Monocarboxylate Transporter 1 (MCT1) is an independent prognostic biomarker in endometrial cancer

Ethical approval for this study was obtained from NRES Committee London - Fulham (reference 12/LO/0364). It was conducted in accordance with the conditions and principles outlined in the EU Directive 2001/20/EC and Good Clinical Practice, including the Data Protection Act 1998. Tumour tissues were obtained from 90 sequential EC patients who underwent hysterectomy (between 2011 and 2013) and donated their tumours for future research at St Mary’s Hospital, Manchester, UK. All participants provided written, informed consent for their clinical data and tumour samples to be stored anonymously and used for future research. Their paraffin-embedded tumour specimens were retrieved from the pathology archives and cut into 4-μm serial sections for immunohistochemical analysis. The median age of the cohort was 67 years (IQR range 57.7-74 years) and there were 47 endometrioid (EEC) and 43 non-endometrioid (Non-EEC) tumours. The non-EEC group comprised tumours of carcinosarcoma (n = 11), clear-cell (n = 9), serous (n = 6), mixed (n = 16) and undifferentiated (n = 1) histology. Mixed tumours were clear cell/high grade endometrioid (n = 13), clear cell/ high grade serous (n = 2) or high grade serous/ endometrioid (n = 1). Clinico-pathological data were obtained from patients’ hospital medical records, original pathology results and death certificates, where appropriate, including age at diagnosis, Body Mass Index (BMI), tumour histological type, grade, Federation of Gynecology and Obstetrics (FIGO) 2009 stage, tumour size, lymphovascular space involvement (LVSI), depth of myometrial invasion, last follow-up date, date of recurrence, type of recurrence, death and cause of death (Table 1). The average follow up time was 32.4 months and there were 28 recurrences and 22 deaths, of which 12 were EC-specific.

All IHC was performed using a fully automated IHC platform Leica BOND-MAX together with Bond™ Polymer Refine Detection kit (DS9800) and on-board retrieval system. The detection kit is based on a biotin-free, polymeric horseradish peroxidase (HRP)-linker antibody conjugate system for the detection of tissue-bound mouse and rabbit IgG and some mouse IgM primary antibodies. It is intended for staining sections of formalin-fixed, paraffin-embedded tissue on the Bond™ automated system. The sections were quenched using hydrogen peroxide and subjected to MCT1 (Santa Cruz, sc-365,501, 1:100 dilution), MCT4 (Santa Cruz, sc-376,140, 1:100 dilution) and CD147 (Fitzgerald Industries International, Clone UM-8D6, 1:100 dilution) primary antibody according to standard validated Protocol F written by Leica. Negative (isotype) controls were used for quality assurance. Following addition of post primary reagents the primary antibody was refined and visualized using the substrate chromogen, 3,3′-Diaminobenzidine tetrahydrochloride hydrate (DAB) via a brown precipitate. Tissue sections were counterstained with haematoxylin to visualize cell nuclei and were permanently mounted.

Immunoreaction in sections was evaluated for whole cell staining in whole tumour sections. A semi-quantitative scoring system was devised to take into account differences in staining intensity and distribution in different tumours. The observed staining intensities were given a score 0–3, from zero intensity staining (0) to high intensity staining (3), each staining intensity was multiplied by the percentage of tumour cells staining positively at each intensity. This gave a score range of 0-300. The tumours were separated into two groups as ‘low expression <200’ and ‘high expression ≥200’. Immunohistochemical evaluation was performed by two blinded independent observers and discrepancies were settled by consensus to determine the final score.

All data was stored and analysed using SPSS statistical software (version 22). Comparisons between staining positivity and clinico-pathological characteristics were made using Chi- square (χ²), Fishers exact test (as appropriate) or Mann Whitney U test (for continuous variables). Pearson correlation coefficients were calculated to assess the association between metabolic markers and/or clinico-pathological characteristics. Overall survival (OS), cancer specific survival (CSS) and recurrence-free survival (RFS) were analysed using Kaplan-Meier curves and compared using the Log-rank (Mantel-Cox) tests. Analyses adjusted for previously identified prognostic factors were performed using Cox proportional hazards models using a large covariate set (grade, FIGO 2009 stage, tumour size, LVSI and depth of myometrial invasion). Further, given the low event number a simpler model adjusted for grade and stage was also fitted. As CSS and RFS endpoints had fewer events a model which just adjusted for grade as a stratification variable was used for these endpoints. In all statistical analyses threshold for significance was p < 0.05.

In this study, immunohistochemical evaluation of MCT1, MCT4 and CD147 was performed in 90 EC (47 endometrioid [EEC] and 43 non-endometrioid [non-EEC]) tumours. All aforementioned markers were expressed at varying levels. Representative images of tumour sections stained for MCT1, MCT4 and CD147 are shown in Fig. 1. MCT1 was expressed in the cytoplasm, the plasma membrane or in both locations. MCT4 and CD147 were always expressed in the plasma membrane accompanied by some degree of cytoplasmic staining (Fig. 1).

MCT1 and MCT4 staining patterns were highly variable both within and between tumours. MCT1 was generally observed in peripheral zones (Fig. 2a) whereas MCT4 was expressed in central zones of the same tumour (Fig. 2b). Strong CD147 expression was mainly observed in those areas that were heavily stained with MCT1 (Fig. 2c). In contrast, in areas with strong MCT4 expression CD147 expression was weaker (Fig. 2c). Stromal and myometrial expression of MCT1 and MCT4 was also seen (Fig. 2d-i). In some, stromal and glandular staining displayed high contrast; malignant glands showed strong MCT1 expression whilst stromal staining was negative (or mild) or vice versa. In others, MCT1 and MCT4 expression was similar across stromal and glandular tumour compartments. Although the nucleus is not a usual location for MCT1 (based on current knowledge of its function), nuclear MCT1 expression was present in more than half of the tumours (present 58.9%, absent 40.1%; Fig. 2j-k).

There was a wide range in the intensity and distribution of staining, which was scored as described in the methods section. Briefly, tumours were recorded as high expressers when their score was equal to or exceeded 200; tumours with scores below this were regarded as low expressers. Interestingly, for MCT1, using the score of 200 resulted in 53.9% of tumours scoring low and 46.1% of tumours scoring high. Using the same 200 score cut off high MCT4 expression was observed in 32.2% of tumours and high levels of CD147 expression were observed in 64.4% of the tumours. There was a significant association between expression of CD147 and MCT1 (p = 0.003) but not MCT4 (p = 0.207) in our cohort (Pearson correlation coefficients = 0.32 and 0.14, respectively). There was no significant association between MCT1 and MCT4 expression (p = 0.93).

There were no significant associations between the clinico-pathological characteristics of the tumours and expression levels of either MCT1 or MCT4 (Table 1). CD147 expression was significantly associated with FIGO 2009 stage (Pearson Correlation Coefficient = 0.24), although this association would not be considered significant if allowance was made for the number of characteristics tested.

When the expression of these three metabolic markers was correlated with patient survival parameters (recurrence free, cancer specific and overall survival), MCT1 was identified as a prognostic marker in EC. Using an unadjusted Log-rank test, patients with high MCT1 expression showed reduced recurrence free and cancer specific survival and a significantly reduced overall survival (Fig. 3a-c, respectively).

Amongst survival outcome covariates (grade, FIGO stage, tumour size, LVSI and depth of myometrial invasion), grade and FIGO 2009 stage were the most significant prognostic factors in our study (Table 2). In order to evaluate whether MCT1 expression was an independent prognostic factor, MCT1 expression was tested in a large adjusted model with all of the listed covariates and a simpler model with just the major predictors, grade and stage (Table 2). In these analyses, MCT1 remained statistically significant after adjustment for all six possible covariates as well as in the simpler stage and grade-adjusted model (Table 2).

Due to the limited number of events observed for recurrence free and cancer specific survival, full adjustment for covariates was not possible for these endpoints. Therefore, we used a simpler adjustment for the most important predictor of survival in our cohort, which was grade, and fitted this as a stratified Log-rank test and Cox regression model.

When the overall cohort was stratified according to grade (I/II vs III), similar to the unadjusted model, an increased trend towards reduced recurrence free, cancer specific and overall survival was observed in patients with high MCT1 expressing grade III (mainly composed of Non-EEC) tumours (Fig. 3d-f). This effect reached statistical significance for overall survival (Table 3). There were too few events in patients with grade I/II tumours for meaningful statistical analysis (only composed of EEC tumours, shown in Additional file 1: Figure S1). Grade stratified Cox proportional hazard analysis performed on other markers (CD147 and MCT4) showed an increased risk of earlier time to event for patients with high CD147 expression but this effect did not reach statistical significance. There were no significant associations between these two metabolic markers and any of the survival parameters evaluated in this study (Table 3).