Background
Up to 20% of Hodgkin lymphoma (HL) patients are either refractory to treatment (primary refractory) or experience relapse within four years (early relapse) of achieving complete remission (CR), and includes patients who experience progressive disease and patients with a particularly poor prognosis for other reasons [
1]. Only half of HL patients survive for two years if front line therapy fails, and autologous hematopoietic stem-cell transplant (ASCT) is only 50% curative [
2]. Although the International Prognostic Score was introduced to improve the risk stratification of patients [
3], its applicability is limited for predicting high risk cHL patients, regardless of clinical stage. While patients in this group may benefit from analysis of the tumor-associated macrophage marker CD68, which can be used to predict adverse outcomes of cHL [
4], the prediction is controversial [
5]. The antibody conjugate drug brentuximab vedotin targets CD30. In clinical trials, brentuximab vedotin therapy improved clinical outcomes for relapsing and refractory classical HL (RR-cHL) patients by producing survival times that were 6 months longer than for patients on the conventional treatment arm [
6]. This increased survival could perhaps be due to increased chemoresistance that can result from heavy pre-treatment. Therefore, the availability of biomarkers that identify patients who will have a poor outcome to conventional frontline therapy will permit more aggressive treatment of these patients, improving their prognosis.
Classical HL is a monoclonal lymphoid neoplasm that in almost all instances appears to be derived from (post-) germinal center B cells [
7‐
9]. The immunohistochemical (IHC) hallmark of HL tumor cells is CD30 antigen expression [
10]. The morphological phenotype of cHL comprises an unusually small number (<2%) of mononuclear Hodgkin (H) cells and multinucleated Reed-Sternberg (RS) cells residing in an extensive inflammatory background, which is mostly composed of T cells, histocytes, eosinophils, plasma cells, and macrophages [
10]. This inflammatory background in the tumor microenvironment is maintained by Hodgkin’s and Reed-Sternberg cell (HRS)-derived chemokines and cytokines that recruit the tumor microenvironment cellular components [
11‐
14]. The composition of the tumor microenvironment or the molecular phenotype of the HRS cells, or both, is thought to determine the relative aggressiveness of cHL at an individual level [
10].
At presentation, about 10–15% of cHL cases have extranodal involvement [
15], which is a negative prognostic factor even for patients with limited stage disease [
16]. Extranodal involvement, whether primary or secondary, indicates lymphatic and hematogenous spread of the disease [
15]. Therefore, neoplastic HRS cells could reasonably be assumed to occur in peripheral blood, albeit at levels not detectable by present diagnostic techniques, thus resulting in circulating tumor cell (CTC) involvement in HL. CTCs are frequently associated with poor clinical outcomes for solid [
17] and liquid tumors [
18]. Despite the limited number of cases (about a dozen over the past 100 years) of CTCs in the peripheral blood of HL patients, most were associated with either primary refractory or relapsing disease. In addition, well-established cell lines that have contributed tremendously to the understanding of HL were derived from primary HL tumor cells isolated from extranodal sites: peripheral blood [
19], bone marrow [
20], or pleural fluid [
21] of refractory or relapsing patients. These findings suggest that primary HL tumor cells can escape the physical barrier of the tumor microenvironment into the circulation to access extra-nodal destinations. The limited evidence indicating the presence of HRS cells among peripheral blood leukocytes (PBL) may be a consequence of their low proliferative index, the terminally differentiated status of the RS cells and their lack of mobility, or the propensity of these malignant cells to form a solid tumor mass [
22]. These characteristics have hampered investigations aimed at identifying HRS-derived biomarkers in peripheral blood for high risk, poor outcome, primary refractory, and early relapsing cHL patients.
Discussion
The survival time for high risk, unfavorable cHL (early relapse, progressive disease, primary refractory) ranges from 0 to less than 4 years [
1,
25], a very poor prognosis indeed, considering the high cure rate enjoyed by HL patients with current standard therapy. Second line treatments which include ASCT, and chemotherapy plus radiation do not significantly improve the prognosis of this group of patients. Therefore, it is essential to identify biomarkers that can help predict, prior to treatment which cHL patients belong in the high risk group so that appropriate treatment options with potentially better outcome can be implemented. The results presented here suggest that coexpression of FGF2 and SDC1 by CD30+ cells identify this group of patients.
Previous studies showed advanced stage, advanced age, and bulky disease as important risk factors for poor outcome [
16]. However, a contingency analysis performed on the HL database from the Tissue Repository of the Hackensack University Medical Center showed no association of any of these risk factors, including treatment history (all p > 0.1) with patient outcome. These data suggest that there may be undetermined molecular pathways that are altered in subsets of NS-cHL patients who are predisposed to be primary refractory or experience multiple relapses shortly after frontline treatments. To improve the specificity of potential biomarkers that may assist in pre-selecting poor outcome patients prior to treatment, we used bioinformatic data mining to derive a list of over 150 genes that represent pathways for metastasis, apoptosis, cell proliferation, tumorigenesis and angiogenesis. Expression screening data for these genes showed a consistent and robust overexpression of FGF2 and SDC1 in HL cell lines that were originally derived from primary HRS cells isolated from extranodal sites of refractory or relapsing HL patients. Qualitative scoring by IHC on lymphoma tissue arrays showed that FGF2 and SDC1 expression were indeed specific to the HL tumor microenvironment. Further analyses by qRT-PCR showed overexpression of either gene in poor outcome samples, while additional IHC on the poor outcome samples showed regions of CD30+ cells where FGF2 and SDC1 were strongly expressed. Double immunofluorescence staining of samples from poor outcome biopsies showed large subsets of CD30+ cells that expressed either FGF2 or SDC1. qRT-PCR and IHC evaluation of CD68 expression confirmed the clinical status of the biospecimens (poor outcome). The metastatic markers MMP9 and TGFβ1 were shown to be overexpressed in poor outcome patient samples, including their overexpression by subsets of CD30+ cells, suggesting metastasis by these HRS cell subsets. qRT-PCR analyses of PBL showed that CD30 and CD15 (the gene which encodes the protein that transfers fucose to N-acetyllactosamine polysaccharides to generate fucosylated carbohydrate structures) were transcriptionally upregulated in the untreated, poor outcome group compared to other clinical groups. Concurrently, markers representing circulating T cells (CD4 and CD8), B cells (CD19 and CD38), and monocytes (CD14 and CD63) were significantly downregulated in the untreated, poor outcome group, indicating that CD30 and CD15 upregulation was not a consequence of their expression by other common circulating lymphocytes. In the untreated, poor outcome group, transcription of FGF2 and SDC1 was upregulated, most likely by CD30+/CD15+ cells. Taken together, these data indicate that in untreated, poor outcome patients, a subset of CD30+ cells that express high levels of FGF2 and SDC1 transcripts, perhaps HRS cells, made their way into the circulation, and may be responsible for the poor outcome generated in primary refractory and early relapsing NS-cHL patients.
The expression of either FGF2 or SDC1 seen in our study is only partially consistent with previous reports. A previous immunoblot analysis showed no expression of FGF2 by HL cell lines KM-H2 and L428, although the same study did detect FGF2 expression in primary HRS cells from HL tumor biopsy samples [
26]. In contrast, our qRT-PCR data showed that FGF2 is transcriptionally upregulated in both KM-H2 and L428 cell lines. In HL, it appears as though FGF2 transcript translation in HRS cells is only induced
in vivo. Although perhaps not completely relevant to HL, elevated FGF2 mRNA is thought to be involved in tumor development and progression, as was demonstrated in acoustic neuromas [
27]. Also, there is discordance among studies of SDC1 expression by HRS cells. Studies have reported that the percentage of SDC1-positive HRS cells varies from 0 to 50% among cHL cases [
28‐
31], which is consistent for a post-germinal center origin. These differences may be due to variations in the fixation techniques and SDC1 antibody clones used; for example, some investigators contend that a much higher percentage of SDC-1 positive cells is seen in frozen material compared to formalin fixed paraffin embedded material. Our double immunofluorescence staining on fresh frozen sections from PO group patients showed large subsets of HRS cells that costained with anti-CD30 (clone Ber-H2) and anti-SDC1 (clone BB4 and a polyclonal antibody from Sigma-Aldrich).
The upregulation of FGF2 and SDC1 by putative CD30+ cells observed in our study may be the result of unregulated, uncontrolled expression of these genes in HRS cells from PO patients. Dysregulation of either FGF2 or SDC1 signaling alone or together has been associated with a variety of malignancies, including those associated with poor prognosis. Disruption of FGF2 expression results in elevated serum FGF2 levels, which is an independent poor prognostic factor for lymphoma, lung cancer, and sarcoma patients [
32‐
35]. In addition, elevated levels of FGF2 in serum have been reported for non-Hodgkin lymphoma (NHL) patients with poor prognosis [
36], shortened survival, and higher risk for mortality [
35]. Also in lymphoma, FGF2 overexpression in diseased tissue biopsy samples is associated with chemoresistance and inferior progression free and overall survival [
37]. Kowalska et al. (2007) showed that elevated FGF2 serum levels correlated with the erythrocyte sedimentation rate, which is a poor prognostic factor in HL [
16,
38]. At a molecular level, FGF2 binds to multiple membrane bound receptors in human cancers, including SDC1 [
39], and this receptor binding can trigger multiple signaling pathways, including those involved in cell proliferation and survival [
40,
41]. Although FGF2 may not always mediate SDC1 expression in cancers, SDC1 overexpression, at either a tissue or serum level, has been reported for multiple tumor types including solid tumors [
42‐
44], lymphomas, and in a number of lymphoproliferative disorders [
29,
30,
45‐
47]. In some instances, SDC1 overexpression is an adverse indicator for both solid and hematological malignancies [
43,
44,
48‐
50]. High levels of FGF2 and SDC1 in the same patient have important clinical implications. Multiple myeloma patients and small cell lung cancer patients with high serum levels of soluble SDC1 and FGF2 have poor prognosis and shortened survival [
51]; high serum levels of soluble SDC1 and FGF2 are also important clinical features of high risk, primary refractory, early relapsing cHL, and untreated poor outcome patients in our study. However, the clinical significance of co-upregulation of SDC1 and FGF2 in serum of HL patients has yet to be explored.
Our results also revealed that a large number of CD68+ tumor-associated macrophages were present in the tumor microenvironment of poor outcome tissue samples in which the CD30+ cells overexpressed both SDC1 and FGF2. A number of previous reports showed that CD68+ tumor-associated macrophages are a poor outcome marker of cHL [
52‐
54]. Therefore, simultaneous overexpression of FGF2 and SDC1 by CD30+ cells can be used as a molecular signature to identify high risk, poor outcome cHL patients.
The downregulation of markers representing circulating T cells, B cells and monocytes in the PBL of untreated poor outcome patients in our study is consistent with lymphocytopenia, a negative prognostic factor in multiple cancer types [
55]. However, lymphocytopenia alone may not adequately predict poor prognosis, and as such better biomarkers are needed. Two recent reports indicated that the ratio of the absolute lymphocyte count to the absolute monocyte count (ALC/AMC) is an independent prognostic marker that can be used for stratifying high vs. low risk cHL [
56‐
58]. Although not demonstrated in hematological malignancies, lymphocytopenia implies a depleted immune system that lacks adequate immune surveillance, which could play an important role in aggressive tumor metastasis. Indeed, both lymphocytopenia and circulating tumor cells were shown to be independent prognostic factors in the metastasis of breast cancer, carcinomas, sarcomas, and lymphomas; their presence is associated with an extremely poor clinical outcome [
55,
59]. Such a scenario may play a role in extranodal involvement of HL, and may identify unfavorable high risk patients irrespective of disease stage. In such a setting, this population of HL patients may benefit from therapies to restore immune function prior to frontline therapy.
In some ways, the co-upregulation of FGF2 and SDC1 seen in tissue biopsies and PBL (albeit at the mRNA level) of PO NS-cHL patients in our study may be related to certain features of multiple myeloma. Several studies found shared features between HRS cells and plasma cells, including those of multiple myeloma and their normal counterparts, despite the differences in disease behavior [
30,
60‐
62]. Like plasma cells, HRS cells typically not only lack expression of B-cell surface markers, but they are also the only other lymphocytes that occasionally express SDC1 [
30,
60‐
63]. Although B lymphocyte–induced maturation protein 1 (Blimp-1), which is a transcription factor required for plasma cell differentiation [
62], was not part of our study, a fraction of HRS cells also express this protein. There is the possibility that subsets of, if not all, HRS cells and multiple myeloma plasma cells share a common ancestral precursor [
64], although the majority of HRS cells shed their plasma cell signature (e.g., SDC1 expression) [
62]. Those HRS cells that continue to express SDC1 are perhaps more aggressive than their SDC1 negative counterparts, thus contributing to the aggressive nature of poor outcome cHL, and producing the sort of unfavorable prognosis that is typical of primary refractory and early relapsing cHL patients. Of additional interest is the coexistence of HRS cells with aggressive multiple myeloma [
65], or their appearance after treatment of multiple myeloma [
66].
Our data also showed that the established metastatic markers MMP9 and TGFβ1 were overexpressed by subsets of CD30+/FGF2+/SDC1+ cells in tissue biopsy samples from PO patients. HRS cells produce activated TGFβ1 in primary tumor tissues, predominantly in nodular sclerosing HL [
67], while MMP9 overexpression is associated with adverse clinical outcomes in HL [
68]. As such, HRS cells that harbor the FGF2+/SDC1+ immunophenotype and express both MMP9 and TGFβ1 are the cells most likely to be shed from the tumor microenvironment. Thus, the molecular interplay of FGF2, SDC1, MMP9, and TGFβ1 may play a role in HL metastasis.
Finally, our results revealed that subsets of circulating cells transcriptionally upregulate CD30, CD15, FGF2, and SDC1 in the untreated poor outcome group. The main subset here could be putative HRS cells or some other variant of these neoplastic cells that are SDC1+/FGF2+ and overexpress MMP9 and TGFβ1. While characteristic HRS cells are not typically found in PBL, the metastatic and hematogenous spread of HL is suspected in cases diagnosed with extralymphatic and extranodal involvement [
15]. Therefore, a variant of HRS cells that do not exhibit the classical phenotype displayed by nodal HRS may be in the circulation of untreated poor outcome patients, perhaps due to a difference in the microenvironment (PBL versus lymph node). In other settings, either normal cells or circulating tumor cells may express important transcripts that are translated only when the appropriate microenvironment prevails, and thus the cell phenotype may also change. This concept is evident during development during which the zygote produces maternal RNAs that are later translated into functional proteins at each stage during embryogenesis. At least two studies of HL patient subsets suggest a similar occurrence. In vitro experiments by Zucker-Franklin and colleagues (1983) and Sitar et al. (1994) showed that RS-like cells can be generated from cultured peripheral mononuclear blood cells (PMBC) from HL patients [
69,
70]. Zucker-Franklin et al. observed RS-like cells only in HL samples (including early stage disease), and not PMBC of NHL, mycosis fungoides, or of control samples, suggesting that giant cell formation from PMBC is limited to HL cases. Sitar and colleagues showed that 10% of the giant RS-like cells were CD30+ and EBV-positive [
70].
Methods and materials
The BioXM software platform (Sophic Alliance, Rockville, MD) was used to mine potential biomarkers for Hodgkin’s lymphoma using the National Cancer Institute (NCI) Cancer Gene Index, which contains 7,000 cancer genes and 2,200 biomarker genes. These genes were annotated and validated from 18 million Medline abstracts and 24,000 Hugo genes from over 80 databases, using a combination of algorithmic methods (Biomax Informatics, Munich, Germany) that included natural language processing (NLP), Biomarker Role Codes, the NCI Cancer Thesaurus, and Karp’s Evidence Codes [
23]. The identification of potential biomarkers was performed by initiating queries on BioXM with a combination of search terms including Hodgkin’s disease, lymphoma, cancer, biomarker, overexpression, up-regulation or down-regulation, and differentially-expressed. The bioinformatics-guided search generated 151 potential HL biomarkers (Table
2).
Cell lines and cell culture
The Hodgkin’s lymphoma cell lines KM-H2, HD-MY-Z, HDLM-2, L-591, and SUP-HD1 were obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). L-428, L-1236, and L-540 cells were generous gifts provided by Dr. Volker Diehl (University of Cologne, Germany). U-H01 and DEV cells were kind gifts from Dr. S. Brüderlein (University Hospital Ulm, Germany) and Dr. Debora De Jong (Netherlands), respectively. KM-H2, L-428, HD-MY-Z, and L-1236 cells were cultured in 90% RPMI 1640 supplemented with 10% fetal bovine serum (FBS). SUP-HD1 cells were grown in 80% McCoy’s 5A medium containing 20% FBS. HDLM-2, L-540, and L-591 cells were grown in 80% RPMI 1640 supplemented with 20% FBS. U-H01 cells were grown in Iscove’s MDM and RPMI 1640 (4:1) supplemented with 20% FBS. All culture media contained 2 mM L-glutamine, penicillin (100 U/ml), and streptomycin (0.1 mg/ml). Cultures were maintained at 37°C with 5% CO
2. The clinical characteristics of each cell line were previously documented and are presented in Table
3. DEV, KM-H2, and SUP-HD1 cells were derived from relapsing cases. HD-MY-Z, L1236, L428, and U-H01 cells were from refractory patients.
Table 3
Characteristics of HL cell lines
DEV | relapse | Pleural fluid |
HDLM2 | n/a | Pleural fluid |
HD-MY-Z | refractory | Bone marrow |
KM-H2 | relapse | Pleural fluid |
L1236 | refractory/relapse | Peripheral blood |
L428 | refractory | Pleural fluid |
L540 | n/a | Bone marrow |
L591 | n/a | Pleural fluid |
SUP-HD1 | relapse | Pleural fluid |
U-H01 | refractory | Pleural fluid |
RNA isolation and cDNA synthesis
Total RNA from cell lines and peripheral blood (PBL) of HL patients was isolated using Trizol (Invitrogen, Carlsbad, CA). RNA from archived FFPE tissue sections was extracted using RNeasy (Qiagen, CA) according to the manufacturer’s instructions. The RNA concentration was spectrophotometrically determined at A260 (ThermoElectro Corporation). Total RNA integrity was checked by resolution on a 2% agarose gel under denaturing conditions. cDNA was generated using the SuperScript III RT First-Strand cDNA Synthesis Kit (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol. Oligo-dT primers were used to generate cDNA from cell lines and PBL-derived RNA, and random hexamers were used for generating cDNA from RNA obtained from FFPE sections.
Polymerase chain reaction (PCR)
Primer sets used for each gene were generated using online primer tools (University of Massachusetts;
http://biotools.umassmed.edu/bioapps/primer3_www.cgi) (Table
2). Primers were designed to have lengths of 18 to 27 nt with Tm = 60°C and 45 to 65% GC content, and were synthesized by a custom primer service provided by Invitrogen. Each primer pair was confirmed to generate a single discrete band by end-point PCR (BioRad DNA Engine Peltier Thermal Cycler) using cDNAs generated from normal spleen tissue. End-point PCR conditions consisted of denaturation at 95°C for 30 seconds, annealing at 55°C for 30 seconds, and primer extension at 72°C for 1 minute. The primer pairs were designed to generate a PCR fragment of 150–170 bp for cell line- and PBL-derived cDNA, and 70–100 bp for FFPE-derived cDNA (Table
4). The PCR products were resolved on a 2% agarose gel and visualized with ethidium bromide staining using a BioRad Imager. For qRT-PCR, each reaction consisted of 43 ng cDNA, 10 mmole primers and 10 μl 2X Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) in a final volume of 20 μl, which was placed in a MicroAmp Fast Optical 96-Well Reaction Plate designed for use with the ABI7900 PCR system (Applied Biosystems). The reaction was performed using the standard mode (initial denaturation at 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute). Each qRT-PCR reaction was done in triplicate, and each data set was analyzed with ABI7900 software. The amount of target mRNA was normalized to the expression levels of the housekeeping gene GAPDH. For cell lines, CD19 was used as control. For PBL analysis, the expression levels of CD14/63, CD38/19, and CD4/8 were compared against their expression in monocytes, CD19+ B cells, helper T cells, and cytotoxic T cells, respectively, of healthy donors (Miltenyi Biotech). Pooled normal cDNA (n=20) was used as a control for gene expression analysis of FFPE tissue-derived cDNA. The ΔΔCt method was used to calculate the fold-change relative to controls.
Table 4
Primer sets for each gene used in this study
A. Primer sets used on PBL samples. |
GAPDH | catggcctccaaggagtaag | aggggtctacatggcaactg |
CD4 | atgtggcagtgtctgctgag | cctagcccaatgaaaagcag |
CD8 | cagagctacccgcagagttc | ctccaaccctgacttgctgt |
CD30 | ccaacttagctgtcccctga | ctgggaccaatgctgttctc |
CD15 | gcaggtgggactttgttgtt | ccaaggacaatccagcactt |
CD19 | ttctgcctgtgttcccttg | cacgttcccgtactggttct |
CD38 | agatctgagccagtcgctgt | aaaaaggcttccgtctctgg |
CD14 | gagctcagaggttcggaaga | ttcggagaagttgcagacg |
CD63 | aaccacactgcttcgatcct | aatcccacagcccacagtaa |
FGF2 | tgctcagcagtcaccatagc | cttgaggtggaagggtctcc |
SDC1 | cttcacactccccacacaga | ggccactacagccgtattct |
B. Primer sets used on FFPE tissues. |
GAPDH | cctcaacgaccactttgtca | ccctgttgctgtagccaaat |
TGFβ | gtacctgaacccgtgttgct | cacgtgctgctccactttta |
MMP9 | ggcgctcatgtaccctatgt | gccattcacgtcgtccttat |
CD30 | gaagctccacctgtgctacc | ggtctggaatccacaagctc |
CD68 | tgacacccacggttacagag | gtggttttgtggctcttggt |
SDC1 | taggacctttccaccacagc | gaggctgcttcagtttggag |
FGF2 | tgaggctgagaggtcaaggt | ctctgttgcctaggctggac |
Selection of clinical samples
The selection criteria of peripheral blood samples were based on the response to front line therapy (Table
1). Twenty five nodular sclerosing cHL patient samples registered in the database at the Hackensack University Medical Center were categorized into: 1) good outcome chemo-naïve, untreated, relapse-free/disease-free > 4 years (n=12); 2) poor outcome chemo-naïve (untreated), primary refractory or early relapse (n=7); 3) chemo-exposed (pretreated), multiple relapses (n=6). Formalin-fixed, paraffin-embedded (FFPE), and fresh frozen (FF) lymph nodes from different HL stages and subtypes were obtained from Thomas Jefferson University, the Tissue Repository of the Hackensack University Medical Center, and Proteogenex (Culver City, CA). Biospecimens with the relevant clinical characteristics were grouped into good outcome (GO, relapse free/disease free > 4 years, n=20) and poor outcome (PO, shortened survival— death 2 to 3 years after diagnosis). A lymphoma tissue array was obtained from US Biomax (Rockville, MD).
Immunohistochemistry
FFPE and fresh frozen lymph nodes from different stages and subtypes of HL were purchased from US Biomax and Proteogenex. FFPE sections (5 μm) mounted on slides were dewaxed twice with Histochoice clearing agent (Amresco, Solon, OH) for 10 minutes each, then sequentially hydrated in 100%, 90%, 80%, 70%, and 50% ethanol followed by equilibration in PBS for 5 minutes each. All antigen retrievals were carried out in a 95°C water bath for 20–30 minutes (depending on the antigen) using high pH (pH 9) buffer (DAKO) for FGF2, SDC1, MMP9, and CD68, or low pH (pH 6) buffer (DAKO) for CD30, TGFβ1, and CD20. The sections were cooled for 20 minutes at room temperature and then washed twice with PBS for 5 minutes. Endogenous peroxidases were quenched by incubating the sections in 3% H2O2 solution in PBS for 10 minutes followed by rapid washes in PBS at room temperature. A hydrophobic PAP pen (Vector Labs, Burlingame, CA) was used to make a dam around the sections, which were then blocked at room temperature for 2 hours with 1% BSA containing 5% swine serum in PBS, followed by overnight incubation with primary antibodies at 4°C. Monoclonal antibodies for CD30 (clone Ber-H2, DAKO), SDC1 (clone BB4, Abd Serotec), CD68 (clone PG-M1, DAKO), and CD20 (clone L26, DAKO) were used at dilutions of 1:20, 1:40, 1:50, and 1:100, respectively. Rabbit polyclonal antibodies for FGF2 (Santa Cruz), TGFβ1 (Santa Cruz), and MMP9 (DAKO) were used at dilutions of 1:200, 1:200, and 1:100, respectively. Stained sections were washed three times in PBS/0.1% Tween-20 for 5 minutes each and then once in PBS for 5 minutes. Signal detection was carried out using an LSAB kit according to the manufacturer’s instructions (DAKO), with minor modifications. Briefly, sections were incubated in Biotinylated Link for 30 minutes at room temperature and washed three times in 0.1% PBST for 5 minutes each. Sections were then incubated in streptavidin-HRP for 30 minutes and washed as described above. Signals were visualized by incubating the slides in a solution of 1 ml substrate buffer with 1 drop chromogen, and immediately rinsed in tap water. The sections were counterstained with hematoxylin (Vector Labs) for 22 seconds and immediately washed in tap water before mounting with Aqua Mount (Vector Labs). Photomicrographs of stained tissues were generated with an Axio Cam MRc camera coupled to an Axio Imager Microscope (Carl Zeiss, Thornwood, NY). Positive control slides included tonsil for CD20, CD68, and SDC1, and ALCL for CD30 (on lymphoma array). For qualitative scoring, no staining was assigned a score of 0, weak staining 1, moderate staining 2, and intense staining, 3.
Immunofluorescence
Double immunofluorescence analysis was performed on 5 μm FFPE and OCT-embedded 8 μm fresh frozen (FF) tissue sections that were mounted on positively-charged frosted slides (Histoserv, Germantown, MD). FFPE sections were processed similarly to the preparation used for IHC. OCT-embedded FF sections were thawed at room temperature for 20 minutes, rinsed briefly in PBS, and then fixed in 3.7% formaldehyde (Electron Microscopy Sciences, PA) for 20 minutes at room temperature. The remaining steps for immunofluorescence signal detection were carried out using a TSA Detection system (Invitrogen) according to the manufacturer’s instructions. Monoclonal and polyclonal signals were detected with Alexa Fluor 488 and Alexa Fluor 546, respectively. The antibodies used were the same as for IHC, except that a SDC1 rabbit polyclonal antibody (Sigma-Aldrich) was used for CD30-SDC1 double staining. Slides were counterstained with Hoechst 33342, visualized with a Leica DMI 6000B inverted microscope, and analyzed using Leica MM AF software, version 1.5 (Leica Microsystems). Slides were independently reviewed and verified by two pathologists.
Statistics
Data analyses were performed using SAS 9.1.3, StatView 5, or JMP 4. Contingency and likelihood ratio analyses were used to determine the independence of staging and prognosis. The mean fold-change for each sample was determined from triplicates of the qRT-PCR data. Analysis of variance (ANOVA) and F statistics were used to determine differences between the means of the poor outcome group and other outcome groups. Fisher’s protected least significant difference (PLSD) was used to determine pair-wise significant differences between group means.
Competing interests
The authors have no conflict of interest to declare.
Authors’ contributions
RG and KSS designed the study and wrote the manuscript. RG, JP, CK, SS, NH, and SV performed the experiments; TT and MT provided clinical samples and performed experiments; KN provided bioinformatics software and analyses; RG, AG, PB and KSS analyzed the data; AG and AP reviewed the paper and provided advice. All authors read and approved the final manuscript.