immunostaining for cd31 and cd34 kaposi sarcoma

I Clin Pathol 1995;48:1011-1016

Immunostaining for CD31 and CD34 in

Kaposi sarcoma

R Russell Jones, G Orchard, B Zelger, E Wilson Jones

AbstractAims-To evaluate antibodies directedagainst CD31 (JC70/A) and CD34(QBEND/10 and anti-HPCA-1) more ex-tensively in Kaposi sarcoma; to assess theirvalue in routine diagnosis; and to comparethem with the traditional endothelial cellmarkers Ulex europaeus agglutinin 1(UEA-1) and factor VIII related antigen.Methods-Twenty four cases of Kaposisarcoma were studied retrospectively. Allspecimens had been fixed in formalin andembedded in paraffin wax. The antibodieswere applied using the Streptavidin biotintechnique in all cases except for UEA-1,for which an indirect two stage methodwas used involving peroxidase conjugatedanti-ulex as the secondary antibody.Results-Tumours were classified intothose showing angiomatoid or lymph-angiomatoid elements and spindle celllesions. Universal labelling of all lesionsand virtually all elements within lesionswas seen with the anti-CD34 antibodiesQBEND/10 and HPCA-1. Labelling ofspindle cells was less consistent withJC70/A but both markers were superiorto the traditional endothelial cell markersUEA-1 and factor VIII related antigen.Conclusions-These data confirm that Ka-posi sarcoma is a tumour of endothelialcell origin. They shed further light on thehistogenesis of this complex tumour anddemonstrate that immunostaining forCD34 and CD31 can be used as an aid todiagnosis in routinely processed tissue.(JClin Pathol 1995;48:1011-1016)

Keywords: Kaposi sarcoma, CD31, CD34.

St John's Institute ofDermatology,St Thomas'sHospital, LondonR Russell JonesG OrchardE Wilson Jones

Department ofDermatology,University ofInnsbruck, AustriaB Zelger

Correspondence to:Dr R Russell Jones,St John's Institute ofDermatology, St Thomas'sHospital, Lambeth PalaceRoad, London SEI 7EH.

Accepted for publication20 July 1995

Immunohistochemistry has been used ex-

tensively in the study of malignant vasculartumours; firstly, to confirm their endothelialcell origin and, secondly, to aid diagnosis inroutine histopathology. However, many of themarkers which have been used for researchpurposes cannot be applied to routinely pro-cessed tissue because the antigenic de-terminants do not survive formalin fixation or

embedding in paraffin wax. Thus our ownstudies using monoclonal antibodies such as

PAL-E and EN4,'2 while valuable in confirmingthe endothelial cell nature of Kaposi sarcoma

and other malignant vascular tumours, cannotbe used for evaluating routinely processedspecimens. These restrictions do not apply tothe traditional endothelial markers, Ulexeuropaeus agglutinin 1 (UEA- 1) and factor VIII

1011

related antigen. However, studies of Kaposisarcoma and angiosarcoma have produced con-flicting results, particularly in relation to factorVIII related antigen.34Monoclonal antibodies directed against

CD34 and CD31 have excited interest in thestudy of vascular tumours. While they are notendothelial cell specific, they are widely ex-pressed in vascular endothelium, particularlyin pathological states. There is also some doubtwhether they will label lymphatic endo-thelium.5`8 Only limited studies of Kaposi sar-coma have been carried out using antibodiesdirected against CD34 or CD31 and some ofthese have used cryostat sections. Our studyaimed to evaluate these markers more ex-tensively in Kaposi sarcoma, to assess theirvalue in routine diagnosis and to compare themwith the traditional endothelial cell markersUEA-1 and factor VIII related antigen.

MethodsTwenty four cases of Kaposi sarcoma wereretrieved from the files of St John's Derma-topathology department, St Thomas's Hos-pital, London, and the Department ofDermatology, University ofInnsbruck, Austria.All specimens had been fixed in formalin andembedded in paraffin wax. The endothelialcell markers included UEA-1 (Sigma, Poole,Dorset, UK; diluted 1 in 50) and four mono-clonal antibodies: anti-factor VIII related an-tigen (Dako, High Wycombe, UK; diluted 1in 40); QBEND/10 (Oxoid, Basingstoke, UK;diluted 1 in 50), JC70/A (Dako; diluted 1 in10); and anti-HPCA-l (Becton Dickinson,Gosport, UK; diluted 1 in 25). The Strep-tavidin biotin technique was used in all casesexcept for UEA-1, for which an indirect twostage method was used involving peroxidaseconjugated anti-ulex as the secondary antibody.All tissue sections were blocked for endogenousperoxidase for 10 minutes prior to im-munolabelling using a 3% hydrogen peroxidein methanol solution. Sections for factor VIIIrelated antigen and UEA-1 labelling were tryp-sinised in the normal manner for 15 minutes(Sigma). JC70/A labelling required proteasedigestion (Sigma Type X bacillus therm-oproteolyticus rokko) for 30 minutes. QBEND/10 and HPCA-1 did not require predigestion.

ResultsHISTOLOGYTraditionally, Kaposi sarcomas are divided intopatch, plaque or nodular lesions. In our study13 biopsy specimens were patch or plaque

on July 8, 2022 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.48.11.1011 on 1 N

ovember 1995. D

ownloaded from

http://jcp.bmj.com/

Russell Jones, Orchard, Zelger, Wilson Jones

(Al

I i- %\Ie

It

*iI-

', !i .,"

t },

(B) /

..9,

1F-

;S 4

(CF -

_

_ *

Figure 1 (A) Patch stage lesion showing angiomatoid foci with early spindle cell formation (original magnification,x 10). (B) High power of angiomatoid focus (original magnification, x 40). (C) Plaque stage lesion showing a mixedpattern consisting of angiomatoid foci and lymphatic slits in the surrounding connective tissue (ornginal magnification,x 25). (D) Plaque stage lesion showing lymphangiomatoid changes which have merged to produce an angiosarcoma-likeappearance (original magnification, x 10).

lesions, and 11 were nodular. However, froma histogenetic point of view it is more in-formative to concentrate on the different ele-ments within a Kaposi lesion, particularlyduring the earliest phases of development. Atthis stage two types of lesion can be seenhistologically. Angiomatoid lesions are smallfoci of vascular proliferation associated with afew connective tissues cells, a variable in-flammatory cell infiltrate, and some red cellextravasation. The angiomatoid foci usuallycontain a central vessel with prominent en-dothelial cells surrounded by one or two thin-walled vessels containing erythrocytes and re-

sembling small capillaries (fig IA and IB). Thelumina of these small vessels can become di-latated so that adjacent lesions are separatedonly by a thin endothelial cell lining, the so-called "back to back" capillaries.

The second type of lesion consists of thin-walled, bloodless vessels dispersed within theconnective tissue. These vessels are angulatedrather than rounded and resemble small lymph-atics. This so called lymphangiomatoid variantis seen commonly, but not exclusively, in HIVrelated diseases. Occasionally, the lymphatic-like vessels merge to produce a dissection ofcollagen, similar in appearance to that seen inwell differentiated angiosarcomas (fig 1D).'Both lymphangiomatoid and angiomatoid

changes may be seen within the same lesion(fig 1C). Of the 13 patch/plaque stage lesionsstudied, two showed lymphangiomatoidchanges only, four showed angiomatoidchanges only, and seven showed a mixed pat-tern. Areas of spindle cell proliferation aresometimes seen in these early lesions, par-ticularly in relation to angiomatoid foci (fig IA).

Immunocytochemical labelling profile offive monoclonal antibodies in 24 cases of Kaposi sarcoma

Lesion Factor VIII related antigen UEA-1 JC70OA QBENDIO HPCA-1

Lymphangiomatoid (n 9)reactivity (R) - + + + + + +number (N) 0 2 5 9 9

Angiomatoid (n = 11)central vesselsR ++ ++ + ++ ++

capillariesR ++ + =/ ++ ++N Blush effect 5 (minority) 6 11 11

Early spindle cells (n =6)R +/ - + + +N 3 0 6 5 5

Nodular (n = I 1)sieve-like areasR + +l- ++ +l- ++ +1N 0 1 3 10 1 10 1

spindle cell areasR + + +++ + +N 0 1 5 10 3 8 3

1012

cR.

\ V's

:4


http://jcp.bmj.com

/J C


ovember 1995. D

ownloaded from

http://jcp.bmj.com/

Immunostaining for CD31 and CD34 in Kaposi sarcoma

_() -_2>-v @ * %--r£s>^ .* ,,. jU,~~~~~~~~~~~

v 11,*,,_ -'

' . , v'#x f' 't

! a' 2r}__

4.4- '1..4 *'

4. 9- .14. j

I -

-F

.4.'

-4 I

I p

- III' ,

(C)

i .I

_

--4-,t~~ ~ ~ ~~~~v .

t4 I

Figure 2 (A) Sieve-like areas with adjacent large back to back capillaries (original magnification, x 10). (B) J'C70/A(CD31) labels well differentiated vessels, but the smaller capillary-like vessels and the back to back capillaries are

negative (original magnification, x 10). (C) UEA-1 shows a similar pattern of labelling to JC701A. Capillaries are

negative but contain positively stained red blood cells, which can give an appearance offalse positivity (originalmagnification, x 40). (D) Equivalent area to (A) and (B). QBEND/10 shows positive labelling of all vessels andcapillary-like channels within the sieve-like area. The dilatated back to back capillaries are only weakly positive (originalmagnification, x 10).

In this study six of the 13 early lesions showedspindle cell changes. Eventually, however, thedifferent foci expand and merge to form a

-~~~~~~~1

'(',.J W-q .-N". , t~

. I., _

_: ~ v

'a

4

!_S v-- 4,/

'4Af. w .

, ~ K4-1

- X i F

Figure 3 Angiomatoid focus with surroundispaces (mixed pattern). (A) Diffuse factor aantigen labelling of angiomatoid focus. The islits are negative (original magnification, x

J'C70/A (CD31) showing positive labelling (differentiated central vessel and adjacent capthe angiomatoid focus. The surrounding lym,also weakly positive (original magnification,

network of tissue containing a mixture of vas-

cular and spindle cell elements. In some areas

numerous small capillaries are cut in cross-

section to give an appearance which has beenlikened to a sieve (fig 2A). In other areas sheetsof spindle cells separated by slit-like spaces

containing erythrocytes are evident. These are

the histological features classically associatedwith nodular lesions of Kaposi sarcoma. Inlater lesions spindle cells may predominate withsieve-like areas absent or confined to the per-

iphery of the nodule.

IMMUNOCYTOCHEMISTRYThe staining profile seen with the differentelements in the 24 lesions of Kaposi sarcoma

studied is summarised in the table. Figures 2to 5 illustrate the important findings.

-*- Factor VIII related antigen was consistentlyC...,. positive only in the angiomatoid foci. In ad-

'~-; dition to staining of the central vessels andsurrounding capillaries, there was also diffusionof factor VIII related antigen leading to a blush

effect and non-specific staining of the sur-

rounding connective tissue cells (fig 3A). Sim-ilar foci were seen within nodular lesions, butfactor VIII related antigen did not stain lymph-angiomatoid lesions, sieve-like areas, or spindlecells (fig 4A).

Staining with UEA-1 was also largely neg-ing slit-like ative. Apart from central vessels within an-

rymphatic-ike giomatoid foci there were no elements

20). (B) consistently stained by this marker. Only oneof well nodular lesion showed positivity in sieve-like'illaries within

phatic slits are and spindle cell areas. However, staining of red, x 20). cells within nodular lesions often gave a false

(D)'

44'

'I') .

N4

-

1013

."I


http://jcp.bmj.com

/J C


ovember 1995. D

ownloaded from

http://jcp.bmj.com/

1014~~~~~~~~~~~~~~~~~~~~~~~~~~Russell_Jones, Orchard, Zelger, Wilson Jo6nes

.1

-' 4. 4- 45 4''

4.E''4

S .' ,. *

A'.. 44) A

- . .4.4 IF

. , ,. p

.p4A *' ..

p.

(B) 'r!

UrC.).

I-i

Ae~~~~~~'4'~~~~~~~~~~~~~~~~~~~~~~~~~O

Figure 4 Pseudo-angiosarcoma lesion. (A) Factor VIII related antigen negative with positive normal central vessel(original magnification, x 25). (B) UEA-1 negative with positive normal central vessel (original magnification, x 20)(C) QBEND/110 positive for all cells lining the lymphangiomatoid channels (original magnification, x 25).(D) J7C70/A (CD3J1) positive for all lining cells and some intraluminal cells (original magnification, x 25).

'(A) ,B -

I

..- %"a t.

%. 'i

.,n .,%.....

. . k

.". II ? i t.

iD)

1444

4 4bk ~ 4r~~a

46

0- 4J.0

44 A~~~~~~~~~~~~~0

Figure 5 Spindle cell changes. (A) QBEND/l0. Early spindle cell formation. The cells lining the slit-like spaces in an

early spindle cell formation are particularly well demonstrated (original magnification, x 20). (B) Equivalent area to

(A). JYC70IA. The spindle cells are positive but the labelling is granular and cytoplasmic rather than membranous. The

slit-like spaces are poorly defined (original magnification, x 20). (C) Nodular lesion of Kaposi sarcoma. Spindle cells

are positive with HPCA-l (original magnification, x 40). (D) Early spindle cell formation, positive with J7C70/A-notegranular cytoplasmic labelling (original magnification, x 40).

appearance of positivity (fig 2C). A few capil-

laries in angiomatoid foci and in two of nine

lymphangiomatoid lesions stained with UEA-

(fig 4B).

By contrast, the CD34 antibodies provided

exceptionally consistent labelling of all ele-

ments within the lesions of Kaposi sarcoma.

The luminal surfaces of all vascular elements

were particularly clearly delineated by these

markers which often permitted the appreciation

of vessels not apparent on haematoxylin and

eosin stained sections. There were no sig-

' . ....... ..

.....:-.:

N...-.l.

4". AW-% .

4/~l,' e

....44..

--44'

".4

1014

.1.

I 14

I11 ! ",

.1

...

1%

....

I., -9-


http://jcp.bmj.com

/J C


ovember 1995. D

ownloaded from

http://jcp.bmj.com/

Immunostaining for CD31 and CD34 in Kaposi sarcoma

nificant differences between the staining pro-files for the two CD34 antibodies, QBEND/10 and HPCA-1. All lymphangiomatoid lesionswere labelled (fig 4C) as well as all vessels andcapillaries within angiomatoid lesions. Con-nective tissue cells in early angiomatoid lesionswere not labelled but all lesions showing earlyspindle cell formation except one were positivefor CD34 (fig 5A). In nodular lesions the sieve-like areas were particularly well visualised withonly one tumour showing weak staining in theseareas (fig 2D). CD34 also labelled spindle cells(fig 5C) but in three predominantly spindlecell lesions the labelling was patchy, particularlytowards the centre of the nodule.The CD31 antibody JC70/A was less con-

sistent than the CD34 markers. Thus, only fiveofnine lymphangiomatoid lesions were labelledby this antibody (fig 5D). Staining of an-giomatoid foci was limited to the central vesselswith only a few capillaries showing weak label-ling (fig 3B). In nodular lesions the sieve-likeareas were also poorly stained by this antibody(fig 2B). However, a few of the connectivetissue cells within the angiomatoid foci andall of the six lesions showing early spindleformation were stained by CD31 (fig 5B). Bycontrast, labelling of spindle cells within nod-ular lesions was weak and patchy in five casesand negative in six. CD31 labelling was cy-toplasmic and granular and less easy to definethan CD34 (fig 5D).

DiscussionPrevious studies of Kaposi sarcoma using anti-bodies directed against CD31 and CD34 havebeen limited. Ramani et alP in a study of 40cases reported positive CD34 labelling oflymphangiomatoid elements with variablestaining of spindle cell elements. Parums et alreported negative CD31 labelling in spindlecells of four cases of Kaposi sarcoma, whereasNickoloffreported positive labelling on cryostatsections.8 The data presented here demonstratethat CD34, and to a lesser extent CD3 1, anti-bodies are more reliable at labelling Kaposisarcoma lesions than the traditional endothelialcell markers, UEA-1 and factor VIII relatedantigen, and can be used on routinely processedtissue. For Kaposi sarcoma CD34 seems to besuperior to CD3 1. All cases of Kaposi sarcomalabelled with QBEND/10 and HPCA-1,whereas four of nine lymphangiomatoid lesionsand the majority of nodular lesions were neg-ative with JC70/A. Furthermore, the patternof labelling for QBEND/10 was more clearlydefined than the diffuse pattern seen withJC70/A.

These findings contrast with our own studiesin angiosarcoma where CD31 was found to be amore reliable marker than CD34.'0 Our findingsalso contrast with those of Suster and Wong'"who reported essentially negative staining withHPCA-1 in 20 cases of Kaposi sarcoma. Thisdiscrepancy may be due to technical factors aswe used the same antibody but at a 1 in 25 di-lution compared with the 1 in 200 dilution usedin the study by Suster and Wong. It has beensuggested that QBEND/10 and HPCA-1 label

different epitopes on CD34." Our findings,however, do not support this hypothesis in thatsimilar labelling of Kaposi lesions was foundwith both monoclonal antibodies.

Staining with the pan endothelial cell markerUEA-1 was less reliable in Kaposi sarcomathan either CD34 or CD3 1. Only one of 11nodular lesions and two of nine lymph-angiomatoid lesions labelled with UEA-1. Bycontrast, UEA-1 is a reliable marker for an-giosarcoma in formalin fixed tissue. In a pre-vious study'0 it labelled 18 of 19 cases in bothformalin and PLP (periodate lysine para-formaldehyde) fixed tissue.Our findings demonstrate that CD34 is the

best marker for labelling Kaposi sarcoma lesionsin routinely fixed tissue and shed some light onthe histogenesis of Kaposi sarcoma. Thus, thespindle cells seen in early patch and plaque stagelesions are CD31 positive in all cases and CD34positive in virtually all cases. By contrast, in nod-ular lesions of Kaposi sarcoma the majority ofspindle cells are still CD34 positive but a sig-nificant number are now CD31 negative. In latelesions where spindle cells predominate, CD34labelling also becomes more variable. This pro-vides evidence of de-differentiation and dem-onstrates that loss of vascular markers isimportant within spindle cell lesions.CD34 is a cell surface protein coded for by

the CD34 gene on chromosome lq/32. It isexpressed by human haematopoietic cells ofboth the myeloid and lymphoid series as well asendothelial cells. It may have a role in regulatingearly events in blood cell differentiation or itmay function as an adhesion molecule in bothendothelial cells and haematopoietic progenitorcells. CD34 is said to be present on normalvenular but not lymphatic endothelium, andhas been reported as negative in lymph-angiomas.6 If so, this would strongly favour avascular origin for the spindle cell elements inKaposi sarcoma. However, Ramani et al5 havereported that five of eight cases of lymph-angioma were focally positive with QBEND/10 and it is possible that adhesion moleculesare expressed more strongly in pathologicalstates. In our own studies some sections showedpositive labelling of lymphangiomatoid ele-ments with negative labelling of adjacent "nor-mal" lymphatics. Similar considerations applyto CD3 1. Again, CD31 is thought to labelvascular rather than lymphatic endotheliumand is reportedly absent in lymphangiomas andlymphoepithelial cysts.7 However, Nickolofffound CD31 labelling on both vascular andlymphatic epithelium using cryostat sections,8and in the present study some normal lymph-atics (that is, those containing small valves)labelled with CD31.

It would seem therefore that in Kaposi sar-coma CD34 and CD31 markers do not helpto distinguish between cells derived fromlymphatic or vascular endothelium. Rather itsuggests that the immunocytochemical profileof the endothelial derived cells in Kaposi sar-coma are unlike those found in normal tissue.Rather than interpreting the immunocyto-chemical findings in terms of "derivationfrom", it might be more appropriate to regard

1015


http://jcp.bmj.com

/J C


ovember 1995. D

ownloaded from

http://jcp.bmj.com/

Russell Jones, Orchard, Zelger, Wilson Jones

the endothelial cells of Kaposi sarcoma aspathological tissue which has differentiatedaway from normal endothelium. Certain ele-ments such as the angiomatoid foci still re-semble vascular tissue, but by the time spindlecells have come to predominate it is no longerpossible to relate endothelial cells to theirnormal counterparts. In addition, themultifocal nature of Kaposi sarcoma indicatesthat more than one type of endothelial cellcould be involved in its histogenesis.

1 Russell Jones R, Spaull J, Spry C, Wilson Jones E. His-togenesis of Kaposi's sarcoma in patients with and withoutacquired immune deficiency syndrome (AIDS). _7 Cliti Pa-thol 1986;39:742-9.

2 Holden C, Spaull J, Williams R, Spry C, Russell Jones R,Wilson Jones E. The detection of endothelial cell antigensin cutaneous tissue using methocam and periodate lysineparaformaldehyde fixation. .7 Immunol Methods 1986;91:45-52.

3 Leader M, Collins M, Patel J, Henry K. Staining for factorVIII related antigen and Ulex europaeus agglutinin 1 (UEA-1) in 230 tumours. An assessment of their specificity forangiosarcoma and Kaposi's sarcoma. Histopathologly 1986;10:1153-62.

4 BurgdorfW, Muhai K, Rosai J. Immunohistochemical iden-tification of factor VIII-related antigen in endothelial cellsof cutaneous lesions of alleged vascular nature. Am JT ClinPathol 1981;75:169-71.

5 Ramani P, Bradley NJ, Fletcher CDM. QBEND/10, a newmonoclonal antibody to endothelium: assessment of itsdiagnostic utility in paraffin sections. Histopathology 1990;17:237-42.

6 Sankey E, More L, Dhillon A. QBEND/10: A new im-munostain for the routine diagnosis of Kaposi's sarcoma.J3Pathol 1990;161:267-71.

7 Parums DV, Cordell JL, Micklem K, Heryet AR, GatterKC, Mason DY. JC70. A new monoclonal antibody thatdetects vascular endothelium associated antigen onroutinely processed tissue sections. J Clin Pathol 1990;43:752-7.

8 Nickoloff B. PECAM-1 (CD31) expressed as proliferatingendothelial cells, stromal spindle-shaped cells and dermaldendrocytes in Kaposi's sarcoma. Arch Dermatol 1993;129:250-1.

9 Gange RW, Wilson Jones E. Lymphangioma-like Kaposi'ssarcoma. BrJ7 Dermatol 1979;100:327-34.

10 Orchard GE, Zelger B, Wilson Jones E, Russell Jones R. Animmunohistochemical assessment of 19 cases of cutaneousangiosarcoma. Histopathology 1996 (in press).

11 Suster S, Wong T-Y. On the discriminatory value of anti-HPLA-1 (CD34) is the differential diagnosis of benignand malignant cutaneous vascular proliferations. Am JDermatopathol 1994;16:355-63.

12 Svanholm H, Neilsan K, Horge P. Factor VIII relatedantigen and lymphatic collecting vessels. 1/rchows Arch APathol Antat Histopathol 1984;404:223-8.

1016


http://jcp.bmj.com

/J C


ovember 1995. D

ownloaded from

http://jcp.bmj.com/

immunostaining for cd31 and cd34 kaposi sarcoma

Documents