Identification of stromal proteins specifically involved in cancer is of considerable interest , since the "reactive stroma" of desmoplastic reactions is thought to create a permissive and supportive environment contributing to cancer progression . Moreover, stromal proteins that are selectively expressed in cancer tissues are prime candidates for tumour targeting strategies because (i) they are expected to be more accessible than intracellular proteins, (ii) they are often present in high quantities and iii) tumour ECM components are generally more stable than tumour-associated cell surface antigens . Considering that ~30% of patients experience relapse and die of complications due to progressive disease and/or treatment, the identification of proteins from HL stroma could not only provide the basis for a better understanding of the disease, but also lead to the discovery of new diagnostic and therapeutic markers allowing earlier detection and treatment of relapsed and refractory lymphoma by e.g. chemoprevention  or antibody based anti-cancer treatments .
The clinical value of antibody-based targeted therapies (targeting specifically the stroma) in HL patient has been elegantly demonstrated in a recent study using a monoclonal antibody directed against the EDB alternative spliced variant of fibronectin, known as L19 antibody . In light of these promising results, it is clear that the discovery of novel accessible biomarkers in HL is a necessary step to develop a large panel of targeting-based therapies. Using a recently described chemical proteomic method , we have identified potentially accessible proteins that are of special interest for imaging technologies and targeted therapies in the Hodgkin's disease. Our method led to the identification of a total of 1431 proteins in HL and RLH. We used mass spectrometry as a screening tool, and only one two dimensional liquid chromatography run was performed per sample. Analytical completeness was certainly not reasonably achievable with one single 2D-LC run , but was not absolutely necessary to unravel new potential biomarkers (the assumption was made that a single run was likely to be sufficient to detect the most abundant proteins, and the quantitative aspect, despite analytical completeness issues, was still helpful for biomarker selection). IHC was the mandatory step to validate these potential biomarkers. Data obtained from mass spectrometry and IHC experiments on identical samples are most of the time consistent, but some discrepancies do exist (see Additional File 3). These differences can be explained by i) analytical completeness of the mass spectrometry data as stated above, ii) heterogeneity in the sample (one half for mass spectrometry processing, one half for IHC experiments) and iii) the specificity and sensitivity of antibodies used in IHC experiments. Our mass spectrometry methodology and IHC are therefore rather complementary techniques, and it can be foreseen that IHC results would more likely compare to mass spectrometry experiment such as MRM (Multiple Reaction Monitoring) that could look specifically for versican, fibulin-1 and periostin in each sample.
The complete list of identified proteins included extracellular and plasma membrane proteins, but also include a non-negligible fraction of intracellular proteins. The reasons for this observation have already been discussed elsewhere [15, 28]. Focusing on ECM proteins, we identified several proteins already known to be associated with human HL. These proteins were consequently not further validated by immunohistochemistry. For instance, hemopexin protein was found elevated in the serum of patients with HL ; eosinophil peroxidase has already been used for radio-immunodetection of Hodgkin's disease ; coagulation factor XIII was found to be expressed by macrophages and involved in the stabilization of fibrin deposits in the tumour stroma of HL samples ; and nucleoside diphosphate kinase A expression was found to be a prognostic factor for classical HL .
We also identified proteins without direct relationship with HL demonstrated in the literature. Among these proteins, we showed that the extracellular proteins versican, fibulin-1 and periostin were actually overexpressed in the extracellular matrix after immunohistochemical analyses.
The staining pattern of expression was somewhat similar for the 3 proteins in most HL samples tested. This result is interesting, since the three proteins are expressed by different cells: versican is known to be expressed by fibroblasts and only rarely by cancer cells, fibulin-1 was shown to be expressed mainly by cancer cells and only by some fibroblasts , and periostin is expressed by both fibroblasts and cancer cells. Moreover the staining was mainly specific to fibrosis. While only scarce information is available about versican in fibrosis, except in the lung  (where it may influence early repair processes), fibulin-1 deposits were already reported in rat liver fibrosis . Periostin, by altering the deposition and attachment of collagen, is a critical regulator of fibrosis .
Among these three potential biomarkers, two of them, namely versican and fibulin-1, were almost absent in RLH. Although periostin displayed a very strong expression in HL, this protein was indeed found expressed at lower levels in RLH. It would be of interest to assess the serum concentrations of each of these ECM proteins in patients with HL in order to determine how valuable seric dosage of these proteins may be for diagnostic purposes.
S100A8 was expressed in the ECM of both RLH and HL. The protein was found to be expressed in cells of the immune system (e.g. monocytes, eosinophils, neutrophils). Since these cells are typically recruited in HL lesions, the increased expression of this protein in this diseased state was expected. Results of mass spectrometry were anyway confirmed for the four proteins tested in IHC.