Gene interfered-ferroptosis therapy for cancers

Although some effective therapies have been available for cancer, it still poses a great threat to human health and life due to its drug resistance and low response in patients. Here, we develop a ferroptosis-based therapy by combining iron nanoparticles and cancer-specific gene interference. The expression of two iron metabolic genes (FPN and LCN2) was selectively knocked down in cancer cells by Cas13a or microRNA controlled by a NF-κB-specific promoter. Cells were simultaneously treated by iron nanoparticles. As a result, a significant ferroptosis was induced in a wide variety of cancer cells. However, the same treatment had little effect on normal cells. By transferring genes with adeno-associated virus and iron nanoparticles, the significant tumor growth inhibition and durable cure were obtained in mice with the therapy. In this work, we thus show a cancer therapy based on gene interference-enhanced ferroptosis.

Although some effective therapies have been available for cancer, it still poses a great threat to human health and life due to its drug resistance and low response in patients. Therefore, various cancer therapies have been being developed to induce cancer cell death through one form of cell death such as apoptosis, necrosis, necroptosis, and pyroptosis1. Recently, ferroptosis has attracted increasing attention as another form of cell death. Ferroptosis was identified as an iron-dependent nonapoptotic cell death in 20122. Subsequently, it was found that ferroptosis is triggered by iron-catalyzed lipid peroxidation initiated by nonenzymatic (Fenton reactions) and enzymatic mechanisms [lipoxygenases (LOXs)]3. In the former case, hydroxyl radicals are directly generated by the Fenton reaction between ferrous iron (Fe2+ ) and hydrogen peroxide and initiate nonenzymatic lipid peroxidation4,5.

Since being found, ferroptosis has attracted considerable attention due to its potential role as a target for novel therapeutic anticancer strategies6,7,8. Because ferroptosis is a form of regulated cell death, the immune system may function in part through ferroptosis to prevent tumorigenesis9. More interestingly, it was found that ferroptosis can propagate among cells in a wave-like manner, exhibiting a potent killing effect on neighboring cells10,11,12. Another attractive point of ferroptosis is its potential to overcome drug resistance such as chemoresistance13,14,15,16,17,18,19, and improve cancer immunotherapy20. Therefore, the antitumor effects of ferroptosis has been being rapidly and widely explored in variant cancers21. However, most of these studies focused on finding various ferroptosis inducers of compounds for inhibiting or depleting system xc−, GPX4 and CoQ1022. These compounds may be challenged by the same resistance problem as the traditional cancer drugs20. For example, IREB2 was identified as one high-confidence gene in the erastin-induced ferroptosis, however, when REB2 was silenced, the expression of iron metabolism-related genes would change, making cells resistant to ferroptosis2,23,24.

Currently, iron oxide nanoparticles (IONPs) have been successfully used as MRI contrast agent in cancer diagnosis and anemia treatment25,26,27. However, IONPs have not yet been used in the cancer treatment. Nevertheless, many studies have shown that IONPs can release ferrous (Fe2+ ) or ferric (Fe3+ ) in acidic lysosome in cells. The released Fe2+ participates in the Fenton reaction to generate toxic hydroxyl radicals (•OH), one of reactive oxygen species (ROS), to induce ferroptosis of cancer cells11,28,29. Therefore, the ferroptosis-driven nanotherapeutics for cancer treatment become increasingly attractive30. Moreover, it was found that nanoparticle-induced ferroptosis in cancer cells eliminates all neighboring cells in a propagating wave10. However, cells are sensitive to iron concentration, and a little fluctuation in concentration can cause a great response31. Due to the importance of iron to cells, cells have evolved out a set of mechanisms to maintain intracellular iron homeostasis32. The intracellular iron is thus under delicate regulation to keep iron homeostasis33,34,35. Cells are therefore capable to effectively store and export the intracellular excess iron ions released from internalized IONPs. For example, in previous study, we found significant up-regulation of transcription of important genes responsible for exporting intracellular iron ion in cells treated by a DMSA-coated Fe3O4 nanoparticle36. Therefore, the ferroptosis induced by a single IONPs treatment is too little to be utilized to treat cancer in clinics.

Recently, an interesting study reported an IONPs-induced ferroptosis that was enhanced by a genetic change. In the study, it was found that a nanoparticle iron supplement, ferumoxytol, has an anti-leukemia effect in vitro and in vivo in leukemia cell with a low level of FPN33. This antitumor effect resulted from the inability of the type of leukemia cell to export the intracellular Fe2+ generated from ferumoxytol, which makes the type of leukemia cell susceptible to the elevated ROS generated by Fe2+ through Fenton reaction. However, this study relies on the rare naturally occurred low expression of ferroptosis-related genes such as FPN, still not resolving the problem how to artificially lower the expression of these genes in tumors of patients, so that this IONPs-induced ferroptosis therapy can be applied to more wide types of cancers.

Nuclear factor-kappa B (NF-κB) is a sequence-specific DNA-binding transcription factor that is over activated in almost all cancers37,38. Therefore, to inactivate the protein, we designed a gene expression vector that can express an artificial microRNA targeting NF-κB RELA/P65under the control of a NF-κB-specific promoter39. This promoter consists of a NF-κB decoy sequence and a minimal promoter and thus its transcriptional activation activity is dependent on the NF-κB activity in cells39. We therefore name it as DMP, meaning Decoy-Minimal Promoter. Our previous results verified that DMP could be used to realize cancer cell-specific gene expression by depending on the NF-κB over-activity in cancer cells40,41. By using the promoter, we developed two kinds of cancer gene therapy. One is a cancer immunotherapy induced by an artificial neoantigen displayed on cancer cell surface40. The other is a cancer gene therapy induced by cutting telomere via CRISPR/Cas9 or oncogenic mRNAs via Cas13a41,42. In these studies, we successfully controlled neoantigen and Cas9 or Cas13a expression selectively in cancer cells by the DMP promoter.

In this work, based on these previous studies, we deduce that inhibition of the genes exporting intracellular iron ions should be lethal to the IONPs-treated cancer cells, because the inhibition would prevent cell from exporting IONPs-produced iron ions, which can thus enhance IONPs-induced ferroptosis. Base on this speculation, we develop a cancer therapy named as gene interfered-ferroptosis therapy (GIFT) by combining the DMP-controlled gene interference tools with the DMSA-coated Fe3O4 nanoparticles (FeNP). By specifically knocking down expression of two iron metabolic genes, FPN and LCN2, with the DMP-controlled CRISPR/Cas13a and microRNA (miRNA) in cancer cells, the FeNP treatment induced dramatic ferroptosis in a wide variety of cancer cells that represent various hematological and solid tumors. However, the therapy has little effect on normal cells. By using both viral (adenovirus associated virus, AAV) and non-viral (PEI-modified Fe3O4nanoparticles, FeNC) vectors, the growth of different xenografted tumors in mice were also significantly inhibited by the therapy.

Conceptualization of GIFT

The principle of GIFT is schematically illustrated in Fig. 1a. GIFT consists of a gene interfering vector (GIV) and FeNP. GIV is composed of a promoter named DMP and downstream effector gene. DMP is a NF-κB-specific promoter that consists of a NF-κB decoy and a minimal promoter (Fig. 1b). Because NF-κB is a transcription factor that is widely over-activated in cancers, the expression of effector gene can be activated in cancer cells by NF-κB binding to DMP (Fig. 1a). In contrast, the effector gene cannot be expressed in normal cells due to lack of NF-κB (Fig. 1a). Therefore, DMP is a cancer cell-specific promoter. When the GIV of DMP-controlled CRISPR/Cas13a or miRNA is transfected into cancer cells, Cas13a or miRNA can be expressed. The expressed Cas13a protein can associate with guide RNA (gRNA) transcribed from a U6 promoter to form Cas13a-gRNA complex. The expressed miRNAs can associate with RNA-induced silencing complex (RISC). Both Cas13a-gRNA and miRNA-RISC complexes can target mRNA of interest to knockdown the expression of target gene in cancer cells.

a Schematic show of CRISPR/Cas13a- and miRNA-based GIFT. DCUg DMP-Cas13a-U6-gRNA, DM DMP-miRNA, U6p U6 promoter, gRNA guide RNA, DMP NF-κB decoy-minimal promoter, Fe3O4 NP Fe3O4 nanoparticle. b Schematic show of DMP and NF-κB dimer. c QPCR detection of NF-κB RELA/P65 expression in various cells (n = 3 biologically independent samples). The statistical significance was obtained by comparing the data of cancer cells with normal cells. Data are presented as mean ± standard deviation (SD) and analyzed by one-way ANOVA with Tukey’s test.

Full size image

In this study, we selected two genes related to iron metabolism, FPN and LCN2, as target genes. The functions of both FPN and LCN2 in cells are related to the efflux of iron ions43,44,45,46,47,48. In our previous study, we found the expression of the two genes were significant up-regulated when cells were treated with FeNP36. Therefore, we speculated that knocking down the expression of the two genes in cancer cells by the transfected GIVs of DMP-controlled CRISPR/Cas13a or miRNA would prevent cells from exporting the intracellular iron ions produced by the internalized FeNP. This can result in accumulation of iron ions and cause significant increase of intracellular ROS level, thus leading to significant ferroptosis in cancer cells. In normal cells, because Cas13a or miRNA cannot be produced, the expression of the two genes cannot be affected, which allows the normal cells to actively export the intracellular iron ions produced by FeNP and maintain the iron homeostasis, leading to little effects from FeNP treatment.

Expression of NF-κB RELA in various cells

NF-κB is widely activated in nearly all types of tumor cells. However, it is expected that its expression is not the same among cancer cells49,50. Because the intracellular NF-κB activity is critical to our strategy. Therefore, we first detected the expression of NF-κB RELA/P65 in 3 leukemia cell lines (KG-1a, HL60 and WEHI-3), 16 solid tumor cell lines, and 7 normal cell lines (HL7702, MRC-5, HMEC-1, NIH-3T3, L929, GES-1, and MCF12A). The results demonstrated that NF-κB RELA/P65 highly expressed at different levels in all cancer cells, but did not in all normal cells (Fig. 1c). Therefore, DMP should drive a cancer cell-specific gene expression.

Effects of FeNP on cell viability

To evaluate the cytotoxicity of FeNP that was stable at the long-time storage at 4 °C (Supplementary Fig. 1a), we dynamically measured the cell viability of three leukemia cells, a solid tumor cell (HepG2), and two normal human cells that were treated by various concentrations of FeNP for 5 days. The results showed FeNP had no significant toxicity to all cells below the dose of 50 μg/mL (Supplementary Fig. 1b). However, when the dosage was over 50 μg/mL, FeNP showed significant toxicity to all cells including normal cells (Supplementary Fig. 1b). Therefore, we used the FeNP at the concentration of 50 μg/mL for subsequent investigation, which is equivalent to the dose of intravenous injection of 3 mg/kg in rodents33.

Antitumor effect of GIFT in vitro

To investigate the antitumor effects of GIFT, we first treated leukemia cells by GIFT. A human leukemia cell (KG-1a) was first transfected by various plasmid vectors and then treated by FeNP. The cell viability was detected by a acridine orange and ethidium bromide (AO&EB) dual staining at three time points. The results showed that only the combinations of all GIVs with FeNP caused the significant time-dependent cell death (Fig. 2a and Supplementary Fig. 2). All plasmids and FeNP alone and the combinations of negative control plasmids (pDCUg-NT and pDM-NT) with FeNP did not significantly affect cell viability (Supplementary Fig. 2). Especially, the co-expression of GIVs (pDCUg-FL and pDM-FL) showed the most significant cancer cell killing effect when combining with FeNP (Supplementary Fig. 2), suggesting a synergistic effect of co-interfering two genes. This same effect of GIFT was then observed on a human solid tumor cell HepG2 (Fig. 2a and Supplementary Fig. 3).

To investigate the broad spectrum of GIFT in killing cancer cells, we next similarly treated a variety of cancer cells representing different leukemia and solid tumors in human and mouse, including 2 leukemia cells (HL60 and WEHI-3) and 14 solid tumor cells (A549, HT-29, PANC-1, SKOV3, MDA-MB-453, C-33A, BGC-823, SGC-7901, MGC-803, KYSE450, KYSE510, B16F10, Hepa1-6, and CT-26) with the co-expressed GIVs (pDCUg-FL and pDM-FL). The results revealed that the combinations of the two GIVs with FeNP produced the significant time-dependent killing effects in all tumor cells (Supplementary Figs. 4‒19). Similarly, the vectors and FeNP alone and the combinations of negative control plasmids (pDCUg-NT and pDM-NT) with FeNP did not significantly affect all cells at any treatment time (Supplementary Figs. 4‒19).

To investigate the cancer cell specificity of GIFT, we next treated seven normal human and mouse cells (HL7702, MRC5, GES-1, HMEC-1, L929, MCF12A, and NIH-3T3). The results showed that all vectors alone and their combinations with FeNP did not significantly affect all these cells (Fig. 2a and Supplementary Figs. 20‒26), which is consistent with low NF-κB expression in these cells (Fig. 1c). To further validate the key role of NF-κB activation in GIFT, we transfected these cells with pDCUg-FL and pDM-FL respectively and then induced them with a NF-κB activator, TNF-α. The cells were then treated by FeNP. We found that these cells were also significantly killed by GIFT with the TNF-α inducement (Fig. 2a and Supplementary Figs. 20‒26). This confirms that cells can be killed by GIFT only when NF-κB is activated, which is also confirmed by the treatment of HEK-293T cell with GIFT. The HEK-293T cell is a human embryonic kidney cell that expresses large T antigen after being transfected with a virus. Although the cell is not considered as a cancer cell, its NF-κB expression is significantly activated (Fig. 1c). Therefore, the same GIFT effects as cancer cells were seen in this cell (Supplementary Fig. 27).

To further characterize the cell viability under the various treatments, we counted the numbers of alive cells in the AO&EB-stained images with ImageJ and identified statistical significance with GraphPad. The result showed that only the combinations of all GIVs with FeNP caused the significant time-dependent cancer cell death (Fig. 2b and Supplementary Figs. 28 and 29). After a treatment of 72 h, few cells remained alive, indicating the strong cancer cell killing effect of GIFT. The cell viability was further confirmed by detecting cell death, which also revealed the efficiency and specificity of GIFT leading cancer cells to die (Fig. 2c and Supplementary Figs. 30‒32). The death detection also revealed the synergistic effect of co-interfering two genes (Fig. 2c and Supplementary Figs. 30‒32).

To evaluate the interfering effect of DMP-Cas13a/U6-gRNA (DCUg) and DMP-miR (DM) systems, we next detected the expression of FPN and LCN2 genes in the treated cells. The results revealed that the expression of the two genes were significantly knocked down at both mRNA (Fig. 3a) and protein (Fig. 3b) levels by the targeting gRNAs/miRNAs in all detected cancer cells (KG-1a, HL-60, and HepG2). However, no significant changes were found in the HL7702 cell, further indicating the cancer cell specificity of the designed gene interfering systems. To further explore the cancer cell-specific expression of effector gene, we detected Cas13a mRNA in the treated cells. The results revealed that Cas13a only expressed in all cancer cells treated by pDCUg vectors (Fig. 3a).

Cells were transfected by plasmid vectors and cultured for 24 h, and then incubated with or without 50 μg/mL of FeNP. Gene expression, ROS production, and iron content of cells were detected at 48 h post FeNP administration. a QPCR analysis of mRNA expression. b Western blot assay of protein expression. Samples were run on the same blot. The representative image and quantified optical density are shown. c Measurement of ROS levels. The fluorescence shift and quantified fluorescence intensity are shown. a.u., arbitrary units. d Measurement of iron content. c, d only show the data of four cell lines (KG1a, HepG2, HL7702, and MRC5). The representative images of flow cytometry detection of ROS are shown in Supplementary Fig. 33. The detection results of ROS and iron content of other two cells (HL60 and WEHI-1) are shown in Supplementary Fig. 34. c, d use a same set of symbols. Data are presented as mean ± SD (n = 3 biologically independent samples) and analyzed by one-way ANOVA with Tukey’s test.

Full size image

GIFT antitumor by ferroptosis

It has been reported that iron-based nanomaterial can up-regulate ROS levels through the Fenton reaction28. To investigate whether ROS was produced in the FeNP-treated cells, we measured the ROS levels in four cells under the various treatments. The results revealed that the combinations of GIVs with FeNP resulted in the highest levels of ROS in all cancer cells (Fig. 3c and Supplementary Figs. 33 and 34a). However, the FeNP alone and the combinations of negative control vectors with FeNP only resulted in a little increase of ROS in all cancer cells (Fig. 3c and Supplementary Figs. 33 and 34a). In contrast, all the same treatments only resulted in a little increase of ROS in two normal cells (HL7702 and MRC5) (Fig. 3c and Supplementary Figs. 33 and 34a). Nevertheless, when two normal cells were induced by TNFα, the combinations of GIVs with FeNP immediately resulted in high ROS levels in two normal cells (Fig. 3c and Supplementary Fig. 33).

To further confirm the origin of ROS, we next measured the iron content in the treated cells. The results revealed that the intracellular iron content of four cancer cells significantly increased under the treatment of FeNP together with GIVs (Fig. 3d and Supplementary Fig. 34b). However, the FeNP alone and the combinations of negative control vectors with FeNP only resulted in a limited increase of iron content (Fig. 3d and Supplementary Fig. 34b). Similarly, when two normal cells were induced by TNFα, the combinations of GIVs with FeNP significantly resulted in high iron content in two normal cells (Fig. 3d). These data are consistent with the ROS levels in the four cancer cells under the same treatment. These data also show that the iron efflux was significantly inhibited by the knockdown of FPN and LCN2in cancer cells (Fig. 3a, b), which leads to a great increase of intracellular ROS and iron contents (Fig. 3c, d).

To confirm the mechanism underlying GIFT is ferroptosis, we next treated the HepG2 cell with GIFT in the presence of various inhibitors including ferroptosis inhibitors (ferrostatin-1, Fer-1; liproxstatin-1, lipro-1), iron chelator (deferoxamine, DFO), cell-permeable analog of cysteine (N-acetylcysteine, NAC), apoptosis inhibitor (ZVAD-FMK, ZVAD), necrosis inhibitor (Necrostatin-1s, Nec1s), and autophagy inhibitor (Bafilomycin A1, BA1). The cell viability was detected with Cell-Titer-Glo 2.0. The results showed that the GIFT treatment significantly reduced the cell viability (Fig. 4a). The treatment of cells with Fer1, lipro-1, DFO, and NAC markedly reduced the GIFT-induced cell death, whereas inhibitors of apoptosis, necroptosis, or autophagy had little impact on cell death (Fig. 4a). Moreover, the cell viability was most significantly rescued by the cotreatment with Fer1, DFO, and NAC (FDN) (Fig. 4a). The similar results were also obtained in the cancer cells MDA-MB-453, CT-26, KG-1a, PANC1, and WEHI-3, and the TNFα-induced normal cells HMEC-1 and MRC-5 (Supplementary Fig. 35).

To further confirm the mechanism underlying GIFT is ferroptosis, we next detected the lipid ROS (a hallmark of ferroptosis) by using a lipid oxidation indicator, C11-BODIPY. It was found that only the GIFT treatment significantly induced lipid ROS in the HepG2 cell (Fig. 4b, c and Supplementary Fig. 36). Additionally, the increased lipid ROS could be reverted by Fer1, lipro-1, DFO, and NAC, but not by ZVAD, Nec1s, and BA1 (Fig. 4b, c and Supplementary Fig. 36). Moreover, the cotreatment with Fer1, DFO, and NAC (FDN) most strongly reverted the GIFT-induced lipid ROS (Fig. 4b, c and Supplementary Fig. 36). The similar results were also obtained in the cancer cells MDA-MB-453, KG-1a, WEHI-3, PANC1, and CT-26, and the TNFα-induced normal cells MRC-5 and HMEC-1 (Supplementary Figs. 37‒44). The colony formation assay of four cell lines (HepG2, MDA-MB-453, CT-26, and PANC1) revealed that the GIFT treatment could nearly eradicate the colony formation capability of various tumor cells; however, the treatment of Fer1 and cotreatment of Fer1, DFO and NAC (FDN) could restore some colonies (Fig. 4d, e and Supplementary Figs. 45 and 46). These results indicated that the mechanism underlying GIFT is the designed gene-interfered ferroptosis.

Virus-based GIFT antitumor in vivo

To explore the in vivo antitumor effects of GIFT, we next cloned the GIVs into AAV to prepare recombinant viruses (rAAV). The rAAVs were tested by transfecting three cells (KG-1a, WEHI-3, and HL7702). The results showed that the combinations of gene interfering rAAVs (rAAV-DCUg-FL and rAAV-DM-FL) with FeNP resulted in significant death of two cancer cells (Supplementary Fig. 47). However, all rAAVs and FeNP alone and the combinations of negative control rAAVs (rAAV-DCUg-NT and rAAV-DM-NT) with FeNP did not significantly affect the growth of all cancer cells (Supplementary Fig. 47). In human normal cell HL7702, all treatments caused no significant cell death (Supplementary Fig. 47).

To investigate the in vivo antitumor effect of GIFT, we next treated the cancer cell xenograft mouse with GIFT. The mouse leukemia cell WEHI-3 was subcutaneously transplanted into BALB/c mice to make tumor-bearing mice. Two batches of animal treatments were performed. The dosage of different viruses and FeNP were 1 × 1010 vg/mouse and 3 mg/kg, respectively. In the first batch of animal treatment, six groups of tumor-bearing mice were treated with a single intravenous injection of separate rAAV and FeNP. The results showed that only the rAAV-DCUg-FL + FeNP treatment significantly inhibited tumor growth (Supplementary Fig. 48a, b). In the second batch of animal treatment, five groups of tumor-bearing mice were treated with a single intravenous injection of mixed rAAV and FeNP. The results indicated that only the rAAV-DM-FL + FeNP treatment significantly inhibited tumor growth (Supplementary Fig. 48c, d). These results indicate that both separate and mixed intravenous injection can be used by GIFT. Therefore, to simplify the drug administration, we used the mixed intravenous injection in the following studies. To further characterize the in vivo antitumor effect of GIFT, we detected the abundance of DNA of GIV and mRNA of effector gene Cas13a and two target gene FPN and LCN2 in various tissues by qPCR. The results showed that the GIV DNA distributed in all detected tissues, especially in tumor (Supplementary Fig. 48e), and Cas13a was only expressed in tumors (Supplementary Fig. 48f). The expression of two target genes (FPN and LCN2) was only significantly knocked down in tumors (Supplementary Fig. 48g, h).

To improve the treatment effect, we subsequently treated tumor-bearing mice of WEHI-3 cell with multiple intravenous injections of rAAV-DM-NT/FL + FeNP (1 × 1010 vg/mouse plus 3 mg/kg FeNP). The tumor-bearing mice were intravenously administered three times with phosphate-buffered saline (PBS), rAAV-DM-NT + FeNP, and rAAV-DM-FL + FeNP (Supplementary Fig. 49a), respectively. The results indicated that the tumor growth was more significantly inhibited (Supplementary Fig. 49b). Moreover, the treatment with rAAV-DM-FL + FeNP significantly improved the survival of mice (Supplementary Fig. 49c). To further improve the treatment effect and check if tumor could be eradicated by GIFT, we next treated tumor-bearing mice of WEHI-3 cell with multiple intravenous injections of rAAV-DM-NT/FL + FeNP at a higher dosage (2 × 1010 vg/mouse plus 3.6 mg/kg FeNP). The tumor-bearing mice were intravenously administered three times with PBS, rAAV-DM-NT + FeNP, and rAAV-DM-FL + FeNP (Fig. 5a), respectively. As a result, the rAAV-DM-FL + FeNP treatment more significantly inhibited tumor growth (Fig. 5b–e) and splenomegaly (Fig. 5f, g). The same treatment also more significantly prolonged the survival time of mice (Fig. 5h–k). However, no significant body weight change was observed during treatment in both mice used for measuring tumor inhibition (Fig. 5b) and survival (Fig. 5i), suggesting the good safety of GIFT reagents. Additionally, the addition of NAC to drinking water eliminated tumor inhibition (Fig. 5a–g) and survival improvement (Fig. 5h–k) of the rAAV-DM-FL + FeNP treatment. Furthermore, the in vivo antitumor effect of GIFT could be reverted by a more specific ferroptosis inhibitor, liproxstatin-1 (Supplementary Fig. 50). These data indicated that GIFT inhibits tumor by ferroptosis, consistent with the in vitro results. To further characterize the treatment, we also detected the iron content and the abundance of DNA of GIV and mRNA of RELA, FPN, and LCN2 in various tissues. The results indicated that the rAAV-DM-FL + FeNP treatment significantly increased the iron content of tumors (Supplementary Fig. 51a), but not other tissues (Supplementary Fig. 51b). The GIV DNA distributed in all detected tissues, especially in tumor (Supplementary Fig. 51c). RELA was only highly expressed in tumors (Supplementary Fig. 51d) and two target genes FNP and LCN2 were only significantly knocked down in tumor (Supplementary Fig. 51e, f). Furthermore, the expression of two tumor growth markers, CD31 and Ki67, was also significantly downregulated by the treatment of rAAV-DM-FL + FeNP (Supplementary Fig. 51g, h), which is consistent with the H&E staining of tumor sections (Supplementary Fig. 51i). Finally, the safety of the in vivo treatment was confirmed by the H&E staining of the main organs (Supplementary Fig. 52a) and detection of blood biochemical markers including white blood cells (WBC), red blood cells (RBC), platelet (PLT), and hemoglobin (HGB), hepatotoxicity markers including alanine aminotransferase (ALT), aspartate aminotransferase (AST), and alkaline phosphatase (ALP), and kidney injury markers including blood urea nitrogen (BUN), creatinine (Cr), and uric acid (UA) (Supplementary Fig. 52b).

To further confirm the observed antitumor effect of GIFT, we treated a new xenograft tumor model with GIFT. The tumor-bearing BALB/c mice were produced with the xenograft CT-26 cell, a widely used mouse colorectal cancer cell line. The tumor-bearing mice were intravenously administered three times with rAAV-DM-NT + FeNP and rAAV-DM-FL + FeNP at the same dose (2 × 1010 vg/mouse plus 3.6 mg/kg FeNP), respectively (Fig. 6a). The results revealed that the similar significant tumor growth inhibition (Fig. 6b–g) and survival improvement (Fig. 6h–k) were reproduced on the new tumor model. Detections of iron content of tumors (Supplementary Fig. 53a), abundance of rAAV DNA (Supplementary Fig. 53b), and mRNA of RELA (Supplementary Fig. 53c), FPN (Supplementary Fig. 53d), and LCN2(Supplementary Fig. 53e) obtained the similar results that were observed in the WEHI-3 model. The antitumor effects were also confirmed by the transcription of CD31(Supplementary Fig. 53f) and Ki67 (Supplementary Fig. 53g) and the H&E staining of tumor sections (Supplementary Fig. 53h). The safety of the in vivo treatment was again confirmed by the H&E staining of the main organs (Supplementary Fig. 54a) and detection of blood biochemical markers, hepatotoxicity markers, and kidney injury markers (Supplementary Fig. 54b).

To determine whether the GIFT could induce antitumor effect to the metastatic tumors, we finally treated the lung metastatic melanoma formed by intravenously injected B16F10 cells in C57BL/6J female mice. The tumor-bearing C57BL/6J mice were intravenously administered three times with rAAV-DM-NT + FeNP and rAAV-DM-FL + FeNP at the same dose (2 × 1010 vg/mouse plus 3.6 mg/kg FeNP), respectively (Fig. 7a). The results indicated that the rAAV-DM-FL + FeNP treatment significantly reduced tumor burden of lung (Fig. 7b–h). Similarly, rAAV-DM-FL + FeNP treatment significantly improved the survival of mice (Fig. 7i–k). Moreover, the rAAV-DM-FL + FeNP treatment significantly decreased splenomegaly (Supplementary Fig. 55a, b) and white blood cells (Supplementary Fig. 55c). Although the rAAV distributed in all tissues (Supplementary Fig. 55d), the GIFT reagents did not cause significant toxicity to the major organs (Supplementary Fig. 55e–h).

To further confirm the safety of GIFT treatment of such multiple intravenous injection, we then treated two groups of healthy BALB/c mice with three times intravenous injection of PBS and rAAV-DM-FL + FeNP (2 × 1010 vg/mouse plus 3.6 mg/kg FeNP), respectively (Supplementary Fig. 56a). The results revealed that the rAAV-DM-FL + FeNP treatment did not cause significant loss of body weight (Supplementary Fig. 56b). Moreover, although the rAAV distributed in all tissues (Supplementary Fig. 56c), the GIFT reagents did not cause significant toxicity to the major organs (Supplementary Fig. 56d–g). These data together with those detected in tumor-bearing mice (Supplementary Figs. 52, 54, and 55) demonstrate the safety of the GIFT treatment.

FeNC-based GIFT antitumor in vitro and in vivo

Finally, to find whether iron nanoparticle itself can be used to transfer GIVs as AAV, we selected a PEI-modified Fe3O4 nanoparticle (FeNC) as a DNA transfection agent. Two batches of FeNC (FeNC-1 and FeNC-2) were evaluated by two experiments. In the first experiment, plasmids were added to FeNC-1 to prepare FeNC-1@DNA. The KG-1a and HepG2 cells were first treated with FeNC-1@DNA (0.5 μg FeNC-1) just for transfecting GIVs and then treated with FeNP (50 μg FeNP). The results showed that only FeNC-1@pDM-FL/pDCUg-FL + FeNP resulted in significant cell death (Supplementary Figs. 57 and 58), whereas FeNC-1@pDM-FL/pDCUg-FL (0.5 μg FeNC-1) and FeNP alone and FeNC-1@pDM-NT/pDCUg-NT + FeNP did not significantly affect the cell growth (Supplementary Figs. 57 and 58). These data indicate that FeNC can be used as gene transfection agent in GIFT. In the second experiment, to simplify the GIFT reagents, we removed FeNP and just treated the KG-1a cell with FeNC@DNA. The results indicated that the treatment of FeNC@pDM-FL (50 μg FeNC) induced significant cell death (Supplementary Fig. 59), whereas FeNC (50 μg FeNC) and pDM-FL alone did not significantly affect the cell growth (Supplementary Fig. 59). To investigate the stability of FeNC@DNA, we also treated the KG-1a cell with the FeNC@pDM-FL kept for 24 h. The results showed that FeNC@DNA still had similar cancer cell killing effect (Supplementary Fig. 59). The followed lipid ROS and cell viability detections also indicated that FeNC@pDM-FL made cell death by ferroptosis (Supplementary Figs. 60 and 61).

To further investigate the antitumor effect of FeNC@DNA, we treated the tumor-bearing mice of WEHI-3 cell with a single intravenous injection of FeNC@pAAV-DCUg-FL/DM-FL and other reagents as controls. The results indicated that only the treatment of FeNC@pAAV-DCUg-FL/DM-FL induced significant tumor growth inhibition (Supplementary Fig. 62a, b). The detection of plasmid DNA revealed that the vector DNA distributed in all detected tissues (Supplementary Fig. 62c). The detection of Cas13a indicated that this effector gene was only expressed in tumors (Supplementary Fig. 62d). Moreover, the transcription of two target genes FPN and LCN2 were only significantly knocked down in tumors (Supplementary Fig. 62e, f). To further improve the antitumor effect of FeNC@DNA, we treated the tumor-bearing mice of WEHI-3 cell with two times intravenous injections of FeNC@pAAV-DM-FL and FeNC@pAAV-DM-NT, respectively (Supplementary Fig. 62g). The results indicated that the tumor growth was more significantly inhibited (Supplementary Fig. 62h–j). The results also showed that the pAAV-DM-FL + FeNC treatment only significantly increased the iron content of tumors (Supplementary Fig. 62k).

Pharmacokinetics and biodistribution

Finally, we investigated the pharmacokinetics and biodistribution of rAAV-DM-FL, FeNP, pAAV-DM-FL, and FeNC. The results indicated that the concentration of rAAV-DM-FL, FeNP, pAAV-DM-FL, and FeNC in blood became approximately 9.91%, 5.41%, 2.7%, and 4.44% of the injected dose per gram of tissue (% ID/g) at 24 h post injection, respectively (Supplementary Fig. 63). The in vivo half-life time of them was 11.1 h, 5.0 h, 5.2 h, and 4.9 h, respectively (Supplementary Fig. 63). The detection of in vivo biodistribution of rAAV-DM-FL, FeNP, pAAV-DM-FL, and FeNC in tumors and major organs indicated that all these reagents were significantly sequestered by the reticuloendothelial system in the liver and spleen compared with other organs (Supplementary Fig. 63). The 29.5%, 24.3%, 20.4%, and 23.4% of rAAV, FeNP, pAAV, and FeNC (%ID/g) were accumulated in the tumors at 24 h post injection, respectively (Supplementary Fig. 63).

Sufficient concentration of systemically delivered iron oxide nanoparticles in tumors is important for GIFT. This study demonstrates that two used iron nanoparticles can concentrate to tumors (Supplementary Figs. 51a, b and 63b, d) and GIFT thus significantly increased the iron content of tumors (Supplementary Figs. 51a, 53a and 62k). Because the used nanoparticles were not modified with any tumor-specific ligands, the nanoparticles should accumulate to tumors via passive enhanced permeability and retention effect (EPR) and active transport through trans-endothelial pathways51. The latter is recently reported accounting for the majority of nanoparticle accumulation in tumors51. Because FeNP is modified with no tumor-specific ligands, it can be generally used to treat three different kinds of tumors (Figs. 5–7). At present, few highly specific biomakers were identified on tumor cell surface, especially on solid tumors. Therefore, preparing nanoparticles with highly tumor specificity is still difficult. On the other hand, if a targeting ligand specific to a kind of tumor is modified on nanoparticles, this kind of nanoparticles may be unsuitable for treating other tumors because different tumors have different targets52, which can complicate the preparation of nanoparticles for treating different tumors. Anyway, the nanoparticles modified with a tumor-specific ligand may improve the therapeutical effect of GIFT to the tumor by accumulating more nanoparticles. GIFT has its unique benefits by using iron oxide nanoparticles and AAV. Iron oxide nanoparticles are the only approved metallic nanoparticles for clinical use because of their demonstrated safety. The approved applications of iron oxide nanoparticles include cancer diagnosis, cancer hyperthermia therapy, and iron deficiency anemia53. Moreover, AAV is also an approved gene vector for clinical use and now more and more gene therapies employ it for its safty54,55.