Correlation between Differential Expression of m6A and Prognosis of

Uterine Corpus Endometrial Carcinoma

Mengqi Zuo

Shanghai Tianjiabing Secondary School, Shanghai, China

Keywords: Uterine Corpus Endometrial Carcinoma (UCEC), Overall Survival (OS), m6A, Prognosis Prediction.

Abstract: UCEC, known as Uterine corpus endometrial carcinoma, is one of the most common types of gynecologic

malignancy worldwide. Notwithstanding great focus has been put on the treatments of UCEC recently, both

the incidence rate and mortality rate of UCEC are still increasing. The 5-year overall survival (OS) rate in

early-stage UCEC ranges from 74 to 91%. Chemotherapy and hormone therapy are viable treatment options

for patients with metastasis or recurrence. However, not all patients benefit from these. For advanced stage

III or IV disease, the 5-year OS rates are 57–66% and 20–26%, respectively. The most common form of post

transcriptional RNA modification, N6-methyladenine (m6A) has attracted increasing interest in cancer

pathogenesis and progression. The differential expression of m6A could be an important clue in the area of

prognosis. Thus, we aimed to identify the correlations between m6A expressions and prognosis of UCEC,

and build a prognostic gene signature in UCEC. In this study, firstly, we filtrated and analysed the gene

expression in RNA sequence and the matched clinical information of UCEC patients from The Cancer

Genome Atlas (TCGA) database. Second, we determined that several m6A regulatory genes had a significant

negative impact on patient survival. By using the Statistical Product and Service Solutions (SPSS) and R-

studio, we built both a univariate Cox regression model and a multivariate Cox regression model. In the end,

we discovered these four m6A gene expressions that had a significant association with the UCEC patient

survival data: VIRMA, METTL14, HNRNPC and FTO. Whereas the multivariate Cox regression model’s

analysis suggested that risk score might be an independent prognostic indicator for the overall survival of

patients with UCEC (p-value ¡0.05). In conclusion, m6A regulator could be an effective and reliable

biomarker for future UCEC prognosis prediction and it deserves further research.

1 INTRODUCTION

Uterine Corpus Endometrial Carcinoma (UCEC), a

common gynaecologic malignancy worldwide, is

defined as an epithelial neoplasm originating from the

endometrium. According to recent research, it is

estimated that there will be 66,570 new cases and an

estimated 12,940 people will die of this disease in

2021 worldwide. Recently, increasing attention has

been paid to adjuvant therapy and targeted therapy in

the overview of the main research progress on UCEC.

Indeed, great advances were made in the treatments

of UCEC. However, the incidence and mortality rates

are still increasing globally. Under this circumstance,

it is crucial to identify novel clinical potential

prognostic biomarkers and therapeutic targets to

improve the patients’ survival of UCEC.

To date, various post-transcriptional RNA

modifications have been discovered and identified as

an epigenetic regulation mechanism in cells and play

a crucial role in a variety of biological diseases,

especially cancers. N6-methyladenine (m6A) mRNA

modification, being the most abundant form of RNA

modification in eukaryotes, has attracted increasing

interest recently. M6A modification relies on a series

of enzymes, which are named “writers”

(methyltransferases), “erasers” (demethylases), and

“readers” (m6A-binding proteins), that can add,

remove, or preferentially bind to m6A functional

sites, thereby altering important biological functions.

The mechanism of m6A in cancer pathogenesis and

progression has been reported in various studies. For

example, researchers found that METTL3, a type of

methyltransferase, acts 2 as an oncogene in lung

cancer and nasopharyngeal carcinoma (NPC).

METTL3 enhances translation of epidermal growth

factor receptor (EGFR). In lung squamous cell

carcinoma, METTL3 interacts with eukaryotic

Zuo, M.

Correlation between Differential Expression of m6A and Prognosis of Uterine Corpus Endometrial Carcinoma.

DOI: 10.5220/0011375000003443

In Proceedings of the 4th International Conference on Biomedical Engineering and Bioinformatics (ICBEB 2022), pages 979-984

ISBN: 978-989-758-595-1

979

translation initiation to accelerate tumorigenicity by

promoting translation of oncogenic mRNAs, such as

Bromodomain-containing protein 4 (BRD4).

In this study, we first analyzed the gene

expression data in RNA sequences and matched

clinical information of UCEC patients from The

Cancer Genome Atlas (TCGA) database. The mRNA

expression levels of a total of 16 m6A regulators were

significantly correlated with different patient data.

Multivariate Cox regression and survival test analysis

suggested that the risk score based on the p-value

(¡0.05) might be an independent prognostic indicator

for the overall survival of patients of UCEC.

2 MATERIALS AND METHODS

2.1 Data Acquisition and Processing

Clinical data of UCEC, including gene expression

RNAseq FPKM, phenotype, and surviving data, were

downloaded from the UCSC Xena website

(https://xenabrowser.net/datapages/). Then they were

processed and sorted by Excel. The download time

was July 2021. There were originally 185 cases

included within the HTseq FPKM data. By using the

VLOOKUP function, the corresponding gene stable

ID was matched between the surviving data and gene

expression data. After excluding the false data and

null data, there were exclusive cases that were closely

correlated and matched with the patient and surviving

data. Each of the 20 m6A gene expression data was

extracted; then they were processed and categorized

into two conditions-low or high gene expression.

Meanwhile, age, overall survival (OS) time, Clinical

M and Clinical T, these 4 categories and their

following data were selected from the phenotype data

acting as another set of variables. Considering periods

usually ended as the patient’s age reached above 45,

the patient age data were differentiated into two

categories: lower or equal to the age of 45, or higher

than the age of 45.

2.2 Proportional Hazard Regression

Model

We denote that 𝑓(𝑡) being the probability density

function (PDF), ℎ(𝑡) being the hazard ratio, and 𝑆(𝑡)

being the survival.

𝑆

(

𝑡

)

=1−𝐹(𝑡) (1)

ℎ

(

𝑡

)

()

()

(2)

ℎ

(

𝑡

)

=ℎ



(𝑡) ∙ 𝑒𝑥𝑝



𝑋𝛽



(3)

In this model we do not assume that the hazard

ratio changes by time with each patient. The

assumption is that the hazard ratio is proportional to

each risk group.

After we processed and categorized the data by

using Excel, we used the Statistical Product and

Service Solutions (SPSS) and language 𝑅 to

complete further research. By using the Cox

regression model and the survival analysis, we

discovered associations between data sets. After the

data is analyzed, we can use a function to describe the

risk factor.

𝑌 = 0.781 ×

(

𝑅𝐵𝑀15𝐵

)

+ 0.781 ×

(

𝑉𝐼𝑅𝑀𝐴

)

0.878 ×

(

𝐼𝐺𝐹2𝐵𝑃2

)

+ 0.798 ×

(

𝐻𝑁𝑅𝑁𝑃𝐴2𝐵1

)

0.622 ×

(

𝐼𝐺𝐹2𝐵𝑃1

)

+ 0.799 ×

(

𝑌𝑇𝐻𝐷𝐹3

)

+ 0.606 ×

(

𝐼𝐺𝐹2𝐵𝑃3

)

+ 0.663 ×

(

𝐻𝑁𝑅𝑁𝑃𝐶

)

+ 0.750 ×

(

𝑅𝐵𝑀15

)

+ 0.686 ×

(

𝑅𝐵𝑀𝑋

)

+ 1.060 ×

(

𝑀𝐸𝑇𝑇𝐿14

)

0.793 ×

(

𝑌𝑇𝐻𝐷𝐶2

)

+ 0.721 ×

(

𝑀𝐸𝑇𝑇𝐿3

)

+ 0.947 ×

(

𝑍𝐶2𝐻13

)

+ 0.621 ×

(

𝑊𝑇𝐴𝑃

)

+ 0.584 ×

(

𝑌𝑇𝐻𝐷𝐹1

)

1.204 ×

(

𝑌𝑇𝐻𝐷𝐶1

)

+ 0.819 ×

(

𝐹𝑇𝑂

)

+ 0.726 ×

(

𝑌𝑇𝐻𝐷𝐹2

)

+ 0.749 × (𝐴𝐿𝐾𝐵𝐻5) (4)

In this equation, the expression of all twenty m 6A

genes and their hazard ratio are shown.

Then we can determine whether each m6A has a

high or low gene expression. By categorizing each

high or low expression of m6A, we can make a risk

stratification system. In addition, we have added the

clinical data as another variable to judge whether the

gene expression is an independent risk factor itself or

not. We added the patient age, clinical T and clinical

M as three new covariations to the Cox regression

model. After the results came out, we determined that

VIRMA, HNRNPC, METTL14 and FTO were

independent risk factors themselves and were not

influenced by patient data (patient age, clinical T and

clinical M).

3 MATERIALS AND METHODS

Table 1 shows the basic tendency of my data. Those

data based on 1 and 0 were analyzed. Since the OS

time and clinical T are not pure 0s and 1s data, I will

use mean standard deviation to express OS time

clinical T.

Based on the data and the progress in R-studio, I

made this Kaplan-Meier plot. Age=0 means that the

ICBEB 2022 - The International Conference on Biomedical Engineering and Bioinformatics

980

recorded age is equal to or below 45. Age=1 means

that the recorded age is above 45.

Through the Kaplan-Meier plot, we can clearly

see that patients with age equal to or below 45 have a

higher survival probability. Especially when the time

reached 1000-2500 days of getting UCEC, the two

sets of data drew a significant distance.

Table 1: Variable stratified survival.

0 1 p

n 51 28

RBM15B = 1 (%) 21 (41.2) 18 (64.3) 0.084

VIRMA = 1 (%) 22 (43.1) 17 (60.7) 0.208

IGF2BP2 = 1 (%) 22 (43.1) 17 (60.7) 0.208

HNRNPA2B1 = 1 (%) 19 (37.3) 20 (71.4) 0.008

IGF2BP1 = 1 (%) 19 (37.3) 20 (71.4) 0.008

YTHDF3 = 1 (%) 26 (51.0) 13 (46.4) 0.879

IGF2BP3 = 1 (%) 22 (43.1) 17 (60.7) 0.208

HNRNPC = 1 (%) 19 (37.3) 20 (71.4) 0.008

RBM15 = 1 (%) 16 (31.4) 23 (82.1) <0.001

RBMX = 1 (%) 22 (43.1) 17 (60.7) 0.208

METTL14 = 1 (%) 30 (58.8) 9 (32.1) 0.042

YTHDC2 = 1 (%) 26 (51.0) 13 (46.4) 0.879

METTL3 = 1 (%) 22 (43.1) 17 (60.7) 0.208

ZC2H13 = 1 (%) 27 (52.9) 12 (42.9) 0.534

WTAP = 1 (%) 23 (45.1) 16 (57.1) 0.430

YTHDF1 = 1 (%) 25 (49.0) 14 (50.0) 1.000

YTHDC1 = 1 (%) 26 (51.0) 13 (46.4) 0.879

FTO = 1 (%) 29 (56.9) 10 (35.7) 0.118

YTHDF2 = 1 (%) 22 (43.1) 17 (60.7) 0.208

ALKBH5 = 1 (%) 24 (47.1) 15 (53.6) 0.750

OS.time (mean (SD)) 1765.06 (1087.45) 914.68 (669.02) <0.001

age = 1 (%) 28 (54.9) 16 (57.1) 1.000

ClinicalM = 1 (%) 3 ( 5.9) 12 (42.9) <0.001

ClinicalT (mean (SD)) 2.04 (0.80) 3.04 (1.14) <0.001

Correlation between Differential Expression of m6A and Prognosis of Uterine Corpus Endometrial Carcinoma

981

Figure 1: Kaplan Meier Plot.

Figure 2: Proportional hazard model.

ICBEB 2022 - The International Conference on Biomedical Engineering and Bioinformatics

982

Table 2: Proportional hazard model.

coef exp(coef) se(coef)

z Pr(>|z|)

RBM15B 0.33 1.39 0.78

0.42 0.68

VIRMA 1.78 5.92 0.78

2.28 0.02

IGF2BP2 -0.32 0.73 0.88

-0.36 0.36

HNRNPA2B1 1.48 4.41 0.80

1.86 0.06

IGF2BP1 0.93 2.53 0.62

1.49 0.14

YTHDF3 -0.86 0.42 0.80

-1.08 0.28

IGF2BP3 0.33 1.39 0.61

0.54 0.59

HNRNPC 1.78 5.96 0.66

2.69 0.01

RBM15 0.66 1.93 0.75

0.88 0.38

RBMX -1.13 0.32 0.69

-1.64 0.10

METTL14 -2.79 0.06 1.06

-2.63 0.01

YTHDC2 1.15 3.15 0.79

1.45 0.15

METTL3 0.37 1.44 0.72

0.51 0.61

ZC2H13 -0.11 0.89 0.95

-0.12 0.91

WTAP 0.28 1.33 0.62

0.45 0.65

YTHDF1 0.4. 1.49 0.58

0.68 0.50

YTHDC1 1.52 4.55 1.20

1.26 0.21

FTO -2.47 0.08 0.82

-3.02 0.00

YTHDF2 -1.16 0.31 0.73

-1.60 0.11

ALKBH5 0.31 1.36 0.75

0.41 0.68

age 0.04 1.04 0.67

0.06 0.96

ClinicalM -0.18 0.84 0.82

-0.21 0.83

ClinicalT 0.91 2.52 0.39

2.40 0.02

In Table 2, compared with patients who did not

express VRIMA, the hazard ratio of patients who

expressed VRIMA was increased by 5.92.

The hazard ratio increases by 5.92 for the patients

with HNRNPC expression compared to the patients

without HNRNPC expression.

The hazard ratio increases by 5.92 for the patients

with METTL14 expression compared to the patients

without METTL14 expression.

The hazard ratio increases by 5.92 for the patients

with FTO expression compared to the patients

without FTO expression.

4 DISCUSSION AND

CONCLUSIONS

UCEC - Uterine Corpus Endometrial Carcinoma OS

- Overall Survival, defined as the time from

randomization to death from any cause, is a direct

measure of clinical benefit to a patient. It is also a

good standard primary end point to evaluate the

outcome of procedure that is assessed in oncologic

clinical trials.

In conclusion, m6A regulator could be an

effective and reliable biomarker for future UCEC

prognosis prediction and it deserves further research.

Indeed, among the 21 m6A genes, only 4 of them

were closely related to the stage and the risk level.

Correlation between Differential Expression of m6A and Prognosis of Uterine Corpus Endometrial Carcinoma

983

This means the prognosis with m6A still has some

limitations. Also, there are still some uncertainties in

this research. For example, RBMX, IGF2BP1, and

other m6A genes. But that is not to say that it is not

beneficial.

In fact, the number of genes and samples included

in this study are limited. In further study, data

enrichment should be used to have a more accurate

and reliable result. The 4 m6A genes we had

identified their different gene expression having an

impact on patient survival data: VIRMA, METTL14,

HNRNPC and FTO. These 4 m6A genes could be

used as a biomark for the prognosis of UCEC

patients.

REFERENCES

Choe J, Lin S, Zhang W, Liu Q, Wang L, Ramirez-Moya J,

Du P, Kim W, Tang S, Sliz P, et al. mRNA

circularization by METTL3-eIF3h enhances translation

and promotes oncogenesis. Nature. 201c8; 561: 556–

60.

James J Driscoll, Oliver Rixe, 2009. Overall survival: still

the gold standard: why overall survival remains the

definitive end point in cancer clinical trials. [online]

pubmed. Available at:

https://pubmed.ncbi.nlm.nih.gov/19826360/ [Accessed

29 November 2021].

Murali R., Soslow R.A. and Weigelt B. (2014)

Classification of endometrial carcinoma: more than two

types. Lancet Oncol. 15, e268–e278 10.1016/S1470-

2045(13)70591-6

Zhou Z, Lv J, Yu H, Han J, Yang X, Feng D, Wu Q, Yuan

B, Lu Q, Yang H. Mechanism of RNA modification

N6-methyladenosine in human cancer. Mol Cancer.

2020 Jun 8;19(1):104. doi: 10.1186/s12943-020-

01216-3. PMID: 32513173; PMCID: PMC7278081.

ICBEB 2022 - The International Conference on Biomedical Engineering and Bioinformatics

984