Waging Warfarin against Tuberculosis: A Review of Dutton et al, 2010

A review of Dutton et al. Inhibition of bacterial disulfide bond formation by the anticoagulant warfarin. PNAS (2010) vol. 107 (1) pp. 297-301 PMCID: PMC2806739

This PNAS study from a few years ago looks at the potential antibacterial use of a popular anticoagulant, warfarin. Although my attention was drawn to this article at first through a journal club, I think that this is an interesting paper to review for several reasons. First, it is a paper published in a relatively high profile journal, and it also note worthy because it is research that has a more direct connection to clinical applications and drug development. 

This work has also been reviewed by other blogs and science writers (in a more timely fashion than I have done). For example, see a guest post in Small Things Considered blog written by two graduate students entitled "All is fair in love and warfarin". These students do a good job of summarizing the article, especially for a wider audience. A more in-depth review (and slightly more critical) entitled "Warfarin: Not Just Rat Poison" is also worth a read.

Here, I would like to add my commentary to this work, going further than the above two reviews / summaries in pointing out some flaws. The very rationale of this work is questionable, and might provide a glimpse into the misalignment of public interest and the motivation of the scientists. This review also happens to continue an unwitting trend of reviews on anti-tuberculosis drug papers, following my review of Shi et al, 2011, entitled "Grainy Westerns and Fuzzy Logic".

Manuscript Highlights

1. High doses of warfarin (and lower doses of the anticoagulant phenindione) are capable of inhibiting the growth of Mycobacterium smegmatis and tuberculosis (only warfain data for M. tb)

2. Mycobacterium VKOR, the homolog of the human target of Warfarin, is active when expressed in E. coli. The activity of this enzyme in these experiments is sensitive to Warfarin.

3. Surprisingly, inhibition of Mycobacterium VKOR by warfarin does not appear to be the reason why this drug inhibits M. tuberculosis growth. Should either VKOR or Warfarin really be the focus of drug development?

Synopsis:

In this study, Dutton and colleagues explore the use of the anticoagulant warfarin to inhibit the Mycobacterium tuberculosis VKOR enzyme (MtbVKOR). This work could lead to the development of novel, VKOR-targeting anti-tuberculosis therapeutics. This bacterial enzyme, MtbVKOR, is the ortholog of the human target of warfarin, Vitamin K epOxide Reductase (VKOR). While human VKOR is involved in the clotting pathway and bacterial VKOR is thought to be responsible for intracellular disulfide bond formation, the underlying enzymatic activity for these orthologs are very similar. In a previous study, the authors demonstrated that MtbVKOR is active in disulfide bond formation when expressed in E. coli, thereby providing a convenient system for studying the properties of this enzyme in vivo.

Utilizing this heterologous E. coli / MtbVKOR expression system, the authors demonstrate that high doses of warfarin can inhibit the activity of the Mycobacterium enzyme in this setting. This demonstrates that the bacterial and human VKOR enzymes share more than just some sequence similarity. Selection and analysis of warfarin-resistant variants of MtbVKOR supports this conclusion. Finally, the authors demonstrate that warfarin has antibacterial properties against Mycobacterium species, although they cannot identify VKOR as the target in this case.

Select 'Read More' to see the rest of the review.

A Critical Review:

This work argues that bacterial VKOR represents a novel target for the development of novel anti-tuberculosis drugs. Tuberculosis is a terrible disease that was a major health concern before the development of antibiotics. Now, with the emergence of antibiotic resistant strains of Mycobacterium tuberculosis, the causative agent of the disease (which used to be referred to as 'the consumption'), as well as tuberculosis infections in immune compromised individuals, there is a real need for the development of novel anti-tuberculosis treatments. Does this study lead the way to targeting disulfide bond formation? The evidence provided is limited, only 3 figures published along with the main text of the PNAS article. I shall go through them in turn. Most importantly, at the end, I will highlight how the final result and other information makes VKOR a questionable choice for antibiotic development.

First, the authors attempt to demonstrate (as they had done in a previous study) that MtbVKOR can function in disulfide bond formation in E. coli, effectively substitute for the endogenous DsbB enzyme (Figure 1). This evidence here is fairly clear; in a strain lacking DsbB, a plasmid borne His-6 epitope tag (N-terminal) version of MtbVKOR is necessary for oxidation of DsbA (as determined from mobility shifts on a gel for oxidized and reduced forms of DsbA). Addition of 5mM warfarin partially inhibits (about 50%) this oxidation activity, but only at intermediate levels of MtbVKOR expression. This is an important point that will be revisited; doubling the level of MtbVKOR inducer (25mM to 50mM) removes or masks the inhibitory effect of warfarin in this system. The protein levels of the enzyme can be seen in the bottom panel of the western in Figure 1A. It is difficult to assess the difference between protein levels under the two induction conditions, since the 50mM western blot may be overexposed. In Figure 2B, the protein levels of MtbVKOR induced with 10, 50 and 250mM IPTG can be observed. Even under the intermediate expression levels of MtbVKOR, warfarin has no effect on the oxidation of FlgI protein. The oxidation of FlgI, an E. coli flagellar protein which must be present and oxidized for the bacteria to be motile.

The authors make use of the oxidation-dependent motility of E. coli to select warfarin resistant variants of MtbVKOR. However, as mentioned before, His tagged MtbVKOR can lead to oxidized FlgI even in the presence of warfarin. The authors claim that His tagged MtbVKOR is expressed at stronger levels than the untagged version of the enzyme, and that this increased expression is the reason for the inability to inhibit motility with warfarin. However, the authors never demonstrate that expression of MtbVKOR is indeed lower when the His tag is not present. Of course, without the His-tag, it may have been technically difficult to detect the levels. What is demonstrated, however, is that motility is reduced (maybe about 50%) when untagged MtbVKOR is expressed with 50mM IPTG compared to the His-tagged version. It is not clear why this level of IPTG is utilized, when lower amounts show reduced expression with the His-tagged version in other experiments.

Four warfarin resistant MtbVKOR mutants were isolated using the motility assay. Three of them have mutations which the authors suggest are similar to warfarin resistant mutations found in the human VKOR. One of these, a Asp 50 to Glu mutation, is a conservative change and it is surprising it would have such a phenotype. The fourth resistance mutation is a change to the -1 region, not within the coding region of the enzyme. The authors speculate that this mutation alters MtbVKOR expression, which seems reasonable in light of the sensitivity the assays display to the protein levels. Still, it would have been beneficial if this hypothesis was actually tested.

To confirm the warfarin resistant nature of these mutations, the authors did not utilize the established DsbA assay. It is not clear why they did not utilize this assay, which would have been much better since it is then directly comparable to the first figure. Instead, in Figure 2, the authors show the results of a different assay, in which oxidation by MtbVKOR (WT or variants) or DsbB (not shown) are able to support beta-galactosidase activity. It is difficult to compare the ~50% decrease in oxidized DsbA with the ~10 fold increase in galactosidase activity with the two different forms (tagged or untagged, respectively) of MtbVKOR in the presence of warfarin. Telling, however (and not shown) is that galactosidase in the presence of MtbVKOR and 5mM warfarin is less than 25% of the activity observed with DsbB. I would question the real extent of warfarin inhibition of MtbVKOR.

The expression of the three warfarin resistance mutations is also analyzed in Figure 2B. Strong expression is observed, and it is similar to WT, but only at 250mM IPTG, which is much stronger than tested in any other assay. At this level of inducer, differences in expression may not be detectable. Much more informative would have been a western comparing wild-type and all variants at 5, 10, 25, or 50mM IPTG. Perhaps the variants (including the D50E and -1 mutants) simply have higher expression, not altered sensitivity to warfarin.

Finally, in Figure 3 (and the corresponding text), the authors claim that Mycobacterium growth is inhibited by warfarin. This is a very important (set of) result(s), and one that I considered leading this review with. If Mycobacterium was unaffected by warfarin, then the study would lose much significance (the effect on VKOR would be not be useful for antibiotic development). Instead, Mycobacterium (both smegmatis and tuberculosis) growth is inhibited by by warfarin at low millimolar concentrations. This is a rather high dose of warfarin; as antibacterial agent in this system, the authors report phenindione as being more effective (250µM necessary for inhibition of M. smegmatis; not tested for tuberculosis, although the reason for this omission is not given).

Importantly, warfarin still inhibits the growth of  M. smegmatis with DsbB, or MtbVKOR 'warfarin resistant' variants in place of the native MtbVKOR. The authors suggest two explanations:

1. Warfarin may have additional essential targets in mycobacteria

2. VKOR is not accessible to warfarin in mycrobacteria (Authors admit this is less likely)

As a stand-alone explanation, point #2 makes no sense. It is not even a relevant point to consider, since point #1 is likely to be the reason why mycobacteria growth is inhibited by warfarin. The possibility that the warfarin resistant variants were selected due to altered expression, but the warfarin sensitivity DsbB complemented strain makes this unlikely.

This last result, that mycobacteria are sensitive to warfarin but in a VKOR independent way, undermines the conclusion of the work; namely, that VKOR represents a novel class of targets for the development of anti-tuberculosis drugs. Instead, researchers and drug developers would benefit from trying to find the cause for the warfarin sensitivity of mycobacteria, rather than focusing on VKOR and other reductase enzymes.

Even focus on warfarin might be questionable. Warfarin is an anti-coagulant, and thus would have unwanted side effects if used in as an antibacterial agent. Even a similar drug, optimized for MtbVKOR (and not the human enzyme), might still have warfarin-like side effects on a patient. Towards this point, MtbVKOR can substitute in the human for native VKOR (Tie 2011). What's more, warfarin has been shown to negatively interact with several antibacterial agents, such as rifampicin (Almog, 1988; Martins, 2013). In light of this, I must question why did the authors choose to search for anti-tuberculosis agents based upon warfarin? Most antibiotics were developed because they target processes or structures unique to bacteria. The ideal antibiotic would target bacteria, and bacterial pathogenesis (so as to not effect the natural flora) and have minimal side effects and human targets. For example, targeting trans-translation (as reported by Shin et al, which I have reviewed previously) makes more sense from this perspective.

Starting with a known human drug seems to me to be a poor choice. However, it is a telling one for other reasons. Warfarin is a popular, FDA approved anti-coagulant. Another, less commonly prescribed anti-coagulant, phenindione, was reported by the authors to have superior antibacterial properties. Why not then concentrate on phenindione for a drug search? Why is the attention focused on warfarin? The key to this answer may lie in the more adverse side effects with phenindione. Since warfarin is more popularly prescribed, and it is safer, the focus on warfarin seems to me to suggest that this drug, or something slightly improved, would make an effect anti-tuberculosis drug. A sort of "this-is-almost-ready-for-market" argument, which is superficially easier to do with warfarin rather than phenindione but in reality is not valid for either drug.

My speculation about the author's motivations and impulses behind how they have communicated the study may be wildy incorrect. I think they are worth consideration, however, and may serve as an example of research being done because it is 'hot' or 'sexy', rather than being a serious attempt at serving public interest and antibiotic development. It may be a subtle example of this, for sure, but it is important to point out when science is being done in a fashion that doesn't make sense for the people that are funding it. I may not be the first to see this paper through this prism, but I am probably the first to publicly acknowledge it. Others have either considered this, and chosen not to follow the lead of this paper (for example, warfarin and VKOR are passed over by Blumenthal et al. 2010), or instead of chosen to focus on their own 'hot' or 'sexy' research angle.

In this review, I do not wish to disparage the authors of this study in particular or the scientific enterprise in general. The current study definitely has merit (whether or not it is worthy of PNAS is a separate question). My only point (apart from technical criticisms) is that it has been overhyped and suggest further research down a relatively unproductive avenue. 

References:

Almog et al. Complex interaction of rifampin and warfarin. (1988) South Medical Journal. Vol 81 (10) pp. 1304-6 PMID: 3175736

Martins et al. Rifampicin-warfarin interaction leading to macroscopic hematuria: a case report and review of the literature. (2013) BMC Pharmacol Toxicology. Vol 14 (27) PMID: 23641931

Blumenthal et al. Simultaneous analysis of multiple Mycobacterium tuberculosis knockdown mutants in vitro and in vivo. (2010) PLoS One. Vol 5 (12) PMID: 21203517

Tie et al. Mycobacterium tuberculosis vitamin K epoxide reductase homologue supports vitamin K-dependent carboxylation in mammalian cells. (2011) Antioxid. Redox. Signal. Vol 16 (4) PMID: 21939388