TGFB1 polymorphisms and TGF‐β1 plasma levels identify gastric adenocarcinoma patients with lower survival rate and disseminated disease

TGF‐β1 is involved in tumour growth. Four TGFB1 SNPs and TGF‐β1 production by stimulated PBMC were determined in seventy‐eight gastric adenocarcinoma patients. In addition, TGF‐β1 levels were measured in the plasma of further thirty patients. rs1800471‐G/C genotype was prevalent in patients (20.7%) compared to controls (8.4%), as it also was the rs1800468 SNP‐G/A genotype in stage IV patients (20.7%) compared to stage I, II and III patients, combined (10.3%). Conversely, the T/T rs1800469 SNP‐T/T genotype was absent in the former group and present in 19.0% in the latter. Furthermore, the rs1800469‐C/rs1800470‐T (CT) haplotype was found in 15.0% of stage IV patients as compared to 3.0% of the remaining patients (3.0%) and also identifies patients with worse five‐year life expectancy (P = .03). TGF‐β1 synthesis by stimulated PBMCs was significantly lower in patients with the risk SNPs or haplotype, compared to the alternative genotype. Finally, TGF‐β1 plasma levels were lower in patients with worse life expectancy. Analysis of TGFB1 SNPs and measurement of plasma TGF‐β1 levels serves to identify patients at risk of developing a more aggressive disease.

forms of gastric cancer. [2][3][4][5] The identification of patients with poor prognosis allows the adoption of more radical therapeutic or surgical approaches upon diagnosis.
Gastric adenocarcinoma is the main type of gastric tumours and accounts for more than 90% of cases. Complex genetic and environmental factors interact and, together, lead to its initiation and progression. These factors, along with the high rate of somatic mutations, 6 can play an important role in the malignancy and low survival rate of gastric cancer.
The tumour microenvironment is involved in the development and evolution of this pathology. The immune system, as the principal mediator of the inflammatory response taking place at the gastric epithelium in response to infectious or other agents, may play a dual role in the progression of the malignancy, whether fostering or limiting the tumour growth and dissemination. Inflammation and DNA damage are mutually related, as each one can be the initiator of the other, leading to mutations, hyperplasia and, finally, neoplasia. 7 However, immune cells also play a protective role in the cancer immunoediting hypothesis, whereby the immune system can recognize and eliminate tumours in the initial stages of the pathology.
TGF-β1 is a cytokine that plays a dual role in gastric cancer, either promoting cancer development, acting as a factor that inhibits immunosurveillance and promotes epithelial-mesenchymal transition (EMT) and metastases or, alternatively, suppressing tumour growth by directly inhibiting cell cycle progression, leading to growth arrest and activation of apoptotic pathways. 8,9 TGF-β1 can be synthetized by healthy epithelia and, in the tumour microenvironment, by immune and tumoural cells. Studies focused on TGF-β1 at the gene or protein level were carried out by several authors in other types of cancer, but they are scarce in gastric cancer.
The gene that codifies this cytokine, TGFB1, includes several single nucleotide polymorphisms (SNPs). In this study, we focused on four of these SNPs: rs1800468, rs1800469 (both located in the promoter), rs1800470 and rs1800971 (both in the exon 1, that codifies for the signal peptide of the protein). These polymorphisms can modify the expression of the protein, either affecting the joining of transcription factors to the promoter region of the gene 10,11 or modifying the effective secretion of the molecule to the extracellular medium. [12][13][14] In fact, genetic polymorphisms of the TGFB1 gene were already associated with increased or reduced risk of development and evolution of other types of pathologies mainly related to the immune response and inflammation processes. [15][16][17][18][19] Thus, we studied these polymorphic variants in a population of patients with gastric adenocarcinoma in different clinical stages of development (TNM-7th edition 2009 UICC/AJCC), 20 as well as in a population of non-affected controls. Moreover, we wished to assess the effect of the different SNPs on the levels of the TGF-β1 produced by PBMC. Finally, we measured TGF-β1 plasma levels to determine whether we could identify patients with worse prognostic.
Results achieved would contribute to establishing TGFB1 gene SNPs as new possible diagnostic and prognostic biomarkers as well as clarifying the actual role of this cytokine in gastric cancer.

| Patients
One-hundred and eight patients with gastric adenocarcinoma undergoing surgery (at Hospital Universitario Príncipe de Asturias, Alcalá de Henares, Madrid, Spain) were involved. This cohort was split as follows: seventy-eight patients were subjected to DNA extraction and subsequent TGFB1 SNPs analysis and measurement of TGF-β1 production by purified PBMCs, and thirty patients were used to measure plasma TGF-β1 levels.
Patients were classified according to TNM-7th edition 2009 (UICC/AJCC) criteria and grouped in stages I (early cancer), II, III (locally advanced disease) and IV (disseminated disease) ( Table 1).

| Control patients
Two sources of control individuals were used: 1. Fifty-four non-cancer individuals were included as a control group. DNA was obtained either from blood or saliva samples.
Methods to isolate DNA are described in the corresponding section.
2. In addition, and to assess differences in the distribution of the SNPs between patients and controls in a larger cohort, a group of 106 individuals from the same location (Iberian population in Spain), genotyped for the TGFB1 polymorphisms previously mentioned and included in the 1000 Genomes Project database, 21 were added.

| Biological samples
Tissue Upon surgery, a sample of gastric tumoural or distal tissue was obtained from each patient.

Blood
Blood was obtained from gastric cancer and controls in EDTAcontaining tubes, to isolate PBMC and extract DNA, and plasma samples.

DNA isolation
DNA from tissue samples and peripheral blood was carried out using the Nucleon BACC kit (GE Healthcare), following the manufacturer instructions. DNA from saliva was isolated with the kit Oragene DNA (DNAgenotek) and cleaned with the reagent prepIT L2P (DNAgenotek), following the manufacturer instructions.

| Genetic studies
DNA samples were then employed for the genotyping of four SNPs of the TGFB1 gene: rs1800468, rs1800469, rs1800470 and rs1800471. To assess the results obtained, three different approaches were used in all patients tested. Genotyping data were considered consistent when at least two of the three methods used yielded concordant results.

| PCR-RFLP
Primers and PCR conditions used for each SNP are shown in Table 2.
DNA samples were amplified and resolved in 2% agarose gels and, after ensuring the correct amplification, PCR products of the SNPs rs1800468, rs1800469 and rs1800470 were further digested with the enzymes HpyCH4IV, Bsu36I and PstI-HF (New England Biolabs), respectively.

| Taqman assay
Genotyping was also carried out by means of the allelic discrimi-

| Statistical analysis
The data of the SNPs sequencing were analysed with the software SNPStats, 22 This software allows to assess Hardy-Weinberg Equilibrium (exact test), chi-square test, OR estimation analysis of association between polymorphisms and disease applying logistic regression models, that consider the dominant, recessive and co-

| TGFB1 genetics in cancer susceptibility and progression
Hardy-Weinberg equilibrium was confirmed for each SNP and group of individuals included in the study and combinations thereof. Also, linkage disequilibrium was assessed between the following pairs of SNPs: rs1800468-rs1800469, rs1800468-rs1800470, rs1800469-rs1800470, rs1800469-rs1800471 and rs1800470-rs1800471 (data not shown).

Patients vs controls
Upon comparison with the control group, patients showed no significant differences in the frequency of the rs1800468, rs1800469 and rs1800470 polymorphisms (P > .05 in all instances, data not shown).
However, when we compared the rs1800471 SNP in our cohort of patients with a genetically related control group from the 1000 Genomes Project database, a significant increase in the frequency of the G/C genotype in the group of patients (20.7%) was observed compared to the control group (8.4%, P = .027, OR 2.84) (Table 3A), revealing that this polymorphism is involved in the pathogenesis of gastric cancer.
Comparison with the fifty-four healthy donors' allelic distribution did not showed significant differences, maybe because of the low number of samples in study.

Inter-patient comparisons (early stages vs disseminated disease)
Differences were also found in two of the SNPs studied (rs1800468  (Table 3B).
We then analysed the frequency of the different extended haplotypes formed by the combination of the SNPs studied. The ACTG haplotype (of the rs1800468, rs1800469, rs1800470 and rs1800471 SNPs, respectively) was present in 15.0% of the stage IV patients as compared to 3.0% of stage I, II and III patients (P = .02, OR 7.65, Figure 1).
Because of the low number of patients bearing the rs1800468-G/A genotype and, thus, the ACTG haplotype, we decided to use only the rs1800469 and rs1800470 SNPs to conform the risk haplotype rs1800469-C/rs1800470-T (CT) to make subsequent statistics.

TGFB1 polymorphism and survival curves
To see whether TGFB1 polymorphisms had any effect on the patient's survival, we decided to test the SNPs rs1800469, rs1800470

| TGF-β1 production upon lymphocyte stimulation
Once the genetic profile of the TGFB1 gene was studied in our group of patients, we decided to test whether the polymorphisms analysed TA B L E 3 Polymorphisms were analysed according to all possible inheritance models for each SNP (but the SNPs rs1800468 and rs1800470, because of the absence of patients bearing the second homozygous genotype A/A and C/C, respectively). ᵃNo inheritance model because of the lack of homozygous genotypes (A/A in rs1800468 and C/C in rs1800471).

Inheritance Model Genotype
Significant results are highlighted in bold.

| TGF-β1 plasma levels, disease progression and survival
Next to measuring TGF-β1 production by PBMC from patients, we We then sought to determine a cut-off point above which we could predict the outcome of a given patient and found that patients with less than 35 ng/mL of TGF-β1 in plasma showed a significant lower survival rate (N = 12, 58%) than patients with levels above that cut-off (N = 18, 94%, Log-rank test P = .026) ( Figure 4B).
These results must be taken with caution because they have been obtained certainly with a reduced number of patients and need to be confirmed in a larger group.

| D ISCUSS I ON
The quest for markers able to pinpoint gastric adenocarcinoma patients prone to disseminated metastases is relevant. For clinicians facing a newly diagnosed patient, the use of such tools will be advantageous when establishing a therapeutic approach. In this sense, the study of the TGFB1 SNPs presented may be of major interest in clinical practice.

| Genetic studies
We found an association between rs1800471 polymorphisms and the development of gastric cancer. Particularly, the frequency of the G/C genotype is more than double in the group of patients with gastric cancer (20.7%) than in controls (8.4%; 1000 Genomes Database P = .027), suggesting that possessing the rs1800471-G/C genotype increases the risk of developing gastric cancer. Moreover, the C allele of this SNP had been linked with a decreased expression of the cytokine TGF-β1. 12,13 As other authors demonstrated, 23 Altogether, these results indicate that the genetic markers herein studied are involved in the progression to advanced stages of the disease, as occurs in other epithelial type malignancies (colorectal cancer). 25,26 Research on the relationship between the TGFB1 gene polymorphisms and the development of gastric cancer focused mainly on disease susceptibility in case-control studies, comparing these SNP frequencies between a cohort of patients and a control group. 27,28 Normally, no group subdivisions were carried out according to the stage of the patients and, if they were, stage I and II patients were taken as a group and stage III or IV patients as another group.
However, we believe that a division according to the absence (stage I, II and III) or presence (stage IV) of advanced disseminated disease is more sound clinically and in terms of prognosis, 29 although it may substantially limit the number of patients available to carry out inter-group comparisons.
As previously mentioned, these single base polymorphisms are In conclusion, analysis of TGFB1 SNPs identified individuals at risk of developing gastric cancer, presenting with a more aggressive form of disease (stage IV) and reduced life expectancy.

| TGFB1 markers and survival
We describe a TGFB1 haplotype linked to life expectancy. Thus, survival of stage I, II and III patients bearing the haplotype rs1800469-C/rs1800470-T is significantly lower than that of patients bearing the alternative haplotype ( Figure 3C).
A possible explanation is that in patients bearing this haplotype in homozygosis, the levels of TGF-β1 synthesized, known to affect tumour progression, could be lower than in individuals with the alternative haplotype, who are thus protected against the progression of the tumour.

| Cytokine studies
By performing PBMC stimulation of carefully rs1800468-A, rs1800469-C and rs1800470-T genotyped patients, we found that carriers of these polymorphisms were less able to produce TGF-β1, in keeping with published data. It is then conceivable that this TGF-β1 production may occur in the tumour environment in patients with the above-mentioned alleles and the combined haplotype thereof ( Figure 2).
These data lend further support to the notion that the disseminated disease-linked genotype and haplotype leads to lower TGF-β1 expression and concomitant faster progression of the disease; this haplotype is thus a genetic signature of patients at risk of developing a more severe form of the disease.
In line with the idea above, we went on to measure plasma TGF-β1 levels (obtained before treatment and resection of the tumour) in a group of patients. Lower levels were found in patients with a survival time lower than 5 years (28.34 ng/mL ± 8.60, N = 6) compared to patients above this time point (48.34 ng/mL ± 4.89, N = 24, P = .03) ( Figure 4A) and, in our hands, 35 ng/mL plasma TGF-β1 is and adequate cut-off point to assess survival of patients ( Figure 4B).
Taken together, these data support the notion that TGF-β1 acts as a tumour suppressor in this malignancy and that high levels of this cytokine limit the growth and aggressiveness of this tumour.
In summary, we report here that metastases-prone gastric adenocarcinoma patients show higher frequency of TGFB1 SNPs associated with lower TFG-β1 production. These SNPs could help identify patients at risk of developing advanced disseminated disease and lower life expectancy, requiring a more aggressive surgical and therapeutic approach early on. In addition, this finding could also explain the proposed dual role of this cytokine on tumour progression: in the initial stages of the disease, low amounts would render individuals more susceptible to an aggressive evolution of the disease. In keeping with this notion, we found that TFG-β1 plasma levels were lower in patients with reduced life expectancy.
Therefore, we suggest the analysis of SNPs rs1800468, rs1800469 and rs1800470 and TGF-β1 plasma levels, as markers to identify gastric cancer patients with poor prognosis.

| Human rights statement and informed consent
All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation F I G U R E 3 TGF-β1 production and TGFB1 polymorphisms. PBMC from genotyped individuals was stimulated, and TGF-β1 production was measured by ELISA. A significant difference in TGF-β1 production was found between individuals rs1800469 C/C and C/T or T/T SNPs (C/C 268.4 pg/mL vs C/T 387.1 pg/mL, P = .04 or T/T 380.5 pg/mL, P = .03), as well as between rs1800470 T/T and C/C individuals (T/T 270.4 pg/mL vs C/C homozygous patients 365.8 pg/mL P = .03). When comparing Haplotypes, individuals bearing the rs1800469-C/ rs1800470-T haplotype in homozygosis showed a significant lower expression of TGF-β1 than individuals with the alternative haplotype rs1800468-T/rs1800470-C (281.6 pg/mL vs 380.5 pg/mL, P = .04) (institutional and national) and with the Helsinki Declaration of 1964 and later versions. Informed consent to be included in the study, or the equivalent, was obtained from all patients.
Association of TGFB1 polymorphisms to the absence (stages I, II and III) or presence (stage IV) of metastasis in gastric adenocarcinoma patients: SNP rs1800468-G/A showed a significant augment in patients with metastasis (type IV) compared to non-metastatic ones (stage I, II and III, OR 3.71 P = .04). Remarkably, no rs1800469-T/T individuals were found in stage IV patients, a genotype associated with a higher activity of the TGFB1 promoter, and either the codominant or recessive model showed a significant difference (P = .03 and .008, respectively) in the distribution of this SNP between metastatic and non-metastatic patients.

ACK N OWLED G EM ENTS
This work was supported by grants from Instituto de Salud Carlos III (PI12/01663 and PI18/00626), with funds from the European Union (Fondo Europeo de Desarrollo Regional FEDER). IJ is a grant recipient of a Universidad Complutense de Madrid-Real Colegio Complutense Harvard grant, (Ayudas para contratos predoctorales de personal investigador en formación CT18/16).

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest. Resources; Visualization.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

R E FE R E N C E S F I G U R E 4
Plasma TGF-β1 levels and survival rate. A, TGF-β1 levels in plasma are lower in patients with a reduced survival time: Patients with a survival time below 5 y have a significant decrease of plasmatic levels of TGF-β1 (28.3 ± 8.6 N = 6) compared to patients above that cut-off 48.3 ± 4.9 N = 24, P = .03). B, Survival curves of patients with different levels of TGF-β1 (TGF-β1 > 35 ng/ mL vs TGF-β1 < 35 ng/mL). Log-rank test detects a statistically significant difference in the survival rate of individuals with less than 35 ng/mL (58%) and patients with levels above this cut-off (94%, log-rank test P = .026). Kaplan-Meier method was used to estimate the 5-y survival function. Stage I, II and IV patients have been included in this analysis. No stage III patients were available