Ribosomal protein L23 negatively regulates cellular apoptosis via the RPL23/Miz-1/c-Myc circuit in higher-risk myelodysplastic syndrome

Ribosomal protein (RP) L23 is a negative regulator of cellular apoptosis, and RPL23 overexpression is associated with abnormal apoptotic resistance in CD34+ cells derived from patients with higher-risk myelodysplastic syndrome (MDS). However, the mechanism underlying RPL23-induced apoptotic resistance in higher-risk MDS patients is poorly understood. In this study, we showed that reduced RPL23 expression led to suppressed cellular viability, increased apoptosis and G1-S cell cycle arrest. Gene microarray analysis comparing RPL23-knockdown and control cells identified an array of differentially expressed genes, of which, Miz-1, was upregulated with transactivation of the cell cycle inhibitors p15Ink4b and p21Cip1, and Miz-1’s functional repressor, c-Myc, was downregulated. Cells derived from higher-risk MDS patients demonstrated consistently increased expression of RPL23 and c-Myc and decreased Miz-1 expression compared with cells from lower-risk patients. In conclusion, Miz-1-dependent induction of p15Ink4b and p21Cip1 was depressed with decreased Miz-1 and increased c-Myc expression under conditions of elevated RPL23 expression, leading to apoptotic resistance in higher-risk MDS patients. Because RPL23 is encoded by a target gene of c-Myc, the RPL23/Miz-1/c-Myc regulatory circuit provides a feedback loop that links efficient RPL23 expression with c-Myc’s function to suppress Miz-1-induced Cdk inhibitors and thereby leads to apoptotic resistance in higher-risk MDS patients.

other signalling pathways was observed in tMDS patients, and among these, RPL23 was overexpressed in tMDS patients compared with sMDS patients. Moreover, analyses of AML evolution and of the overexpressed genes in tMDS versus sMDS revealed that RPL23 overexpression was potentially involved in disease progression. A report from Li demonstrated significantly elevated RPL23 mRNA expression levels in higher-risk MDS patients, and these levels were inversely correlated with the percentage of apoptotic CD34+ cells, leading to disease progression and poor survival 14 .
This study further detected the RPL23 expression levels in MDS patients with different levels of risk and provided evidence supporting the anti-apoptotic effects of RPL23 in the BM cells of higher-risk MDS patients.

Results
RPL23 knockdown suppresses the viability of SKM-1 and K562 cells. We detected comparatively higher RPL23 mRNA expression in the SKM-1 [an acute myeloid leukaemia cell line established in the leukaemic phase during the progression from MDS to AML (MDS/AML)] and K562 [human chronic myelogenous leukaemia (CML) cell line] cell lines than in other leukaemic cell lines, e.g., HEL (human erythroleukaemia cell line), HL60 (human acute promyelocytic leukaemia cell line) and NB4 (human acute promyelocytic leukaemia cell line) (Supplementary Fig. S1a). Therefore, we selected the SKM-1 and K562 cell lines for subsequent biological function studies. To investigate the role of RPL23 in apoptotic-resistant, higher-risk MDS patients, the LV-RPL23-RNAi vector was designed and transfected into SKM-1 and K562 cells. The knockdown efficiency was verified by qRT-PCR (p < 0.001 for both cell lines; Supplementary Fig. S1b) and western blotting (p < 0.001 for SKM-1 and p = 0.005 for K562; Supplementary Fig. S1c and S1d) through comparison between the KD (vectors carrying LV-RPL23-RNAi-GFP) and NC (LV-RPL23-NC-GFP) groups. We measured the effects of RPL23-KD on cellular viability by performing CCK-8 assays ( Fig. 1a and 1b). Twenty-four hours after efficient transfection, no changes were detected between the KD and NC groups (p > 0.05). However, the viability of SKM-1 cells decreased from 0.89 ± 0.02 to 0.69
Global gene expression profiles. To obtain insights into the molecular changes in response to RPL23 knockdown, the global changes in gene expression were examined by performing a DNA microarray analysis of RPL23-KD and RPL23-NC SKM-1 cells. As described in the Methods, a supervised unpaired T-test analysis using the Limma package produced a volcano plot depicting alterations in gene expression (Fig. 2a). Specifically, the results revealed 753 differentially expressed genes (DEGs), including 305 upregulated genes and 448 downregulated genes in the KD groups. The data from the gene expression arrays have been deposited in the NCBI Gene Expression Omnibus under GEO Accession Number GSE95348. A hierarchical clustering (HC) of the samples based on the transcript expression levels clearly segregated the transcripts into two distinct groups consistent with their treatments (Fig. 2b). Among the 753 DEGs, Miz-1 (FC = 6.53, p = 2.24E-06), CDKN1A (p21 Cip1 ; FC = 2.98, p = 1.60E-05), and CDKN2B (p15 Ink4b ; FC = 2.57, p = 6.46E-05) were upregulated, whereas c-Myc (FC = −9.81, p = 2.62E-05) was downregulated. Additionally, PTEN (FC = 3.62, p = 3.87E-06) and PIK3CG (FC = −13.07, p = 8.62E-05) were notably differentially expressed in the RPL23-KD groups.
Gene set enrichment analysis (GSEA). We then explored whether the DEGs mapped to coherent functional gene sets and pathways. The array data were interpreted by performing a GSEA analysis through annotation with the C2: Kyoto Encyclopedia of Genes and Genomics (KEGG) pathways and C5: Gene Ontology (GO) gene sets. The DEGs were assigned to multiple C5: GO gene sets categorized by biological process, cellular component and molecular function [15][16][17] . The major biological processes associated with these DEGs were apoptotic processes, cell cycle and programmed cell death, consistent with the growth inhibition observed during our functional experiments. The following cellular component gene sets were enriched in RPL23-KD cells: nucleolus, Scientific RepoRts | 7: 2323 | DOI:10.1038/s41598-017-02403-x intracellular organelles and nucleus, covering genes associated with ribosomal biogenesis within the nucleus as well as nuclear stress 5 . The enriched molecular function gene sets mainly involved transcriptional regulation. Detailed information is provided in Fig. 2c. Pathway annotation based on the KEGG database indicated that pathways in cancer (NES = 2.09, nominal p < 0.001, FDR q = 0.038) and apoptosis (NES = 1.65, nominal p = 0.033, FDR q = 0.199) were highly enriched, and the leading-edge subsets included Miz-1, CDKN1A (p21 Cip1 ), c-Myc, PIK3CG, PTEN and other apoptosis-related genes (e.g., BCL2, FAS, and CDK6; Fig. 2d and 2e). Validation of screened mRNAs identified from the microarray analysis through qRT-PCR and western blotting of SKM-1 cells. We performed qRT-PCR and western blotting analyses of RPL23-KD/ NC SKM-1 cells to verify the expression profiles obtained in the microarray analysis and to further investigate changes in the related protein levels. The relative mRNA expression levels are shown in Fig. 3a: RPL23 (0.02 ± 0.01, p < 0.001), Miz-1 (2.50 ± 0.14, p = 0.009), c-Myc (0.58 ± 0.08, p = 0.031), p15 Ink4b (2.44 ± 0.23, p = 0.003), p21 Cip1 (6.46 ± 0.37, p = 0.005), and PI3KCG (0.15 ± 0.05, p = 0.004). The mRNA expression levels of the RPL23-KD/ NC groups were normalized to those of the WT group. p15 Ink4b and p21 Cip1 were transcriptionally activated following RPL23 knockdown, and this effect was accompanied by increased Miz-1 and decreased c-Myc expression. Our western blotting results also revealed altered protein levels, specifically increased expression of Miz-1 (p < 0.001) and its downstream target molecules, p21 Cip1 (p < 0.001) and p15 Ink4b (which exhibited a mild trend: p = 0.09) and decreased expression of c-Myc (p < 0.001) and PIK3CG (p < 0.001), as shown in Fig. 3b and 3c. The phosphorylation of AKT, a downstream effector of PI3K/AKT signalling, was also weakened (p = 0.008), indicating suppression of PI3K/AKT signalling activity.

Expression of c-Myc overcomes RPL23-KD-induced viability suppression and apoptotic tendency.
To determine whether c-Myc plays an important role in RPL23-knockdown-mediated biological The cellular viability was measured twenty-four hours after coinfection using CCK-8 (Fig. 4a). The growth suppressive effect of RPL23-KD was overcome by CMV-driven expression of c-Myc (p < 0.01). Note that c-Myc overexpression alone resulted in a higher proliferative ratio than that of the RPL23-KD-c-Myc-OE group (p < 0.05). Forty-eight hours after superinfection, cellular apoptosis was measured by Annexin V single-staining of specific cells ( Fig. 4b and 4c), and induced expression of c-Myc alleviated the high apoptotic potential mediated by depletion of L23 (from 36.33 ± 4.55% to 14.07 ± 1.50%, p = 0.002). Furthermore, the induced expression of c-Myc alone in SKM-1 cells displayed a lower apoptotic ratio (5.42 ± 1.68%, p = 0.003) in the absence of LV-RPL23-RNAi. Detailed information of the gating protocol used for the flow cytometry data is described in Supplementary Fig. S3.   Supplementary Information (Supplementary Fig. S5). Each assay was performed in triplicate. Unpaired Student's t tests were used to calculate all p values shown throughout the figure. The data are expressed as the means ± S.E.M. WT: wild type; NC: RPL23-NC; KD: RPL23-KD. * p < 0.05; ** p < 0.01; *** p < 0.001. different levels of risk were performed (the patients' information is listed in Supplementary Table S1). The RPL23, Miz-1 and c-Myc mRNA expression levels, which were calculated using the 2 −△△Ct method (mean ± S.E.M), were evaluated in 97 MDS samples (54 lower-risk and 43 higher-risk MDS patients) and compared with the levels in corresponding samples from 21 normal donors (Fig. 5a-5c). The RPL23 expression levels in the lower-and higher-risk MDS patients were 1.05 ± 0.07 and 1.88 ± 0.17, respectively (p < 0.001), and the c-Myc expression levels were 1.03 ± 0.07 and 1.86 ± 0.09 in the lower-and higher-risk MDS patients, respectively (p < 0.001). Miz-1 demonstrated higher expression in lower-risk MDS patients compared with higher-risk patients (1.88 ± 0.20 vs. 0.69 ± 0.06, p < 0.001) and normal controls (p = 0.001). We also observed lower Miz-1 expression in higher-risk patients than in normal donors (p = 0.035). A Pearson's correlation analysis was performed to analyse the RPL23 and c-Myc mRNA levels (Fig. 5d), and RPL23 expression was found to be positively correlated with c-Myc expression (r = 0.703, p < 0.001). Additionally, a relatively weak negative correlation was found between the RPL23 and Miz-1 mRNA levels (r = −0.25, p = 0.01).
An IHC analysis of target protein expression was performed by determining the mean density of positive staining. The mean densities for RPL23 in BM haematopoietic cells from lower-and higher-risk MDS patients were 0.14 ± 0.01 and 0.21 ± 0.01 (p < 0.001), whereas the normal donors exhibited no difference compared with lower-risk patients (0.13 ± 0.01, p = 0.44). The mean densities for c-Myc in the cells from lower-and higher-risk MDS patients were 0.13 ± 0.00 vs. 0.21 ± 0.01, respectively (p < 0.001), whereas that obtained for the cells from the normal donors was 0.12 ± 0.00. The expression intensity of Miz-1 was 0.37 ± 0.02 in lower-risk MDS patients, whereas higher-risk MDS patients exhibited decreased Miz-1 expression (0.23 ± 0.01, p < 0.001), and this decreased expression was even lower than that observed in normal donors (0.29 ± 0.01, p < 0.001; Fig. 5e-5h).

Discussion
RPL23 is a protein component of the 60S large ribosomal subunit and is believed to perform multiple auxiliary extraribosomal functions; additionally, RPL23 negatively regulates apoptosis by suppressing the Miz-1-induced transcription of the cell cycle inhibitors p15 Ink4b and p21 Cip1 . As reported before, RPL23 is overexpressed at the mRNA level in higher-risk MDS patients, and elevated RPL23 expression is inversely associated with the apoptosis ratio in CD34+ BM cells, which might lead to disease progression and adverse prognosis. However, the mechanism underlying this effect remains uncharacterized.
In this study, knockdown of RPL23 was found to suppress SKM-1/K562 cellular viability by strongly inducing apoptotic cell death and G1/S phase arrest. To further investigate the molecular mechanisms involved in these phenomena, we performed a gene microarray analysis of RPL23-KD and RPL23-NC samples and detected elevated expression of Miz-1 and its target molecules p15 Ink4b and p21 Cip1 as well as decreased expression of c-Myc. To further elucidate the potential regulatory mechanisms involving c-Myc/Miz-1, a rescue experiment using SKM-1 cells co-infected with LV-RPL23-RNAi and CMV-c-Myc-expressing plasmid vectors was performed, and the results demonstrated that the induction of c-Myc expression could overcome the higher tendency to apoptosis mediated by RPL23-KD, supporting the conclusion that c-Myc is an important downstream regulator of RPL23-mediated abnormalities in apoptosis. The analysis of the patients' data confirmed this deregulation and correlated the RPL23/Miz-1/c-Myc expression levels with the risk levels of MDS patients.
Compared with patients presenting normal RPL23 expression levels, higher-risk MDS patients with increased RPL23 expression demonstrated decreased Miz-1 expression, which resulted in suppression of Miz-1-induced p15 Ink4b and p21 Cip1 expression in their BM haematopoietic cells. p15 Ink4b and p21 Cip1 downregulation promotes cell cycle progression in the BM haematopoietic cells of MDS and myeloid disease patients and transforms their tendency to undergo apoptosis into apoptosis resistance [18][19][20] . Consistently, lower-risk MDS patients exhibited elevated Miz-1 expression compared with normal donors, which potentially accounted for the BM haematopoietic cells' tendency toward apoptosis observed in lower-risk MDS patients. Additionally, elevated c-Myc expression was also observed in higher-risk MDS patients, and in vitro experiments and microarray analyses indicated that RPL23 knockdown was potentially associated with decreased c-Myc expression as well as suppressed PI3K/ RPL23, which is encoded by a gene that is upregulated by c-Myc, has an essential function in restricting Miz-1-dependent cell-cycle arrest 24 . Our study suggests a regulatory feedback loop in which RPL23 could reinforce c-Myc-dependent oncogenic functions. Elevated expression of RPL23 could decrease the Miz-1-dependent expression of p21 Cip1 and p15 Ink4b , thereby increasing the ability of c-Myc to promote cell-cycle progression. In turn, the activation of c-Myc function in the context of RPL23 overexpression might further stimulate RPL23 expression. Thus, the regulation of Miz-1 activity by RPL23 represents a positive feedback mechanism that couples efficient RPL23 expression with the function of c-Myc to suppress Miz-1-induced Cdk inhibitors and thereby promote cell-cycle progression (Fig. 6). In higher-risk MDS, RPL23 overexpression might confer growth advantages and resistance to apoptosis through this positive feedback loop, potentially leading to AML evolution.
According to our functional experiments, the knockdown of RPL23 expression in SKM-1 and K562 cells changed the cellular proliferation rate, the apoptotic ratio and the cell cycle distribution. Considering the known roles of Miz-1 and c-Myc in cellular senescence, we may wonder whether the RPL23-KD cells exhibit altered cellular senescence. The β-galactosidase staining of RPL23-NC/KD and wild-type SKM-1/K562 cells revealed that the percentages of SA-gal-positive cells (senescent signals) did not show any significant differences among the three groups in both cell lines (Supplementary Fig. S4). The following explanations might account for the undetected senescent signals observed in our study. (1) The SKM-1 and K562 cell lines, which were subjected to RPL23 knockdown and senescence analysis, are both haematological cancer cell lines. Cancer cells show the characteristics of immortalization and absence of senescence; thus, these two cell lines, including the RPL23-knockdown and blank/negative control groups, barely showed any senescent signals. (2) MDS is a clonal haematopoietic stem cell disorder that is characterized by bone marrow cell dysplasia and ineffective haematopoiesis 1 . Cellular senescence is a feature of this disease (e.g., observed in BM mesenchymal stromal cells (MSCs) of MDS patients) but is not the dominant feature of haematopoietic cells, particularly in patients whose disease has evolved to AML or in patients with chromosomal instability (e.g., patients with an isolated monosomy 7 showed significantly longer telomeres and weak senescent signals 25,26 ). It remains to be investigated whether cellular senescence might be involved in the potential mechanisms resulting in RPL23 overexpression MDS patients or is a subordinate aspect. Interestingly, RPL23 has been reported to functionally inhibit the HDM2 ubiquitin ligase and thereby activate p53, leading to growth inhibition and anti-tumour effects in cases of gastric cancer [27][28][29] . In our RPL23-KD SKM-1 cells, however, decreased RPL23 expression was correlated with the induction of apoptosis, and our western blotting results did not reveal any alternations in p53 expression. Furthermore, the downstream target gene of p53, HDM2, was not transcriptionally activated, as demonstrated by our qRT-PCR analysis ( Supplementary  Fig. S1e second panel and S1f). To the best of our knowledge, the SKM-1 cell line, which was established in the leukaemic phase during the progression from MDS to AML, is karyotypically abnormal (including a 17p deletion) and is characteristic of the acquisition of p53 mutations (missense and silent point mutations) 30,31 . Mutation of the p53 gene is associated with complex karyotypes, reduces overall survival and plays a role in the evolution of MDS to AML [30][31][32] . Therefore, the loss of normal p53 gene function, i.e., the allelic inactivation of 17p and the point mutation of the other allelic p53 gene in the SKM-1 cell line, might give rise to undetected p53 alterations and RPL23-HDM2-p53 signal activation, as previously reported. Regarding the clinical cases of elevated RPL23 expression in higher-risk MDS patients, the two opposing roles of RPL23 in cellular apoptosis might be the result of heterogeneity in individual cells and disease stages in different MDS patients, an effect that is complicated by different karyotype abnormalities and p53 mutation scenarios. Furthermore, the RPL23 expression levels of our patients presented significant variation among higher-risk MDS patients, and some patients with a higher-risk phenotype exhibited normal or lower RPL23 expression levels. However, the overall expression trend was higher among higher-risk MDS patients than among lower-risk MDS patients and normal donors. As reported previously, DBA and 5q-MDS are associated with inherited (DBA) or acquired (5q-MDS) ribosomal protein haploinsufficiency through ribosomal protein gene mutations or karyotype abnormalities (e.g., chromosome loss) 33,34 . Therefore, the heterogeneity of RPL23 expression levels in MDS patients might result from the haploinsufficiency of RPL23 protein production in patients with different related chromosomal abnormalities and complex gene mutations.
In summary, RPL23 expression levels have been identified as an independent predictor of prognosis of MDS patients, regardless of patients' age, IPSS score, or haemoglobin levels 14 . Based on these findings, RPL23 dysregulation involves a novel molecular mechanism in MDS pathogenesis and represents a potential prognostic biomarker and novel therapeutic target in patients with MDS. Cellular functional assays. The cellular viability was measured by performing a CCK-8 assay (Dojindo, kamimashiki gun Kumamoto, Japan) following the manufacturer's instructions. Briefly, SKM-1 and K562 cells infected with LV-RPL23-RNAi or LV-NC, as well as untreated cells, were seeded into 96-well plates (4,000 cells/ well) in a volume of 100 μl. Each experiment was performed with four replicates. After 24, 48, 72, and 96 h, 10 μl of CCK-8 solution was added to each well, and the plate was incubated at 37 °C for 2-4 h. The absorbance at 450 nm was measured with a microplate reader (Biotech, NY, USA). The percentage of viable cells was calculated using BM trephine biopsy and immunohistochemistry. BM trephine biopsy (BMTB) was performed by experienced pathologists, and BM trephines were fixed in 4% paraformaldehyde, decalcified with 5% EDTA and embedded in paraffin blocks. Briefly, 4-μm-thick paraffin-embedded specimens were deparaffinized, hydrated, heated for antigen retrieval and incubated in 3% H 2 O 2 to block endogenous peroxidase activity. The slides were treated with goat serum to prevent non-specific binding. The following primary antibodies were used: a mouse mAb against c-Myc (5605, CST, 1:800; cytoplasmic staining), a rabbit pAb against Miz-1 (sc-22837, Santa Cruz, 1:50; cytoplasmic staining) and a rabbit pAb against RPL23 (16086-1-AP, ProteinTech, 1:50; cytoplasmic staining). A mouse/rabbit immunohistochemistry (Enhanced Polymer) kit (PV-9000-D; ZSGB-BIO, China) was used to detect the primary antibodies. The sections were stained with a DAB kit. Counterstaining was performed with haematoxylin. Images of the slides stained for the target proteins were scanned at 40× magnification using an optical microscope (Olympus Co., Tokyo, Japan). We evaluated the staining of representative fields using Image-Pro Plus v6.0 software (Media Cybernetics, USA). After background removal via optical density calibration, the segmentation option was used to capture the areas of interest (AOIs) in each image with the HSI colour model parameter settings (H:0-30, S:0-255, and I:0-230), which covered the majority of the DAB-positive areas. The integrated optical densities (IODs) and positively stained regions in each image were measured and integrated as the IOD (sum) and the area (sum). The intensity of target protein expression was measured as the density mean = IOD (sum)/area (sum) (optical density/mm 2 of AOI).