Carrimycin structure and synthesis
Carrimycin, also known as bitespiramycin and shengjimycin, was approved recently by the China National Medical Products Administration. It is produced by recombinant Streptomyces spiramyceticus, which harbors a 4″-O-isovaleryltransferase gene (ist) from Streptomyces thermotolerans [2, 20]. Carrimycin’s composition is heterogenous and mostly consists of 4″-O-isovalerylspiramycins (ISP) I, II, III and trace amounts of 4″-O-acylspiramycin components (Fig. 1A).
Genomic sequencing of the Carrimycin-producing bacterial strain revealed a biosynthetic gene cluster of approximately 90 kb (Genbank accession number MH460451) (Fig. 1B). The gene cluster has a Spiramycin biosynthetic gene bundle as well as heterologous ist genes for 4″-O-isovalerylation of Spiramycin. Carrimycin’s structural backbone is a polyketide, putative platenolide I including ethylmalonyl-CoA, methylmalonyl-CoA and methoxymalonyl-CoA, which depends on polyketides synthase (PKS) and malonyl-CoA for its biochemical composition. After several post-PKS tailoring steps, including glycosylation, oxidation, acylation, and isovalerylation, a heterogenous mixture of ISP I, II and III results, differing only by acyl substitutions of the hydroxyl group on carbon 3 (Fig. 1C).
ISP I was isolated and purified from Carrimycin to a purity level of 98.47%, as detected by HPLC on ODS column (Fig. S1A). Its chemical structure was elucidated by spectra methods (Fig. S1B-H). The molecular formula C48H82N2O15 was established by HRMS spectrometry at m/z 927.5800 [M + H]+ (calculated 927.5788) (Fig. S1C). The NMR data were collected by Bruker Avance III. According to the 1H-NMR (Fig. S1D, Supplementary Table 1), 13C-NMR (Fig. S1E, Supplementary Table 1), 1H-1H cosy (Fig. S1F), and HSQC spectra (Fig. S1G), there is a sixteen-membered macrolide skeleton in ISP I along with an isovaleryl moiety and three deoxyhexoses forosamine, mycaminose and mycarose. The HMBC correlations indicated that the forosamine is connected to C-9, the mycaminose is connected with C-5, and the mycarose is connected to mycaminose with (1 → 4) glycosidic bond (Fig. S1H, Supplementary Table 1). The isovaleryl moiety is connected to C-4′′ of mycarose based on the correlation between 4.45 (m, 1H, H-4′′) and 172.09 (C-1′′′′) in HMBC spectrum. Therefore, the structure of ISP I was unambiguously assigned as 4′′-O-isovalerylspiramycin I (Fig. S1B).
ISP I suppresses tumorigenesis and metastasis
To assess potential cytotoxicity, we treated glioblastoma cell lines with serial doses of Carrimycin main components, ISP I, II and III and assessed cell viability; the 50% inhibitory concentration (IC50) values showed that ISP I was most potent (Fig. 2A and B, and Fig. S2A to D). We focused our attention on ISP I.
Multiple tumor cell lines, including glioblastoma, renal cell carcinoma (RCC), and meningioma were sensitive to ISP I’s cytotoxic effect (Fig. 2A and B, and Figs. S4A and B, and S6A and B). Distribution across the cell cycle and apoptosis was assessed by flow cytometry, followed by EdU/DAPI or Annexin V/PI staining, separately. Flow analysis showed ISP I caused cell cycle arrest and induced dose-dependent apoptosis (Fig. 2C and D, and Figs. S3A and B, and S4C to F). Moreover, immunoblotting showed that cleaved PARP, a marker of apoptosis, increased in ISP I-treated LN229 cells and U251 cells (Fig. S3C). Likewise, ISP I increased cleaved caspase 3 and Bax but reduced Bcl-2 in a dose dependent manner (Fig.S3C). Furthermore, transcriptome profiling and Gene Set Enrichment Analysis (GSEA) of ISP I-treated LN229 cells both demonstrated ISP I was able to promote cell apoptosis (Fig. S3D and E).
We next investigated the tumor-suppressing effect of ISP I in vivo, using a human glioblastoma (LN229) xenograft intracranial mouse model (Fig. 2E). NSG mice were inoculated with luciferase-expressing cells in the right frontal cortex. After 7 days, intracranial tumor growth was confirmed with in vivo bioluminescence imaging; mice were randomized into ISP I or DMSO (control) groups. Mice treated with ISP I have significantly reduced tumor growth compared to controls (Fig. 2F and G). We confirmed these results in two additional xenograft models, using RCC (786-O) and meningioma (IOMM) cells (Fig. S5A to C and Fig. S6C to E). In all xenograft mouse models, ISP I treatment was well tolerated and animals maintained their bodyweight (Fig. S7A to C). Finally, we assessed ISP I’s anti-tumor effect in metastatic cancer. We examined two different, syngeneic murine lung metastases models: melanoma (B16) and mammary carcinoma (4T1). C57BL/6 mice were injected intravenously with B16 cells and randomized into ISP I-treated or saline-treated (control) groups (Fig. 2H). The lung metastasis mammary carcinoma model was established by injecting BALB/c mice with 4T1 cells into their second mammary fat pads. Seven days after tumor inoculation, BALB/c mice were randomized into ISP I-treated or saline-treated (control) groups (Fig. S8A). In both lung metastases models, ISP I significantly reduced lung tumor burden. After 12 days of ISP I treatment, B16-bearing mice had significantly reduced the number of lung metastases compared to saline-treated mice (Fig. 2I and J). Similarly, 4T1-bearing mice treated with ISP I had significantly fewer lung tumor nodules after 49 days of treatment compared to saline-treated mice (Fig. S8B and C). Collectively, these findings demonstrated that ISP I inhibited tumor growth in vitro and, more importantly, exerts potent and lasting antitumor effects in several different primary and metastatic tumors.
ISP I targets to selenoprotein H
Given ISP I’s cytotoxicity across a variety of tumor cell lines, we sought to identify ISP I’s molecular target. We performed drug affinity responsive target stability (DARTS) assays in LN229 cells (Fig. 3A shows assay strategy). In DARTS, when ISP I bound a target protein, it created a stable conformational structure that inhibited proteases. Mass spectrum analysis showed that selenoprotein H (SELH), a thioredoxin-like protein with glutathione peroxidase activity [7, 21], was one of the most abundant primary protein presenting in ISP I-treated LN229 cells (Supplementary Table 2). Using a thermo-stability assay, we showed that ISP I protected SELH over a range of increasing temperatures (Fig. 3B and Fig. S9A), suggesting that ISP I targeted SELH.
Peroxidases are conserved enzymes catalyzing redox reactions that reduce hydroperoxides. They include several families with different active redox centers containing selenocysteine residues (e.g., selenoprotein), cysteine thiols (e.g., thioredoxin and thiol peroxidase), or heme cofactors (e.g., catalase) [22]. SELH has a conserved CXXU motif with redox function (cysteine separated by two other residues from selenocysteine) corresponding to the CXXC motif in thioredoxins [7]. To verify the specificity of ISP I’s targeting of SELH, we designed a surface plasmon resonance (SPR) assay to assess the interaction of ISP I with peroxidases synthesized in bacteria, including SELH, Thioredoxin (TrxA), Catalase (KatA), and Thiol peroxidase (TpX). ISP I tightly bound to SELH but not the other peroxidases (Fig. 3C and Fig. S10A to C). ISP I reduced SELH protein expression in a dose-dependent manner in multiple tumor cell lines (Fig. 3D and E, and Fig. S9B). Cycloheximide (CHX) chase assay confirmed that treatment with ISP I decreased SELH protein half-life, which suggested that ISP I promoted SELH protein degradation (Fig. 3F and G). Collectively, these results showed that ISP I targeted SELH and promoted SELH protein degradation.
To confirm that ISP I inhibits cell growth via a SELH-dependent mechanism, we generated SELH-deficient LN229 and RCC cell lines (786-O and RCC4, respectively) with CRISPR-Cas9. SELH-deficient cells resist ISP I treatment when compared to wild-type cells (Fig. 3H and Fig. S9C). Using selective siRNA knock-down, SELH inhibition significantly decreased cell growth rate, retards cell proliferation, and increased apoptosis (Fig. 3I to K, and Figs. S9D to G, and S11A and B). To evaluate whether SELH regulates tumor growth in vivo, NSG mice were orthotopically injected with either wild-type or SELH-deficient LN229 cells. Mice implanted with SELH-deficient LN229 cells have an almost twofold increase in median overall survival compared to mice with wild-type LN229 cells (control) (SELH-deficient LN229 group vs wild-type LN229 group: 51 vs. 28 days, Fig. 3L). SELH’s regulatory tumor growth effect was validated in the melanoma lung metastasis model. C57BL/6 mice were injected with either wild-type B16 cells or SELH-deficient B16 cells (Fig. 3, M to O). Twelve days after tail-vein inoculation, mice with SELH-deficient B16 cells have significantly fewer lung tumors when compared to mice injected with wild-type B16 cells (Fig. 3N and O). Thus, ISP I inhibited growth of primary and metastatic tumors through inhibition of SELH.
ISP I disrupts intracellular redox homeostasis
Because selenoproteins can protect against reactive oxygen species [5, 6], we investigated whether ISP I altered redox homeostasis in vitro. First, we assessed ISP I’s effect on glutathione peroxidase activity. ISP I-treated LN229 cells exhibit significantly reduced glutathione peroxidase activity (Fig. 4A). ROS-Glo H2O2 assay confirms that ISP I-treated tumor cells have increased intracellular ROS levels (Fig. 4B and Fig. S12A). MitoSOX staining, followed by flow cytometry, shows that ISP I-treated cells generate higher intracellular ROS, in a dose-dependent manner (Fig. 4C and D, and Fig. S12B and C).
We assessed whether ISP I alters antioxidant downstream signaling pathways. Transcriptome profiling and GSEA of ISP I-treated LN229 cells both demonstrated widespread changes in oxidative pathways as well as specific upregulation of the nuclear factor erythroid 2-related factor 2 (NFE2L2/NRF2) (Fig. 4E and F), a transcription factor that regulates redox homeostasis by binding to the regulatory regions of antioxidant response elements (ARE) [23]. NFE2L2 expression was significantly upregulated in ISP I-treated cells; ISP I induced an increase in the downstream targets of NFE2L2 signaling, HMOX1 and NQO1 (Fig. 4G and H, and Fig. S12D and E). We measured NFE2L2 expression in glioblastoma cells that either overexpressed SELH or had SELH expression silenced. Overexpression of SELH decreased NFE2L2 protein levels, while NFE2L2 proteins were increased in SELH-silenced cells (Fig. 4I). NFE2L2 mRNA was slightly increased in SELH-silenced U251 cells (Fig. 4J). Of note, SELH-silenced cells had a significant increase in HMOX1 mRNA, consistent with an increase in NFE2L2 protein (Fig. 4K). By suppressing SELH, ISP I induced ROS accumulation and activated antioxidative signaling, including the NFE2L2 pathway.
ISP I triggers DNA damage and R-loop formation
Since increased intracellular levels of ROS cause oxidative DNA damage, which induces genomic instability and inhibits cell cycle progression [24], we assessed whether treatment with ISP I augmented DNA damage in cancer cells. Immunofluorescence staining and western blot analyses revealed that ISP I-treated cells showed higher quantities of γH2AX, a sensitive marker of DNA damage (Fig. 5A and B, and Fig. S12F and G). To further understand the role of ISP I-induced ROS in DNA damage, we performed Immunofluorescence staining and western blot analyses for γH2AX in ISP I-treated LN229 cells and U251 cells with the presence of N-acetylcysteine (NAC), a ROS scavenger. Both analyses showed that supplementing NAC (10 mM) rescued increased γH2AX to the basic level of that in the cells without ISP I treatment (Fig. S13A to C), suggesting that ISP I-induced DNA damage was due to elevated ROS generation.
To assess ISP I’s effect within the nucleus, we evaluated R-loop formation in cancer cells. R-loops are DNA-RNA hybrid structures composed of a displaced single-stranded DNA hybridized with the nascent RNA transcript [25]. R-loops are strongly induced by ROS within the nucleus and can trigger DNA damage, genome instability, and cell cycle arrest [26]. We quantified R-loop formation using a human bone osteosarcoma cell line (U2OS) or a glioblastoma cell line (U251) overexpressing the V5-tagged catalytically-inert RNase H (R-loop marker), a dynamic strategy for R-loop profiling [27]. ISP I-treated cells demonstrated increased R-loop accumulation in a dose-dependent fashion (Fig. 5C and D, and Fig. S14A and B). To assess DNA replication, we stained tumor cells with 5-ethynyl-2′-deoxyuridine (EdU). ISP I reduced EdU incorporation into DNA in a dose-dependent manner (Fig. 5C and I, and Fig. S14A and C). Co-immunofluorescence staining with antibodies specific for nucleolin (a marker of nucleolus), γH2AX and EdU revealed that treatment with ISP I not only consistently increased DNA damage and reduced DNA replication but also altered R-loop localization (Fig. 5C, E to J). R-loop foci concentrated in nucleoplasm instead of the nucleolus in ISP I-treated U251 cells (Fig. 5C). R-loop foci specifically localized at DNA damage sites, not DNA replication sites (Fig. 5G and J). Furthermore, we saw the delocalization of nucleolin in ISP I-treated U251 cells, evidence of nucleolar stress (Fig. 5C). Collectively, ISP I significantly augmented intracellular levels of ROS, which decreased DNA replication and led to DNA damage and R-loop formation.
ISP I impairs nucleolar rRNA transcription via JNK2/TIF-IA pathway
We investigated ISP I’s ability to induce the nucleolar stress response. LN229 cells were immunostained with primary antibodies against nucleolar proteins: Fibrillarin, Nucleophosmin (NPM1), and RNA Polymerase I (POLI). Immunofluorescence imaging revealed increased nucleolar protein dispersion in ISP I-treated LN229 cells and a pronounced decrease in SELH expression within the nucleolus (Fig. 6A). NPM1 is a critical nucleolar protein involved in several key regulatory pathways, including mRNA transport, ribosomal biogenesis, chromatin remodeling, apoptosis, and genome instability [10, 28]. NPM1 protein level fell and p53 (an apoptotic marker) increased in ISP I-treated LN229 cells in a dose-dependent manner (Fig. 6B). GSEA profiling confirmed the induction of the p53 pathway (Fig. 6C and D).
To assess ribosomal biogenesis alterations and probe potential links with nucleolar dysfunction, we analyzed pre-rRNA transcription. ISP I reduced pre-rRNA (POLI transcript) in a dose-dependent manner (Fig. 6E). The quantities of mature 18S RNA (rRNA cleavage) and pre-GAPDH mRNA (POL II transcript) were not altered (Fig. S15A); conversely, pre-tRNA 13 (POL III transcript) and some mature tRNAs (tRNA) were increased (Fig. S15A and B). Stress-dependent inhibition of POLI transcription is mediated through inactivation of TIF-IA, a POLI-specific co-activator, via specific phosphorylation of TIF-IA at a single threonine residue (Thr 200) by JNK2 [11]. Phosphorylation at Thr 200 impairs the interaction of TIF-IA with POLI thereby abrogating initiation complex formation and rRNA synthesis [11]. JNK2 belongs to the family of stress-activated protein kinases that play a crucial role in the cellular response to oxidative stress [29]. ISP I increased JNK2 phosphorylation in a dose-dependent manner, indicating elevated cellular ROS-induced JNK2 activation (Fig. 6F). Although ISP I-treatment led to a slightly decreased trend in POLI and TIF-IA expression, co-immunoprecipitation assays showed that the physiological interaction between TIF-IA and POLI was significantly disrupted in ISP I-treated or SELH-deficient LN229 cells (Fig. 6G), suggesting that activated JNK2 mainly impaired TIF-IA and POLI interaction in SELH-deficient cells. To confirm reduced POLI initiation complex formation and rRNA transcription in SELH-deficient cells, chromatin immunoprecipitation (CHIP) assay was performed in ISP I-treated or SELH-deficient LN229 cells. CHIP assay showed reduced POLI recruitment to the promoter and the coding regions of rDNA (5.8S, 28S and 5’ETS) in ISP I-treated or SELH-deficient LN229 cells (Fig. 6H). This suggested that ISP I induced suppression of SELH impaired POLI-rDNA interaction by activating the ROS/JNK2/TIF-IA pathway. These findings demonstrated that ISP I disrupted ribosomal biogenesis by suppression of POLI transcription but does not alter the mature RNA machinery.