RNA viruses vs. DNA synthesis: a general viral strategy 1 that may contribute to the protective antiviral effects of selenium

9 The biosynthesis of DNA inherently competes with RNA synthesis because it depends on the reduction 10 of ribonucleotides (RNA precursors) to 2’-deoxyribonucleotides by ribonucleotide reductase (RNR). 11 Hence, RNA viruses can increase viral RNA production in cells by partially blocking the synthesis of 12 DNA, e.g. by downregulating the mammalian selenoprotein thioredoxin reductase (TR), which 13 normally acts to sustain DNA synthesis by regenerating reduced thioredoxin, a hydrogen donor for 14 RNR. Computational and preliminary experimental evidence supports the hypothesis that a number of 15 pathogenic RNA viruses, including HIV-1, Ebola, Zika, some flu viruses, and SARS-CoV-2, target TR 16 isoforms by antisense. TR knockdown would create a host antioxidant defect that could be partially 17 rectified by increased selenium intake, or be exacerbated by selenium deficiency, contributing to viral 18 pathogenesis. There are several non-selenium-dependent means that viruses might also exploit to slow 19 DNA synthesis, such as targeting RNR itself, or components of the glutaredoxin system, which serves 20 as a backup redox system for RNR. HIV-1 substantially downregulates glutathione synthesis, so it 21 interferes with both the thioredoxin and glutaredoxin systems. Computational results suggest that, like 22 Ebola, SARS-CoV-2 targets TR3 by antisense. TR3 is the only TR isoform that includes an N-terminal 23 glutaredoxin domain, so antisense knockdown of TR3 may also affect both redox systems, favoring 24 RNA synthesis. In contrast, some DNA viruses encode their own glutaredoxins, thioredoxin-like 25 proteins and even RNR homologues – so they are doing just the opposite, favoring DNA synthesis. 26 This is clear evidence that viruses can benefit from shifting the RNA:DNA balance to their advantage. 27


Introduction 28
It is not a coincidence that the vast majority of the most notorious emerging and pandemic viruses, 29 from the coronaviruses that cause to Ebola,HIV,avian influenza,Zika,30 Dengue, West Nile, Chikungunya, yellow fever, Eastern Equine Encephalitis, Norvirus, Nipah and 31 Hantaviruses, as well as the less exotic measles, mumps, hepatitis viruses A and C, common cold and 32 enteroviruses, and many more, all have RNA genomes. DNA viruses such as herpes viruses, 33 adenoviruses and papillomavirus can cause very serious disease, but other than smallpox, DNA viruses 34 have not historically been associated with mass pandemics that can cause deaths in the millions. Nor 35 do they (or other potential pathogens like bacteria, fungi and parasites) mutate anywhere near as fast 36 as RNA viruses [1], so they tend to be more genetically stable, rather than a moving target for vaccine 37 and antiviral drug design. 38 This is a provisional file, not the final typeset article Thus, among the viruses, RNA viruses appear to be particularly well suited as agents of new emerging 39 virus outbreaks and global pandemics, because of several unique characteristics that enable rapid 40 adaptation. First, their very small genome size (typically between 10 and 30 thousand nucleotides) 41 allows for fast replication, easily attaining multiple generations within a 24 hour period [2]. Second,42 their RNA polymerases are highly error-prone, due to lack of proof-reading ability (with a few notable 43 exception like the nsp14 3′-5′ exoribonuclease of coronaviruses and other Nidovirales), so that their 44 mutation rate is not only many orders of magnitude higher (~10 6 ) than host DNA-based genomes, but 45 is also substantially higher (100-fold or more) than typical DNA viruses [2,3]. This accelerated 46 evolutionary capability enables them to adapt following species transfer, in order to optimize the 47 required host receptor tropism to attain a foothold in the new host population. It also enhances their 48 ability to continuously evade immune surveillance, as illustrated by the need for the production of new 49 seasonal flu vaccines every year. These considerations and more have been succinctly reviewed by 50 Carrasco-Hernandez et al [ between various animal species with varying degrees of ease or difficulty, without the need for a blood-58 eating insect as an intermediary. The frequency of such inter-animal transmissions is much higher for 59 RNA viruses than for DNA viruses [4]. 60 If the greatest zoonotic and pandemic threats we face are from RNA viruses, to fully understand their 61 pathogenic mechanisms and possible ways to reduce the severity of their impact, we must seek to 62 understand the modi operandi that they have developed as a consequence of their fundamental 63 characteristics as RNA viruses. Of these, none is more fundamental than the simple fact that RNA 64 viruses need the cells they infect to make RNA in copious amounts, to enable the formation of as many 65 viral progeny as the system can bear. Herein, perhaps, lies a vulnerability. 66 2 DNA biosynthesis depletes the pool of RNA precursors: a critical role for selenium 67 Although new evidence may offer alternatives to the RNA World Hypothesis [5], which posits that 68 DNA evolved later than RNA [6], the fact remains that for all life on earth, DNA biosynthesis is an 69 add-on to RNA biochemistry, so that 2'-deoxyribonucleotides can only be made from ribonucleotides. 70 Hence, DNA synthesis inevitably depletes the pool of ribonucleotide precursors that an RNA virus 71 would need for copying its RNA for new virus production. This means that RNA viruses can increase 72 viral RNA production by partially blocking the synthesis of DNA. There are various ways that they 73 could manage to do that, most of which may be utilized to a varying extent by different RNA viruses. 74 But one of the best ways to slow DNA synthesis involves selenium, and that is the focus of this 75 commentary, as it can help to explain a lot of previous observations about RNA viruses and selenium. 76 The thioredoxin system is a key redox cycle involved in the reduction of ribose to deoxyribose, in 77 which thioredoxin serves as a hydrogen donor for ribonucleotide reductase (RNR). To sustain that 78 redox cycle, thioredoxin reductase (TR), a selenium-containing enzyme in mammals, is essential. 79 Hence, TR is a perfect target for an RNA virus to slow down DNA synthesis. Specifically, antisense 80 targeting of TR isoforms would be an elegant way for an RNA virus to partially inhibit DNA synthesis 81 to enhance viral RNA synthesis, so that there will be more RNA to make into new viruses. As an 82

RNA viruses vs. DNA synthesis
3 essential component of TR, selenium thus could be considered a natural antagonist of RNA viruses,83 which casts a new light on an extensive body of literature linking selenium status to the incidence, 84 morbidity and mortality of a number of RNA viral infections (as reviewed, [7][8][9]). 85

3
The role of selenium in COVID-19 follows a pattern seen with many RNA viruses 86 The recent demonstration by Zhang et al. of a highly significant association between the outcome of 87 SARS-CoV-2 (SCoV2) infection and previously documented regional selenium (Se) status in Chinese 88 cities [10] is just the latest example of a role for selenium that has been reported for a variety of RNA 89 viruses and reverse transcribing viruses with an RNA stage (HIV-1 and Hepatitis B virus) going back 90 four decades. That these cases form a consistent pattern for the involvement of selenium in the 91 incidence, progression or outcome of a variety of viral infections is attested by the fact that over the 92 last several decades, this phenomenon has been the subject of a considerable number of independent 93 reviews, of which I will cite only a few of the most recent [7][8][9]. 94 In some cases, selenium compounds have been found to have direct antiviral activity either in cell 95 culture (e.g., for influenza and oncogenic retroviruses [11,12]) or in an animal model (e.g., mouse 96 mammary tumor virus, coxsackievirus and influenza [13][14][15]), or a clinical benefit in a human viral 97 disease, e.g. HIV-1 (as reviewed in [9]) and epidemic hemorrhagic fever linked to hantavirus infection 98 [16]. In other examples, the frequency of cases of infection, viral pathogenicity or disease progression 99 has been found to be associated with either low Se status in patients (HIV-1, influenza), or with a 100 geographic area in which Se deficiency was endemic due to low soil Se content (Coxsackievirus,  101 hepatitis B and hantavirus), as reviewed by various authors [7][8][9]17]. For the viral infections in each of 102 the latter examples, the increased mortality risk associated with low selenium status or reduced intake 103 in the affected geographic region was significantly reduced by selenium supplementation in every case. 104

4
The discovery and significance of regions of antisense complementarity between RNA 105 virus mRNAs and host mRNAs encoding isoforms of thioredoxin reductase (TR) 106 As my group first reported in regard to HIV-1 and the Zaire Ebolavirus (EBOV) [17], and later for Zika 107 [18], the possibility that those RNA viruses target thioredoxin reductases (TR) by antisense is supported 108 by computational RNA:RNA hybridization results and preliminary experimental data, in the form of 109 gel shift assays with DNA oligonucleotides. We initially discovered those interaction sites in HIV-1 110 and EBOV because in both cases they were proximal to highly conserved UGA stop codons 111 (potentially encoding selenocysteine) that terminate the HIV-1 nef and EBOV nucleoprotein open 112 reading frames. Although years earlier we had identified (by sequence analysis), cloned and expressed 113 an HIV-1 encoded frameshift variant of the viral gp120 envelope protein and showed that it encoded a 114 functional glutathione peroxidase (GPx, the prototypical selenoprotein), we had to incorporate a 115 mammalian selenocysteine insertion sequence (SECIS) element in the construct in order to express the 116 viral GPx as a selenoprotein [19]. We were never able to identify a functional SECIS element encoded 117 by an RNA virus. Thus, the discovery of the improbable juxtaposition of a highly conserved viral UGA 118 codon with a nearby region of strong antisense complementarity to a host selenoprotein immediately 119 suggested a viral mechanism for capture, by "antisense tethering interactions" (ATI), of a host SECIS 120 element [17]. This mechanism could enable the recoding of the viral UGA stop codon as selenocysteine, 121 to form a low-abundance extended selenoprotein variant of the known viral protein. In retrospect this 122 is not at all surprising, because viruses contain only the barest elements of the machinery of life, 123 primarily what they need to get in and out of cells and to replicate their RNA or DNA; they hijack all 124 the cellular machinery for almost everything else. So it makes sense that HIV and EBOV might also 125 hijack SECIS elements. However, because that capture involved an antisense interaction, there is a 126 4 This is a provisional file, not the final typeset article direct implication that this could cause knockdown of host TR1 or TR3 levels as "collateral damage" 127 -but perhaps it isn't collateral damage at all, perhaps it is also deliberately benefiting the virus. And 128 the most obvious benefit would be via the role of TR in DNA synthesis. 129 We have now demonstrated selenium-dependent readthrough of both of those UGA codons, in HIV-1 130 nef and the EBOV nucleoprotein, and a role for TR1 in the mechanism in the case of nef, via GFP 131 reporter gene assays [20,21]. The fact that in database searches these and other RNA virus mRNAs 132 consistently show a preference for antisense targeting of TR over other viral selenoproteins like GPx 133 supports the supposition that the knockdown of the targeted TR isoforms likely to result from such 134 interactions might also benefit an RNA virus, via the role of TR in DNA synthesis [18]. Figure 1  This correlation between selenium status and the outcome of yet another RNA virus infection raises 149 the obvious question, could a similar mechanism involving antisense targeting of TR be at work in 150 SCoV2? As shown in Figure 2, a similar analysis identified two SCoV2 regions with antisense matches 151 to human TR3, both having 22 base pairs in a stretch of 23 or 24 nucleotides (equivalent to a high 152 affinity microRNA interaction), with each having only one GU base pair (which are common in RNA 153 helices). The first of these regions (Figure 2A), just before base 5000 in the coronavirus genome, is 154 particularly significant, because it is proximal to a predicted -1 ribosomal frameshift site leading to a 155 region with a single in-frame UGA (potential selenocysteine) codon that is only a few hundred bases 156 upstream from the anti-TR3 antisense site, in the SCoV2 genome of almost 30,000 nucleotides 157 (Supplementary Material Figure S1). Equally compelling is the fact that the targeted site around base 158 2100 in the human TR3 mRNA is in its 3′-UTR, only 150 bases from the SECIS element that enables 159 the recoding of UGA as selenocysteine; capture of this element is thus a likely factor driving the 160 evolution of this interaction. All of these features were found to be completely conserved in a set of 161 almost 1000 SCoV2 isolates available in Genbank and included in a search on 5-14-2020, with the 162 exception of a few viral isolates which proved to have single-base sequencing misreads (e.g. N rather 163 than A,T,C or G) within this region, contributing to a slightly lower alignment score. Thus, in addition 164 to predicting the knockdown of TR3 mRNA and/or protein levels in SCoV2 infected cells, this example 165 perfectly fulfils the requirements for the viral selenoprotein expression mechanism we proposed for 166 HIV-1 nef and the EBOV nucleoprotein: a >20 base long antisense match to a TR isoform within a few 167 hundred bases or less of an accessible in-frame UGA codon [17]. In HIV-1 and EBOV, the nearby UGA 168 codon was accessible as the stop codon of a known gene, enabling an extended protein variant; in 169 SCoV2, the potential coding UGA is accessed via a programmed ribosomal frameshift that was 170 identified by an unbiased algorithm ( Figure S1). The targeting of the TR3 isoform by SCoV2 is similar 171 5 to what we reported for EBOV, and is also what is computationally predicted for mumps virus ( Figure  172 1), whereas HIV-1, influenza and Zika all preferentially target TR1 (Figure 1). 173 TR3 is sometimes called the "testicular" form of TR, because that tissue is where TR3 mRNA levels 174 are highest. But according to the Human Protein Atlas [22], even though mRNA levels are highest in 175 the testes, TR3 protein levels are as high or higher in the lung and GI tract, which are major sites of 176

RNA viruses vs. DNA synthesis
SCoV2 replication. The Atlas data also show that the ACE2 receptor used by SCoV2 is expressed at 177 high levels in the testes. Significantly, testicular mumps infection has long been known to be a potential 178 complication in males, and in the 2014 EBOV outbreak, cases of persistent EBOV infection of the 179 testes were identified in patients presumed to have recovered [23]. Because of the high levels of ACE2 180 receptor there, SCoV2 could also target the testes. So all three of these TR3-targeting viruses appear 181 to at least have the potential to infect the tissue in which TR3 is most highly expressed in human males. 182 6 The glutaredoxin system and non-selenium dependent inhibition of DNA synthesis 183 The thioredoxin system seems particularly critical for DNA synthesis in certain cell types and 184 conditions, such as during T cell proliferation [24]. But there is a backup system for DNA synthesis, 185 the glutaredoxin system, which uses glutathione rather than thioredoxin as its hydrogen/electron donor 186 [25]. Significantly, TR3 is unique among TR isoforms in that it contains an N-terminal glutaredoxin 187 domain, so it can function in both the thioredoxin and glutaredoxin systems to sustain DNA synthesis. 188 Thus, antisense-mediated knockdown of TR3 could be an effective general strategy for RNA viruses 189 because of its ability to partially interfere with both redox systems that provide electrons to RNR for 190 reduction of ribonucleotides. 191 The glutaredoxin system is one of the various non-selenium dependent means mentioned earlier (i.e., 192 not involving TR isoforms), by which an RNA virus could slow down DNA synthesis. Antisense 193 targeting of RNR subunits, or glutaredoxin isoforms, or enzymes involved in glutathione synthesis, 194 could all potentially achieve a similar goal, alone or in combination with anti-TR based mechanisms. 195 Possible examples of these can be found, one of the most convincing being the inhibition of glutathione 196 synthesis by HIV-1, which would inhibit the ability of the glutaredoxin system to provide electrons to 197 RNR. There is an extensive body of evidence dating to the mid-1980s of a progressive deficit of 198 reduced glutathione (GSH) in AIDS patients (reviewed in section 2.1.2. of [26]), and real-time PCR 199 analysis has shown an 89% knockdown of glutathione synthetase (GSS) in HIV-1 infected 200 macrophages [27]. This may be driven by antisense targeting of GSS mRNA by HIV-1, as suggested 201 by the antisense BLAST hit shown as Figure S2A. Thus HIV-1 may be an example of simultaneous 202 interference in both the thioredoxin system (by TR1 knockdown) and the glutaredoxin system (by GSS 203 knockdown). Simultaneous blockade of both redox systems may prove to be necessary in order to 204 significantly favor RNA synthesis. Significantly, the very large genome size of some DNA viruses, 205 particularly poxviruses, affords them the luxury of encoding their own glutaredoxins, thioredoxin-like 206 proteins, and even RNR homologues [28], which serve in part to facilitate viral DNA synthesis, as well 207 as thiol reduction for viral assembly and other purposes. That pretty much proves the case that viruses 208 can benefit by shifting the RNA:DNA balance in their favor, and that a variety of mechanisms could 209 be used to achieve this goal. 210 In regard to the possible antisense targeting of glutaredoxins by RNA viruses, some of the strongest 211 identifiable matches are between regions of glutaredoxin-2 (GLRX2) and respiratory syncytial viruses 212 (also known as orthopneumovirus Subgroup A), as well as GLRX2 vs. Eastern Equine Encephalitis 213 Virus (EEEV), shown in Figure S2 B-D. It is more difficult to find good examples of potential viral 214 antisense targeting of RNR, which if it exists seems much less common, and the potential interactions 215 6 This is a provisional file, not the final typeset article less convincing. One possible explanation for this is that, since there is no backup enzyme for RNR, 216 its knockdown could risk shutting down essential DNA repair processes. 217 Overall, TR isoforms may be ideal targets for RNA viruses because on the one hand, the thioredoxin 218 system appears to be the predominant electron donor for RNR, particularly in the cell cycle S phase 219 [25], but even if TR1 was totally blocked, the glutaredoxin system assures a basal level of DNA 220 synthesis that may be necessary for continued cell viability. And if the viral agenda also includes the 221 expression of its own selenoprotein module, such as a viral GPx [19,29], antisense targeting of TR 222 isoforms is an ideal choice, because it achieves 2 goals simultaneously, by TR knockdown to increase 223 RNA synthesis, while simultaneously exploiting the ATI mechanism for SECIS capture [17]. This 224 would be very typical of how viruses operate, to do more with less, by encoding multifunctional RNAs 225 and proteins. 226

Discussion and conclusions 227
Given the diversity of viruses and possible mechanisms, it is clear that some RNA viruses may interfere 228 in selenium-based mechanisms more than others, and there could even be significant variation in this 229 regard between different subytpes and strains of a given virus. For example, the predicted anti-TR1 230 interaction shown for a bird flu strain in Figure 1 is an exceptionally strong interaction, not seen at that 231 level of significance for other common strains of influenza A. However, selenium status has been 232 linked in various ways to influenza virus pathogenicity, as recently reviewed [7,9], so the potential role 233 of anti-TR1 interactions in the pathogenesis of influenza merits further investigation. In regard to 234 expected knockdown of TR isoforms by the antisense mechanism, this may occur at the protein level 235 without visible changes in TR mRNA levels. As discussed previously, based on precedents from 236 microRNAs, inhibition of protein synthesis without degradation of the targeted mRNA is actually the 237 expected result if the RNA:RNA base pairing is imperfect, i.e., with more gaps and bulges [30]. 238 However, if the base pairing is almost perfectly continuous, like those predicted for SCoV2 vs. TR3 in 239 Figure 2, it is more likely that knockdown may be observed at both the mRNA and protein levels. But 240 if there are typical structural irregularities in stem regions of the RNA:RNA interaction (as seen for 241 HIV-1:TR-1 in Figure 1), a failure to observe mRNA knockdown via qRT-PCR or microarray does not 242 necessarily rule out this mechanism. This point is validated by the fact that cellular levels of TR1 243 protein are in fact substantially decreased in HIV-1 infected cells [31] (consistent with our antisense 244 results, Figure 1 and ref. [17]), but TR1 is not a gene that has been reported to be downregulated by 245 HIV-1 at the mRNA level in microarray studies. So this may be a case of antisense disruption of protein 246 synthesis primarily at the ribosomal level. 247 To summarize the major theme of RNA viruses vs. DNA synthesis as it relates to selenium, the central 248 basis is that in mammals, TR enzymes are selenoproteins, so selenium is an essential component of 249 TR; hence, as part of the thioredoxin system, selenium plays an important role in the eternal 250 competition between DNA and RNA synthesis. This implies that, even in the absence of specific 251 antisense or other targeting of TR by an RNA virus, a more universal sensitivity to selenium status 252 could still exist for this class of viruses. Under conditions of selenium deficiency sufficient to 253 substantially decrease TR protein levels, DNA synthesis may be at least somewhat disfavored, 254 conferring an advantage to RNA viruses. The converse may also be true -that a more replete selenium 255 status may tend to enhance DNA synthesis, creating less favorable conditions for RNA viral replication 256 by depletion of ribonucleotides, thereby providing a protective antiviral benefit. 257 It should be emphasized, however, that there are a multitude of possible mechanisms by which 258 selenium can influence viral infections, involving both host and viral factors; this just happens to be 259 7 one that particularly applies to RNA viruses as a class. For example, the importance of selenium to the 260 immune system has been reviewed many times (recently, here [32]), and there are specific roles of 261 selenium in human biology that may be relevant to the symptomatology of certain viral infections, e.g. 262 a role in blood clotting, that could be relevant for observed thrombosis in COVID-19, as well as in 263 viral hemorrhagic fevers [9]. The recent identification of human GPx1 as a possible binding partner for 264 the SCoV2 M pro protease [33] raises the possibility of host selenoprotein knockdown by proteolysis. 265 Consistent with that possibility, remarkably, there is an instance of an exact match to the SCoV2 M pro 266 protease cleavage consensus sequence LQ/A near the very C-terminal of human TR1, which could 267 enable M pro to clip off 5 amino acids including the C-terminal redox center of TR1, with the catalytic 268 selenocysteine in the penultimate position. Thus, we may have instances of targeting by SCoV2 of two 269 different isoforms of TR, one by proteolysis (TR1) and one via antisense knockdown (TR3). But the 270 common theme is direct viral interference with the host selenoproteome. 271 In conclusion, considering the new evidence for a significant correlation between selenium status and 272 reported COVID-19 outcomes [10], and computational evidence presented here for antisense targeting 273 of human TR3 mRNA by SCoV2 (Figure 2)   These include previously published interactions for the EBOV nucleoprotein mRNA vs TR3, the HIV-386 1 nef 3′ region vs TR1, and Zika mRNA vs. TR1 [17,18]. The asterisk indicates the 3′-UGA stop codon 387 of HIV-1 nef, where selenium-dependent readthrough occurs [21]. Additional predicted interactions 388 with either TR1 or TR3 are shown for a strain of avian influenza and mumps virus. All of these 389 interactions were initially identified as DNA/DNA +/-matches using BLAST, then confirmed at the 390 RNA level using the RNAHybrid program [34], and finally confirmed to be sufficiently strong as to 391 overcome internal folding energies of the individual RNA strands using the IntaRNA program [35], as 392 described previously [17]. The Genbank accession numbers and regions for the sequence fragments 393 shown are given in the relevant references, the others are: Bird flu vs human TR1: the antisense match 394 is between the genomic negative sense strand of H9N2