Transcriptional pausing induced by ionizing radiation enables the acquisition of radioresistance in nasopharyngeal carcinoma

Abstract

Lesions on the DNA template can impact transcription via distinct regulatory pathways. Ionizing radiation (IR) as the mainstay modality for many malignancies elicits most of the cytotoxicity by inducing a variety of DNA damages in the genome. How the IR treatment alters the transcription cycle and whether it contributes to the development of radioresistance remain poorly understood. Here, we report an increase in the paused RNA polymerase II (RNAPII), as indicated by the phosphorylation at serine 5 residue of its C-terminal domain, in recurrent nasopharyngeal carcinoma (NPC) patient samples after IR treatment and cultured NPC cells developing IR resistance. Reducing the pool of paused RNAPII by either inhibiting TFIIH-associated CDK7 or stimulating the positive transcription elongation factor b, a CDK9–CycT1 heterodimer, attenuates IR resistance of NPC cells. Interestingly, the poly(ADP-ribosyl)ation of CycT1, which disrupts its phase separation, is elevated in the IR-resistant cells. Mutation of the major poly(ADP-ribosyl)ation sites of CycT1 decreases RNAPII pausing and restores IR sensitivity. Genome-wide chromatin immunoprecipitation followed by sequencing analyses reveal that several genes involved in radiation response and cell cycle control are subject to the regulation imposed by the paused RNAPII. Particularly, we identify the NIMA-related kinase NEK7 under such regulation as a new radioresistance factor, whose downregulation results in the increased chromosome instability, enabling the development of IR resistance. Overall, our results highlight a novel link between the alteration in the transcription cycle and the acquisition of IR resistance, opening up new opportunities to increase the efficacy of radiotherapy and thwart radioresistance in NPC.

Introduction

Radiotherapy is an essential treatment for ∼50% of all malignancies, particularly non-metastatic solid tumors like nasopharyngeal carcinoma (NPC), which is a geographically unbalanced malignancy prevalent mostly in Southeast Asia (Chen et al., 2019). High-energy ionizing radiation (IR) causes tumor regression by generating a variety of lesions in living cells. In addition to oxidative damages, every Gray (Gy) of IR induces roughly 1000 single-strand breaks (SSBs) and 40 double-strand breaks (DSBs) to the nuclear DNA, which elicit most of the cytotoxicity after IR exposure (Ward, 1988; Goldstein and Kastan, 2015). Thus, comprehensive characterization of cellular responses to IR-induced DNA damages has been the central theme in the field to improve treatment efficacy, as alterations in these responses often underlie the development of radioresistance, which significantly limits the clinical benefits.

As the rate-limiting step of gene expression, the dynamic process of transcription underlies many cellular responses toward intrinsic and extrinsic stimuli. A full transcription cycle consists of four sequential phases—initiation, pausing, elongation, and termination—which are regulated by post-translational modifications on the C-terminal domain (CTD) of the Rpb1 subunit of RNA polymerase II (RNAPII) (Harlen and Churchman, 2017). After initial recruitment to promoter regions by transcription factors and general transcription factors, RNAPII CTD is phosphorylated at serine 5 (S5) of the Y₁S₂P₃T₄S₅P₆S₇ repeat by the TFIIH-associated kinase CDK7, escaping from promoters and then quickly becoming paused after transcribing 20–60 nucleotides downstream of the transcription start site (TSS) (Chen et al., 2018; Core and Adelman, 2019). The positive transcription elongation factor b (P-TEFb), a cyclin-dependent kinase complex consisting of CDK9 and CycT1, phosphorylates serine 2 (S2) of RNAPII CTD, thereby driving the paused RNAPII to enter elongation phase (Harlen and Churchman, 2017; Core and Adelman, 2019; Noe Gonzalez et al., 2021). As a master regulator governing global transcriptional elongation, P-TEFb is subject to complex controls. The majority of P-TEFb in fast-growing cells is sequestered in an inactive form by hexamethylene bisacetamide-inducible proteins (Hexim), binding with 7SK small nuclear ribonucleoprotein (snRNP) (Zhou et al., 2012). The CycT1 subunit of P-TEFb can undergo liquid–liquid phase separation, forming a compartment that entraps a sufficient amount of CDK9 with RNAPII CTD to promote transcriptional elongation (Lu et al., 2018). In fact, the promoter-proximal pausing of RNAPII and its release have been an emerging nexus for regulation in many biological processes (Core and Adelman, 2019).

Transcription is intrinsically connected with DNA damage responses, as unscheduled encounters of RNAPII with different lesions on the DNA template can lead to catastrophic consequences such as transcriptional failure and genome instability. During the repair of bulky single-strand DNA damages induced by ultraviolet (UV) light, RNAPII functions as a damage sensor. UV damage triggers an immediate release of paused RNAPII into gene bodies to detect local lesions and initiate transcription-coupled nucleotide excision repair (Lavigne et al., 2017). Following this initial phase, RNAPII is then cleared from the damaged DNA template and degraded by ubiquitination, which leads to a global shutdown of transcription (Noe Gonzalez et al., 2021). In response to double-strand DNA damages, a repressive chromatin state mediated by histone H2AK119ub is formed around the damaged sites to halt local RNAPII activity and prevent transcription errors (Machour and Ayoub, 2020). Recently, we reported that, upon DNA damages, the phase separation of CycT1 was disrupted by PARP-1-mediated poly(ADP-ribosyl)ation, inhibiting global transcription to ensure transcription fidelity and genome stability (Fu et al., 2022). At certain stimulus-inducible genomic loci, breaks in the DNA template can promote transcriptional elongation dependent on topoisomerase II (Bunch et al., 2015; Madabhushi et al., 2015). How radiotherapy-induced complex DNA damages modulate the cellular transcriptional dynamics and whether it influences the treatment efficacy have not been investigated.

Here, we report that the transcriptional pausing-related S5 phosphorylation (pS5) is increased in recurrent NPC samples from patients who have received radiotherapy and in IR-resistant NPC cell lines, indicating an accumulation of paused RNAPII. Downregulation of RNAPII pausing level by stimulation of P-TEFb activity or chemical inhibition of the pS5 kinase CDK7 overcomes IR resistance in both cell culture and a xenograft model. Remarkably, the poly(ADP-ribosyl)ation of CycT1 that disrupts its phase separation is elevated in IR-resistant NPC cells, and mutation of major poly(ADP-ribosyl)ation sites of CycT1 decreases RNAPII pausing and restores IR sensitivity. Chromatin immunoprecipitation followed by sequencing (ChIP–seq) analyses of RNAPII distributions confirm the increased transcriptional pausing not only in NPC cells that have developed IR resistance but also in cells shortly after being exposed to IR, pinpointing a group of genes involved in radiation response and cell cycle control that are regulated by the paused RNAPII. These genes include several known therapeutic resistance factors, such as TP53, ATRX, JUNB, and EIF4E, as well as a new radioresistance candidate, the NIMA-related kinase NEK7. Knockdown of NEK7 leads to the increased radioresistance in IR-responsive cells, and overexpression of NEK7 sensitizes the IR-resistant cells to irradiation. Taken together, our results reveal an unprecedented function of transcriptional pausing in the acquisition of cellular resistance to IR, suggesting that the transcription cycle is a promising target for the development of new radiosensitizers to improve treatment response and prevent resistance.

Results

RNAPII pS5 increases in recurrent NPC patient samples and IR-resistant cell lines

During transcription, the paused RNAPII is enriched in pS5 and the elongating RNAPII is enriched in S2 phosphorylation (pS2), mainly catalyzed by the kinase activity of TFIIH and P-TEFb, respectively (Figure 1A). To investigate the change in RNAPII transcriptional status during the development of resistance to IR, we collected five paired pathologic slices of matched primary and recurrent tissues from the same NPC patients and performed RNAPII pS5 antibody-based immunohistochemistry (IHC). By calculating the IHC scores (Jian et al., 2021), we found that the recurrent tumor samples manifested a small but steady increase in the overall pS5 level compared with the paired primary tumors (Figure 1B), suggesting that an increase pool of paused RNAPII was associated with radioresistance in clinical settings.

Figure 1.

Elevated RNAPII pS5 in recurrent NPC patient samples and NPC cells developing IR resistance. (A) Schematic of RNAPII CTD phosphorylation in transcription. S5P, phosphorylated S5; S2P, phosphorylated S2. (B) Left: representative IHC staining for RNAPII pS5 in five paired histopathologic slices of primary and recurrent NPC. Scale bar, 50 μm. Right: semiquantitative comparison of RNAPII pS5 IHC scores between five paired primary and recurrent tumors. Data represent mean ± SD. *P < 0.05 by paired t-test. (C) Immunofluorescence (purple) probing RNAPII pS5 and pS2 in CNE2 and CNE2-IR cells. DNA was visualized by DAPI staining (blue). Scale bar, 20 μm. (D) Western blot detecting RNAPII pS5 and pS2 levels in CNE2 and CNE2-IR cells, with tubulin as the loading control. (E) RNAPII pS5 level in different IR-responsive and IR-resistant cancer cell lines from independent experiments. (F) Western blot detecting RNAPII pS5 level in NPC cell line at 48 h after exposure to the indicated doses of irradiation. (G) Quantification of RNAPII pS5 level in the indicated cells after irradiation from independent experiments. Error bars represent SD. ns, not significant, *P < 0.05, **P < 0.01 by t-test.

To ascertain the relationship between the altered transcriptional status of RNAPII and radioresistance, we further analyzed two previously validated NPC cell lines, CNE2 and Hone1, and their IR-resistant derivatives (Tan et al., 2018; Shen et al., 2021). Immunofluorescence and western blotting results showed a marked increase of pS5 and an accompanying decrease of pS2 in the IR-resistant CNE2-IR cells compared with the parental IR-responsive CNE2 cells (Figure 1C and D). A similar increase in pS5 of RNAPII was observed in the IR-resistant Hone1-IR cells relative to the parental line (Figure 1E). These observations suggested that an increase in transcriptional pausing indicated by the pS5 level of RNAPII was a shared property of NPC cells developing IR resistance.

IR-resistant cells are derived from parental IR-responsive cell lines by repeated exposure to IR. To test whether IR exposure could effectively increase the pS5 level of RNAPII, we irradiated the cells with different radiation doses and, after 2 days, detected the pS5 level by western blotting (Figure 1F and G). Both the IR-responsive and IR-resistant cell lines displayed upregulated pS5 levels after IR treatment, indicating that the kinetics of transcription were sensitive to IR and irradiation increased the pool of paused RNAPII. This increase was, at least in part, reserved in the IR-resistant cells and likely contributing to the acquisition of radioresistance.

Reduction of RNAPII pS5 attenuates IR resistance

To test whether the elevated transcriptional pausing contributed to the development of IR resistance, we modulated the transcription cycle using two different strategies and evaluated the impacts on IR responsiveness. Hexim1 forms a ribonucleoprotein complex with 7SK snRNP to sequester and inactivate P-TEFb (Figure 1A), and inhibiting the function of Hexim1 can promote P-TEFb-dependent transcriptional elongation (Michels and Bensaude, 2018). We first knocked down the expression of Hexim1 in CNE2-IR cells to reduce pausing and promote elongation (Figure 2A). The CNE2-IR cells with Hexim1 knockdown showed a significant decrease in cell viability upon IR treatment (Figure 2B). This result suggested that increasing the pool of active P-TEFb could attenuate IR resistance.

Figure 2.

Modulation of RNAPII pS5 reverses IR resistance. (A) qPCR analyzing Hexim1 mRNA level in CNE2-IR cells infected with lentiviruses carrying empty vector as control (NT) or shRNA targeting Hexim1 (#1 or #2). Cell lysates were collected on Day 5 after infection. Error bars indicate SD. (B) The combined inhibitory effect of Hexim1 knockdown with radiotherapy in CNE2 and CNE2-IR cells. Error bars indicate SD. **P < 0.01, ***P < 0.001 by t-test. (C) Cell viability assay of CNE2 and CNE2-IR cells treated with different concentrations of THZ1 or relative DMSO control. The viability was measured on Day 5 using MTT. (D) Western blot detecting RNAPII pS5 level in the indicated experimental groups. (E) The combined inhibitory effects in CNE2 and CNE2-IR cells treated with 150 nM THZ1 and IR. Error bars indicate SD. ***P < 0.001 by t-test. (F) Schematic of the experimental procedure testing IR resistance in a xenograft tumor model. (G) Tumor volumes in different experimental groups after 4 Gy IR treatment. Data represent mean ± SD. n = 5. (H) Immunofluorescence staining detecting RNAPII pS5 (green) level in the frozen sections from the mouse subcutaneous tumors. Cell nucleus was visualized by DAPI staining (blue). Scale bar, 1000 μm and 10 μm (inset).

RNAPII S5 is phosphorylated by the TFIIH-associated CDK7 kinase, and THZ1 is a potent covalent inhibitor of CDK7 (Ali et al., 2009; Kwiatkowski et al., 2014). We next investigated whether chemical inhibition of CDK7 could influence the responsiveness to radiation. When treated with THZ1, CNE2-IR cells were more sensitive relative to CNE2 cells, with IC₅₀ of 209.6 nM versus 1027.0 nM, respectively (Figure 2C). Given that THZ1 can also inhibit CDK12 at high concentrations (Kwiatkowski et al., 2014), we treated the cells with 150 nM THZ1 to minimize off-target effects. Western blotting results confirmed that the pS5 level was downregulated by THZ1 treatment (Figure 2D). Moreover, combining THZ1 treatment with irradiation effectively reduced the cell viability of CNE2-IR cells (Figure 2E).

To further test whether the downregulation of RNAPII pS5 by THZ1 treatment could reverse radioresistance in vivo, we seeded CNE2-IR cells subcutaneously into nude mice and generated a xenograft model (Figure 2F). After the tumor volume reached 200 mm³, mice in the experimental groups were injected with vehicle or THZ1 and received 4 Gy irradiation. THZ1 was then administrated daily and the tumor volume was measured thrice a week. As shown in Figure 2G, radiation alone only mildly limited the growth of CNE2-IR tumors, whereas radiation combined with THZ1 treatment completely blocked the tumor enlargement. We dissected the tumors at the end of the experiment and prepared frozen sections for immunostaining with RNAPII pS5 antibody (Figure 2H). As expected, THZ1 treatment downregulated the RNAPII pS5 level in the CNE2-IR tumor tissues. These results suggested that inhibiting CDK7 activity to lower the RNAPII pS5 level could weaken IR resistance.

Disruption of CycT1 phase separation by poly(ADP-ribosyl)ation underlies IR resistance

DNA strand break after radiotherapy is accompanied by the activation of PARP1, which can poly(ADP-ribosyl)ate many substrate proteins including CycT1, disrupting its liquid–liquid phase separation and inhibiting transcription (Figure 3A). We first assessed the total poly(ADP-ribosyl)ation level in CNE2 and CNE2-IR cells with or without IR treatment. The poly(ADP-ribosyl)ation level was increased in both cells immediately after IR exposure (Figure 3B). More interestingly, poly(ADP-ribosyl)ation in CNE2-IR cells was significantly higher than that in CNE2 cells even without IR treatment, indicating that IR-resistant cells harbored a higher baseline poly(ADP-ribosyl)ation level. We immunoprecipitated the endogenous CycT1 from CNE2 and CNE2-IR cells and observed that, in the IR-resistant CNE2-IR cells, the poly(ADP-ribosyl)ated CycT1 was increased (Figure 3C). To investigate whether inhibiting PARP1 and downregulating the poly(ADP-ribosyl)ation level could restore the IR sensitivity of CNE2-IR cells, we treated cells with the PARP inhibitor Olaparib. CNE2-IR cells were more sensitive to Olaparib compared with CNE2 cells, with IC₅₀ of 3904 nM versus 6403 nM, respectively (Figure 3D). Furthermore, Olaparib treatment effectively reduced the poly(ADP-ribosyl)ation of CycT1 (Figure 3E), and the combination of Olaparib with radiation markedly decreased the cell viability of CNE2-IR cells (Figure 3F), suggesting that downregulating the poly(ADP-ribosyl)ation of CycT1 by Olaparib could reverse IR resistance.

Figure 3.

Disruption of CycT1 phase separation by poly(ADP-ribosyl)ation underlies IR resistance. (A) Schematic of poly(ADP-ribosyl)ation disrupting phase separation of CycT1 after IR treatment. (B) Western blot detecting poly(ADP-ribosyl)ation level in CNE2 and CNE2-IR cells after exposure to the indicated doses of irradiation, with tubulin as the loading control. (C) Western blot detecting poly(ADP-ribosyl)ation level of immunoprecipitated CycT1 in CNE2 and CNE2-IR cells. (D) Cell viability assay in CNE2 and CNE2-IR cells treated with different concentrations of Olaparib. The viability was measured after 4 days using MTT. (E) Western blot detecting poly(ADP-ribosyl)ation level of immunoprecipitated CycT1 in 10 Gy-irradiated CNE2-IR cells, pre-treated with or without 1000 nM Olaparib for 1.5 h. (F) Cell viability assay detecting the combined inhibitory effects in CNE2 and CNE2-IR cells treated with 1000 nM Olaparib and IR. Error bars indicate SD. ***P < 0.001 by t-test. (G) CNE2-IR cells were depleted of CycT1 (CycT1 KD) and reconstituted with stably expressed CycT1-WT or CycT1-Mut2. Western blot detecting CycT1 level in control, CycT1 KD, and the reconstituted cells. (H) Western blot detecting RNAPII pS5 and pS2 levels in CycT1-WT or CycT1-Mut2-reconstituted CNE2-IR cells. (I) Cell viability assay detecting cell survival after exposure to the indicated doses of IR in CNE2-IR cells reconstituted with CycT1-WT or CycT1-Mut2. n = 3. ***P < 0.001 by t-test. (J) Schematic of the experimental procedure testing IR resistance in a xenograft tumor model. (K) Tumor volumes of the indicated experimental groups. Data represent mean ± SD. n = 5. ***P < 0.001 by t-test. (L) Tumor weights of the indicated groups at experimental termination. ***P < 0.001 by t-test.

To directly investigate whether the upregulated poly(ADP-ribosyl)ation of CycT1 in CNE2-IR cells was involved in the maintenance of radioresistance, we knocked down the expression of endogenous CycT1 in CNE2-IR cells by shRNA and substituted it with wild-type CycT1 (CycT1-WT) or a mutant CycT1 (CycT1-Mut2) in which all the major poly(ADP-ribosyl)ation sites were mutated to alanine (Figure 3G). CycT1-Mut2 can no longer be poly(ADP-ribosyl)ated and is resistant to DNA damage-induced disruption of liquid–liquid phase separation (Fu et al., 2022). In the CNE2-IR cells expressing CycT1-Mut2, RNAPII pS5 was downregulated and pS2 was upregulated, compared with that in the cells expressing CycT1-WT (Figure 3H). Consequently, the CNE2-IR cells expressing CycT1-Mut2 were more sensitive to irradiation than the cells expressing CycT1-WT (Figure 3I).

To further assess the contribution of CycT1 poly(ADP-ribosyl)ation to radioresistance in vivo, we seeded the CNE2-IR cells expressing either CycT1-WT or CycT1-Mut2 subcutaneously into nude mice. The tumors received 4 Gy irradiation (2 Gy per day) on Day 17 and Day 18 after the inoculation and were measured every 3 days (Figure 3J). On Day 25, which was 7 days after IR treatment, the tumors formed by the CNE2-IR cells expressing CycT1-Mut2 started to regress, whereas those expressing CycT1-WT continued to grow (Figure 3K). We dissected and weighed the tumors at the end of the experiment and found that the tumors from the CNE2-IR cells expressing CycT1-Mut2 were significantly smaller than those expressing CycT1-WT after irradiation (Figure 3I), confirming that the mutation of CycT1 poly(ADP-ribosyl)ation sites restored the radiosensitivity of the IR-resistant cells. Together, these results suggested that the upregulation of CycT1 poly(ADP-ribosyl)ation in the IR-resistant cells, which reduced the P-TEFb activity by disrupting its phase separation and thus increased the pool of paused RNAPII, was crucial for the maintenance of radioresistance.

Many genes involved in radiation response and cell cycle control are regulated by IR-induced transcriptional pausing

To understand the molecular underpinnings of the impacts of transcriptional pausing on the development of radioresistance, we performed ChIP–seq analyses with different RNAPII antibodies and characterized their genome-wide distributions (Figure 4A). We evaluated the short-term or persistent effects of IR on transcription by comparing the ChIP–seq results of CNE2 cells exposed to 4 Gy irradiation or CNE2-IR cells, respectively, to those of CNE2 control. A pausing index (PI) can be calculated for each gene using ChIP–seq data generated with the total RNAPII antibody (Day et al., 2016; Figure 4B). The overall PI in the irradiated CNE2 cells was increased relative to that in the untreated CNE2 cells (Figure 4C), and similarly, the overall PI was significantly higher in the radioresistant CNE2-IR cells compared to the CNE2 control (Figure 4D; Supplementary Table S2). We further analyzed the distributions of RNAPII pS5 and RNAPII pS2. There was a global increase of RNAPII pS5 around the TSS in both the irradiated CNE2 cells and the radioresistant CNE2-IR cells (Figure 4E and F). Concomitantly, RNAPII pS2 decreased both at the TSS and across the whole gene body region (Figure 4G and H), validating the increase of pS5 and decrease of pS2 shown by western blots and immunofluorescence (Figure 1C and D).

Figure 4.

IR-induced global increase in transcriptional pausing. (A) Schematic showing the experimental procedures for ChIP–seq and RNA-seq analyses. (B) Calculation of PI using RNAPII ChIP–seq data. (C and D) Violin plots comparing PI between CNE2 and 4 Gy-irradiated CNE2 cells (C) or between CNE2 and CNE2-IR cells (D). **P < 0.01, ***P < 0.001 by Kolmogorov–Smirnov test. (E–H) Profiles of the RNAPII pS5 (E and F) and RNAPII pS2 (G and H) ChIP–seq signals on genomic regions spanning 5 kb upstream of the TSS and 5 kb downstream of the transcription end site (TES) of the human RefSeq genes in the indicated cells.

To further elaborate the IR-induced alterations in transcriptional pausing, we identified 2895 genes showing increased PI in the irradiated CNE2 cells compared to the untreated control and 1133 PI-increased genes in the radioresistant CNE2-IR cells relative to the IR-responsive CNE2 cells. A total of 331 genes were shared between these two groups (Figure 5A; Supplementary Table S3). Gene ontology (GO) analysis of the 331 genes showed that many of these genes were involved in the response to DNA damage stimuli and cell cycle regulations (Figure 5B; Supplementary Table S4). The observed increase in RNAPII pausing could affect gene expression. We performed RNA sequencing (RNA-seq) analysis with CNE2 and CNE2-IR cells and found that 83 out of the 331 genes displayed the reduced transcription in the radioresistant CNE2-IR cells (Figure 5C; Supplementary Table S5). Of note, several genes previously reported as therapeutic resistance factors, such as TP53, JUNB, ATRX, and EIF4E, were identified in our analysis (Carter et al., 2016; Koschmann et al., 2016; Fan et al., 2017; Hientz et al., 2017). The NIMA-related kinase NEK7, which has been reported to participate in centrosome duplication as well as NLRP3 inflammasome activation (Yissachar et al., 2006; Kim et al., 2011; Sharif et al., 2019; Yang et al., 2020), was also identified in our analysis (Figure 5D). We then designed different primer sets targeting the TSS and gene body regions of TP53 and NEK7 according to the ChIP–seq results and validated by quantitative polymerase chain reaction (ChIP–qPCR) that the pS5 of RNAPII increased at the TSS and the pS2 decreased at the gene body in the irradiated CNE2 cells or the radioresistant CNE2-IR cells (Figure 5E). We also confirmed by qPCR that the mRNA abundance of NEK7 was downregulated in CNE2-IR cells, likely as a consequence of the observed accumulation of paused RNAPII around the TSS (Figure 5F).

Figure 5.

IR-induced transcriptional pausing on genes involved in cell cycle control. (A) Venn diagram of genes with increased transcriptional pausing. (B) GO analysis of the overlapping 331 genes. (C) Volcano plot showing the expression of the 331 genes with increased pausing. Upregulated (red) and downregulated (blue) genes were determined with the cut-off values of adjusted P < 0.05 and |Log₂(fold change)| > 0.585. (D) Genomic snapshots of ChIP–seq and RNA-seq signals at the NEK7 and TP53 loci. (E) ChIP–qPCR of RNAPII pS5 around the TSS and RNAPII pS2 in the gene body regions of TP53 and NEK7 genes in the indicated cells. (F) qPCR results showing the expression of NEK7 in the indicated cells. Error bars indicate SD. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001 by t-test.

Downregulation of NEK7 contributes to radioresistance via influencing mitotic spindle formation

To investigate the involvement of NEK7 in the development of radioresistance, we knocked down its expression using shRNA in CNE2 cells and tested IR sensitivity. TP53 was included as a positive control (Figure 6A and B). While the control CNE2 cells manifested a severe decline in viability in response to the increasing doses of irradiation, TP53 or NEK7 shRNA-treated cells gained significant radioresistance. We then performed the reciprocal experiment by overexpressing NEK7 in CNE2-IR cells and observed an increase in radiosensitivity (Figure 6C and D). These results together justified that NEK7 was a new radioresistance regulator.

Figure 6.

Downregulation of NEK7 contributes to radioresistance via influencing mitotic spindle formation. (A) Cell survival curves of CNE2-IR or CNE2 cells infected with lentiviruses expressing control shRNA (NT) or shRNA targeting TP3 or NEK7 after irradiation with different doses. n = 3. ***P < 0.001 by t-test. (B) qPCR results showing the knockdown efficiencies in CNE2 cells. Error bars indicate SD. (C) Survival curves of CNE2, CNE2-IR, or NEK7-overexpressing CNE2-IR cells after irradiation with different doses. n = 3. ***P < 0.001 by t-test. (D) qPCR results showing the mRNA level of NEK7. Error bars indicate SD. (E) Flow cytometry analysis of the cell cycle using propidium iodide. Error bars indicate SD. ns, not significant, ***P < 0.001 by t-test. (F) Time-course cell cycle analysis on synchronized CNE2 and CNE2-IR cells. (G) Fluorescent staining of the indicated cells in mitosis. Mitotic centrosomes were labelled with anti-γ-tubulin (purple), spindles with anti-α-tubulin (green), centromeres with anti-ACA (red), and chromosomes with DAPI (blue). Scale bar, 10 μm. (H) Quantification of multipolar spindles in the indicated cells. n = 50. ***P < 0.001 by chi-squared test. (I) Confocal images from live-cell imaging experiments. Chromosomes were visualized with mCherry-H2B (red) and spindles with GFP-tubulin (green). The numbers on the top right represent elapsed time (min). Scale bar, 10 μm. (J) Quantification of the mitotic duration. Error bars indicate SD. ***P < 0.001 by t-test.

Altered cell cycle control and increased genome instability are key enabling processes in tumor progression (Hanahan, 2022). We then characterized the cell cycles of CNE2 and CNE2-IR cells and investigated the impacts of NEK7 knockdown using flow cytometry. The radioresistant CNE2-IR cells manifested a G2/M arrest compared to the IR-responsive CNE2 cells (Figure 6E). We further performed time-course cell cycle analysis with synchronized cells and confirmed that CNE2-IR cells took longer than CNE2 cells to proceed from S phase to next G1 phase (Figure 6F). Of note, the cell cycle alteration in the CNE2 cells with NEK7 knockdown resembled that in CNE2-IR cells (Figure 6E).

We next analyzed the mitotic process using immunofluorescence staining. Both CNE2-IR and NEK7 shRNA-treated CNE2 cells displayed defects in the formation of bipolar mitotic spindles (Figure 6G), with the ratio of multipolar spindles greatly exceeding that in the control CNE2 cells (Figure 6H; 54% in CNE2-IR, 40% in CNE2 with NEK7 knockdown, and 8% in CNE2). To further investigate the impact of the observed mitotic spindle defects on the progression of mitosis, we performed live-cell imaging using ectopically expressed GFP-tubulin and mCherry-H2B to visualize the mitotic spindle and chromosomes, respectively (Figure 6I). The duration of mitosis was ∼2 h in CNE2 cells, whereas in CNE2-IR or NEK7 shRNA-treated CNE2 cells, the mitotic phase was significantly extended (Figure 6J; 113 ± 8 min in CNE2, 255 ± 18 min in CNE2-IR, and 270 ± 19 min in CNE2 with NEK7 knockdown), with increased errors in chromosome segregation. These results suggested that the delay in mitosis was in part responsible for the prolongation of G2/M phase and centrosomal activity of NEK7 contributed to the development of radioresistance via affecting the accuracy of chromosome segregation.

Discussion

The development of radioresistance in NPC is polymodal and involves many intracellular and extracellular alterations (Buckley et al., 2020). These alterations include but are not limited to changes in the cell cycle, DNA damage response, and transcription. It is intriguing how NPC cells initiate these different alterations in response to irradiation. In this study, we reported an IR-induced increase of global transcriptional pausing, enabling the development of radioresistance. IR exposure upregulated the pausing-related RNAPII pS5 level, which was partially reserved during the development of radioresistance. Inhibition of TFIIH-associated kinase CDK7 to lower the pS5 level could attenuate IR resistance; meanwhile, stimulation of the P-TEFb activity to promote transcriptional elongation, either by knockdown of Hexim1 or by mutation of poly(ADP-ribosyl)ation sites of CycT1, restored the radiosensitivity of IR-resistant cells. Moreover, we found many genes involved in distinct biological processes, such as radiation response and cell cycle control, were sensitive to the IR-induced transcriptional pausing. Of particular interest, the expressions of TP53 and NEK7 were subject to such control, leading to their downregulation in the IR-resistant cells. Knockdown of NEK7 led to increased radioresistance in IR-responsive cells; reciprocally, overexpression of NEK7 sensitized the IR-resistant cells to irradiation, validating the NIMA-related kinase NEK7 as a new radioresistance factor (Figure 7). Our observations suggest that the IR-induced transcriptional pausing could be one initiating event to trigger downstream alterations leading to the development of radioresistance in NPC. It will be compelling to test in future studies whether this mechanism holds true in other types of cancer.

Figure 7.

Working model. Upregulation of the poly(ADP-ribosyl)ation of CycT1 after irradiation disrupts its phase separation and thus increases the pool of paused RNAPII. This IR-induced transcriptional pausing decreases the expression of many genes such as TP53 and NEK7, leading to the development of IR resistance.

DNA DSBs induced by IR cause the majority of the cytotoxic effects. Previous studies on DSBs and transcription have suggested a two-way relationship. On the one hand, DSBs can cause local accumulation of inhibitory H2AK119ub and transcriptional repression (Shanbhag et al., 2010; Pankotai et al., 2012; Dellino et al., 2019; Machour and Ayoub, 2020). On the other hand, transcriptional elongation at certain genomic loci requires DNA breaks (Bunch et al., 2015; Madabhushi et al., 2015), and the release of paused RNAPII can promote DSBs and cancer-associated translocation, especially around the boundaries of topologically associating domains (Dellino et al., 2019). Our study demonstrates that IR leads to a global increase in RNAPII pausing, which adds another layer of regulation to our understanding of the transcriptional plasticity in DNA damage response. Interestingly, the increase in transcriptional pausing persists in cells that have acquired IR resistance, and it is necessary for the maintenance of their radioresistant state. These findings highlight the transcription cycle as a promising target for the development of new radiosensitizers to improve treatment response and prevent resistance in NPC, and one such candidate is CDK7 inhibitors. The recurrent NPC patient tumor samples displayed an elevated pS5 level of RNAPII, and treatment with the CDK7 inhibitor THZ1 lowered the pS5 level of the IR-resistant CNE2-IR cells and reversed their radioresistance, indicating that THZ1 can be enlisted as a new class of adjuvant chemo drugs to enhance the efficacy of radiotherapy toward NPC. Consistently, a recent report also suggested that the combination of CDK7 inhibition with radiation is a viable therapeutic strategy for Myc-driven medulloblastoma (Veo et al., 2021). Yet, it is noteworthy that THZ1 shows additional inhibitory activity to CDK12/13 at higher concentration (Kwiatkowski et al., 2014). More preclinical and clinical studies are needed to further advance this emerging paradigm.

In addition to NEK7, whose downregulation contributes to the development of radioresistance likely via increasing genomic instability and promoting tumor evolution. Our results also reveal that TP53 is subject to the regulation of IR-induced transcriptional pausing. In response to genotoxic stress, p53 is phosphorylated by ATM, ATR, and other kinases in the signaling cascades, preventing the MDM2-mediated ubiquitination and proteasomal degradation. The stabilized p53 accumulates in the nucleus and activates the expression of several hundred genes, inducing cell cycle arrest and DNA repair or cell death if the DNA damage is irreparable or sustained (Friedel and Loewer, 2022). How p53 activation results in different cell fates has drawn intensive attention in the field (Lees et al., 2021; Rizzotto et al., 2021; Abuetabh et al., 2022; Friedel and Loewer, 2022; Okazaki, 2022). One prevailing model is based on the dynamics of p53. Pulsatile p53 accumulation often facilitates cell cycle arrest and DNA repair, promoting cell survival; sustained p53 accumulation, however, leads to apoptosis and senescence (Carlsen and El-Deiry, 2021; Stewart-Ornstein, 2021; Friedel and Loewer, 2022). The dynamics of p53 upon DNA damages are context-specific and best characterized at the post-translational level. For example, the phosphatase WIP1 can dephosphorylate p53, counteracting the kinase activities in DNA damage response and limiting p53 accumulation (Friedel and Loewer, 2022). To what extent the transcriptional regulation of TP53, particularly that imposed by the IR-induced RNAPII pausing reported here, modifies p53 dynamics is currently unclear. We surveyed our ChIP–seq and RNA-seq data from NPC cells and found that the majority of DNA damage response genes upstream of p53, as well as the genes participating in the non-homologous end joining and homologous recombination DNA repair pathways, were not regulated by the IR-induced transcriptional pausing. However, many genes downstream of p53 manifested reduced expression levels in the IR-resistant CNE2-IR cells, accompanying the observed downregulation of TP53 transcription (Supplementary Table S6). Known as the guardian of the genome, TP53 is the most frequently mutated gene in human cancers. IR-resistant NPC cells manifested more frequent chromosome segregation errors and a higher baseline poly(ADP-ribosyl)ation level. The downregulation of TP53 expression by IR-induced RNAPII pausing might help the IR-resistant cells to survive with the sustained intrinsic DNA damage signals. Future elaborations of p53 dynamics during the development of radioresistance are needed to fully understand its function in enabling the decision-making of tumor cells.

What can be the molecular mechanism underlying the IR-induced increase in transcriptional pausing? P-TEFb governs the pause–release of RNAPII, and its subunit CycT1 undergoes liquid–liquid phase separation to form a compartment that concentrates CDK9 with RNAPII CTD, thus promoting efficient transcriptional elongation (Lu et al., 2018). IR-induced DNA damage activates PARP1, which poly(ADP-ribosyl)ates CycT1, disrupting its ability to undergo phase separation and reducing the pool of elongating RNAPII (Fu et al., 2022). As a result, the pool of paused RNAPII increases, downregulating the expression of several target genes and enabling the development of IR resistance. This molecular model highlights that the elevated PARP-1 activity is central to the acquisition of resistance to radiotherapy. In agreement with this, it has been reported that PARP-1 is overexpressed in NPC cells and its inhibition can enhance radiotherapy (Chow et al., 2013), and our data demonstrated that the PARP-1 inhibitor Olaparib can restore the radiosensitivity of IR-resistant NPC cells. In fact, in several types of cancer, PARP inhibitors have already entered phase I or phase II clinical studies as radiosensitizers (Lesueur, 2017). Results from these investigations will unravel their real potential in clinical settings. Additionally, the involvement of CycT1 phase separation in regulating the sensitivity to radiotherapy reported here also opens up unlimited possibilities to target disease-related liquid–liquid phase separation events in future drug discoveries (Alberti and Dormann, 2019).

Materials and methods

Cell culture

The NPC cell lines CNE2 and Hone1 were obtained from Dr Rong Tan and Ms Na Li (Xiangya Hospital, Central South University). The corresponding IR-resistant cell lines were validated as previously described (Fu et al., 2019). All the cells were authenticated by short tandem repeat (STR) profiling and tested for mycoplasma contamination. CNE2, Hone1, and 293T were cultured with DMEM/High Glucose medium (Biological Industries) containing 10% fetal bovine serum (VISTECH, SE100-B) in a humidified incubator with 5% CO₂ at 37°C.

IHC

All experiments involving human tissues were approved by the Medical Ethics Committee of Central South University, and informed consent was obtained from the patients. A total of 910 patients diagnosed with NPC at the Second Xiangya Hospital over the past 5 years were screened, and 14 patients who had recurrences after radiotherapy confirmed by surgical pathology were identified. These patients met the following criteria: low-differentiated squamous cell carcinoma at the initial treatment, no history of other malignancies or systemic diseases, complete radical radiotherapy treatment in the Second Xiangya Hospital, complete clinical data, and paraffin-embedded tissue samples available. Only five of these cases had histologically intact paired histopathologic slices of both primary and recurrent NPC. IHC was performed to detect the level of RNAPII pS5 in these five pairs of nasopharynx cancer samples using the RNAPII phospho S5 antibody (1:10000, Abcam, ab5408). The IHC score was evaluated by two independent investigators blinded to the histopathologic features and clinical characteristics using the intensity and proportion of positively stained tumor cells as previously described (Jian et al., 2021).

Construction of plasmids and RNA interference

Plasmids expressing wild-type or mutant CycT1 or expressing shCycT1 were described previously (Fu et al., 2022). NEK7 cDNA was kindly gifted by Prof. Rong Tan (Tan et al., 2017). The cDNA was amplified by PCR and ligated into the pCDH-CMV-MCS-EF1-Puro vector with EcoRI and BamHI. The shRNA oligos were synthesized (Tsingke Biotechnology) and cloned into the pLKO.1-TRC vector. The shRNA plasmid and packaging vectors (pMD2.G and psPAX2) were transiently co-transfected into 293T cells using Poly(ethyleneimine) solution (Sigma, P3143). The target cells were treated with lentivirus supernatant and 8 μg/ml polybrene for 24 h, followed by 1 μg/ml puromycin (Selleck, s7417) selection for 3 days. On Day 5 after lentivirus infection, cell experiments were performed, and the cell lysates were prepared. All the sequences of shRNA and primers are listed in Supplementary Table S1.

Tumor cell viability analysis after irradiation

Cells were irradiated with the indicated X-ray dose at room temperature at 200 cGy/min with a linear accelerator X-Rad 225, (Precision X-Ray). For drug–radiation combination studies, cells were cultured in fresh media containing 0.15 μM THZ1 (Selleck, S7549) or corresponding DMSO control for 12 h prior to irradiation. The viability was measured 96 h post-IR treatment using MTT (Sigma, M5655).

For the xenograft tumor model, 5 × 10⁶ cells were suspended in 100 μl Dulbecco's phosphate-buffered saline (DPBS, Biological Industries) and injected subcutaneously into the flank of 6-week-old female BALB/c nude mice (Hunan SJA Laboratory Animal). The tumors were measured thrice weekly with a digital caliper, and the volume was calculated using the formula: 0.5 × (length × width²). For in vivo administration, THZ1 was dissolved in 5% dextrose water containing 10% DMSO or the vehicle and then injected intraperitoneally the day before irradiation. The mice were irradiated at 4 Gy under 2,2,2-tribromethanol anesthesia (Avertin). Only the tumor was irradiated, and the rest of the body was shielded by lead. All the animal experiments were approved by the Medical Ethics Committee of Central South University, and conducted according to the Guidelines of Animal Handling and Care in Medical Research in Hunan Province, China.

Western blotting and immunoprecipitation

For western blotting experiments, cells were harvested 48 h after irradiation treatment. Cells were washed twice with cold DPBS (Biological Industries) and then lysed in sample buffer (2% sodium dodecyl sulfate, 10% glycerol, and 62.5 mM Tris–HCl, pH 6.8) supplemented with 1× protease inhibitor cocktail (Sigma, P8340), sodium fluoride (10 mM, Sigma, 450022), and sodium orthovanadate (1 mM, Sigma, 450243). After sonication, the protein concentration was determined using a BCA assay (Beyotime, P0009). The following primary antibodies were used: rabbit anti-poly-ADP-ribose binding reagent (1:1000, Millipore, MABE1016), rabbit anti-CycT1 antibody (1:1000, Santa Cruz Biotech, sc-10750), mouse anti-RNAPII antibody (1:3000, Abcam, ab817), mouse anti-RNAPII phospho S5 antibody (1:3000, Abcam, ab5408), rabbit anti-RNAPII phospho S2 antibody (1:5000, Abcam, ab5095), and horseradish peroxidase-conjugated mouse anti-tubulin antibody (1:5000, Cell Signaling Technology, 12351S). The corresponding secondary antibodies (Thermo Fisher Scientific) were used at 1:5000. The signal was detected with ECL substrates (Millipore, WBKLS0500).

For immunoprecipitation, cells were collected 10 min after 10 Gy irradiation. Olaparib (MedChem Express, HY-10162) was administrated 1.5 h before irradiation. Whole-cell extracts were prepared and incubated overnight with anti-CycT1 antibodies or control rabbit IgG, and then incubated with protein A beads (Invitrogen) at 4°C. The immunoprecipitants were extensively washed before western blotting with the anti-poly-ADP-ribose binding reagent (Millipore, MABE1016) as previously described (Fu et al., 2022).

ChIP–seq and ChIP–qPCR

A previously described ChIP protocol was used with some modifications (Yu et al., 2022). Briefly, 48 h after receiving 4 Gy irradiation, the cells were cross-linked with 1% formaldehyde for 10 min, quenched with 125 mM glycine for 5 min, rinsed with ice-cold PBS twice, and scraped in 1 ml PBS. After centrifugation, the pellet was progressively resuspended in 500 μl Lysis Buffer I, II, and III. The cell lysate was sonicated until the DNA was sheared to 200 bp, quenched, and centrifuged. Then, 50 μl supernatant was reserved as input, and the rest was incubated overnight at 4°C with the magnetic beads bound with 4 μg anti-RNAPII (Abcam, ab817), anti-RNAPII phospho S5 (Abcam, ab5408), or anti-RNAPII phospho S2 (Abcam, ab5095) antibody. The beads were washed four times with Wash Buffer. The DNA was eluted with 100 μl Elution Buffer. The cross-links were reversed by incubating in a heating oscillator at 65°C for 2 h. The sample was then incubated with 4 μl of 25 mg/ml RNase (Magen) overnight at 65°C, and subsequently with 2 μl of 10 mg/ml proteinase K (Sigma, P6556) at 55°C for 4 h. The DNA was purified by AMPure XP Reagent (Beckman, A63881) and resuspended in 20 μl nuclease-free ddH₂O. ChIP–seq library was prepared using NEBNext® ChIP-Seq Library Prep Reagent Set (New England Biolabs, E7645S). The library was sequenced on Illumina NovaSeq 6000 (Novogene).

ChIP–seq raw reads were filtered using trim_galore v0.6.0 and aligned to the hg38 genome assembly using Bowtie2 v2.3.5.1 with default parameters. Duplicate reads were removed using MarkDuplicates from the gatk package v.4.1.4.1. The PI was calculated as previously described (Day et al., 2016). The RefSeq gene model was downloaded from the UCSC Genome Browser database (https://genome.ucsc.edu). ChIP–seq and input reads were calculated using bedtools coverage v2.26.0, mapped to the TSS region (–50 bp to + 300 bp relative to the TSS) and the gene body region (TSS + 300 bp to + 3 kb past the TES) for each annotated RefSeq isoform. The density of reads was normalized by the region length and by the mapped filtered read numbers multiplied by 1 million (rpm/bp). The input was then subtracted and PI was calculated as the ratio between RNAPII densities in the TSS and gene body regions. For multiple RefSeq isoforms of the same gene, the one with the strongest RNAPII ChIP–seq signal in the TSSR, at least 0.001 rpm/bp, was selected. All those genes with PI > 2 were defined as paused, and the rest were non-paused. The bigwig tracks were generated using bamCompare from deeptools. Negative values were set to zero. IGV v.2.4.13 was used to visualize the bigwig tracks. ChIP–seq profiles were created by computeMatrix and plotProfile in deeptools.

For ChIP–qPCR, primer sets targeting the TSS and gene body regions of TP53 and NEK7 were designed according to the ChIP–seq results, and qPCR was performed using the SYBR Green qPCR Master Mix (SolomonBio, QST-100) on the QuantStudio 3 Real-Time PCR system (Applied Biosystems). Primers are listed in Supplementary Table S1.

RNA-seq and RT–qPCR

RNA was extracted using TRIzol (Life Technologies, 87804). Libraries were prepared using the mRNA-Seq Sample Preparation Kit (Illumina) and sequenced on an Illumina NovaSeq platform (Novogene). Raw reads were filtered using trim_galore and then mapped to hg38 genome assembly using STAR v2.7.1a. Differential expression analysis was performed using the Bioconductor package DESeq2. For RT–qPCR, RNA was converted to cDNA using the PrimeScript RT Reagent Kit (TaKaRa, RR037A) and subjected to qPCR. All the primers are listed in Supplementary Table S1.

Flow cytometry assay

Cells were treated with 2 Gy or 4 Gy irradiation and collected 48 h post-IR treatment. The cells were washed once with DPBS, fixed with cold 70% ethanol for 24 h at −20°C, centrifuged at 500× g for 5 min to remove ethanol, and washed with DPBS twice. Then, the cells were treated with 100 unit/ml RNase and stained with 50 μg/ml propidium iodide (Sigma, P4170) at 37°C for 30 min in the dark. The samples were subjected to flow cytometric analysis by a flow cytometer (BD Biosciences). For time-course cell cycle analysis, cells were synchronized to G1/S phase by the double thymidine block procedure (Chen and Deng, 2018) and then released into fresh medium for the indicated time before collection.

Immunofluorescence and live-cell imaging

Cells were quickly rinsed with pre-warmed DPBS, fixed with cold methanol for 10 min at −20°C, permeabilized in 0.5% Triton X-100 for 10 min, blocked with 5% bovine serum albumin (BSA) for 30 min, and incubated overnight at 4°C with anti-RNAPII phospho S5 (1:1500, Abcam, ab5408), anti-RNAPII phospho S2 (1:3000, Abcam, ab5095), anti-γ-tubulin (1:1000, Sigma, T6557), anti-α-tubulin (1:1000, Abcam, ab18251), or anti-ACA human centromere antibody (1:3000, gifted by Prof. Xuebiao Yao at University of Science and Technology of China). Then, the cells were rinsed with DPBS thrice and incubated at room temperature for 1 h with Alexa Fluor 488, 546, or 647 secondary antibodies (1:400, Thermo Fisher Scientific). For frozen sections, mouse tumor tissue was immediately embedded in optimum cutting temperature media (SAKURA Tissue-Tek® O.C.T. Compound 4583) and frozen at −80°C. Cryostat sections (12 μm) were made and kept at −80°C. Slides were fixed in acetone for 5 min at room temperature, rinsed with DPBS twice, and blocked with 5% BSA for 30 min. The slides were then incubated overnight at 4°C with anti-RNAPII phospho S5 (1:1000), rinsed thrice, and incubated for 1 h at room temperature with Alexa Fluor 488 secondary antibody (1:400). DNA was visualized by DAPI staining (0.5 μg/ml, Sigma, D9542). The samples were mounted with SlowFade™ Diamond Antifade Mountant (Thermo Fisher Scientific, S36963).

For live imaging, cells in glass-bottom dish (NEST, 801001) were transfected with pEGFP-tubulin and mCherry-H2B using Lipofectamine 3000 (Invitrogen) and synchronized with the double thymidine block procedure. Cells were then released into thymidine-free medium for 8 h and imaged in PeCon environmental microscope incubator (ZEISS) at 37°C and 5% CO₂. Images were collected on the LSM 880 confocal microscope using a 63× oil immersion objective lens with Airyscan mode (ZEISS).

Statistical analyses

The experiments were carried out at least three times. Data were presented as mean ± standard deviation (SD). Statistical analysis and survival fraction analysis were performed using GraphPad Prism 9. Flow cytometry data were analyzed using ModFit LT 4.1. The statistical details of each experiment were indicated in the respective figure legends. Student's t-test or chi-squared test was performed to evaluate significant differences between two groups. Kolmogorov–Smirnov test was used to evaluate the difference in the PI between two groups. P-values were presented as star marks in figures: *P < 0.05, **P < 0.01, ***P < 0.001.

Data availability

ChIP–seq and RNA-seq datasets were deposited to the NCBI GEO database under the accession number GSE185423. All custom scripts are available from the authors upon request.

Funding

This project has been supported by grants from the National Natural Science Foundation of China (32170821 and 92153301 to K.Y. and 32101034 to F.C.), the Ministry of Science and Technology of the People's Republic of China (2021YFC2701202), Department of Science & Technology of Hunan Province (2021JJ10054 and 2019SK1012 to K.Y., 2021JJ41049 to C.Y., and the Innovative Team Program 2019RS1010), and Central South University (the Innovation-driven Team Project 2020CX016). K.Y. is supported by the National Thousand Talents Program for Young Outstanding Scientists.

Conflict of interest: none declared.

Author contributions: conceptualization: K.Y. and N.Z.; methodology: H. Liu, H.F., C.Y., F.C., L.L., K.Y., and N.Z.; validation: N.Z., C. Huang, F.C., and L.L.; software: H. Liu and C. Huang; formal analysis: H. Liu, C.Y., and K.Y.; investigation: H. Liu, H.F., C.Y., N.Z., and C. Huang; resources: C. Hu, Z.X., Z.Z., H. Lu, K.Y., and N.Z.; data curation: H. Liu, C.Y., and K.Y.; writing original draft: H. Liu and K.Y.; writing review and editing: K.Y.; visualization: H. Liu, C.Y., N.Z., C. Huang, and K.Y.; supervision: K.Y.; project administration: K.Y. and L.L.; funding acquisition: K.Y.