Aberrant Lymphatic Endothelial Progenitors in Lymphatic Malformation Development

Lymphatic malformations (LMs) are vascular anomalies thought to arise from dysregulated lymphangiogenesis. These lesions impose a significant burden of disease on affected individuals. LM pathobiology is poorly understood, hindering the development of effective treatments. In the present studies, immunostaining of LM tissues revealed that endothelial cells lining aberrant lymphatic vessels and cells in the surrounding stroma expressed the stem cell marker, CD133, and the lymphatic endothelial protein, podoplanin. Isolated patient-derived CD133+ LM cells expressed stem cell genes (NANOG, Oct4), circulating endothelial cell precursor proteins (CD90, CD146, c-Kit, VEGFR-2), and lymphatic endothelial proteins (podoplanin, VEGFR-3). Consistent with a progenitor cell identity, CD133+ LM cells were multipotent and could be differentiated into fat, bone, smooth muscle, and lymphatic endothelial cells in vitro. CD133+ cells were compared to CD133− cells isolated from LM fluids. CD133− LM cells had lower expression of stem cell genes, but expressed circulating endothelial precursor proteins and high levels of lymphatic endothelial proteins, VE-cadherin, CD31, podoplanin, VEGFR-3 and Prox1. CD133− LM cells were not multipotent, consistent with a differentiated lymphatic endothelial cell phenotype. In a mouse xenograft model, CD133+ LM cells differentiated into lymphatic endothelial cells that formed irregularly dilated lymphatic channels, phenocopying human LMs. In vivo, CD133+ LM cells acquired expression of differentiated lymphatic endothelial cell proteins, podoplanin, LYVE1, Prox1, and VEGFR-3, comparable to expression found in LM patient tissues. Taken together, these data identify a novel LM progenitor cell population that differentiates to form the abnormal lymphatic structures characteristic of these lesions, recapitulating the human LM phenotype. This LM progenitor cell population may contribute to the clinically refractory behavior of LMs.


Introduction
Vascular anomalies are a heterogeneous group of lesions with arterial, venous, or lymphatic components that develop in utero or shortly after birth.These lesions are classified into malformations and tumors, based on histological classification, endothelial cell morphology, and clinical behavior [1][2][3].Vascular malformations are further classified based on the cellular subtype of the malformation, with lymphatic malformations (LMs) consisting of abnormal lymphatic vasculature.LMs are subdivided on the basis of morphology, and include macrocystic (lumen >1cm), microcystic (lumen <1cm), mixed macrocystic and microcystic (mixed), and diffuse LMs (referred to as generalized lymphatic anomalies, GLA) [2][3][4].
The lymphatic vasculature functions in maintenance of interstitial fluid balance, mounting immune responses, and uptake of lipids and lipid-soluble nutrients from the intestines.Consequently, individuals with LMs are subject to significant morbidities resulting from disruption of these essential functions, including lymphedema, lymphatic fluid pooling (chylous ascites and chylothorax), and intralesional bleeding.Mass effects of large LMs can impair vital functions, such as cervicofacial lesions that cause airway obstruction or impingement on the eye.Superinfection of tissues in which lymphatic flow is impaired can lead to overwhelming sepsis.Despite this significant burden of disease, the pathobiology of these lesions is poorly understood.
Here, we identify a previously undescribed lymphatic malformation progenitor cell (LMPC) population in LMs that line the aberrant lymphatic vessels and reside in adjacent parenchyma in patient tissues.In LM tissues, LMPCs co-expressed the stem cell marker, CD133, and the lymphatic endothelial cell (LEC) marker, podoplanin.CD133 + cells isolated from LM patient tissues and fluid aspirates were multipotent, and expressed markers of stem cells, circulating endothelial precursor cells, and LECs.We compared CD133 + LM cells to CD133 − LM cells isolated from LM fluids.Relative to CD133 + LM cells, CD133 − LM cells had significantly lower expression of stem cell markers, maintained expression of circulating endothelial precursor markers, and expressed increased levels of differentiated LEC markers.Unlike CD133 + LM cells, CD133 − LM cells were not multipotent, suggesting that CD133 − LM cells represent differentiated lymphatic malformation endothelial cells (LMECs).We demonstrate that LMPCs recapitulate the LM phenotype in a mouse model.When xenografted in mice, CD133 + LM cells differentiated into LECs that formed aberrant lymphatic vessels, morphologically and histologically similar to those observed in LM patient tissues.Taken together, these data suggest a progenitor cell origin for human LMs.

Clinical Samples
Resected tissues and aspirated fluids were acquired from pediatric LM patients (infants to adolescents).For histologic and molecular characterization, tissues were fixed in 4% paraformaldehyde, incubated in 20% sucrose/PBS, and frozen in OCT or fixed in formalin and paraffin-embedded.Cells were isolated immediately from resected tissues and aspirated fluids.Description of LM specimens and methodologies performed are presented in Table S1 in S1 File.

Cell Culture
LM cells were isolated from tissues or centrifuged fluids using the anti-CD133 bead selection system (Miltenyi Biotec) as described [11,12].Cells were maintained in EGM-2 media (Lonza) supplemented with 18% FBS on fibronectin-coated plates.Fluorescence-activated cell sorting (FACS) of CD133 + live cells was performed to generate CD34-positive or negative populations, and podoplanin-positive or negative populations.Hemangioma stem cells (HemSC) and human dermal lymphatic endothelial cells (HdLECs) were isolated and maintained as described [11][12][13].Human bone marrow-derived mesenchymal stem cells (MSCs) were purchased and maintained as described by the manufacturer (Lonza).

Quantitative RT-PCR (qPCR)
RNA was isolated (RNeasy Mini Kit, Qiagen) and cDNA synthesized (SuperScript First-Strand Synthesis System, Invitrogen) [17].qPCR with Sybr Green Master Mix (ABI) was performed in triplicates using a CFX96 PCR Cycler (Bio-rad).Gene-specific PCR products cloned into pDrive (Stratagene) served as standards and reactions were normalized to β-actin.Primers are listed in Table S3 in S1 File.

Ethics Statement
For human studies, de-identified tissues and aspirated fluid were collected from discarded surgical specimens.As protected health information (PHI) was neither stored nor disclosed to the researchers, this study was deemed exempt by Columbia University IRB (AAAA7338).Therefore, there was no verbal or written consent required.Mouse studies were approved by Columbia University IACUC (AC-AAAC1609 and AC-AAAG5852).

LM tissues contain CD133 + stromal and lymphatic endothelial cells
Involved and uninvolved tissues resected from a patient with a mixed cervicofacial LM were stained for LEC markers, podoplanin and LYVE1.In uninvolved tissues, morphologically normal lymphatic vessels stained strongly for LYVE1 and podoplanin (Fig. 1A).In LM tissues, cells lining the irregularly dilated or ectatic lymphatic channels stained weakly for LYVE1 and strongly for podoplanin (Fig. 1A, Fig. IA in S2 File).In the LM tissue parenchyma, stromal cells stained weakly for podoplanin.These podoplanin + stromal cells had a mesenchymal morphology, potentially consistent with that of an LM progenitor cell.
Circulating endothelial precursor cells have been shown to express CD133 [18,19].We determined CD133 and podoplanin expression in LM tissues from 7 patients, as well as in control neonatal foreskin (postnatal day 1), and 3 patient-matched uninvolved tissues.In neonatal foreskin, lymphatic vessels expressed CD133 consistent with the lymphatic vessel immaturity and active remodeling proposed to occur shortly after birth (Fig. 1B, Figs.IA, IB in S2 File) [20].In contrast, CD133 expression was very low in the mature lymphatic vessels in the uninvolved tissue (Fig. 1B).The ectatic lymphatic endothelium and the cells lining the lymphatic channels in LMs expressed both CD133 and podoplanin (Fig. 1B).In the parenchyma, two different sub-populations of CD133 + cells were observed, CD133 + /podoplanin low and CD133 + / podoplanin − cells.Parenchymal CD133 + cells were detected in all LM tissues evaluated (mixed cervicofacial LM (n = 2), macrocystic subcutaneous LMs (under the skin; n = 2), macrocystic mesenteric LM (n = 1), Gorham's LM (n = 1), generalized lymphatic anomalies (GLA; n = 1) (Fig. 1B, data not shown).CD133 expression in the lymphatic endothelium in LMs was similar to the immature neonatal lymphatic vasculature, suggesting arrested or defective development of LM lymphatics.In contrast, CD133 expression was only weakly expressed in normal LECs, consistent with their mature status.CD133 + /podoplanin + cells were observed in several different subtypes of LMs from various anatomic locations.Despite this anatomic and subtype heterogeneity, our data indicates that these CD133 + /podoplanin + LM cells are common to multiple lymphatic anomalies.
In murine inflammatory and tumor lymphangiogenesis, a subset of LECs has been suggested to arise from a myeloid/monocyte cell [14,22,23].The expression of the myeloid marker, CD45, and the monocyte marker, CD11b, was assessed in CD133 + and CD133 − LM cells.CD11b expression was absent or low in a majority of the podoplanin + cells, irrespective of their CD133 status (Fig. IIA in S2 File).CD45 expression was not observed (data not shown).
We assessed transcript expression of the stem cell markers, Oct4, NANOG and Sox2, as well as the endothelial precursor marker, c-Kit.Expression of c-Kit, Oct4, and NANOG was detected in CD133 + LM cells, and significantly lower or absent in CD133 − LM cells (Fig. IIB in S2 File).Sox2 was not expressed in either LM cell population (data not shown).
FACS and immunostaining were performed on patient-matched CD133 + and CD133 − LM cells to assess expression of the mature endothelial markers, VE-cadherin and CD31.All CD133 − LM cells expressed CD31, where as its expression was inconsistently observed in CD133 + LM cells (Fig. 3B, 3C).Although VE-cadherin expression was observed in majority of CD133 − LM cells by immunostaining, its expression was poorly localized to adherens junctions as compared to HdLECs (Figs. 3C bottom panels).Consistent with the immunostaining of the CD133 − LM cells isolated from a GLA, FACS for VE-cadherin did not detect cell surface expression (Figs. 3B, 3C).
Patient-matched CD133 + and CD133 − LM cells were immunostained for the lymphatic endothelial markers, LYVE1 and podoplanin.Expression of podoplanin was higher in CD133 − LM cells relative to CD133 + LM cells (Fig. 3D).When compared to HdLECs, CD133 − LM cells expressed lower levels of LYVE1 and higher levels of podoplanin (Fig. 3D), similar to the abnormal lymphatic endothelium in LM tissues (Fig. 1A) and the transcriptional analyses (Fig. 3A).
Overall, expression of LEC genes was lower in CD133 + LM cells relative to CD133 − cells, whereas CD133 − LMs were more similar to HdLECs.Based on our assessment of markers, we designated CD133 + LM cells as lymphatic malformation progenitor cells (LMPCs), and CD133 − LM cells as lymphatic malformation endothelial cells (LMECs).

LMPCs are multipotent and can be induced to differentiate down multiple lineages
We hypothesized that LMPCs are progenitor-like and predicted that they would be multipotent.We thus assessed the ability of CD133 + LMPCs to be induced down multiple cell lineages.5 LMPC populations and 3 LMEC populations were cultured in induction media for adipocytes, osteoblasts, and vascular smooth muscle cells (VSMCs).After two weeks in adipogenic medium, LMPCs, but not LMECs, differentiated into adipocytes as assessed by Oil Red O staining for lipid accumulation (Fig. 4A LMPCs cultured in mural cell differentiation media for two weeks expressed alpha smooth muscle cell actin (αSMA), consistent with VSMC differentiation (Fig. 4C).LMPCs maintained in growth medium expressed higher levels of the pericyte marker, NG2, than those in the mural cell differentiation media (Fig. 4C).We assessed whether NG2 expression was observed in LM tissues and found that CD133 + cells co-expressed NG2 (Fig. IVA in S2 File).Cultured CD133 + LMPC populations expressed NG2, while NG2 expression was lower in patientmatched CD133 − LMECs (Figs.IVB, S4C in S2 File).Thus, NG2 appears to be a marker of LMPCs and its expression was lower in LMECs relative to LMPCs.
FACS analyses indicated that LMPCs are a heterogeneous population of cells (Fig. 2).To determine if CD34 + or podoplanin + LMPCs are multipotent, three early passage LMPC populations were live cell sorted for CD34 and podoplanin, and positive and negative populations collected.Sorted LMPCs were cultured in adipogenic, osteogenic and mural cell differentiation media for 2 weeks and cell differentiation assessed.CD34 + and podoplanin + LMPCs differentiated into fat, bone and VSMCs (Figs.VA-C in S2 File).CD34 − or podoplanin − cells from two LMPC populations failed to expand after FACS.CD34 − or podoplanin − cells isolated from a subcutaneous LM did differentiate down the bone pathway, but poorly differentiated down fat or VSMC pathways (data not shown).Thus, multipotency was observed in the CD34 + and podoplanin + LMPC populations.

LMPCs differentiate into LMECs in vitro
To determine if LMPCs could differentiate into LMECs, we cultured five LMPC populations and three CD34 + or podoplanin + LMPC populations in LEC differentiation media.After two weeks, LMPCs strongly expressed VE-cadherin and podoplanin, while CD31 and LYVE1 expression were modestly upregulated (Figs.5A, 5B; Fig. VD in S2 File).In contrast, expression of all four endothelial and lymphatic endothelial proteins was low in LMPCs maintained in growth media.In LMPCs grown in LEC differentiation media, VE-cadherin was cytoplasmic and not localized to the adherens junctions, and LYVE1 was poorly expressed, similar to LMECs (Figs. 3C, 5A, 5B).In all LMPCs induced down the LEC fate, podoplanin and VEGFR-2 transcripts were significantly upregulated, but LYVE1 was not induced (Fig. 5C).Similar to the heterogeneity observed in the different LMEC populations, LEC markers, VE-cadherin, Prox1, and VEGFR-3 were only upregulated in a subset of the LMPCs in differentiation media.Although not normally expressed by LECs, the blood endothelial cell-specific VEGFR-1 was

LMPCs recapitulate the LM phenotype in mice
To determine whether isolated LMPCs have the capacity to form lesions in vivo that phenocopy human LMs, six CD133 + LMPC populations isolated from cervicofacial (CF, n = 2), mesenteric (Mes, n = 2), and subcutaneous (SC, n = 1) LMs, and a GLA were suspended in Matrigel and implanted into GFP-expressing immunocompromised mice.Matrigel implanted alone or Matrigel with normal neonatal HdLECs or MSCs served as controls.The proposed progenitor cells of infantile hemangioma, HemSCs (hemangioma stem cells), were also assessed in the xenograft model.Implants were removed at 3 and 5 weeks, and sections were H&E stained.Similar to patient LM tissue, large channels and ectatic lymphatic vessels were observed in the LMPC implants (Fig. 6A).In both LMPC implants and LM patient tissues, sloughing of the endothelium into the dilated lymphatic channels was observed.Unlike LMPCs isolated from lymphatic anomalies, HemSCs from a blood vessel anomaly, infantile hemangioma [11], did not develop dilated channels.Aberrant vessels were not observed in Matrigel implants, or implants that contained MSC or HdLEC.
LMPC and control implants were immunostained for GFP to identify host-derived cells, and human-specific podoplanin to visualize LM-cell derived lymphatic vessels.In HdLEC implants, lymphatic vessels stained strongly for podoplanin and were observed to lie between host-derived GFP + adipocytes (Fig. 6B).In LMPC implants, the cells lining the dilated channels stained positive for podoplanin and negative for GFP, demonstrating that the abnormal lymphatic vessels were derived from the LMPCs (Fig. 6B).Podoplanin expression was not observed in MSC or Matrigel controls (Fig.

VI in S2 File).
To further characterize the podoplanin + vessels in the LMPC implants, patient-matched CD133 + LMPC and CD133 − LMEC implants were stained for additional lymphatic endothelial specific proteins, LYVE1, Prox1 and VEGFR-3.Similar to patient LM tissues, expression of LYVE1 was variable in the cells lining the dilated lymphatic vessels of LMPC and LMEC implants (Fig. 7A).The aberrant lymphatic vessels in LMPC and LMEC implants also stained positive for Prox1 and VEGFR-3 (Figs. 7B, 8).The expression levels and patterns of podoplanin, LYVE1, Prox1 and VEGFR-3 were similar in the dilated vessels of LMPC and LMEC implants.In the xenograft model, LMPCs differentiated into LMECs and recapitulated the LM phenotype.
However, pericytes are defined as negative for the endothelial precursor marker, CD34, which is expressed by LMPCs [32,33].Thus, the cell-type or organ of origin for LMPCs remains undefined.
Podoplanin + /VEGFR-3 + /CD31 + LECs isolated from LM tissues have previously been shown to develop abnormal podoplanin + /LYVE1 low lymphatics in Matrigel implants in mice [34].We demonstrate that CD133 + LMPCs that are podoplanin + /VEGFR-3 + /CD31 − , as well as CD133 − LMECs that are podoplanin + /VEGFR-3 + /CD31 + , develop ectatic aberrant lymphatics that phenocopy the lymphatic vessel morphology in LM tissues.Like LMEC implants and LM tissues, lymphatic vessels in LMPC implants express podoplanin, VEGFR-3, and Prox1, and The abnormal dilated lymphatic vessels in LM patients and LMPC implants shared some similarities with the immature lymphatic vessels in the neonatal dermis.Unlike mature lymphatics in uninvolved tissues, the immature lymphatic vessels in neonatal tissue expressed high levels of CD133, similar to LM vessels.In contrast, LYVE1 expression was discontinuous and reduced in the aberrant lymphatic vessels in LM tissues and LMPC or LMEC xenografts.Lowlevel LYVE1 expression was also observed in LMECs isolated from patients or differentiated from LMPCs in vitro.LYVE1 function in the lymphatic vasculature is poorly understood, and embryonic lymphatic vessel growth appears unaffected in LYVE1 knockout mice [35].However, closer analysis of the vessel morphology of adult mice has demonstrated that LYVE1 knockout mice have distended or disorganized lymphatics in the intestines and liver [36].LYVE1 was recently shown to bind FGF-2 and function to suppress FGF-2-induced LEC proliferation, migration, and invasion [37].Thus, the defect in LYVE1 expressions observed in LMECs may contribute to the abnormal dilated lymphatic phenotype observed in LM lesions.
Unlike HdLECs that formed uniform and nondilated lymphatics in the mouse model, LMPCs differentiated into LMECs that formed ectatic lymphatic vessels, with sloughing of the abnormal lymphatic endothelium into vessel lumens.This sloughing may be due to defective LEC-LEC interactions.Consistent with this notion, LMECs were abundant in the LM patient fluid aspirates.We posit that this LMEC shedding may be secondary to defects in VE-cadherin function.Although VE-cadherin was expressed in LMECs or LMEC differentiated from LMPCs, it was often not present in adherens junctions.This improper localization of VEcadherin may be secondary to defects in other components of the adherens junctions, such as αor β-catenin.
Patients with well-circumscribed macrocystic LMs can often be effectively treated by surgical excision of the affected area, or by sclerotherapy.However, a majority of LMs present with wholly or partially microcystic or diffuse disease that responds poorly to surgery and sclerotherapy, with recurrence observed in 22-59% of LM patients [5,6].The presence of lymphatic progenitor cells, after surgical or ablative chemical treatment, may explain this high rate of recurrence.Thus, therapeutic interventions that target both the abnormal lymphatic endothelium and the progenitor cell population in LMs may be required for clinical efficacy.We propose that identification of LMPCs and the development of a mouse LM model that we describe here will provide critical tools to allow for development of treatment options for patients with LMs.

Fig 4 .
Fig 4. Differentiation of LMPCs into fat, bone and smooth muscle cells.(A) Oil Red O staining of LMPCs isolated from macrocystic cervicofacial (Macro CF) LM and microcystic mesenteric (Micro Mes) LM after 2 weeks in growth media (control) or adipogenic media.(B) Alkaline phosphatase (Alk Phos) staining of LMPCs isolated from macrocystic Macro CF LM and Micro Mes LM after 2 weeks in growth media (control) or osteogenic media.(C) NG2 and alpha smooth muscle actin (αSMA) staining of LMPCs isolated from Micro Mes LM after 2 weeks in growth media (control) or mural cell differentiation (Diff) media.Scale bars: 50μm.doi:10.1371/journal.pone.0117352.g004