Limitations in the current tumour classification and grading for NEN subtypes have highlighted the lack of predictive and prognostic markers [12]. The routinely used immunochemical markers CgA, synaptophysin and CD56 confirm a neuroendocrine diagnosis and the proliferation marker Ki67 is used a key indicator for the grading of NENs [8]. These markers have some limitations including their lack of correlation to clinical outcomes and their presence in non-tumour tissue. The routinely used markers represent only a few of the molecules that characterise the tumour phenotype. Thus, additional markers are required to more comprehensively define NEN subtypes.
In this study we used quantitative real-time reverse transcription PCR (qRT-PCR) to measure the mRNA levels of twenty seven cancer markers associated with proliferation, metabolic activity, invasive potential and metastasis, in tissue from a patient with a G1 pancreatic NET. Comparison of mRNA marker levels from tumour tissue with adjacent non-tumour tissue from the same patient provided insights into the changes associated with acquisition of the cancer phenotype. The inclusion of matched controls enabled a comparative analysis, in comparison to many previous studies reporting tumour data only, which has limited value by itself when the status of the non-tumour tissue is unknown. Of the twenty seven markers, CgA had a transcript abundance that was 80 times higher in tumour relative to non-tumour. CD56 and p53 were 6.5-fold more abundant in tumour relative to non-tumour tissue and β -catenin, PDX1 and CK20 were 2–3 times as abundant in tumour than non-tumour tissue. The profile of the tumour tissue indicated an overall reduction in mRNA levels of the analysed markers relative to non-tumour tissue. Ki67, PCAD and CK7 expression was reduced up to 10-fold in the tumour relative to non-tumour with 2-9-fold reductions in CD24, CD44, CD31, CD49 MENA, ECAD, EPCAM, CDX2, CK6 and CK13. This suggests an overall loss of activity of these genes in the tumour. The most distinctive features of the tumour tissue were a gain in CgA, CD56, p53, PDX1, CK20 and a loss of Ki67, PCAD and CK7.
The observed increases in mRNA transcripts of the mesenchymal markers vimentin and β -catenin, decreased transcripts of the cell adhesion markers E-cadherin and P-cadherin and loss of differentiation markers CDX2, CK6, CK7 and CK13 are hallmarks of transition to a mesenchymal phenotype tumour cells. The epithelial to mesenchymal transition is a process by which normal epithelial cells lose their cell:cell and cell:matirix adhesion and acquire the migratory and invasive characteristics of a cancer cell [14].
This study indicates that qRT-PCR can distinguish phenotypic differences between tumour and non-tumour tissue. Advantages of qRT-PCR are that it is a semi-quantitative tool that generates numerical values for marker expression so comparisons can be made between different patients, as each marker is measured relative to a housekeeping gene. Other benefits are that it is quick and multiple markers can be tested in a single run. qRT-PCR is cheap, requiring only a few enzymes, buffers and primers. The same set of identical primers can be used in multiple labs thereby reducing variabilities between labs. A further advantage of qRT-PCR is that the RNA is extracted from the whole tissue and same extract used for all marker analyses. The extract is representative of the whole tissue rather than a portion of the tissue. PCR avoids the use of an antibody and the variability associated with antibodies to the same protein but from different sources.
qRT-PCR measures mRNA levels markers relative to a housekeeping gene or control gene. This is selected to be stably expressed under different experimental conditions and for its expression to remain constant in different tissues and not influenced by the presence of disease so the expression of other genes can be measured relative to it. β-actin and GAPDH are commonly used housekeeping genes as their expression has been found to remain constant under different conditions [15, 16]. However, studies indicate that some housekeeping genes could be better than the others for a particular sample, as the expression of these and various other housekeeping genes can be different in different sample cohort. Hence, Hence, we tested the β-actin and GAPDH housekeeping genes
and selected the one with least variation. We found β-actin to have less variation between tumour and non-tumour, compared to GAPDH for the pancreatic NET sample. Therefore, β-actin was selected as the housekeeping gene for analysis and future experiments.
To provide a profile of the different markers expressed in tumour and non-tumour tissue, we measured mRNA levels of markers in each of these tissues relative to each other using DAXX as a reference as this gene was expressed in similar levels in tumour and non-tumour. This data showing the pattern of expression of different markers in tumour and non-tumour tissue provides a profile that may have application in defining individual patients. CgA for example is seen to be significantly low in the normal tissue profile while it has a massive increase in the tumour tissue. Future studies are required to validate these data using a larger number of patients and determine whether such profiles have potential for stratification of G1 tumours. A larger cohort may establish the possibility to Identify markers that are evidently different between normal and tumour to draw out a reference profile for different grades of NETs.
We compared qRT-PCR results with IHC. Fifteen of the twenty seven markers showed the same expression trends between tumour and non-tumour tissue and twelve showed differences in trends between tumour and non-tumour tissue. The most notable differences between the qRT-PCR and the IHC results were Ki67, MENA, ECAD and the cluster of differentiation markers CD24, CD44 and CD31. Our data showed Ki67 was higher in tumour relative to non-tumour using IHC but qRT-PCR analysis indicated Ki67 mRNA was reduced in tumour tissue relative to the non-tumour. Previous studies comparing Ki67 protein with Ki67 mRNA levels indicated variable correlations amongst different tumour samples [17]. Variability in Ki67 expression has been attributed to changes in Ki67 levels with different stages of the cell cycle [18]. It is possible that the low levels of Ki67 found in our study may correspond with regions of the tumour containing non-cycling cells, given that IHC only utilises a small proportion of the total tumour mass. qRT-PCR for Ki67 was found to be more accurate than IHC in a breast cancer study in predicting patient response [19]. ECAD expression associated with breast cancer metastasis is understood to be regulated by cell surface integrin α3β1. α3β1 inversely affects the mRNA expression level of ECAD and this is counter balanced by the protein expression [20].
Our study showed that for four of the seven CD markers (CD24, CD44, CD31, CD14) mRNA and protein data did not correlate. This is consistent with a previous study where concordance between gene and protein expression for a range of CD antigens in normal and prostrate tumour samples was poor to moderate (Pearson correlations ranged from 0 to 0.63), attributed to low levels of protein expression, sample preparation as well as the real biological differences between protein and mRNA expression [21]. CD24 functions in cell adhesion and signaling, where high expression is associated with increased proliferation and invasion in pancreatic, colorectal and lung cancer but decreased proliferation and invasion in breast cancer cells [22]. The lack of direct correlation between mRNA and protein levels may be related to the complex regulatory pathways between CD24 transcription and translation. CD31 is an endothelial marker used as an indicator of blood vessels. Its expression may relate to the amount of vascular tissue which can vary considerably depending on the microscopic fields selected for IHC.
Synaptophysin was not detected by IHC and was only barely detectable by qRT-PCR, showing a slight increase in tumour relative to non-tumour. The levels of this transcript may not have been sufficient to produce detectable levels of protein. IHC staining for synaptophysin varies between different NENs and was reported as positive for approximately 60% of gastrointestinal neuroendocrine carcinomas [23]. EPCAM mRNA levels were lower in the tumour tissue than non-tumour and the IHC label was also lower in tumour relative to non-tumour however both tumour and non-tumour were categorised as 1–25% stained and not sufficiently different to be distinguishable. CD49 showed less mRNA in tumour than non-tumour. This pattern was also observed in IHC where staining was slightly less in tumour than non-tumour but both tumour and non-tumour showed < 25% positive staining.
IHC is the traditional method for tumour diagnosis, essential for establishing tissue morphology and providing information for grading of tumour biopsies. IHC provides information about morphology of tissue that cannot be obtained with PCR. IHC is a qualitative and relatively time-consuming procedure in which the scoring technique may dependent on the scorer. Small changes of a few percentage in the Ki67 index can affect the grading of the tumour which itself may have an influence on the type of treatment regime chosen. While the new WHO NEN guidelines have improved the stringency of IHC methods [8], the heterogeneity of tissue, differences in preparation of tissue, differences in counting approaches (automated versus manual) as well as variations in the sensitivity and specificity of antibodies adversely affects the prognostic potential of IHC [24]. IHC requires further development to become an immunoassay, not simply a stain [25].