Low-carb tissue factor?

Tissue factor (TF), the cell-surface protein responsible for initiating blood clotting following vascular injury, is a small and seemingly uncomplicated protein, but it has generated a surprising number of controversies. TF is an integral-membrane protein whose extracellular domain consists of either 217 or 219 amino acids (for the human protein), followed by a single transmembrane domain and a short cytoplasmic tail [1]. The main function of TF is to bind factor VIIa (FVIIa) and serve as the regulatory subunit of the two-subunit enzyme that triggers blood clotting (the TF:FVIIa complex; see Fig. 1). The extracellular domain of TF has three Asn residues (Asn 11, 124 and 137) that occur in consensus sequences for attachment of N-linked carbohydrates (Asn-X-Ser/Thr). N-linked carbohydrate chains are found attached to all three of these Asn residues in recombinant human TF expressed in mammalian cells [2,3], and TF purified from human placenta [4]. Recombinant human TF produced in E. coli lacks such carbohydrate chains, but was reported by Paborsky et al. in 1989 to have a specific activity at least as high as that of recombinant TF produced in 293 cells (a human cell line) [3]. This result therefore indicated that the carbohydrate chains of human TF are dispensable for its procoagulant activity and paved the way for using bacterially-expressed TF in a variety of studies. When my own lab began making recombinant TF in bacteria in the mid-1990s, we also found, like Paborsky et al., that bacteria-expressed TF had specific activities in clotting assays that were at least as high as those of TF purified from human brain [5]. Since recombinant TF (and sTF, the isolated extracellular domain of TF) can be produced inexpensively and at high yield from bacterial cultures, in the 2+ decades since the Paborsky et al. study appeared, a large number of studies have used recombinant TF produced in bacteria, ranging from detailed biochemical and enzymatic studies to x-ray crystal structures. Furthermore, some commercial thromboplastins that are widely used to perform Prothrombin Time (PT) clotting assays have, as their active ingredient, recombinant tissue factor made in bacteria. It is probably a fair statement that much of what we now know about the functional properties of TF has come from studies using recombinant human TF and/or sTF produced in bacteria.

Figure 1.

X-ray crystal structure of the sTF:FVIIa complex, showing the three Asn residues in TF to which N-link glycans are attached. TF is rendered in gray and FVIIa in white. In the “front” view, the active site of FVIIa (with an inhibitor shown in stick rendering) faces the viewer. In the “back” view, the complex has been rotated about 180 degrees. The presumptive location of the membrane is indicated by the shaded rectangle. The structure is from PDB file 1DAN [11].

The idea that the carbohydrate chains of TF were dispensable for procoagulant function was challenged in a recent study by Krudysz-Ambio et al., who reported that the ability of the TF:FVIIa complex to activate factor X, or even small peptidyl substrates, was markedly affected by the presence of the N-linked carbohydrate chains on TF [4]. In particular, they reported that TF:FVIIa complexes prepared using TF purified from human placenta exhibited k_cat values several-fold higher than when TF was produced in bacteria or insect cells. When they enzymatically removed the carbohydrate chains from placental TF, its specific activity dropped to that of TF produced in bacteria or insect cells. (Removing the N-linked carbohydrate chains from TF produced in insect cells, on the other hand, did not affect its activity.) These authors concluded that at least one of the three N-linked carbohydrate chains of wild-type TF purified from human placenta is required for full activity, and furthermore, that full activity apparently requires a type of N-linked carbohydrate produced when TF is made in human but not insect cells.

In the present issue, Kothari et al. [6] address this controversy by examining the activity of TF lacking one, two or all three of its N-linked carbohydrate chains. They showed that treating human monocytes with tunicamycin caused the cells to express non-glycosylated TF but did not diminish the levels of TF protein expression. Interestingly, the TF activity on such cells (measured as rates of factor X activation by TF:FVIIa) was not diminished, suggesting that non-glycosylated TF is fully functional. They followed up by expressing recombinant TF in which the three Asn residues to which N-linked carbohydrates are attached were mutated to Ala (creating all three single mutants and double mutants, as well as a triple mutant lacking all three N-linked glycans). The TF mutants were expressed in CHO cells (a mammalian cell line) and cell-surface expression levels were quantified by measuring the binding of radiolabeled FVIIa and anti-TF antibody. Mutating these three Asn residues, singly or in combination, had no effect on cell-surface expression of TF or on the rate of factor X activation by the TF:FVIIa complex. This result contradicts the findings of Krudysz-Ambio et al. [4] but is consistent with the much earlier report from Paborsky et al. [3].

In addition to its well-known function as the triggering agent for blood clotting, TF also exhibits signaling properties, chiefly through the ability of the TF:FVIIa complex to activate protease-activated receptors (PARs) via limited proteolysis [1,7]. Kothari et al. [6] examined whether the N-linked carbohydrate chains of TF played a role in the signaling properties of TF, by expressing wild-type or mutant TF in cultured endothelial cells, and then examining the ability of the resulting TF:FVIIa complexes to transduce intracellular signals via cleavage of PAR-2. Interestingly, none of the Asn mutations, singly or in combination, significantly reduced TF-mediated signaling, indicating that N-linked carbohydrate chains of TF are dispensable for its signaling functions as well.

Kothari et al. [6] also examined whether the N-linked glycans of TF played a role in TF encryption/decryption on the cell surface. It has long been known that TF on the surface of healthy, resting cells has much less activity while retaining full ability to bind to FVIIa (TF encryption). Lysing the cells, damaging the plasma membrane, or treating cells with calcium ionophores or redox reagents causes elaboration of full TF procoagulant activity (TF decryption). The molecular basis of TF encryption/decryption is controversial [8]. Kothari et al. examined this question by treating transfected endothelial cells with calcium ionophore or HgCl₂, but found that wild-type and mutant TF activities were decrypted to the same extent.

It is always difficult to know where the truth lies when faced with two reports from highly regarded laboratories that come to opposite conclusions. A particular strength of the Krudysz-Ambio et al. study [4] is that it includes extensive biochemical characterization of TF produced from three sources, including, with follow-up papers from the same lab [9,10], extensive characterization of the glycans. A concern, however, is that the TF preparations were not produced in the authors’ own lab, but were obtained from others, with little information provided on how the proteins were produced. Could it be that the three TF preparations used by Krudysz-Ambio et al. varied in quality in ways that were unrelated to their carbohydrate contents? Furthermore, the obvious next step was to produce recombinant versions of TF in which the attachment sites for the three N-linked glycans were mutated, in order to identify which one(s) of the glycans were responsible for the increased activity of human placental TF over the two forms of recombinant TF that the authors tested. A particular strength of the Kothari et al. [6] study is that they made just such TF mutants and compared the mutants side-by-side in the same cell types (including both rodent and human cell lines). Repeated tests performed by these authors failed to find any role for the N-linked glycans of TF, confirming Paborsky et al. [3]. On the other hand, none of the resulting TF preparations were purified and studied biochemically; instead, the activities of the proteins were examined in cells, and the study relies on quantitation of TF by antibody and FVIIa binding. In this vein, Kothari et al. obtained puzzling data when they tried to quantify the levels of mutant versus wild-type TF by immunoblots, which gave data that did not agree with ELISA studies or cell-surface binding studies, and an explanation for this discrepancy was not found.

In conclusion, the study by Kothari et al. [6] in the present issue has carefully examined the role of each of the three N-linked glycans of human TF and can find no contribution of any of them to cell-surface expression of TF, to TF procoagulant activity, or to the ability of TF to transduce intracellular signals via PAR-2 activation. We have doubtless not heard the end of this story, however, especially given the track record of TF to generate and sustain scientific controversies.