Transcriptional Analysis Of a Tissue-Specific Factor VIII Knockout Model Demonstrates The Importance Of Endothelial Factor VIII Synthesis

Definitively identifying the cells responsible for synthesis of coagulation factor VIII (F8) has proven to be a challenge. Transplantation studies demonstrate that as an organ liver is the major, but not exclusive source of plasma F8. Within the liver F8 expression has been variously attributed to hepatocytes, and/or liver sinusoidal endothelial cells, and/or Kupffer cells. Extrahepatic transcription of F8 mRNA appears to be nearly ubiquitous at a low level throughout the body. Previous studies have relied upon retrospective post-expression detection of F8 protein or mRNA using a variety of immunochemical, in situ, and cell isolation techniques, but continuing controversy speaks to the difficulties in localizing expression of a trace protein such as F8.

We used a rather different, pre-emptive approach to address the question of F8 synthesis. We developed a conditional F8 knockout (KO) mouse model that allows inactivation of the F8 gene, thus preventing expression, in specific cell types. Exons 17/18 of the F8 gene were flanked by LoxP sites (floxed) resulting in their excision in cells expressing Cre recombinase. Tissue-specific Cre-expressing mouse strains were cross-bred with floxed (F8^F) mice to generate tissue-specific F8-KO models. Embryonic Cre expression resulted in a new F8^KOstrain displaying a severe hemophilia A phenotype. A hepatocyte-specific F8-KO has completely normal plasma F8 levels, while each of 3 endothelial cell (EC)-Cre models displays a reduced-F8 phenotype that correlates in severity with endothelial Cre efficiency. Presumably due to a shared hemangioblast progenitor, Cre is expressed with similar efficiency in both EC and hematopoietic cells in these models. Plasma F8 is undetectable in the most efficient EC-KO model. In contrast, a highly efficient hematopoietic F8-KO model presents with only modestly reduced F8 levels, likely due to off-target effects.

RNA analysis revealed that the F8^KO allele produces 2 alternatively spliced transcripts in roughly equivalent amounts. The 1st transcript represents the predicted exon 16/19 splicing event. In the 2nd transcript, 46bp at the 5’ end of intron 16 are retained due to the same cryptic splice site observed in the Kazazian exon 17-disrupted F8^null model. Combined, the 2 F8^KO allele transcripts are present at ∼1/8 to 1/5 of normal levels in the F8^KO strain. No normal F8 transcripts are present. In the phenotypically normal hepatocyte-KO model ∼70% of total liver gDNA is converted to the F8^KO allele, indicative of very efficient hepatocyte Cre activity, yet almost exclusively normal F8 mRNA is present, with only traces of F8^KO message. This is consistent with endothelial synthesis as our further results indicate. For the 3 EC-KO models, plasma F8 levels were correlated with hepatic levels of normal F8 mRNA, and inversely correlated with F8^KO transcripts. Excessive F8^F to F8^KOconversion in the hematopoietic-Cre model suggests variable loss of tissue-specificity.

In the most efficient, functionally hemophilic EC-KO model, ∼20% of liver gDNA is converted to the F8^KO allele, in good agreement with the expected number of hepatic EC, and F8^KO mRNA is present at ∼10% of normal liver levels. With undetectable plasma F8, the continued production of normal F8 mRNA at a similar low level (∼10%) by the remaining 80% of Cre-negative, presumably non-endothelial hepatic cells, was unexpected. In addition to liver we found both normal and F8^KO message only in kidney and perhaps brain. As expected, only F8^KOmRNA was found in spleen and bone marrow, but the presence of exclusively normal mRNA in heart, intestine, testis, lung, and thymus, at relatively normal (low) levels, was surprising. The persistence of widespread transcriptional “expression” of F8, albeit in a functionally hemophilic mouse, is reminiscent of the near-ubiquitous presence of low level F8 transcription in normal mice. This low level transcription apparently does not support functional plasma F8 production, at least not in these EC-KO mice.

In summary, our results support the hypothesis that synthesis of F8 is a function of endothelial cells, both in the liver and presumably elsewhere. Neither hepatocytes nor hematopoietic cells appear to contribute significantly to steady-state plasma F8 levels. Transcriptional analysis of normal and F8^KO-specific transcripts provides further support for the localization of F8 expression to endothelial cells.