CD34 and Factor XIIIa-Positive Microvascular Dendritic Cells and the Family of Fibrohistiocytic Mesenchymal Tumors

To the editor: We read with interest and greatly enjoyed the article by Dr. Narvaez and colleagues (1), who studied the immunophenotype of CD34+ dermal dendritic cells. They concluded that CD34+ cells in the dermis represent uncommitted mesenchymal cells distinct from the other dendritic cell types in the skin, notably the factor XIIIa+ dendrophages or histiocytes. We wholeheartedly agree with this conclusion and applaud the thorough immunophenotyping of these cells performed by the authors. One of us (J.S.S.), with various colleagues, has advanced a theory that supposes interactions between the two types of microvascular dendritic cells: the undifferentiated CD34+ fibroblasts, which we believe serve as cellular raw material for tissue remodeling, and the often fibroblast-shaped dendritic histiocytes that express coagulation factor XIIIa (FXIIIa). Both cells first appear in the dermis and connective tissue anlagen during early embryogenesis (2-5). These two markers have been used to differentiate among dermal spindle cell tumors. This has been the thrust of most of the literature on this subject, and in practical terms this is where most of the attention should be focused, in aiding the practitioner to accurately diagnose these lesions. We have noted that CD34+ and FXIIIa+ putative "primitive mesenchymal cells" appear to interact in formation of a diverse group of dermal (6-9) and other fibrovascular soft tissue tumors (10-18) and believe that together these cells form an attractive unifying concept for the histogenesis of a broad array of visceral and soft tissue neoplasms including Kaposi's sarcoma (KS), CD34+ gastrointestinal stromal tumors (GIST), solitary fibrous tumor (SFT)/hemangiopericytoma, dermatofibrosarcoma protuberans (DFSP), and others (6-18). In a recent letter in these pages, we expanded these observations (19) to include a myxoid DFSP and two examples of myofibroblastic dermatofibroma, a variant thoroughly delineated by Dr. Zelger and his colleagues (20). We also discussed our case report of a CD34+ angiomatoid malignant fibrous histiocytoma (AMFH) similar to two CD34+ cases that we were happy to see reported by Dr. Morgan and his colleagues (21) in contrast to one recent large CD34-negative series of 97 AMFH (22). We appreciate the time and interest taken by Drs. Morgan (21A) and Zelger (20A) to reply to our letter and would like to take the opportunity to elaborate on some issues raised by them and to briefly review for the interested reader current knowledge of the basic science of CD34 and Factor XIIIa. Dr. Morgan's remarks regarding the importance of staining methodology with the QBend10 clone of CD34 were particularly relevant (21A). QBend10 is the clone we also use, and it is widely used by others. The epitope recognized by QBend10 and many other clones of CD34 antibodies are quite formalin- and processing- resistant (23), and it appears that they are best "left alone" without benefit of heat-induced epitope retrieval or incubations at elevated temperatures. Like Dr. Morgan, we too wonder about methodology when the results of a series do not jibe with our own observations. In a similar situation, we have studied numerous myxoid neoplasms and found that they are often strikingly composed of CD34+ dendritic fibroblasts admixed with FXIIIa+ histiocytes (6,10-15,18,19). Ockner et al. recently published a study of genital angiomyofibroblastoma and compared these with aggressive angiomyxoma and a diverse group of myxoid tumors, none of which, in their hands, marked with CD34 (24). However, the CD34 clone that they used is, as far as we can tell from manufacturer's lists, for use in flow cytometry and these workers employed microwave antigen retrieval, both of which might account for the lack of CD34 reactivity and disparity in our respective results in myxoid tumors. This again underscores the importance of carefully considering methodology when reporting results and comparing various series to study the biology of tumors and to establish diagnostic criteria for their recognition. We greatly enjoyed the description of dermatofibromas (DF) with myofibroblastic differentiation by Dr. Zelger and his colleagues (20) and believe that their emphasis on this potential diagnostic pitfall among spindle cell tumors is neither trivial nor oversimplified (20A), just taken from a different viewpoint-that of the diagnostic pathologist who must accurately diagnose and prognosticate on a lesion's behavior for the patient. We, on the other hand, unencumbered by the responsibility of making a diagnosis or of administering an academic department and training residents in all of the complexities of dermatohistopathology, are afforded the luxury of trying to dissect immunohistochemically what lesions cross our community hospital bench and to theorize regarding cellular interactions based on the basic science work of many other investigators. We are sorry that Dr. Zelger finds our descriptions of various mesenchymal tumors inscrutable or arcane, but are mindful of the fact that his perspective must of necessity vary from those of a histotechnologist with a histogenetic hobby. In attempting to conceptualize a histogenetic and nosologic scheme for a family of fibrohistiocytic tumors, and to perhaps define putative subsets of dermal stem cells, we have studied the seminal work of many investigators (1-3,23,25-28). As pointed out by the excellent immunophenotypic study of CD34+ dendritic cells by Narvaez and colleagues, the CD34+ fibroblast is an attractive candidate for a dermal stem cell (1). CD34 was named the human progenitor cell antigen and is expressed on hematopoietic precursor cells, many endothelia, and on ubiquitous dendritic fibroblasts in collagenous connective tissue (1,23,28). In hematopoietic cells, CD34 expression is associated with phenotypic plasticity and may be involved in maintenance and activation of multipotent stem cells throughout life. It is not implausible to think CD34 performs a similar function in stromal dendritic cells, constituting a marker for a stromal "reserve cell". CD34 expression disappears from hematopoietic cells with early lineage commitment (23,29), and a similar phenomenon occurs in fibroblasts when they attain a differentiated (e.g., myofibroblastic) phenotype (4,13,15,16,19). CD34's structure features a highly glycosylated extracellular mucin domain that could function as a receptor ligand, as well as an intracellular domain with cytoplasmic phosphorylation sites (23). In hematopoietic cells, engaging of CD34 is involved in signal transduction controlling cell adhesion and actin polymerization (29). To our knowledge this remains unexplored in stromal cells. Other than L-selectin on lymphocytes binding to high endothelial venule CD34, the ligands for CD34 remain unknown (30). Additionally, CD34's range of signaling mechanisms and pathways remain obscure (29). In matrix biology, the interactions between stromal cells and extracellular matrix ligands and growth factors are complex. In stromal dendritic cells, it appears likely that CD34 could have numerous ligands in the extracellular matrix corresponding to its multiplicity of glycosylation patterns. This could result in different signaling cascades resulting in the various CD34+ fibroblast phenotypes. For wound repair, such putative stromal reserve cells would need to be able to elaborate first a provisional or myxoid matrix and must later attain a myofibroblastic and finally a (keloidal) collagenous fibroblastic phenotype. This is borne out in the spectrum of CD34+ fibrovascular spindle cell tumors recapitulating such tissue response patterns within and among themselves. In many of these tumors, including DFSP, GIST, SFT, and KS, clonal proliferations of CD34+ spindle cells persist and overtake and outnumber the FXIIIa+ cells. Overexpression of the apoptosis inhibitor Bcl-2 recently has been consistently demonstrated in many of the CD34+ spindle cell lesions (31). Why CD34 disappears from more committed cells remains a mystery; as Aiba et al. point out, CD34 is also lost after a number of cell cycles, and this has hampered the study of CD34+ dendritic cells in vitro (32). This explains the absence of CD34 reactivity in all cultured spindle-shaped cells explanted from normal dermis, DFSP, KS, and DF recently reported by Bonish at al. (33). Accordingly, we believe that trying to correlate in vivo neoplastic tissue patterns with growth factor, matrix ligand, receptor, and cytokine profiles in CD34+ fibroblast and factor XIIIa+ histiocytes may be at least as fruitful in understanding dermal biology as in vitro methods. This is currently technically difficult in all but frozen tissues, but hopefully modern catalyzed reporter deposition signal amplification methods might soon help us to use archival paraffin blocks to localize these labile substances in neoplastic and normal cells. The FXIIIa+ cells are often discounted as "reactive immune histiocytes". Dr. Zelgers' remarks rightly emphasized the temporal flux of cellular populations and phenotypic transformations in dynamically evolving dermal tumors, a subject on which we have speculated often. Normal adult dermis has widely scattered FXIIIa+ histiocytes and these cells greatly accumulate in tumors and remodeling tissues. We now believe that these FXIIIa+ cell populations arise mainly because of autocrine (dermal FXIIIa+ cells) and/or paracrine (fibroblast) chemotaxis of peripheral blood monocytes that proliferate and differentiate into FXIIIa+ dendrophages locally. They do this, we believe, mainly to effect phenotypic and functional changes on the accompanying fibroblasts and their matrix rather than to perform any antigenpresenting function which has been the main function often ascribed to them (27). We have shown FXIIIa+ dendritic cells cycling in a variety of neoplasms (6-9,11,15,16,19), but not in reparative lesions. We wonder conceptually, at least in certain neoplasms, whether some FXIIIa+ cells might not differentiate directly from CD34+ fibroblast-like precursor cells just as the monocyte/macrophage lineage arises from CD34+ hematopoietic cells in the bone marrow. We have found clear-cut coexpression of CD34 and FXIIIa thus far only in one hemangiopericytoma (17) and an overlap of reactivity suggestive of such a lineage relationship in other lesions (11,15). FXIIIa+ cells also express certain reticuloendothelial markers suggesting that they may be involved in signaling between the immune system and the mesenchyme. FXIIIa+ cells express the CD4 receptor and may be targets of infection for human immunodeficiency virus (27). Gray and associates observed that FXIIIa+ cells are especially numerous in patch and plaque lesions of Kaposi's sarcoma (KS), whereas in later nodular KS tumors the FXIIIa+ cells are overtaken by a proliferation of CD34+ spindle cells that show varying endothelial specialization (34). It is tempting to speculate that stimulation by FXIIIa+ cells is somehow in part responsible for the development of clonal proliferations of CD34+ dendritic cells. We also feel that this CD34+ dendritic cell is perhaps a more attractive "cell of origin" for KS than endothelium. In earliest patch and plaque KS, it is the dermal microvascular adventitial cells (CD34+ fibroblasts and FXIIIa+ histiocytes) that appear activated, not vascular endothelium (unpublished observations). We thus believe that KS basically is a multifocal immunemodulated angioformative fibrohistiocytic proliferation. It has been said that a tumor is a wound that won't heal. Without cell spreading and adhesion, wounds don't heal and many tumors don't grow. FXIIIa cross-links various matrix proteins such as fibrin, fibronectin, and collagens (35). Fibronectin forms a conduit upon which fibroblasts spread and migrate during tissue remodeling and during tumor growth. It has been shown that proper polymerization of fibrin and fibronectin by FXIIIa is essential for fibroblast migration (36). Recent exciting work also showed that FXIIIa, similar to fibronectin alone, can by itself modulate fibroblasts through integrin-mediated signaling pathways resulting in formation of actin+ stress fibers and novel spreading and adhesion of cultured fibroblasts as well as in various types of tumor cells (37). This explains our repeated observation in vivo of up to 50% admixture of proliferating FXIIIa+ histiocytes in many CD34+ and actin+ myofibroblastic tumors (4,13,15,16,19) wherein FXIIIa+ cells presumably exert their putative juxtacrine effects on fibroblast phenotype in as yet unexplored ways. FXIIIa has been shown to modulate collagen synthesis by and growth of fibroblasts (26,38), so it is obvious that FXIIIa+ dendritic histiocytes are major players in normal morphogenesis (2-4) and in some forms of neoplasia. This brief review will hopefully clarify why we continue to believe that the almost ubiquitous CD34+ fibroblasts and factor XIIIa+ dendritic histiocytes or dendrophages (27) are attractive candidates for interactive subsets of dermal (stromal) stem cells. We feel that this system is also an attractive target for research aimed at perhaps abrogating tumor cell invasion and metastasis in epithelial malignancies (39), manipulating and augmenting wound healing, and controlling the growth of inoperable or stubbornly recurrent fibrohistiocytic and fibrovascular sarcomata. Jeffrey S. Silverman HT,HTL,QIHC (ASCP) Ali Tamsen, M.D.