Characterization and visualization of murine coagulation factor VIII-producing cells in vivo

Coagulation factors are produced from hepatocytes, whereas production of coagulation factor VIII (FVIII) from primary tissues and cell species is still controversial. Here, we tried to characterize primary FVIII-producing organ and cell species using genetically engineered mice, in which enhanced green fluorescent protein (EGFP) was expressed instead of the F8 gene. EGFP-positive FVIII-producing cells existed only in thin sinusoidal layer of the liver and characterized as CD31high, CD146high, and lymphatic vascular endothelial hyaluronan receptor 1 (Lyve1)+. EGFP-positive cells can be clearly distinguished from lymphatic endothelial cells in the expression profile of the podoplanin− and C-type lectin-like receptor-2 (CLEC-2)+. In embryogenesis, EGFP-positive cells began to emerge at E14.5 and subsequently increased according to liver maturation. Furthermore, plasma FVIII could be abolished by crossing F8 conditional deficient mice with Lyve1-Cre mice. In conclusion, in mice, FVIII is only produced from endothelial cells exhibiting CD31high, CD146high, Lyve1+, CLEC-2+, and podoplanin− in liver sinusoidal endothelial cells.


Identification of FVIII-producing cells.
We attempted to identify EGFP-positive organs responsible for FVIII production in F8 Δ -KI mice by flow cytometry. We focused on CD31-positive endothelial cells in organs because endothelial cells are reported to be a major source of FVIII production 12,13 . CD31-expressing liver endothelial cells were the only cell population to express EGFP, and no EGFP-positive endothelial cells were detected in the lung, kidney, spleen, intestines, lymph nodes, or bone marrow (Fig. 2a). We further examined the characteristics of EGFP-positive endothelial cells in the liver. Liver endothelial cells were separated into three populations according to their CD31 and CD146 expression profile (Fig. 2b). EGFP expression in F8 Δ -KI  [16][17][18][19][20][21][22][23][24][25][26][27][28] in F8 Δ knock-in mice. PCR was performed using the indicated F and R primers in (a). www.nature.com/scientificreports/ mice was mainly observed in the CD31 high CD146 high population, but expression levels of 0.03% and 4.84% were observed in the CD31 -CD146population and CD31 mid CD146 mid population, respectively (Fig. 2c). Lymphatic endothelial cells have previously been reported to be FVIII-producing cells 20 , and many of their cellular characteristics are common to sinusoidal endothelial cells. The EGFP-positive cells in CD31 high CD146 high population also expressed Lyve1, a marker for sinusoidal endothelial cells and lymphatic endothelial cells (Fig. 2d). Low levels of Lyve1 expression were also observed in the CD31 mid CD146 mid population with low EGFP expression ( Supplementary Fig. S1a). We focused on the expression of podoplanin as a marker for lymphatic endothelial cells to identify the different characteristics between FVIII-producing cells and lymphatic endothelial cells 21 . Although mAb was useful to detect podoplanin in lung cells ( Supplementary Fig. S1b), podoplanin was not expressed in the CD31 high CD146 high Lyve1 + population of the liver, and EGFP-positive cells were detected in the podoplanin − Lyve1 + population (Fig. 2e). LSECs reportedly express several members of CLEC members [22][23][24] . EGFP-positive cells in the CD31 high CD146 high Lyve1 + population expressed CLEC-2, a physiological receptor for podoplanin, abundantly expressed on platelets 25 (Fig. 2e). These data suggest that FVIII-producing sinusoidal endothelial cell in the liver can be clearly distinguished from lymphatic endothelial cells in the expression profile of podoplanin and CLEC-2.
Localization of FVIII-producing cells. We employed immunohistochemistry for EGFP expression to identify the histological localization of FVIII-producing cells. Immunohistochemical analysis did not detect EGFP expression in liver parenchymal cells (hepatocytes) or around the central vein ( Supplementary Fig. S2). Intravital microscopy clearly identified EGFP expression in the liver of F8 Δ -KI mice, revealing that EGFP-positive cells were localized to liver endothelial cells constructing thin sinusoids (Fig. 3a, Supplementary Movies 1 and 2).
We stained liver sections with antibodies against endothelial markers, such as CD146, Lyve1, CLEC-2, and VWF ( Fig. 3b-e and Supplementary Fig. S3), to further determine the profile of EGFP-positive cells in the liver. CD146-positive endothelial cells were found to be present not only in the sinusoid but also in the central veins (Fig. 3b). Lyve1 and CLEC-2 were only expressed in sinusoidal endothelial cells, whereas the dominant expression of VWF was observed in the central veins ( Fig. 3c-e). Taken together, it was observed that EGFP-positive cells localized in the sinusoids and coexpressed CD146, Lyve1, CLEC-2, and VWF.

FVIII-producing cells during development.
We next investigated EGFP expression in the liver during the embryonic development of F8 Δ -KI mice (Fig. 4a). No EGFP expression was detected in the whole fetus at embryonic day (E)10.5. Of note, EGFP expression in CD31 high CD146 high Lyve1 + CLEC-2 + liver endothelial cells surged from 5% at E12.5 to 80% at E14.5. Furthermore, its expression reached more than 90% at birth (Fig. 4a). These results suggest that FVIII in CD31 high CD146 high Lyve1 + CLEC-2 + liver endothelial cells expressed in the later phase of embryonic development.
We further examined mRNA expression profiles in Lyve1-positive cells during embryonic development. We isolated CD31 high CD146 high Lyve1 + cells from the liver obtained from E12.5 and E17.5 mouse embryos and compared mRNA expressions by microarray analysis (Supplementary Fig. S4a and Fig. 4b). Of 45,037 probes examined, 8.0% and 7.4% of the genes were found to be significantly decreased or increased more than twofold, respectively (Fig. 4c). We analyzed the expression of coagulation factors and endothelial markers. We confirmed the decrease in CD34 expression and significant increase in Fcgr2b and F8 (Fig. 4d). The genes that were upregulated at E17.5 also included Mcam (CD146) and Lyve1, which were used as cell surface markers for cell sorting. The expressions of other coagulation factors and Vwf were not changed. We also analyzed Clec family members and found dominant expression of Clec2d, Clec14a, Clec1b, and Clec4g ( Supplementary Fig. S4b). The upregulated genes such as Clec1b, Clec4g, and Stab2 were also specifically expressed in CD31 high CD146 high Lyve1 + liver endothelial cells of adult mice ( Supplementary Fig. S5).

Conditional deletion of F8 in Lyve1-positive endothelial cells. Previous reports suggested mating
with Tie2-Cre mice, but not Alb-Cre mice, suppressed FVIII:C expression in F8 conditional deficient mice 12,13 . We crossed F8 conditional deficient mice with Lyve1-Cre mice to further confirm the importance of sinusoidal endothelial cells for FVIII production. Because Lyve1 is expressed in sinusoidal endothelial cells but not vascular endothelial cells 26 , Lyve1 promoter enables to more specifically express Cre in sinusoidal endothelial cells than Tie2 promoter. We employed F8 fl/fl conditional deficient mice (F8 tm1Rmnt/J ) possessing 2 loxP sites flanking exons 17-18 13 . Male offspring of crosses between female F8 fl/fl conditional deficient mice and male Lyve1-Cre mice showed a marked decrease in FVIII:C by Cre expression compared with male littermates without Cre expression (Fig. 5). These data clearly support our idea that FVIII was produced predominantly from LSECs.

Discussion
Details of the exact organ and endothelial cell types involved in FVIII synthesis have been controversial for decades 3,4,7-9,15-18 . To resolve this issue, we generated F8 flox knock-in mice in which Cre induction led to the elimination of F8 and expression of EGFP to aid the identification of organs and cells responsible for FVIII production. After the systemic induction of Cre, we could clearly identify FVIII-producing cells as EGFP-positive cells in mouse.
We found that the liver was the only organ to produce FVIII, and sinusoidal endothelial cells exhibiting CD31 high , CD146 high , Lyve1 + , and CLEC-2 + were responsible for FVIII production. This endothelial fraction was not observed in other organs, such as the lung, kidney, spleen, intestines, lymph nodes, and bone marrow. Although previous observation revealed that lymphatic endothelial cells were the major source of extrahepatic FVIII production 20 , EGFP-positive cells could not be detected in lymph nodes and spleen in this study. Our www.nature.com/scientificreports/ findings are consistent with the fact that the FVIII:C of patients with hemophilia A reaches a normal range soon after liver transplantation 3,4 . Sinusoidal endothelial cells and lymphatic endothelial cells substantially share similar cellular and phenotypic characteristics, for example, minimal basement membranes and expression profile of Lyve1 + , VSP-1 + , and Reelin + and CD34 −27 . In addition, some members of C-type lectin receptor family such as CLEC-4G and CLEC-4M are expressed in both cells 23,24 . In contrast, the FVIII-producing sinusoidal endothelial cells could be completely distinguished from lymphatic endothelial cells by expression profiles of podoplanin and CLEC-2; sinusoidal endothelial cells express CLEC-2, whereas lymphatic endothelial cells express its ligand podoplanin. It is very interesting that two molecules acting as receptor-ligand relationship could separate the characteristics between these cells. Podoplanin on lymphatic endothelial cells plays an essential role for the blood/lymphatic separation and lung development through the interaction with CLEC-2 on platelet surface 25 . Further analysis will be required to identify the ligand for CLEC-2 on sinusoidal endothelial cells because the physiological role of CLEC-2 on sinusoidal endothelial cells still remains unknown. We did not observe FVIII-producing endothelial cells until E12.5 and found that sinusoidal endothelial cells expressing FVIII surged at E14.5 during embryonic development. These data suggest that the FVIII production system matures in the later phase of liver development. EGFP expression profile in sinusoidal endothelial cells just after birth was similar to those in adult liver. This is consistent with previous clinical observation that plasma FVIII:C in newborn is the same as the adult, whereas levels of vitamin K-dependent coagulation factors at birth are significantly lower than those of adults 28 . Sinusoidal development follows a series of processes 29 . First, sinusoidal endothelial cells are lined by a laminin-rich basal membrane, then start to fenestrate by stereotypic differentiation at E14-15, and finally undergo maturation of their anatomical structure 29,30 . Our investigation of the profile of FVIII-producing cells could help identify the factor(s) responsible for sinusoid maturation because the expression of FVIII seems to occur during the later phase.
VWF and FVIII circulate within the blood as a tight complex in which VWF stabilizes the structure of FVIII and protects it from proteolytic degradation 31 . It was previously reported that VWF is mainly expressed in capillary endothelial cells but not in lymphatic endothelial cells, whereas FVIII is expressed in lymphatic endothelial cells, suggesting a differential pattern of expression between the proteins 20 . However, EGFP expression by the F8 gene promoter was not observed in lymphatic endothelial cells in this study. We observed predominant VWF expression in endothelial cells of the central vein; hence, our data also suggest that vascular endothelial cells are the main source of circulating VWF. However, it should be noted that most VWF molecules exist in a nonbinding form with FVIII in the circulation because plasma concentration of VWF is significantly higher than those of FVIII (50 nM and 1 nM, respectively) 32 . In this study, sinusoidal endothelial cells also expressed VWF at lower levels, and VWF colocalized with EGFP-positive cells in F8 Δ knock-in mice. It is conceivable that formation of the FVIII/VWF complex occurs within the cells. Indeed, human endothelial cells stimulated with vasopressin in vitro secreted cell-anchored ultralarge VWF strings covered with bound FVIII 19 . Further analysis is needed to clarify the involvement of VWF in the endothelial cell secretion of FVIII.
The current adeno-associated virus (AAV) vector-mediated gene therapy for hemophilia expresses the target coagulation factor in hepatocytes [33][34][35] , which are non-physiological cells for producing FVIII. A recent study revealed that the plasma activity of FVIII ectopically expressed from hepatocytes by AAV-mediated gene therapy was 1.3-to 2.0-fold higher in a one-stage coagulation assay than in a chromogenic substrate assay 36 . The F8 gene sequence is identical, indicating that FVIII has undergone unique post-transcriptional modifications in hepatocytes. Ectopic post-transcriptional modifications have been suggested to affect the folding, trafficking, and secretion of FVIII in hepatocytes and alter cell viability and function via endoplasmic reticulum stress 37 . Moreover, FVIII expression in hemophilia A gene therapy tends to gradually decrease over time compared to that in hemophilia B. Hemophilia A gene therapy also required the administration of huge amounts of AAV vectors in clinical trials, even though the molar concentration of FVIII is much lower than that of FIX 33 . Gene therapy targeting FVIII-producing cells may resolve these problems. Liver sinusoid endothelial cells are almost quiescent under the physiological state 38 , representing an ideal target for AAV-mediated gene therapy. Therefore, we are currently developing strategies to efficiently introduce AAV vectors into FVIII-producing cells in vivo for hemophilia A gene therapy and genome editing. Furthermore, it has been reported that the administration of sinusoidal endothelial cells expressing FVIII generated from patient-derived induced pluripotent stem cells  www.nature.com/scientificreports/ (iPSCs) improved the hemophilia A phenotype in mice 39 . Deriving sinusoidal endothelial cells from the patient's gene-corrected iPSCs may also provide an attractive source for cell-based hemophilia A therapy. This study has some limitations. First, we must consider the species differences to extrapolate the implications of this study to human biology. It is possible that the mechanism of FVIII synthesis differs among species. Indeed, several reports suggest the existence of a compensatory mechanism by other organs to maintain the plasma FVIII. In a porcine model of fulminant hepatic failure caused by a total hepatectomy, plasma FVIII levels increased in the absence of the liver 40 . In addition, previous observations in dogs and humans suggest that the liver may www.nature.com/scientificreports/ not be the unique organ that produces FVIII. In fact, when the liver of a hemophilic dog was transplanted into a normal dog, the recipient did not develop the same phenotype as the hemophilic animal 41 . In humans, FVIII production was sustained in a transplant recipient of a hemophilia A donor liver 42 . Second, our study clearly identified FVIII-producing cells in vivo, but we could not assess the cellular localization and release mechanism of FVIII. Because we can only look at EGFP protein instead of FVIII, our data could not consider the protein interaction of FVIII. FVIII localization within the cells might be changed by the interaction with other proteins.
Although previous reports suggest that FVIII is packaged with VWF in Weibel-Palade bodies in endothelial cells 19 , the existence of these bodies in sinusoidal endothelial cells is controversial 43 .
In conclusion, this study suggested that FVIII in mice is mainly produced from LSECs with expressions of CD31 high , CD146 high , Lyve1 + , and CLEC-2 + . Moreover, we found that FVIII production in the liver starts and matures during the later phase of embryonic development. These data help resolve the controversy about which organ-derived endothelial cell types synthesize FVIII. The next challenge will be to elucidate the precise mechanism by which FVIII release is regulated, with the ultimate aim of developing gene therapy to efficiently transduce FVIII-producing cells in vivo.

Dissociation of cells from primary tissue and flow cytometry. Mice anesthetized with isoflurane
were perfused with a perfusion buffer (10 mM HEPES, 140 mM NaCl, 6.5 mM KCl, 5 mM CaCl 2 , pH 7.4) and their tissues were removed. Bone marrow cells and spleen cells were isolated as previously described 48  Immunohistochemical analysis. Mice anesthetized with isoflurane were perfused with 50 mL PBS. For double staining of EGFP with endothelial markers, liver tissues were fixed with 4% paraformaldehyde, incubated with PBS containing sucrose (10-20%), and then frozen in the presence of Tissue-Tek O.C.T. Compound (Sakura Fintek Japan, Tokyo, Japan) in dry ice/ethanol. Tissue sections were blocked with 5% donkey serum and then incubated with an anti-EGFP polyclonal antibody (MBL Co.) and a specific antibody for endothelial markers as follows: anti-CD31 mAb (clone 390; BioLegend), anti-CD146 mAb (clone ME-9F1; BioLegend), anti-CLEC-2 mAb (clone 2A2B10; kindly provided from Dr. Suzuki-Inoue), anti-Lyve1 mAb (clone ALY7; Thermo Fisher Scientific), or anti-VWF polyclonal antibody (code ab11713; Abcam, Cambridge, UK). Sections were then incubated with a species-specific secondary antibody conjugated with Alexa Fluor 594 (Thermo Fisher Scientific) and anti-rabbit IgG conjugated with Alexa Fluor 488 (for the detection of EGFP) for 2 h at 4 °C. Slides were mounted with VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA, USA). Immunofluorescence staining was observed and photographed using a confocal microscope (Leica TCS SP8; Leica Microsystems, Wetzlar, Germany). Intravital microscopy. Intravital microscopy was performed as reported previously 49 . Briefly, anesthetized mice were injected with rhodamine B isothiocyanate-Dextran (5 mg/body; Sigma-Aldrich) and Hoechst 33342 (3 mg/body; Thermo Fisher Scientific). A resonance scanning confocal microscope (Nikon A1R; Nikon, Tokyo, Japan) was used to obtain sequential images of the liver.
Isolation of liver endothelial cells of fetal mice. C57BL/6J fetus at E12.5 and E17.5 were sacrificed, and then livers were removed. After treatment of liver tissues with collagenase/dispase and DNase I, the cells were incubated with anti-CD16/CD32 antibody and then stained with fluorescent-labeled antibodies against CD31, CD146, and Lyve1. Dead cells were stained with 7-AAD. A BD FACSAriaII Special Order Research Product (BD Biosciences) was used to isolate the CD31 high CD146 high Lyve1 + cells.
Microarray analysis. The total RNAs were isolated from CD31 high CD146 high Lyve1 + cells in fatal liver (E12.5 and E17.5) by NucleoSpin RNA kit (MACHEREY-NAGEL GmBH & Co, Duren, Germany). mRNA expressions were analyzed by GeneChip Mouse Genome 430 2.0 Array (Thermo Fisher Scientific) and detected by GeneChip Scanner 3000 7G (Thermo Fisher Scientific). Transcriptome Analysis Console software (Thermo Fisher Scientific) was used for data analysis.
Measurement of FVIII activity. FVIII activity (FVIII:C) was measured using an activated partial thromboplastin time (APTT)-based one-stage clotting-time assay on an automated coagulation analyzer (Sysmex CA-500 or CS-1600 analyzer; Sysmex Corp., Kobe, Japan). Clotting time was measured by APTT (Thrombocheck APTT, Sysmex Corp.) after mixing the test mouse plasma with human FVIII-deficient plasma (Thrombocheck FVIII, Sysmex Corp.). FVIII:C was determined based on a calibration curve prepared using pooled plasma from wild-type C57BL/6J mice. www.nature.com/scientificreports/ Statistical analyses. Data are presented as means ± standard error of the mean of at least three independent experiments. Statistical significance was determined by 2-tailed Student's t tests using GraphPad Prism version 8 (GraphPad Software, San Diego, CA, USA) or Microsoft Excel (Microsoft, Redmond, WA, USA). A value of P < 0.05 was considered to be statistically significant.

Data availability
The original data in this study are available upon request from the corresponding author.