Despite the years of routine use of lymph node dissection in breast cancer management, inspired by Halstedt , our understanding of the role of lymph nodes in the metastatic process is still marginal. The analysis of pre-selected subpopulations of cancer cells found in lymph nodes metastases could provide insights into biological background of cancer progression. Our work explored patterns of conversion in classical and EMT-related biomarker status between primary breast tumors and corresponding synchronous axillary lymph node metastases to determine whether phenotypic variability is associated with different clinical outcome.
Numerous studies have shown discordant expression of classical molecular markers, ER, PgR and HER2 between the PT and both lymph nodes and distant metastases [28–32]. Discordance in ER and PgR status has been demonstrated within the range of 10-32% and 34%–41% cases, respectively [28, 33–35] and 3%–24% for HER2 status [28, 32, 34–36] when PT were compared with metastatic relapse. Generally, lower discordance rates have been observed in LNM, as showed in individual studies [30, 31, 37] and confirmed in recent meta-analysis on HER2 status in primary and metastatic cancer, including 26 studies of 2520 subjects . However, there are also reports showing the opposite, as exemplified by the study of Aitken et al., who found different breast/node status of at least one receptor (ER, PgR or HER2) in almost 47% of cases .
Receptor discordance has been frequently associated with poor survival [32, 34, 35], what has been attributed to the inappropriate use of hormone and targeted therapy prescribed based on the characteristics of PT  or selection of tumors with more unstable phenotype and therefore more aggressive behavior . Adverse impact of receptor discordance was abolished, when treatment was modified according to the results of metastatic biopsy . In general, the cancer management scheme in 14-20% patients was changed on the basis of metastasis biopsy [28, 33, 39]. The growing body of evidence supports reassessment of ER, PgR and HER2 at the time of relapse diagnosis to tailor the most effective treatment for each patient at all times .
Until now, main research focus has been put on the discordance of ER, PgR and HER2 status between the PT and metastasis, as discussed above. Some studies examined cell proliferation, differentiation and apoptosis markers [30, 37, 41]. However, to the best of our knowledge there is no data available on the status of EMT regulators: TWIST1, SNAIL and SLUG in LNM compared to PT. The clinical outcome in relation to these biomarker status in lymph nodes has not yet been reported.
Our results show frequent conversion of TWIST1, SNAIL and SLUG status between PT and LNM, occurring both at mRNA and protein level (within the range of 28-53% of cases). Expression levels of EMT-TFs were significantly higher in LNM compared to PT. This could be explained by the higher frequency of pre-selected aggressive subpopulations of cancer cells resulting from EMT present in LNM than in the PT, where these aggressive cells constitute a minority of cells [16, 17] and therefore might not be easily captured. It seems that indeed molecular profile of LNM might be a surrogate marker of aggressive phenotype of the PT, as postulated .
We have found that both the elevated level of TFs in LNM and their negative-to-positive switch were associated with poor clinical outcome. Increased TWIST1 and SNAIL expression in LNM correlated with shorter OS (P = 0.04 and P = 0.02, respectively) and DFS (P = 0.02 and P = 0.01, respectively), whereas their expression in PT had no prognostic impact. Negative-to-positive conversion of SNAIL status also correlated with worse survival compared to unchanged status (OS: HR = 4.6; 1.1-18.7; P = 0.03; DFS: HR = 3.8; 1.0-48.7; P = 0.05). Numerous studies show poor prognostic impact of increased expression of TWIST1, SNAIL and SLUG in PT [12–14, 42–44]. The contradictory results have also been reported . But no study have examined their clinical outcome in lymph nodes until now.
We also described the protein status of EMT markers – E-cadherin and vimentin in PT and LNM. We have observed no or little change in E-cadherin and vimentin status between PT and LNM. Our results and the results of other groups  show that E-cadherin is at least as frequently expressed in LNM as in primary tumors and it is not downregulated in LNM, what would be expected if EMT was involved. However, this observation does not exclude the possibility that E-cadherin undergoes transient downregulation during EMT and is re-expressed at the stage of circulating tumor cells to facilitate adhesion and metastases formation . Similar to other groups we observed that vimentin is rarely expressed in LNM  or is reduced in comparison to the PT . However, expression of vimentin in metastases was also shown to be heterogeneous, metastasis size and side dependent , what supports the notion that vimentin expression is transient and environment-controlled. As the observed change in EMT-TFs status between PT and LNM is a new finding, a question arises if it results from true biological variation or from inconsistent measurement.
There are many hypotheses that could explain biomarkers change between PT and corresponding metastasis at biological level. Clonal selection, with subsequent clonal expansion during tumor progression, has been proposed as the mechanism inducing the differences in the genetic composition of primary and metastatic breast cancer [50, 51]. Clonal selection may be related to intra-tumor heterogeneity [52, 53] and/or to various selective pressures such as the immune surveillance of the host, stromal or growth factor interactions, nutritional deficiencies, hypoxia, and therapy. Indeed, Niikura reported recently a significantly higher discordance rate in HER2 status among women who received chemotherapy than among those who did not . To the biological variation between primary and metastatic tumor could also contribute independent evolution of an early stem cell clones in both sites, instead of a linear progression from the PT to metastasis .
The discordance between PT and LNM may also be caused by inconsistent measurement resulting from numerous technical issues . For example, recent meta-analysis revealed 15 pre-analytical variables capable of impacting IHC, including fixation delay, fixative type, time in fixative, reagents and conditions of dehydration, clearing, and paraffin impregnation, and conditions of slide drying and storage . Analytical procedures of antigen retrieval, immunostaining and interpretation of results add additional variance to the final result . To ensure maximal consistency, in our study all specimens were from single cancer hospital, examined by the same two observers with the same protocol of staining and scoring for PT and LNM. Additionally, independent observers were blinded to the clinical data and patient outcome.
The results should be, however, interpreted with caution because there is a possibility of selection bias due to retrospective character of the study. Another limitation might be the small sample size. But even with that number of patients the study had still sufficient power to detect reliably the 20% difference in biomarker status, and all of the discussed differences are above that value. It has been claimed that to reduce reporting errors the use of confirmatory test is recommended . To strengthen our study we confirmed increase of EMT-TFs at both mRNA and protein level with RT-qPCR and IHC, respectively. Taking into consideration discussed biological and technical issues, it seems that current study demonstrates true biological variation in TWIST1, SNAIL and SLUG in PT and LNM.
To make gene expression most reliable we have carefully validated the method we used. Since gene expression profiling in lymph nodes could only be examined in formalin-fixed, paraffin-embedded (FFPE) material, we have undertaken the methodological substudy to assess the reliability of obtained results of RT-qPCR. We aimed to investigate the feasibility of RT-qPCR-based gene expression profiling of low-level transcripts TWIST1, SNAIL and SLUG in FFPE tissues compared to matched frozen counterparts.
Expression of genes associated with EMT is transient and space-limited , what makes them specially difficult to study in clinical setting. Technical difficulties that interfere with reliable gene expression analysis can be particularly prominent in FFPE samples, in which RNA is degraded and chemically modified, resulting in lower RNA Integrity Number (RIN) values and rendering it inaccessible for amplification.
In order to examine to what extent tissue fixation influences RT-qPCR based methods we have measured sensitivity of the method in use by defining limits of detection and quantification (LOD and LOQ). No sample presented relative gene expression value of TWIST1 lower than calculated LOD and LOQ, which confirms the reliability of our data, as experimental results less than the theoretically possible LOD should never be reported.
The overall profile of TWIST1, SNAIL and SLUG expression was similar to that generated with well-preserved RNA from matched FF tissue. We presume however, that decreased functionality of RNA (due to degradation or modification) is the underlying factor responsible for increased intra- and interassay variation performed on FFPE samples. This process did not affect the sensitivity of the method as the assay measures relative quantity of the gene expression, which does not change due to similar degree of reference gene and examined genes degradation.
Thus, despite low RIN values RNA from FFPE samples may work fine in quantitative PCR in terms of specificity, sensitivity and reproducibility, even for low-level transcripts, as exemplified by TWIST1, SNAIL and SLUG.