Model

Background:

Virtually few accurate and robust prediction models of gastric cancer (GC) early diagnosis exist that may aid physicians in making clinical decisions. We aimed to develop a diagnostic prediction model of GC by incorporating biomarker G3BP1 and luciferase reporting system.

Method:

Based on the gene expression profiles of G3BP1 in GC cell line, we proposed a statistical model constructed by a luciferase designed plasmids. The construction of model involves several steps, including data collection, data preprocessing and evaluation.

The GTPase activating protein (SH3 domain) binding protein 1 could interacts with the Ras-GTPase-activating protein, thereby playing an important role in the Ras signaling pathway which is crucial for cell growth, differentiation, and survival. G3BP1 has been implicated in the regulation of cell proliferation and apoptosis. The interaction with various signaling molecules can influence cell survival and growth, which is particularly relevant in cancer biology. According to its role in stress response and signaling transduction, G3BP1 is a promising biomarker for GC early diagnosis.

Step-by-step design of the statistical model

1.Data collection

1.1Gastric cancer cell line

In this experiment, two different GC cell lines (AGS and MGC-803) were selected, and one normal cell line (GES-1) was chosen as health control. AGS is a cell line exhibiting epithelial morphology isolated from gastric adenocarcinoma. MGC-803 is a hybrid cell line of Hela and gastric cancer cells from an Asian individual with poorly differentiated gastric mucinous adenocarcinoma. Besides, GES-1 is a normal human gastric epithelial cell line.

1.2Biomarker data

The expression level of G3BP1 was quantified in GC cell lines using real-time qPCR. Total RNA was extracted from whole cell lysates via the QIAGEN RNeasy kit with on-column NDase I treatment (Qiagen). cDNA was reverse transcribed from 1 μg of RNA using the Invitrogen reverse transfection kit according to the manufacturer’s instructions. For dsRNA enrichment, RNA was first treated for 30 min with 50 μg of ml-1 RNase A (Qiagen) in high salt concentration (NaCl, 0.35 M) to prevent dsRNA degradation. After treatment, RNase A was removed by ethanol precipitation and the product was resuspended in sterile water. Gene-specific primers for SYBR Green real-time PCR were either obtained from previously published sequences or designed by PrimerBlast (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). The mRNA expression levels of G3BP1 were quantified in five technical replications on a StepOnePlus Cycler (Applied Biosystems) with 10 ng of cDNA in a 20-μl reaction volume using ThermoFisher’s Power SYBR Green Master Mix.

1.3the construction of luciferase reporting system

Based on the function of G3BP1 in regulating mRNA with highly structures in the 3'UTR region, we designed and synthesized two recombinant plasmid systems (Luciferase system and EGFP system) for detecting the expression level of G3BP1 in patient cells. Four recombinant plasmids were cloned into the downstream regions of firefly luciferase and EGFP, respectively, by cloning the highly-structured fragment (HSU) and the mutant fragment (MUT) without highly structures. The recombinant plasmids were then transfected into three cell lines, GES-1, AGS, and MGC-803, using the Opti-MEM transfection kit. Luciferase activity and EGFP fluorescence intensity were detected using a spectrophotometer and fluorescence microscope.The luciferase activity and EGFP fluorescence intensity are negatively correlated with the content of G3BP1 in cells.

2.results

2.1 Biomarker evaluation

First, this study evaluated the G3BP1 upregulation in GC cell lines (Table 1). The qPCR with reverse transcription (RT-qPCR) confirmed a significant upregulation of mRNA G3BP1 in GC cell lines (AGS and MGC-803) rather than in normal human gastric epithelial cell line. However, the expression level of G3BP1 did not show a significant difference between GC cell lines, indicating that G3BP1 was generally dysregulated in different gastric cell lines (Figure 1). In summary, these data confirmed that G3BP1 was a potential biomarker for GC.

Figure 1. Gene expression of G3BP1 was assessed by RT-qPCR in cell lines.

Table 1. real-time qPCR data of G3BP1

Cell type	Rep 1	Rep 2	Rep 3	Rep 4	Rep 5
GES-1	0.975539	0.86796	0.970079	1.217442	1.007755
AGS	1.490177	1.831077	1.533925	2.958479	1.953414
MGC-803	1.709916	1.981862	1.97573	2.113312	1.945205

2.2 luciferase data preprocessing

The activity ratio of double luciferase (firefly luciferase [FLUC] and renilla luciferase [RLUC]) were collected and calculated (Table 2). Besides, the enhanced green fluorescent protein (eGFP) is also used to report the experiment results (Table 2).

Table 2. raw data of luciferase reporting system

Cell type	Plasmid	FLUC	RLUC	Activity ratio	eGFP
GES-1	CTL	5364	15691	0.3418	1500632
GES-1	CTL	4750	14983	0.3171	1354177
GES-1	CTL	4250	14630	0.2905	1415055
GES-1	CTL	4463	16048	0.2781	1467049
GES-1	MUT	3416	13529	0.2524	1583793
GES-1	MUT	4889	17172	0.2847	1492877
GES-1	MUT	3737	15789	0.2367	1476780
GES-1	MUT	3933	12742	0.3086	1412478
GES-1	HSU	4325	12642	0.3421	1413776
GES-1	HSU	5282	19833	0.2663	1353230
GES-1	HSU	3742	12716	0.2942	1406070
GES-1	HSU	3579	15378	0.2327	1308666
MGC-803	CTL	15010	43367	0.3461	1373540
MGC-803	CTL	13979	40254	0.3472	1385270
MGC-803	CTL	17664	58931	0.2997	1445934
MGC-803	CTL	15354	54374	0.2824	1367696
MGC-803	MUT	25114	90591	0.2772	1342005
MGC-803	MUT	35651	112669	0.3164	1390080
MGC-803	MUT	24203	74577	0.3245	1336585
MGC-803	MUT	65058	188129	0.3458	1428417
MGC-803	HSU	17266	98698	0.1749	404993
MGC-803	HSU	11745	53392	0.22	506038
MGC-803	HSU	11080	72314	0.1532	451537
MGC-803	HSU	26330	116247	0.2265	421352
AGS	CTL	6754	20496	0.3295	1280530
AGS	CTL	7863	21549	0.3649	1228251
AGS	CTL	5586	18647	0.2995	1232706
AGS	CTL	10125	26286	0.3852	1153998
AGS	MUT	3211	9879	0.3251	1197769
AGS	MUT	2080	5860	0.3549	1228068
AGS	MUT	4715	14185	0.3324	1131863
AGS	MUT	4904	11972	0.4096	1237845
AGS	HSU	795	3782	0.2102	368462
AGS	HSU	346	4125	0.084	310629
AGS	HSU	3710	17992	0.2062	404993
AGS	HSU	3915	16176	0.2421	376970

2.3 Statistical model construction

To evaluate the predicting effects of the luciferase reporting system, we performed multivariate Cox regression analysis. The analysis used cell line type as dependent variable (normal cell versus cancer cell). The results revealed the significant results of the activity ratio of double luciferase (P < 0.05) and eGFP level (P=0.05) (Table 3). Therefore, the statistical model using the coefficient estimate form the Cox regression was defined as:

Y = 39.2xDoubleRatio - 0.0000242 x eGFP

Table 3. multivariate Cox regression analysis of cancer cell type

Variables	β	S.E	Z	P	OR (95%CI)
Variables	β	S.E	Z	P	OR (95%CI)
Intercept	2.16E+01	1.55E+01	1.391	0.1643	2.32e+09 (1.23 ~ 1.76e+26)
DoubleRatio	3.92E+01	1.96E+01	1.998	0.0457	1.05e+17 (357.08 ~ 2.45e+37)
eGFP	-2.42E-05	1.25E-05	-1.938	0.0526	0.99 (0.99 ~ 1.00)

Furthermore, we used cell line type (GES-1 versus MGC-803 versus AGS) as dependent variable to construct the statistical model. The results confirmed the significant results of the activity ratio of double luciferase (P < 0.05) and eGFP level (P < 0.05) (Table 4). These results further validated that the luciferase reporting system was more sensitive to different cancer cell lines. The statistical model were robustness.

Table 4. multivariate Cox regression analysis of cell line type

Variables	β	S.E	Z	P	OR (95%CI)
Variables	β	S.E	Z	P	OR (95%CI)
Intercept	-8.24E-01	1.85E+00	-0.444	0.65696	0.43 (8.68e-03 ~ 21.46)
DoubleRatio	2.73E+01	1.08E+01	2.527	0.01151	6.96e+11 (6.01e+03 ~ 4.66e+22)
eGFP	-6.07E-06	2.03E-06	-2.988	0.00281	0.99 (0.99 ~ 1.00)

To exclude the potential effects of plasmid, this analysis used groups of plasmid as dependent variable (CTL versus MUT versus HSU). The results revealed that the results of activity ratio of double luciferase (P = 0.09) and eGFP level (P = 0.18) were not significant (Table 5). These results clarified that the luciferase signal was not influenced by plasmid.

Table 5. multivariate Cox regression analysis of plasmid process

Variables	β	S.E	Z	P	OR (95%CI)
Variables	β	S.E	Z	P	OR (95%CI)
Intercept	9.40E+00	5.10E+00	1.844	0.0652	1.21e+04 (1.85e+1 ~ 3.64e+10)
DoubleRatio	-1.68E+01	1.00E+01	-1.676	0.0937	5.12e-08 (3.22e-18 ~ 2.50)
eGFP	-3.51E-06	2.66E-06	-1.32	0.1867	9.99 (9.98 ~ 1.00)

Summary

This study possessed an excellent discriminative ability to identify patients of gastric cancer at high mortality risk. Our statistical model revealed that the luciferase reporting system was significantly associated with disease diagnosis in accuracy and robustness.