Procollagen 1 expression in atypicalfibroxanthoma and other tumors

Kristin Jensen1, BarbaraWilkinson2, Nina Wines2 andSteven Kossard2

1Department of Pathology, Stanford UniversityMedical Centre, and2Skin and Cancer Foundation Australia,Darlinghurst, New South Wales, Australia

Steven Kossard, FACD, Skin and Cancer FoundationAustralia, 277 Bourke Street, Darlinghurst NSW2010, New South Wales, AustraliaTel: þ61 2 9360 4480Fax: þ61 2 9331 1244e-mail: skossard@scfa.edu.au

Accepted September 5, 2003

Background: Procollagen (PC) is secreted by fibroblasts into theextracellular matrix, where it is cleaved to form collagen. The rat anti-humanPC-1monoclonal antibodyhas been reported to reactwith atypicalfibroxanthoma (AFX), a poorly differentiated but usually benign skin lesioncommon in elderly patients. We have studied PC-1 staining in 50 tumorswithAFXhistological features (four ofwhichwere subsequently reclassifiedas non-AFX tumors) to confirm this prior observation. In addition,wehaveinvestigated PC-1 in other skin tumors, particularly thosewith spindled cellor sclerosing/desmoplastic morphologies.Method: Archival material was retrieved and sections were preparedand immunostained with PC-1 as well as a panel of antibodies, includingS-100 and MNF-116 (cytokeratins 5, 6, 7, 8, 17, and 19).Results: PC-1 staining was strongly positive in 40 of 46 (87%) AFXs.Three AFXs displayed weak staining with PC-1 even after repeat stainingof 10 tumors that were initially weak. Three additional tumors stainedwithboth PC-1 and MNF-116 and were classified as AFX-like squamous cellcarcinoma (SCC). One tumor with AFX-like histology was PC-1 negativeand S-100 positive and was classified as an AFX-like melanoma. Positivestaining in tumor cells was observed in three of nine (33%) desmoplasticmalignant melanomas, three of eight (38%) desmoplastic SCCs, zero of 10(0%) desmoplastic trichoepitheliomas, zero of 10 (0%) morpheic basal cellcarcinomas, and zero of 10 (0%) sclerosing sweat duct carcinomas.Conclusion: PC-1 is a useful antibody in a diagnosticimmunohistochemical panel when investigating AFX and AFX-liketumors; however, good technical quality and careful interpretation arenecessary when using a panel of antibodies, particularly to keratin andS-100 protein, for optimal accuracy.

Jensen K, Wilkinson B, Wines N, Kossard S. Procollagen 1 expression inatypical fibroxanthoma and other tumors.J Cutan Pathol 2004; 31: 57–61. # Blackwell Munksgaard 2004.

Procollagen (PC) is synthesized by fibroblasts andother epithelial and stromal cells and consists of thecollagen protein with additional globular amino acidends rich in disulfide bonds. Upon secretion from thecell, these additional ends are cleaved so that collagencan become incorporated into the extracellularmatrix, where it is an almost ubiquitous structuralprotein. The production of PC by fibroblasts is influ-enced in part by transforming growth factor-b andtumor necrosis factor-a, and gene expression is underboth transcriptional and post-transcriptional regu-lation. The expression and secretion of PC to become

collagen in the extracellular matrix is an importantfeature of wound healing and tissue repair processesto which the desmoplastic stroma of a malignancyhave sometimes been compared.1

Type I collagen composes 80–85% of dermalcollagen; some studies have suggested that type Icollagen plays an important role in fibroticconditions such as post-irradiation skin damage.2

One in vitro model of photodamage showed inhibitionof PC-1 synthesis in the presence of damaged orpartially digested collagen fragments, the degree ofwhich appeared to lessen after the fragments were

J Cutan Pathol 2004: 31: 57–61 Copyright # Blackwell Munksgaard 2004Blackwell Munksgaard. Printed in Denmark

Journal of

Cutaneous Pathology

57

processed.3 This in vitro model also suggested a rela-tionship between fibroblast morphology (possibly sec-ondary to mechanical tension) and PC synthesis.Atypical fibroxanthomas are benign lesions com-

posed of spindled cells set in a variably collagenousstroma. Although some atypical fibroxanthomas(AFXs) feature spindled cells with relatively blandcytology, others can contain epithelioid to enlarged,bizarre cells and numerous mitotic figures, includingatypical figures, and show marked pleomorphism.Because these lesions typically occur in elderly individ-uals, the differential diagnosis includes entities suchas poorly differentiated squamous cell carcinoma(SCC), spindled melanoma, and other poorly differ-entiated aggressive neoplasms. There has not been asatisfactory positive immunohistochemical markerfor AFX, and it is considered a diagnosis of exclusion.AFXs have been recently reported to show expressionof PC-1 by monoclonal antibody staining.4 This seriesexamines a large number of AFXs to confirm thisfinding and to describe in further detail, the stainingcharacteristics and interpretive considerations.Uncommon examples of PC-1-positive poorly differ-entiated SCCs and rare melanomas may closelyresemble AFX histologically but can be distinguishedby positive keratin or S-100 antibody staining,respectively.It has been hypothesized that certain tumors which

feature a desmoplastic stroma may overexpress PC.1,5

For instance, Moy et al.5 demonstrated increased typeI and type III PC messenger RNA in sclerosing basalcell carcinomas (BCCs). We have therefore also com-piled several examples of spindle cell lesions and/orlesions featuring a prominent desmoplastic, scleros-ing, or fibrotic stroma to study PC-1 expression inthese lesions.

Materials and methods

The archival files of the Skin and Cancer FoundationAustralia (Darlinghurst, New South Wales, Australia)from 1992 to 2002 were searched for cases diagnosedas AFX, desmoplastic malignant melanoma, desmo-plastic SCC, morpheic BCC, sclerosing sweat ductcarcinoma, and desmoplastic trichoepithelioma.These diagnoses were made on the basis of establishedmorphologic criteria and/or characterized immuno-histochemical profiles. A summary of patients’ demo-

graphical information is provided in Table 1.Formalin-fixed, paraffin-embedded tissue blockswere sequentially sectioned at 4mm and stained witha panel of antibodies using standard immunohisto-chemical techniques. Each tumor was stained withthe rat anti-human PC-1 amino-terminal monoclonalantibody (dilution 1 : 350 clone M-58) (Chemicon,Temecula, CA, USA). Selected tumors were addition-ally stained with antibodies to S100 (dilution 1 : 500clone S-100) (DAKO, Carpinteria, CA, USA), and akeratin cocktail consisting of cytokeratins 5, 6, 7, 8, 17,and 19 (dilution 1 : 100 clone MNF-116) (DAKO).Prior to immunostaining, the slides stained for PC-1and keratin were enzyme digested with protease for4min and 2min, respectively. Slides were counter-stained and coverslipped according to standardimmunohistochemical methods.

Results

Forty of 46 (87%) AFXs showed positive reactivitywith the PC-1 antibody, with localization to thecytoplasm (Fig. 1). Initially, 10 tumors had difficultyin interpreting weak reactivity, but this was reducedto three after repeat staining. Nine of the positivetumors showed only focal reactivity. The positivelystaining cases included AFXs with both predominantepithelioid cell (Fig. 2) and spindle cell (Fig. 3) patternsof differentiation. Large bizarre tumor cells, eventhose undergoing mitoses, were often labeled by theantibody. Labeling, however, did not appear to be afunction of cell morphology. Three additionaltumors with the morphological features of AFXstained positively with MNF-116 (Fig. 4) as well asPC-1 (two strong and one weak). One tumor withmorphological features of AFX stained positive forS-100 but was negative for PC-1. On the basis ofthe positive keratin or S-100 markers, these tumorswere classified as AFX-like SCCs or AFX-like mela-nomas, respectively.Three of nine desmoplastic melanomas (confirmed

by strong S-100 reactivity in each of the tumors) andthree of eight desmoplastic SCCs (confirmed bystrong-positive reactions with the pan-keratin anti-body, MNF-116) showed positive cytoplasmic reactiv-ity with PC-1 (Fig. 5). One tumor originally diagnosedas desmoplastic SCC, which was particularly spindledin appearance, showed strong PC-1 positivity and did

Table 1. Patients’ demographics

Diagnosis Sex ratio (male : female) Mean age Site (head/neck)

Atypical fibroxanthoma 29 : 17 75 36/50Desmoplastic malignant melanoma 3 : 6 59 5/9Desmoplastic squamous cell carcinoma 6 : 2 76 8/8Sclerosing sweat duct carcinoma 5 : 5 65 6/10Morpheic basal cell carcinoma 3 : 7 62 8/10Desmoplastic trichoepithelioma 4 : 6 43 10/10

Jensen et al.

58

not react withMNF-116. In light of this immunopheno-typic profile, we reclassified this tumor as an AFX.None of the morpheic BCCs, sclerosing sweat duct

carcinomas, or desmoplastic trichoepitheliomas (10examples of each tumor, all confirmed by strongpositive reactions with the pan-keratin antibody,MNF-116) showed reactivity with PC-1 except instromal fibroblasts (Fig. 6).The appropriate external control slides were posi-

tive for PC-1, and all negative cases showed PC-1expression in stromal fibroblasts, which providedpositive internal control. Immunohistochemical stain-ing profiles are summarized in Table 2.

Discussion

Putative classification of AFX as a fibrohistiocytictumor is supported by PC-1 expression by tumorcells. Interestingly, this expression was shown notonly in the spindled cells, but also in the epithelioidand large, bizarre cells, including those undergoingmitoses within the tumors. This finding varies some-what from the results of Varani et al.,3 in that thoseauthors found variable PC expression based on fibro-blast morphology; however, their in vitro studies

explored PC expression in photodamaged fibroblasts.We believe that the consistent PC-1 expression inmorphologically distinct cells of AFX is most likelysecondary to the proliferation of cells derived from acommon stem cell.Increased PC-1 expression has been observed in

other tumor types, namely, uterine carcinosarcomas,breast carcinomas, and experimental anaplasticthyroid carcinomas.6–8 Kauppila et al.6 showed thatmalignant, undifferentiated cells of uterine carcino-sarcomas (cells of both epithelial and mesenchymalorigin) expressed PC-1, and that the nearby collagenproduced by these cells lacked the mature, fullycrossed-linked structure of collagen produced bybenign stromal fibroblasts. This group also observednewly formed collagen (lacking the mature, cross-linked structure) in poorly differentiated breast car-cinomas, although PC expression was limited tostromal fibroblasts and was not seen in the malignantcells.7 These separate findings led this group to pos-tulate that the abnormal collagen matrix produced byeither the tumor cells themselves or by tumor-influenced stromal cells is important in stromalinvasion by tumor. Interestingly, Dahlman et al.8

found that collagen I expression by experimentally

Fig. 1. Pleomorphic spindle and epithelioid tumor cells demonstrating

procollagen 1 positivity.

Fig. 2. Detail of cytoplasmic procollagen 1 stain within pleomorphic

epithelioid cells.

Fig. 3. Spindle cell variant of atypical fibroxanthoma highlighted by

procollagen 1 stain.

Fig. 4. Atypical fibroxanthoma-like squamous cell carcinoma with

positive MNF-116 keratin stain in epithelioid dermal keratinocytes.

PC-1 in atypical fibroxanthoma

59

derived anaplastic thyroid carcinoma cell lines wasinversely related to tumorigenicity, although thesestudies were performed in vitro. Certainly, these andother studies underline the critical role that PC-1plays in structural integrity and suggest that normaland abnormal expression of PC-1, by tumor and/orstromal cells, is somehow inherent to stromal invasionby tumor cells. Continued characterization of PC-1expression and maturity of collagen matrix willfurther elucidate this function.A large percentage of AFX (87%) showed positive

staining for PC-1. Although this is an encouragingdiscovery in the search for a positive diagnostic markerof AFX, a significant (one-third or more in oursmall series) percentage of desmoplastic melanomaand desmoplastic SCC also showed positive reactivitywithin tumor cells. These results are similar to thosedescribed earlier, although we obtained a slightlyhigher rate of PC-1 positivity in desmoplastic mela-noma than other investigators.2

We would therefore caution against the use of PC-1 staining in isolation of other immunohistochemicalmarkers. Likewise, we would consider the positive

reactivity of other antibodies, such as keratin andS-100, as outweighing positive PC-1 staining in the finaldetermination of tumor type. Mar et al.4 describe onecase in their series of 49 AFXs that showed nearlyequal expression of PC-1 and keratin, which theyconsidered AFX on morphologic grounds. Weencountered three such cases, but we think it mostprudent to classify these as AFX-like SCCs. In arecent report, an example of an AFX that was initiallynegative or had weak-keratin staining later recurredand was strongly positive, at which time lymph nodemetastasis was also noted.9 We also encountered onemelanoma with morphological overlap with the spindlevariant of AFX that was PC-1 negative and S-100positive, and this could be classified as an AFX-likespindle cell melanoma.As with all diagnostic immunohistochemical stains,

careful localization of antibody staining should beestablished. Distinction between positive PC-1 stain-ing in spindled tumor cells and positive reactivity inbackground fibroblasts becomes critical, especially inthe setting of a desmoplastic stroma containing acti-vated fibroblasts. In BCCs and desmoplastic tricho-epithelioma, stromal fibroblasts demonstrated strongexpression of PC-1, and this contrasted with thenegative basaloid keratinocytes. Careful determin-ation of cytoplasmic staining, including staining of thedendritic processes of fibroblasts, will assist in theinterpretive process.Increased PC-1 messenger RNA has been demon-

strated in sclerotic (morpheic) BCCs as compared tonodular BCCs and control fibroblast populations.5 Inaddition, poorly differentiated ovarian tumors, produ-cing strong local and remote fibroproliferative reac-tions, have been shown to contain scattered cellswhich express type I PC by in situ hybridization.1

Given these results, one might hypothesize thattumors with a desmoplastic stroma may containtumor cells which express PC-1. Although three ofnine desmoplasticmelanomas and three of eight desmo-plastic SCCs did show tumor cell expression ofPC-1, the majority of desmoplastic, sclerosing, andfibrotic tumors we examined did not show any tumorcell expression, and none of the PC-1-expressingAFXs in our series featured a desmoplastic stroma.In addition, none of the tumor cells in morpheicBCCs studied here showed reactivity with PC-1 byimmunohistochemistry. The study of Moy and col-leagues5 utilized in situ hybridisation with cDNAprobes for corresponding messenger RNA withintumor tissue. Perhaps, a less mature, non-secretedPC-1 product is detected by immunohistochemistryin tumor cells of AFXs and other tumors, a productthat is functionally unable to produce a normal orsclerotic stroma.We were therefore unable to predict PC-1 positiv-

ity, intensity of staining, or cell-type staining on the

Fig. 5. Desmoplastic melanoma demonstrating procollagen 1-positive

spindle melanocytes

Fig. 6. Infiltrative basal cell carcinoma with procollagen 1 labeling

confined to stromal fibroblasts.

Jensen et al.

60

basis of tumor morphology in any of the tumors.Presumably, some aspect of tumor biology independ-ent of phenotypic alteration may influence PC-1expression. The results seen by Moy and colleagues5

in morpheic BCCs may in fact be the consequence offibroblast activation by the tumor and fibroblasticexpression of PC-1, rather than overexpression bytumor cells themselves.In summary, we have confirmed PC-1 expression

in AFXs and detailed the interpretive challengesinherent in its use. The discovery of a positive anti-body marker for AFX is promising as a diagnostictool in the differential diagnosis of dermal-basedspindle cell lesions. Although PC-1 is a highly usefulmarker in the sense of positive staining in AFXs, it isnot entirely specific, and some tumors may show onlyweak or focal staining that may need to be repeated ifinconclusive, and some tumors may remain negative.PC-1 labeling should always be used and interpretedin the setting of an immunohistochemical panel inorder to adequately exclude diagnostic mimics thatmay appear AFX like.

References

1. Kauppila S, Saarela J, Stenback F, Risteli J, Kauppila A, Risteli L.

Expression of mRNAs for type I and type II procollagens in

serous ovarian cystadenomas and cystadenocarcinomas. Am J

Pathol 1996; 148: 539.

2. Riekki R, Parikka M, Jukkola A, Salo T, Risteli J, Oikarinen A.

Increased expression of collagen types I and III in human skin

as a consequence of radiotherapy. Arch Dermatol Res 2002;

294: 178.

3. Varani J, Perone P, Fligiel SEG, Fisher GJ, Voorhees JJ.

Inhibition of type I procollagen production in photodamage:

correlation between presence of high molecular weight col-

lagen fragments and reduced procollagen synthesis. J Invest

Dermatol 2002; 119: 122.

4. Mar N, Ruben B, McCalmont T. Procollagen I: a useful

immunoperoxidase reagent for the diagnosis of atypical fibrox-

anthoma. J Cutan Pathol 2002, 29: 577 [Abstract].

5. Moy RL, Moy LS, Matsuoka LY, Bennett RG, Uitto J. Selec-

tively enhanced procollagen gene expression in sclerosing

(morphea-like) basal cell carcinoma as reflected by elevated

Proa1(I) and Proa1(III) procollagen messenger RNA steady-

state levels. J Invest Dermatol 1988; 90: 634.

6. Kauppila S, Stenback F, Kacinski BM, Carcangiu M, Risteli J,

Risteli L. Characterization of type I collagen synthesis and

maturation in uterine carcinosarcomas. Cancer 1999; 86:

1299.

7. Kauppila S, Stenback F, Risteli J, Jukkola A, Risteli L. Aber-

rant type I and type III collagen gene expression in human

breast cancer in vivo. J Pathol 1998; 186: 262.

8. Dahlman T, Lammerts E, Bergstrom D, et al. Collagen type I

expression in experimental anaplastic thyroid carcinoma: regu-

lation and relevance for tumorigenecity. Int J Cancer 2002; 98:

186.

9. Gray Y, Robidoux HJ, Farrell DS, Robinson-Bostom L.

Squamous cell carcinoma detected by high-molecular-weight

cytokeratin immunostaining mimicking atypical fibroxanthoma.

Arch Pathol Lab Med 2001; 125: 799.

Table 2. Summary of immunoperoxidase (IPX) results

Histological diagnosis Procollagen 1 (PC-1) S100 MNF-116

Atypical fibroxanthoma (AFX) 40/46* strong; nine focally positive,10 initially weak, three weak afterrepeated testing

One positive with AFX-likehistology (1/50) but negativefor PC-1; classified asAFX-like melanoma

Three positive AFX-likepathology (3/50) PC-1,two strong,one weak;classified as AFX-like SCC

Desmoplastic SCC 3/9 0/8 8/9One reclassified as AFX

Morpheic basal cell carcinoma 0/10 0/10 10/10Desmoplastic trichoepithelioma 0/10 0/10 10/10Sclerosing sweat duct carcinoma 0/10 1/10 10/10Desmoplastic malignant melanoma 3/9 9/9 0/9

*‘True’ AFX: all 46 negative for S-100 and MNF-116.

PC-1 in atypical fibroxanthoma

61