Endothelial cells have many characteristics in common, but significant morphological and functional differences exist between endothelial cells from different anatomic sites. The specific glomerular endothelial (GEn) cell transcript repertoire is unknown. We sought to determine whether endothelial cells derived from bovine glomeruli display a distinct transcriptional profile compared with bovine aortic endothelium (BAE) under identical conditions. Serial analysis of gene expression (SAGE), which includes known and unknown transcripts, was used to make the comparison. The GEn and BAE SAGE libraries contain 36,844 and 26,452 total tag sequences, respectively. Among 6,524 unique tag sequences represented at least 2 times in the 2 libraries, 2,094 (32%) were matched to well-characterized bovine cDNA sequences (358 tags) or expressed sequence tags (EST). Identification of the human homolog was achieved for 1,035 of these tags. Forty-two tags were differentially expressed in GEn. For 25 of these, the bovine cDNA or EST, and for 17 the human homolog was identified. Among all transcripts with a known bovine and human tag, seven were expressed at levels more than 10-fold higher in cultured GEn cells compared with all other SAGE libraries. The transcript “DKFZp564B076” was localized by in situ hybridization to glomerular endothelium in vivo and was shown by real-time RT-PCR to be highly abundant in glomeruli compared with aortic intima. This work supports the concept that differences in the transcriptional profile of endothelial cells from distinct origins are observed under otherwise equivalent conditions. Furthermore, we have identified the first known transcript predominant in glomerular endothelium in vivo.
- gene expression
- microvascular endothelium
- macrovascular endothelium
- serial analysis of gene expression libraries
the endothelium forms a single layer of cells lining the lumen of all blood vessels and regulates hemostasis, leukocyte recruitment, and vasomotor tone. Although all endothelial cells share many basic characteristics, significant morphological and functional differences exist between endothelial cells in different locations. For example, the renal glomerular endothelium is densely perforated by fenestrae and therefore highly permeable to water and small solutes (32), whereas the endothelium of the brain microvasculature forms tight junctions and is relatively impermeable to water and solutes (19). The glomerular endothelium also has a particular susceptibility to activation and injury in the hemolytic uremic syndrome (27) and in thrombotic thrombocytopenic purpura (24). There are also differences in microvascular endothelial cell phenotypes within the same organ. Although both renal peritubular and glomerular endothelial (GEn) cells are fenestrated, the plasmalemma vesicle-associated protein (PV-1) found in the stomatal diaphragm of endothelial cell caveolae is observed in the bridging diaphragm of peritubular but not GEn cell fenestrae (34). While such studies suggest that the endothelial cell phenotype is determined by the microenvironment, other studies indicate that the microenvironment is not the sole determinant of location-specific endothelial cell differentiation. In this regard, only brain microvascular endothelial cells developed tight junctions, and only aortic endothelial cells expressed abundant Weibel-Palade bodies under identical culture conditions (8). Similarly, we reported that Weibel-Palade body density increased in response to shear stress in aortic but not GEn cells under equal culture conditions (29). Also, when various types of tumors were grafted in brain or subcutaneous tissue in mice, the ultrastructure of capillary endothelium invading the tumor was determined mainly by the tumor and did not match the phenotype of the vasculature at the site of implantation. Nevertheless, endothelia from different host sites invaded the tumors at different rates because of differences in vascular endothelial growth factor (VEGF) receptor expression (31). This strongly suggests that the microenvironment influences endothelial cell phenotype, yet the transcriptional repertoire of endothelial cells does not revert fully to a common phenotype when the cells encounter the same microenvironment.
Published systematic evaluation of potential differences in the transcriptional repertoire of distinct, quiescent endothelial cells has been limited to “serial analysis of gene expression” (SAGE; see Ref. 35) of tumor vs. normal endothelium in the colon (33). Here we sought to determine whether normal endothelial cells derived from renal glomeruli display a distinct transcriptional repertoire compared with aortic endothelium when subjected to identical culture conditions. SAGE was used because it enables quantitative comparison of frequencies of already known and unknown transcripts between distinct cell populations. Our data demonstrate that significant transcriptional differences exist between glomerular and other endothelial cells. Using SAGE data from cultured cells and microdissected glomeruli (5), we were able to identify one transcript that also is expressed at much higher levels in GEn cells in vivo compared with other endothelial and nonendothelial cells.
MATERIALS AND METHODS
All bovine aortic endothelium (BAE) and GEn cells were clonal isolates free of contaminating cells, as determined by dual labeling with Hoechst 33342 and acetylated low-density lipoprotein (LDL) uptake, and expressed von Willebrand factor antigen (3). BAE cells were initially isolated from calf thoracic aorta as described (2) and propagated on gelatin-coated plates in RPMI 1640 containing 15% FBS, and 5 U/ml penicillin and 5 μg/ml streptomycin (P&S). GEn cells derived from bovine kidneys as previously described (3) require growth supplements for cloning and propagation. They were therefore cultured on gelatin-coated plates in RPMI 1640 containing 8 ng/ml acidic fibroblast growth factor, 20% FBS, 0.1 μg/ml sodium heparin, and P&S until 10 days before RNA isolation. Upon reaching confluency, both quiescent BAE and GEn cells were then maintained under identical conditions in RPMI 1640 containing 10% FBS, and P&S for 10 days before RNA harvest.
Cells were lysed directly in P100 culture dishes with 7 ml of the TRIzol reagent, and total RNA was extracted as described (7). The RNA quality was verified by electrophoresis through formaldehyde agarose gels and quantitated by measuring absorption at 260 nm. For construction of the SAGE libraries and for RT-PCR, poly(A)+ RNA was isolated from total RNA using an Oligotex-kit from Qiagen (Hilden, Germany). This step was repeated one time to minimize contamination by ribosomal RNA. mRNA (5–8 μg) was used for double-strand (ds)-cDNA synthesis (cDNA Synthesis Kit; Invitrogen, Carlsbad, CA) using biotinylated 18-mer oligo(dT) primer to allow subsequent immobilization of 3′-cDNA ends on streptavidin beads.
Construction of SAGE libraries.
SAGE libraries were constructed from streptavidin-labeled cDNA of GEn and BAE cells at passage 9. NlaIII was the anchoring and BsmFI the tagging enzyme (35). PCR amplification of the ditags was performed using biotinylated linker-specific primers in 180 reactions for 26 cycles. “No template” controls remained negative after 35 cycles of PCR amplification. All DNA purification steps were performed using a Qiaquick Purification Kit (QIAGEN). PCR-based dideoxy chain termination sequencing was then performed using an M13 forward primer and a Beckman DTCS sequencing kit (95°C for 25 s; 45°C for 35 s; 60°C for 4 min) on a TECHNE (Cambridge, UK) Genemate cycler. Sequencing was performed on a Beckman CEQ 2000, eight-channel capillary electrophoresis sequencer (Beckman-Coulter, Fullerton, CA). The “call threshold” was set at 0.9, which signifies a ≥90% probability of a correct base call. All other steps and all linker or primer sequences were according to the original SAGE protocol developed by Velculescu et al. (35) and available at www.sagenet.org. Electronic extraction of tags from concatemer sequences and analysis of tag frequency were performed using the SeqTools software (Dr. Soren Rasmussen, Copenhagen, Denmark, http://www.seqtools.dk). All possible linker sequences were electronically subtracted from the final tag list.
Identification of transcripts from tags.
For tag-to-gene matching, the bovine reliable tag database “SAGEmap_tag_ug-rel” ([ftp.ncbi.nih.gov][pub][sage][map][Bt][NlaIII]) was imported into a Filemaker Pro Database, followed by matching of the GEn and BAE libraries by tag sequence. To verify tag-to-gene matches for all tags presented in Supplemental Table A (supplemental tables for this article may be found at http://ajprenal.physiology.org/cgi/content/full/00076.2004/DC1), the electronically identified bovine cDNA or expressed sequence tag (EST) was examined individually for appropriate tag orientation and location. To identify homologous human cDNA sequences, bovine cDNA and EST sequences containing GEn or BAE SAGE tags were matched using the “blastn” algorithm against the nonredundant GenBank database. Matches were checked for orientation and distance of the bovine SAGE tag from the poly(A)+ tail of the longest human sequence. For these, the gene name and Unigene symbol were recorded. Tag-to-gene matches were assigned four levels of certainty as shown in supplemental Table A.
For each unique tag sequence, its proportion in the GEn and BAE cell libraries was compared by χ2 analysis, using absolute tag counts (23). The critical χ2 value is 3.85 for P < 0.05, 6.65 for P < 0.01, and 10.83 for P < 0.001. The GEn-to-BAE ratio of expression frequency (tag count/100,000) was also established for each tag. All differentially expressed tags were examined using a database subroutine to find single nucleotide mismatches within tags.
Human endothelial cell SAGE libraries.
To evaluate available endothelial cell SAGE libraries for endothelium-predominant tags and to compare tag frequencies between bovine and human endothelial cell libraries, two strategies were employed. First, three endothelial cell SAGE libraries available at SAGEmap [SAGE Duke human dermal microvascular endothelium (HMVEC): gene expression omnibus sample (GSM) 706; SAGE Duke HMVEC + VEGF: GSM707; SAGE Hemangioma 146: GSM1516] and two endothelial cell SAGE libraries generated from freshly isolated colonic endothelium or colon carcinoma endothelium (kindly provided by Drs. K. Kinzler and B. Vogelstein; see Ref. 33) as well as 132 human nonendothelial cell SAGE libraries were imported into our database.
The human SAGE libraries were then matched against the SAGE Genie Hs.best_tag database at http://cgap.nci.nih.gov/SAGE by tag sequence (where Hs indicates Homo sapien). SAGE-Genie uses what is currently the most accurate bioinformatics tool to achieve tag-to-gene matches (4). The GEn and BAE libraries were imported into the same database by matching human Hs. Unigene symbols. During the course of this study, new SAGE libraries were submitted by others into the SAGEmap database. For the final comparison of putative Gen-predominant tags, four additional human endothelial cell libraries (normal liver endothelium: GSM14798; normal breast endothelium: GSM14754; endothelium from breast carcinoma: GSM14748; and microscope-dissected mammary gland ductal in situ carcinoma endothelium DCIS2: GSM688) were used, as noted where appropriate in the text and in Table 5. Similarly, the total number of human SAGE libraries found in SAGE Genie, and used for the analysis in Table 1, had increased to 208 as of February 2004. SAGE libraries can be retrieved by GSM accession number at http://www.ncbi.nlm.nih.gov/geo.
Single-stranded cDNA (ss-cDNA) was generated from 10 μg total RNA (Single Strand cDNA Synthesis Kit; Invitrogen) from GEn and BAE cultures at passage 13 distinct from those used for production of the SAGE libraries, but otherwise cultured and prepared identically. Serial dilutions of BAE and GEn ss-cDNA were used to determine the dilution at which the template for both samples resulted in equivalent PCR amplification of the ubiquitous transcript glyceraldehyde-3-phosphate dehydrogenase (GAPDH), as quantified and compared by densitometry (AlphaEase FC Software; Alpha Innotec, San Leandro, CA). The conditions, primer sequences, and the corresponding expected PCR-product size for each transcript represented by a tag are shown in Table 1. ss-cDNA was similarly generated using RNA derived by TRIzol extraction from freshly isolated bovine glomeruli and bovine aortic intima.
An ABI Prism 7500 Sequence Detector thermal cycler was used for quantitative real-time PCR using SYBR Green PCR Master Mix. cDNA was produced using the RevertAid H minus First Strand cDNA Synthesis Kit (no. K1623; Fermentas, Burlington, ON). Primers were designed using Primer Express 1.5 software (Applied Biosystems, Foster City, CA). For each primer pair, a standard curve using serial dilutions of ss-cDNA from glomeruli was used to verify equivalent efficiency of amplification. Calculation of the expression ratios was performed using the delta cycle time method previously described (14).
Bovine kidney cortex (0.5 × 0.5 mm) was incubated in 30% sucrose overnight, frozen in optimal cutting temperature medium (OCT; Sakura Finetek, Torrance, CA), and stored at −80°C. Frozen sections (8 μm) were thaw mounted on Superfrost slides, air-dried, and fixed for 10 min in acetone at 4°C. Slides were washed in PBS (3 × 5 min) at room temperature (RT) and blocked for 30 min with 3% BSA in PBS, followed by incubation with the primary antibodies for 1 h at RT. Mouse anti-CD 146 IgG (clone P1H12; MAB16985 Chemicon International, Temecula, CA) was used for detection of melanoma cell adhesion molecule (MCAM) at a dilution of 1:250 in 3% BSA. Podocin was detected with rabbit anti-human podocin IgG (PODO11-A; Alpha Diagnostic International, San Antonio, TX) at a dilution of 1:50 in 3% BSA. Slides were washed (3 × 5 min) in PBS and incubated for 1 h in the dark with 1:500 dilutions of Alexa Fluor 594 goat anti-mouse IgG and Alexa Fluor 488-labeled goat anti-rabbit IgG (Molecular Probes). Slides were washed three times with PBS, mounted with ProLong Gold antifade (Molecular Probes), and viewed on a Zeiss LSM 510 laser scanning confocal microscope. Images were captured using a ×20 objective.
In situ hybridization.
Bovine kidney cortex (0.5 × 0.5 mm) was frozen in OCT and stored at −80°C. Frozen tissue sections (8 μm) were thaw-mounted on Superfrost slides, air-dried, and fixed for 1 h in 4% paraformaldehyde (PFA; pH 7.0), washed (3 × 5 min) with PBS containing 0.1% Tween 20 (PBT), dehydrated through a series of 25, 75, and 100% methanol in PBT, and stored at −20°C. Digoxigenin-labeled riboprobes were generated from plasmids (pCRII-TOPO vector; Invitrogen) containing nucleotides 126–316 of Bos taurus MCAM (GenBank accession no. U89327) and nucleotides 44–445 of the Bos taurus DKFZp564B076 homolog (GenBank accession no. CB451539) with an Ambion Maxiscript in vitro transcription kit. Specimens were rehydrated in methanol through the reverse order of dehydration steps, washed (2 × 5 min) with PBT, and incubated in 10 μg/ml proteinase K for 10 min. After being rinsed with PBT, slides were fixed in 4% PFA for 15 min, washed (2 × 5 min) in PBT, acetylated for 10 min in 0.1 M triethanolamine (pH 8.0) containing 0.25% acetic anhydride, and washed (2 × 5 min) in PBT. After rinsing in a 1:1 mix of PBT-hybridization solution [50% formamide, 1.3× SSC (0.15 M NaCl + 0.015 M sodium citrate, pH 5), 5 mM EDTA (pH 8), 50 μg/ml wheat germ tRNA, 0.2% Tween 20, 0.5% CHAPS, and 100 μg/ml heparin], slides were incubated in hybridization solution at 52°C for 1 h and then covered with 100 μl digoxigenin-labeled probe (5 ng/ml) in hybridization buffer for 12–16 h at 52°C. No-probe control sections were covered with hybridization solution only. Posthybridization washes (30 min each) were performed at 62°C with a 2:1 mix of hybridization solution-posthybridization solution (50% formamide, 1.3× SSC, 1% SDS), with a 1:1 mix of hybridization solution-posthybridization solution, two times with hybridization solution, with a 1:1 mix of posthybridization solution-TNT (0.1 M Tris·HCl, pH 7.5, 0.15 M sodium chloride, and 0.5% Tween 20), and finally two times with TNT at RT. To remove residual ss RNA probe, specimens were treated with RNase A (20 μg/ml) and RNase T1 (100 U/ml) in RNase buffer (0.5 M NaCl, 0.01 M Tris·HCl, and 1 mM EDTA, pH 8.0) for 40 min at 37°C, followed by a 30-min wash in RNase buffer at the same temperature. After being washed (2 × 5 min at RT) with TNT, slides were blocked with TNTB [TNT plus 0.2% Boehringer blocking reagent solution (Roche Diagnostics)] for 1 h (RT), and 15 min at 4°C. Tissue sections were then incubated overnight at 4°C with Anti-Digoxigenin-AP (Roche Diagnostics) diluted 1:1,000 in TNTB. Slides were rinsed two times with TNT and washed (6 × 20 min, RT) in TNT. After equilibration in detection buffer (0.1 M Tris·HCl and 0.1 M NaCl, pH 9.5), samples were incubated in NBT/BCIP solution (1:100 dilution in detection buffer; Roche Diagnostics) for 1–6 h and protected from light. Color development was stopped with 10 mM Tris·HCl and 1 mM EDTA, pH 8.0. Sections were examined using a Zeiss Axioplan 2 microscope, and digital pictures were taken with a ProgRes C14 camera (Jenoptik) using ×40 dry and ×100 oil objectives.
The bovine endothelial cell SAGE databases.
The GEn cell SAGE library contains 36,844 total tags, with 16,699 unique tag sequences. The BAE cell SAGE library contains 26,452 total tags, with 12,122 unique tag sequences. These databases have been deposited at www.ncbi.nlm.nih.gov/geo; the accession number for the GEn database is GSM3036, for the BAE database GSM3037. Among 6,524 unique tag sequences represented at least 2 times in the 2 libraries, 2,094 (32%) were matched electronically either to well-characterized cDNAs (358 tags) or to EST sequences in the bovine SAGEmap_tag_ug-rel database. Identification of the human homolog was achieved for 1,035 of the tags observed at least two times in the two libraries. Details are shown in Fig. 1.
All bovine tags identified, along with the respective tag counts for the GEn and BAE cell libraries, the Bt. (Bos taurus) GenBank accession number, Bt. sequence name, Hs. (Homo sapiens) homolog name, and Hs. Unigene symbol are shown in Supplemental Table A as a partial bovine endothelial cell transcriptome. An excerpt of the bovine endothelial cell SAGE tag list for which gene identification has been obtained is shown in Table 2. As expected, the frequency of the GAPDH SAGE tag as well as tags for nine other “housekeeping” transcripts were similar in the two libraries (Table 3).
Endothelial cell predominant transcripts.
Five human endothelial (SAGE Duke HMVEC: GSM706; SAGE Duke HMVEC + VEGF: GSM707; SAGE Hemangioma 146: GSM1516, freshly isolated colonic endothelium and colon carcinoma endothelium, n = 345,909 tags) and 132 human nonendothelial SAGE libraries (n = 5,813,438 tags) were matched by tag against the SAGE Genie Hs.best_tag database at http://cgap.nci.nih.gov/SAGE. Among 25,543 unique tags observed at least two times in human endothelial cell SAGE libraries, a secure tag-to-gene match was found for 8,698 tags (34%). For each tag, a frequency (absolute tag count/100,000 tags) ratio of endothelial vs. nonendothelial cell libraries and the respective χ2 value were calculated. Using an arbitrary cut off of a five-fold higher tag frequency in endothelial vs. nonendothelial cells and a critical χ2 value of 6.65 (P < 0.01), we catalogued 131 endothelium-predominant transcript tags, shown in Supplemental Table B.
We next sought to establish a method to distinguish endothelial from nonendothelial cell SAGE libraries (Fig. 2, Table 3). We found that tags representing any single, well-established endothelial cell-specific transcript, for example Tie-1 (Fig. 2A), were not exclusive to endothelial cell SAGE libraries. However, when the sum of frequencies for 10 endothelial cell predominant tags, for which both human and bovine SAGE tags are known (Table 3), was calculated, it was significantly higher in the endothelial cell libraries, averaging 96.12 ± 11.22 (n = 5, mean ± SE) compared with 4.45 ± 0.71 (n = 132, mean ± SE) in the nonendothelial cell libraries (Fig. 2B, P < 0.001, t-test, assuming unequal variance), without overlap.
By contrast, the sum of SAGE tag frequencies for 10 ubiquitous tags (Table 3) averaged 1,272 ± 140 and 1,113 ± 37 in the endothelial and nonendothelial cell SAGE libraries, respectively (P = 0.33, Fig. 2C). In the bovine glomerular and aortic endothelial cell SAGE libraries, the sum of the frequencies of the 10 endothelial cell marker tags was 198 and 174, well above that observed in nonendothelial cells. The sum of the ubiquitous tag frequencies (934 and 1,365 in glomerular or aortic endothelial cell libraries, respectively) was not different from the sum of ubiquitous tag frequencies in human nonendothelial cell libraries.
Differential abundance of a subset of SAGE tags.
Among the 6,524 unique bovine SAGE tags observed at least two times in the two libraries, 42 tags were more highly expressed in GEn than in BAE cells (ratio GEn-BAE ≥ 5, P ≤ 0.01, Fig. 1). One differentially expressed tag, GCGCCCCTGC, was observed at a frequency of 325:100,000 in the GEn library and not at all in the BAE cell SAGE library. However, this tag differs from the GCGCCCCTTC tag observed at a frequency of 91:100,000 in the BAE, but not at all in the GEn cell library that differed by only one nucleotide, and both match two distinct bovine myosin heavy chain class 1 cDNA sequences in GenBank, thus representing a single nucleotide polymorphism (SNP) within the tag sequence. For 24 of the remaining differentially expressed tags, in which SNP could not explain differential expression, the corresponding bovine cDNA or EST was identified, and the human homolog was matched to 17 of these (Fig. 1). For a more comprehensive analysis, 13 tags with an absolute count of 8 in the GEn and 0 in the BAE cell SAGE libraries, respectively (χ2 = 6.16), for which the human homolog had been identified, were added to the analysis. Hence, SAGE analysis has identified 30 transcripts that appear to predominate in GEn cells compared with BAE cells in culture.
SAGE predicts differential expression.
For 11 tags from the set of differentially expressed tags (Table 4), semiquantitative PCR was performed using RNA from GEn and BAE cells distinct from those used to establish the SAGE libraries but otherwise cultured identically. Among these RT-PCR verified transcripts, three for which only bovine EST's, but not the human homolog is known, are included. For each of the transcripts whose tag count suggested higher expression levels in GEn cells compared with BAE cells, PCR confirmed higher levels of expression in the GEn cells. Similarly, for the cystatin B tag, whose frequency was much higher in BAE than in GEn cells, RT-PCR showed a higher level of transcript expression in BAE compared with GEn cells (Fig. 3, A–E). As control for equal template input, GAPDH was amplified equally from serial dilutions of GEn and BAE cell cDNA (Fig. 3F). To further confirm the quantitative nature of the bovine endothelial cell SAGE libraries, real-time RT-PCR was performed for the well-characterized endothelial cell transcript eNOS, observed at a frequency ratio of 4:1 in the GEn vs. the BAE SAGE libraries. GAPDH was the internal control. The difference in GEn eNOS to BAE eNOS transcript abundance of 3.54-fold was consistent with the difference observed by SAGE (data not shown).
Predominant expression of the DKFZp564B076 in GEn cells in vivo.
To determine whether any of the transcript tags observed at a level eightfold or higher in GEn compared with BAE cells could also be GEn predominant in vivo, we next examined the frequency of the corresponding human tags in all 208 SAGE libraries available for analysis at SAGE Genie (Table 5), in 9 endothelial cell SAGE libraries, those in the public domain and those previously reported by St. Croix et al. (Ref. 33 and Table 5), and in the one human glomerular SAGE library (GEO accession no. GSM10419; see Ref. 5). Seven tags [echinoderm microtubule-associated protein like 2 (EMAP-2), cathepsin C, hypothetical protein DKFZp564B076, major vault protein, the zinc finger protein ZDHHC14, solute carrier family 29, and adenine phosphoribosyltransferase] were observed at frequencies that were more than 10-fold higher in the GEn SAGE library compared with other human nonglomerular libraries and endothelial libraries. Of the seven, the SAGE tag for the DKFZp564B076 transcript was also observed at a much higher frequency in a SAGE library prepared from microdissected glomeruli (5) compared with its frequency in other tissues or endothelial cells. To determine whether DKFZp564B076 expression localizes to endothelium in glomeruli, CD146 (MCAM) in situ hybridization and immunofluorescence as the endothelial cell marker was compared with the in situ hybridization pattern of DKFZp564B076. MCAM immunofluorescence in bovine glomeruli followed a lobular pattern and, as expected, did not overlap with podocin immunoreactivity (Fig. 4, A–C). In situ hybridization with an MCAM antisense mRNA probe similarly showed a lobular pattern in bovine glomeruli (Fig. 4, D–F). The sense probe was similar to the no-probe control (data not shown). The pattern of DKFZp564B076 antisense mRNA hybridization was essentially the same as that observed for MCAM (Fig. 4, G–I). Real-time RT-PCR was performed using RNA derived from freshly isolated bovine aortic intima and glomeruli to quantify mRNA abundance. Using GAPDH as the internal standard, the abundance of glomerular and aortic intima mRNA of Tie-2, cystatin B, ε-COP, and podocin were compared first. All but podocin were expressed at higher levels in aortic intima than in glomeruli (Tie-2: 3.39 ± 0.80-fold; cystatin B: 11.55 ± 2.74-fold; and ε-COP: 1.45 ± 0.21-fold greater; mean ± SE, n = 4, Fig. 4J). Conversely, podocin was highly expressed in glomeruli and could not be amplified from aortic intima (Fig. 4J). The DKFZp564B076 transcript abundance was consistently (n = 4) higher in glomeruli compared with a very weak expression in aortic intima. We corrected for differential loading of endothelial cell-derived RNA using Tie-2 as the internal standard. DKFZp564B076 transcript abundance was 58.89 ± 4.04-fold higher in glomerular endothelium than in aortic endothelium.
This study used SAGE to determine whether endothelial cells derived from renal glomeruli display a distinct transcriptional profile compared with aortic endothelium. We detected significant transcriptional differences between glomerular and aortic endothelial cells. Moreover, GEn cells in culture expressed several genes at much higher levels than any other cell. These included EMAP-2, cathepsin C, hypothetical protein DKFZp564B076, major vault protein, the zinc finger protein ZDHHC14, solute carrier family 29, and adenine phosphoribosyltransferase. Among these, DKFZp564B076 was also highly expressed in glomeruli in vivo, where it was localized to glomerular endothelium. We postulate that DKFZp564B076 represents a GEn cell predominant transcript. The function of its putative product remains to be elucidated.
SAGE characterizes the transcriptome of cells and tissues and defines differences in transcript frequencies in mRNA pools by extracting transcript-specific 10-bp sequence “tags” (35). The frequency with which a given tag is present in a SAGE library reflects the abundance of its transcript (15). The likelihood of finding differences depends on the size of the SAGE libraries being compared and the frequencies of the differentially represented tags (18). In addition, SAGE allows for unbiased gene discovery because all tags for known and unknown transcripts are collected.
A successful SAGE-based approach for identification of differentially expressed transcripts in the kidney was previously reported. El-Meanawy and colleagues (11) constructed a small SAGE library with 3,868 tags from kidneys of a glomerulosclerosis-prone mouse strain and found that 1,453 of these tags were unique kidney transcripts. Similarly, Virlon et al. (36) found that the tag for uromodulin was highly enriched in a thick ascending limb SAGE library, and the tag for aquaporin-2 was differentially expressed in the collecting duct. Chabardes-Garonne et al. (5) compared SAGE libraries from eight regions of human nephron, including glomerulus, and defined differentially expressed tags. So far, no SAGE libraries have been established for the distinct cell populations of the renal glomerulus.
In regard to endothelial cells, SAGE has been used to compare tumor endothelium in colon carcinoma with the endothelium of surrounding normal tissue, and transcripts specific to the tumor endothelium have been characterized (33). Also, suppression subtractive cloning has revealed 22 transcripts preferentially expressed in high endothelial venule endothelial cells (13). Nevertheless, as stated in a recently published comprehensive review on endothelial heterogeneity, in spite of numerous studies comparing endothelial cells from distinct origins for isolated characteristics, there have been no detailed quantitative comparisons of transcript profiles between quiescent endothelial cells from different anatomic sites (1). For instance, a study comparing endothelial cells from brain, lung, adipose tissue, or aorta demonstrated differences in the growth rate, survival in culture, presence of tight junctions, and the expression of Weibel-Palade bodies (8). Also, Nitta et al. (26) suggested that endothelial cells cultured from bovine glomeruli express factor VIII-related antigen and take up acetylated LDL but do not contain the Weibel-Palade body, unlike endothelial cells derived from large vessels. Such work predicted transcriptional differences in cultured endothelial cells from different locations, but transcriptomes for different endothelial cell subtypes, and in particular for the glomerular endothelium, have not been described so far.
In the current study, the expression frequency of individual endothelial cell marker tags (e.g., Tie-1) did not fully separate endothelial from nonendothelial cell libraries (Fig. 2A), most likely because of the presence of some endothelial cell-derived transcripts in bulk tissues. Hence, finding an endothelial cell tag in a library cannot be taken as evidence that the library is an endothelial cell library. We therefore calculated the sum of the frequencies of 10 endothelial cell marker tags (Table 3) and found that the sum of the frequencies of these tags readily distinguished endothelial from nonendothelial cell libraries. The sum of the frequencies of these endothelial cell marker tags was comparably high in our BAE and GEn libraries (Fig. 2B), verifying their endothelial cell origin. As expected, the sum of SAGE tag frequencies for 10 ubiquitous tags that served as the control (Table 3) was not different in the endothelial and nonendothelial cell SAGE libraries (Fig. 2C). We suggest that calculation of the sum of cell-specific tag frequencies may be a useful tool to describe SAGE libraries derived from homogeneous cell populations.
The most abundant tags in both the GEn and BAE libraries represent transcripts of housekeeping genes and mitochondrial enzymes, as is the case in most other SAGE libraries (Supplemental Table A). The GEn and BAE libraries also contained many tags representing endothelial cell transcripts, including those found in other endothelial cell libraries. The list of endothelial cell predominant tags is shown in Supplemental Table B. These tags are defined by a tag frequency ratio in endothelial vs. nonendothelial cells that is greater than 5:1 and by a χ2 value >6.65 (P < 0.01). PV-1, CD34, platelet/endothelial cell adhesion molecule (CD31), eNOS, P-selectin, MCAM, and von Willebrand factor were among these. There also are transcripts not previously reported to be endothelium predominant, such as the transcription factor mesenchyme homeo box 1, the transcriptional repressor SNAIL homolog 1, the fatty acid-binding protein 4, the paralemmin palmdelphin, and type IV collagens α1 and α2 (Supplemental Table B). The role of these newly described endothelium-predominant transcripts in defining the endothelial cell phenotype will require further study in the future.
In the bovine endothelial cell SAGE libraries, tags for thymosin-β4, calmodulin 1, osteonectin, laminin receptor-1, connective tissue growth factor, and many cytoskeletal proteins are highly expressed (Supplemental Table A). An excerpt of the bovine endothelial cell transcriptome, for which gene identification has been made, is shown in Table 2. It is our hope that the information in Supplemental Tables A and B can serve as a resource in the study of endothelial cells.
The tags observed more frequently in GEn cells compared with BAE cells could potentially represent GEn cell-predominant transcripts. To further narrow the list of potential GEn-predominant transcripts, and to make the findings BAE independent, we compared the frequencies of the most differentially observed GEn SAGE tags with 208 human SAGE libraries and with all available endothelial cell SAGE libraries (9 SAGE libraries). Among all transcripts with an identified bovine and human tag, 7 were expressed at least 10-fold higher in GEn cells compared with all 208 human SAGE libraries and all human endothelial cell SAGE libraries. Given that SAGE tags for these seven transcripts, namely, hypothetical protein DKFZp564B076, EMAP-2, cathepsin C, major vault protein, the zinc finger protein ZDHHC14, solute carrier family 29, and adenine phosphoribosyltransferase, appeared preferentially expressed by cultured GEn cells, we sought evidence that they might also be highly expressed in glomeruli in vivo. To this end, we again utilized comparison with a SAGE library [in this case, that produced for human glomeruli by Chabardes-Garonne et al. (5)]. We reasoned that the SAGE tag for any transcript that is predominant in glomerular endothelium in vivo should be found at a higher frequency in whole glomeruli compared with other tissues. Tags for six of the potential Gen-predominant candidates in Table 5 were either not observed in the glomerular SAGE library or were present at levels similar to that in other human SAGE libraries, suggesting that the high abundance of these transcripts in GEn cells might be restricted to in vitro culture. This would not be surprising, since St. Croix et al. (33) had previously noted large differences in the transcriptional profile of cultured endothelial cells when they were compared with freshly isolated endothelial cells. Nevertheless, the appropriate human SAGE tag representing the putative GEn-predominant transcript DKFZp564B076 was observed at very high copy number in the glomerular SAGE library, and its frequency in glomeruli was much higher than that in other human SAGE libraries. Using in situ hybridization, we found that the pattern of DKFZp564B076 expression in glomeruli is essentially the same as that of the endothelial cell-specific marker MCAM. MCAM, previously shown to be highly endothelial cell specific (33), was chosen because pilot studies showed that the mouse monoclonal P1H12 anti-MCAM antibody was highly cross-reactive with bovine endothelium in vivo and with cultured bovine endothelial cells in vitro. To quantify differences between aortic and glomerular endothelium, real-time RT-PCR was performed using template RNA from freshly isolated bovine glomeruli and bovine aortic intima. With GAPDH serving as the internal standard, the DKFZp564B076 transcript was enriched some 14- to 20-fold in glomeruli, and even more so when the endothelial cell marker Tie-2 was used to correct for nonendothelial cell RNA admixture in the glomerular RNA preparation (Fig. 4). The fact that the DKFZp564B076 SAGE tag was not detected in the BAE cell library reflects the fact that a library of this size (26,452 tags) has limited sensitivity for rare tags. The putative protein product of DKFZp564B076 (Genbank nucleotide accession no. AL049313, BC020923, protein accession no. AAH20923) is predicted to be 69 amino acids in length, without known functional domains. Of interest, the genomic sequence is located on chromosome 6p12.1 just upstream of the chloride intracellular channel 5. Whether the DKFZp564076 transcript in GEn cells produces a protein product or represents regulatory RNA will require further investigation.
To conclude, this study represents the first in-depth transcriptional analysis and comparison of renal GEn cells with endothelial cells from other vascular beds. To our knowledge, we identified the first GEn cell-predominant transcript in vivo.
The postdoctoral fellowship of G. Sengoelge at The Johns Hopkins University was funded by the Austrian Science Fund (Fonds zur Förderung der wissenschaftlichen Forschung) with Grant no. J1462. Research stipends for W. Fierlbeck and J. Sorensson were from the German Research Foundation (Deutsche Forschungsgemeinschaft) and the Council of Natural Sciences, Sweden, respectively. This study was funded by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-50674 and Grants CEG63108 and MOP641814 from the Canadian Institutes of Health Research to B. J. Ballermann, who also holds the Tier I Canadian Research Chair in Endothelial Cell Biology.
We thank Veronica Senchak for skilful technical assistance. The assistance of Mathilde J. Sector, Dr. Afschin Soleiman, and Dr. Gere Sunder-Plassmann in preparation of this manuscript is gratefully acknowledged. Dr. Alian Liu provided invaluable advice for in situ hybridization studies. We are grateful to Custom Packers, Edmonton, for providing bovine tissues for our studies.
Supplemental Tables A and B can be found at http://ajprenal.physiology.org/cgi/content/full/.00076.2004/DC1
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 the American Physiological Society