The most compelling data pertain to Keshan disease, a classical Se-deficiency disease manifested as a non-obstructive cardiomyopathy. Chinese investigators suspected an infectious agent might be a cofactor, and eventually isolated coxsackievirus from the hearts of Keshan disease victims; the combination of the virus and Se deficiency produced cardiomyopathy in mice.⁶ Recently, Beck and coworkers have shown that in Se-deficient mice, even a normally nonvirulent strain of coxsackievirus B3 can produce myocarditis similar to that seen in Keshan disease⁷ (and references therein). Significantly, during passage through Se-deficient mice, the virus mutates into a more virulent strain that is pathogenic even in normal animals on Se-adequate diets.

Along similar lines, it is of considerable interest that Ziegler has pointed out a correlation between high rates of endemic Kaposi’s sarcoma (KS) in African subsistence farmers and geographic regions in Africa where the soils are of volcanic origin.⁸ These include regions surrounding the entire East African Rift Valley and the Nigeria-Cameroon border. It is widely documented that low Se levels in plants and Se deficiency syndromes of livestock are common in areas with soils of volcanic origin: the Rift Valley is a typical example. Furthermore, Se deficiency in humans has been specifically documented in northern Zaire (e.g.⁹). Since recent evidence strongly suggests that KS involves a novel herpesvirus, this association of KS in Africa with low Se areas suggests a possible analogy to Keshan disease and coxsackievirus. Significantly, we have also found very large ORFs with start codons and up to 11 in-frame UGA codons in herpesviruses like cytomegalovirus and Epstein Barr virus (reported at 8th ICAR, Taylor et al.⁵), suggesting that some herpesviruses may also be “Se-dependent”.

Rather than being indirect (e.g. involving a nonspecific antioxidant effect), the possibility that some antiviral effects of Se might involve virally-encoded selenoproteins has apparently not been considered until very recently.^3,5 Even though first demonstrated about ten years ago, it has still has not become widely appreciated that SeC can be encoded by the UGA codon, which usually serves as a stop codon in the genetic code. Conventional analyses of potential protein coding regions in genes still do not usually discriminate UGA from the other two stop codons, and thus they fail to reveal that proteins might be encoded in regions containing UGA codons. Such regions are routinely assumed to be inactive due to the presence of stop codons, which is probably true in the vast majority of cases, because efficient SeC incorporation is only possible when the mRNA contains a cis-acting signal known as a SeC insertion sequence.¹⁰ However, as shown in Figure 2, such consensus SeC insertion sequences capable of forming the required characteristic stem-loop RNA structures are present in several Ebola mRNAs that also encode UGA-rich ORFs.

We recently reported a similar potential for selenoproteins to be encoded in HIV-1^3,5,11 and in coxsackievirus B3,¹¹ in regions overlapping known genes. In both cases, the link between Se deficiency and the associated viral diseases (AIDS and viral myocarditis, respectively) is strongly supported by an extensive body of literature (reviewed in^2,3,7,12).

In the Ebola virus genome (Zaire strain), there are several ORFs with highly significant clusters of inframe UGA codons. Overlapping the major nucleoprotein (NP) gene, there are two ORFs in the -1 reading frame, containing 17 and 11 UGA codons, respectively (Figure 1). The first ORF has excellent potential to be expressed by a ribosomal frameshift from the NP coding region, due to the presence of an “ideal” heptameric shift sequence and an RNA pseudoknot (PK) 8 bases downstream (Figure 3A). This frameshift site comes very near the beginning of the ORF, and could permit the formation of a fusion protein consisting of the N-terminal 314 residues of NP fused to a 181 residue C-terminal module potentially containing 16 SeC residues, encoded in the -1

Potential Selenoprotein Genes in Ebola Virus

Potential Selenoprotein Genes in Ebola Virus

from an edited or spliced RNA. There is a definite splice acceptor site very near the beginning of the ORF, a CAGA sequence preceded by a pyrimidine rich sequence and an upstream “CURAY” sequence (CUGAC). There are various potential splice donors in the large NP mRNA that could bring this region inframe to the main NP ORF or the upstream selenoprotein ORF with 16 UGAs. Since there are no reports of Ebola replication and transcription in the nuclei of infected cells, this ORF and the associated potential splice sites may be mere artifacts. However, Borna disease virus provides a precedent for nuclear replication/transcription and RNA splicing of a negative non-segmented single stranded RNA virus.¹³ Thus, we cannot rule out the possibility that splicing of this Ebola mRNA could occur,

e.g. in the unknown “reservoir” species that is the natural host for Ebola virus. Furthermore, RNA editing can also bring such an “out of frame” ORF into frame, and RNA editing is known to occur in a number of viruses, including Ebola Zaire.

There are several additional potential selenoprotein ORFs overlapping the first 6 genes of Ebola virus, including the vp24, vp30, vp35 and vp40 regions, all of which have potential SeC insertion sequences in their mRNAs (shown for vp30 and vp35 in Figure 2). Because the sequence has not yet been released, we are unable to report on the polymerase coding region, where we have consistently found potential overlapping selenoprotein genes in a number of other viruses. On the Ebola minus strand genomic RNA there are also potential SeC insertion sequences and several UGA-rich ORFs (up to 9 UGAs), some with start codons, and some potentially expressed from spliced genomic RNAs. On both plus and minus strands, some of these potential genes have start codons in the context of Kozak-like sequences, suggesting they may be programmed to bind ribosomes and initiate protein synthesis. All these data will be presented in detail in a subsequent publication.

If viruses like HIV-1, coxsackievirus B3 and Ebola do encode selenoproteins, why does all the evidence suggest that dietary Se inhibits viral replication, whereas Se deficiency triggers replication? Why would Se not “feed” the virus? The answer must lie in how viruses use Se.

As discussed previously,³ due to the inefficiency of frameshifting and SeC insertion mechanisms, these hypothetical viral selenoproteins could only be formed in very small amounts. Thus, in most cases they are not likely to be major structural proteins; some might have regulatory roles, acting in the midphase of the life cycle, and might not even be packaged in virions. If even one such selenoprotein were involved in negative feedback on replication (a repressor type function), decreased levels of that protein would provide the virus a way to respond to low Se levels by leaving the cell in search of a new host. By such a mechanism, the virus could satisfy a basal dependence on Se by escaping from a cell where Se levels had become dangerously low.

Since Se is an essential antioxidant, critical as a component of glutathione peroxidase in blood cells and liver cells (the very cell types that Ebola and many other viruses prefer to infect), very low Se levels are potentially associated with oxidative stress, lipid peroxidation and cell death. Thus, viral survival might be enhanced by the stimulation of replication under low Se conditions.

At the same time, host/viral competition for a limited amount of Se - particularly in a malnourished host - could significantly contribute to pathogenesis. This could be particularly acute with Ebola virus, due to the unprecedented high Se requirement implicit in the ORF with 16 UGA codons.

Dietary Se is also known to have immunopotentiating effects (reviewed in¹²). Thus, in addition to any direct effects exerted via (hypothetical) viral selenoproteins, Se deficiency can also weaken the immune system’s ability to fight viral infection, permitting increased replication, rapid mutation, and facilitating the emergence of more virulent strains, as Beck et al. suggest in the case of coxsackievirus.⁷

Given the unique dependence of selenoprotein genes upon a trace nutrient whose availability varies widely in geographical areas and host populations, the presence and activity of such genes would very likely be strain specific, as we suggested for coxsackievirus.¹¹ This could help explain why some viral strains can be very virulent, even when significant subsets of indigenous populations have antibodies to a similar, presumably much less virulent viral strain, which appears to be true even for filoviruses.^14,15

Certainly, prolonged Se deficiency in a host population could eventually lead to the inactivation and loss of any viral seleno-protein genes. Whether that loss would lead to more virulent strains, or whether those strains might undergo a compensatory attenuation by passage through the host population, would be difficult to predict.

However, in the case of Ebola virus, there is some reason to think that the presence of a gene with an exceptionally high Se demand could be a factor in the pathogenicity of specific viral strains. This is supported by the striking observation that in the Ebola Reston strain, which was devoid of pathogenicity in the 3 humans that were infected, there is no equivalent to the major potential selenoprotein gene overlapping the NP gene in Ebola Zaire (ORF1 in Figure 1). In the Ebola Reston NP mRNA, the UGA-rich ORFs are disrupted by non-UGA stop codons, there are fewer UGA codons, no analogous frameshift sites or PKs, and no SECIS element in the 3'-UTR. Thus there is no way that this potential selenoprotein gene could be expressed in Ebola Reston. This is a definite major difference at the gene level between these strains, which have previously been considered to be very close genetically. This potential NP-associated selenoprotein gene is also absent in Marburg virus, which also has a lower mortality rate than Ebola Zaire.

Since the hypothetical selenoprotein overlapping the Ebola Zaire NP gene could only be expressed as an NP fusion protein, it is possible that it could be formed as an NP variant comprising as much as a few percent of the total NP present in virions (more likely a fraction of a percent), in which case it might be possible to detect selenium in Ebola Zaire virions. This percentage would be expected to decrease in late infections if cellular stores of SeC became depleted. In regard to the possible function of such a viral selenoprotein, it is tempting to speculate that it might provide some type of antioxidant protection to the Ebola virions in a rapidly degenerating cellular environment.

Ebola is classified as a “hemorrhagic fever” virus, and produces the characteristic hemorrhaging due to the formation of blood clots (“disseminated intravascular coagulation”), leading to the obstruction and rupture of small blood capillaries. For this reason, counterintuitively, the anticoagulant drug heparin has been used to reduce the bleeding in Ebola patients.

It is very well documented that Se plays a significant role in the regulation of blood clotting via its effects on the thromboxane/ prostacyclin ratio. Se has an anti-clotting effect, whereas Se deficiency has a pro-clotting or thrombotic effect.¹⁶ Se deficiency has been associated with thrombosis and even hemorrhaging, which has been documented in a number of animals with severe Se deficiency (often artificially induced), but is almost never seen in humans, probably because such an extreme Se deficiency is rarely attained due to the diversity of human diets.

Thus, the possibility that a rapid depletion of Se due to the formation of viral selenoproteins could be a factor contributing to the severity of the hemorrhagic symptoms is mechanistically very feasible. Our analysis suggests that severe Ebola infections could produce an artificial and extreme Se depletion, resulting in extensive cellular damage due to lipid peroxidation, combined with enhanced thrombosis. This could also contribute to the associated immune deficiency that has been observed in Ebola infections.

To our knowledge, indicators of Se status and lipid peroxidation have never been examined in Ebola patients. However, Se has apparently been used with great success by the Chinese in the palliative treatment of an infectious hemorrhagic fever.¹⁷ Although this did not involve Ebola virus, there are a number of different hemorrhagic fever viruses, and they may share common mechanisms. This example provides yet another reason to expect that pharmacological doses of Se may also have some benefit in Ebola infections.

In the light of the extensive data on the antiviral effects of Se, the association between coxsackievirus and Keshan disease, and the geographic correlation for KS proposed by Ziegler, it is certainly intriguing that a number of emerging viruses have emanated from these same regions of Africa, that are potentially low in Se. By providing compelling theoretical evidence for the existence of selenoprotein genes in a number of viruses, now including Ebola virus, we have attempted to provide a unifying theoretical model to explain some of these disparate observations.

Taken as a whole, these observations and theoretical findings suggest the basis for a new paradigm in antiviral chemotherapy: the use of nutritional factors to alter the dynamics of the virus-host interaction so as to reestablish a balance in which the natural host defenses can be more effective. In essence, this is the fundamental concept of orthomolecular medicine, so perhaps this is not such a “new” paradigm after all. What is new and exciting is that this simple concept may be more widely applicable, to more virulent viral diseases, and a broader range of vitamins and minerals - in this case Se -than previously thought possible.

Finally, because SO2 reacts with Se compounds in soil, making it more difficult for plants to absorb, it has long been suspected that fossil fuel burning and acid rain may be contributing to a gradual decrease of Se in the food chain.¹⁸ Thus, like deforestation in jungles and rain forests, the resulting alterations in global Se cycling and distribution may be yet another example of how human activity possibly contributes to the emergence of new viral diseases. Ultimately, it is only a deeper understanding of the impact of these human activities on both microbes and their hosts that will empower us to rectify the resulting imbalances in our shared ecosystem.