Enisamium is an inhibitor of the SARS-CoV-2 RNA polymerase and shows improvement of recovery in COVID-19 patients in an interim analysis of a clinical trial

Pandemic SARS-CoV-2 causes a mild to severe respiratory disease called Coronavirus Disease 2019 (COVID-19). Control of SARS-CoV-2 spread will depend on vaccine-induced or naturally acquired protective herd immunity. Until then, antiviral strategies are needed to manage COVID-19, but approved antiviral treatments, such as remdesivir, can only be delivered intravenously. Enisamium (laboratory code FAV00A, trade name Amizon®) is an orally active inhibitor of influenza A and B viruses in cell culture and clinically approved in countries of the Commonwealth of Independent States. Here we show that enisamium can inhibit SARS-CoV-2 infections in NHBE and Caco-2 cells. In vitro, the previously identified enisamium metabolite VR17–04 directly inhibits the activity of the SARS-CoV-2 RNA polymerase. Docking and molecular dynamics simulations suggest that VR17–04 prevents GTP and UTP incorporation. To confirm enisamium’s antiviral properties, we conducted a double-blind, randomized, placebo-controlled trial in adult, hospitalized COVID-19 patients, which needed medical care either with or without supplementary oxygen. Patients received either enisamium (500 mg per dose) or placebo for 7 days. A pre-planned interim analysis showed in the subgroup of patients needing supplementary oxygen (n = 77) in the enisamium group a mean recovery time of 11.1 days, compared to 13.9 days for the placebo group (log-rank test; p=0.0259). No significant difference was found for all patients (n = 373) or those only needing medical care (n = 296). These results thus suggest that enisamium is an inhibitor of SARS-CoV-2 RNA synthesis and that enisamium treatment shortens the time to recovery for COVID-19 patients needing oxygen.


Introduction 73
Severe acute respiratory coronavirus 2 (SARS-CoV-2) is an important human 74 pathogen and the causative agent of COVID-19. Vaccines are available to prevent 75 the spread of SARS-CoV-2, and several antiviral strategies, such as treatment with 76 ratio of 2:2:1 to form a nsp12/7/8 complex. Next, we incubated the nsp12/7/8 141 complex with a hairpin template (Fig. 1F) in the presence of 0.5 mM of each 142 nucleotide triphosphate (NTP) and varying concentrations of enisamium or the 143 previously identified enisamium metabolite VR17-04. Enisamium inhibited nsp12/7/8 144 activity at relatively high concentrations, with an IC 50 of 26.3 mM (Fig. 1G). By 145 contrast, VR17-04 had an estimated IC 50 of 0.98 mM on the hairpin template (Fig.  146 1G). These IC 50 values are within an order of magnitude of those observed for the 147 inhibition of the influenza virus RNA polymerase (13). Moreover, the IC 50 value for 148 VR17-04 on SARS-CoV-2 nsp12/7/8 is similar to remdesivir triphosphate in the 149 presence of 0.5 mM NTPs in a comparable assay (14). A 10-fold reduction of the 150 NTP concentration in the assay lowered the VR17-04 IC 50 value to 0.029 mM (Fig. 151 1G), suggesting that VR17-04 is competing with NTP incorporation. Collectively, 152 these data suggest that RNA synthesis by the SARS-CoV-2 nsp12/7/8 complex is 153 inhibited by enisamium and its metabolite VR17-04 in vitro. Previous studies (13, 15) and our data in Fig. 1A, suggest that enisamium acts 158 through metabolite VR17-04 and inhibits the activity of the influenza A virus and 159 SARS-CoV-2 nsp12 RNA polymerases. However, the mechanism of RNA synthesis 160 inhibition by VR17-04 is not fully understood. We hypothesised that the additional 161 OH-group of VR17-04 could support hydrogen bond formation with adenine and 162 cytosine ( Fig. 2A), creating two hydrogen bonds in total, while enisamium would form 163 only one hydrogen bond with these bases. We expected that VR17-04 would not 164 form hydrogen bonds with guanine or uridine (Fig. S2), suggesting that the inhibitory 165 effect of VR17-04 would be dependent on the template sequence. 166 The CO and OH groups of VR17-04 can adopt trans and eclipsed 167 conformations that are characterized by three dihedral angles: φ 1 , φ 2 and φ 3 (Fig. 2B). 168 Dihedral angles φ 1 and φ 2 are also present in enisamium, but φ3 involves different 169 atoms (Fig. S4). In VR17-04, dihedral angle φ 3 is ~180° in the trans conformation 170 and ~0° in the eclipsed conformation (Fig. 2B). Based on quantum chemical 171 calculations, the eclipsed conformation has a lower energy than the trans 172 transformation (Table S1), and only the eclipsed conformation would be compatible 173 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 21, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 with cytosine or adenosine binding (Fig. 2B). To confirm that VR17-04 can adopt the 174 eclipsed conformation in solution, we measured the 1 H NOESY spectrum of VR17-175 04 in water and found a correlation between the protons 'HN' and 'H5' that is 176 compatible with a φ3 value of 0° (Fig. 2C). The chemical shifts of the selected 177 protons are reported in Table S2. Additionally, we observed a correlation between 178 the HN and the CH 2 , or the ortho aromatic protons of the Ph group (Fig. 2C). These 179 observations suggest that VR17-04 can adopt a conformation that would be 180 compatible with hydrogen bond formation with cytosine and adenine. 181 To further investigate whether VR17-04 can inhibit the activity of the SARS-182 CoV nsp12 RNA polymerases through base-pair interactions with the template, we 183 docked enisamium or VR17-04 into the SARS-CoV-2 nsp12/7/8 complex bound to 184 template RNA and remdesivir monosphosphate (PDB 7bv2, Fig. 2D). Prior to 185 docking, we removed the remdesivir monosphosphate from the complex, and used 186 in silico mutagenesis to change the uridine in the active site to cytosine or adenine. 187 After selection of the most significant poses, we found that both enisamium and 188 VR17-04 can be accommodated in the +1 position of the nucleotide binding pocket 189 ( Fig. 2E-G), in a position similar to remdesivir monosphosphate (Fig. 2D). VR17-04 190 was specifically coordinated through hydrogen bond interactions with the unpaired 191 cytosine residue of the template RNA. Nsp12 residues K545 and K555 were 192 predicted to preserve the VR17-04 position in the catalytic cavity (Fig. S4). In 193 addition, our modelling suggests that VR17-04 can form a stacking interaction with 194 the -1 base of the nascent strand. By contrast, enisamium docked in the same +1 195 nascent strand position, but only formed one hydrogen bond with the cytosine in the 196 +1 template position (Fig. S4). 197 To estimate the binding stability of enisamium or VR17-04 in the nsp12 active 198 site, we performed molecular dynamics (MD) simulations of enisamium or VR17-04 199 in the nsp12 active site (see Material and Methods for specifics). Our MD simulations 200 predict that VR17-04 favourably binds the unpaired cytosine in the +1 position of 201 nsp12/7/8 complex, maintaining two hydrogen bonds (Watson-Crick base pair) as 202 seen by the preserved hydrogen bonds distances (Fig. 2 E-G) and the coplanarity 203 angle (Fig. 2H, I). By contrast, our simulations predict that enisamium binds less 204 stably to the nsp12/7/8 complex, based on the ~2-fold difference in distance between 205 enisamium and cytosine compared to VR17-04 and cytosine (Fig. 2E, F; Table S3), 206 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 21, 2021. ; https://doi.org/10.1101/2021.01.05.21249237 doi: medRxiv preprint and a higher estimated Poisson-Boltzmann free energy for the enisamium binding 207 (43.6 kcal/mol) compared to VR17-04 binding to either cytosine (-19.8 kcal/mol) or 208 adenine (-14.8 kcal/mol) (Table S4). 209 To investigate if VR17-04 inhibits the SARS-CoV nsp12/7/8 RNA polymerase 210 complex in a sequence-specific manner, we analysed the extension of a hairpin 211 template containing a cytosine or adenine in the +1 position of the template (Fig. 2J) 212 in the presence or absence of VR17-04. We found that at high VR17-04 213 concentrations, the nsp12/7/8 RNA polymerase complex was prevented from 214 incorporating GTP and UTP (Fig. 2J). When we subsequently moved the cytosine 215 residue to the +2 nt position of the template down and introduced a uridine at the +1 216 position, ATP was incorporated in the presence of VR17-04 but GTP was not ( Fig randomization. Randomized COVID-19 patients were treated with either placebo or 229 enisamium iodide for 7 days. The chosen primary endpoint of the trial was time-to-230 recovery, and recovery was defined as an improvement in the SR baseline status by 231 2 SR score values (e.g., a change from SR 4 to SR 6). 232 In accordance with the study protocol, an interim analysis on all patients in the 233 1 0 faster time-to-recovery with the enisamium-treated patients compared to the 241 placebo-treated for the whole period of recovery, starting 7 days after randomization 242 (Fig. 3D). The estimated median time-to-recovery was 13 days for the placebo-243 treated patients and 11 days for the enisamium-treated patients. The mean time-to-244 recovery was 13.9 days for the placebo-treated group and 11.1 days for the 245 enisamium-treated group. The log-rank test showed a significant advantage in 246 recovery time in favour of enisamium treatment at interim stage (P = 0.0259). The 247 maximum time-to-recovery was reached on day 21 for the enisamium-treated group. 248 Not all patients recovered in the placebo-treated group. For these patients, the 249 recovery time was displayed as 29 days (Fig. 3D). The rapid global spread of SARS-CoV-2 necessitates development of effective 258 therapeutic interventions, and the most promising short-term strategy is to repurpose 259 existing drugs. In this study we showed that enisamium, which is approved for use 260 against influenza in 11 countries, can inhibit SARS-CoV-2 RNA synthesis ( Fig. 1). 261 Moreover, we showed that enisamium and its metabolite, VR17-04, inhibit the RNA 262 synthesis activity of the SARS-CoV-2 nsp12/7/8 complex ( Fig. 1, 2). Molecular 263 dynamics analysis and in vitro activity assays suggest that VR17-04 prevents GTP 264 and UTP incorporation into the nascent RNA chain (2). 265 It was previously reported that enisamium inhibits the influenza A virus RNA 266 polymerase activity in vitro with a relatively high IC 50 value of 46.3 mM (13). This 267 inhibition was improved 55-fold by addition of a hydroxyl group in the compound 268 VR17-04 (13). We find that SARS-CoV-2 nsp12/7/8 complex activity is inhibited by 269 enisamium and VR17-04 with similar IC 50 values to the influenza A virus RNA 270 polymerase at similar NTP levels ( Fig. 1) (13). Furthermore, the IC 50 value for VR17-271 04 is similar to remdesivir triphosphate in a similar in vitro assay used (14). 272 Remdesivir triphosphate is the active metabolite of remdesivir, which has shown 273 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Our docking and molecular dynamics simulations suggest that VR17-04 can 276 bind a template cytosine or adenine base in the active site of the SARS-CoV-2 RNA 277 polymerase. This hypothesis is supported by our NOE experiments performed in 278 water, which indicate that VR17-04 can adopt an eclipsed conformation in solution. 279 Only the eclipsed conformation is compatible with the sequence-specific inhibition 280 we observed in our in vitro polymerase reactions. We cannot fully exclude the 281 possibility that VR17-04 can also bind bases in the nascent RNA strand. However, 282 we think that this alternative explanation does not adequately explain our data. 283 We observed that the inhibitory effect of enisamium was more pronounced in 284 NHBE cells than in Caco-2 cells, which is in line with previous influenza A virus 285 experiments (13, 15) and suggest that enisamium is more readily metabolised into 286

VR17-04 in primary bronchial epithelial cells compared to adenocarcinoma cells. 287
Interestingly, the interim phase III clinical trial results that we disclose here reveal a 288 faster improved recovery in COVID-19 patients needing supplementary oxygen, 289 strongly suggesting that our in vitro data are well-aligned with the data from the 290 clinical trial. We cannot exclude that enisamium or its metabolite has effects beyond 291 RNA polymerase inhibition and that these additional effects contribute to the COVID-292 19 patient recovery. 293 Overall, our results strongly suggest that enisamium metabolite VR17-04 294 inhibits RNA synthesis by the SARS-CoV-2 nsp12/7/8 complex. Together with the 295 interim phase III clinical trial findings that enisamium improved the recovery of 296 COVID-19 needing supplementary oxygen (SR 4) by more than 2 days, our 297 observations raise the possibility that enisamium could be used a viable therapeutic 298 option against SARS-CoV-2 infection. Moreover, unlike remdesivir, enisamium does 299 not require intravenous administration, which would be advantages for its use 300 outside of a hospital setting. Together with observations that enisamium can inhibit 301 other RNA virus infections, and DNA virus infections (13,15,17), these results here 302 suggest that it can act as broad-spectrum polymerase inhibitor. 303 304 305

SARS-CoV-2 infections 307
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

HCoV-NL63 NHBE infections 321
MatTek's EpiAirway System (MatTek; Ashland, MA) consisted of differentiated 322 NHBE cells that were cultured to form a multilayered, highly differentiated model that 323 closely resembles the epithelial tissue of the respiratory tract. The cells from a single 324 donor (No. 9831) were used for assay consistency. The apical surface of the cells 325 was exposed to a humidified 95% air/ 5% CO2 environment. The basolateral 326 medium was changed, and the mucin layer was washed every 24-48 hours. NHBE 327 cells were inoculated via the exposure of the apical side to HCoV-NL63. After 1 hour 328 of incubation with virus in a water-jacketed 37°C incubator with a 5% CO2 supply, 329 the viral inoculum was removed from the cells. The apical side of the cells was 330 washed once prior to infection. After viral inoculation, enisamium chloride or control 331 media were added to the apical side of the cells and the basal media compartment 332 and incubated with the cells for 1 hour. After a 1-hour incubation, the drug containing 333 media was removed from apical and basal chambers. Growth medium alone or 334 growth medium with enisamium chloride were added to the bottom chamber, and 335 cells were incubated for 48 hours. At the termination of the experiments, cells were 336 washed twice, then 1 ml Trizol (Invitrogen) was added to each well for RNA isolation. 337 Total RNA was isolated from cells using Trizol per the manufacturer's instruction. 338 Two-step RT-qPCR was performed using HCoV-NL63 N gene-specific primers/probe 339 (forward primer: CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Conformational characterization and geometry optimization 374
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Molecular docking 380
The docking simulation was performed using Autodock 4.2 software (21). The 381 geometry of the ligands was previously optimized by DFT B3LYP/6-31G*. The 382 geometry of the receptor RdRp was extracted from the PDB 7bv2 (22). The catalytic 383 site of the nsp12/7/8 complex include the template RNA strand, the nascent strand 384 RNA, the pyrophosphate moiety [O 3 P-O-PO 3 ] -4 (pyr) and two Mg +2 ions, whose 385 positions were included in 7BV2. The two Zn +2 ions that were co-crystallized in 7BV2 386 were also included in the models. The co-crystallized inhibitor remdesivir was 387 removed from the complex, while its position and the contacts with the template RNA 388 trough the unpaired uracil base in the +1 position were used to guide the molecular 389 docking. The unpaired uracil base was subsequently mutated in cytosine (Cyt) or 390 adenine (Ade) in Pymol 2.3.4 (Schrodinger LLC), generating two different 391 complexes, identified as nsp12/7/8(Cyt) and nsp12/7/8(Ade) respectively. Next, 392 Gasteiger charges were calculated for both ligands (VR17-04, enisamium) and the 393 receptor complexes, and used as parameters of the docking simulation (23). In the 394 docking simulation VR17-04 and enisamium were described by five and four 395 rotational degrees of freedom, respectively. The docking gridbox was built by 396 orthogonal hedges of length between 60 to 80 points. The gridbox centre was set to 397 the NH 2 -group of the target cytosine or adenine residues of the template RNA 398 strand, respectively, and further set-up to fit the space between the R555 and K545, CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 21, 2021. ; https://doi.org/10.1101/2021.01.05.21249237 doi: medRxiv preprint VR17-04 (or enisamium) and the unpaired cytosine or adenine. The second criterium 408 was based on the possibility to figure out interferences of the ligand with the catalytic 409 mechanism, for example by interaction with key residues of the nsp12/7/8 complex. 410 The selected poses were further ranked by preliminary MD simulation (approximately 411 20 ns) in explicit solvent, to predict the stability of the interaction, and to estimate the 412 Poisson-Boltzmann free energy of binding, this last property was used to obtain a 413 final selection and ranking of the poses. 414 415

Molecular dynamic simulation 416
Explicit solvent MD simulation were run using NAMD 2.12 (24) software, applying the 417 Amber force field (ff14SB) (25). The TIP3P (26)  docking) and pyr were optimized using the quantum chemical approach DFT 424 B3LYP/6-31G*; the corresponding partial charges were estimated fitting the 425 electrostatic potential that was calculated at level of theory B3LYP/6-31G*/RHF/6-426 31G*. This procedure is in accord to the standard required by the Amber force field. 427 The quantum chemistry software GAMESS (27)  MD simulations were run fixing the number of particle (N), the absolute temperature 438 (T), and the pressure (P) applied to the cell hedges. The absolute temperature was 439 300 K and maintained with a Lowe-Andersen thermostat, while the pressure on the 440 cell box hedges was set as P = 1.01325 bar and preserved by the Nosé-Hoover-441 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 21, 2021. ; https://doi.org/10.1101/2021.01.05.21249237 doi: medRxiv preprint Langevin piston algorithm. The first MD simulation stage was run to adjust the 442 simulation cell box density, allowing the relaxation of all the inter-molecule distances, 443 i.e., the solute-solute, solute-solvent, and solvent-solvent distances. The cell density 444 equilibration stage was run restraining the atoms of the solute to their initial position 445 (energy minimized geometry of the complex) applying a harmonic restraint. In this 446 stage of cell density equilibration, the water molecules were left free to move. The and OH belong to VR17-04, while the remaining NH 2 and N belong to the opposite 467 cytosine (or adenine). Alternatively, the enisamium-cytosine pair require the following 468 atoms CO-NH 2 -N-CH to define φ, in this case CO and CH belong to enisamium, 469 while NH 2 and N belong to the opposite cytosine. In fact, values of this angle around 470 '0' indicate that the 'base pair' contact between VR1704 the cytosine (or adenine) is 471 coplanar, a geometric condition favouring the Watson-Crick 'base-pair' interaction 472 between VR17-04 and the base. VMD 1.9.3 (28) was used for the MD simulation 473 trajectory visualization and image creation. 474 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Estimation of Poisson-Boltzmann free energy of binding 476
In a system that evolve in accord to the complex formation reaction (29) In eq. 3 the E MM correspond to the potential energy of the system, as described by 488 the force-field; G sol is the polar solvation energy, estimated by the Poisson-489 Boltzmann equation (30); G nopol is the non-polar solvation energy, estimated by the 490 solvent-accessible surface area, a method included in the MMPBSA. T is the 491 absolute equilibrium temperature, while S MM is the molecular entropy of the system. 492 The sum of the first three terms on the right-hand side of the eq. 3 is conveniently 493 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 21, 2021. ; https://doi.org/10.1101/2021.01.05.21249237 doi: medRxiv preprint 1 8 The diagnosis of COVID-19 was based on body temperatures of ≥ 37.8 °C and 508 laboratory confirmed RT-PCR SARS-CoV-2 test from pharyngeal swabs or sputum. 509 According to intent-to-treat (ITT) definition and the actual recruitment status, 373 510 patients from 15 Ukrainian hospitals could be included into the interim analysis. 511 The study was conducted in accordance with the Declaration of Helsinki, ICH-512 GCP and the national laws and regulations in Ukraine. All patients singed an 513 informed consent prior to study participation. The study was approved by the Ethics 514 Commission of the Regional Clinical Hospital of Ivano-Frankivsk Regional Council on 515 12.05.2020. The registration number of the study in ClinicalTrials.gov is 516 NCT04682873. 517 518

Treatment 519
Patients were randomized at a 1:1 ratio to receive either 500 mg enisamium iodide 520 or matching placebo 4 times daily every 6 h for 7 full days. Supporting treatment 521 could be given at the investigator's discretion. All patients treated with either placebo 522 or enisamium, having a laboratory-confirmed SARS-CoV-2 at randomization and 523 providing any valid efficacy data after initiation of treatment were included in the 524 Intent-to-Treat evaluation set (ITT). 525 526

Data Collection 527
Baseline clinical data of the patients were recorded in the electronic case report 528 forms including demographics, medical history, previous and ongoing medication, 529 body height and weight and results of physical examination. During the study up to 530 day 29 or until discharge, symptom severity rating, vital signs, assessment of 531 COVID-19 symptoms, safety lab results, concomitant medication and adverse events 532 were recorded. The patient's and investigator's judgment on safety and efficacy were 533 collected independently. RT-PCR tests on SARS-CoV-2 at defined intervals and at 534 the discretion of the investigator were carried out. 535 536

Primary Outcome 537
The primary outcome was measured as the time from the day of randomization (day 538 1) to an improvement of at least two score points (from the status at randomization) 539 on the severity rating (SR) scale in days. The SR scale (WHO scale) is defined as 1 540 (8) -Death; 2 (7, 6) -Hospitalized, on invasive mechanical ventilation or 541 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Statistical Analysis of Interim Data. 552
The unblinded interim analysis was carried out by an Independent Data Monitoring 553 Committee based on a pre-specified Charter. At this stage, the primary outcome was 554 pre-planned to be tested by using the two-sided log-rank test stratified by centre. 555 Because of the low patient number in some centres the stratification was omitted for 556 gaining primary interim results. The primary outcome was defined as an 557 improvement from baseline of at least two SR points and was evaluated for all ITT 558 patients at the interim stage as well as for subgroups with baseline score of SR 4 559 and 5 separately. The interim analysis used the promising zone approach according 560 to (31) that was implemented according to (32). For this method no adjustment of the 561 type-I error rate is needed. Conflict of interest 575 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 21, 2021.

610
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted January 21, 2021. ; https://doi.org/10.1101/2021.01.05.21249237 doi: medRxiv preprint 1 e  t  a  l  .  ,  E  n  i  s  a  m  i  u  m  r  e  d  u  c  e  s  i  n  f  l  u  e  n  z  a  v  i  r  u  s  s  h  e  d  d  i  n  g  a  n  d  i  m  p  r  o  v  e  s  p  a  t  i  e  n  t   611   r  e  c  o  v  e  r  y  b  y  i  n  h  i  b  i  t  i  n  g  v  i  r  a  l  R  N  A  p  o  l  y  m  e  r  a  s  e  a  c  t  i  v  i  t  y  .  M  e  d  R  x  i  v  (  2  0  2  0

643
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 21, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 21, 2021. ; https://doi.org/10.1101/2021.01.05.21249237 doi: medRxiv preprint