Toll-like receptor 4 single-nucleotide polymorphisms Asp299Gly and Thr399Ile in head and neck squamous cell carcinomas

Tissue specimens of 188 consecutive HNSCC were collected by the Department of Pathology, University hospital Essen, Germany. All patients were diagnosed and treated at the Department of Otorhinolaryngology, University Hospital Essen, Germany (1995-2002); treatment decisions were based on consensus recommendations from oncologists, radiotherapists and head and neck surgeons, which were based on treatment guidelines of treatment at the time. All patients gave written informed consent for research use of the tissues and for participating in the research project. The study was conducted according to the Declaration of Helsinki. Tissues were obtained during diagnostic or therapeutic surgery.

Overall, ninety nine (53%) patients received cisplatin/5-fluorouracil-based chemotherapy regimens and radiation up to 70 Gy as adjuvant therapy after surgery. Seventeen (9%) patients received primary radio-chemotherapy. Follow-up was performed regularly; median follow-up in patients still alive at analysis was 50 months (range, 0 to 128 months). Relapse data were available for all patients: 60 (32%) experienced disease recurrence and 89 (47%) death. Complete therapeutic regimens are listed in Table 1 and 2.

Due to poor or lack of sufficient material for PCR or IHC or absence of complete clinicopathological data, the initial sample of 188 patients of the total collective was split into three groups: a group of 138 for analysis of TLR4 Asp299, a group of 62 for analysis of TLR4 Thr399 (39 patients were analyzed for both SNPs), and a group of 78 patients with HNSCC for TLR4 expression analysis (43/78 were also genotyped for TLR4 Asp299; 20/78 for TLR4 Thr399 - see Table 3).

Routinely formalin-fixed and paraffin-embedded tumor tissue blocks were retrieved from the files of the Institute of Pathology (University Hospital of Essen, Germany) and processed using the tissue microarray (TMA) technology. In short, tumor tissue cores of 3 mm in diameter were removed from the area of interest from each donor block using a hollow needle skin biopsy punch (PFM, Cologne, Germany) and inserted into recipient blocks in a precisely spaced, array pattern. One tissue core of each normal thyroid and kidney tissues in preset position in each block served as control tissue and helped with the orientation.

5 μm TMA sections were cut and mounted on SuperFrost^® Plus slides (Menzel, Braunschweig, Germany). IHC was performed using the Dako Autostainer Plus System (DakoCytomation, Carpinteria, CA, USA). After antigen retrieval (water bath at 95°C; 20 min in citrate buffer), TMA slides were immunostained by the TLR4 (H-80) rabbit polyclonal antibody (sc-10741, dilution 1:100, Santa Cruz Biotechnology Inc., Sant Cruz, CA, USA). Antibody visualisation was performed using the anti-mouse IgG detection kit (EnVision+, DakoCytomation, Carpinteria, CA, USA) according to the manufacturer's recommendations.

Stained sections were reviewed by one of the authors (AB). The percentage of tumor cells showing a positive membranous/cytoplasmatic staining and the intensity of staining were assessed. Cases with complete lack of staining were scored as negative, a weak membranous/cytoplasmic reaction in 1-50% was classified as 1+, moderately strong reactions in up to 80% of tumor cells were scored 2+, whereas moderate to strong membranous/cytoplasmic immunostaining of > 80% of tumor cells were classified as 3+ (Figure 1). Inherent positivity of capillary endothelial cells and mononuclear inflammatory cells in the stroma served as positive control; for negative control purposes the incubation step with the primary antibody was omitted.

As described earlier [17], DNA samples were extracted from 10- μm sections of formalin-fixed, paraffin-embedded tumor tissue. The germline mutations TLR4 Asp299Gly (rs4986790) and Thr399Ile (rs4986791) were analyzed in all patients using polymerase chain reaction restriction fragment length polymorphism (PCR-RFLP). For rs4986790 (TLR4 8552A > G), PCR was performed with the forward primer 5'-CTG CTC TAG AGG GCC TGT G-3' and the reverse primer 5'-TTC AAT AGT CAC ACT CAC CAG-3', resulting in a 140 bp fragment. After denaturation at 95°C, 38 cycles of DNA amplification were performed using Taq DNA Polymerase 2× Master Mix RED (Ampliqon-Biomol, Hamburg, Germany) at 95°C for 30 s, 61°C for 30 s and 72°C for 30 s. Digestion with BccI at 37°C (New England Biolabs Inc., Ipswich, MA, USA) and results in fragments of 77 bp and 63 bp for the G-allele vs. 140 bp for the A-allele (no digestion) separated on a 2.5% agarose gel were analysed. To genotype for rs4986791 (TLR4 8852C > T), PCR was performed with the forward primer 5'-CTA CCA AGC CTT GAG TTT CTA G-3' and the reverse primer 5'-AAG CTC AGA TCT AAA TAC CT-3'. After denaturation at 95°C, 38 cycles of DNA amplification were performed using Taq DNA Polymerase 2× Master Mix RED (Ampliqon-Biomol, Hamburg, Germany) at 95°C for 30 s, 53°C for 30 s, and 72°C for 30 s. The resulting 110 bp PCR products were digested using the restriction enzyme BslI at 55°C and analyzed on a 2.5% agarose gel. The unrestricted products represent the TT genotype; the completely restricted products (89 and 21 bp) represent the CC genotype.

Electrophoresis was performed using SYBR Safe^® DNA Gel Stain (Invitrogen Corporation, Carlsbad, CA, USA) for visualization under UV light. Correctness of genotyping has been ensured by concomitantly analyzing DNA samples from human volunteers whose genotypes have already been confirmed by direct sequencing. Re-genotyping of both polymorphisms in 40 randomly chosen samples revealed complete concordance with previous results.

While the TLR4 Asp299Gly genotype was evaluable in 138 patients, the TLR4 Thr399Ile genotype was only evaluable in 62 patients. This was due to a low amount of and strongly degraded DNA in the available paraffin-embedded tumor tissue probably because of unbuffered paraffin on the tumor cells in more than 10 years old paraffin-embedded tissue samples or a high guanine-cytosine content in the gene region for Thr399, which hampers amplification. Therefore every sample was tested four times but utilizable DNA-products were available only for those 62 patients. Due to the reduced quality of samples other methods for genotyping (e.g. direct sequencing, pyrosequencing or TaqMan-genotyping) were not considered.

The two genotype distributions were tested for deviations from Hardy Weinberg equilibrium (both two-sided exact p-values were 1.0). Associations between clinical tumor characteristics and TLR4 genotype were assessed either by non-parametric Wilcoxon-Mann-Whitney tests in case of quantitative variables or by generalized Fisher's exact test for categorical variables in 2 × m tables. Time to events was calculated as the difference between primary diagnosis and either the date of the clinical assessment where the respective event occurred or last clinical assessment in case of censoring. While survival probabilities were graphically assessed by the Kaplan Meier method (including a log-rank test for inference in the figures), uni- and multiple cox regression analyses were used for the statistical analyses. In the multiple regression model variables with p > .1 in the univariate model were excluded to address estimation concerns. Model diagnostic of the proportional hazards (PH) assumption for the TLR4 genotypes comprised both graphical and formal investigations - none of which indicated strong evidence for a deviation from the PH assumption. Confidence intervals were calculated with coverage of 95% level (95%CI) and accordingly the level α for each test was 0.05 (two-sided). Unless otherwise mentioned, all reported p-values are nominal and two-sided.

In the present primary HNSCC cohort, 125 patients (90.6%) showed a homozygous TLR4 genotype for aspartate at aminoacid location 299, and 13 patients (9.4%) had a TLR4 Asp299Gly variant (minor allele frequency (MAF) ~4.7%). We observed no evidence for a deviation from Hardy-Weinberg equilibrium (HWE; p = 1.0; two-sided exact test). The genotype distribution is in accordance with previous reports [13, 15], which describe a carrier frequency of ~7% in both healthy controls and gastric cancer patients of the Caucasian population. Regarding the other SNP (Thr399Ile) 51 out of 62 (82.3%) of our patients were homozygous for threonine and 11 heterozygous (17.7%) for threonine and isoleucine alleles (MAF ~ 8.9%; p = 1.0; two-sided exact test for deviations from HWE).

No evidence for associations was found between clinical tumor characteristics or histopathological characteristics and TLR4 Asp299Gly genotype (Table 1). For the TLR4 Thr399Thr genotype the explorative statistical analysis indicated a positive correlation between AJCC tumor stage and Thr399Thr genotype only (p < 0.01; Table 2).

Sixteen percent of HNSCC tumors showed low (score 1+), 49% moderate (2+), 9% strong (3+), and 26% showed no TLR4 staining (Figure 1; Table 3). TLR4 staining (all scores) showed a diffuse and fine granular cytoplasmatic pattern. Distinct membrane staining was observed in some tumors but never without cytoplasmatic staining. TLR4 scores did not significantly correlate with clinicopathologic variables, in particular there was no correlation between TLR4 expression patterns and disease-free or overall survival (data not shown).

TLR4 genotype showed no evidence for an association with TLR4 protein expression phenotype (IHC; Table 3). Altered grouping of the expression values (low/high for grade 0/1 or 2/3) or TLR4 genotype (wild-type for both SNPs vs. any heterozygous variant) had no impact on this observation.

Our analysis revealed a significant association between TLR4 Asp299Gly genotype and recurrence of disease with a hazard ratio (hr) of 2.37 for a reduced disease-free survival (DFS; 95%CI: 1.05-5.33; p = 0.04; Figure 2A). Also, overall survival (OS) was significantly associated with Asp299Gly genotype with a hazard ratio of 2.00 for reduced survival (OS; 95%CI: 1.02-3.92; p = 0.04; Figure 2B; Table 4).

For the other SNP a similar pattern was observable (Figure 3); in case of DFS patients with the Thr399Ile variant displayed a significantly higher risk for disease advancement (hr = 4.97; 95%CI: 2.00-12.37; p = 0.0006; Figure 3B).

Next, we considered clinicopathological variables (age, sex, smoking, AJCC stage) in univariate cox models for overall survival. Afterwards we jointly included clinicopathological variables in addition to TLR4 Asp299Gly genotype status in a multivariable cox model (Table 4). Though a similar result pattern was observed for the TLR4 Thr399Ile variant, we decided to limit the displayed analyses to TLR4 Asp299Gly due to the too small sample size for the Thr399Ile variant. Even after correcting for clinicopathological variables TLR4 Asp299Gly genotype status was an independent prognostic factor of overall survival with a hazard ratio of 2.02 for reduced survival (95%CI: 1.01-4.06; p = 0.05; Table 4).

Based on the observed correlation of TLR4 genotype and applied primary therapy (Table 1 and 2), we also explored the additional impact of the use of adjuvant systemic therapy in the survival analysis (as main and interaction effect with TLR4 Asp299 genotype in the multivariate model of Table 4). According to this analysis, the interaction term indicated no evidence for an interaction (p = 0.18) which most likely reflects that the sample was statistically underpowered to detect an interaction. Displaying the relationship between TLR4 Asp299Gly genotype, use of adjuvant systemic therapy and course of disease graphically, we observed no evidence for significant survival differences between TLR4 genotypes in patients without adjuvant systemic therapy. However, with adjuvant systemic therapy, patients with wild-type genotype showed significantly longer DFS (p = 0.004 by log-rank test; Figure 4).