|Home | About | Journals | Submit | Contact Us | Français|
Friend leukaemia integration‐1 (FLI‐1) antibody is a useful marker for Ewing's sarcoma/primitive neuroectodermal tumour (EWS/PNET) and vascular tumours. However, it is also expressed in subsets of lymphoblastic lymphoma, Merkel cell carcinoma (MCC) and desmoplastic small round cell tumour (DSRCT).
To determine expression of FLI‐1 in various benign and malignant neoplasms, by immunohistochemical analysis on 4323 tumours using multiple tumour microarrays, as well as on whole sections.
FLI‐1 was expressed in 46/62 EWS/PNETs, 2/3 olfactory neuroblastomas, 7/102 small cell carcinomas of the lung, 10/34 MCCs, 1/14 rhabdomyosarcoma, 19/132 non‐Hodgkin's lymphomas, 2/3 DSRCTs, and in 53/74 benign and malignant vascular tumours. In addition, 27/508 squamous cell carcinomas, 19/837 adenocarcinomas, 10/400 urothelial bladder cancers, 1/40 basal cell carcinomas, 3/29 liposarcomas, 1/40 glioblastoma multiforme and 9/29 medullar carcinomas of the breast expressed FLI‐1. The sensitivity and specificity of FLI‐1 to distinguish EWS/PNET from all types of malignancies were 74.2% and 96.0%, respectively. Finally, the sensitivity and specificity of FLI‐1 to distinguish EWS/PNET from other small round cell tumours (SRCTs) were 74.2% and 91.6%, respectively.
This study was the first to show that FLI‐1 can be seen in a variety of solid tumours, some of which had never been explored before. This finding should be kept in mind, especially when using FLI‐1 as a marker for finding the primary origin of poorly differentiated metastatic tumour. Finally, despite the expression of FLI‐1 in numerous malignancies, it is still considered to be highly sensitive and specific in distinguishing EWS/PNET from other tumour types in general and from other SRCTs in particular.
Ewing sarcoma/primitive neuroectodermal tumour (EWS/PNET) belongs to a group of highly aggressive malignant tumours named small round cell tumours (SRCTs).1 Almost 80% of EWS/PNET cases present the fusion gene Ewing sarcoma (EWS)/FLI‐1 (friend leukaemia integration‐1) resulting from the balanced translocation t (11; 22) (q24; q12), which includes the N‐terminal transactivation domain of the EWS gene and the C‐terminus DNA‐binding domain of the FLI‐1 gene.2 FLI‐1 antibody is a polyclonal commercially available antibody directed against the C‐terminus of FLI‐1 protein‐binding domain. In normal tissues, FLI‐1 was found to be restricted to haematopoietic cells and endothelial cells. FLI‐1 was mainly expressed in EWS/PNET with a specificity of over 90%, and later on it was added to CD99 as a useful marker in the histological diagnosis of EWS/PNET.3,4 However, further studies showed that FLI‐1 was frequently seen in various tumour types, including vascular tumours, lymphoblastic lymphoma, Merkel cell carcinoma (MCC) and desmoplastic small round cell tumour (DSRCT).3,4,5,6 Furthermore, few studies on FLI‐1 immunoexpression in tumour types other than those mentioned above were found. In those studies, FLI‐1 was not expressed in any of the cases analysed.3,4,5,6 However, those studies were hampered by a small number of samples and by a very limited selection of tumour types.3,4,5,6 Hence, the aim of this study is to define the expression of FLI‐1 using FLI‐1 polyclonal antibody in a large number of benign and malignant tumours. Our results will shed some light on the possibility of FLI‐1 expression in a variety of tumours, which could be of major importance when using FLI‐1 to evaluate poorly differentiated tumours, especially when they are metastatic and of unknown origin. To accomplish this aim, a multiple tumour microarray technique has been used, which is a high‐throughput, efficient and practical technique for evaluating the expression of immunohistochemical markers.7,8,9
A formalin‐fixed, paraffin‐wax‐embedded multiple tumour tissue microarray was used. The microarray was constructed as described previously.7,8,9,10 In addition, whole sections from 26 EWS/PNET, 28 MCC and three DSRCT cases were included in the study. Whole sections from normal tissues were also included. An H&E‐stained section was evaluated for the presence of the tumour by light microscopy. Sections of 4 μm thickness were processed for immunohistochemical analysis. Endogenous peroxidase was blocked with 0.3% hydrogen peroxidase for 30 min. Antigen retrieval was carried out in a high pH buffer for 3 min in a steamer/cooker. Subsequently, sections were incubated with FLI‐1 polyclonal (C‐19) antibody (clone sc‐356, 1:50, Santa‐Cruz, California, USA) at room temperature for 30 min. A biotin‐free horseradish peroxidase enzyme‐labelled polymer of the Envision plus detection system was added (Dakocytomation). The diaminobenzidine complex was used as the chromogen detection reagent. In negative controls, a normal rabbit serum was used instead of the primary antibody. Nuclear staining was required in order to consider the FLI‐1 staining positive. Evaluation of the immunohistochemistry slides was performed semiquantitatively by two pathologists (PM‐F and WB), who were not aware of the original histological diagnosis, using a double‐head microscope. The scores were reviewed, and, whenever a discrepancy was noted between the first and second readings, a third pathologist (RP) was asked to review the cases. The three pathologists reached an agreement on the final scoring. For scoring, intensity and percentage of positive cells were taken into consideration. The intensity was classified into three categories: weak, moderate and strong. A cut‐off of 5% positive tumour cells was used to define positive results, as described previously.4 However, all cases had >50% positive tumour cells.
Table 11 summarises FLI‐1 immunoexpression in normal human tissues. In normal tissues, FLI‐1 expression was restricted to haematopoietic cells and endothelial cells. It was negative in all other normal tissues.
Table 22 gives the summary of FLI‐1 expression in 4323 benign and malignant tumour types.
FLI‐1 expression in the SRCT group was as follows: 46/62 (46.8%) EWS/PNET, 10/34 (29.4%) MCC, 7/102 (6.9%) small cell carcinoma of the lung, 19/132 (14.4%) non‐Hodgkin's lymphoma (NHL), 1/14 (7.1%) rhabdomyosarcoma (RMS), 2/3 (66.7%) DRSCT and 2/3 (66.7%) olfactory neuroblastoma (ONB) cases. FLI‐1 was expressed in 53/74 (71.6%) of all vascular neoplasms, including 3/4 (75.0%) angiosarcoma, 19/29 (65.5%) Kaposi's sarcoma, 7/12 (58.3%) haemangiopericytoma and 24/29 (82.7%) capillary haemangioma. In addition, there were 27/508 (5.3%) squamous cell carcinomas, 19/837 (2.3%) adenocarcinomas, 10/400 (2.5%) urothelial carcinomas of the bladder, 3/29 (10.3%) liposarcomas, 1/40 (2.5%) glioblastoma multiforme and 1/40 (2.5%) basal cell carcinoma of the skin (fig 1A–D). Furthermore, 9/29 (31.0%) medullary carcinomas of the breast showed immunoexpression of FLI‐1. In more than half of the negative MCC cases there was a cytoplasmic background staining, although, once again, those cases were interpreted as negative, owing to the absence of nuclear staining for FLI‐1 (fig 22).
Table 33 illustrates the sensitivity, specificity and predictive value of FLI‐1 in distinguishing EWS/PNET from all other malignancies, as well as from tumours belonging to the SRCTs and others that might enter in the differential diagnosis with EWS/PNET.
We found FLI‐1 to be highly sensitive in distinguishing EWS/PNET from all types of malignancies in general, and from germ cell tumours, as well as from other tumours belonging to the group of SRCTs in particular, with a value of 74.2% for each. On the other hand, FLI‐1 was highly specific in distinguishing EWS/PNET from all types of malignancies in general, and from germ cell tumours, as well as from other SRCTs in particular, with values of 96.0%, 100% and 66.7%, respectively. FLI‐1 had a low positive predictive value (25.6%) in distinguishing EWS/PNET from all other solid tumours, but had a very high negative predictive value of 99.5%. On the other hand, FLI‐1 had a high positive predictive value (100%), as well as negative predictive value (90.2%), in distinguishing EWS/PNET from germ cell tumours. Finally, the positive and negative predictive values of FLI‐1 to distinguish EWS/PNET from other SRCTs were 66.7% and 94.0%, respectively.
EWS/PNET is a small, blue, round cell tumour with a very characteristic t(11,22) chromosomal rearrangement, which results in fusion of the EWS and FLI‐1 genes. The resultant fusion protein promotes oncogenesis, which is necessary for continued growth of tumour cell lines. Studies show that loss of the COOH‐terminal domain can attenuate the ability of EWS/FLI‐1 to promote anchorage‐independent growth. Furthermore, cells with EWS/FLI‐1 COOH deletion mutants show a progressive decrease in tumour oncogenesis and loss of round cell morphology on histological examination. The above data led to the suggestion that the C‐terminus of FLI‐1 seems to have a crucial functional role in EWS/FLI‐1 oncogenicity.3 Since the commercial availability of FLI‐1, few comprehensive studies on its expression in benign and malignant tissues have been published. In 1999, Nilsson et al11 were the first to demonstrate FLI‐1 expression using sc‐356 polyclonal antibody in EWS/PNET cell lines and in all five formalin‐fixed paraffin‐wax‐embedded EWS/PNET cases. Soon after, Folpe et al4 explored its expression in a series of SRCTs. They showed that FLI‐1 was expressed not only in 71% EWS/PNET but also in various SRCTs, including 7/8 (88%) lymphoblastic lymphomas, 1/1 DSRCT, 0/1 poorly differentiated synovial sarcoma, 0/32 RMS, 0/3 neuroblastoma, 0/3 Wilms' tumour, 0/8 ONB and 0/1 mesenchymal chondroblastoma, indicating that FLI‐1 can be expressed in a variety of SRCTs but its expression is still highly specific (90%) for EWS/PNET. Since then, FLI‐1 has been used as a marker for EWS/PNET, and its use has been repeatedly described in numerous case reports.12,13,14,15 Another work, by Llombart‐Bosch et al,3 found FLI‐1 expression in 16/19 (84%) EWS/PNETs, 4/5 (80%) NHLs (large cell type, unspecified lineage), 2/9 (22.2%) neuroblastomas and in 3/6 (50%) undifferentiated synovial sarcomas. In our series of SRCTs, FLI‐1 was found in 46/62 (74.2%) EWS/PNETs, 2/3 (66.7%) ONBs, 2/3 (66.6%) DSRCTs, 1/14 (7.1%) RMS and 7/102 (6.9%) small cell carcinomas of the lung. In addition, and similar to the results by Folpe et al, we found FLI‐1 to be almost always negative in Wilms' tumours (n=47) and neuroblastomas (n=31), findings that are both important and very useful in the differential diagnosis of SRCTs in children. Finally, FLI‐1 was negative in our series of 148 cases of germ cell tumours, which might enter into the differential diagnosis with EWS/PNET. From these data, we concluded that FLI‐1 is a highly sensitive and specific marker in distinguishing EWS/PNET from other SRCTs, as well as from germ cell tumours. Further, we showed that FLI‐1 has high positive predictive and negative predictive values in distinguishing EWS/PNET from other SRCTs, as well as from germ cell tumours. Also, 19/132 (13.1%) NHLs expressed FLI‐1, and it was seen in different NHL subtypes, including follicular, Burkitt's, diffuse large B‐cell and peripheral T‐cell lymphoma. Thus, FLI‐1 is seen not only in lymphoblastic lymphoma but also in other lymphoma subtypes. FLI‐1 is expressed in normal endothelial cells, and the FLI‐1 gene has been reported to play an important role in the embryological development of blood vessels.16 The purpose of using FLI‐1 in the diagnosis of vascular tumours was explored, and FLI‐1 seemed to be expressed in 94% of cases.5,17 In the present study, we found FLI‐1 to be expressed in 71.6% of all vascular neoplasms.
There are two studies in the literature discussing FLI‐1 expression in tumours other than SRCTs and vascular neoplasms. In the first report, all 60 cases of non‐vascular tumours were negative for FLI‐1 expression. These cases were 0/16 sarcoma, 0/7 melanoma and 0/45 overall carcinoma, including 3 squamous cell carcinomas (SCC), 12 breast ductal carcinomas, 21 adenocarcinomas (pancreatic (n=4), pulmonary (n=5), ovarian papillary serous (n=5), uterine endometrioid (n=6) and colonic (n=1)), 3 renal cell carcinomas, 1 hepatocellular carcinoma, 3 salivary mucoepidermoids and 2 pulmonary carcinoids. In the second report, however, FLI‐1 monoclonal antibody (GI‐46‐222, BD Pharmingen) was used, and the results were as follows: all EWS/PNET (n=15) and vascular tumour (n=45) cases, 2/5 MCCs and 1/10 malignant melanoma (MM) were strongly positive for FLI‐1, weak expression was seen in 3/5 MCCs, 3/10 synovial sarcomas, 5/10 malignant melanomas, 6/10 lung adenocarcinomas, 1/10 breast carcinoma, and all DSRCTs (n=5), RMS (n=10), high‐grade pleomorphic sarcomas (n=10) and colon carcinomas (n=10) were negative for FLI‐1.18 On exploring FLI‐1 expression in a large number of tumours, we found it to be present in a small percentage of a variety of solid tumours, such as 27/508 (6.9%) SCCs, 19/837 (2.3%) adenocarcinomas, 10/400 (2.5%) urothelial bladder carcinomas, 9/29 (31.0%) medullary carcinomas of the breast, 1/40 (2.5%) glioblastoma multiforme and 1/40 (2.5%) basal cell carcinomas of the skin. Despite its expression in a variety of malignant tumours, FLI‐1 is still a highly sensitive and specific marker to distinguish EWS/PNET from all types of malignancies. Furthermore, and in this context, it did not have a high positive predictive value but still showed a high negative predictive value. Thus, our study is the first to evaluate a large series of types of tumours using FLI‐1 polyclonal antibody, and its expression by the tumours described above should be kept in mind when using FLI‐1 as a marker to find the primary origin of a metastatic poorly differentiated tumour.
Finally, although Llombart et al found FLI‐1 expression in 18/20 (90%) MCCs, we found it in 10/34 (29.4%) MCCs. The difference between the two studies could not be due to difference in techniques; it might be because FLI‐1 was carried out on whole sections, and because the samples were interpreted by three pathologists blinded to the tumour types being evaluated. After nuclear expression of FLI‐1 was scored by the pathologists, the samples were re‐evaluated in light of the discrepancy between Llombart's study and ours. In our results, the majority of MCC cases showed negative nuclear staining for FLI‐1 (fig 22).
We found FLI‐1 polyclonal antibody in a variety of tumours including EWS/PNET, ONB, small cell carcinoma of the lung, NHL (B and T cell), RMS, DSRCT, vascular tumours and MCB, and subsets of cases of MCC, SCC, adenocarcinomas, urothelial carcinomas and basal cell carcinoma. In conclusion, and despite the expression of FLI‐1 in a variety of malignancies, it is still considered to be highly specific and sensitive marker in distinguishing EWS/PNET from other tumours. However, we should be aware of its expression in the above tumour types whenever this marker is used.
We thank Charles LeVea, MD for his critical review of the manuscript. We also thank Mrs Joan Natiella for her histopathological skills, Mr Tim Dolan for his help in searching the archives and Mr Doug Nixon for his illustration expertise.
DSRCT - desmoplastic small round cell tumour
EWS/PNET - Ewing's sarcoma/primitive neuroectodermal tumour
FLI‐1 - friend leukaemia integration‐1
MCC - Merkel cell carcinoma
NHL - non‐Hodgkin's lymphoma
ONB - olfactory neuroblastoma
RMS - rhabdomyosarcoma
SRCT - small round cell tumour
Competing interests: None declared.