Expression of factor VIII by murine liver sinusoidal endothelial cells.

Factor VIII (fVIII) is the procoagulant plasma glycoprotein that is missing or decreased in hemophilia A. The cellular origin of fVIII synthesis is controversial. Liver transplantation cures hemophilia A, demonstrating that the liver is a major site of fVIII synthesis. We detected fVIII mRNA in purified populations of murine liver sinusoidal endothelial cells (LSECs) and hepatocytes, but not Kupffer cells. LSECs and hepatocytes contained comparable numbers of fVIII mRNA (40 and 70 transcripts per cell, respectively) by quantitative competitive reverse transcriptase-polymerase chain reaction analysis. There was not detectable mRNA for factor IX, a hepatocyte marker, in the LSEC preparation, nor was there detectable mRNA for von Willebrand factor, an endothelial cell marker, in the hepatocyte preparation. This excludes the possibility that detectable fVIII mRNA is due to cross-contamination in the hepatocyte or LSEC preparations. Primary cultures of LSECs were established in which fVIII mRNA levels were indistinguishable from purified LSECs. LSECs secreted active fVIII into the culture medium. This finding represents the first demonstration of homologous expression of fVIII mRNA and protein in cell culture and should facilitate studies of fVIII gene regulation. Additionally, LSECs potentially are targets for a fVIII transgene during gene therapy of hemophilia A.

containing DMEM/F-12, 15% FBS, 250 g/ml dibutyryl cAMP was added. Initially, the cells were 30 -40% confluent and grew to confluence in 1 or 2 days. Each well contained approximately 1 ϫ 10 5 cells at confluence. Confluent cultures of human umbilical vein endothelial cells and human dermal microvascular endothelial cells were purchased from the Emory Skin Cell Center and were also maintained in DMEM/F-12, 15% FBS, 250 g/ml dibutyryl cAMP. For studies of fVIII secretion, confluent monolayers were washed three times with HBSS and then maintained in AIM-V, a serum-free cell culture medium for 2 days.
Freshly isolated hepatocytes were plated onto collagen-coated Costar six-well plates (962 mm 2 per well) at 6 ϫ 10 5 cells per well for 30 min in DMEM/F-12, 10% FBS. Nonadherent cells were washed off with HBSS. Hepatocytes were maintained in medium containing DMEM/ F-12 plus 10% FBS for 2 days prior to assay for fVIII secretion.
Characterization of Cells-Purified cells were prepared for differential staining and light microscopy by centrifugation onto microscope slides using a Cytospin 3 Cell Preparation System (Shandon Scientific, Cheshire, United Kingdom). Cells were stained using a modified Wright-Giemsa stain (Diff-Quik, Baxter, McGaw Park, IL).
Freshly isolated LSECs and trypsin/EDTA-solubilized cultured LSECs were characterized further by flow cytometry. Cells were diluted to 1 ϫ 10 5 cells/ml with Dulbecco's PBS, 3% FBS and incubated with saturating concentrations of dye-labeled specific antibody, isotype control antibody, or wheat germ agglutinin. The cells were washed once and resuspended in Dulbecco's PBS, 3% FBS. Data collection and analysis were done using a Becton Dickinson FACSort flow cytometer and CellQuest software.
RT-PCR Reactions-Total RNA from hepatocyte, Kupffer cell, and LSEC preparations and from cultured LSECs was isolated using a RNeasy Mini Kit (QIAGEN, Santa Clarita, CA). RNA was quantitated spectrophotometrically at 260 nm using an extinction coefficient of 25 ml/mg/cm. Reverse transcriptase (RT) reactions were conducted using a First-Strand cDNA Synthesis Kit (Pharmacia), 0.2 g of total RNA template and primers specific for fVIII, von Willebrand factor (vWf) or factor IX (Table I).
The resulting cDNA fragments were amplified by PCR using Taq DNA polymerase (Promega, Madison, WI) and the primers listed in Table I. Samples were denatured for 2 min at 94°C, followed by 28 cycles of denaturation for 1 min at 94°C, annealing for 2 min at 55°C, and elongation for 2 min at 72°C. Reactions were completed by a final elongation step for 5 min at 72°C. The products were subjected to 1.5% agarose gel electrophoresis and visualized using ethidium bromide.
FVIII mRNA levels in LSECs and hepatocytes were quantitated by a modification of published competitive RT-PCR procedures (13)(14)(15). In this method, a known concentration of fVIII-specific cRNA is added to the RT reaction, which produces a cDNA that competes for the native fVIII cDNA during PCR amplification. The point of equivalence, where cRNA-derived and mRNA-derived cDNA products are equal, determines the number of mRNA molecules in the sample.
To initially test the method, two regions of fVIII were amplified using cRNAs and PCR primers corresponding to the A3 and C2 domains of murine fVIII. The PCR products contained intron-spanning sequences to avoid potential signals from contaminating genomic DNA. The cRNAs were identical to endogenous fVIII mRNA except for an internal 50-bp deletion so that the resulting cRNA-derived PCR product could be distinguished from the endogenous fVIII PCR product electrophoretically. This produced 500/450-and 400/350-bp products for the A3specific and C2-specific PCR reactions, respectively.
The loop-out cDNA products were cloned into pBluescript II phagemid (Stratagene, La Jolla, CA) using PstI restriction sites that had been incorporated into the PCR primers (underlined). The sequence of the cDNA products was verified by dideoxy sequencing. The cRNA was produced by T3 RNA polymerase-catalyzed in vitro transcription off pBluescript using a Ribomax Transcription Kit (Promega) and purified using a RNeasy Mini Kit. The molar concentration of cRNA was calculated from the absorbance at 260 nm and the molecular weight.
The cRNA-dependent inhibition of endogenous fVIII cDNA synthesis was quantitated by densitometry of ethidium bromide-stained products. Gel photographs were obtained using a Hewlett-Packard 6200C ScanJet scanner and were analyzed using digitization and quantitation software (UN-SCAN-IT, Silk Scientific, Orem, UT). Plots of the ratio of mRNA-derived and cRNA-derived products as a function of cRNA concentration were linear. The point of equivalence was determined by linear regression analysis.
FVIII Activity Assay in Endothelial Cell Culture Supernatants-The measurement of fVIII in cell culture supernatants was measured using a plasma-free chromogenic assay that measures the thrombin-activated fVIII-dependent rate of factor X activation by factor IXa as described previously (16,17). The reaction components included limiting activated factor VIII, 0.5 nM porcine factor IXa, 425 nM porcine factor X, and 50 M phospholipid. The initial rate of factor Xa formation was measured using the chromogenic substrate Spectrozyme Xa (American Diagnostica, Greenwich, CT). The results were compared with a standard curve prepared using human recombinant VIII of known coagulant activity (provided by Hyland-Immuno Div., Baxter Healthcare, Duarte, CA).

Isolation of Liver Cell
Populations-Hepatocytes were isolated from liver cell suspensions by low speed centrifugation. They were identified as large (20 -25 m), frequently binucleate cells with basophilic cytoplasm. The preparation was less than 5% contaminated by other cell types. Kupffer cell preparations were obtained using anti-CD11b magnetic bead cell sorting and identified by their relatively large size (10 -12 m), eccentric nuclei, and numerous vacuoles. CD11b-positive preparations contained approximately 80% Kupffer cells/monocytes and 20% granulocytes. LSECs were isolated from the CD11bnegative population by anti-ICAM-1 magnetic bead cell sorting. They were identified by their relatively small size (8 -10 m), clear cytoplasm, and oval nuclei (Fig. 1A). LSEC preparations typically were 90% pure and contained Kupffer cells (2%) and red blood cells (8%) as contaminants.
Analysis of the LSEC preparation by flow cytometry demonstrated that the cells expressed cell surface markers ICAM-1 and PECAM-1 and contained a subpopulation of the cells that bound wheat germ agglutinin (Fig. 2). LSECs did not stain with the monocyte/macrophage-specific marker CD14 or the endothelial marker VCAM-1 (data not shown). These findings are similar to previously described phenotypic characteristics of murine LSECs, which were ICAM-1 ϩ /PECAM-1 ϩ /VCAM-1 ϩ and contained two subpopulations that differentially bound wheat germ agglutinin (18,19). The reason for the lack of VCAM-1 staining in our LSEC preparation is not known, although the cells became VCAM-1 ϩ after cell culture.
Quantitation of FVIII mRNA by Competitive RT-PCR-Ini- tial experiments demonstrated that fVIII mRNA could be identified by RT-PCR in LSECs and hepatocytes, but not Kupffer cells (data not shown). A quantitative competitive RT-PCR method was developed to measure mRNA levels using a fVIIIspecific cRNA as a source of cDNA competitor as described under "Experimental Procedures." In this method, the concentration of fVIII mRNA can be determined by finding the concentration of cRNA that produces equal amounts of mRNA-derived and cRNA-derived PCR (13).
Separate competitive RT-PCR analysis of total liver was performed on regions corresponding to the A3 and C2 domains of the fVIII cDNA (Table I). Levels of 6.5 ϫ 10 5 and 7.4 ϫ 10 5 fVIII mRNA transcripts per g of total RNA were obtained using the A3-specific and C2-specific primers, respectively (data not shown). These results were within the experimental error of the method and verified that the region of fVIII se-lected for amplification did not bias the measurements. Further experiments were performed using A3-specific primers. Fig. 3 shows the results of a competitive fVIII RT-PCR assay of purified LSECs. Lanes 2-7 show that the concentration of 500-bp PCR product derived from endogenous fVIII mRNA is inversely related to the concentration of added competitor cRNA, which produces a 450-bp product. Regression analysis of the band intensities yielded a value of 3.3 ϫ 10 6 fVIII mRNA transcripts/g of total cellular RNA. Similar experiments were done using a hepatocyte preparation. The results of several experiments with LSECs and hepatocytes are shown in Table  II. Purified LSECs contain approximately 5-fold more fVIII transcripts/g of total cellular RNA than hepatocytes. Because hepatocytes contain more total RNA per cell than LSECs, hepatocytes contain approximately twice as many fVIII transcripts per cell than LSECs.
FVIII Expression in Cultured LSECs-Primary monolayer cultures of LSECs were established by growing cells on gelatin in the presence of DMEM/F-12, 15% FBS, dibutyryl cAMP. LSECs were spindle-shaped and displayed long cytoplasmic extensions prior to confluence (data not shown). As they approached confluence, they formed a polygonal, flat "cobblestone" monolayer ( Fig. 1B) that is characteristic of cultured endothelial cells (20). Cultured LSECs were positive by flow cytometry for ICAM-1, PECAM-1, VCAM-1, and bound wheat germ agglutinin (data not shown). Competitive RT-PCR analysis of fVIII mRNA of cultured LSECs (Fig. 4) yielded levels that were indistinguishable from purified LSECs (Table II). Additionally, like purified LSECs, cultured LSECs expressed vWf, but not factor IX (Fig. 4). Cultured LSECs maintained their phenotypic characteristics when trypsinized, split 1:1, and regrown to confluence. They did not survive a second passage under the growth conditions described under "Exper-

5124-5144
a Numbering is based on the cDNA sequences of murine fVIII (Gen-Bank accession number L05573) and murine factor IX (GenBank accession number M23109). vWf numbering is by homology to the human vWf cDNA sequence (47).
imental Procedures." The fVIII-dependent rate of factor Xa formation in LSEC supernatants was significantly greater than medium alone (Fig. 5). The measured rate corresponds to 1.4 ϫ 10 Ϫ2 units/ml of coagulant activity using human fVIII as a standard. In contrast, rates of factor Xa formation due to human umbilical vein endothelial cells or dermal microvascular endothelial cells were not significantly above background. FVIII activity was not detected in primary cultures of hepatocytes (data not shown).

DISCUSSION
Hepatic fVIII gene expression was studied by RT-PCR using purified populations of murine LSECs, hepatocytes, and Kupffer cells. We detected fVIII mRNA in LSECs and hepatocytes, but not Kupffer cells. The absence of factor IX mRNA, a hepatocyte marker, in purified LSECs (Fig. 3) excluded the possibility that the fVIII mRNA was due to contaminating hepatocytes. Conversely, the absence of vWf mRNA, an endothelial cell marker, in the hepatocyte preparation excluded the possibility of a false positive fVIII signal due to contaminating LSECs.
We also identified fVIII mRNA in cultured LSECs (Fig. 4). This represents the first demonstration of homologous expression of fVIII mRNA in primary cell culture. In contrast, previous studies of fVIII expression in cell culture have been conducted by transfecting fVIII gene fragments into liver cellderived cell lines (21,22) or Chinese hamster ovary cells (23,24). Our results should facilitate studies of fVIII gene regulation under more physiological conditions. There has been considerable controversy regarding which type of liver cell synthesizes fVIII. FVIII mRNA was identified in a human hepatocyte preparation by RNase protection assay (7). However, the preparation was contaminated with LSECs and Kupffer cells. In the same study, FVIII mRNA was not detected in a liver sinusoidal cell preparation. FVIII protein was localized by immune electron microscopy to the rough endoplasmic reticulum of both human hepatocytes and LSECs   FIG. 5. Secretion of fVIII from cultured LSECs. Endothelial cell monolayers were incubated in serum-free medium for 2 days. Supernatants were collected from LSEC, human umbilical vein endothelial cells (HUVEC), or dermal microvascular endothelial cells (MEC) and assayed for fVIIIa-dependent activation of factor X by factor IXa as described under "Experimental Procedures." The data represent the mean and standard deviation from cultures obtained from three separate preparations. (12). In contrast, fVIII was localized immunohistologically to LSECs but not hepatocytes (9,10,25,26). FVIII activity was detected in rat LSECs, but not hepatocytes (11). FVIII mRNA levels were not determined in that study.
Our results show that LSECs and hepatocytes make similar amounts of fVIII mRNA (40 versus 70 transcripts per cell, respectively, Table II). The ratio of hepatocytes to LSECs in liver is approximately 3 to 1 (27,28). Thus, the ratio of hepatocyte to LSEC steady-state fVIII mRNA transcripts in liver is approximately 5 to 1, suggesting that LSECs may contribute 15-20% of the normal hepatic synthesis of fVIII. FVIII levels rise to normal after liver transplantation in patients with hemophilia A, during which there can be no extrahepatic synthesis of fVIII (3). This indicates that LSECs potentially synthesize hemostatically significant amounts because fVIII levels in the 15-20% range substantially ameliorate the hemostatic defect in hemophilia A. In fact, fVIII levels actually are increased in fulminant hepatic failure (29 -31), which is associated with a profound loss of protein synthesis by hepatocytes. In this setting, levels of all other hepatic coagulation and fibrinolytic factors, including fibrinogen, prothrombin, factors V, VII, IX, X, XI, XII, XIII, prekallikrein, high molecular weight kininogen, protein C, plasminogen, antithrombin III, and a2-antiplasmin are decreased. Up-regulation of fVIII synthesis by LSECs may occur under these circumstances.
The amount of fVIII that is secreted by cultured LSECs is consistent with significant synthesis in vivo. There are approximately 8 ϫ 10 10 LSECs in adult human liver. Synthesis of 1.4 ϫ 10 Ϫ2 units of fVIII per 10 5 cells over 48 h (Fig. 5) would correspond to total LSEC expression of 5,600 units/day in vivo. By comparison, the estimated daily synthesis of fVIII in adults is roughly 3000 units because total circulating fVIII is approximately 3000 units and turnover occurs approximately daily.
In contrast, we did not detect fVIII activity in cultured hepatocytes. The expected levels are relatively low (1.4 ϫ 10 Ϫ2 units/ml observed in LSECs corresponds to 12 pM). Cellular uptake, which occurs during heterologous expression of fVIII by Chinese hamster ovary cells (23), or degradation by a protease secreted by hepatocytes, could account for the lack of detectable activity.
The identification of fVIII in LSECs raises the question of whether endothelial cells from other tissues contribute significantly to fVIII synthesis. FVIII mRNA has been detected in spleen, lymph node, heart, brain, lung, kidney, testes, muscle, and placenta (7,16,32), which is consistent with a common endothelial cell origin. However, fVIII has not been identified in cultured endothelial cells from human umbilical vein and other tissues (7,33). This is consistent with our finding that human umbilical vein or dermal microvascular endothelium does not contain detectable fVIII mRNA or activity (Fig. 5). Furthermore, hemophilia A is not cured by kidney transplantation (34,35), which further indicates that significant fVIII synthesis is not a general property of endothelium. The inability of bone marrow transplantation to cure hemophilia A (36) also excludes cells of the monocyte/macrophage system as a source of fVIII synthesis, which is consistent with our finding that Kupffer cells do not contain detectable fVIII mRNA.
However, several observations suggest that the extrahepatic synthesis of fVIII can be clinically significant. The most compelling finding is that liver transplantation from hemophilia A dogs to normal dogs does not produce hemophilia A (6). Additionally, spleen transplantation has been reported to produce increased fVIII levels in human (36 -39) and canine hemophilia A (40), although other investigators have not observed this in the canine system (5, 6). Overall, these findings, combined with the unequivocal demonstration of endothelial synthesis of fVIII in the present study, are most consistent with the hypothesis that both LSECs and nonhepatic endothelial cells contribute significantly to fVIII synthesis.
FVIII circulates bound noncovalently to vWf. In contrast to fVIII, vWf has been identified throughout the vascular endothelium and is widely used as an endothelial cell marker. Interestingly, vWf mRNA levels in liver are low relative to other tissues (41). Circulating fVIII protein levels are regulated by vWf (see Refs. 1 and 2, for reviews). Infusion of vWf into patients with severe von Willebrand disease leads to a rapid increase in circulating fVIII levels (42,43). This increase occurs without an increase in synthesis of fVIII mRNA (44). This indicates that vWf increases secretion of stored fVIII in this condition. Whether vWf influences fVIII mRNA and/or protein secretion under normal conditions is unknown. The identification of fVIII and vWf synthesis in the same cell type (Figs. 3 and 4) raises the possibility of coordinate gene regulation in vivo. The availability of cultured LSECs that synthesize both fVIII and vWf should facilitate studies in this area.
Hemophilia A is an attractive target for gene therapy. Our finding that the LSEC can support substantial synthesis of fVIII make it a potentially attractive host for fVIII synthesis. Portal vein infusion of a suitable vector could deliver fVIII directly to LSECs. Alternatively, fVIII could be introduced into cultured LSECs ex vivo, followed by autologous transplantation. Subsequent expression of fVIII by transduced LSECs under physiological conditions, particularly with respect to regulatory control by vWf, could offer a superior approach to the management of hemophilia A.