Profile of snoRNAs in CRC tissues and adjacent normal tissues
To ascertain whether snoRNAs were aberrantly expressed in CRC, five pairs of CRC tissues and para-tumor tissues were subjected to snoRNA sequencing using an Illumina NovaSeq PE150 platform. Clean reads were aligned to human snoRNA sequences collected from the snoDB database 15. The Pearson correlation matrix plot indicated that snoRNA expression was highly dissimilar between tumor and normal colon tissues (Fig. S1A). A total of 423 snoRNAs were identified (Fig. 1A), among of which 36 snoRNAs were upregulated and 36 were downregulated (|fold-change|>1.5, p < 0.05) in CRC tumor tissues compared with adjacent normal tissues (Figs. 1B and S1B). These snoRNAs ranged in length from a few tens of nt to a few hundreds (Fig. S1C-D). Mapping all differentially expressed snoRNAs (DESs) to their chromosomal locations revealed that they were almost evenly distributed on most chromosomes (Fig. 1C). Of note, 62.50% (45/72) of these DESs were SNORDs and 34.72% (25/72) were SNORAs (Fig. 1D). As there is no functional annotation database for eukaryotic snoRNAs, in order to initially assess the functions of these snoRNAs, we performed GO (gene ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment analysis of their host genes. GO and KEGG analysis showed that the host genes of these snoRNAs were mainly involved in translation in the cytoplasm and were mostly important components of the ribosome (Fig. 1E-F).
SNORD11B is significantly upregulated in CRC with poor prognosis
SNORD11B (snoRNA ID: URS000075ED7D) has a predicted 18S rRNA Nm modification site and was ranked second among DESs with a differential expression multiple of 3.512 (p = 0.002) in the sequencing results. Its host gene is NOP58 (nucleolar protein 58), which encodes a key protein involved in Nm modification. Therefore, we chose SNORD11B for further study. The 112-nt SNORD11B is located on chromosome 2, region 202291317–202291428, and is derived from intron 8 of its host gene NOP58 (Fig. 2A). SNORD11B is a classical SNORD with a conserved D box (CUGA) and C box (AUGAUGA). First, we examined the expression level of SNORD11B in 76 pairs of CRC tumor tissues and paired para-normal colon tissues and found that SNORD11B was significantly upregulated in CRC tumor tissues (Fig. 2B), consistent with the snoRNA-seq results. As SNORD11B is derived from the intron of NOP58, its high expression in CRC may be correlated with the level of NOP58. As previously reported, NOP58 is upregulated in CRC 16. In our cohort, NOP58 was also significantly highly expressed in the 76 pairs of CRC tissues (Fig. 2C), and we found a significant positive correlation between NOP58 and SNORD11B expression in these tissues (Pearson’s r = 0.2747, p = 0.0163) (Fig. 2D). We also examined the expression levels of SNORD11B in five CRC cell lines and normal human intestinal epithelial cells (HIEC) and found that the expression levels of SNORD11B were markedly higher in the CRC cell lines than in HIEC. SNORD11B expression was particularly high in HCT116 and HCT8 (Fig. 2I), so these two cell lines were selected for subsequent experiments.
SNORD11B is a potential diagnostic and prognostic biomarker in CRC
The receiver operating characteristic (ROC) curve and the area under curve (AUC) values showed that SNORD11B levels in tissues had diagnostic efficacy for CRC (AUC = 0.6784, p = 0.003) (Fig. 2E). And we collected plasma from 39 normal subjects (age 57.67 ± 14.74 years, without underlying gastrointestinal diseases) and 57 CRC patients (age 64.74 ± 11.29 years) and found that SNORD11B levels in plasma of CRC patients were notably higher than those in normal healthy subjects (Fig. 2F), and SNORD11B levels in plasma had better efficacy for CRC diagnosis compared with that in tissue (AUC = 0.8862, p < 0.001) (Fig. 2G). We also examined CEA and CA199 levels in the plasma of these participants and demonstrated that the diagnostic efficacy of SNORD11B was slightly better than that of CEA and obviously better than that of CA199 (Fig. S2A-C). These results suggest that SNORD11B could be used as a biomarker for CRC diagnosis.
To further analyze the correlations between SNORD11B expression and clinical characteristics of CRC patients, we downloaded SNORD11B expression data from The Cancer Genome Atlas colon adenocarcinoma collection (TCGA_COAD) from the SNORic database, as well as clinical characteristics and survival/prognosis information for TCGA_COAD patients from the UCSC Xena website. We integrated all the data and used chi-square tests or Fisher's exact tests to show that there were significant correlations of SNORD11B expression with colon cancer histology (p = 0.035), lymph node metastasis (p = 0.027), and TNM stage (p = 0.003) (Tables S1-2). Unfortunately, multivariate analysis by Cox proportional hazards regression model of clinical variables and overall survival showed that SNORD11B level was not an independent prognostic risk factor (Fig. 2H). Nevertheless, the Kaplan–Meier survival curve and log-rank test analyses revealed that elevated SNORD11B expression was significantly associated with shortened overall survival (p = 0.0234) (Fig. 2I). We subsequently analyzed the patients included in the cohort of this study; owing to a lack of relevant data on patient prognosis, the correlation between SNORD11B and each clinical feature was assessed by chi-square test or Fisher's exact test. As shown in Tables S3-6, there were significant correlations between tissue SNORD11B levels and lymphatic invasion (p < 0.001) and TNM stage (p = 0.017), as well as plasma SNORD11B (lymphatic invasion, p = 0.003; TNM stage, p < 0.001). These results suggest that SNORD11B level could reflect poor prognosis of CRC patients to some extent and could thus serve as a potential prognostic marker.
SNORD11B promotes CRC malignant progression in vitro and in vivo
Aberrantly high expression of SNORD11B may have a synergistic role in the development of CRC. Therefore, to explore the biological function of this snoRNA, we designed antisense oligonucleotides (ASOs) to interfere with its expression in CRC cell lines. The knockdown efficiencies of two ASOs targeting SNORD11B were validated by real-time quantitative PCR (RT-qPCR) after transfection of HCT116 and HCT8 cells (Fig. 3A,B). As expected, Cell Counting Kit-8 (CCK8) and 5-ethynyl-2′-deoxyuridine (EdU) assays showed that the proliferation of CRC cells was significantly inhibited after SNORD11B silencing (Fig. 3C-E), and knockdown of SNORD11B abrogated the colony formation ability of CRC cells (Fig. 3F). Transwell assays showed that the invasive behavior of CRC cells was strongly reduced after SNORD11B knockdown (Fig. 3G). In addition, flow cytometry results indicated that SNORD11B silencing significantly increased early and late apoptosis levels of CRC cells (Fig. 3H-I). Conversely, the opposite results were obtained after four-fold overexpression of SNORD11B in normal colonic epithelial cells (HIEC) (Fig. S3). To further validate the pro-cancer role of SNORD11B in CRC, a xenograft tumor assay was performed. We designed three single guide RNAs (sgRNAs) targeting SNORD11B and cloned them into a LentiCRISPRv2 vector, selecting the one with the best efficiency (knockdown efficiency of about 80%) (Fig. S4A). sgSNORD11B did not affect the expression of NOP58 (Fig. S4B), so it was prioritized for further analysis. HCT116 control cells and cells with stable SNORD11B knockdown were injected subcutaneously into the left and right sides, respectively, of the rear of the axilla of 5-week-old nude mice. The weights and volumes of subcutaneous tumors in the SNORD11B-knockdown group were dramatically reduced (Fig. S4C-E) compared with controls. Immunohistochemistry analysis showed that proportions of Ki67-positive cells were significantly decreased in the SNORD11B-knockdown group (Fig. S4F). These results confirm that SNORD11B has an oncogenic role in CRC in vivo and in vitro.
SNORD11B mediates Nm modification at the G509 site of 18S rRNA
According to previous studies, 38 Nm modification sites have been reported in human 18S rRNA 17,18, and Nm modification at these sites involved the participation of SNORDs. By searching the snoRNA-related databases snOPY 19 and snoRNA Atlas 20, we found that SNORD11B had been predicted to mediate Nm modification at the G509 site of 18S rRNA, and that SNORD11B could complementarily base-pair with sequences 5 nt upstream and 8 nt downstream of the G509 site on 18S rRNA, resulting in a stable interaction (Fig. 4A). Moreover, the G509 site is located exactly at the 5-nt position upstream of the D box of SNORD11B, consistent with the classical rules of SNORD-mediated Nm modification of target rRNAs 21. However, no study has yet confirmed that SNORD11B can mediate Nm modification at the G509 site of 18S rRNA; therefore, an RTL-PCR (Reverse transcription at low dNTP concentrations followed by PCR) assay was performed as previously described 22. This method provides a semi-quantitative assessment of Nm modifications between two RNA samples by comparing the PCR efficiency after RT reactions with high or low concentrations of dNTP. RT is inhibited at the Nm site under low dNTP conditions but not under standard high dNTP conditions, and the efficiency of the RT reaction can be quantified by PCR (Fig. 4B). We found that the level of the product (174 bp) of PCR with F1/R as primer was significantly increased after SNORD11B interference when the RT products obtained at a low concentration of dNTP (1 µM) were used as a template, whereas there was no difference when the RT products obtained at a high concentration of dNTP (40 µM) were used as the template (Fig. 4D), indicating that SNORD11B silencing led to increased RTL-PCR efficiency for 18S rRNA. Consistent results were obtained in cells with stable knockdown of SNORD11B via CRISPR–Cas9 (Fig. 4F), and the opposite results were observed in HIEC with SNORD11B overexpression (Fig. 4H). Previous studies have reported that Nm modification plays an integral part in rRNA precursor processing and maturation 17,23, so we designed specific primers for total 18S rRNA and its precursor (Fig. 4C). We found an elevation of immature 18S rRNA levels in CRC cells after SNORD11B interference or knockout (Fig. 4E,G) and a reduction after exogenous upregulation of SNORD11B in HIEC (Fig. 4I). We also found that levels of immature 18S rRNA were reduced in most of the CRC tumor tissues with upregulated expression of SNORD11B, of a total of 12 pairs of tissues (Fig. S5A). These results confirm that SNORD11B promotes maturation of 18S rRNA processing, and that this effect is most likely to be mediated by promotion of Nm modification at the G509 site of 18S rRNA.
Given that 18S rRNA is the core component of the small subunit of eukaryotic ribosomes and its maturation affects ribosome assembly and subsequent protein translation, we first analyzed the relative position of 18S rRNA Gm509 in the global mature ribosome structure using PyMOL software. As shown in Fig. S5B, Gm509 of 18S rRNA is far from the ribosome decoding center, which limits its effect on overall translation. To further assess the effect of SNORD11B-mediated 18S rRNA Nm modification on translation, we performed a puromycylation assay. We found that the global translation level of cells was not changed after intervention in SNORD11B expression (Fig. 4J-L). These results imply that the effect of SNORD11B on translation is limited.
SNORD11B suppresses the expression of let-7-family miRNAs
To find targets of SNORD11B other than 18S rRNA, we conducted in silico prediction of SNORD11B interaction factors using the RNAinter database 24. The results revealed weak binding interactions between SNORD11B and let-7-family miRNAs in several datasets. The expression of most let-7-family miRNAs was significantly elevated after SNORD11B knockdown (Figs. 5A-B and S6A-B), whereas in HIEC cells with SNORD11B overexpression, let-7-family miRNAs were all significantly reduced in expression (Fig. S6D). Western blots showed that the expression levels of most downstream target genes of let-7-family miRNAs were significantly reduced after SNORD11B knockdown (Figs. 5C and S6G), in contrast to the results for SNORD11B-overexpressing HIEC cells (Fig. S6H).
We then selected let-7a-5p, which was regulated by SNORD11B under different treatments with relatively high fold change, as a representative of let-7-family miRNAs for subsequent experiments. Let-7a-5p was markedly downregulated in 61 pairs of CRC tumor tissues (Fig. 5D). There was also a significant negative correlation between let-7a-5p and SNORD11B expression in these tissues (Pearson’s r = − 0.4058, p = 0.0012) (Fig. 5E). If SNORD11B interacted directly with let-7a-5p, it would be expected to be localized in the cytoplasm; however, RNA fluorescence in situ hybridization (FISH) assays demonstrated that SNORD11B was located in the nucleus (Figs. 5G and S6F). We separated the cytoplasmic and nuclear fractions of HCT116, HCT8, and HIEC cells. Using U6 snRNA as the nuclear control transcript and GAPDH as the cytoplasmic control, we confirmed that SNORD11B was mainly distributed in the nucleus, whereas let-7a-5p was predominantly expressed in the cytoplasm (Figs. 5F and S5E). This demonstrated that SNORD11B did not act directly on let-7a-5p, but that it could act on the primary transcript of let-7a-5p (pri-let-7a) in the nucleus, leading to decreased levels of let-7a-5p and accelerating the malignant progression of CRC.
Nucleotides 84–94 of SNORD11B could complementarily pair with pri-let-7a (Fig. 5H). The expression of pri-let-7a was significantly elevated after SNORD11B knockdown and reduced after SNORD11B overexpression (Figs. 5I and S6I). RNA stability analysis showed that the half-life of pri-let-7a was prolonged in HCT116 and HCT8 cells after SNORD11B knockdown (Fig. 5J-K) and was shortened in HIEC cells after SNORD11B overexpression (Fig. S6J).
Taken together, these results show that SNORD11B can affect pri-let-7a stability and inhibit the expression of the tumor suppressor gene let-7a-5p to promote CRC development.
SNORD11B mediates Nm modification of pri-let-7a and inhibits splicing
Notably, we found that the region where pri-let-7a binds to SNORD11B was similar to 18S rRNA, (upstream of the D box), so we speculated that the G base on the pri-let-7a corresponding to the 5 nt upstream of the SNORD11B D box may have undergone the same Nm modification as the G509 site of 18S rRNA. We mapped this binding region to the let-7a-5p host gene, lncRNA MIRLET7A1HG (miRlet-7a-1/let-7f-1/let-7d cluster host gene), and the 220–230 nt region of MIRLET7A1HG was found to be complementarily base-paired with the 84–94 nt region of SNORD11B; moreover, Nm modification may occur at the G225 site of MIRLET7A1HG (Fig. 6A). To identify whether this site underwent Nm modification, we designed site-specific primers: MeUA-RT (3' without Nm modification site G225) and MeA-RT (3' with Nm modification site G225) (Figs. 6B and S7A) and performed RTL-qPCR experiments as described previously 22 (Fig. 6C). In HCT116 and HCT8 cells after SNORD11B knockdown, the relative levels of qPCR products in the MeUA-RT with low dNTP group were significantly increased, whereas those in the MeA-RT group were unchanged (Figs. 6D-E and S7B-C). The opposite results were obtained in HIEC with SNORD11B overexpression (Fig. S7D). These findings strongly suggest that SNORD11B mediates Nm modification at the G225 site of MIRLET7A1HG.
For further validation, we performed FISH experiments to explore that fibrillarin (FBL), the key enzyme mediating Nm modification8, co-localized with pri-let-7a in the nucleolus (Fig. 6G-H). We then performed RNA immunoprecipitation (RIP)-qPCR experiments to verify the binding of pri-let-7a to FBL. According to the results, anti-FBL significantly recruited pri-let-7a, and this enrichment decreased to a level comparable with that of the IgG group after SNORD11B knockdown (Fig. 6I-J), indicating that FBL interacted with pri-let-7a with the involvement of SNORD11B. As the G225 site and the SNORD11B-binding region are located in the stem-loop structure of let-7a precursor, we hypothesized that the Nm modification of this site might affect the binding of DGCR8/DROSHA to pri-let-7a, thereby inhibiting splicing and maturation. We therefore performed anti-DGCR8 RIP-qPCR and found that the binding of DGCR8 to pri-let-7a was significantly enhanced upon SNORD11B knockdown, whereas DGCR8 did not bind to SNORD11B (Fig. 6K-L).
These findings suggest that SNORD11B attenuates the binding of DGCR8 to pri-let-7a by mediating Nm modification at the G225 site, thereby inhibiting the splicing and maturation of pri-let-7a in the nucleus, decreasing the expression of mature let-7a-5p, and promoting the malignant progression of CRC.
SNORD11B promotes CRC malignancy via let-7a-5p
To further verify that SNORD11B promotes CRC malignancy by inhibiting let-7a-5p, we conducted rescue experiments. We transfected SNORD11B-knockdown HCT116 and HCT8 cells with a mimic or inhibitor against let-7a-5p to overexpress or knock down the expression of let-7a-5p in the target cells. Transfection with mimic or inhibitor did not affect the expression of SNORD11B (Figs. 7A and S8A); the transfection efficiencies are shown in Figs. 7B and S8B. SNORD11B knockdown followed by further inhibition of let-7a-5p expression partially counteracted the inhibition of cell proliferation and invasion, and suppressed the increase in apoptosis caused by SNORD11B silencing, whereas followed by further let-7a-5p overexpression exacerbated the opposite results. (Figs. 7C-J and S8CJ). We verified the expression of let-7a-5p downstream genes by western blot assay and found that simultaneous knockdown of SNORD11B and let-7a-5p expression significantly enhanced the expression of target genes compared with SNORD11B knockdown alone, which was contrary to the results for SNORD11B-knockdown and let-7a-5p-overexpressing cells (Figs. 7K-L and S8K-L). These results suggest that SNORD11B promotes the malignant progression of CRC by inhibiting let-7a-5p and subsequently enhances the expression of oncogenes downstream of let-7a-5p.