Alterations in gene expression associated with changes in the state of endothelial differentiation

Differentiation (1995) 58:217-226 Ontopeny, Neoplssia and DifferentiationTherapy

0 Springer-Verlag I995

Alterations in gene expression associated with changes in the state of endothelial differentiation David T. Shima1.2, Kim B. Saunders’, Anne GougosI, Patricia A. D’Amore’Jd I

2

Laboratory for Surgical Research, The Children’s Hospital, Boston, MA 021 15, USA Program in Biological and Biomedical Sciences, Harvard Medical School, Boston, MA 021 15, USA Department of Pathology, Harvard Medical School, Boston, MA 021 15, USA

Accepted in revised form: 19 October 1994

Abstract. The endothelium maintains a developmental plasticity which allows rapid phenotypic change in response to extracellular signals during normal processes, such as corpus luteum formation and wound healing, and in pathologic processes, such as tumor angiogenesis. Endothelial cells (EC) in culture have been very useful for investigating various aspects of endothelial growth and behavior. In spite of documented similarities between EC in vitro and the endothelium in vivo, many characteristics of the vessel endothelium are lost when the cells are placed into culture. We have undertaken to identify differences in gene expression between differentiated vessel endothelium and dedifferentiated EC. We utilized a new technique called differntial display which compares polymerase chain reaction (PCR)-amplified mRNA from two (or more) cell populations. Endothelium scraped directly from freshly obtained aortas, and demonstrated to be free of contaminants, were used as the source of differentiated RNA, whereas proliferating, primary explanted EC grown for five days in the presence of basic fibroblast growth factor (bFGF) provided a pool of ‘dedifferentiated’ RNA. Using differential display, we have observed numerous reproducible differences in gene expression. To confirm that the expression differences visualized by differential display represented actual differences in gene expression, we isolated vessel-specific and culture-specific cDNA tags for additional analysis. Three cDNA tags specific to vessel endothelium were cloned and sequenced, and compared to nucleotide and protein databases. Two of the clones (A1 and 2.5) displayed no significant sequence similarity, whereas a third clone (A2) is nearly identical to a human expressed sequence tag (EST) and has significant sequence similarities to a plant and Xenopus ubiquitin-like protein. Northern and/or in situ hybridization analysis of the A1 and A2 genes confirmed their restricted expression to the vessel endothelium. Correspondence to: P.A. D’ Amore, Laboratory for Surgical Research, Enders 1061. 300 Longwood Avenue, Boston, MA 02 1 15. USA

The expression of Al by the endothelium in vivo is not simply a function of growth state, as cultured cells did not express A l even when grown to postconfluence. One other cDNA fragment, selected as a culture-induced gene, was identified by sequence analysis as the bovine homologue of laminin B I , and Northern analysis confirmed that expression was induced upon culturing of EC. Use of differential display to study endothelial gene expression will allow us to investigate the molecular mechanisms that underlie initiation and maintenance of endothelial differentiation. .---

Introduction

The development of methods for the isolation and culture of vascular endothelial cells (EC) has made it possible to study various aspects of endothelial growth and metabolism. EC in culture retain a subset of their in vivo characteristics, most notably a contact-inhibited growth pattern, the synthesis of von Willebrand’s factor, the ability to take up modified low-density lipoprotein [29], and the expression of PECAM (CD31) [3]. In spite of these similarities, evidence indicates that many of the properties that characterize the differentiated vessel endothelium are lost upon culturing and subsequent ‘dedifferentiation’ of EC in vitro. For example, although EC in culture are widely known to be contact-inhibited, their labeling index is significantly higher [ I31 than that of the normal endothelium in vivo [14, 251; 3-5% in vitro vs. O.OI-O.l% in vivo. Differences in structural specializations are also noted between EC in culture and those in vivo. The endothelium of a number of microvascular beds is discontinuous or fenestrated. However, once cultured these EC lose their fenestrations and become continuous. Similarly, EC of the blood brain barrier form extensive tight junctions and have a number of specialized surface receptors as the basis for the barrier function. Once placed into culture, many of these features are lost.

SCAT 1

W

W

-

-,-.

-. --. .c..3.

01 Y

m

4

W

u

DI 2

scAT1

I

DI 2

Fig. 1. a Clumps of endothelial cells (EC) scraped from bovine aortas, cytospun onto slides, fixed and visualized by light microscopy. b EC scraped from bovine aortas and cultured in Dulbeccos modified Eagle’s medium with 10% calf serum (DMEMIIOCS) with 2 ng/ml basic fibroblast growth factor (bFGF) until confluent. c Cells cultured as in b incubated overnight in 2 pg/ml dilacetylated-LDL (dil-acLDL) and visualized by epifluorescence. d Fluorescence-activated cell sorting (FACS) analysis of diI-acetyla-

ted-LDL uptake by cultured aortic scrapes. Primary EC cultured as in b were incubated for I2 h with dil-acLDL (2 pg/ml), rinsed, detached by trypsinization and examined by FACS. Results are from three separate isolations, and include a culture of smooth muscle cells as a negative control. Each condition is screened for light scatter (/@-hand graph of each panel), and log fluorescence intensity as a function of cell number (right-hmd graph). Magnification of a-c is 500x

Cultured EC have been reported to have elevated expression of many genes relative to endothelium in vivo. An examination of mRNA isolated from EC of human umbilical vein, as well as from bovine aorta, reveal that the in vivo levels of platelet-derived growth factor (PDGF) B chain mRNA are extremely low [6]. In contrast, PDGF mRNA levels in cultured cells are chronically elevated [6, 271. Growth of cultured EC to confluence, a condition assumed by many researchers to mimic the in vivo state, does not lead to a reduction in PDGF expression. However, c-sis mRNA levels are reduced in EC induced to form three-dimensional capillary-like tubes [ 161. EC in culture have also been shown to express high steady state levels of mRNA for a variety of genes including basic fibroblast growth factor (bFGF) and thrombospondiri, whereas EC in vivo have low to negligible levels [ 191. Taken together, these data indicate dramatic differences in the profile of gene expression between endothelium in vivo and cultured EC. Moreover, it is likely that a subset of these and other as yet uncharac-

terized alterations in gene expression contribute to the loss of the differentiated EC phenotype associated with tissue culture. One approach to understanding the factors required for the initiation and maintenance of the differentiated phenotype of vessel endothelium is to identify genes expressed by the endothelium in vivo, but not by cultured EC. Such genes/proteins can be examined for their potential role in the regulation of the EC differentiated state, or could be used as markers to gain insight into the microenvironmental determinants which control their expression. Until recently, the methods available to distinguish genes expressed differentially by a particular cell population included subtractive hybridization and differential hybridization. Though remarkably successful in some cases (e.g. myo D, [8]), these techniques, for the most part, lack the sensitivity necessary to identify low abundance mRNAs. Recently, a novel method, termed differential display, was described to identify and clone cDNAs for differentially-expressed mRNAs. Beginning

219

with two pools of total RNA from similar but non-identical cell populations, subsets of mRNAs are amplified by reverse transcription and polymerase chain reaction (PCR). The 3' PCR primer, designed to utilize the polyadenylated tail on most mRNAs, is anchored at the 3' end of the single-stranded cDNA. Two additional nucleotides are added such that one primer recognizes one-twelfth of the total cDNA pool. The second upstream primer is arbitrary in sequence and 10 basepairs in length. Differences in the two cell populations are detected in a side-by-side comparison of the amplified subsets of cDNAs that are separated and visualized on a denaturing, sequencingsize polyacrylamide gel. We have applied this technology to compare gene expression in quiescent vessel endothelium to that of proliferating EC with the goal of identifying genes that are markers of and/or involved in the regulation of endothelial growth and differentiation.

R-actin VSM-act in

EtBr

Methods lsoltrtiori crnd characterixtion o j pure endothelium from blood ve.s.sr1.s. Bovine calf aortas were used as a source of differentiated endothelium. Aortas were obtained fresh from the slaughterhouse (Research 87, Boston, Mass., USA), and immediately processed for isolation of total RNA from endothelium. Following exhaustive rinsing, the lumens of the aortas were gently scraped with a sterile scalpel blade, taking care not to penetrate the internal elastic lamina. The scraped endothelium was immediately placed in guanidine isothiocvanate buffer and total RNA isolated using a CsCl step gradient. Alternatively, primary EC removed by scraping were plated in Dulbecco's modified Eagle's medium (DMEM) with 10% calf serum (JHR Biosciences, Lexana, KS) (DMEM/IOCS) in the presence of bFGF (2 ng/ml) for 5 days before total RNA was isolated. (Cell culture reagents were obtained from GIBCO Laboratories, Grand Island, N.Y.). To determine if the EC obtained from the scrapes were free from contaminants, isolates were grown in culture for 7 days to allow for the proliferation of possible smooth muscle cells (SMC), and then assessed for purity by morphology, von Willebrand's factor immunostaining (not shown), and uptake of dil-acetylated-LDL (Biomedical Technologies, Stoughton, Mass., USA). As additional evidence of the purity of the endothelium obtained by scraping, total RNA processed from aortic scrapes, primary cultures of EC, and ECsmooth muscle cell co-cultures (as a positive control) were analyzed by Northern blot for alpha-smooth muscle actin (SM-actin) expression.

Fig. 2. Total RNA (10 pg) from smooth muscle cell (SMC)-EC co-culture, vessel endothelium, and primary EC were analyzed by Northern blot using a probe. which recognizes 0-actin and SM-actin. Ethidium b r m i d e staining of the denaturing gel is shown bdow to demonstrate approximately equal loading of RNA in each lane

I

Di~i~rrritial displuj and cloning methods. Total RNA from directly-scraped endothelium and proliferating primary EC was processed for differential display as follows. Total RNA (50 pg) was treated with DNase in the presence of placental RNase inhibitor to remove contaminating genomic DNA. After phenollchloroform extraction and ethanol precipitation, samples were redissolved at a concentration of 200 ng/pl in sterile dH20. RNA (200 ng) from each sample was reverse transcribed with 300 U MMLV reverse transcriptase in the presence of 2.5 p M dT12XlX2primer (XI can be either dA, dG or dC; X2 can be any base e.g. S ' w C A 3 ' ) , and 20 pM dNTPs. After reverse transcription, the single-strand cDNA (designated the RT mix) was aliquoted into ten different PCR reactions, each containing a unique arbitrary 10-mer primer. Approximately 2 pl of RT mix was added to 18 pl of PCR mix including: 2.5 p M T12XlX2primer (same as in RT reaction), 0.5 p M arbitrary 10-mer (Operon Technologies, Palo Alto, Calf., USA), 4 p M dNTPs with 0.5 pM 35SdATP(1200 Ci/mmole) (New England Nuclear, Boston, Mass., USA) and I U Taq DNA polymerase (Perkin Elmer, Boston, Mass., USA). Cy-

cling parameters (40 cycles) were 94" C for-30 s, 42" C for I min, 72" C for 30 s and a final 5 min extension step at 72" C. Products were separated on a denaturing 6% polyacrylamide sequencinglength gel and visualized by autoradiography. The majority of amplified fragments was shared between the two gene pools being compared. Fragments unique to the gene pool of interest were isolated by excising the band from the dried gel, rehydrating gel slices in boiling dH20, crushing the gel, and vigorously vortexing. The fragments were reamplified using the PCR conditions described above, omitting radiolabel. PCR products were cloned, using a T/A Cloning kit (Invitrogen), into the EcoRl site of the pCRlOOO plasmid. Dideoxy sequencing was performed [23j, and probes for Northern blot analysis were generated either by asymmetric PCR labeling to generate full length single-strand (antisense) cDNA probes [ 28 j, or by random hexamer labeling of gel-purified insert cDNA [ 101. Unless otherwise noted, molecular biology grade chemical rcagents were purchased from Sigma Chemical Co. (St. Louis, MO., USA) and modifying enzymes were obtained from BRL (Gaithersburg, MD., USA). Northern blot nncrlvsis. Total RNA was separated in formaldehyde-agarose gels and transferred to Hyhond nylon (Amersham, Arlington Heights, Ill., USA) by capillary blotting. The nylon was baked 1 h or UV cross-linked (Stratagene), prehybridized, hybridized and washed as has been described 1201. Loading of RNA was assessed by stripping blots and hybridizing with a cDNA probe for 28s rRNA, or by visualization of ethidium bromide stained gels. Many standard genes used to normalize for RNA loading (glyceraldehyde-3-phosphatedehydrogenase or p-actin) show dramatic differences in expression between directly-scraped endothelium and primary EC, and cannot be used for normalization. I n situ hybridization. Cytospins were prepared using EC scraped from vessels as described above, and fixed with 4% formaldehyde for 10 min. Prior to hybridization the cells were subjected to pro-

220

vc vc

a

teinase K (2 pg/ml) (Boehringer Manheim, Indianapolis, IN., USA) digestion at 37" C for 10 min. After rinsing in PBS, the cytospins were incubated at RT with 0.2M HCI for 10 min, 0.1% Triton XIOO/PBS for 5 min and 10 mM dithiothreitol (DTT) at 45" C for 10 rnin followed by a further 30 rnin in the presence of 10 m M iodoacetamide. Amino groups were acetylated in 0.25% acetic anhydride/triethanolamine-HCI and the cells were then washed once in 2xSSC and dehydrated. Antisense and sense RNA probes were synthesized by in vitro transcription (Promega, Madison, WI., USA) with incorporation of 3sS-UTP (New England Nuclear, Boston, MA., USA) using DNA templates described above. Approximately 8x105 cpm of denatured riboprobe per 20 pl hybridization solution were added to slides and covered by a 22x22 mm glass coverslip. Hybridization was carried out for 18 h at 53" C in the presence of I00 pg/ml tRNA and 300 mM DTT in chambers humidified by 4xSSC/50% formamide. Cells were washed several times for 5 rnin in 2xSSC/50% formamide, briefly exposed to 0.05% Triton X100/2xSSC and rinsed in 2xSSC. all washes in the presence of 10 mM DTT. The cells were incubated in 26 pg/ml of RNAse A (United States Biochemical. Cleveland, OH., USA)12xSSC/5 mM DTT for I h at 37°C and subsequently in 0.05% sodium pyrophosphate/2xSSC/I mM DTT for 5 h at 37" C. A higher stringency wash was carried out in OSxSSC for 15 rnin at 50" C. Slides were

b

Fig. 3. a A representative autoradiograph of a differential display of endothelium from vessels versus primary culture EC. Lrrries A-J represent comparisons of the two RNA pools amplified with different prima combinations. The negative control for genomic DNA contaminants was performed under identical conditions as lane F. except for the omission of reverse transcriptase. b Comparison of endothelium isolated from vessels (v) versus primary culture ( c ) EC. Dashes denote examples of banding pattern differences that would be exploited by molecular cloning

coated with Kodak NTB- I autoradiographic emulsion, exposed for 3 weeks and visualized using a Zeiss Axiophot microscope fitted with a darkfield condenser.

Results Isolation and churacterization of pure endothelium from blood vessels

EC scraped from bovine aortas were affixed to slides by cytospinning (Fig. la). EC similarly scraped from aortas were placed into culture in the presence of bFGF (2 ng/ml) until they reached confluence (Fig. 1 b) and were then incubated overnight in dil-acetylated low density lipoprotein (diI-acLDL) (Fig. 1 c). Fluorescence-activated cell sorting (FACS) analysis of cultures incubated in diIacLDL revealed that cells isolated from the aorta were >99% acLDL positive, indicating negligible contamination with SMC (Fig. Id). A probe which recognizes vari-

22 I

ous actin isoform genes hybridized to two mRNA species, corresponding to P/g-actin (-2.1 kb) and SM actin (-1.8 kb), in the EC-SMC co-culture (positive control). Endothelium scraped from aorta, and primary cultures of EC from scrape had undetectable levels of SM-actin mRNA (Fig. 2). Interestingly, endothelium scraped from aortas expressed low steady-state levels of p-actin mRNA when compared with primary cultured cells, suggesting that the quiescent vessel endothelium has a lower demand for p-actin synthesis than proliferating, migrating cultured EC. I~ij/ierentialdispluy of vessel endothelium \'er.su.sprirnury cultured EC

After determining that vessel and cultured cell isolates of EC were homogeneous, total RNA was used for differential display of their gene expression. Figure 3a shows an example of a sequencing gel used to visualize the cDNAs amplified by the differential display protocol. Each pair of lanes (denoted by letters A-J) results from amplification of RNA from endothelium scraped from an aorta ( I st lane) and RNA from EC in primary cultures (2nd lane) with a unique combination of 5' 10mer and 3' poly-dT primers. The individual bands represent tags for different mRNAs, allowing for a sideby-side comparison of the RNA pools from these two cell populations (Fig. 3b). The negative control lane represents differential display of vessels and cultured EC, with the omission of reverse transcriptase from the cDNA synthesis step. This is an important control as without the production of cDNA, only RNA and contaminating genomic DNA are present. Because Taq polymerase only efficiently amplifies DNA, bands that appear in this control group would be derived from genomic DNA contaminants in the RNA isolates. Identijcution and conjirmution of differential gene expression

The goal of these studies was to establish the methodology and feasibility for rapidly identifying, cloning and sequencing differentially-expressed genes. Using only ten different primer combinations, theoretically representing about 2-5% of the total mRNAs in a cell [ 181, we observed numerous reproducible differences in differential display fragments between endothelium in vivo and EC in culture. To demonstrate that these differences represented true differences in gene expression, we focused on four cDNA fragments: three unique to endothelium scraped directly from the vessel and one fragment unique to cultured EC. These fragments were cut from the dried gel, eluted and amplified as described in Methods. Three cDNA tags were selected on the basis of their unique expression in vessel endothelium. Two of these A l and A2 (Fig. 4, top panel) were used to synthesize probes for Northern blot analysis of vessel endothelium RNA and cultured primary EC RNA (Fig. 4, bottom pan-

1

2

-

Fig. 4. (Upper panel) Portion of a differential display gel indicating the cDNA tags for A l (lane I ) and A2 (lane 2) and (Lower panel) corresponding Northern blot analysis confirming that the expression of A I and A2 is restricted to vessel endothelium. First lane in both panels is vessel endothelium and the sccond lane is EC proliferating in culture

1 2 3 4

-28s-

A1

-28s-

5.2

28 S

Fig. 5. Comparison of mRNA levels for the A l (vessel-specific) clone and the culture-induced clone 5.2 (laminin BI). Total RNA (10 pg) was isolated from sparse EC, post-confluent EC, freshly scraped bovine aortic EC, or bovine lung. The same blot was used for A I , 5.2, and 2 8 s rRNA probe (bottom panel) analysis

el). The A1 probe hybridized to an approximately 4.8 kb mRNA that was found exclusively in vessel endothelium. The A2 probe hybridized to an approximately 2.5 kb mRNA that was also present in the vessel endothelium and not in growing primary EC. Thus far, we have not been able to detect an mRNA on a Northern blot corresponding to the 2.5 vessel-specific clone. To determine if the expression of the A l mRNA was regulated by changes in the EC growth state, total RNA

222

Fig. 6. In situ hybridization of vessel endothelium with the vessel-specific probe A I . Vessel endothelium was scraped from aortas, colonies were fixed to slides using a cytospin apparatus (see Fig. I a) and processed as in Methods. Cytospins hybridized with (a) A l riboprobe and (b) sense control

was obtained from sparse and confluent EC cultures. The A I probe hybridized to a 4.8kb mRNA present only in endothelium directly scraped from fresh bovine aortas; no signal was observed in RNA isolated from either culture condition (Fig. 5). These data support the hypothesis that local signals, which are unique to the vessel microenvironment and not reproduced under traditional cell culture conditions, control the expression of genes required to maintain the differentiated phenotype of vessel endothelium. In addition, A l mRNA could not be detected in total RNA from bovine lung. Since the lung is a capillary-rich tissue, it may be that Al expression differs between large and small vessel endothelium. To insure that A l was expressed by the scraped vessel endothelium and not by an extremely low abundance, high expressing contaminating cell type, the A l cDNA tag was

used for in situ hybridization of endothelium obtained by scraping the vessel. Antisense RNA probes corresponding to A l were localized to clusters of scraped EC, with minimal labeling of non-cellular regions in each field (Fig. 6a). Sense control probe hybridized with a low, uniform, background pattern over cellular and non-cellular regions of each field (Fig. 6b). Culturing endothelium not only leads to loss of genes, but activates a new repertoire of genes whose encoded proteins may help cells to adapt to the culture microenvironment. The clone designated 5.2 was identified by differential display as a culture-induced gene. Northern analysis confirmed that expression of this gene was significantly elevated in cultured EC relative to their in vivo counterparts (Fig. 5).

223

a

A 5261 A T G G T A T M G A A G G T A G T T T M T ~ C ~ C T G A A G A 5 3G1 0 wuae IIIIII I IIIII IIIIII 1 G A T T G C C M G C C G A A G A G 2 2 bovine/DI

............................

IIIIIIII IIIII IIIIII 5201 ATGAAAAGTATAAAAAAGTAGTTTMTTGCWCTGAAGAG

5 2 5 0 human

5 3 1 1 T C A G U G A T G C C A ~ G T G A G C T ~ A C 5360 ~ T ~ ~ C IIIIIIIIIIIIII IIIIIIII ll IIIIIIIIIIIIIIIIIIIIII 23 TCAGCAGATGCCAGAA-GUWTGCTA-TGAA~C 72 IIIII IIIIIIIIIIIIIIIII IIIIIIIIIIIIIIIIIIIIIIIIII 5 2 5 1 T C A G C T ~ T G C C A G G C C ~ T G C T A - T ~ ~ 5300 C

5 3 6 1 A C T C T T G G C T C M G C T M C A G C M G C T C C A G C T G T T ~ G A C T T A W 5410 I l l IIIIIIIIIII IIIIIIII II II II I1 IIII IIIIIII 1 3 A C T T T T G G C T C M G C A A A C A G C T G C M C T C C T ~ ~ T T T1 2A2~ I I I I I IIIIIIIIIII I I I I I IIIIIIII l l IIIIIIIIIIIII 5 3 0 1 TCTTTTAGCTCMGCAAATAGCAAGCTGCAACTGCTCAAAGATTTAW 5350 5 4 1 1 G A A A A T A T G A G M C M T C ~ T A C T T A G M G AAAAGC T .......... 9450 IIIIIIIIIII IIIIIIIIII I 1 2 3 GAAAATATGAGGACMTCAAAATATTT.. 149 IIIIIIIIII IIIIIIIIII I 5 3 5 1 GAAAATATGMGACMTCAAAGATACTTAGMGATAAAGCTCAAGAATTA 5 4 0 0

.....................

A1 sequence I 10 I 20 I 30 1 CTTGATTGCC TAGAGATCAG AGACAGTCAG 61 AACAAMTAT TTTGTTTTCA TTTCTGCCGA 1 2 1 CAAGGAGGGA ATTTAGAAAG CAAAGCAACT 1 0 1 TTCTCACAGC I 10 I 20 I 30

I 40 I 50 I 60 AATAGCCTGG CTGCCTGTCC TCCTGCACTT GGGAAAGTTA AGCTACAAAG GTCCCCATTC GCGGTTTTTC CCCGAGGGTC CCTAAGATAA I

40

I

50

I

60

I

40

I

50

I

60

A2 sequence I 10 I 20 I 30 1 CACAAGTAGT MTTTATATT CTCTGAATTT 6 1 TGATTTTTGC TGCAGCTTCT GCTTTGTAAT 1 2 1 GGTGAAGTCC ACAAAACAAA TTTCCACATC

I

I

10

20

I

30

TTTCAGCCAC AACMCTGGA TTCTCTTTTC CATATGGACA GTTGTGCTTG TCAGAGTAAC GGCAATCAAG I 40 I 50 I 60

2 . 5 sequence

I 10 I 20 I 30 I. CATAACCAAA GACTTTMGA AGAAGAAGTA 6 1 ACAAAACCTT TGAAAATGAG GAAGTAGATA 1 2 1 ATAAGMCTA MATATTTAA TTTGGCAATC I 10 I 20 I 30

I 40 I 50 I 60 ATTTATGAAT AATAATATGT CATAGAATGA AAGAGGAAAT ACAAGGAGAG TTATTATAAA AAGAAGCCGA A I

40

I

50

I

60

b 0

OD R2

TGRTTCCCGATGTCGRAATTTGTTTTGTGGRCTTCACCCGTTRCTCTGRCRRGCRCRRCTG

humEST

TGCCTGCCGATGTGCAAATTTGTTTTGTGGRCTTCRCCCTTRCTCTGRCRRGCRCRRCTG

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

II 5 5 5 2 TAGACGGGCTCGGTGACCAAGGTAAACCACACGCGCAAACCGACGCAGTC 5 6 0 1 I I I I1 I I I I I I 1 6 2 . . . . .GACCAAACGAAAATCATTACCTTTTAAAAACTGACTTAATAATGA 1 1 0 II Ill IIII I I l l IIIIIIIIIIIIIII IIIII 5 5 0 1 GAGGTGAACAAGGTAAAACAACTACATTTTAAAAACTGACT TAATGC 5547

TCCRTATCATTRCARAGCAGRRGCTGCRGCRRRRRTCRGRRRtlGRGRRTCCRGTTGTTGT

Ill IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

...

TCCGTATCATTACARAGCAGRRGCTGCRGCRRRRRTCRGRRRRGRGRtlTCCRGTTGTTGT

5 6 0 2 ATCTACAAATMCCCATCATCTATTTAATGTTTTTAACCACCTACTTTTG 5 6 5 1 I IIIIII I I IIIIIIIIIIIIIIIII IIII IIIII 1 1 1 TCTTCAAAATAAAACTCMCCTATTTAATGTTTTTAATCACCACATTTTG 6 8 IIIIIIIIIIIIIII IIIIIIIIIIIIIIIIIIIIII IIIIIII 5 5 4 0 TCTTCAAAATAAMCATCACCTATTTMTGTTTTTAATCA CATTTTG 5594

...

. . . . . . . . . . . . 5609

5 6 5 2 TATGCAGTTAAATAAAAGACATTGGTTTTGTATAAACA IIIIIIIIIIIIIIII Ill I IIII I I I 6 1 TATGGAGTTAMTAAAGTACAAAGCTTTTATAAGCCGAATTCTGCAGATA Ill IIIIIIIIIIIII I 5 5 9 5 TAT.GAGTTAAATAAAGCCC

..............................

11

TCCATCATGCTCGCGGC .................................

C P = 9.6

x

18-18

10. 5613

DD A2

151

PUPR3

92

CRCGNLFCGLHRYSDKHNCPYDYKRERRtlKIRKENPUUUREKIQRI 17 CRCWLFC HRY+D+H+C YDYK I +ENPUU A K I ++ CRCGDLFCREHRYTDRHDCSYDYKTUGRERIRRENPUUKRRKIUKU 137

1

Fig. 7. Nucleotide sequence comparison of clone 5.2 with mouse and human laminin BI, (A) corresponding to laminin B1 open reading frame (ORF) and (B) laminin B 1 3’ untranslated regions. Comparisons were performed using the BLAST algorithm [4]

Sequence analysis. The vessel- and culture-specific cDNA tags were sequenced and these were compared with Genbank database sequences. All clones contained flanking sequence corresponding to the particular 10mer primer used for PCR amplification. The initial sequence obtained from clone 5.2 was nearly identical to nucleotide sequence in the coding region of the human and mouse laminin B1 gene (Fig. 7A). Further sequence analysis (obtained by sequencing the opposite end of the clone) led to alignments of another region of the 5.2 clone with similar stretches in the 3’ untranslated region of the mouse and human laminin B1 gene (Fig. 7B). These data led us to conclude that we had cloned a fragment (-450 bp total) of the bovine laminin B1 gene. This conclusion is supported by the results of the Northern blot analysis, in which the probe prepared from clone 5.2 hybridized to an mRNA of approximately 6.5 kb (Fig. 5 ) , which corresponds to the reported size of mouse laminin B1 mRNA [24]. Laminin, an extracellular matrix molecule which has been shown to promote EC proliferation and migration, is enriched at sites of

P

-

1.5

x

18-15

DO R2

157

xsnUBQ

617

DCRCGNLFCGLHRYSDKHNCPYD~KRERRtlKIRKENPUUUtlEKI~RI 17 +CRCtN FC HRY++ H+C YDYK + + NPU++R K + + I ECRCGNNFCARHRYAETHDCTYDYKTHGRRLLTERNPUIIRPKLPKI 69

Fig. 8. a Nucleotide sequence of endothelium-specific clones. Sequencing was performed by the dideoxy’ method [23]. b Comparison of A2 nucleotide sequence with the human expressed sequence tag (hum EST00848). (c) Comparison of A2 ORF with plant PVPR3 and Xenopus ubiquitin-like protein. Comparisons were performed using the BLAST algortihm

capillary sprouting during corneal neovascularization [l 1 1 and is necessary for the formation of EC sprouts in 3-dimensional Matrigel cultures [ 121. These data are consistent with our identification of laminin B1 as a gene associated with the proliferative, invasive, and dedifferentiated phenotype of cultured EC. The sequence of A1 (Fig. 8A) as well as 2.5 and A2, two other vessel-specific genes (Table 1 ), were compared with the nucleotide and protein databases available through the Genetics Computer Group (GCG) sequence analysis program [4]. Clones A 1 and 2.5 yielded no continuous stretch of sequence similarity. Differential display usually results in amplification of fragments proximal to polyadenylation sites. Therefore, most sequence information obtained is in the generally non-conserved

224

Table 1. Differential display clones Name

Size (bp)

mRNA (kb)

Sequence similarities

Culture-specific: 5.2

450

6.5

human & mouse laminin B 1190% nt

Ve.s.se1-specific: Al A2

191 160

4.8

none

2.5

99% identical to human EST;

2.5

161

ND

Both are 80% similar to plant pathogen response gene and gene for Xenopus ubiquitin-like protein in only ORF none

ND, none detected; EST, expressed sequence tag; ORE open reading frame

3’ untranslated region, making predictions of sequence similarity and function difficult. Additional information will be obtained by cloning full length cDNAs by rapid amplification of cDNA ends (RACE) and/or by screening a cDNA library produced from directly-scraped vessel endothelium mRNA. The clone designated A2 (Fig. 8B) was nearly identical to a human cDNA expressed sequence tag (EST00848) that was isolated from fetal hippocampal mRNA [I]. The A2 clone encodes a single open reading frame (ORF), and this region has highly significant similarity ( P value of 9.6~10-’8)to a plant pathogen-response gene (PVPR3) [26] and to Xenopus ubiquitinlike protein (Fig. 8C). Moreover, the proposed ORF for A2 terminates at the same position as the aligned PVPR3 gene, further supporting the conclusion that these two genes are related. Little is known regarding the function of the proteins encoded by the pathogen response or ubiquitin-like genes.

Discussion The environment of EC in culture mimics that of a wound with the loss of contact inhibition requiring migration, proliferation and many other repair and remodeling processes. The in vivo correlate of this behavior is at least partially reflected in the invasive phenotype of EC during developmental and pathological neovascularization. Whereas the endothelium in vivo usually re-establishes its differentiated phenotype following injury, EC in culture do not. Due to the limitations of culture in reproducing the cues available to EC in vivo, the phenotype of EC in culture is not that of the endothelium, a tissue, but is instead simply a monolayer of cells. The signals that lead to the changes in EC gene expression upon tissue culture are unknown. The microenvironment experienced by the endothelium in vivo is complex and dramatically different from that experienced

by the cells in culture. Whereas cells in vivo are bathed in plasma and exposed to circulating cellular elements, cultured cells are exposed to serum, which is known to contain a variety of growth factors, cytokines and other effectors. Similarly, EC in vivo sit on a deformable basement membrane while the EC in culture grow on a rigid substrate of chemically modified plastic. Although cultured EC make an extracellular matrix, its composition differs from that synthesized by the cells in vivo [22]. Cells in culture are exposed to oxygen concentrations that vary, depending on their location in the vasculature (from 15% in the aorta to 3-5% at the level of the capillary). On the other hand, cultured cells are routinely grown at 21% oxygen. Finally, whereas cells in culture are in a static environment, EC in vivo experience a variety of mechanical forces including shear stress, a function of vessel diameter, and transmural stretch, resulting from the local blood pressure. In light of the numerous and significant differences between the milieu of cells in vivo and in vitro, it is perhaps not surprising that cultured EC display a phenotype, which though ‘endothelial’ (as evidenced by their growth pattern and staining for von Willebrand’s factor), differs significantly from the endothelium in vivo. We . postulate that the endothelium, once established in culture, l a c h the appropriate cues, and reverts to a more dedifferentiated state. This reversion of quiescent, differentiated cells to a proliferating, dedifferentiated state in tissue culture presumably recapitulates events that occur in vivo when EC proliferate during development, wound healing and in pathologic angiogenesis. As a corollary of this hypothesis, we postulate that reinstatement of the appropriate environmental signals would lead to the reinduction of the differentiated state. There are a number of potential explanations for the differential expression of genes between EC in situ and in culture. First, the genes differentially expressed by the endothelium may be entirely related to cell quiescence. However, if this were the case, one would expect that culturing EC under conditions leading to contact-inhibited growth would lead to increased gene expression. Though it is very likely that the expression of some genes will correlate with or regulate quiescence, our preliminary data suggest that this will not be the case for all differentially expressed genes. Our studies examining the expression of the Al gene indicate high levels of expression in endothelium scraped from aortas, but virtually none in quiescent, confluent cultured EC. A second explanation for the differential expression of genes between endothelium scraped from blood vessels and EC in culture, is that such genes are either markers of the differentiated endothelium (i.e. are ‘downstream’) and are involved in controlling homeostasis in the endothelium, or are involved in the regulation and/or maintenance of EC differentiation. The absence of these genes in quiescent, cultured cells would support our contention that true differentiation does not occur in EC in vitro under our current culture conditions. However, restoration of environmental signals has been demonstrated to restore some of the dif-

225

ferentiated phenotype. Although adrenal cortical EC lose their fenestrations in vitro, culturing the cells on extracellular matrix produced by kidney epithelial cells leads to the formation of fenestrations [21]. In addition, there is a large body of literature which suggests that tight junctions of blood brain barrier EC that are lost upon culture may be re-induced both in vivo [ 151 and in vitro [S] by the presence of astrocytes or astrocyte-conditioned medium. Similarly, coculture of EC with astrocytes leads to the expression of gamma glutamyl transpeptidase, a brain capillary-specific enzyme [9]. A precedent for this process of reversible differentiation is provided in the well-characterized system of the mammary epithelium. During lactation, mammary epithelial cells in vivo form highly specialized ductules which express a number of well-studied proteins, including casein and whey acidic proteins. When placed in culture the mammary epithelium dedifferentiates to form a flattened epithelial monolayer. Manipulation of culture conditions, including the replacement of the rigid plastic substrate with a malleable collagen substrate, leads to the partial differentiation of these cells, as indicated by the expression of casein [ 171. Additional alterations of the epithelial growth environment, permitting the cells to form tube-like three dimensional structures, leads to the further expression of the differentiated phenotype as evidenced by the synthesis of whey acidic protein [7]. Findings in this system provide some insight into the cues that may be important in regulating endothelial differentiation into tissue endothelium. In addition to providing precedent for an important role of extracellular signals such as matrix [2], the observations in the mammary epithelium also suggest that the process of differentiation is represented by a continuum of alterations in gene expression, rather than an all-or-none switch. If regulation of endothelial differentiation parallels that of mammary epithelium, we might expect to be able dissect, using in vitro systems, the relative contribution of various environmental signals such as the extracellular matrix, soluble factors and heterotypic cell contact. Elucidation of the variables that govern EC growth and differentiation will not only permit us a greater understanding of the control of these processes, but will also assist us in the design of culture conditions that better re-create the in vivo microenvironment. Acknowled~ements.The authors gratefully acknowledge Dr. Peng

Liang (Dana Farber Cancer Institute) for generously sharing his expertise on differential display and its modifications prior to publication. This work was supported by NIH grants EY05318 and CA45548 to PD’A.

References I . Adams MD, Dubnick M, Kerlavage AR, Moreno RF, Kelley JM, Utterback TR, Nagle JW, Fields CA, Venter JC (1992) Sequence identification of 2375 human brain genes. Science 3551632-634 2. Aggeler J, Ward J, Blackie LM, Barcellos-Hoff MH, Streuli CH, Bissell MJ (1991) Cytodifferentiation of mouse mammary epithelial cells cultured on a reconstituted basement mem-

brane reveals striking similarities to development in vivo. J Cell Science 99:4074 17 3. Albelda SM, Muller WA, Buck CA, Newman PJ (1991) Molecular and cellular properties of PECAM- I (endoCAM/CD3 I ) : A novel vascular cell-cell adhesion molecule. J Cell Biol 114:1059-1068 4. Altschul SF, Glsh W, Miller W, Myers EW, Lipman DJ (1990) Basic local alignment search tool. J Mol Biol 215:403410 5. Arthur FE, Shivers RR, Bowman PD (1987) Astrocyte-mediated induction of tight junctions in brain capillary endotheliurn: an efficient in vitro model. Dev Brain Res 36: 155-159 6. Barrett TB, Gajdusek CM, Schwartz SM, McDougall JK, Benditt EP (1984) Expression of the sis gene by endothelial cells in culture and in vivo. Proc Natl Acad Sci USA 8 I :67724774 7. Chen LH, Bissell MJ (1989) A novel mechanism for whey acidic protein gene expression. Cell Regul I :45-54 8. Davis RL, Weintraub H, Lassar AB (1987) Expression of a single transfected cDNA converts fibroblasts to myoblasts. Cell 51:987-1000 9. DeBault LE, Cancilla PA (1980) GGTP in isolated brain endothelial cells: induction by glial cells in vitm. Science 207:653-654 10. Feinberg AP, Vogelstein B (1984) Addendum. A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal Biochem I37:266-267 1.1. Form DM, Pratt BM, Madri JA (1986) Endothelial cell proliferation during angiogenesis: In vitro modulation by basement membrane components. Lab Invest 5552 1-530 12. Grant DS, Tashiro K-I, Segui-Real B, Yamada Y, Martin GR, Kleinman HK (1989) Two different laminin domains mediate the differentiation of human endothelial cells into capillay-like structures in vitro.Cell 58:933-943 .13. Haudenschild CC, Dahniser D, Folkman J, Klagsbrun M (1976) Human vascular endothelial cells in culture: lack of response to serum growth factors. Exp Cell Res 98:175182 14. Hobson B, Denekamp J (1984) Endothelial proliferation in tumours and normal tissues: Continuous labelling studies. Br J Cancer 49:4054 I3 15. Janzer RC, Raff MC (1987) Astrocytes induce blood-brain barrier properties in endothelial cells. Nature 325:253-257 6. Jaye M, McConathy E, Drohan W, Tong B, Duel T, Maciag T (1985) Modulation of the s i s gene transcript during endothelial cell differentiation in v i m . Science 228:882-885 7. Lee EY-H, Lee W-H, Kaetzel CS, Parry G, Bissell MJ (1985) Interactions of mouse mammary epithelial cells with collagen substrata: Regulation of casein gene expression and secretion. Proc Natl Acad Sci USA 82:1419-1493 8. Liang P, Pardee AB ( I 992) Differential display of eukaryotic messenger RNA by means of the polymerase chain reaction. Science 257:967-97 I 19. Liaw L, Schwartz SM (1993) Comparison of gene expression in bovine aortic endothelium in vivo versus in vitro. Arteriosclerosis and Thrombosis I3:985-993 20. Maniatis T, Fritsch EF, Sambrook J (1982) Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N Y 21. Milici AJ, Furie MB, Carley WW (1985) The formation of fenestrations and channels by capillary endothelium in vitro. Proc Natl Acad Sci USA 82:6181-6185 22. Sage H, Pritzl P, Bornstein P (1981) Characterization of cell matrix associated collagens synthesized by aortic endothelial cells in culture. Biochemistry 20:435-442 23. Sanger F, Nicklen S, Coulson AR (1977) DNA sequencing with chain-terminating inhibitors. Proc Natl Acad Sci USA 74:5463-5467 24. Sasaki M, Kato S, Kohno K, Martin GR, Yamada Y (1987) Sequence of the cDNA encoding the laminin B 1 chain reveals

226 a multidomain protein containing cysteine-rich repeats. Proc Natl Acad Sci USA 84:935-939 25. Schwartz SM, Benditt EP (1977) Aortic endothelial cell replication I. Effects of age and hypertension in the rat. Circ Res 4 1 :248-255 26. Sharma YK, Hinojos CM, Mehdy MC (1992) cDNA cloning, structure and expression of a novel pathogenesis related protein in bean. Mol Plant Microb Interact 5:89-95 27. Starksen NF, Harsh GR, Gibbs VC, Williams LT (1987) Regulated expression of the platelet-derived growth factor A chain

gene in microvascular endothelial cells. J Biol Chem 262: I438 I - I4384 28. Sturzl M, Roth WK (1990) “Run-Off ” synthesis and application of defined single-stranded DNA hybridization probes. Anal Biochem 185:164-169 29. Voyta JC, Via DP, Butterfield CE, Zetter BR (1984) Identification and isolation of endothelial cells based on their increased uptake of acetylated-low density lipoprotein. J Cell Biol 99:2034-2040