A Unique Microvascular Phenotype Shared by Juvenile Hemangiomas and Human Placenta FREE

JUVENILE HEMANGIOMAS are the most common tumors of infancy, affecting approximately 10% of children. These benign vascular tumors vary from small and innocuous to large and deforming (Figure 1A-C), but share a remarkably predictable biological behavior: lesions appear within weeks of birth (or less commonly are congenital), proliferate rapidly, then spontaneously involute. Although many hemangiomas can be left to regress without intervention, even involuted lesions may be disfiguring, with loose, atrophic skin covering fatty tumor residuum (Figure 1D). Active therapeutic approaches include corticosteroid and laser therapy, surgical resection, and cosmetic surgical correction of postinvolutive residua. Pathogenesis has remained obscure.

We recently reported that the erythrocyte-type glucose transporter protein, GLUT1, is highly expressed by juvenile hemangiomas at all stages of their evolution.¹ GLUT1 expression represents an intrinsic feature of the committed endothelial phenotype of these lesions and enables their immunohistochemical distinction from other histologically similar benign vascular tumors and anomalies. Use of this marker has increased diagnostic accuracy, allowing study of juvenile hemangiomas as a coherent, singular entity.

GLUT1 is immunohistochemically undetectable in the vasculature of normal skin and subcutis, but is highly expressed in normal endothelia at sites of blood-tissue barriers, including brain, eye, nerve, and placenta,^{2 - 6} suggesting possible pathogenic associations between juvenile hemangiomas and these tissues. We sought to further define the antigenic phenotype of juvenile hemangiomas, and to compare this with other tissues that normally demonstrate high GLUT1 expression phenotype. A panel of antibodies, directed against endothelial and basement membrane antigens selectively expressed in neural and/or placental tissues, was applied to a large series of juvenile hemangiomas—representing all stages of their evolution—and other vascular lesions, and the patterns of expression were compared with those of normal human brain, placenta, and skin. We further determined immunoreactivity of the aforementioned tissues for Lewis Y antigen (LeY), an oligosaccharide normally expressed by certain epithelia and by activated T cells, as LeY immunoreactivity has been reported in a small series of proliferative-phase hemangiomas.⁷ We find that capillaries of juvenile hemangioma are unique in sharing a stable, distinctive immunophenotype with placental fetal microvessels. These results suggest an exciting new paradigm for our understanding of the pathogenesis of this most common tumor of infancy.

All vascular lesions excised at Arkansas Children's Hospital, Little Rock, during the past 3 years (1997-1999), for which both frozen and paraffin-embedded tissue samples were available (111 cases), were reviewed. Hematoxylin-eosin–stained sections were reviewed blindly by 2 of us (P.E.N. and M.C.M.), and pathological diagnoses made independently.^{1 ,8} Two intracranial and 4 intramuscular lesions (all malformations) were excluded, leaving 66 juvenile hemangiomas, 26 malformations, and 13 pyogenic granulomas, with greater than 95% diagnostic concordance between reviewers. Diagnostic discrepancies, all minor, were resolved by joint review. Twenty-eight additional cases, for which only paraffin sections were available, were also included: 6 tufted angiomas, 7 epithelioid hemangioendotheliomas, 1 infantile kaposiform hemangioendothelioma, and 14 angiosarcomas. Clinical information was obtained by chart review after immunohistochemical analysis.

Samples of fresh human term placenta were taken within 30 minutes of delivery from central villus parenchyma. Samples of fresh human brain, spinal cord, and truncal skin were collected from 4 patients (aged 1-15 years) at autopsy within 4 to 8 hours post mortem. Four additional samples of normal skin were taken from margins of resected pediatric skin lesions of nonvascular origin. As examples of normal, reactive capillary proliferation, granulation tissue was collected from 1 healing surgical wound, 2 decubitus, and 1 case of ulcerative colitis. Six malignant pediatric tumors of nonvascular origin (to evaluate tumor neovasculature) included renal medullary carcinoma (n = 1), desmoplastic small blue cell tumor (n = 1), neuroblastoma (n = 2), and osteosarcoma (n = 2).

Cryosections were collected at −20°C on sialanized slides, fixed in acetone at 20°C for 1 minute, and air-dried prior to rehydration in Tris-buffered isotonic sodium chloride solution containing 0.05% Triton-X100. Paraffin sections were deparaffinized, rehydrated, and subjected to citrate buffer antigen retrieval¹ ; all sections were protein-blocked¹ before incubation with primary antibodies under optimal conditions (Table 1). Bound primary antibody was detected using a peroxidase kit (LSAB+; DAKO Corporation, Carpinteria, Calif) using diaminobutyric acid chromagen (DAB+, DAKO).¹ Negative controls were processed in parallel without primary antibody. Normal tissue immunoreactivities provided internal positive controls, except for LeY, for which sections of LeY-immunopositive oral mucosa were included in each run.

Immunoreactivities were scored blindly by one of us (P.E.N.) as none, weak, moderate, or intense (intense meaning greater than or equal to control immunoreactivity). For LeY, occasional weak immunoreactivity in a perinuclear, hof-type pattern was discounted. Only membranous and/or diffuse cytoplasmic-membranous LeY immunoreactivity was scored as positive. For merosin, occasional immunoreactivity in a discontinuous "stringy" pattern in connective tissue around larger veins and arteries in normal skin and subcutis and malformations, was discounted. Only circumferential, basement membrane–like patterns of merosin (and laminin) immunoreactivity were scored as positive.

Tissue cubes (1 mm³) from central portions of 3 routinely submitted juvenile hemangiomas (patients aged 22 days, 5 months, and 3 years) and from terminal villus-containing portions of 2 full-term placentas were promptly fixed in glutaraldehyde and processed for electron microscopy.

Patients (aged 22 days to 7 years) with both proliferative and involutive phase hemangiomas were well represented, with sex and site distributions similar to those reported in large epidemiological studies^{15 - 16} (Table 2). Age ranges for patients with other vascular lesions (excluding angiosarcoma) overlapped that of the hemangioma group.

Specific microvascular GLUT1 immunoreaction was present in normal brain and placental chorionic villi (and placental trophoblast), but not in normal skin and subcutis, granulation tissue, or neovasculature of nonvascular malignant tumors (Table 2, Figure 2B and H, Figure 3B, and Figure 4B). Intense microvascular, entirely endothelial, GLUT1 immunoreactivity was present in all juvenile hemangiomas tested (Table 2, Figure 4, Figure 5A). No lesional GLUT1 expression was detected in any of the malformations, pyogenic granulomas, tufted angiomas, or hemangioendotheliomas (Table 2, Figure 6). Weak-to-moderate, focal GLUT1 immunoreaction was present in 5 of 14 angiosarcomas (Table 2, Figure 6K).

Vascular immunoreaction for LeY or FcγRII was not present in normal skin and subcutis or brain (except for perivascular macrophages and microglia)¹⁷ (Figure 2). Placental intravillous fetal capillaries showed intense endothelial immunoreactivity for LeY and FcγRII (Figure 4). Epithelia of normal and marginal skin showed variable immunoreactivity for LeY (but not for FcγRII), ranging from none (approximately 25% of cases) to moderate (in 10%). Strong LeY immunoreactivity within oral glandular and mucosal epithelia served as positive controls. Patient blood type did not affect LeY immunoreactivity (data not shown). Skin FcγRII immunoreactivity was restricted to dermal macrophages, epidermal Langerhans cells, and inflammatory cells.

Marked lesional endothelial immunoreactivity for FcγRII and LeY, in a diffuse, granular cytoplasmic-membranous pattern, similar to that seen in placenta, was present in all 66 juvenile hemangiomas (Table 2, Figure 4 and Figure 5). No LeY or FcγRII immunoreactivity was found in adjacent native dermal capillaries, or in arterioles or arteries, either within or adjacent to hemangiomas. There was no lesional immunoreaction for LeY or FcγRII in any of the malformations, pyogenic granulomas, granulation tissues, or tumor neovasculatures (Table 2, Figure 3). There was no LeY immunoreactivity in tufted angiomas or the kaposiform hemangioendothelioma, and only weak-to-moderate LeY immunoreactivity (in 2%-5% of tumor cells) in 2 of 7 epithelioid hemangioendotheliomas and 5 of 14 angiosarcomas (Table 2, Figure 6). FcγRII immunoreactivity could not be assessed in these latter lesions, for which frozen material was unavailable.

Strong merosin (α2-laminin) immunoreactivity was seen in vascular basement membranes of brain and placental chorionic villi but not in the vasculature of normal skin and subcutis, as reported previously^{18 - 21} (Figure 2 and Figure 4). Merosin was highly expressed in nonvascular basement membranes of trophoblast, nerve, skeletal muscle, and dermal-epidermal junctions. α-Laminin ("classic" laminin), in contrast, decorated all vascular and nonvascular basement membranes (Figure 2 and Figure 4).

Intense merosin immunoreactivity was present in all juvenile hemangiomas, in a broad, continuous, bandlike pattern encircling lesional capillaries (Table 2, Figure 4 and Figure 5). Dermal capillaries of normal marginal skin, as well as intralesional and extralesional arterioles, were immunonegative for merosin, although occasional weak-to-moderate immunoreaction was seen in arterial internal elastic laminae. α-Laminin immunoreactivity was present in all basement membranes, including basement membranes of lesional capillaries and adjacent normal large and small vessels.

No lesional merosin immunoreactivity was present in malformations, pyogenic granulomas, granulation tissue, or tumor neovasculature (Table 2, Figure 3). Intense α-laminin immunoreactivity was present in all lesional and nonlesional vascular and epithelial basement membranes in these specimens (Figure 3). Frozen tissue was not available for tufted angiomas, hemangioendotheliomas, or angiosarcomas, precluding assessment of merosin or laminin immunoreactivity.

Moderate-to-strong endothelial immunoreactivity for the endothelium-specific, pinocytotic vesicle-associated antigen PAL-E was observed in all specimens with available frozen tissue, except for normal brain, confirming preservation of endothelial antigenicity and findings of previous studies.^{14 ,22} Endothelial CD31 immunoreactivity, used to assess antigenicity in paraffin sections, was present in all cases.

Juvenile hemangiomas (patients aged 22 days to 3 years) and placental microvessels shared ultrastructural features of (1) continuous, nonfenestrated endothelia with occasional uniform, small, subplasmalemmal vesicles; (2) small numbers of closely apposed, encircling pericytes; and (3) continuous, redundant basement membrane, varying in thickness, adjacent to or encircling interspersed collagen fibrils (Figure 7). Early hemangioma endothelia were plump, whereas those of older (involuted) lesions and of term placental capillaries were relatively thin. Basement membrane multilamination was more developed in hemangiomas of older children.

We report a constellation of shared markers of cellular specialization that imply a unique relationship between juvenile hemangioma and placental fetal microvessels. These findings characterize juvenile hemangioma as a distinct pathological entity, intrinsically different from other vascular tumors and anomalies to which it has been compared. In light of the perinatal or congenital presentation of hemangiomas and their distinctive pattern of limited growth and involution, the antigenic similarities between these lesions and placental microvessels led us to hypothesize 2 possible pathogenic mechanisms: (1) an origin from invading angioblasts that aberrantly differentiate toward the placental microvascular phenotype in the mesenchyme of skin and subcutis or (2) an origin from embolized placental cells.

The first hypothesis proposes colonization of receptive (possibly abnormal) mesenchyme by angioblasts aberrantly "switched" toward the placental endothelial phenotype by genomic alterations or abnormal inductive influences. These initially dormant cells, perhaps reflected in the blush or blanched spot of nascent hemangiomas, might be driven to proliferate on loss from the fetal circulation, coincident with late gestation or birth, of factors negatively modulating angiogenesis in utero, or in response to emerging positive angiogenic influences in the perinatal period. Alternatively, as delineated by our second hypothesis, embolic placental endothelial cells, already committed to the GLUT1/merosin/LeY/FcγRII phenotype, might reach fetal tissues from chorionic villi through right-to-left shunts characteristic of the normal fetal circulation. Because intravascular shedding of placental cells is likely increased by placental injury, this hypothesis could explain the heightened incidence of hemangiomas seen following chorionic villus sampling,^{23 - 25} as well as the observation that intentional placental trauma, produced during embryoscopy prior to elective pregnancy termination, results in rapid development of fetal ecchymotic lesions.²⁶

These hypotheses invite consideration of angiogenic control mechanisms specifically manifest in utero to explain the explosive perinatal growth phase of hemangiomas. Whereas profound angiogenesis occurs within placental villi until near term, little angiogenesis occurs in maternal tissues or in fetal tissues after organogenesis. Factors that may orchestrate this control include vascular endothelial growth factor (VEGF), placental growth factor (PlGF, a placenta-derived structural homologue of VEGF²⁷ ), Flt-1 (a high-affinity transmembrane receptor for both PlGF and VEGF), and sFlt-1 (a placenta-derived, truncated, soluble form of Flt-1 produced by alternative splicing^{28 - 30} ). Both PlGF and VEGF induce angiogenesis in vivo and induce mitotic activity in cultured endothelial cells.³¹ sFlt-1 is released into maternal blood and amniotic fluid during gestation, and may, as a competitive sink for both PlGF and VEGF, protect maternal and fetal tissues from overvascularization.^{28 - 30} Loss of these and other placental factors at birth is likely to dramatically alter the balance of negative and positive angiogenic factors in the fetus, possibly allowing unrestrained proliferation of embolized or aberrant rests of cells expressing the placental endothelial phenotype. This idea recalls the suggestion of Boon et al³² that hemangioma growth might correlate with decreased levels of trophoblast-derived interferon. The limited period of growth and subsequent involution of hemangiomas could reflect a programmed limit to mitosis in placental-type endothelial cells, appropriate to the 9-month life span of human placenta, or perhaps vascular stabilization imposed by maturing pericytic-endothelial associations and basement membranes.

Hypotheses concerning the pathogenesis of juvenile hemangiomas must also address the strong predilection of these lesions for skin and subcutis (rarely viscera), and for the head.^{15 - 16} Tissue- and region-specific mesenchyme receptive for angioblasts or embolic placental cells might reasonably be invoked. Head mesenchyme is derived from neural crest,³³ and, unlike the mesoderm-derived mesenchyme found elsewhere, does not show intrinsic angioblast development, relying instead on invasion by exogenous, migrating angioblasts and vascular sprouts from surrounding tissues for vascularization.³⁴ Angioblast migration in chick-quail chimeras is unusually prominent in the head, suggesting heightened production of motility or chemotactic factors in that region.³⁵

The vascular antigens co-expressed by juvenile hemangioma and placenta are probably best regarded as markers of cellular specialization, rather than of cellular proliferation or immaturity, since they persist in late-stage hemangiomas and are not associated with other forms of neovascularization. Ultrastructural studies reported herein and by others³⁶ support this contention by demonstrating well-developed basement membrane structure and pericytic-endothelial associations, consistent with vascular stability,³⁷ in both proliferative- and involutive-phase hemangiomas. GLUT1 is widely expressed early in embryonic development, but fetal endothelial GLUT1 expression rapidly disappears except in microvessels of developing neuroepithelial tissues, where expression increases.^{3 ,38 - 39} In mature brain, endothelial zonula occludens–type junctions, a component of the blood-brain barrier, necessitate high GLUT1 expression for adequate blood-brain glucose transport.⁴⁰ In placenta, GLUT1 expression is required throughout gestation for adequate maternal-fetal glucose transfer,^{4 ,41} and is independent of developmental stage.⁴² High GLUT1 expression by juvenile hemangiomas may permit tumor growth by providing fuel for mitosis.

Merosin (α2-laminin) expression has been previously reported for 2 (of 3) juvenile hemangiomas.⁴³ Our results show universal expression of merosin in a large series of GLUT1-positive hemangiomas representing all stages of proliferation and involution and a wide spectrum of patient ages. Vascular basement membrane merosin expression is normally restricted to the nervous system and eye¹⁹ and placenta.²¹ The merosin (α2) chain, 1 of at least 10 genetically distinct laminin chains, is also present in certain nonvascular basement membranes, including those of placental trophoblast, Schwann cells, striated muscle, and skin, and is abundant in the mesangial matrix of mature renal glomeruli.^{13 ,20 ,44 - 46} The known heterogeneity of laminin subtypes in basement membranes and the highly restricted expression of merosin suggest tissue-specific and developmental stage–specific functional diversity. Merosin is essential, for instance, for stability and survival of fused myoblasts in vitro, apparently due to apoptosis inhibition,⁴⁷ and merosin deficiency results in severe muscular dystrophy.⁴⁸ This suggests that merosin might facilitate growth of hemangiomas by inhibiting apoptosis.

The selective, persistent expression of LeY and FcγRII we find in endothelia of hemangiomas and placental vessels is of unclear functional significance. LeY is a difucosylated type 2 chain oligosaccharide expressed on cell surfaces in a tissue- and stage-specific manner. Changing glycosylation patterns, mediated by differential display of LeY and other specific cell surface oligosaccharides, appear to be important in cell-cell recognition and adhesion during development and have been associated with malignant transformation and progression.^{49 - 51} Expression of LeY correlates with apoptotis in some studies^{52 - 53} (but not others⁵⁴ ); the constant LeY expression by hemangiomas throughout proliferation and involution reported herein does not support a close association with apoptosis in these lesions. FcγRII is a low-affinity Fc receptor that is normally expressed on macrophages, Langerhans cells, platelets, and various leukocytes^{12 ,55} and binds only aggregated IgG. Normal endothelial FcγRII expression is limited to placenta^{12 ,56} and hepatic sinusoidal lining cells (possibly Kupffer cells).⁵⁷ This suggests that endothelial FcγRII expression may be uniquely advantageous in placenta, perhaps assisting clearance of immune complexes and/or maternal-fetal IgG transport.^{12 ,58} Its potential effect on the natural biology of hemangiomas remains obscure.

In summary, we report a distinct immunophenotypic pattern consisting of 4 functionally unrelated markers of cellular specialization uniquely coexpressed by fetal microvessels of human placenta and juvenile hemangiomas at all stages. We present 2 hypotheses that might explain the shared molecular differentiation of placental vessels and juvenile hemangiomas, and that might explain some of the behaviors and known associations of the latter common, sometimes devastating, lesions. These findings suggest possible new preventive strategies and therapeutic avenues, in particular the pharmacological use of angiogenic modulators that may normally decrease in the fetal circulation on separation from the placenta at birth.

Accepted for publication February 8, 2001.

This study was supported in part by the University of Arkansas for Medical Sciences, Departments of Pathology and Otolaryngology–Head and Neck Surgery.

This work was presented in part at the 89th Annual Meeting of the United States and Canadian Academy of Pathology, New Orleans, La, March 29, 2000, and at the 13th International Workshop on Vascular Anomalies in Montreal, Quebec, May 10, 2000.

Corresponding author and reprints: Paula E. North, MD, PhD, Department of Pediatric Pathology, Arkansas Children's Hospital, 800 Marshall St, Little Rock, AR 72202.