INTRODUCTION
Despite the approval of one intravenous (remdesivir) and two oral (molnupiravir and paxlovid) anti-SARS-CoV-2 drugs, the armamentarium against SARS-CoV-2 and related coronaviruses remains thin. Monoclonal antibodies have proven effective as prophylaxis and therapy (
1,
2) but must be given intravenously or subcutaneously in a clinical setting. Initially effective products have rapidly become irrelevant in the face of newly emerging viral variants (
3). While highly effective at preventing severe disease if given early during infection in outpatient settings (
4), remdesivir has been mostly given during hospitalization (
5,
6). Its use during early disease is limited due to its intravenous formulation. Molnupiravir, an oral drug, is 30% effective at preventing hospitalization when given early during infection (
7). Paxlovid, another oral drug, is highly efficacious (
8), but side effects, drug interactions (
9), and rebound of symptoms and viral shedding upon treatment cessation (
10,
11) can limit therapy success. The recent descriptions of paxlovid resistance
in vitro (
12–16) suggests there may be at some point no highly effective oral agent to prevent hospitalization in high-risk infected people. Thus, there remains a need to develop SARS-CoV-2 prophylactic and therapeutic regimens that are suitable for self/home administration, via inhaled or oral routes, particularly for infected people who are newly diagnosed and early in their COVID-19 course (
17). Ideally, these treatments should be relatively inexpensive so they can be widely used by infected people and their contacts around the globe.
It is well established that the most effective drug-based therapies for chronic, persistent RNA viruses include a combination of two or three drugs (
18,
19). Drug combinations comprise the standard of care for chronic RNA virus infections such as HIV and HCV infections and demonstrate promise for treating acute RNA virus infections, including infections with filoviruses (
20–22), arenaviruses (
23), influenza viruses (
24–27), and, most recently, SARS-CoV-2 (
28–31; for a review, see reference
32). A key feature of drug combinations is that they may confer bioactivities beyond additivity, such as multiplicative or synergistic effects, by targeting different steps of the viral life cycle. (Paxlovid is a combination treatment consisting of nirmatrelvir [PF-07321332], which inhibits the SARS-CoV-2 3C-like protease, and ritonavir, which slows the metabolism of nirmatrelvir [
33]. However, paxlovid does not leverage dual drug target sites.) Synergistic antiviral activity may permit dose reductions of each drug in the combination, thereby reducing the potential for clinical side effects while allowing increased clinical efficacy. Moreover, synergistic drug combinations can bring
in vitro drug levels that inhibit virus by 50% (IC
50) into
in vivo pharmacokinetic (PK) ranges. For chronic RNA virus infections such as HIV and HCV infections, drug combinations critically limit emergence of drug resistant viral mutants (
18,
19).
Combination therapies may be of particular use for immunocompromised patients with SARS-CoV-2 infection. More potent treatments are likely needed for patients who may not be able to exert adequate immune pressure on the virus (
34–36). Moreover, SARS-CoV-2 evolution in immunosuppressed patients has become a major public health concern since variants of concern (VOCs) may arise during SARS-CoV-2 infections that persist at high viral load for months (
37). Novel VOC continue to extend the pandemic by undermining vaccine efforts. There is also a greater window for the emergence of drug-resistant mutants in this setting.
Oral drug combination regimens are often preferred to allow outpatient treatment early during disease as prompt treatment after symptom onset is associated with better outcomes for SARS-CoV-2 virus (
1,
2,
4,
7), influenza virus (
38), Ebola virus (
39), HIV (
40), and zoster virus (
41) infections. Here, we demonstrate the synergistic potential of combining host-targeting antivirals (HTAs) that either target SARS-CoV-2 entry (camostat, nafamostat, and avoralstat) or an HTA (brequinar) that targets SARS-CoV-2 replication with a DAA replication inhibitor (molnupiravir) to block SARS-CoV-2 infection in Calu-3 lung epithelial cells. All four agents (three oral and one inhaled) synergized with the oral agent molnupiravir, with the pairs camostat plus molnupiravir and brequinar plus molnupiravir demonstrating the strongest synergy, which was similar to the synergy observed with the all-DAA combination of molnupiravir plus nirmatrelvir. When the three component drugs, brequinar, camostat, and molnupiravir were combined, even greater antiviral efficacy and potency was observed. Pharmacodynamic (PD) modeling of the
in vitro data suggest drug concentrations at which high potency can be achieved, which may fall within observed plasma values of these agents, hence suggesting possible
in vivo effects.
DISCUSSION
Besides paxlovid and molnupiravir, there are presently no approved and easy to administer (e.g., oral, intranasal, inhaled) antiviral medications to prevent or ameliorate SARS-CoV-2 infection. Because SARS-CoV-2 continues to evolve rapidly (
12–14), drug combination studies such as described in this report provide one step toward more effective drug-based control of the ongoing and possibly future pandemics. Effective drug combinations against HIV and HCV are comprised of two or three DAAs (
18,
19). Similarly, we show that combining the two currently approved DAAs, molnupiravir plus paxlovid, confers synergistic suppression of SARS-CoV-2 infection in human lung cells, as we predicted (
32), and as recently reported by others (
60,
61), including a report that the combination of molnupiravir plus paxlovid appears superior to either drug alone in a mouse model of SARS-CoV-2 (
62). We also show that combining HTAs with a DAA provides a similar level of antiviral synergy as an all-DAA combination: three TMPRSS2 inhibitors examined, when combined with molnupiravir, synergistically suppress WT SARS-CoV-2 and the VOC tested. Thus, TMPRSS2 is a potential target for an HTA-containing drug combination, as we have shown here with camostat, nafamostat, and avoralstat. Moreover, investigational inhibitors such as N-0385 (
63) and enoxaparin (
64) inhibit TMPRSS2 function by distinct mechanisms. In addition to TMPRSS2 cleavage, the SARS-CoV-2 Spike protein is also processed by the cellular enzyme furin. As such, novel furin inhibitors have been shown to synergistically inhibit SARS-CoV-2 infection when combined with camostat (
65). Moreover, a phase 2 randomized, double-blind, placebo-controlled clinical trial of oral camostat showed accelerated overall symptom resolution despite no reduction in intranasal viral RNA (
66), further supporting the notion that combining other drugs with camostat might enhance clinical and antiviral efficacy. Hence, new TMPRSS2 and furin inhibitors may be worth testing as drug combinations (
67,
68).
Brequinar combined with molnupiravir conferred strong antiviral synergy against SARS-CoV-2 infection of Calu-3 cells. Our findings agree with recent studies showing that this HTA+DAA combination is synergistic
in vitro and conferred superior antiviral effects in mice, although brequinar was administered via the intraperitoneal, not the oral, route in the recent report (
31). Brequinar inhibits dihydroorotate dehydrogenase (DHODH), a ubiquitous host enzyme that is required for
de novo pyrimidine synthesis (
69). It was recently shown that brequinar, when combined with dipyridamole, an inhibitor of the pyrimidine salvage pathway (which can be activated when DHODH is inhibited by brequinar), also confers synergistic suppression of SARS-CoV-2 (
70). Thus, dipyridamole may provide added benefit to brequinar-containing drug combinations.
We provide a mathematical model that faithfully reproduces experimental data and projects potency against SARS-CoV-2 at combinatorial drug concentrations that were not specifically tested experimentally. This model will serve as the basis for more comprehensive models that include PD data as shown here with human PK for the individual drugs and with SARS-CoV-2 dynamics, to accurately project human
in vivo potency (
71), as we recently did for synergistic drug combinations for Ebola virus (
58). Combined PK, PD, and viral dynamics modeling can be conducted for the drug combinations tested here, as well as for others proposed in the literature (
32,
72) or revealed in future studies. Lastly, with the triple combination of two HTAs (camostat and brequinar) and a DAA (molnupiravir), we demonstrate the potential of adding a third drug to enhance the potency of a drug pair, echoing the fact that certain drug regimens against HIV and HCV contain three drugs. The next stages of our work will involve further modeling (
58) and testing of promising oral drug combinations in mouse models of SARS-CoV-2.
Paxlovid and molnupiravir have 89 and 30% efficacies, respectively, defined as prevention of hospitalization or death (
7,
8,
73,
74). While current orally available drugs for SARS-CoV-2 are highly effective at reducing hospitalization, they are not effective for all indications (including postexposure prophylaxis). Although the issue of paxlovid resistance during monotherapy (
12–16) has not yet fully emerged clinically, preexisting drug-resistant mutants have been found in therapy naive patients (
75,
76). Moreover, resistance to remdesivir has also been observed
in vitro (
77,
78) and clinically (
79). Thus, current antivirals are not guaranteed to maintain effectiveness against future variants. We propose that the drug combinations described here may demonstrate increased efficacy
in vivo and may be of particular importance in immunocompromised hosts who have a difficult time clearing the virus. As investigators are testing combinations of monoclonal antibodies plus a polymerase inhibitor for SARS-CoV-2 (
80) and small molecule drug combinations for influenza (
27), we advocate for further modeling and
in vivo testing of oral small molecule drug combinations containing both HTAs and DAAs versus SARS-CoV-2, VOC, other coronaviruses, and other viruses for pandemic response and preparation (
32).
MATERIALS AND METHODS
Chemicals, cell culture, and live virus.
Vero E6 and Calu-3 cells were maintained in standard medium (minimum essential medium; Gibco, catalog no. 11095) supplemented with 9% fetal bovine serum (FBS; HyClone, SH3007103) and 1% penicillin-streptomycin (Gibco, catalog no. 15140); HEK293T cells expressing human ACE2 and TMPRSS2 (
55) (here called 293TAT cells), provided by Carol Weiss, were maintained in Dulbecco modified Eagle medium (DMEM; Gibco, catalog no. 11995), 9% FBS (HyClone, SH3007103), 1% penicillin-streptomycin (Gibco, catalog no. 15140), 1% nonessential amino acids (Gibco, catalog no. 11140), and 20 mM HEPES (Gibco, catalog no. 15630). HTA and DAA compounds were purchased from commercial vendors as outlined in Table S1 in the supplemental material.
Wild-type, infectious SARS-CoV-2 was obtained from BEI Resources (isolate USA-WA1/2020 NR-52281). The SARS-CoV_2 VOC stocks of beta and delta viruses were generated as follows. The beta virus (B.1.351 lineage), isolate hCoV-19/South Africa/KRISP-K005325/2020 (NR-54009; GISAID: EPI_ISL_678615) was originally obtained from BEI Resources, NIAID, NIH, and further expanded in Vero-TMPRSS2 cells in DMEM containing 2% heat-inactivated FBS (HI-FBS). The delta virus (GISAID accession number EPI_ISL_13858636) was isolated from local Seattle nasopharyngeal swab samples of a SARS-CoV-2-infected patient stored in viral transport medium, kindly provided by Alex Greninger. Briefly, the viruses in the VTM were first expanded using Vero-ACE2-TMPRSS2 cells in DMEM containing 2% HI-FBS. The initial crude expansion was then plaque-purified in Vero-ACE2-TMPRSS2 cells. The plaque-purified viruses were sent for amplicon sequencing (Swift Biosciences) to confirm their genome sequence. Virus was further amplified in Vero-TMPRSS2 cells. Virus titers were measured by plaque assays on Vero-TMPRSS2 cells.
Drug treatment and infection of cells with live virus.
Compounds were added to cell plates and 2 h later, live virus (SARS-CoV-2 WA1 at a multiplicity of infection [MOI] of 0.1 or delta or beta VOC at MOIs of 0.01) was added to each infected well, or virus-free medium was added to each uninfected well. Plates were incubated at 37°C and 5% CO2 for 48 or 96 h 293TAT cells or Calu-3 cells, respectively. The assay readout was the cell viability based on a CellTiter-Glo assay (Promega).
Cell viability assay.
For drug dose-response studies, we measured cell viability in parallel wells using the CellTiter-Glo 2.0 assay (Promega). The assay measures the number of viable cells in culture by quantifying ATP, which indicates the presence of metabolically active cells. The contents of each well were aspirated, followed by the addition of 50 μL of phosphate-buffered saline (Gibco, catalog no. 10010031) and 50 μL of CellTiter-Glo reagent. The plate was then shaken for 2 min, incubated at room temperature for 10 min, and read for luminescence on a BioTek Synergy H4 plate reader.
Drug combination assays.
Checkerboard assays of two drug combinations were performed as described earlier (
23). Briefly, two drugs were tested in dose responses consisting of dimethyl sulfoxide (DMSO) control and five concentrations of each drug, yielding six concentrations of drug A and six concentrations of drug B. Combining the two separate drug dose responses creates a matrix or checkerboard of 36 dose-combinations of the two drugs. Calu-3 (
n = 50,000) or 293TAT (
n = 10,000) cells were seeded into each well of 96-well black, clear-bottom plates in the cell line’s respective medium. Each two-drug concentration tested in a 6 × 6 checkerboard assay was performed in duplicate for 293TAT cells or triplicate for Calu-3 cells. Cells were infected, and parallel wells were not infected to assess the toxicity of the drugs. Both plates were incubated at 37°C and 5% CO
2 for 48 h (for 293TAT cells) or 96 h (for Calu-3 cells) prior to the CTG assay. For three drug combination studies, we employed a method that samples the diagonal space of the three-dimensional checkerboard (
23,
59). Briefly, camostat, brequinar, and molnupiravir (EIDD-1931) were mixed at top concentrations of 1.4, 50, and 0.28 μM, respectively, to produce one-, two-, and three-drug combinations. Stock concentrations were serially diluted in 10% increments and added to Calu-3 cells at 2 h prior to infection. After 96 h, cell viability was measured by a CellTiter-Glo assay (Promega).
Data analyses.
For single-drug experiments, drug concentrations were log transformed, and the concentrations of drug(s) that inhibited virus by 50% (i.e., IC
50) and the concentrations of drug(s) that killed 50% of cells (i.e., CC
50) were determined via nonlinear logistic regressions of log(inhibitor) versus response-variable dose-response functions (four parameters) constrained to zero bottom asymptote by statistical analysis using Prism 9 (GraphPad Software, Inc.), as described previously (
23). Input data for these equations consisted of relative light units (RLU) generated by CellTiter-Glo assays. The selectivity index was calculated by dividing the CC
50 by the IC
50. The antiviral efficacy was calculated by comparing the RLU from virus-infected (IFX) cells treated with drug to the average of the infected cells treated with solvent (DMSO) and expressed as a percentage relative to the virus-induced cytopathic effect: RLU of noninfected/DMSO-treated cells – RLU in infected DMSO treated cells, using the following equation:
Cell viability was calculated by comparing the RLU from noninfected cells treated with drugs to the noninfected cells treated with DMSO:
Dose-response data from checkerboard assays were analyzed in SynergyFinder3, an open-access platform for multidrug combination synergies (
56). Several combination parameters were reported from SynergyFinder3, including the average Bliss Synergy Score of the entire dose-response matrix and the maximum synergistic area (MSA), which corresponds to the maximum Bliss score calculated over an area of nine doses of the two compounds in a checkerboard experiment (i.e., 3 × 3 dose-response submatrix). Synergy is defined as when the observed inhibition is greater than that predicted by multiplicative Bliss independence at a given set of drug concentrations. For triple-drug experiments, the three-drug combination was compared to the three two-drug combinations of the drugs that comprise the triple combination using SynergyFinder3.0. Finally, SynToxProfiler (
57) was used to evaluate the toxicity profile of the drug combinations in noninfected cells via checkerboard assays. For this, toxicity was calculated by subtracting the viability results above from 100 (to convert to percent inhibition values), and checkerboard toxicity data were uploaded to SynToxProfiler (see Fig. S2).
Mathematical modeling.
We previously modified the Bliss independence PD model (
81) to directly assess the potency of drug combinations across different drug concentrations. The purpose of the model was to recapitulate observed experimental data and then make projections about combinatorial drug potency at concentrations that were not measured experimentally. The model can then be synchronized with PK models to predict the percentage of new cell infections being prevented at any point during the drug dosing interval. The model includes the parameters IC
50 (the drug concentration at which 50% of infections are prevented), Hill coefficients of both drugs (
h1 and
h2, the slope of the dose-response curve), and an exponent (
a) for the assessment of synergy (
58), as expressed in the following equation:
ACKNOWLEDGMENTS
This article is dedicated to the memory of Hugo R. Rosen, a talented clinician-scientist, collaborator, colleague, and friend.
J.M.W. was partially supported by NIH grant AI114776. S.J.P. was partially supported by a Washington Research Foundation Technology Commercialization Grant and the Department of Laboratory Medicine and Pathology (University of Washington). T.A. was partially supported by Academy of Finland grants 345803 and 340141. S.B.B. was partially supported as an Open Philanthropy Awardee of the Life Sciences Research Foundation. M.G. and T.-Y.H. were supported by NIH/NIAID grant AI151698. S.P. was supported by BMBF grants (01KI2006D, 01KI20328A, and 01KX2021), by the Ministry for Science and Culture of Lower Saxony (14-76103-184, MWK HZI COVID-19), and by the German Research Foundation (DFG; PO 716/11-1 and PO 716/14-1).
We thank Carol Weiss for the 293TAT cells and Alex Greninger for the delta clinical sample. The following reagent was deposited by the Centers for Disease Control and Prevention and obtained through BEI Resources, NIAID, NIH: SARS-related coronavirus 2, isolate USA-WA1/2020, NR-52281. The following reagent was obtained through BEI Resources, NIAID, NIH: SARS-related coronavirus 2, isolate hCoV-19/South Africa/KRISP-K005325/2020, NR-54009, contributed by Alex Sigal and Tulio de Oliveira.