www.elsevier.com/locate/humpath

Human Pathology (2013) 44, 2302–2311

Original contribution

Pathogenesis of human hemangiosarcomas andhemangiomas☆

Liping Liu MD, PhDa, Satoko Kakiuchi-Kiyota PhDb, Lora L. Arnold MSa,Sonny L. Johansson MD, PhDa, David Wert BS a, Samuel M. Cohen MD, PhDa,⁎

aDepartment of Pathology and Microbiology, University of Nebraska Medical Center, 983135 Nebraska Medical Center,Omaha, NE 68198-3135, USAbCompound Safety Prediction, Pfizer, Inc, Groton, CT 06340, USA

Received 5 March 2013; revised 16 May 2013; accepted 17 May 2013

HreVW

pc

0h

Keywords:Hemangiosarcoma;Hemangioma;Hematopoietic stem cells;Endothelial progenitorcells;Immunohistochemistry

Summary Hemangiosarcomas are uncommon aggressive vascular tumors that have recently becomethe focus of attention because several chemicals and pharmaceuticals increase their incidence in mice.The relevance of these mouse vascular tumors to humans is unclear. In the present study, we semi-quantitatively evaluated the expression profiles of hematopoietic stem cell markers (CD117 [c-kit],CD133, CD34, and CD45), endothelial cell markers (vascular endothelial growth factor receptor 2,CD31, and factor VIII–related antigen), and a myeloid lineage cell marker (CD14) in human heman-giosarcoma (n = 12) and hemangioma (n = 10) specimens using immunohistochemistry. CD133 wascompletely negative in almost all cases of hemangiosarcomas and hemangiomas. Most hemangio-sarcomas, but not hemangiomas, stained for CD117 and CD45. Both groups diffusely expressed CD34,vascular endothelial growth factor receptor 2, and factor VIII–related antigen; however, hemangiomashad more intense and diffuse CD34 and factor VIII–related antigen expression compared with heman-giosarcomas, whereas CD31 was positive in all hemangiosarcomas but only half of the hemangiomas.CD14 staining was negative in most hemangiosarcoma and hemangioma cases. Our results indicate thatmultipotential bone marrow–derived hematopoietic stem cells or early endothelial progenitor cells(EPCs) expressing CD117, CD34, and CD45 are involved in hemangiosarcoma formation, whereashemangiomas originate from late EPCs or differentiated endothelial cells, which have lost the expres-sion of most hematopoietic stem cell markers. This contrasts with our previous results that demonstratedthat both hemangiosarcomas and hemangiomas in mice may be derived from early EPCs that are notcompletely differentiated.© 2013 Elsevier Inc. All rights reserved.

Abbreviations: EC, endothelial cell; EPC, endothelial progenitor cell;&E, hematoxylin and eosin; PPAR, peroxisome proliferator–activatedceptor; RTU, ready to use; UEA-1, Ulex europaeus agglutinin-1;EGFR2, vascular endothelial growth factor receptor 2; vWF, vonillebrand factor.☆ Statement of interest: Funding for the immunostains was provided, in

art, by a grant from Pfizer, Inc, Groton, CT. Dr Cohen has previouslyonsulted for Pfizer, Inc, but on matters unrelated to this report.⁎ Corresponding author.E-mail address: scohen@unmc.edu (S. M. Cohen).

046-8177/$ – see front matter © 2013 Elsevier Inc. All rights reserved.ttp://dx.doi.org/10.1016/j.humpath.2013.05.012

1. Introduction

Hemangiosarcomas are rare aggressive malignant tumorsin humans and usually have no known cause [1,2]. The onlyknown causes of hemangiosarcomas in humans include afew rare genetic disorders, previous irradiation, and exposureto certain genotoxic agents such as vinyl chloride and thoro-trast [3,4]. Recently, it was reported that a wide range of

2303Pathogenesis of human hemangiosarcomas and hemangiomas

nongenotoxic chemical and pharmaceutical agents such asperoxisome proliferator–activated receptor γ agonists in-creased the incidence of hemangiosarcomas in rodents,primarily in mice in 2-year carcinogenicity studies [3,5].Human hemangiomas are benign proliferations of bloodvessels, can occur in several tissues, and are relativelycommon tumors [6]. In mice, vascular tumors occurfrequently, with incidences in controls ranging from 2% to5% [3].

The pathogenesis of hemangiosarcomas is not wellunderstood in humans or animals. It has been proposedthat the malignant endothelial cells (ECs) are derived eitherfrom differentiated ECs that developed malignant potentialthrough mutations or from transformed bone marrow–derived endothelial progenitor cells (EPCs) [7]. Numerousstudies show that adult bone marrow and peripheral bloodcontain EPCs that are capable of differentiating into matureECs [8,9]. Bone marrow–derived EPCs are consideredhematopoietic in origin, expressing hematopoietic stem cellmarkers including CD117 (c-kit), CD133, CD34, and CD45.The mobilization of EPCs from the bone marrow is initiatedby the activation of matrix metalloproteinase-9, resulting indetachment of early stem and progenitor cells expressingCD117 from the bone marrow stromal niche and theirsubsequent movement to the vascular zone of the bonemarrow [8–10]. CD133 is expressed in hematopoietic stemcells and EPCs but absent in mature ECs [8]. Beaudry et al[11] demonstrated that circulating EPCs appear to have lostCD117 expression while retaining CD133 expression inmice. Expression of CD34 is not restricted to hematopoieticstem cells or early EPCs but is also present on late EPCs andmature ECs at lower levels [8,9]. CD45 is expressed inhematopoietic stem cells and in both lymphoid and myeloidcell lineages from progenitor to mature cells [12]. EarlyEPCs in the bone marrow or in blood just after migration intothe circulation begin to express the EC marker vascularendothelial growth factor receptor 2 (VEGFR2) in additionto hematopoietic stem cell markers [8,9]. During differen-tiation from EPCs to mature ECs, the circulating EPCs losethe expression of CD117 and CD133 [8–10] and begin toexpress mature EC markers such as CD31 and factor VIII–related antigen (or von Willebrand factor) [8,9]. Recently,the contribution of myeloid lineage cells in angiogenesis hasbeen reported. Yoder et al [13] showed that CD45- andCD14 (monocyte/macrophage cell surface antigen)-expres-sing cells also expressed EC markers such as CD31,VEGFR2, factor VIII–related antigen [14], and Ulexeuropaeus agglutinin-1.

We demonstrated that mouse spontaneous and chemicallyinduced hemangiosarcomas and hemangiomas are bothderived from EPCs expressing CD34, VEGFR2, and CD31but are negative for CD45, factor VIII–related antigen, andCD14. Factor VIII–related antigen expression is known tooccur later than CD31 expression in EPCs. Therefore, thelack of factor VIII–related antigen expression may indicatethat these tumor cells are arrested at a stage before complete

differentiation. Different from mouse EC tumors [15], caninehemangiosarcoma cell lines expressed CD117, CD133,CD34, and, in some cases, CD45 and CD14 [7]. However,their expression pattern in canine EC tumors in vivo is notyet available. In humans, one study showed that CD117expression was observed in hemangiosarcomas (15%) [4].Some case reports demonstrated that CD45 expression isnegative in hemangiosarcomas [16–19], whereas CD34,VEGFR2, CD31, and factor VIII–related antigen expressionhas been demonstrated in hemangiosarcomas and hemangi-omas [4,16,17,20–22]. However, expression of CD133 andCD14 in human EC tumors is not well understood.

In the present study, we evaluated the expression profilesfor hematopoietic stem cell markers (CD117, CD133, CD34,and CD45), EC markers (VEGFR2, CD31, and factor VIII–related antigen), and a myeloid cell marker (CD14) in humanhemangiosarcomas and hemangiomas. These tumors werealso stained with the p53 tumor suppressor protein becausethis marker has been present in most hemangiosarcomas inhumans, but not in human hemangiomas or in any mousevascular tumor types [3]

2. Materials and methods

2.1. Selection of cases and controls

Pathology files of the University of Nebraska MedicalCenter were searched for cases of hemangiosarcoma (fromyears 2000 toMarch 2013) and hemangioma (from years 2011to 2012). Twelve cases of hemangiosarcomas and 10 cases ofhemangiomas (including 1 epithelial type) were identified,and formalin-fixed and paraffin-embedded tissue specimenswere retrieved from the tissue archives of the University ofNebraska Medical Center. Clinical data including age, sex,and site(s) of involvement were available in all cases(Table 1). The study was approved by the institutional reviewboard of the University of Nebraska Medical Center. Repre-sentative hematoxylin and eosin (H&E)–stained histologyslides from hemangiosarcoma and hemangioma specimens areshown in Fig. 1A and B, respectively.

2.2. Immunohistochemistry

Immunohistochemical procedures including antibodiesand negative controls for primary antibodies used in ourstudy are summarized in Table 2. Formalin-fixed andparaffin-embedded tissues were deparaffinized and rehy-drated through graded alcohols. Endogenous peroxidaseactivity was quenched with 3% hydrogen peroxide, followedby heat-induced antigen retrieval using Leica Bond Max(Leica Biosystems, Buffalo Grove, IL) for 20 minutes at100°C in citrate buffer, pH 6.0; ER1 (Leica Biosystems) forCD117 and VEGFR2 staining or EDTA, pH 9.0; ER2 (LeicaBiosystems) for CD133, CD45, CD34, CD31, CD14, and

Table 1 Summary of patients' clinical information

Tumors Patient ID Age a Sex Location

Hemangiosarcomas 1 81 F Skin, chest2 48 F Breast3 48 M Leg4 60 M Kidney5 85 F Soft tissue, right

forearm6 60 F Breast7 65 F Breast8 99 M Skin, forehead9 82 F Breast10 77 F Skin, chest11 34 M Pleura12 70 M Skin, face

Hemangiomas 13 86 F Breast14 17 M Muscle15 36 M Soft tissue16 66 M Skin17 10

moF Skin

18 49 M Liver19 62 M Skin20 81 F Bone21 47 F Sinus22 b 41 M Bone

Abbreviations: F, female; M, male.a Age in years except for patient 17, which is in months.b Diagnosed as having epithelioid hemangioma.

2304 L. Liu et al.

p53 staining; or for 5 minutes at 37°C in ER1 for factor VIII–related antigen staining. Additional blocking of nonspecificbinding was performed using 10% normal serum from thehost species of the secondary antibody (Leica Biosystems)for 10 minutes at room temperature for CD133, CD34,VEGFR2, CD31, CD14, CD117, and p53 staining or normalserum contained in VECTASTAIN Elite ABC Kit (RabbitIgG; Vector Laboratories, Burlingame, CA) for factor VIII–related antigen staining, as directed by the manufacturer. No

Fig. 1 Hemangiosarcomas (A) and hemangiomas (B) stained by H&Espindle cells with significant atypia, forming anastomosing immaturestructures lined by bland ECs.

serum block was performed for CD45 staining. Sectionswere incubated with primary antibodies or negative controlsfor 30 minutes at room temperature and then incubated withantirabbit polymer or antimouse polymer (Leica Biosystems)for all antibodies. Refine Detection System (Leica Biosys-tems) containing 3,3′-diaminobenzidine was used for thedetection of target proteins, resulting in brown staining.Slides were then counterstained with hematoxylin, dehy-drated, and coverslipped.

2.3. Immunohistochemical evaluation

Immunohistochemically stained sections were semiquanti-tatively evaluated by clinical pathologists (L.L., S.L.J., andS.M.C.) for the presence/absence of staining, the percentage oftumor cells expressing positive staining, and the overallstaining intensity of positively stained tumor cells and weresubsequently classified into one of the following grades: 0, nopositive staining in any of the tumor cells; 1+, positive stainingin 25% or less of the tumor cells; 2+, positive staining in 26%to 50% of the tumor cells; 3+, positive staining in 51% to 75%of the tumor cells; and 4+, positive staining in 76% to 100% ofthe tumor cells. The overall staining intensity was graded asweak (w), moderate (m), or strong (s).

3. Results

For each immunostain, no major differences were ob-served between males and females or between differenttissues containing the vascular tumors, in terms of stainingpattern and/or intensity (data not shown).

3.1. Hematopoietic stem cell markers

Eight of 12 hemangiosarcoma cases were positive forCD117 staining with moderate to strong intensity (Fig. 2A

(×10). The hemangiosarcomas were hypercellular and composed ofvascular channels. Hemangiomas showed anastomosing vascular

Table 2 Antibodies and negative controls used for immunohistochemistry

Description ofantibodies andblocking reagents

Manufacturer Stock concentrationof primary antibodies

Antigenretrieval

Primaryantibodydilution

Secondaryantibodies

Internal positivecontrol cells

CD 117 Cell Marque (117R-18) RTU ER1 RTU LeicaPolymer

GISTNormal rabbit serum Jackson ImmunoResearch

(001-000-001)PROM1 (CD133) Millipore (MAB4399) 1.0 mg/mL ER2 1:100 Leica

PolymerNot applicable

Normal rabbit serum Jackson ImmunoResearch(001-000-001)

CD45 Leica (PA0042) RTU ER2 RTU LeicaPolymer

LeukocytesNot used N/ACD34 Leica (PA0212) RTU ER2 RTU Leica

PolymerVascularendotheliumNormal rabbit serum Jackson ImmunoResearch

(001-000-001)VEGFR2 Cell Signaling (2479) Unknown ER1 1:200 Leica

PolymerVascularendotheliumNormal rabbit serum Jackson ImmunoResearch

(001-000-001)CD31 Leica (PA0250) RTU ER2 RTU Leica

PolymerVascularendotheliumNormal rabbit serum Jackson ImmunoResearch

(001-000-001)Factor VIII–relatedantigen

Cell Marque (250A-18) RTU ER1 RTU LeicaPolymer

Vascularendothelium

Normal rabbit serum Jackson ImmunoResearch(001-000-001)

CD14 Leica (NCL-CD14-223) Unknown ER2 1:100 LeicaPolymer

NeutrophilsNormal rabbit serum Jackson ImmunoResearch

(001-000-001)p53 DAKO (IR616) RTU ER2 RTU Leica

PolymerTumor cells

Normal rabbit serum Jackson ImmunoResearch (001-000-001)

NOTE. Cell Marque, Rocklin, CA; Millipore, Auburn, CA; DAKO, Carpinteria, CA; Cell Signaling, Danvers, MA; GeneTex, Irvine, CA; JacksonImmunoResearch, West Grove, PA.Abbreviations: ER1, Leica Bond Epitope Retrieval Citrate pH 6.0; ER2, Leica Bond Epitope Retrieval EDTA pH 9.0; GIST, gastrointestinal stromal tumor;N/A, not applicable; RTU, ready to use.

2305Pathogenesis of human hemangiosarcomas and hemangiomas

and Table 3). In contrast, only 3 of 10 hemangioma cases hadweak CD117-positive staining (Fig. 2B and Table 3). Wewere unable to find a positive internal control for CD133because of the absence of expression of the marker in ourspecimens. However, we used pancreatic tissue for ourpositive control, which showed dot-like staining in the apicalside of the acini. In our study samples, only 1 hemangio-sarcoma specimen showed scattered CD133-positive cells(Table 3). Nine cases of hemangiosarcomas (Fig. 2C andTable 3) and all cases of hemangiomas (Fig. 2D and Table 3)were positive for CD34 staining, exhibiting intense mem-brane staining on most tumor cells. In general, the percentageof CD34-positive tumor cells in hemangiomas was higherthan that in hemangiosarcomas. The staining intensity ofCD34 in both groups was similar to the internal vascularcontrols (Fig. 2C and D). The staining pattern of CD45 washard to identify in 3 cases in the solid parts of hemangio-sarcomas because of the interference of residual CD45-expressing leukocytes (Fig. 2E, inset). However, CD45tumor cell expression was observed in at least 7 cases ofhemangiosarcoma (Fig. 2E and Table 3). Nine hemangioma

cases were completely negative for CD45 staining (Fig. 2Fand Table 3), despite positive membrane staining of internalcontrol cells (leukocytes) within these sections.

3.2. EC markers

The percentage of cells staining positive for VEGFR2was similar between hemangiosarcomas and hemangiomas(Table 3); more than 75% of cells stained in most cases.However, the staining intensity for VEGFR2 was stronger inthe hemangiosarcoma cases (Fig. 3A and Table 3) than thehemangioma cases (Fig. 3B and Table 3). CD31 tumor cellstaining was observed in all cases of hemangiosarcoma(ranging from moderate to strong; Fig. 3C and Table 3),although only 5 hemangioma cases were positive for CD31(ranging from weak to strong; Fig. 3D and Table 3). FactorVIII–related antigen staining was focal and relatively weakin hemangiosarcoma cases (Fig. 3E and Table 3), whereas itwas strong and diffuse in hemangioma specimens (Fig. 3Fand Table 3).

Fig. 2 Immunostains for hematopoietic stem cell markers showed that hemangiosarcomas were positive for CD117 (A) and CD45 (E) andpatchy positive for CD34 (C); hemangiomas were negative for CD117 (B) and CD45 (F) and positive for CD34 (D). Several cases ofhemangiosarcoma showed scattered positive cells for CD45 in the lesions. However, they were interpreted as indeterminate owing to the lackof vasculature in the solid lesion and possible contamination of CD45-expressing leukocytes (E, inset).

2306 L. Liu et al.

3.3. Myeloid lineage cell marker and tumorsuppressor protein p53

Membrane/cytoplasmic staining was observed in neutro-phils (a positive internal control); however, most hemangio-sarcomas (Table 3) and hemangiomas (Fig. 4B and Table 3)were negative for CD14 staining. Only 4 hemangiosarcomasshowed patchy positivity for CD14 (Fig. 4A). p53 positivitywas observed in all cases of hemangiosarcoma (Fig. 4C),whereas most hemangiomas were negative (Fig. 4D), except

in 2 cases (Table 3). The intensity ranged from moderateto strong in all hemangiosarcoma cases, except 2 cases(Table 3), but was very weak in hemangioma cases (Table 3).

3.4. Staining pattern in an epithelioid hemangioma

One case was diagnosed as “epithelioid hemangioma” (avariant of hemagonium). This benign vascular tumor iscomposed of well-formed immature vessels that are linedpredominantly with plump epithelioid ECs [23]. Epithelioid

Table 3 Staining for markers and p53 in human hemangiosarcomas (HS) and hemangiomas (HA)

Tumors PatientID

Hematopoietic stem cell markers EC markers Myeloid cellmarker: CD14

Tumor suppressormarker: p53CD117 CD133 CD34 CD45 VEGFR2 CD31 Factor VIII–

related antigen

HS 1 4+ s 0 4+ s 0 4+ s 3+ s 4+ s 0 4+ s2 0 1+ s 2+ s 0 1+ w 3+ m 1+ w 0 1+ m3 3+ m 0 0 ind 4+ s 4+ s 3+ m 0 2+ m4 0 0 0 3+ s 1+ w to m 1+ m to s 1+ m 2+ m 1+ m5 4+ s 0 4+ s 2+ s 4+ s 4+ s 4+ s 0 3+ m6 4+ s 0 4+ s 1+ s 4+ s 4+ s 4+ m to s 0 1+ w7 3+ m 0 4+ m to s ind 4+ s 3+ m to s 4+ m to s 0 1+ w8 0 0 0 2+ s 4+ s 3+ s 3+ m ind 1+ m9 4+ s 0 4+ s ind 4+ s 3+ m 4+ s 0 1+ s10 1+ m 0 3+ s 2+ s 4+ m to s 4+ s 3+ m to s 1+ m 1+ s11 0 0 1+ m to s 1+ s 4+ s 3+ s 3+ m to s 1+ s 1+ m12 1+ m 0 2+ s 2+ s 4+ m to s 4+ s 3+ s 1+ s 3+ s

HA 13 0 0 4+ s 0 4+ s 0 4+ s 0 014 0 0 4+ s 0 1+ w 0 4+ s 0 015 1+ w 0 4+ s 0 3+ w to m 0 4+ s 0 016 0 0 4+ s 0 4+ m 0 4+ s 0 017 0 0 4+ s 0 4+ s 4+ m to s 4+ s 0 1+ w18 3+ w to m 0 4+ s 0 4+ m 3+ w 4+ s 0 019 0 0 4+ s 0 4+ m 0 4+ s 1+ m 020 0 0 4+ s 0 4+ w to m 2+ s 4+ s 0 020 0 0 4+ s 0 2+ w to m 4+ s 4+ s 0 022 a 1+ w 0 4+ s 2+ s 3+ w to m 1+ w to m 4+ s 2+ m 1+ w

NOTE. Staining for markers in human HS and HA: 0, no positive staining in any of the tumor cells; 1+, positive staining in 25% or less of the tumor cells;2+, positive staining in 26% to 50% of the tumor cells; 3+, positive staining in 51% to 75% of the tumor cells; 4+, positive staining in 76% to 100% of thetumor cells.Abbreviations: w, weak; m, moderate; s, strong; ind, indeterminate.a The patient has epithelioid hemangioma.

2307Pathogenesis of human hemangiosarcomas and hemangiomas

hemangiomas are observed in a wide age range, peaking inthe third to fifth decades, and the incidence in female patientsis higher than that in male patients [23]. In the present study,it was removed from a 41-year-old man who presented with apainful scapular destructive lesion. The patient had no rele-vant history and had been disease-free for 15 months aftersurgery. An H&E-stained slide revealed a cellular lesion,with epithelial-like rather than spindle tumor cells forminganastomosing vasculature structures (Fig. 5A). No readilyidentified mitoses were observed. The tumor exhibited apositive staining pattern similar to hemangiosarcoma forCD117 (Fig. 5B), CD45 (Fig. 5C), and p53 (Fig. 5D), but itsexpression pattern for EC markers was similar to that ofhemangiomas (Table 3).

4. Discussion

In the present study, we showed that expression of CD117and CD45 was observed in most hemangiosarcoma cases,whereas they were negative in most of the hemangiomacases. CD133 was negative and CD34 was positive for both

hemangiosarcomas and hemangiomas. Staining forVEGFR2 and factor VIII–related antigen was positive forboth hemangiosarcomas and hemangiomas. CD31 stainingwas observed in all cases of hemangiosarcomas, althoughonly 5 hemangioma cases were positive for CD31. Mosthemangiosarcoma and hemangioma cases were negative forCD14 staining.

CD133 and CD117 are expressed in immature cells suchas hematopoietic stem cells and early EPCs located in thebone marrow or in blood just after migration into the cir-culation, and their expression is decreased in late progenitorcells and absent in mature ECs [8,9,11]. Expression of CD34is not restricted to hematopoietic stem cells or early EPCs,but also is present on late EPCs and mature ECs at lowerlevels [8,9]. CD45 is expressed not only on hematopoieticstem cells but also on lymphoid and myeloid cell lineagesfrom progenitor to mature cells as a pan-hematopoieticmarker [9,12]. A previous study found that 15% (5/34 spe-cimens) of hemangiosarcomas were positive for CD117 [4].No reports are available regarding the expression of CD133in these tumors in humans. Multiple studies demonstratedpositivity for CD34 on human hemangiosarcomas[17,20,22]. Some case studies reported that immunostaining

Fig. 3 Immunostains for EC markers demonstrated that hemangiosarcomas were strongly and diffusely positive for these 3 markers,VEGFR2 (A), CD31 (C), and factor VIII–related antigen (E); hemangiomas were weakly positive for VEGFR2 (B), partially positive forCD31 (D), and strongly positive for factor VIII–related antigen (F).

2308 L. Liu et al.

in human hemangiosarcomas for CD45 was negative [16–18]. In the present study, in contrast to previous publications,most cases of human hemangiosarcomas showed strong andat least focally positive staining for CD117 and CD45.CD133 was completely negative in almost all hemangiosar-comas (9/10). That relatively few hemangiosarcomas showedpositivity for CD133 is in accord with the fact that it is anearly marker in progenitors or initiating cells [8,9,11], theexpression level of which is below the level of detection byimmunohistochemistry after differentiation or malignanttumor heterogeneity. CD34 was positive with intensity

similar to the internal control (nonneoplastic vasculature)for most hemangiosarcomas (8/12). In contrast, most heman-gioma cases were negative for CD45 and CD117. Similar tohemangiosarcomas, CD133 was negative but CD34 waspositive in all cases of hemangiomas, with intensity similar tothe internal control (nonneoplastic vasculature). These resultssuggest that more immature EPCs expressing CD45 andCD117, in addition to CD34, contribute to hemangiosarcomaformation compared with hemangiomas in humans. Interest-ingly, our results demonstrated that tumor cells in heman-giosarcomas express CD117 but not CD133. If the lack of

Fig. 4 Immunostains for myeloid lineage cell marker CD14 and tumor suppressor protein p53 revealed that several hemangiosarcomasshowed focal CD14 positivity (A), whereas a majority were negative for CD14 and positive for p53 (C); hemangiomas were largely negativefor CD14 (B) and p53 protein (D).

2309Pathogenesis of human hemangiosarcomas and hemangiomas

CD133 expression is caused by loss of expression duringdifferentiation, this would be opposite from the observationof mouse circulating EPCs, suggesting that CD133 is retainedlonger than CD117 [11]. This provides some evidence thatthe mechanism for differentiation from hematopoietic stemcells to the EC lineage in humans differs from that in mice.

Among the 3 EC markers used in our study, VEGFR2 isconsidered to be expressed earliest and can be seen in im-mature cells, which also express CD133 [24]. During thedifferentiation process, these cells lose the expression ofhematopoietic stem cell markers and begin to express otherEC markers, first CD31 and then factor VIII–related antigen[8,25]. In the current study, human hemangiosarcomasexpressed higher levels of VEGFR2 and CD31 but lowerlevels of factor VIII–related antigen staining compared withhemangioma cases. These findings are consistent with ourresults of hematopoietic stem cell staining, demonstratingthat more immature EPCs are involved in hemangiosarcomascompared with hemangiomas. Staining results of EC markersmay indicate that expression levels of VEGFR2, CD31, andfactor VIII–related antigen are determined by the differen-tiation of ECs.

CD14 is a monocyte/macrophage cell surface antigen, andthe contribution of myeloid lineage cells to angiogenesis has

recently been demonstrated [13,26]. In addition, coexpres-sion of a dendritic cell marker and EC markers was observedin human infantile hemangioma [27]. Also, it is possible thatmyeloid-derived cells are involved in the hemangiosarcomaformation in dogs, at least in some cases [7]. A few heman-giosarcomas (4/12) demonstrated CD14 positivity; this sup-ports the possibility of involvement of myeloid-derived cellsin the pathogenesis of the disease.

In contrast to human specimens, murine spontaneous andchemically induced hemangiosarcomas and hemangiomasshowed similar expression profiles: both entities were nega-tive for CD45, factor VIII–related antigen, and CD14, andpositive for CD34 (stronger than internal vasculature con-trol), VEGFR2, and CD31 [15]. Based on the increasedintensity of CD34 and the lack of factor VIII–related antigenin these tumors, we concluded that both types of the mousetumors may be derived from early EPCs that are not com-pletely differentiated [15]. The origin of these 2 vasculartumors appears to be different between mice and humans.For example, although hematopoietic stem cells and/orcirculating EPCs seem to be involved in both mouse andhuman hemangiosarcoma formation, CD45 was positive forhuman vascular tumors but not for mice. In addition, humanhemangiosarcoma cases are positive for factor VIII–related

Fig. 5 A, One case diagnosed as epithelioid hemangioma showed spindle to epithelioid cells with mild atypia forming anastomosis vascularchannels. The neoplastic cells stained positive for CD117 (weak) (B), CD45 (C), and p53 (D).

2310 L. Liu et al.

antigen, indicating that they retain the ability to further dif-ferentiate similar to canines [28], whereas mouse tumor cellsare arrested at a stage before complete differentiation owingto the lack of factor VIII–related antigen. Although mousehemangiomas may consist of early EPCs, human hemangi-omas showed diminished intensity of VEGFR2 and CD31and increased expression of factor VIII–related antigen, astaining pattern of late EPCs or differentiated ECs.

We also investigated the expression of the tumor sup-pressor gene TP53 in human hemangiosarcomas andhemangiomas. Wild-type p53 protein has a short half-life;however, the mutated (inactivated) form of p53 protein isstabilized and can be detected in the nucleus of neoplasticcells [29]. Studies have reported that in humans, immuno-stains for p53 were almost entirely negative in hemangiomas[30], whereas nuclei of malignant cells in hemangiosarcomasstained positive for p53 [18,31]. Furthermore, TP53mutationsin exons 5 to 8 were observed in spontaneous and vinylchloride–induced hemangiosarcomas in humans [32,33].There was moderate to strong nuclear staining for p53 in allhemangiosarcomas in the current study, whereas only 2hemangiomas showed weak staining.

Our findings for the expression profile of the epithelialhemangioma warrant further exploration. This specific entity

exhibited an expression profile similar to hemangiosarco-mas, positive for CD45, CD117, and p53, which indicatesthese cells might be derived from EPCs and might havemalignant potential.

Our study showed relative heterogeneity of the expressionprofile of hemangiosarcomas, although a clear trend of theexpressed markers indicates the different origins of hemangio-sarcoma and hemangioma in humans. The heterogeneity mightbe caused by intratumor clonal diversity, dysregulation ofdifferent proteins because of genomic or epigenetic modifica-tions, or aberrant expression of transcription factors, micro-RNAs, and so on, and even sampling issues of the tumor [34].

In summary, our study indicates that human hemangio-sarcomas are most likely derived from hematopoietic stemcell/EPCs with high malignant potential compared withhemangiomas. Human hemangiomas are formed from lateEPCs or differentiated ECs. Hemangiosarcomas and hem-angiomas appear to be 2 different entities with distinctpathogenesis, and hemangiomas are unlikely to transforminto hemangiosarcomas. There is also no evidence thathuman hemangiosarcomas (rare tumors) arise from heman-giomas (common tumors) or any other benign precursor [3].Thus, our findings raise questions regarding the relevance ofthe mouse model to human vascular tumors.

2311Pathogenesis of human hemangiosarcomas and hemangiomas

Acknowledgments

We gratefully appreciate Cheryl Putnam (University ofNebraska Medical Center) for her assistance with the pre-paration of this manuscript. We also thank Drs Jon C. Cookand Leslie A. Obert (Pfizer, Inc) for advice and commentsregarding this project.

References

[1] Hardisty JF, Elwell MR, Ernst H, et al. Histopathology ofhemangiosarcomas in mice and hamsters and liposarcomas/fibrosar-comas in rats associated with PPAR agonists. Toxicol Pathol 2007;35:928-41.

[2] Krishnan Unni K, Inwards CY, Bridge JA, et al. Malignant vasculartumor. In: Silverberg S, editor. Tumors of the bones and joints. AFIP.Washington, DC: AFIP/ARP; 2005. p. 266-73.

[3] Cohen SM, Storer RD, Criswell KA, et al. Hemangiosarcoma inrodents: mode-of-action evaluation and human relevance. Toxicol Sci2009;111:4-18.

[4] Yonemori K, Tsuta K, Ando M, et al. Contrasting prognostic im-plications of platelet-derived growth factor receptor-beta and vascularendothelial growth factor receptor-2 in patients with angiosarcoma.Ann Surg Oncol 2011;18:2841-50.

[5] Kakiuchi-Kiyota S, Vetro JA, Suzuki S, et al. Effects of thePPARgamma agonist troglitazone on endothelial cells in vivo and invitro: differences between human and mouse. Toxicol Appl Pharmacol2009;237:83-90.

[6] Krishnan Unni K, Inwards CY, Bridge JA, et al. Hemangiomas. In:Silverberg S, editor. Tumors of the bones and joints. AFIP.Washington, DC: AFIP/ARP; 2005. p. 261-5.

[7] Lamerato-Kozicki AR, Helm KM, Jubala CM, Cutter GC, ModianoJF. Canine hemangiosarcoma originates from hematopoietic pre-cursors with potential for endothelial differentiation. Exp Hematol2006;34:870-8.

[8] Hristov M, Weber C. Endothelial progenitor cells: characterization,pathophysiology, and possible clinical relevance. J Cell Mol Med2004;8:498-508.

[9] Urbich C, Dimmeler S. Endothelial progenitor cells: characterizationand role in vascular biology. Circ Res 2004;95:343-53.

[10] Hristov M, Erl W, Weber PC. Endothelial progenitor cells: mobiliza-tion, differentiation, and homing. Arterioscler Thromb Vasc Biol2003;23:1185-9.

[11] Beaudry P, Hida Y, Udagawa T, et al. Endothelial progenitor cellscontribute to accelerated liver regeneration. J Pediatr Surg 2007;42:1190-8.

[12] Hermiston ML, Xu Z, Weiss A. CD45: a critical regulator ofsignaling thresholds in immune cells. Annu Rev Immunol 2003;21:107-37.

[13] Yoder MC, Mead LE, Prater D, et al. Redefining endothelialprogenitor cells via clonal analysis and hematopoietic stem/progenitorcell principals. Blood 2007;109:1801-9.

[14] Harach HR, Jasani B, Williams ED. Factor VIII as a marker ofendothelial cells in follicular carcinoma of the thyroid. J Clin Pathol1983;36:1050-4.

[15] Kakiuchi-Kiyota S, Crabbs TA, Arnold LL, et al. Evaluation ofexpression profiles of hematopoietic stem cell, endothelial cell, andmyeloid cell antigens in spontaneous and chemically inducedhemangiosarcomas and hemangiomas in mice. Toxicol Pathol2012;41:709-21.

[16] Lin BT, Weiss LM, Battifora H. Intravascularly disseminated angio-sarcoma: true neoplastic angioendotheliomatosis? Report of two cases.Am J Surg Pathol 1997;21:1138-43.

[17] Forton GE, Van Parys G, Hertveldt K. Primary angiosarcoma of thenon-irradiated parotid gland: a most uncommon, highly malignanttumor. Eur Arch Otorhinolaryngol 2005;262:173-7.

[18] Terada T. Angiosarcoma of the oral cavity. Head Neck Pathol 2011;5:67-70.

[19] Kuroda N, Hamaguchi N, Inoue K, et al. Application of immunocy-tochemistry to the diagnosis of primary epithelioid angiosarcoma ofthe lung. Med Mol Morphol 2009;42:250-3.

[20] Poblet E, Gonzalez-Palacios F, Jimenez FJ. Different immunoreactiv-ity of endothelial markers in well and poorly differentiated areas ofangiosarcomas. Virchows Arch 1996;428:217-21.

[21] Itakura E, Yamamoto H, Oda Y, Tsuneyoshi M. Detection andcharacterization of vascular endothelial growth factors and theirreceptors in a series of angiosarcomas. J Surg Oncol 2008;97:74-81.

[22] Breiteneder-Geleff S, Soleiman A, Kowalski H, et al. Angiosarcomasexpress mixed endothelial phenotypes of blood and lymphaticcapillaries: podoplanin as a specific marker for lymphatic endothelium.Am J Pathol 1999;154:385-94.

[23] Fetsch J. Epitheloid hemangioma. In: Fletcher CU, KK. Mertens F,editors. WHO classification of tumors of soft tissue and bone. Lyon,France: IARC Press; 2002. p. 159-60.

[24] Peichev M, Naiyer AJ, Pereira D, et al. Expression of VEGFR-2 andAC133 by circulating human CD34(+) cells identifies a population offunctional endothelial precursors. Blood 2000;95:952-8.

[25] Kanayasu-Toyoda T, Yamaguchi T, Oshizawa T, Hayakawa T. CD31(PECAM-1)-bright cells derived from AC133-positive cells in humanperipheral blood as endothelial-precursor cells. J Cell Physiol 2003;195:119-29.

[26] Kim SJ, Kim JS, Papadopoulos J, et al. Circulating monocytesexpressing CD31: implications for acute and chronic angiogenesis.Am J Pathol 2009;174:1972-80.

[27] Ritter MR, Reinisch J, Friedlander SF, Friedlander M. Myeloid cells ininfantile hemangioma. Am J Pathol 2006;168:621-8.

[28] Von Beust BR, Suter MM, Summers BA. Factor VIII–related antigenin canine endothelial neoplasms: an immunohistochemical study. VetPathol 1988;25:251-5.

[29] Hong HL, Ton TV, Devereux TR, et al. Chemical-specific alterationsin ras, p53, and beta-catenin genes in hemangiosarcomas fromB6C3F1 mice exposed to o-nitrotoluene or riddelliine for 2 years.Toxicol Appl Pharmacol 2003;191:227-34.

[30] Iwata J, Sonobe H, Furihata M, Ido E, Ohtsuki Y. High frequency ofapoptosis in infantile capillary haemangioma. J Pathol 1996;179:403-8.

[31] Meis-Kindblom JM, Kindblom LG. Angiosarcoma of soft tissue: astudy of 80 cases. Am J Surg Pathol 1998;22:683-97.

[32] Hollstein M, Marion MJ, Lehman T, et al. p53 mutations at A:T basepairs in angiosarcomas of vinyl chloride–exposed factory workers.Carcinogenesis 1994;15:1-3.

[33] Naka N, Tomita Y, Nakanishi H, et al. Mutations of p53 tumor-suppressor gene in angiosarcoma. Int J Cancer 1997;71:952-5.

[34] Marusyk A, Polyak K. Tumor heterogeneity: causes and conse-quences. Biochim Biophys Acta 1805;2010:105-17.