08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 107

Chapter

Soft Tissue Tumours: the Surgical Pathologist’s 8 Perspective

Roberto Salgado, Eric Van Marck

Contents Recently, using transcriptional gene profiling comple- mented with, e.g. tissue microarrays, new potential ther- 8.1 Introduction ...... 107 apeutic markers of histiotype-specific tu- 8.2 When Meets Radiology ...... 107 mours have emerged, e.g. CD117 (c-kit) and DOG1 for 8.3 When Pathology Meets the Clinics ...... 108 gastrointestinal stromal tumours, PDGFRb in dermatofi- 8.4 When Morphology is not Enough ...... 109 brosarcoma protuberans and EGFR1 for synovial sarco- mas, which in some cases might even be able to predict 8.5 World Health Classification of Soft Tissue Tumours 2002 ...... 111 therapy response and prognosis [1–3]. These advances have led to an increased appreciation of the importance 8.6 To Grade or not to Grade? ...... 112 that histological typing might have in predicting natural 8.7 Reporting Soft Tissue Tumours ...... 114 history and treatment sensitivity of soft tissue tumours. 8.8 Gene Expression Profiling Tumour tissue procurement and exact tissue han- and Tissue Microarrays in Soft Tissue Tumours . . . . . 114 dling are critical issues to provide the pathologist with 8.9 Conclusions and Future Perspectives ...... 115 adequate and well preserved tissue, in order to perform References ...... 115 the necessary ancillary studies for making an exact di- agnosis. Cross-talk between pathologists, radiologists and clinicians is crucial in obtaining this goal.

8.2 When Pathology Meets Radiology 8.1 Introduction The importance for the pathologist of both clinical and The diagnosis of soft tissue tumours has always been radiological information cannot be overemphasised. difficult and controversial. This was mainly due to the Some tumours merely occur in children, e.g. rhab- rarity of these lesions, comprising <1% of all malignant domyosarcomas, while others are rare in children and tumours. Nevertheless, according to the Surveillance, are mostly found in adults, e.g. pleiomorphic liposarco- Epidemiology, and End Results (SEER) database, about ma [4]. 15,000 cases of both bone and soft tissue tumours are Magnetic Resonance Imaging (MRI) and Computed diagnosed annually in the US. This puts soft tissue and Tomography (CT) give the pathologist information bone tumours in the same order of magnitude as about the size, consistency, homogeneity, location, pres- myeloma, gliomas and cervical carcinoma. ence or absence of necrosis, infiltration pattern, inter- Previously, the clinical importance of the exact histo- face with adjacent structures, etc. logical typing of soft tissue tumours by the pathologist Anatomic pathologists rendering a diagnosis on was limited, knowing that grade and stage, and not his- mostly limited material without having the advantages tological type determined treatment options. The of clinical and radiological information might be in a pathologist merely served to grade the lesion and to ex- disadvantageous position in many cases. It is imperative clude malignant non-sarcomatous, benign or pseudotu- that pathologists with a special interest in soft tissue tu- moral lesions. The small number of lesions – in particu- mours do at least have a rudimentary knowledge of the lar of single histological types – and the poorly defined radiological and clinical features of the tumours, in pathological criteria are the major sources of discor- analogy with pathologists interested in bone tumours. dances in diagnostic soft tissue pathology. Even in spe- On the other hand, it is equally important to have a ra- cialist centres about 5–10% of soft tissue lesions remain diologist and clinicians with a keen interest in soft tis- unclassified. sue tumour pathology. 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 108

108 Roberto Salgado, Eric Van Marck

Some soft tissue tumours such as liposarcomas are sion biopsies should only be performed on superficial quite heterogeneous. Therefore, the pathologist might or subcutaneous tumours which are usually smaller aid in choosing which area of the tumour is best suited than 2–3 cm in size. Excision biopsies of larger and/or for sampling, considering that sampling in areas of deep seated masses are best avoided. overt necrosis is best avoided. Sampling near necrotic Open, incisional biopsies are performed in most cen- areas, however, can be meaningful since the presence or tres. Open biopsies enable the pathologist to have absence of necrosis is considered an integral part of enough tissue, not only with a preserved architecture of grading schemes [5]. Meanwhile, in analogy with bone the tumour which is crucial for histological subtyping, tumours, e.g. osteosarcoma and chondrosarcoma, but also for performing ancillary studies, e.g. immuno- imaging features might provide the pathologist with histochemistry, cytogenetic and gene expression profil- spatial features of soft tissue lesions that are highly sug- ing-studies, when deemed necessary. gestive of being the high grade areas of the tumour. For Recently, more emphasis is being placed on the abili- example, chondrosarcomatous tumours having myxoid ty of pathologists to make a reliable diagnosis on rela- areas juxtaposed to cartilaginous areas are highly sug- tively smaller amounts of tissues. This is mainly due to gestive of being of high grade malignancy [6]. Sampling economic reasoning and patient comfort. Core Needle in these areas might be necessary for arriving at an ex- Biopsies (CNB) and Fine Needle Aspiration Biopsies act diagnosis, meanwhile trying to avoid underestima- (FNAB) are temporarily quite popular [7, 8]. The main tion of the malignant potential of the tumour when advantages of both these methods include the feasibili- sampling of non-high grade areas occurs. Sampling is ty in performing these kinds of biopsies in outpatient also imperative for performing additional studies such clinical settings, the minimal risk of complications, the as cytogenetic analysis, as will be explained later. cost-effectiveness and the acceptability of the patient. Estimating the percentage of necrosis preoperatively Moreover, the reduced risk of spilling tumour cells dur- by the radiologist might also be important, since some ing the biopsy procedure and the lack of compromising tumours,e.g.Ewing’s sarcoma and extra-renal rhabdoid the subsequent surgical management are considered tumours, are treated preoperatively, and the pathologist additional features in favour of FNAC or CNB vs open is asked to estimate the percentage of viable tumour tis- biopsy.Spilling of tumour cells and compromising here- sue left after treatment, as a measure of tumour re- with the surgical bed and subsequent therapeutic mea- sponse. This is quite elaborate and subjective since it is sures are often the case when open, incisional biopsies quite impossible, histologically, to distinguish between are performed by inexperienced surgeons. This urges spontaneous or therapy-induced necrosis. that for soft tissue lesions suspicious of being malig- Close collaboration between radiologists and pathol- nant, biopsy procedures should be decided on after in- ogists is important, not only to decide where to sample tense cross-talk between pathologists, radiologists and but also in determining how to sample. clinicians, and should preferable be performed in refer- ence centres. Nevertheless, despite the above-mentioned obvious 8.3 When Pathology Meets the Clinics advantages of sampling limited material, this does not imply that pathologists are willing or able to make a Once the lesion is localised, obtaining a biopsy in the specific diagnosis on small amounts of material. Limit- appropriate tumour area is the next logical step. Ideally, ed sampling might make grading and exact subtyping the biopsy should occur in the centre where the patient quite difficult. Moreover, preoperative chemo- and/or will be treated. It is not surprising that a pathologist of- radiotherapy for some tumours, e.g. Ewing sarcoma, ten requests larger tumour samples or asks for a second might make adequate and reliable histological typing opinion when being confronted with a soft tissue tu- and grading evaluation of resected specimens extreme- mour. This is mainly due to the rarity of the lesions, the ly difficult, if not impossible. lack of adequately defined diagnostic criteria and the Preoperatively, with limited material, especially after “unearthing” of new entities. There are some 50 differ- FNAC, the pathologist is in some cases only able to pro- ent subtypes of soft tissue tumours. Considering the vide diagnoses such as “malignant soft tissue tumour, rarity of the lesions and the difficulties associated in ob- not otherwise specified” or “spindle” or “round cell” tu- taining an adequate biopsy from representative tumour mour.This would imply in some cases that it would even areas, which can only occur once the tumour has been not be possible to know exactly which tumour was investigated thoroughly by adequate imaging means treated, since preoperative therapy might distort mor- (MRI and CT), biopsy procedures in the primary care phology considerably. settings should be discouraged. Biopsy procedures Deep seated lesions, however, are amenable for CNB, without having adequate radiological information giv- avoiding the discomfort of open, incisional biopsy in en by CT or preferable by MRI, should be avoided, un- these specific circumstances. FNAC is indicated mostly less the lesion is small and superficially located. Exci- for assessing local recurrence or metastasis or in cases 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 109

Chapter 8 Soft Tissue Tumours: the Surgical Pathologist’s Perspective 109

where incisional biopsy is difficult to perform. Never- ples stored in culture media for subsequent cytogenetic theless, in some specialised centres, with extensive ex- analysis and fixing tissue in glutaraldehyde for electron pertise in soft tissue tumour pathology, reliable subtyp- microscopic investigation are necessary when morpho- ing might be possible in most of the (untreated) cases logic features on haematoxylin and eosin slides are using FNAC or CNB. Similarly, distinguishing between deemed insufficient. Tumour tissue stored in this man- benign and malignant lesions can be done in these spe- ner is never wasted and can be consulted later on for di- cialist centres, which, as mentioned before, is currently agnostic and research purposes, when appropriate. considered sufficient to initiate treatment. Failure of ex- Yet, in most cases the pathologist does not have the act subtyping does not in all instances withhold initiat- luxury of having enough material to be stored in a way ing therapy, since quite limited information is needed as mentioned before. by the clinician. This information might be rendered by might be beneficial in these limited samples, with FNAB yielding less material than circumstances. Immunohistochemistry is the use of an- CNB. Obviously, when both FNAB and CNB are inade- tibodies specific to epitopes located on the cells of inter- quate, an open and incisional biopsy is mandatory. Nev- est.A range of immunohistochemical markers are avail- ertheless, an open, incisional biopsy is recommended in able nowadays. These antibodies should be applied ac- centres without explicit experience in FNAC or CNB. cording to the morphology of the tumour (Table 8.1), As already mentioned, in this era of targeted therapy, which should enable the pathologist to arrive at an ac- histological subtyping is crucial. Since imaging modali- curate diagnosis. The rationale of the use of these anti- ties cannot reliably distinguish between benign and bodies lies with the assumption that antigens expressed definitely malignant lesions, morphology has the last by the tumour cells are the same as those expressed by word. This can only be accomplished in a reliable man- their normal counterparts. For example, myogenin and ner with substantial material. MyoD-1 seem to be specific for muscle cells, since foetal muscle cells do express these proteins. Similarly, blood vessels are CD31 and CD34 immunoreactive (Figs. 8.1 8.4 When Morphology is not Enough and 8.2), and consequently their tumoral counterparts, the angiosarcomas, are immunoreactive for both these Considering the increased importance of histological proteins. In a number of cases, however, this simplifica- typing and the presence/absence of molecular markers tion does not hold. Some soft tissue tumours, e.g. an- for diagnostic, predictive or prognostic purposes, ade- giosarcomas might have an anomalous expression of quate handling of biopsy material is mandatory. proteins pointing to another lineage, e.g. cytokeratin. It is imperative that radiologists and clinicians need For example, only two soft tissue tumours do have to be aware of not only the total amount of material they “true” cytokeratin expression, namely epitheloid sarco- (need to) sample, but also on the handling of soft tissue ma and synovial sarcoma. These two tumours have so tumour tissue by the pathologist. The tissue needs to be far an undetermined origin. The example of the syn- transported fresh in an unfixed state to the pathology ovial sarcoma is interesting in another respect. Al- department. This allows an accurate and exact orienta- though the name might imply a synovial origin, most tion of the specimen by, ideally, both the surgeon and synovial tumours do not arise in or near joints. Using the pathologist who is going to diagnose the lesion. This transcriptional gene profiling experiments it has been cross-talk is also important for assessing the often com- shown that these tumours might have a neuroectoder- plex resection margins of soft tissue tumour specimens. mal (stem cell?) origin [9]. Another example illustrative Resection margins, be it other than fascia or perios- of the incongruent relation between a normal cell and teum, less than 1 cm are associated with a substantial its tumoral counterpart are the rhabdomyosarcomas. risk of recurrence, prompting additional treatment Some rhabdomyosarcomas are located in bladder and such as. e.g. radiotherapy or wider surgical excision. gall bladder, yet in these organs no striated muscle cells When the tissue arrives at the pathology department, are found. Expression of a single marker should there- critical decisions are made with respect to how the tis- fore not be considered in all cases as indicative of the sue is handled.With limited amounts of material, as ob- origin of a specific tumour. The example mentioned be- tained by FNAC and CNB, the pathologist is deemed to fore of the synovial sarcoma might also be applied to include all tissue for, e.g. formalin fixation and subse- the gastrointestinal soft tissue tumours (GIST), of quent paraffin embedding. Yet, as mentioned before, which a plethora of publications have arisen during the histological typing of the tumour should be the main past years. One of the criteria to diagnose GIST-tu- objective of the pathologist. With small amounts of tis- mours is the presence of the membrane protein c-kit sue, it is sometimes quite difficult, even impossible to (Figs. 8.3 and 8.4). Since this protein is detected on the asses the of the tumour. Moreover, differenti- interstitial cells of Cajal in the intestine, it has been ating between benign and low grade malignant lesions speculated that these cells might be the origin of these is not always possible. Snap-frozen tissue, tissue sam- tumours. However, some tumours are c-kit negative and 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 110

110 Roberto Salgado, Eric Van Marck

Table 8.1. Immunohistochemical panel according to the morphology of the tumour Immunohistochemical panels Spindle cell tumours Cytokeratin, epithelial membrane antigen (EMA), desmin, smooth muscle actin (SMA), h-caldesmon, S-100 Round-cell tumours CD99, myogenin, desmin, cytokeratin, EMA, synaptophysin, FLI-1 Epitheloid tumours Cytokeratin, EMA, CD34, CD31, S-100, FLI-1

Fig. 8.1. Angiosarcoma (magnification: ¥200) Fig. 8.2. Angiosarcoma tumour cells staining for CD31 (magnifi- cation: ¥200)

Fig. 8.3. Gastrointestinal stromal tumour (GIST) (magnification: Fig. 8.4. GIST-tumour cells staining for CD117 (c-kit) (magnifica- ¥200) tion: ¥200)

some do not at all arise in the gastrointestinal tract, e.g. Considering that GIST-tumours were previously cat- the extra-gastrointestinal GIST. C-kit expression in egorised as being of smooth muscle cell or neural ori- most of these cases is related to an acquired gain of gin, the comments above illustrate that identification of function mutation of the c-kit gene. C-kit expression in genetic alterations in histiotype-specific tumours is able GIST-tumours is also related to the therapeutic effec- to change the morphological classification of soft tissue tiveness of imatinib mesylate (Glivec). Intriguingly, tumours. GIST-tumour cells expressing the protein without hav- If sufficient tissue is available, additional investiga- ing the corresponding mutation have less response to tions can be performed such as cytogenetic analysis and imatinib mesylate (Glivec) therapy. electron microscopic studies. 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 111

Chapter 8 Soft Tissue Tumours: the Surgical Pathologist’s Perspective 111

Cytogenetic analysis is the identification of specific 8.5 World Health Classification molecular alterations thought to be specific for certain of Soft Tissue Tumours 2002 [10] tumours. This concept will be discussed elsewhere by A. Sandberg. Once a specific diagnosis is reached the lesion should be Performing electron microscopic investigations for categorized accordingly. Recent advances in cytogenetic soft tissue tumours is not routine in most pathology de- and molecular genetic information, the recognition of partments. This might be due to the pathologists them- new entities and the enhanced understanding of the bi- selves who are unwilling to embrace and further devel- ological behaviour have prompted a new morphological op or invest in “old” techniques such as electron mi- classification of soft tissue tumours. croscopy. In any branch of medicine any technique It is recommended to categorise the lesions in a be- which might offer additional clinical information is dif- nign, intermediate (locally aggressive), intermediate ficult to be discarded,in contrast to electron microscopy (rarely metastasizing) or a malignant category: for soft tissue tumour diagnostics. 1. Benign lesions: e.g. lipoma. These tumours mostly do It is argued that electron microscopy does not add not recur and if recurrence occurs it is most of the much to the diagnosis of soft tissue tumours. time in a non-destructive fashion. In exceptional cir- Specific examples highlight the importance electron cumstances metastasis might occur, as in e.g. benign microscopy might have in soft tissue tumours: fibrous histiocytoma. 1. Distinguishing between malignant peripheral nerve 2. Intermediate (locally aggressive): e.g. kaposiform sheath tumours and monophasic synovial tumours. haemangioendothelioma and desmoid fibromatosis. Monophasic synovial sarcomas are much more These tumours are infiltrative and might recur local- chemosensitive than malignant peripheral nerve ly, sometimes in a destructive fashion when incom- sheath tumours. pletely excised. 2. Distinguishing between mimics of soft tissue tu- 3. Intermediate (rarely metastasizing): e.g. retiform mours such as sarcomatous carcinoma and metastat- haemangioendothelioma. These tumours are locally ic melanoma. aggressive, recur if incompletely excised and may 3. Diagnosing alveolar soft tissue sarcoma in an aspe- give rise to metastasis. Histological features are not, cific clinical or metastatic setting. in most cases, predictive for metastasis. 4. Malignant: e.g. round cell liposarcoma. These tu- All the above-mentioned features highlight a few issues: mours are infiltrative, recur locally if incompletely 1. The pathologist needs to have sufficient material to excised and have a significant risk of metastasis. perform the ancillary studies which he seems neces- sary to arrive at a diagnosis as accurate as possible. Significant changes have occurred in the new WHO- 2. Without sufficient material it is difficult to distin- classification compared with the previous classification guish between benign, low grade and sometimes published in 1994. Major adaptations include: even high grade malignant lesions. 1. Adipocytic tumours: new entities include myolipoma 3. Ancillary immunohistochemical techniques might and chondroid lipoma. Four types of liposarcoma oc- help the pathologist in arriving at a diagnosis. cur: 4. The expression of molecules on soft tissue tumour a) Atypical lipomatous tumour/ well differentiated li- cells does not always imply a corresponding normal posarcoma: these terms are synonyms, although counterpart. the term well differentiated liposarcoma is prefer- 5. The expression of molecules on tumour cells, is able in retroperitoneal and paratesticular areas, sometimes not merely diagnostic, but might in some due to its higher morbidity and mortality in these cases also have prognostic and predictive informa- specific areas. tion. b) Dedifferentiated liposarcoma: previously were 6. It might be argued that the classification of soft tissue these tumours categorised as being definitively tumours by the corresponding normal counterpart of high grade. In the current classification a low- the tumours should be abandoned, considering the grade variant is included. These tumours, although anomalous expression of proteins by some tumour being of the dedifferentiated type, do have a more cells. Some (all?) soft tissue tumours seem to arise protracted clinical course compared with other from somatic stem cells, and not from its normal dedifferentiated soft tissue tumours. counterpart. c) Myxoid liposarcoma: both round-cell liposarcoma and myxoid liposarcoma are considered to be one entity, with the round-cell liposarcoma type being at the less dedifferentiated spectrum of this tu- mour. 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 112

112 Roberto Salgado, Eric Van Marck

d) Pleiomorphic liposarcoma: these are the rarest of nodular fasciitis. Since extra-skeletal myxoid lipomatous tumours and are considered to have a chondrosarcoma does not seem to have chondroid worse prognosis. differentiation, this lesion is placed in the category 2. Fibroblastic/myofibroblastic tumours: keloid is not “tumours of uncertain differentiation”. longer regarded as a soft tissue tumour. New benign entities include, e.g. giant cell fibroblastoma, Gardner fibroma and angiomyofibroblastoma. New malignant 8.6 To Grade or not to Grade? entities include, e.g. sclerosing epitheloid fibrosarco- ma and low-grade fibromyxoid sarcoma. Important- The purpose of grading is to segregate tumours in those ly, the term “hemangiopericytoma” is included in the with a favourable prognosis, amenable to surgical exci- chapter of the solitary fibrous tumours considering sion only, and in those with an unfavourable prognosis, the lack of morphological dissimilarities between amenable to surgical treatment and additional chemo- these two tumour types. therapeutic and/or radiotherapeutic treatment. More- 3. Fibriohistiocytic tumours: pleiomorphic malignant over, adequate grading is considered to be more repro- fibriohistiocytic tumour (MFH) is considered to be a ducible than histological typing, might even circumvent high grade undifferentiated tumour. diagnostic disagreement between pathologists and 4. Smooth muscle tumours: inclusion of GIST-tumours might triage unclassified sarcoma in a low or high risk in different organs, e.g. in the gastrointestinal tract is category. As a matter of fact, most contemporary oncol- considered in the respective WHO-volumes. ogists triage their patients for subsequent treatment by 5. Pericytic/perivascular tumours: only glomus and merely relying on the tumour grade provided by the myopericytoma are included in this category. pathologist. Relying in a way that grading is in all cir- 6. tumours: the embryonal, alveolar and cumstances able to distinguish between benign, low pleiomorphic subtypes are included. grade and high grade lesions, irrespective of histiotype. 7. Vascular tumours: in this category is only the epith- Currently, two grading classifications exist: the eloid haemangioendothelioma included as a malig- National Cancer Institute (NCI)-grading system nant subtype of the haemangioendotheliomas. (Table 8.2) and the Fédération Nationale des Centres de 8. Chondro-osseous: myositis ossificans and fibro-os- Lutte Contre le Cancer (FNCLCC)-grading system, the seous pseudotumours are considered to be variants so called “Trojani system” (Table 8.3). The FNCLCC-

Table 8.2. National Institute of Cancer grading system Common histological types Grade 1 Grade 2 Grade 3 Well-differentiated liposarcoma Pleiomorphic liposarcoma Alveolar rhabdomyosarcoma Myxoid liposarcoma Fibrosarcoma Soft tissue osteosarcoma Deep-seated dermatofibrosarcoma protuberans Malignant fibrous histiocytoma Primitive neuroectodermal tumour Some leioymyosarcomas Alveolar soft part sarcoma Epitheloid haemangioendothelioma Synovial sarcoma Mesenchymal chondrosarcoma Spindle cell haemangioendothelioma Leiomyosarcoma Infantile fibrosarcoma Neurofibrosarcoma Subcutaneous myxofibrosarcoma or or 0–15% necrosis >15% necrosis

Table 8.3. Fédération Nationale des Centres de Lutte Contre le Cancer grading system Differentiation score Mitosis score (per 10 HPF) Necrosis Score 1 Sarcomas resembling adult mesenchymal tissue 0–9 No necrosis Score 2 Sarcomas of certain histiotype >9; <20 <50% necrosis Score 3 Embryonal/undifferentiated sarcomas 20 or more >50% necrosis and sarcomas of uncertain histiotype

HPF: high power field. The total scores for each grade are as follows: grade 1, 2–3; grade 2, 4–5; grade 3; 6–8 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 113

Chapter 8 Soft Tissue Tumours: the Surgical Pathologist’s Perspective 113

grading system is considered to be more reproducible giosarcomas. Some tumours are definitively of low ma- and more apt to classify patients in either grade 1 or lignant potential, e.g. infantile fibrosarcoma and well grade 3, avoiding more frequently the inclusion of pa- differentiated liposarcoma. Some tumours are not yet tients in the “indeterminate” grade 2 category, com- amenable for grading, e.g. malignant giant cell tumour pared with the NCI-system. Both systems are included and hemangiopericytoma. Conversely, some tumours in the latest WHO Soft Tissue Tumour Classification. are amenable for grading, e.g. leiomyosarcoma, myxoid Both are three-tiered systems, assigning different liposarcoma and fibrosarcoma. weights to histiotype, necrosis, mitotic figures, cellulari- When looking at both grading systems it is obvious ty, pleiomorphism and differentiation. that some subjectivity in applying these parameters is Grading should be performed on untreated tumours unavoidable. Although, e.g. mitotic figure counting on representative and well preserved tissue. This should should be performed rigidly in soft tissue tumours, the be the case for all tumours, be it sarcomatous or non- inter-observer reproducibility, equally as e.g. in breast sarcomatous. Previous therapy, whether be it chemo- carcinoma, is in the range of 60–70%, even among soft therapy or radiotherapy, precludes grading in all cir- tissue specialists. Cellularity, mitotic figure counting cumstances since both mitotic figures and the extent of and, e.g. extent of necrosis are all influenced by sam- necrosis can be affected. It is also well known that soft pling, fixation, section thickness, the number of blocks tissue tumours acquire a higher grade in local recur- taken and microscopic field size. Also, as previously rence or metastatic lesions. Grading of visceral sarco- mentioned, having small amounts of material precludes mas should also be avoided. grading in most cases. Considering the intra-tumoral It is a genuine fact that some soft tissue tumours are heterogeneity of soft tissue tumours, sampling error is not amenable at all to exact histological grading. Histo- likely to occur, especially when trying to avoid highly logical typing does not in all cases provide the neces- necrotic areas. It is also difficult to asses macroscopical- sary prognostic and/ or predictive information. Some ly the extent of necrosis since myxoid and hyalinized ar- soft tissue tumours should not at all be graded, especial- eas might mimic necrosis. There is also no consensus as ly considering the inherent malignant potential of these whether the mere presence or the extent of necrosis de- tumours. According to the WHO 2002 Classification of termines prognosis. Similarly, how to measure necrosis, Soft Tissue Tumours grading should not be performed micro- or macroscopically, is not at all defined. on malignant peripheral nerve sheath tumours, an- Counting mitoses is, as previously mentioned, inher- giosarcomas, extraskeletal myxoid chondrosarcoma, ently subjective. Distinguishing mitotic figures from py- alveolar soft part sarcoma, clear cell sarcoma and epith- knosis is sometimes difficult. There is also no universal eloid sarcoma. consensus on which cut-off and microscopic field size to Considering both grading systems it is quite clear use. that some histological typing is crucial for grading soft Some have argued that intra-tumoral (lympho)vas- tissue tumours. As previously mentioned, exact histo- cular invasion should also be searched for. In analogy logical typing should be the main aim of the patholo- with epithelial tumours, there is considerable difficulty gist,yet,it is not always possible,as is the case when hav- in distinguishing artefactual retraction from “true” ing only small amounts of tissue. On the other hand, if a lymphatic or blood vessels. Immunohistochemistry pathologist wants to use one of the grading systems, an might be helpful in this respect. attempt at histological typing is necessary, despite the In the FNCLCC-system differentiation is one the pa- inherent difficulties of soft tissue diagnostics. It is even rameters to asses. As previously mentioned, no normal elaborated that both tumour type and tumour size are counterpart exists for some tumours, e.g. epitheloid able to adequately predict survival,withholding grading sarcoma. in these cases [11]. Also, some tumours, e.g. GIST-tumours do have their Mitotic counts are inherently associated with the own grading system including tumour size, mitotic fig- FNLCC-grading system. However, as mentioned before, ure counting and location [12]. Here also, estimating histological typing is mandatory because some tumours mitotic figure counting is subjected to the same inher- behave in a benign fashion despite having many mi- ent subjectivity as mentioned before. Size should be as- toses, e.g. cellular schwannoma and atypical fibroxan- sessed by the pathologist on a fresh resected specimen. thoma. Some tumours are considered low grade despite It is argued that the FNCLCC- and the NCI-grading the presence of mitotic figures, e.g. infantile fibrosarco- system should be restricted to adult tumours, consider- ma. Conversely, some tumours are inherently malignant ing the unique features of childhood soft tissue tu- despite the lack of mitotic figures and/ or necrosis, e.g. mours. For example, childhood rhabdomyosarcoma- the round-cell liposarcoma. tous tumours are graded according to the system pro- Grading systems can thus not be meaningfully ap- vided by the Intergroup Rhabdomyosarcoma Study plied to all soft tissue tumours.Some tumours are defin- Group where risk assignment, rather than grading is itively malignant, e.g. Ewing sarcoma/MPNET and an- performed by a combination of both exact histology 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 114

114 Roberto Salgado, Eric Van Marck

and the location of the tumour. A modified TNM stag- 8.8 Gene Expression Profiling ing system has even developed for these tumours, with and Tissue Microarrays in Soft Tissue Tumours sites being defined as unfavourable (e.g. bladder, prostate) and favourable (e.g. orbit, head and neck) Soft tissue tumours are heterogeneous, both with re- [13]. spect to the morphology and the clinical behaviour. Al- Importantly, currently no method, e.g. DNA flow cy- though grade and histiotype currently determine treat- tometry, other than histological subtyping and grading ment options and prognosis, prediction of clinical be- provides more prognostic information in soft tissue tu- haviour remains elusive in most circumstances, mours. It is clear that there is no single universal para- especially in individual cases. Tumour growth, even meter applicable to all tumours. In ideal circumstances within the same subtype, ranges from slow growing should grading systems be developed and applied for with low risk of metastasis, to fast growing lesions with each histological subtype. considerable morbidity and mortality. This precludes adequate morphological classification. Furthermore, progress in the adequate identification 8.7 Reporting Soft Tissue Tumours (Table 8.4) of molecules that might be useful for therapeutic target- ing is hampered by the examination of only a small set A template for reporting soft tissue tumours should in- of genes by the more conventional methods such as, e.g. clude the following: Northern and Southern blotting. 1. Histological type. Gene expression profiling might be useful in this re- 2. Grade (1–3) or risk assessment, when applicable. spect. The basis of microarray studies is the labelling of 3. Size, assessed by the pathologists on fresh, unfixed cDNA of the tumour with a fluorescent dye. This la- tissue. belled cDNA is hybridised together with a reference 4. Location, cross-talk with radiologist and surgeon is fluorescent labelled cDNA onto a microarray chip. The crucial, especially when trying to identify infiltration slide is scanned after hybridisation on a gene specific in critical structures, e.g. blood vessels and nerves. probe and the ratio values of the intensity of the sample 5. Margins, especially mentioning those margins <1 cm cDNA vs the reference cDNA is measured and calculat- and whether there is a surrounding fascia or perios- ed. Thousands of mRNA transcripts can be analysed in teum. this manner. 6. Necrosis, if assessed macroscopically it should ideal- A number of different array technologies are avail- ly be on fresh, unfixed and untreated tissue. able, each with its advantages and disadvantages. A 7. Additional investigations, e.g. karyotyping. main review of all aspects of gene expression profiling is beyond the scope of this review. These are covered more extensively in excellent reviews ([14] and references therein). Table 8.4. Template for soft tissue tumour reporting Sample preparation is crucial for these kinds of ex- Template periments. This method measures the abundance of a) Histological type gene transcripts (mRNA). One caveat is the fragility of b) Grade (1–3), or risk assessment mRNA by its rapid degradation by RNAses present in c) Size cells and tissues. Thus, it is crucial that the tissue is stored as rapid as possible at –80 °C when sampling the d) Location tissue. Although formalin fixed and paraffin embedded e) Margin status tissue might also be useful for RNA extraction, this is f) Necrosis more laborious and the risk of not having substantial h) Additional investigations RNA is considerable. This is mainly due to the slow in- activation of RNAses by the fixative. Gene expression profiling has revealed novel sub- types in other tumours, e.g. in breast cancer and malig- nant lymphoma. An illustrative example of the applica- tion of gene expression profiling in soft tissue tumours include a study performed by Nielsen and colleagues. Unsupervised analysis of a set of 5500 genes demon- strated that synovial sarcomas, GIST-tumours, malig- nant peripheral nerve sheath tumours and some leiomyosarcomas clustered in separate groups, in sup- port of the morphological classification of these tu- mours. When applying a supervised analysis, several 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 115

Chapter 8 Soft Tissue Tumours: the Surgical Pathologist’s Perspective 115

transcripts were found to be specifically upregulated in for ancillary studies, such as karyotyping and electron the respective subgroups. For GIST-tumours this in- microscopy. Pathologists should aim at trying to histio- cluded c-kit, for synovial sarcomas EGFR-1 and SSX- type and grade soft tissue tumours, whenever consid- transcripts were upregulated [15]. ered adequate. Clinicians need to be aware that grade Importantly, the identification of these transcripts in does not replace histology. It is mandatory that soft tis- specific tumour histiotypes prompted the development sue tumoral specimens should always be stored in liq- of agents directed specifically at these molecules (Glivec uid nitrogen for future diagnostic and research purpos- directed at c-kit positive GIST-tumours and Iressa at es. Gene expression profiling and tissue microarrays re- EGFR-1 positive synovial sarcoma). Clinical trials are in fine the morphological classification and are currently progress. the best ways to identify new diagnostic, prognostic and Moreover, using gene expression profiling it has also predictive molecules in soft tissue tumours. Clinical tri- been shown that synovial sarcomas clustered on the one als with target specific compounds are ongoing. Cen- hand separately from other tumours, but on the other tralised pathological review is considered critical in hand clustered closely to malignant peripheral nerve these circumstances. sheath tumours and expressed genes that are involved in the migration of neural crest cells, such as, e.g. ephrin-3 and endothelin-3, raising suspicion that syn- Things to remember: ovial sarcomas might be of neuroectodermal origin. 1. Pathological diagnosis of soft tissue tumours is Considering the plethora of data provided by gene difficult. Second opinion or even preferable a cen- expression profiling, many new markers have emerged tralised pathological review should be mandatory which need to be tested and validated on large tumour for difficult cases. sets. 2. Biopsy and diagnosis should be performed in the Tissue microarrays might be useful in this respect. centre were the patient will be treated. Tissue microarrays utilises 0.6-mm core needle biopsies 3. Cross-talk between radiologists, clinicians and of representative areas of the tumour on a paraffin pathologists should occur at all phases of patient block. This core is subsequently re-embedded in a nov- management. el paraffin block. With this method a large number of 4. Clinicians and radiologist need to be aware of the tumour samples, ranging to 1000 samples on a single amount of tissue the pathologist needs for an ac- slide, can be analysed at the RNA, DNA or protein level. curate diagnosis. The GIST-marker DOG-1 was evaluated and validated 5. Grading does not replace histology. in such a manner, following identification by gene ex- 6. Pathologists should aim at trying to histiotype pression profiling [1]. and grade soft tissue tumours. 7. It is mandatory that soft tissue tumoral tissue should always be stored in liquid nitrogen for 8.9 Conclusions and Future Perspectives future diagnostic and research purposes.

Soft tissue tumours are rare tumours. Pathological diag- nosis is difficult. This is mainly due to the poorly de- fined pathological criteria and the more than 50 differ- References ent histological subtypes. Second opinion or, even 1. West RB, Corless CL, Chen X, Rubin BP,Subramanian S, Mon- preferable, a centralised pathological review should be togomery K, Zhu S, Ball CA, Nielsen TO, Patel R, Goldblum JR, mandatory for difficult cases. The Belgian Soft Tissue Brown PO, Heinrich MC, van de Rijn M (2004) The novel marker, DOG1, is expressed ubiquitously in gastrointestinal Registry (BSTNR) is worth mentioning in stromal tumors irrespective of KIT or PDGFRA mutation sta- this respect. The BSTNR is composed partly of experi- tus. Am J Pathol 165(1):107–113 enced soft tissue tumour pathology- and radiology ex- 2. Blay JY,Ray-Coquard I,Alberti L, Ranchere D (2004) Targeting other abnormal signalling pathways in sarcoma; EGFR in syn- perts confirming or even providing an alternative diag- ovial sarcoma, PPAR-gamma in liposarcoma. Cancer Treat Res nosis in submitted cases. Biopsy and diagnosis should 120:151–167 be performed in the centre were the patient will be 3. McArthur G (2004) Molecularly targeted treatment for der- matofibrosarcoma protuberans. Semin Oncol 31(2 Suppl 6): treated. It is imperative that cross-talk between radiolo- 30–36 gists, clinicians and pathologists occurs at all phases of 4. Enzinger and Weiss’s soft tissue tumors. Fourth Edition (2001) patient management. Preoperatively, when choosing the 5. Guillou L, Coindre JM, Bonichon F,Binh Bui N, Terrier P,Collin F,Vilain M, Mandard A, Le Doussal V,Leroux A, Jacquemier J, biopsy technique and the area to sample and postopera- Duplay H, Sasttre-Garau X, Costa J (1997) Comparative study tively, at the bench side. Clinicians and radiologist need of the National Cancer Institute and French Federation of to be aware of the amount of tissue the pathologist Cancer Centers Sarcoma Group Grading Systems in a popula- tion of 410 adult patients with soft tissue sarcoma. J Clin Oncol needs for an accurate diagnosis. Adequate sampling is 15:350–362 mandatory, not only for diagnostic purposes, but also 6. Dahlin’s Bone Tumors. Sixth Edition (1996) 08_DeSchepper_Soft_Tissue 15.09.2005 13:22 Uhr Seite 116

116 Roberto Salgado, Eric Van Marck

7. Sing H, Kilpatrick S, Silverman J (2004) Fine needle aspiration 11. Evans HL (1995) Classification and grading of soft-tissue sar- biopsy of soft tissue tumours. Adv Anat Pathol 11(1):24–36 comas: a comment. Hematol Oncol Clin North Am 9:653–656 8. Gielen J (2004) Magnetic resonance imaging of soft tissue tu- 12. Miettinen M, El-Rifai W, Sobin HL, Lasota J (2002) Evaluation mours. A prospective study of malignancy and tumor pheno- of malignancy and prognosis of gastrointestinal stromal tu- type on MRI. Epidemiological studies and evaluation of char- mors: a review. Hum Pathol 33(5):478–483 acterization parameters. Thesis, University of Antwerp 13. Meyer W, Spunt S (2004) Soft tissue sarcoma of childhood. 9. Allander SV, Illei PB, Chen Y, Antonescu CR, Bittner M, Cancer Treatment Rev 30:269–280 Ladanyi M, Meltzer PS (2002) Expression profiling of synovial 14. Yang YH, Speed T (2002) Design issues for cDNA microarray sarcoma by cDNA microarrays: association of ERBB2, IGFBP2, experiments. Nat Rev Genet 3(8):579–588 and ELF3 with epithelial differentiation. Am J Pathol 161(5): 15. Nielsen TO,West RB, Linn SC, Knowling MA, O’Connel JX, Zhu 1587–1595 S, Fero M, Sherlock G, Pollack JR, Botstein D, van de Rijn M 10. Fletcher CDM, Krishnan U, Mertens F (2002) Pathology and (2002) Molecular characterisation of soft tissue tumours: a genetics. Tumours of soft tissue and bone.World Health Orga- gene expression study. Lancet 359(9314):1301–1307 nization Classification of Tumours