Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 May 7;21(1):97.
doi: 10.1186/s13059-020-02009-z.

Integrative Analyses of the RNA Modification Machinery Reveal Tissue- And Cancer-Specific Signatures

Affiliations
Free PMC article

Integrative Analyses of the RNA Modification Machinery Reveal Tissue- And Cancer-Specific Signatures

Oguzhan Begik et al. Genome Biol. .
Free PMC article

Abstract

Background: RNA modifications play central roles in cellular fate and differentiation. However, the machinery responsible for placing, removing, and recognizing more than 170 RNA modifications remains largely uncharacterized and poorly annotated, and we currently lack integrative studies that identify which RNA modification-related proteins (RMPs) may be dysregulated in each cancer type.

Results: Here, we perform a comprehensive annotation and evolutionary analysis of human RMPs, as well as an integrative analysis of their expression patterns across 32 tissues, 10 species, and 13,358 paired tumor-normal human samples. Our analysis reveals an unanticipated heterogeneity of RMP expression patterns across mammalian tissues, with a vast proportion of duplicated enzymes displaying testis-specific expression, suggesting a key role for RNA modifications in sperm formation and possibly intergenerational inheritance. We uncover many RMPs that are dysregulated in various types of cancer, and whose expression levels are predictive of cancer progression. Surprisingly, we find that several commonly studied RNA modification enzymes such as METTL3 or FTO are not significantly upregulated in most cancer types, whereas several less-characterized RMPs, such as LAGE3 and HENMT1, are dysregulated in many cancers.

Conclusions: Our analyses reveal an unanticipated heterogeneity in the expression patterns of RMPs across mammalian tissues and uncover a large proportion of dysregulated RMPs in multiple cancer types. We provide novel targets for future cancer research studies targeting the human epitranscriptome, as well as foundations to understand cell type-specific behaviors that are orchestrated by RNA modifications.

Keywords: Dysregulation in cancer; Epitranscriptomics; RNA modifications; Tissue specificity.

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Fig. 1
Fig. 1
Evolutionary analysis of RNA modification “writers.” a Detailed overview of the evolutionary history of RMW duplications during eukaryotic evolution. Red stars indicate that proteins do not target RNAs but they are in the same family with an RNA writer protein. Red lines indicate the evolutionary group in which the enzyme has appeared. b Histogram of RMW duplication events throughout eukaryotic evolution. Duplication events were inferred using multiple sequence alignments, coupled to maximum likelihood tree generation, for each family. c, d Maximum likelihood phylogenetic trees of methyltransferase family METTL2A/2B/6/8 (c) and TRMT10A/B/C (d). Cyan squares indicate the node where the duplication occurred. Numbers shown on the branches indicate bootstrapping values
Fig. 2
Fig. 2
Analysis of RMP tissue specificity expression in different species. a Heatmap of z-scaled log(TPM) values of catalytic RNA writer proteins (M: methyltransferases; D: deaminases; P: pseudouridylases) throughout human and mouse tissues. In both human and mouse, testis has the most distinct RMP expression pattern in which many genes show very high expression, whereas other tissues such as colon show moderate expression level of RMPs. b Scatter plots depicting tissue specificity analysis, which have been computed by representing the RMP mRNA expression values in a given tissue (y-axis) relative to the mean mRNA abundance in all tissues (x-axis). Scatter plots show that testis has a significant number of tissue-specific genes in both human and mouse, while colon shows no tissue-specific genes in human and only one in mouse. Tissue-specific genes are labeled in red. c Venn diagram of the conservation of tissue specificity between human and mouse. Out of 26 common tissue-specific genes, 16 of them are specifically expressed in the same tissue. d Principal component analysis of amniote tissues based on the log(RPKM) mRNA expression of their RMPs. The loadings plot (left) shows the contribution of each RMP to the clustering of amniote tissues. The score plot (right) shows the clustering of each tissue, where testis tissue (in red) is the main contributor to the variance of the data, and is found apart from the rest of the amniote tissues for every given species. e Schematic representation of the fate of the 46 RMW duplication events shown in Fig. 1, showing that 89% of them suffered a change in their tissue and/or target specificity
Fig. 3
Fig. 3
Analysis of RMP gene expression during spermatogenesis. a Schematic representation of the four main phases of spermatogenesis: (i) mitotic division of spermatogonia (SPG) into primary spermatocytes (PSC), (ii) meiotic division of PSCs into secondary spermatocytes (SC), (iii) meiotic division SCs into round spermatids (RST), and (iv) spermiogenesis, in which round spermatids (RST) mature into elongated spermatids (EST). b Heatmap of RMP expression levels in mouse testis. RMPs were clustered into 4 groups based on k-means analysis of their normalized average mRNA expression values. c Violin plots of the expression patterns of each of the 4 identified clusters. d RNA median expression barplot and immunohistochemistry of NSUN7, NSUN2, and METTL14, depicting distinct protein expression levels along the different sections of the testis and epididymis, as well as different subcellular localizations. Brown color indicates a specific staining of the antibody whereas blue represents hematoxylin counterstain
Fig. 4
Fig. 4
Expression analysis of RMPs in human tumor-normal paired samples. a Heatmap of z-scaled dysregulation scores of RMPs in tumor-normal paired samples, across 28 cancer types. Positive (red) values indicate upregulation in tumor samples, whereas negative (blue) values indicate downregulation. Genes labeled as red (upregulated) and blue (downregulated) represent top significantly dysregulated genes, which are also individually listed in panel c. b Scatter plot comparing RMP expression levels of matched tumor-normal samples, for the following cancer types: LAML (acute myeloid leukemia) and UCS (uterine carcinosarcoma) BRCA (breast invasive carcinoma) and KIRP (kidney renal papillary cell carcinoma). Values represent median log(TPM) across all patients. Black data points indicate the expression of RMPs, where dysregulated genes are highlighted in red (upregulated) or blue (downregulated). Non-RMP genes are depicted in gray. c Barplot illustrates the number of cancer types in which significantly dysregulated genes are highlighted in red (upregulated) or blue (downregulated). Only RMPs that are dysregulated in more than 2 cancer types are shown. For the full list of dysregulated RMPs, see Table 1. d Boxplots of log(TPM) mRNA expression values of HENMT1 (upper panel) and LAGE3 (bottom panel) across all 28 cancer types analyzed in this work. Green box plots represent normal samples, whereas red box plots represent tumor samples. Tumor-normal pairs highlighted in cyan represent cancer types in which the RMP is significantly downregulated, whereas those highlighted in orange represent those cancer types in which the RMP is upregulated. Error bars represent standard deviation of mRNA expression levels across patients. Each data point represents a different patient sample. Abbreviations: ACC (adrenocortical carcinoma), BLCA (bladder urothelial carcinoma), BRCA (breast invasive carcinoma), CESC (cervical squamous cell carcinoma and endocervical adenocarcinoma), COAD (colon adenocarcinoma), ESCA (esophageal carcinoma), GBM (glioblastoma multiforme), HNSC (head and neck squamous cell carcinoma), KICH (kidney chromophobe), KIRC (kidney renal clear cell carcinoma), KIRP (kidney renal papillary cell carcinoma), LAML (acute myeloid leukemia), LGG (brain lower-grade glioma), LIHC (liver hepatocellular carcinoma), LUAD (lung adenocarcinoma), LUSC (lung squamous cell carcinoma), OV (ovarian serous cystadenocarcinoma), PAAD (pancreatic adenocarcinoma), PCPG (pheochromocytoma and paraganglioma), PRAD (prostate adenocarcinoma), READ (rectum adenocarcinoma), SARC (sarcoma), SKCM (skin cutaneous melanoma), STAD (stomach adenocarcinoma), TGCT (testicular germ cell tumors), THCA (thyroid carcinoma), UCEC (uterine corpus endometrial carcinoma), UCS (uterine carcinosarcoma)
Fig. 5
Fig. 5
Immunohistochemical analysis and prognostic value of RMP expression levels in different cancer types. a, b Immunohistochemical analysis and images of normal and tumor LAGE-3 stained LUSC (lung squamous cell carcinoma), LIHC (liver hepatocellular carcinoma), and PRAD (prostate adenocarcinoma) (a) and HENMT-1 stained HGSC (High-grade serous carcinoma), LUSC, and STAD (stomach adenocarcinoma) TMAs (b). Representative cores and subsets are shown for each tissue and antibody, where the brown color indicates a specific staining of the antibody and blue represents the hematoxylin counterstain. Mean TMA score is plotted for each core, with three cores from different individuals per condition quantified. Two-sided Wilcoxon tests did not yield significant differences in any comparison, p values of all tumor-normal comparisons for each cancer type and antibody are shown in Figure S13. c Heatmap of survival p values of 146 RMPs across 28 cancer types. Survival p values are calculated by comparing the prognosis of patients that express high (upper 50%) versus low (lower 50%) RMP levels. “N” column shows the number of patients included for the analysis of each cancer type. d Individual examples of survival plots where the expression levels of the RMP are predictive of cancer prognosis. p values have been calculated by comparing the survival between patients expressing high levels (yellow, top 50%) versus low expression levels (blue, bottom 50%)

Similar articles

See all similar articles

References

    1. Saletore Y, Meyer K, Korlach J, Vilfan ID, Jaffrey S, Mason CE. The birth of the epitranscriptome: deciphering the function of RNA modifications. Genome Biol. 2012;13:175. doi: 10.1186/gb-2012-13-10-175. - DOI - PMC - PubMed
    1. Zhou J, Wan J, Shu XE, Mao Y, Liu X-M, Yuan X, et al. N6-methyladenosine guides mRNA alternative translation during integrated stress response. Mol Cell. 2018;69:636–47.e7. doi: 10.1016/j.molcel.2018.01.019. - DOI - PMC - PubMed
    1. Safra M, Sas-Chen A, Nir R, Winkler R, Nachshon A, Bar-Yaacov D, et al. The m1A landscape on cytosolic and mitochondrial mRNA at single-base resolution. Nature. 2017;551:251–255. doi: 10.1038/nature24456. - DOI - PubMed
    1. Warda AS, Kretschmer J, Hackert P, Lenz C, Urlaub H, Höbartner C, et al. Human METTL16 is a N6-methyladenosine (m6A) methyltransferase that targets pre-mRNAs and various non-coding RNAs. EMBO Rep. 2017;18:2004–2014. doi: 10.15252/embr.201744940. - DOI - PMC - PubMed
    1. Lim SL, Qu ZP, Kortschak RD, Lawrence DM, Geoghegan J, Hempfling A-L, et al. HENMT1 and piRNA stability are required for adult male germ cell transposon repression and to define the spermatogenic program in the mouse. PLoS Genet. 2015;11:e1005620. doi: 10.1371/journal.pgen.1005620. - DOI - PMC - PubMed

LinkOut - more resources

Feedback