Plasticity of endothelial cells: rapid dedifferentiation of freshly isolated high endothelial venule endothelial cells outside the lymphoid tissue microenvironment

Although all vascular endothelial cells (ECs) share certain common functions, it has become clear that considerable structural and functional heterogeneity exists along the length of the vascular tree and in the microvascular beds of various organs.^1-7  ECs either form a tight continuous monolayer in organs, where they render important barrier functions, such as in the brain (blood-brain barrier), or they form a discontinuous layer with intercellular gaps or fenestrations in organs, where the rapid exchange of fluid, particles, and cells is needed, such as in the kidney, spleen, and bone marrow.^1,6  EC diversity is also reflected at the molecular level, by vessel size–specific, tissue-specific and even disease-specific differences.^2,5,6  For instance, several organ- and tissue-specific EC receptors, called vascular addresses⁸  or vascular signatures,⁹  have recently been identified with in vivo phage display.^5,10  These vascular addresses may turn out to be useful for targeting therapies to blood vessels in various tissues or tumors.^8,9  Tissue-specific and vessel type–specific differences in ECs have also been detected at the transcriptional level by subtractive hybridization technologies^11,12  or gene expression profiling.^13-15  For instance, several tumor-specific EC molecules were identified by comparing gene expression patterns of ECs freshly isolated from blood vessels of normal or malignant colorectal tissue.¹⁴ 

Although the molecular diversity of ECs is now clearly established, the origin of this heterogeneity remains poorly understood. Recent studies have suggested that some degree of EC diversity is maintained in vitro,^7,16-19  but there is an increasing appreciation that tissues “speak” to ECs⁶  and that EC heterogeneity mostly develops through interactions with the microenvironment, either through soluble factors or cell-cell contacts. Numerous recent reports support such a regulation of EC phenotypes.^1,3,4,6,20  Thus, microvascular ECs lose organ-specific properties, such as fenestrations,²¹  when they are isolated from the organ environment and are cultured in vitro for several passages. Conversely, murine ear vessels start to express genes characteristic of heart vessels when heart tissue is transplanted to the ear,²²  suggesting EC plasticity in adult tissues. Modulation of EC specialization by the local tissue microenvironment is even more strikingly shown in studies of the blood-brain barrier, in which in vitro culture and in vivo transplantation experiments^23,24  reveal the ability of astrocytes to induce blood-brain barrier properties in ECs.^20,25  It is of further note that reciprocal signaling from ECs to the surrounding tissue may also occur.^6,26  Thus, 2 recent reports have clearly established the role of EC signals in the development of the pancreas²⁷  and liver,²⁸  and similar signaling may be important for other tissues and organs.⁶ 

To better understand the reciprocal signals between EC and the microenvironment, it is necessary to characterize EC heterogeneity at the molecular level by analyzing tissue-specific EC phenotypes and identifying locally controlled EC genes. Here, we describe a novel strategy that could help to meet these objectives. We focused on ECs from high endothelial venules (HEVs), specialized postcapillary venules controlling lymphocyte recruitment in organized lymphoid and chronically inflamed tissues.²⁰  HEVs are characterized by plump, almost cuboidal ECs (HEVECs) that express high levels of sulfated counterreceptors for the lymphocyte homing receptor L-selectin (CD62L), defined by the HEV-specific, sulfate-dependent epitope for the monoclonal antibody (mAb) MECA-79.²⁹  A critical role of the microenvironment in HEV specialization has been suggested by studies in rodents; when peripheral lymph nodes were deprived of afferent lymph, the HEV converted to a flat-walled EC morphology without MECA-79 reactivity or the ability to support lymphocyte traffic.^30,31 

In this study, we used a combination of reverse transcription–polymerase chain reaction (RT-PCR) and DNA microarray experiments to compare gene expression patterns between HEVECs freshly isolated from human tonsils without any cell culture step (D0 HEVECs) and dedifferentiated HEVECs cultured outside the lymphoid tissue microenvironment for 48 hours (D2 HEVECs). We found a rapid and striking down-regulation of several key HEV genes in the dedifferentiated HEVECs, including DARC¹¹  and fucosyltransferase Fuc-TVII.³²  DNA microarray analysis identified several other candidate HEV genes controlled by the tissue microenvironment. This strategy may be applied to other specialized EC types and may turn out to be useful for further defining EC diversity at the molecular level.

Freshly purified HEVECs (D0 HEVECs) were prepared from palatine tonsils of children (obtained using a protocol approved by the CHU Toulouse Institutional Review Board, after informed consent was provided according to the Declaration of Helsinki) up to 6 hours after tonsillectomy, without any culturing step and over a mean time period of 8 hours. Tonsillar stromal cells were prepared essentially as described.³³  To facilitate HEVEC selection from stromal cell suspension without any cell culture step, we then performed 2 steps of leukocyte depletion with specific antibodies (CD3, CD19, CD45, and CD14) coupled to DynaBeads (Dynal France, Compiègne, France). D0 HEVECs were subsequently selected using the MACS system (Miltenyi Biotec, Paris, France). The cells were labeled with 10 μg/mL mAb MECA-79 (rat immunoglobulin M [IgM] κ light chain; kindly provided by E.C. Butcher, Stanford University, CA) for 15 minutes at 4°C, followed by mouse antirat κ light chain antibody conjugated to MACS superparamagnetic microbeads (mouse antirat κ light chain; 474-01; MicroBeads; Miltenyi Biotec) for 15 minutes at 8°C to 10°C, and finally fluorescein isothiocyanate (FITC)–labeled mouse antirat κ chain IgG1 (01263329; Zymed, San Franscisco, CA) for 5 minutes at 8°C to 10°C. MECA-79–positive HEVECs were then isolated by multiple rounds of selection (1-4 ×) on a medium size (MS) MACS column in phospate-buffered saline (PBS)/1% bovine serum albumin (BSA)/2 mM EDTA (ethylenediaminetetraacetic acid) buffer, according to the manufacturer's instructions (130-042-201; Miltenyi Biotec). HEVECs were finally eluted and counted on a Neubauer slide (Polylabo, Strasbourg, France) under a fluorescence microscope (Carl Zeiss, Le Pecq, France) to assess the final cell number and the final purity. Positive cells were then centrifuged, and the pellets were instantaneously frozen and kept at -80°C.

Culture characterization and preparation of D2 HEVECs were performed as previously described.³⁴  HUVECs were isolated from fresh human umbilical cords as described by Jaffe et al,³⁵  further purified with CD105 microbeads according to the manufacturer's instructions (Miltenyi Biotec), and used without any cell culture step (D0 HUVECs).

Tissue sections from human palatine tonsils were prepared as described in Girard et al.¹¹  Briefly, 4-μm cryosections were incubated with mouse anti-DARC (1:30, clone Fy6; courtesy of R. Horuk, Richmond, CA, and D. Blanchard, Nantes, France), mouse anti–E-selectin (1:500; R&D Systems, Minneapolis, MN), mouse anti–vascular endothelial growth factor receptor-3 (anti–VEGFR-3) (1:300, clone 9D9; kindly provided by K. Alitalo, Helsinki, Finland), or FITC-conjugated rabbit anti–von Willebrand factor (anti-VWF) (1:400; DAKO, Carpinteria, CA), together with MECA-79 (1:30, rat IgM). This was followed by cyanine 3 (Cy3)–conjugated goat antirat IgM (1:100; Jackson Immunoresearch, West Grove, PA) alone or in mixture with biotinylated goat antimouse IgG1 (1:200; Southern Biotechnology, Birmingham, AL) and finally by Cy2-conjugated streptavidin (1:1000; Amersham, Piscataway, NJ). For immunocytochemistry, cytospins of tonsil stromal cell suspensions were prepared, dried overnight (22°C), and fixed in acetone (10 minutes, 22°C). The cytospins were then incubated with a mixture of MECA-79 and mAb Fy6 against DARC, followed by a mixture of Cy3-conjugated goat antirat IgM and biotinylated goat antimouse IgG1, and finally Cy2-conjugated streptavidin. For flow cytometry, single-cell suspensions of tonsil stromal cells obtained after trypsinization were incubated at 4°C with the antibody mixtures described above and were analyzed on a FACScan (Becton Dickinson, San Jose, CA).

Total RNA was extracted from D0 human umbilical vein endothelial cells (HUVECs), D0 HEVEC, and D2 HEVEC samples using the RNeasy Extraction Kit (Qiagen, Courtaboeuf, France). Each sample was preamplified with the SMART Probe Amplification Kit (Clontech, Palo Alto, CA) with 100 ng total RNA. Then 500 ng each cDNA sample was ³²P-labeled with a SMART Probe Labeling Kit and was hybridized to Atlas Human 1.2 I and 1.2 II and Atlas Human Cardiovascular Arrays (7871 and 7734, respectively; Clontech) according to the manufacturer's instructions. Each array was hybridized with at least 2 independent D0 HUVEC, D0 HEVEC, and D2 HEVEC samples to reduce intersample variability/heterogeneity. Each hybridized microarray was then exposed on a PhosporScreen and was scanned on a Storm 640 PhosporImager (Molecular Dynamics, Sunnyvale, CA). Resultant hybridization patterns were analyzed with AtlasImage 1.5 software (Clontech) and AtlasInfo 3.0 (available online at http://atlasinfo.clontech.com/atlasinfo/array-info-action.do?catalog_no=7734-1). Only signals whose intensity equaled more than 2-fold the background level were taken into account. To compare results among arrays, the hybridization signals were normalized with either the sum or the median method of the global normalization mode. Iterative comparisons between different hybridizations were then carried out on Microsoft Excel 97 (Microsoft, Seattle, WA). Only genes showing more than 4-fold normalized hybridization signal differences between D0 HEVEC and D0 HUVEC, or between D0 HEVEC and D2 HEVEC, samples were selected as differentially expressed.

For semiquantitative RT-PCR, the following primer pairs were used: 5′-GAAGAAGGCATTGGGTATGG-3′ and 5′-AGGGCTTCTGCCAGGTTCAG-3′ for Duffy blood antigen (DARC); 5′-AACTCTCAGCACCGCGACTACA-3′ and 5′-GCGTTGGTATCGGCTCTCATTC-3′ for FucTVII; 5′-GCAGCACCCAGATGTTTTCTACCT-3′ and 5′-ACTGAAAGAGGCTGGACTGTCTCC-3′ for LSST; 5′-CTTCTTGCTGTCCCGCTTCAC-3′ and 5′-CTCGGTAGAAGCAGGGGTCAA-3′ for Core1 extending β(1-3)–glucosaminyltransferase; 5′-CCATCCCAGCTATCCTGTTC-3′ and 5′-TCAGTCCTCTTGCAGCCTTT-3′ for CCL-21 (SLC); 5′-CACCTGGACCTCACGCAGCTC-3′ and 5′-ATTGGCACCAGGCCCGAAAC-3′ for type XV collagen; and 5′-ACCACAGTCCATGCCATCAC-3′ and 5′-TCCACCACCCTGTTGCTGTA-3′ for the housekeeping gene GAPDH, used as an internal control. SMART-preamplified D0 HEVEC and D2 HEVEC cDNAs were used as templates in PCR amplification cycles (94°C, 30 seconds; 60 to 65°C, 30 seconds; 70°C, 1 minute)_n with each primer pair. Five microliters of PCR reaction products were set apart for every 2 amplification cycles and were loaded on a 2% agarose gel.

Quantitative real-time PCR was performed for the type XV collagen gene using β-actin as an internal standard to control for variability in amplification. Total RNA extracted from D0 HUVEC, D0 HEVEC, and D2 HEVEC was preamplified with the SMART Probe Amplification Kit (Clontech). Five nanograms each cDNA sample was then used as PCR template under the following conditions: 94°C, 1 minute (94°C, 15 seconds; 60°C, 15 seconds; 72°C, 30 seconds) up to 40 cycles. SYBR Green I PCR assays were performed in duplicate using gene-specific primers, in an iCycler iQ real-time PCR detection system (Bio-Rad Laboratories, Hercules, CA) according to the manufacturer's instructions. An amplification plot was generated for each sample, which displayed an increase in the SYBR Green emission fluorescence detection with each PCR cycle. C_T values were calculated as the PCR numbers at which the fluorescence reached a user-defined threshold. Serial dilutions of a template were used to determine the efficiency of each primer pair and the relative amount of each target gene cDNA. For each sample, the relative expression of type XV collagen in D2 HEVEC and D0 HUVEC was then determined by quantification of type XV collagen values normalized to the endogenous reference (β-actin) and expressed as a percentage of the relative expression in D0 HEVEC.

In situ mRNA hybridization was performed essentially as described.¹⁴  Sense and antisense riboprobes approximately 1 kb in length were labeled with digoxigenin by in vitro transcription with polymerase T7 according to the DIG RNA Labeling Kit (Roche, Mannheim Germany). Frozen sections (6 μm) from human palatine tonsils were fixed in 4% paraformaldehyde/diethyl pyrocarbonate (DEPC)–treated PBS. After treatment with 0.2 N HCl, digestion with pepsin (5% in 0.2 N HCl; S3002; DAKO), and acetylation in acetic anhydride (A-6404; Sigma, St Louis, MO)/0.1 M triethanolamine (T9534; Sigma), the sections were equilibrated in 5 × SSC, prehybridized at 55°C for 2 hours with DAKO mRNA in situ hybridization solution (S3304; DAKO), and hybridized with denatured 200 ng/mL DARC and type XV collagen riboprobes overnight at 55°C.

After stringent wash and single-stranded RNA degradation, sections were blocked with blocking buffer (DIG nucleic detection kit; 1175041; Roche) diluted in maleic acid buffer (0.1 M maleic acid; M0375; Sigma), 0.15 M NaCl (S3014; Sigma), pH 7.5, and 1:20 dilution of rabbit immunoglobulin fraction (X0903; DAKO) for 30 minutes. Hybridized probe was detected using a catalyzed signal amplification kit (Genpoint; DAKO) with 2 rounds of biotinyl tyramide amplification cycles according to the manufacturer's instructions. Hybridization signals were detected through incubation with Fast Red Tris (F4648; Sigma).

For double staining by immunohistochemistry, sections were further saturated with PBS/1%BSA (A2153; Sigma), incubated with mAb Meca79 (10 μg/mL) for 30 minutes, and then incubated with FITC-conjugated antirat κ light-chain mAb (Immunotech, Marseilles, France) for 30 minutes. Sections were finally counterstained with Meyer hematoxylin solution (MHS-32; Sigma) and viewed on a Nikon Eclipse TE300 fluorescence microscope (Nikon, Tokyo, Japan).

We have previously shown that the promiscuous chemokine receptor DARC exhibits an HEV-specific expression pattern in human tonsils almost identical to that of MECA-79 antigens (Figure 1A).¹¹  DARC was first identified on erythrocytes, but it has also been found to be specifically expressed in postcapillary venule ECs of most tissues examined,³⁶  making it one of the best known molecular markers of the postcapillary venule phenotype. In the present study, we found that DARC was rapidly lost when HEVECs were isolated from their natural tissue microenvironment. Plating of stromal cells obtained after human tonsil dispersion resulted in a mixed adherent population containing 15% to 25% HEVECs coexpressing DARC and MECA-79 antigens, as indicated by indirect immunofluorescence staining of cytospins from 24-hour cultured cells (Figure 1B). However, fluorescenceactivated cell sorter (FACS) analysis on different time points of the same tonsil culture, divided in 2 parts, revealed that this population of HEVECs, which is MECA-79– and DARC-positive after 24 hours of culture, rapidly lost DARC expression and only expressed the MECA-79 epitope after 2 days ex vivo (Figure 1C). It is unlikely that these findings resulted from the preferential survival/proliferation of a subpopulation of MECA-79–positive/DARC-negative HEVECs because such a subpopulation was not observed in vivo (Figure 1A). Thus, DARC expression in postcapillary venule HEVECs did not appear to be genetically determined but, rather, controlled by the tissue microenvironment.

To extend these findings to other HEV-specific genes, we compared HEVECs cultured for 2 days (D2 HEVECs) with HEVECs freshly purified from human tonsils without any cell culture step (D0 HEVECs). To obtain highly purified D0 HEVECs suitable for molecular studies, we modified our previously described HEVEC isolation protocol.^33,34  Importantly, the contaminating leukocytes in tonsillar stromal cell suspensions were depleted with magnetic beads (Dynabeads) immunoreactive for CD45, CD3, CD19, and CD14 before multiple rounds of MACS selection with MECA-79–conjugated super-paramagnetic microbeads were performed (see “Materials and methods”). With fresh tonsils (within 6 hours after tonsillectomy), this modified protocol provided highly purified D0 HEVECs containing 95% to 98% MECA-79–positive cells, as revealed by immunofluorescence staining (data not shown). Total RNA was prepared from these freshly purified D0 HEVECs and from D2 HEVECs and were used in semiquantitative RT-PCR experiments to compare the expression of key HEV genes, including fucosyltransferase Fuc-TVII, L-selectin ligand sulfotransferase LSST/HEC-GlcNAc6ST, Core1 extending β(1-3)–N-acetylglucosaminyltransferase (Core1-β3GlcNAcT), and chemokine CCL21. DARC and GAPDH were used as controls to validate our test system. These semiquantitative RT-PCR experiments confirmed the striking down-regulation of DARC within 2 days of culture, whereas GAPDH expression was not affected (Figure 2A). We then examined a possible down-regulation of the α1 → 3 fucosyltransferase Fuc-TVII, which is specifically expressed by HEV in secondary lymphoid organs³⁷  and is known to play a key role in the synthesis of HEV-specific L-selectin ligands.³²  This analysis revealed a complete loss of Fuc-TVII mRNA in D2 HEVECs (Figure 2B). By contrast, expression of the 2 enzymes controlling synthesis of the MECA-79 epitope and sulfation of HEV-specific L-selectin ligands, L-selectin ligand sulfotransferase LSST/HEC-GlcNAc6ST,^38-40  and Core1-β3GlcNAcT⁴¹  was not altered in D2 HEVECs compared with D0 HEVECs (Figure 2C).

Another key gene for the HEV phenotype is that encoding chemokine CCL-21 (also referred to as Exodus 2, 6C-kine, TCA-4, and SLC), which is abundantly expressed in HEV⁴²  and plays an essential role in lymphocyte arrest and migration through HEV walls.⁴³  We found a significant down-regulation of chemokine CCL21 mRNA, with a difference of at least 4 PCR cycles between D0 HEVECs and D2 HEVECs (Figure 2D).

Essentially identical results—that is, the down-regulation of DARC, Fuc-TVII, and CCL21 with no effect on LSST/HEC-GlcNAc6ST and Core1-β3GlcNAcT—were obtained with independent D0 HEVEC and D2 HEVEC samples (data not shown). Together, these findings indicated that human HEVECs rapidly dedifferentiate outside their natural tissue microenvironment and down-regulate expression not only of postcapillary venule-specific markers, such as DARC, but also of key HEV-specific genes such as Fuc-TVII and CCL21.

To identify other HEV genes rapidly down-regulated outside the lymphoid tissue microenvironment, we performed DNA microarray analyses with D0 HEVECs and D2 HEVECs (Figure 3). Hybridizations were carried out with Atlas Human Cardiovascular and Human 1.2 I and 1.2 II Arrays (Clontech), covering a broad range of biologic pathways (approximately 3000 genes). We first selected candidate HEV genes preferentially expressed in HEVECs by comparing D0 HEVECs with D0 HUVECs. As seen in Figure 3, in a second screen, we searched for genes with higher expression levels in D0 HEVECs compared with D2 HEVECs. Genes preferentially expressed in D0 HEVECs (ratio greater than 4) in the 2 screens in Figure 3 were identified as candidate HEV genes rapidly down-regulated ex vivo. The identification of DARC (indicated by an arrow in Figure 3) served as an internal control and validated our experimental strategy. The genes Fuc-TVII, LSST, and Core1-β3GlcNAcT were not covered by the Atlas arrays, and CCL-21, for unclear reasons, gave only a low hybridization signal that was excluded from our analysis because it was under the cut-off used (signal intensities more than 2-fold over the background). In addition to DARC, we identified 22 other genes preferentially expressed in HEVECs compared with HUVECs that appeared to be regulated by the tissue microenvironment (Table 1). We confirmed expression of some of the corresponding proteins in MECA-79–positive HEV in situ: VWF (Figure 4A); E-selectin (Figure 4B), a specific marker of postcapillary venule endothelial cells in inflamed tissues⁴⁴ ; and VEGFR-3/flt-4 (Figure 4C), a marker of lymphatic endothelium found on tumor blood vessels but not on other blood vessels.^45,46  Similarly, the myelin and lymphocyte (MAL) antigen has recently been found to be specifically expressed in situ by HEV in human tonsils and lymph nodes.⁴⁷  In addition to VWF, VEGFR-3, E-selectin, and MAL, other genes were identified that encoded cell surface proteins (adhesion receptors, cytokine receptors, transporters), secreted factors (growth factors/cytokines, extracellular matrix proteins), nuclear factors (DNA binding proteins, transcription factors), intracellular mediators, and apoptosis-associated proteins (Table 1). As did DARC and FucTVII, these candidate HEV genes appeared to be rapidly down-regulated ex vivo after 48 hours, suggesting that their expression in vivo is under the control of the microenvironment.

HEVs have been shown to secrete high levels of extracellular matrix–associated proteins such as Hevin, mac-25/angiomodulin, and trombospondin TSP-1.^11,13,33  Therefore, among the genes identified in the present screen (Table 1), we focused on the gene encoding type XV collagen, a major extracellular matrix component.^48,49  To confirm the expression of type XV collagen in HEVECs in vivo, we first performed in situ hybridization on human tonsil sections (Figure 5A). The staining pattern obtained with antisense probes corresponding to the type XV collagen was similar to that obtained with antisense probes corresponding to the HEV-specific gene DARC. By contrast, no staining was obtained with type XV collagen sense probes. When combining in situ hybridization and immunohistochemistry with the HEV-specific mAb MECA-79 on the same tissue section, we found that type XV collagen mRNA was specifically expressed in MECA-79–positive HEV (Figure 5B). Semiquantitative RT-PCR was then performed to confirm the down-regulation of type XV collagen observed in our microarray experiments (Table 1). We found a significant down-regulation of type XV collagen mRNA, with a difference of at least 4 PCR cycles between D0 HEVECs and D2 HEVECs, and the expression of GAPDH was not affected during this interval (Figure 5C). Type XV collagen mRNA down-regulation was further confirmed by quantitative real-time RT-PCR analysis with β-actin as a standard. This revealed an approximately 12-fold higher expression of type XV collagen mRNA expression in D0 HEVECs than in D2 HEVECs (Figure 5D).

Organ- and tissue-specific EC phenotypes result from the complex interactions of ECs with their microenvironment, including surrounding cells, secreted factors, extracellular matrix, and blood pressure and flow. Therefore, isolating ECs from their natural microenvironment is likely to trigger rapid changes in EC gene expression. Using human HEVECs as a model system, we here show that this indeed is the case. By comparing HEVECs freshly isolated from human tonsils without any culturing step with dedifferentiated HEVECs grown ex vivo for 2 days, we identified several HEVEC genes rapidly down-regulated outside the lymphoid tissue microenvironment. These included vessel type–specific EC genes, such as the postcapillary venule-specific marker DARC, and tissue-specific EC genes, such as the HEV-restricted fucosyltransferase Fuc-TVII. Our results suggested that comparing freshly isolated ECs with counterparts dedifferentiated ex vivo may be a useful strategy to characterize tissue-specific EC phenotypes and to identify EC genes under control of the local microenvironment in other microvascular beds.

Several recent studies have reported that ECs retain some of their characteristics when cultured in vitro for several passages.^7,16-19  For instance, stable differences between cultured blood vessel ECs and lymphatic ECs were described.^16-18  Similarly, global gene expression profiling of 53 cultured EC samples, representing 14 distinct types of ECs, including large-vessel ECs from arteries or veins and microvascular ECs from several distinct tissues, revealed many stable vessel type–specific and tissue-specific differences.⁷  Nevertheless, the relationship between the gene expression heterogeneity observed in cultured ECs and that existing in vivo remains to be established. Our results demonstrated that rapid changes in gene expression programs occur when ECs are removed from their natural tissue microenvironment. Striking changes in HEVEC gene expression were observed as early as after 2 days ex vivo. Therefore, although cultured ECs may retain part of their differentiated gene expression programs in vitro,⁷  they are unlikely to mimic adequately the corresponding in vivo phenotypes.

Rapid dedifferentiation of HEVECs outside their natural lymphoid tissue microenvironment likely occurs because of the loss of tissue-specific inductive signals. Previous studies in rodents on the influence of afferent lymphatic vessel interruption in peripheral lymph nodes have revealed that lymph constituents are required to maintain the specialized features of the HEVEC phenotype, including plump EC morphology, expression of MECA-79–positive sulfated L-selectin ligands, and ability to support L-selectin–mediated binding and lymphocyte traffic in vivo.^30,31  The HEV-specific fucosyltransferase Fuc-TVII was decreased after 5 days and completely lost after 8 days.⁵⁰  Our findings of rapid down-regulation of Fuc-TVII mRNA levels in dedifferentiated human HEVECs (Figure 2) are in agreement with these later findings in mice. Differences in the kinetics of Fuc-TVII down-regulation could be explained by the fact that the human HEVECs were separated from other lymphoid tissue components, whereas the mouse HEVECs were exposed to inductive signals for a few days after ligation of the afferent lymphatics. Taken together, our results showed that the expression of Fuc-TVII in HEVECs is under the control of the lymphoid tissue microenvironment. Given that Fuc-TVII is essential for the biosynthesis of L-selectin ligands in HEVECs and in lymphocyte recruitment in vivo,³²  these results illustrate perfectly the fact that functionally differentiated EC phenotypes such as HEVEC are unlikely to be maintained outside their natural microenvironment.

Although Fuc-TVII mRNA was rapidly down-regulated, we found that mRNA levels of the 2 enzymes controlling synthesis of the sulfated MECA-79 epitope, L-selectin ligand sulfotransferase LSST/HEC-GlcNAc6ST^38-40  and Core1-β3GlcNAcT,⁴¹  were not altered in dedifferentiated HEVECs. This was in agreement with a previous observation that MECA-79 reactivity was maintained for a few days in cultured HEVECs, before complete loss of the epitope after 8 days.³⁴  Interestingly, we noted a significant down-regulation of mRNA for sulfotransferase LSST/HEC-GlcNAc6ST (but not for Core1-β3GlcNAcT) in HEVECs after 8 days in culture (data not shown), which might have contributed to the disappearance of the MECA-79 epitope at that time point. It remains unclear why expression of the MECA-79 epitope and the HEV-specific sulfotransferase LSST/HEC-GlcNAc6ST involved in its synthesis were maintained for a few days ex vivo whereas expression of the HEV-specific Fuc-TVII was rapidly lost. However, our results suggested that HEV-specific enzymes, such as LSST/HEC-GlcNAc6ST^38,39  and Fuc-TVII,³⁷  are regulated by different mechanisms. Therefore, it will be important in the future to further characterize the molecular mechanisms controlling HEV-specific expression of these enzymes, including identification of the HEVEC nuclear transcription factors involved in the regulation of their corresponding genes. An attractive candidate for this function is NF-HEV, a recently described homeodomain-like nuclear factor with preferential expression in HEVECs,¹²  which may play a key role in the induction of HEV-specific genes such as LSST/HEC-GlcNAc6ST and Fuc-TVII.

Among the genes rapidly down-regulated ex vivo, we identified 2 additional genes that have previously been shown to be expressed in HEVECs but not in other blood vessel ECs—chemokine CCL21,⁴²  which plays a critical role in lymphocyte recruitment from the blood,⁴³  and VEGFR-3, a tyrosine kinase receptor for growth-factors VEGF-C and VEGF-D.^45,46  Although the expression of these 2 genes was not completely lost in dedifferentiated HEVECs, both were substantially down-regulated (Figure 2; Table 1). Interestingly, chemokine CCL21 and VEGFR-3 have recently been identified as specific markers of lymphatic ECs^16-18  and appear to be coregulated by the lymphatic endothelium transcription factor Prox-1.^16,51  According to our results, these 2 genes also appear to be coregulated in HEVECs, though Prox-1 has not been shown to be expressed in these cells.

The starting point of our study was the observed rapid loss of DARC expression in cultured HEVECs after 2 days (Figure 1). This was unexpected because DARC is not only an HEV-specific molecule in lymphoid tissues,¹¹  it is one of the best postcapillary venule-specific markers in nonlymphoid tissues,³⁶  suggesting a lack of dependency on regulatory factors selective for lymphoid tissue. Instead, hemodynamic forces (blood flow and pressure) or cellular factors (pericytes) common to postcapillary venules in lymphoid and nonlymphoid tissues may be involved.⁵²  Thus, fluid shear stress generated by blood flow can profoundly influence the EC phenotype by modulating gene expression.⁵³  Because postcapillary venules are believed to differentiate from primitive capillaries in response to steady, slow blood flow and low pressure,⁵²  DARC expression in HEVECs and other venular ECs may be controlled by these conditions in vivo. Alternatively, pericytes may contribute to DARC expression, or specific signaling pathways such as the Tie2-angiopoeitin system may be involved.⁵² 

As is DARC, some of the genes we identified by DNA microarray analyses might also be down-regulated in dedifferentiated HEVECs as a result of absent inductive signals from hemodynamic forces or surrounding cells, such as pericytes. Other genes, however, could be controlled by the lymphoid tissue microenvironment, as is Fuc-TVII. In any case, future studies will be required to validate the candidate HEV genes showing rapid down-regulation ex vivo (Table 1). We initiated the present work by focusing on the extracellular matrix component type XV collagen^48,49  and confirmed its abundant and specific in situ expression in MECA-79–positive HEVECs and its rapid down-regulation ex vivo with an approximately 12-fold message reduction in dedifferentiated HEVECs. Identifying type XV collagen in our screen (Table 1) suggested that extracellular matrix–associated, HEVEC-related proteins are differentially regulated because previous studies have revealed that the expression of hevin, another major secreted component of HEV,³³  is maintained for at least 8 days in cultured HEVECs.³⁴ 

In addition to the 23 genes preferentially expressed in D0 HEVECs compared with D0 HUVECs and D2 HEVECs (Table 1), we identified 17 other genes down-regulated in D2 HEVECs that were of lower interest because they were not preferentially expressed in D0 HEVECs compared with D0 HUVECs (data not shown; complete microarray data available under accession number GSE751 at http://www.ncbi.nlm.nih.gov/geo/). We also excluded from our study 86 genes found to be up-regulated in D2 HEVECs compared with D0 HEVECs. Although the microenvironment is likely to repress the expression of certain genes, it is difficult to identify these genes among the many other genes artificially induced by in vitro culture conditions. Finally, we identified 37 genes preferentially expressed in D0 HEVECs compared with D0 HUVECs that were not down-regulated after 48 hours ex vivo. As previously discussed for MECA-79 and LSST, some of these genes may also be controlled by the microenvironment, but with different kinetics of down-regulation ex vivo. However, others are likely to represent stable HEVEC characteristics, suggesting that (epi)genetic influences may also be important in determining HEV phenotype.

In conclusion, our study revealed that differentiated ECs such as HEVECs rapidly lose their specialized characteristics when isolated from their natural tissue microenvironment. Our approach, based on the comparison of ECs freshly isolated from human tissue, without any culturing step, with their counterparts cultured for 2 days ex vivo may be extended to other differentiated EC phenotypes. In the end, this should facilitate further characterization of EC diversity and identification of other locally regulated tissue-specific EC genes, thereby providing a better understanding of how tissues speak to ECs.

We thank Dr Eugene Butcher (Stanford, CA), Dr Kari Alitalo (Helsinki, Finland), Dr Richard Horuk (Richmond, CA), and Dr Dominique Blanchard (Nantes, France) for reagents. We also thank Luc Aguilar and Laure Americh (Endocube, Labège, France) for help with in situ hybridization. Special thanks go to others members of the Laboratory of Vascular Biology (IPBS, Toulouse, France) for stimulating discussions.