 4A). Therefore, we characterized doxorubicin-induced DNA damage by a comet assay to determine whether Rad18 regulates DNA repair. The results showed that the DNA comet tail moment and tail length in Rad18-knockout cells were extended compared to those in control cells; and these trends were reversed in Rad18-overexpressing cells, suggesting that Rad18 is a key factor in the tolerance of doxorubicin-induced DNA damage (Fig. 4C, D and Supplementary Fig. 4B, C).
Since DSBs are serious consequence of DNA damage caused by doxorubicin, we investigated whether Rad18 regulates DSBs by assessing the foci of γ-H2AX, which is phosphorylated in response to DSBs [23]. We found that Rad18-knockout cells contained a greater number of γ-H2AX-positive foci, and the difference remained significant at 24 h, suggesting that Rad18-knockout cells were unable to efficiently repair DSBs (Fig. 4E). The enhanced accumulation of γ-H2AX in Rad18-knockout cells was further confirmed by western blotting (Fig. 4F). In Rad18-overexpressing cells, H2AX phosphorylation was weakened, suggesting that Rad18 could reduce the accumulation of DSBs (Supplementary Fig. 4D, E).
HR is one of the key pathways to repair DSBs [24]. Our GSEA also showed that the HR pathway was enriched after Rad18 knockout (Fig. 4B), so we hypothesized that Rad18 regulates HR to influence the repair of doxorubicin-induced DSBs. Subsequently, we employed HR-GFP, the DSB repair reporter that allows quantification of the activities of HR. Fluorescence microscopy showed that Rad18 knockout reduced HR-mediated DSB repair efficacy in both Saos-2 and 143B cells treated with doxorubicin, suggesting that Rad18 knockout inhibits DSB repair through inhibition of the HR pathway (Fig. 4G-I). Collectively, we found HR deficiency in Rad18 knockout cells. This deficiency resulted in a delay in DNA damage repair, which aggravated doxorubicin-mediated DNA damage and enhanced doxorubicin cytotoxicity.
Rad18 interacts with MRE11 and enhances MRN complex formation
To gain further insights into the mechanism of Rad18, 143B cells treated with 5 μM doxorubicin for 6 h were collected and lysed, and e immunoprecipitated with an anti-Rad18 antibody. Co-IP proteins were separated using SDS-PAGE (Fig. 5A). The target strip was excavated and for mass spectroscopy (MS). We focused on the proteins related to DNA damage repair, and MRE11, PCNA and RPA were identified. (Supplementary Fig. 5A). Previous studies have shown that PCNA and RPA are definite interacting proteins of Rad18, but neither of them is directly involved in HR. [25, 26] MRE11, a DNA DSB repair protein, is a key initiation of the HR pathway, while the interaction between Rad18 and MRE11 protein has not been elucidated. The phosphatase domain of MRE11 has single-stranded DNA endonuclease and double-stranded DNA exonuclease activities, which are mainly responsible for pruning the ends of DNA broken strands and subsequently initiating the DSB repair pathway [27, 28]. We innovatively found that MRE11 might be a binding partner of Rad18 by IP/mass spectrometry (Fig. 5B and Supplementary Fig. 5A). In addition, we predicted the relationship between Rad18 and MRE11 and HR repair pathways through the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (Supplementary Fig. 5B). Therefore, we speculated that Rad18 may regulate the HR pathway by interacting with MRE11, thereby affecting DNA damage repair (DDR).
To explore the above hypothesis, we first explored the interaction between Rad18 and MRE11. After immunofluorescence staining, we found the colocalization of Rad18 and MRE11 through regression analysis of fluorescence intensity (Fig. 5C, Supplementary Fig. 5C, D). Moreover, an immunoprecipitation (IP) assay of Rad18/MRE11 was performed in 143B and 143B-RES cells. MRE11 was detected from IP products of Rad18, suggesting that Rad18 could interact with MRE11 (Fig. 5D). To further verify the interaction, HEK293T cells were co-transfected with flag-tagged Rad18 and HA-tagged MRE11 expression vectors. Protein complexes were immunoprecipitated with an anti-FLAG antibody. The lysates were examined by western blot using anti-Flag and anti-HA antibodies. Analysis under specific loading conditions revealed that Rad18 interacts with MRE11 (Fig. 5E).
Based on the above conclusions, we then explored whether Rad18 regulates the expression level of MRE11 to affect the HR pathway. The expression levels of MRE11 in OS cell lines showed different trends from Rad18 (Fig. 5F). In addition, the expression level of MRE11 was not changed after knockout or overexpression of Rad18, indicating that Rad18 does not regulate the expression of MRE11 (Fig. 5G, Supplementary Fig. 5E). Considering that MRE11 is the key active region of the MRN (MRE11/RAD50/NBS1) complex, we further studied the influence of Rad18 on the MRN complex [27]. We first observed the enhanced interaction between Rad18 and the MRN complex after treated with 5 μm doxorubicin for 6 h (Fig. 5H). Additionally, the formation of the MRN complex was disrupted after Rad18 knockout, suggesting that Rad18 could promote the formation of the MRN complex (Fig. 5I, Supplementary Fig. 5F). However, we found that regardless of whether Rad18 was knocked out or overexpressed, the expression levels of NBS1, MRE11 and RAD50 did not change significantly (Fig. 5J, Supplementary Fig. 5G). This means that the effect of Rad18 might be promoting the formation of the MRN complex as a whole rather than the expression of the individual components. The MRN complex promotes ATM activation by inducing its autophosphorylation at S1981 [29]. Activated ATM rapidly phosphorylates a large number of substrates in local chromatin, providing scaffolding for assembly of higher-order complexes to repair damaged DNA, which facilitates the HR pathway [29, 30]. Therefore, we detected the phosphorylation level of ATM as a marker of HR activation to reveal the functional status of MRN complexes. We found that ATM phosphorylation was significantly inhibited when Rad18 was knocked out, and it was increased after Rad18 was overexpressed (Fig. 5J, Supplementary Fig. 5G). Immunofluorescence yielded consistent results, indicating that Rad18 is involved in the regulation of ATM phosphorylation (Fig. 5K, Supplementary Fig. 5H). Collectively, we verified that Rad18 interacts with MRE11 to promote the formation of the MRN complex, which facilitates the activation of ATM to promote the HR pathway (Fig. 5L).
Targeted delivery of chemically modified siRad18 by engineered RGD-EXOs sensitized OS cells to doxorubicin treatment
To further improve the efficiency of chemotherapy in vivo, we designed a treatment regimen of Rad18 knockdown combined with doxorubicin chemotherapy. Targeted delivery of siRNA loaded in engineered exosomes has become our first choice [31]. Chemical modification effectively improves the stability of siRNA in blood, but the lack of targeting limits its application. Exosomes have become the focus of drug delivery research, which are characterized by high biosafety, easy preparation, and weak immunogenicity [32, 33]. Previously successful applications of engineered exosomes for the targeted delivery of siRNAs guided our approach [34,35,36]. In addition, RGD-EXOs have been proven to target OS in vivo [37,38,39]. Therefore, we designed a targeted delivery scheme based on engineered RGD-EXOs and chemically modified siRNA (Fig. 6A). We constructed a fusion expression vector of RGD and LAMP2, and LAMP2 carried the integrin-targeting RGD peptide to exosome membrane surface, allowing exosomes to target osteosarcoma cells with high integrin expression. We designed siRad18 to knock down the expression of Rad18 in osteosarcoma tissues. SiRad18 was modified with cholesterol to increase stability and coupled with cy3 groups for tracer. SiRNA was loaded into RGD-EXOs by electrical transfer. After loading, RGD-EXOs solution was injected through the tail vein and delivered siRad18 to OS cells of orthotopic osteosarcoma.
We investigated the efficiency of cholesterol-modified cy3-siRad18 delivered to OS cells by engineered exosomes. First, through electron microscopy, we found that RGD-EXOs loaded with siRad18 and empty RGD-EXOs had no obvious morphological changes and little difference in particle size (Supplementary Fig. 6A). Second, siRad18 could be more efficiently integrated into cells via RGD-EXOs than via transfection alone to facilitate gene knockdown (Supplementary Fig. 6B). Finally, cholesterol-modified cy3-siRad18 delivered by engineered exosomes reduced the mRNA and protein expression of Rad18 (Supplementary Fig. 6B, C). In addition, Rad18 knockdown cells had significantly higher sensitivity to doxorubicin (Supplementary Fig. 6D). These results suggest that cholesterol-modified siRNA delivered by engineered RGD-EXOs can function in vitro.
Distribution detection in vivo showed that cholesterol-modified cy3-siRad18 loaded by RGD-EXOs were widely distributed in the liver and spleen but less in the heart, lung and kidney, which may be related to the rich capillary networks in the liver and spleen. Compared with ordinary exosomes, RGD-EXOs were effectively transported to the site of OS (the tibia) and released siRad18, knocking down Rad18 in OS, suggesting that cholesterol-modified cy3-siRad18 loaded by RGD-EXOs can effectively play the role of targeted delivery and knockdown in vivo (Fig. 6B, C and supplementary Fig. 6E). Moreover, we verified the biosafety of RGD-EXOs and RGD-EXOs-siRad18 in vivo. HE-stained sections of all organs showed no tissue or organ damage, and liver and kidney function analysis showed that control RGD-EXOs and RGD-EXOs-siRad18 had no significant effect on the liver and kidney function of nude mice (Supplementary Fig. 6F, G). Subsequently, we applied a combined treatment in which cholesterol-modified cy3-siRad18 loaded by RGD-EXOs were injected through the tail vein and doxorubicin chemotherapy was injected intraperitoneally to inhibit orthotopic OS in nude mice. The treatment lasted for 4 weeks; RGD-EXOs were injected through the tail vein 2 days earlier than doxorubicin to take full advantage of the effects of Rad18 knockdown (Fig. 6D).
In vivo bioluminescence experiments showed that OS tumor growth in the doxorubicin group and the combined treatment group was significantly inhibited compared with that in the no doxorubicin treatment group. It is worth mentioning that although the tumor volume in the combined treatment group was lower than that in the doxorubicin group, there was no significant difference between the two groups (Fig. 6E, F). However, the tumor weight of the combined treatment group was significantly lower than that of the doxorubicin group (Fig. 6G, H). This may be caused by the inadequate sensitivity of bioluminescence signal measurement and the large difference in signal intensity within the group. These results indicated that cholesterol-modified cy3-siRad18 loaded by RGD-EXOs combined with doxorubicin enhanced the in vivo lethality of doxorubicin against OS. Immunohistochemical staining indicated that engineered RGD-EXOs delivered cholesterol-modified cy3-siRad18 could effectively inhibit the expression of Rad18 in OS. The phosphorylation level of H2AX and the expression level of cleaved caspase-3 in the combined treatment group were significantly higher than those in the doxorubicin group (Fig. 6I). This suggests that RGD-EXOs mediated targeted knockdown of Rad18 aggravated doxorubicin-induced DNA damage and promoted OS apoptosis.