INTRODUCTION

Tumor necrosis factor (TNF) is a proinflammatory cytokine, which plays a significant role in the pathogenesis of a number of diseases, including oncological and autoimmune diseases [1, 2]. The TNF gene is located on chromosome 6 (6p21.3) in a highly polymorphic region of class III HLA histocompatibility complex at a distance 250 kb from the class I HLA-B locus and 850 kb from the class II HLA-DR locus. The TNF gene has a number of functional polymorphisms, of which single nucleotide substitutions in the promoter regions –308(g/a)TNF (rs1800629) and –238(G/A)TNF (rs361525) are most studied [3] (for convenience, ‒308(g/a)TNF and –238(G/A)TNF alleles are designated by lowercase and uppercase letters, respectively). It is known that these polymorphisms are representatives of the extended ancestral haplotypes covering the whole complex of HLA genes. One of the most common haplotypes, AH8.1 (which is found in 9% of cases in the European population), which includes the HLA-A*1, HLA-B*8, HLA-DRB1*3, and –308(a)TNF alleles, is associated with an increased level of TNF synthesis and predisposition to autoimmune diseases [4, 5]. In turn, the –238(A)TNF allele is associated with a low production of the cytokine and forms a unique B57 haplotype with the HLA-B*5701 allele, which is known to be a protective allele in the pathogenesis of HIV-1 infection [6]. It has been suggested that the effect of HLA complex haplotypes can be associated with the level of TNF gene expression; however, on the contrary, the effect of TNF can be caused by a linkage with the HLA complex genes.

TNF can perform multidirectional functions in the pathogenesis of oncological diseases: inducing both apoptosis and proliferation of tumor cells and stimulating tumor infiltration by cells of the immune system and promote the invasion and metastasis of tumor cells [1].

Breast cancer (BC) is the most common oncological disease in women. The antitumor activity of TNF with BC is actively studied in preclinical and clinical studies. Although the clinical use of TNF is limited by systemic toxicity, its use as an adjuvant in combination with in chemo- and radiation therapy is considered promising [7]. However, the effect of the microenvironment and cellular and genomic context on the final effects of TNF should be taken into consideration. The TNF gene polymorphism is studied as a factor of risk and prognosis of BC. Previously, we demonstrated the dependence of overall survival (OS) of BC patients on –308(g/a)TNF and –238(G/A)TNF polymorphisms [8, 9]. In the present study, we tried to answer the question of whether the effect of TNF gene polymorphism on OS is a consequence of cooperation with the gene environment, namely, with the ancestral haplotypes of HLA complex involved in autoimmune processes.

MATERIALS AND METHODS

An archival collection of DNA from 442 primary BC patients with the confirmed histological diagnosis and 327 women without oncological and autoimmune diseases (clinical control group) with known –238(G/A)TNF and –308(g/a)TNF genotypes (partially described previously [8–10]) was used in the work. The mean age of BC patients was 54.2 years (from 23 to 80 years); in the control group, the mean age was 53.7 years (from 19 to 89 years). Data on the stage of the disease were available for 420 BC patients. Since only 2.9% of patients had stage IV BC, these patients were included in the group with stage III BC, and the distribution by stages of the disease was in general as follows: I, 25.0%; II, 46.4%; and III, 28.6%. Information about the condition of BC patients during the observation period since the initial hospitalization was obtained in 278 cases.

The HLA-A*1, HLA-B*8, and HLA-DRB1*03 alleles were determined by the polymerase chain reaction (PCR) method in 412 BC patients; HLA-B*57, in 169 BC patients. The following allele-specific primers were used in the reaction: A1_F: 5'-ggACCAggAgACACggAATA-3', A1_R: 5'-AggTATCTgCggAgCCCg-3'; B8_F: 5'-gACCggAACACACAgATCTT-3', B8_R: 5'-CCgCgCgCTCCAgCgTg-3' [11]; B57_F: 5'-gCTCACATCATCCAggT-3', B57_R: 5'-CgTCTCCTTCCCgTTCTC-3'; B57_ R0101: 5'-ATCCTTgCCgTCgTAggCgg-3', B57_R0102: 5'-ATCCTTgCCgTCgTAGGCAG-3', [12]; DR3_F: 5′-TACTTCCATAAC CaggAggAgA-3', DR3_R: 5′-TgCAgTAgTTgTCCACCCg-3' [13]. The alleles of –238(G/A)TNF were determined by a PCR-RFLP method as described previously [9].

To determine the TNF gene –308/–238 haplotype in samples heterozygous for both studied sites (–308ag/–238AG genotype), a PCR using allele-specific primers was conducted to determine the –308(g/a)TNF polymorphism: TNF-308aF 5'-AATAggTTTTgAggggCATgA-3', TNF-308gF 5'-ATAggTTTTgAggggCATgg-3', TNF-308R 5'-TCTCggTTTCTTCTCCATCg-3' [14], since this region of PCR product includes the position of the –238(G/A)TNF polymorphism (Fig. 1a). The nucleotide sequence of the PCR product of each haplotype was determined by a reverse Sanger sequencing.

Fig. 1.
figure 1

Results of sequencing of the TNF gene double heterozygous –308ag/–238AG genotype. (a) Sequence of PCR product using allele-specific TNF-308a/gF and TNF-308R primers [14] (underlined) includes the sites of single nucleotide substitutions –308(g/a)TNF and –238(G/A)TNF (highlighted in bold). (b) Fragment of PCR product sequence: above, TNF gene –308a/–238G haplotype; below, –308g/–238A haplotype.

Full size image

When analyzing the data obtained, the groups were compared using a two-sided Fisher’s criterion. Pearson’s criterion was used to check the compliance of genotype distribution with Hardy–Weinberg equilibrium. The distribution of the TNF gene –238/–308 haplotypes was determined by a direct calculation. To estimate the risk of disease, the odds ratio was calculated and presented as OR (95% confidence interval, CI). The hazard ratio (HR) (95% CI) was a criterion for estimating the prognostic significance of the trait, the indices of 10-year OS was presented as Mean ± SE%, and the comparison of OS curves was conducted by a Kaplan–Meier method using a Logrank test in the GraphPad Prism program (version 4.00). For all criteria, differences were considered significant upon reaching p < 0.05.

RESULTS

Data on the TNF gene polymorphism were analyzed for 442 BC patients and 327 women from the control group. The distribution of –308(g/a)TNF and –238(G/A)TNF genotypes corresponded to theoretical Hardy–Weinberg distribution (p > 0.05). The distribution of genotypes and haplotypes for these sites had no statistical differences in BC patients and in the control group, and the obtained OR values do not allow to talk about the effect of these polymorphisms on predisposition to BC disease (Table 1).

Table 1.   Distribution of TNF gene genotypes and haplotypes in BC patients (n = 442) and control group (n = 327)
Full size table

It is necessary to note that all seven samples (three from the control group and four from BC patients) that were heterozygous both for the –238(G/A)TNF site and the –308(g/a)TNF site had only one polymorphic substitution in each of the haplotypes of its diploid set (Fig. 1b). Accordingly, the double heterozygous –308ag/–238AG genotype was taken into account in the analysis as containing two haplotypes: ‒308a/–238G and –308g/–238A (Table 1). Thus, except for these samples, the carriers of three TNF gene –308/–238 haplotypes (a/G, g/G, g/A) in fact correspond to the carriers of ag/GG, gg/GG, and gg/AG genotypes. The carriers of the aa/GG genotype (in our sample, 1.4% of BC patients) were included in the group ag/GG for subsequent analysis.

Four hundred and twelve BC patients were tested for the markers of extended AH8.1 haplotype (HLA-A*1, HLA-B*8, and HLA-DRB1*03 alleles) (Table 2). In 8.5% of cases, three markers were found simultaneously, which corresponds to literature data [15]. The HLA-B*57 was detected only in the –238A allele carriers, but not in the carriers of TNF gene –238GG homozygote (74.4 and 0%, respectively; p = 7.1D-22). In addition, samples positive for HLA-B*57, of which 96% (24 out of 25) had the HLA-B*5701 allele subtype, were additionally tested in nested PCR.

Table 2.   Distribution of marker alleles of AH8.1 and B57 haplotypes in BC patients with different TNF genotypes
Full size table

The distribution of marker AH8.1 alleles in the carriers of TNF gene –308/–238 genotypes (gg/GG, ag/GG, and gg/AG) is presented in Table 2. The group of ag/GG carriers differed significantly in the frequencies of the HLA-A*1, HLA-B*8, and HLA-DRB1*03 alleles from the carriers of both gg/GG and gg/AG genotypes, except for the HLA-A*1 in the gg/AG carriers (Table 2). A high frequency of the HLA-A*1 allele (30.3%) was registered for the –238AG genotype, which corresponds to literature data, since the ancestral B57 haplotype also includes the HLA-A*1 alleles along with the minor allele of –238(A/G)TNF polymorphism and the HLA-B*57 allele [15, 16]. It should be noted that, all of the four carriers of the TNF gene double heterozygous –308ag/–238AG genotype were positive for the HLA-B*57, while two of them had all three AH8.1 markers. Thus, as was expected, the –308ag genotype is associated with AH8.1 markers, while the TNF gene –238AG genotype is associated with B57 haplotype markers.

We considered OS of BC patients as a characteristic of the final response of polymorphic alleles. We found no statistically significant differences for the whole sample, and 10-year OS was 65.0 ± 10.6% for the gg/AG genotype, 73.6 ± 3.3% for the gg/GG and 75.0 ± 5.3% for the ag/G (Fig. 2a). For the –308gg/–238AG genotype carriers, a decrease in 10-year OS till 50.0 ± 17.7% was associated with the HLA-B*57, although this dependence was not statistically significant (Fig. 2b). No statistically significant effect of the AH8.1 haplotype on the OS of BC patients was found (Fig. 2c). Since the number of carriers of three marker alleles of the AH8.1 haplotype simultaneously was small (8.5% in our sample), groups for comparison were subsequently formed depending on the presence of at least one of the marker HLA-A*1, HLA-B*8, and HLA-DRB1*03 alleles or in the absence of these alleles (designated as M+ and M–, respectively) (Fig. 2d).

Fig. 2.
figure 2

Overall survival of BC patients depending on TNF gene –308/–238 genotypes and marker alleles of ancestral AH8.1 and B57 haplotypes. Overall survival curves according to Kaplan–Meier. Abscissa axis, observation time, months. Ordinate axis, overall survival, %. (a) TNF gene –308/–238ag/GG, gg/GG, gg/AG genotypes; (b) overall survival curves for carriers of TNF gene gg/AG genotype in the presence or absence of HLA-A*1 and HLA-B*57 alleles; (c) AH8.1+ contains three marker HLA-A*1, HLA-B*8, and HLA-DRB1*3 alleles simultaneously; and (d) M+ contains one or more marker alleles of AH8.1 haplotype.

Full size image

Subsequently, we tried to identify the effect of marker AH8.1 (M+) alleles on the OS of the carriers of the TNF gene gg/AG, gg/GG, and ag/GG genotypes (Fig. 3a). A significant difference was detected only for the ag/GG genotype (10-year OS was 87.2 ± 5.4% in the presence of at least one AH8.1 marker and 56.0 ± 9.9% in the absence of AH8.1 alleles; Logrank test p = 0.0052, HR = 3.8 [1.4; 11.1]) (Fig. 3a). For the gg/AG carriers, the 10-year OS had an even larger, but statistically insignificant, difference in the presence of marker AH8.1 alleles and in their absence (88.9 ± 10.5 and 44.4 ± 16.6%, respectively; Logrank test p = 0.064; HR = 5.9 [0.9; 22.9]) (Fig. 3a).

Fig. 3.
figure 3

Overall survival of BC patients depending on genotype and stage of the disease. (a, c, d) Overall survival curves according to Kaplan–Meier. Abscissa axis, observation time, months; ordinate axis, overall survival, %. (b) Dependence of the occurrence of genotypes on stage of BC. Abscissa axis, stage of BC; ordinate axis, occurrence, %. (a, b, c) Whole sample of BC patients. (d) BC patients with stage II. I, II, III, BC stages; ag/GG, gg/GG, gg/AG, TNF gene –308/–238 genotypes; M+ genotype contains one or more marker alleles of AH.8.1 haplotype.

Full size image

The occurrence of both marker alleles of AH8.1 haplotype and minor alleles of –238(G/A)TNF and ‒308(g/a)TNF polymorphisms did not depend on the stage of BC (Fig. 3b). Three groups of survival curves were noted when the BC stage was included in the analysis of OS dependence on TNF genotypes (Fig. 3c). The carriers of common gg/GG genotype at the stage II of the disease had a high 10-year OS (more than 80%) comparable with the OS of patients with stage I BC. On the contrary, the gg/AG genotype carriers in stage II BC had a low 10-year OS (less than 60%) comparable with stage III BC. The intermediate group included patients with stage II and III BC that are carriers of the ag/GG genotype. At the same time, the maximum difference was noted between the groups of carriers of three TNF genotypes at stage II of the disease (Logrank test p = 0.05; Logrank test for trend p = 0.016) (Fig. 3c).

When analyzing stage II BC, it was found that the 10-year OS of the TNF gene –308ag/–238GG carriers in the presence of at least one AH8.1 allele (M+) was significantly higher than in their absence (90.0 ± 6.7 and 46.2 ± 13.8%, respectively; Logrank test p = 0.0063; HR = 6.6 [1.7; 26.9]). At the same time, the OS curves came in two groups: the ag/GG genotype carriers in the presence of AH8.1 markers and gg/GG carriers regardless the AH8.1 markers had 10-year OS more than 80%; 10-year OS was lower than 50% in the ag/GG carriers in the absence of AH8.1 markers and in the gg/AG carriers (since the frequency of the gg/AG carriers was low, the size of this group did not allow the dependence of OS on AH8.1 markers to be analyzed in this case) (Fig. 3d).

DISCUSSION

The effect of TNF gene polymorphisms on predisposition to oncological diseases has been quite extensively studied. In a metaanalysis of literature data on the effect of –238(G/A)TNF on predisposition to BC (based on eight studies), no association of this polymorphism with BC was detected (including with ranking by ethnicity) [16]. For the –308(g/a)TNF polymorphism, no effect on predisposition to BC was found for the whole sample in a metaanalysis of 20 publications; however, their authors draw the conclusion that the –308a allele may serve as a protective factor in postmenopause, while the –308aa genotype may be a risk factor in premenopause [17]. In our study, we found no association with BC for both –238(G/A)TNF and –308(g/a)TNF genotypes and the TNF gene –308/–238 haplotypes (Table 1).

The effect of the TNF gene polymorphisms on the prognosis of BC is much less studied, while the published data are controversial. The association of the TNF gene –308aa homozygote with an overall and tumor-specific survival was detected [18], while other authors found no such connection for the –238(G/A)TNF and –308(g/a)TNF polymorphisms [19]. In other works, the minor allele of the –308(G/A)TNF polymorphism was associated with both overall and relapse-free survival [20, 21]. Previously, we demonstrated a statistically significant decrease of 5-year OS in the carriers of common –308gg genotype in BC patients with stage III [8] and in the carriers of the minor –238A allele at stage II of the disease [9]. These previously obtained data can be interpreted as a decrease in OS for the –308gg/–238GG and –308gg/–238AG genotype carriers in BC patients with stage III and for the –308gg/–238AG genotype carriers with stage II of the disease presented in this study (Fig. 3c).

When analyzing the dependence of OS curves on the TNF genotype and the stage of the disease, we categorized BC patients into three groups with high, low, and intermediate levels of OS (Fig. 3c), the total indices for which were 86.7 ± 2.9, 70.6 ± 6.4, and 46.3 ± 6.1%, respectively (p < 0.0001). Although the difference in the stages of the disease (a prognostic trait used in the clinic) makes the main contribution to such difference, a modulating effect of the TNF gene polymorphisms on OS (which is most pronounced in stage II of the disease) should be noted. It is possible that the conditions for the manifestation of functional differences of the TNF gene minor alleles are namely created at the stage II of BC. Such conditions can include an imbalance of the immune system with oncological disease. It is known that the expression of class I HLA molecules decreases with the progression of BC; at the same time, the expression of class II HLA appears in the tumor tissue. Class I HLA molecules play a central role in the cellular immune response as an antigen presenting molecules for cytotoxic T lymphocytes. A decrease in the expression of class I HLA up to complete elimination from the surface of tumor cells is a mechanism of the escape of neoplastic cells from antitumor immune surveillance, which leads to dissemination and metastasis and worsens the prognosis of the disease [22]. In turn, class II HLA molecules are required for the presentation of peptides to T helper cells and their expression is responsible for the initiation of the immune response. The presence of such molecules makes the tumor more immunogenic, which yields a good prognosis [23]. Since the AH8.1 haplotype is related to autoimmune processes [6], we supposed that it can affect the intensity of antitumor immune response and BC prognosis.

A maximum effect of the interaction of the TNF gene ag/GG genotype with marker AH8.1 alleles was registered for stage II BC (Fig. 3d), although it was detected for the whole sample. Regarding the gg/GG genotype (which is not affected by the AH8.1 alleles), the presence of marker AH8.1 alleles increases, while their absence decreases OS of the carriers of the TNF gene minor –308a and –238A allele in the total sample (Fig. 3a). However, only a decrease in OS for the ag/GG genotype in the absence of marker AH8.1 alleles can be observed for stage II BC. Thus, if the effect of the gg/GG genotype (presented in 67.9% of our sample of BC patients) is taken as a reference point, the effect of surrogate AH8.1 haplotype, which includes the TNF gene –308a allele and at least one of the HLA-A*1, HLA-B*8, and HLA-DRB1*3 alleles, will be close to it. It can be supposed that the ancestral AH8.1 haplotype maintains regulation and immune balance at the optimal (favorable) level under standard therapy of BC; at the same time, a combination of the TNF gene –308a and –238A alleles with other HLA haplotypes is apparently unfavorable for the prognosis of BC.

A trends toward an increase in OS in the presence of at least one AH8.1 marker was also detected for the carriers of the gg/AG genotype, probably due to the association of this genotype with the HLA-A*1 allele (Fig. 3a). However, in general, the gg/AG genotype carriers have a poor prognosis at stage II BC, apparently due to the association with the HLA-B*57 allele (Fig. 2b). The unique B57 haplotype is known and has been studied in connection with HIV-1 infection. It has been suggested that a low expression of the TNF cytokine in the –238 carriers can promote the protective effect of the HLA-B*5701 allele [24]. An increased level of a number of factors with antiviral activity, including those related to the family of APOBEC3 cytidine deaminases, was found in healthy uninfected HLA-B*57 carriers [25]. APOBEC3-associated mutagenesis has been registered in many tumors; at the same time, only the expression of APOBEC3B correlates with the proliferation of tumor cells, while the functions of other members of this family are related to processes in cells of the immune system [26]. It has been demonstrated that an increased level of APOBEC3B mRNA correlates with a poor prognosis of the disease [27, 28]. The HER2 amplification and the loss of PTEN suppressor gene activates APOBEC3B in vitro and correlates with APOBEC3-associated mutagenesis in vivo [29]. Previously, we found in our studies that the –238(G/A)TNF polymorphism is associated with activating allele HER2 Ile655Val and can be a risk factor of BC for –238AG genotype carriers [10]. Thus, it can be supposed that the increased risk of BC and a decrease in OS in the carriers of minor –238(G/A)TNF allele may be mediated by the association of the HLA-B*57 with the APOBEC3B.

A low frequency of the TNF gene minor –238A allele carriers and, appropriately, small groups of comparison are a limitation of this study. However, although we obtained no statistically significant evidence of the effect of –238(G/A)TNF polymorphism on the disease prognosis, the results seem biologically justified to us and require further detailed study. Apparently, this group of BC patients, as well as -308a allele carriers in the absence of marker AH8.1 alleles, need additional treatment, and the use of immunotherapeutic approaches can be efficient for these cases.

CONCLUSIONS

The results of our study demonstrate that the ‒308(g/a)TNF and –238(G/A)TNF polymorphisms do not affect predisposition to BC diseases, but both of them significantly decrease the OS of BC patients, especially at stage II of the disease. Apparently, the final effect of the TNF gene polymorphisms on the disease prognosis depends on the genomic context. Based on data obtained and the analysis of literature sources, it can be assumed that the mechanisms of the –308(g/a)TNF and –238(G/A)TNF effects differ. Thus, a decrease in OS in patients with the minor ‒238A allele can be mediated by its association with the HLA-B*57, while, on the contrary, a decrease in OS in the –308a carriers is not associated with the autoimmune AH8.1 haplotype. Thus, two genetically determined groups of BC patients having an unfavorable prognosis in conditions of standard antitumor therapy of BC were detected, which requires additional study.