Mycophenolic acid treatment drives the emergence of novel SARS-CoV-2 variants

Significance

Mycophenolic acid (MPA), an immunosuppressant widely used in posttransplant regimens, exhibits antiviral activity by depleting cellular guanosine triphosphate, thereby inhibiting viral replication. However, prolonged exposure to MPA can drive the emergence of novel viral variants with enhanced replication capabilities. Here, we identified specific mutations in severe respiratory syndrome coronavirus 2 that conferred altered viral fitness, allowing for faster replication and increased viral titers despite MPA treatment. Importantly, these mutations have been observed in vivo, suggesting a real-world risk of variant evolution under immunosuppressive treatment. However, these mutations have not yet been identified together in a single infected individual. These findings underscore the importance of vigilant monitoring in immunosuppressed patients, as treatment may inadvertently foster viral variants with a competitive advantage.

Abstract

Mycophenolic acid (MPA) is commonly used in immunosuppressive regimens following solid organ transplantation. We demonstrate that MPA treatment reproducibly inhibits the replication of a range of viruses, including severe respiratory syndrome coronavirus 2 (SARS-CoV-2). Mechanistically, we identified cellular guanosine triphosphate pool depletion as a key mediator of this antiviral effect. Strikingly, this inhibition can be overcome which was correlated with the emergence of three breakthrough mutations in the SARS-CoV-2 genome (S P812R, ORF3 Q185H, and E S6L). Subsequent analyses confirmed that the combination of these mutations conferred accelerated replication kinetics, higher viral titers, and more rapid onset of cytopathic effects, but not MPA resistance. Comparison of global transcriptional responses to infection highlighted dysregulation of specific cellular gene programs under MPA treatment prior to breakthrough mutation emergence. Together, these findings identify viral and host drivers of variant emergence under immunosuppression. They also advocate for close monitoring of immunosuppressed patients, where emergence of novel viral variants with a fitness advantage may arise.

Viral infections pose a significant global threat, contributing to disease, mortality, and economic losses. A large number of deaths are attributed to infections with viral pathogens globally (1, 2). This raises significant concerns, especially given the limited availability of around 90 drugs for treating nine viral species and the approval of only about 15 vaccines for various viral species at present (3). Recent viral outbreaks including Zika (ZIKV), Ebola (EBOV), and the more recent Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) pandemic highlight the urgent need for novel antiviral strategies.

Approximately 22% of the global population display at least one underlying health condition, which puts them at higher risk of adverse outcomes for infectious diseases (4). Indeed, the recent SARS-CoV-2 pandemic highlighted the vulnerability of specific populations with comorbidities toward less favorable disease outcomes (5, 6). Among those, immunosuppressed patients have become the center of attention. In the United States, an estimated 2.7 to 4% of all adults self-report immunosuppression, which refers to a diverse group of individuals with inherited or acquired immune deficiencies (7). This group includes individuals with various malignancies and immunologic diseases, as well as patients who have undergone solid organ transplantation (8). The latter are typically treated with immunosuppressive drugs, which can be classified into different categories, including calcineurin inhibitors (such as Tacrolimus and Cyclosporin A), antiproliferative agents (such as Mycophenolate mofetil and azathioprine), mTOR inhibitors (such as Rapamycin), and steroids (Prednisolone). Mycophenolate mofetil, a prodrug of mycophenolic acid (MPA), is an inhibitor of the inosine monophosphate dehydrogenases (IMPDH) thereby negatively impacting purine biosynthesis and causing cell cycle arrest (9). In addition to being an immunosuppressive drug, MPA has been shown to effectively inhibit the replication of flaviviruses such as dengue virus (DENV) (10, 11) and hepatitis C virus (HCV) (12, 13), as well as the Old World arenavirus lymphocytic choriomeningitis virus (14) and several other viruses including Mpox virus (MPXV) (15–22) in vitro. Laboratory experiments have also demonstrated that MPA can impair the replication of several coronaviruses such as Middle East Respiratory Syndrome coronavirus (MERS-CoV), SARS-CoV-2, and endemic coronaviruses (such as human coronavirus HCoV-OC43 and -NL63) (14, 23, 24). Additionally, MPA was recently shown to effectively inhibit bacterial infections including Chlamydia (25) and also displayed beneficial effects on gastric cancer cells by targeting several KEGG pathways (9). Intriguingly, in animal models including mice and common marmosets, the administration of MPA resulted in enhanced infection rates for SARS-CoV-2 and MERS-CoV. However, the underlying mechanism and potential for viral adaptation is still to be determined.

Despite ongoing debates regarding the severity of the disease (26–28) and mortality rates (27–29), there is a consistent body of evidence reporting prolonged or chronic infections with SARS-CoV-2 in immunocompromised individuals (30–34). Certain reports even proposed that SARS-CoV-2 Variants of Concern (VOCs), which became globally dominant, including the Alpha variant, are thought to have originated from individuals with compromised immune systems (35–37). This phenomenon has been observed in chronic infections with other viruses, such as DENV (38) and hepatitis E virus (HEV) (39). Continuous replication of the virus potentially favors the accumulation of mutations in order to adjust to environmental and immunological pressure and therefore presumably facilitates the emergence of new variants with enhanced epidemic or pandemic potential (35, 40–43). Additionally, immunocompromised patients often require pharmacological interventions to clear infections, which may also further drive viral mutagenesis. Direct acting antivirals such as the nucleotide analogues Remdesivir or Ribavirin (RBV) facilitate lethal mutagenesis during viral replication, which leads to defective or replication-incompetent viral genomes and subsequent viral extinction (44, 45). However, in some cases the selective pressure drives viral mutagenesis leading to the formation of intrahost populations (46, 47). RBV treatment of patients infected with HEV was associated with increased viral heterogeneity, eventually resulting in the emergence of RBV-resistant mutants and/or variants with increased viral fitness, causing therapy failure (46–49). Remdesivir has also been shown to further facilitate viral mutagenesis upon administration of low concentrations (47, 50). The emergence of resistance mutations has also been reported during treatment with antiretroviral therapy of HIV infection (51).

The emergence of novel viral variants presents a risk to the general population by affecting various aspects such as transmission dynamics, public health measures, vaccine efficacy, antiviral treatment, disease severity, and pathogenesis, among others. Therefore, it is crucial to thoroughly investigate potential drivers of mutagenesis and understand the impact of specific mutations. Here, we aimed to explore viral replication under immunosuppressive therapy to better predict and manage the spread of novel variants.

Results

Mycophenolic Acid Inhibits a Wide Range of Viruses.

Immunosuppression can have a profound impact on viral infections (52). To further investigate this, we tested the antiviral activity of MPA, Rapamycin, Cyclosporin A, Tacrolimus and Prednisolone against SARS-CoV-2 infection in vitro (SI Appendix, Fig. S1 and Table S1). To this end, VeroE6 cells were treated with single or combinatorial doses of immunosuppressants at clinically relevant concentrations to mimic long-term immunosuppression, reflecting the common practice of prescribing multiple immunosuppressants to patients (53–60). Viral replication was determined by an end-point dilution assay to determine infectious particles released into the supernatant (SI Appendix, Fig. S1A) and cell associated viral M-gene expression was quantified by qRT-PCR (SI Appendix, Fig. S1B). Consistent with previous reports (10), MPA treatment led to a significant decrease in infectious viral titers (1.5 logarithmic units [log] TCID₅₀/mL), whereas none of the other immunosuppressive drugs exhibited an antiviral effect upon single treatment (SI Appendix, Fig. S1 A and B). None of the drug combinations affected cell viability (SI Appendix, Fig. S1C). Combinations encompassing MPA reduced virus replication, without any synergistic effects with drugs that are frequently coadministered (SI Appendix, Fig. S1 A and B). To validate these findings in a more physiologically relevant system, differentiated primary human airway epithelial cells (hAEC) in air liquid interface (ALI) culture were exposed to single treatments or different combinations of immunosuppressive drugs, and the production of infectious particles (Fig. 1 A–F, Left) and cell viability (Fig. 1 A–F, Right) were monitored over time. In line with previous observations, solely MPA-encompassing treatments reduced the production of infectious progeny, without affecting cell viability (Fig. 1 A–F). These observations confirm an antiviral effect for MPA against SARS-CoV-2 in vitro.

Fig. 1.

Antiviral activity of immunosuppressive drugs in hAECs over time. hAECs were treated with MPA (2.5 µg/mL), Rapamycin (6 ng/mL), Cyclosporin A (150 ng/mL), Tacrolimus (6 ng/mL), and/or Prednisolone (20 ng/mL) in indicated combinations (A–F). After 1 h the cells were infected with SARS-CoV-2 (25,000 PFU) for 2 h and subsequently washed thrice with HBSS. Directly after infection as well as 24, 48, 72, and 96 h post infection, infectious progeny virus was collected from the apical site and subjected to an end-point dilution assay to determine TCID₅₀/mL (left half of each graph). Viral titers were normalized to viral titers derived from the untreated cells (dotted line). Simultaneously, cell culture medium from the basal site was obtained to evaluate cell viability by an LDH assay (right half of each graph). All experiments were performed in three different donors (mean ± SD). MPA—mycophenolic acid. ns—not significant.

To investigate the breath of antiviral activity, we subsequently examined the susceptibility of unrelated RNA and DNA viruses from various families to MPA inhibition. Consistent with prior findings (15–21), we observed a dose-dependent decrease in virus infection upon MPA treatment for the SARS-CoV-2 Omicron variant BA2.86, HCoV-229E, HEV, respiratory syncytial virus (RSV), Influenza A virus (IAV), and MPXV (SI Appendix, Fig. S2). These viruses exhibited comparable susceptibility to MPA treatment, with IC₅₀ values ranging from 0.07 (IAV) to 1.59 µg/mL (MPXV). IAV and MPXV infections were completely abolished at the highest concentration of MPA, while reduced virus production was detected for HCoV-229E, HEV, and RSV. The IC₅₀ value for the Omicron variant BA2.86 was similar to what has been observed for the SARS-CoV-2 B.1.1.70 variant. These findings indicate MPA’s broad activity against diverse viruses is likely due to its targeting of cellular processes rather than specific viral proteins.

Mycophenolic Acid Targets Postentry Stages of the Coronavirus Lifecycle via Depletion of Intracellular Guanosine Triphosphate (GTP) Pools.

Immunosuppressive drugs impact a variety of cellular signaling pathways including JAK/STAT, NF-κB, PI3K/AKT-mTOR, MAPK, and Keap1/Nrf2/ARE pathway (61, 62), thereby modulating host cell immune responses. Although the cellular targets of MPA are well defined, its specific antiviral mechanisms remain elusive. MPA has been shown to affect the expression of cellular structural proteins, fatty acid, and lipid metabolism, as well as nucleotide-dependent processes including the reduction of GTP pools (63). To further study which step of the viral replication cycle was affected, we first focused on virus entry, employing pseudotyped VSV where the native glycoprotein was deleted and replaced with the SARS-CoV-2 Spike protein (VSVΔG + S), encoding a firefly luciferase reporter (64). VeroE6 cells were treated with MPA, Tacrolimus, and/or Prednisolone at different time points relative to infection, including 1 h before infection (pre¹), before and during infection (pre–during²), during the entire experiment (pre–during–post³) or only after infection with VSVΔG + S (post⁴) (SI Appendix, Fig. S3A). Cell viability remained unaffected throughout the experiment (SI Appendix, Fig. S3B). While drug treatment prior to infection or before and during infection did not impact luciferase activity, administering MPA after infection with VSVΔG + S resulted in reduced luciferase activity suggesting that MPA does not play a role during virus entry but affects the expression of the reporter gene (SI Appendix, Fig. S3C). Next, similar time-of-addition experiments were performed using authentic SARS-CoV-2. A549-A/T cells were treated with either the vehicle control (EtOH) or MPA for the specified time periods (Fig. 2A) and viral titers were determined. MPA treatment, as long as administered shortly after infection or for an extended period of time before infection (24 h), led to a significant reduction in SARS-CoV-2 secretion compared to the vehicle control. In contrast, a single pretreatment dose administered 1 h before infection did not alter SARS-CoV-2 titers.

Fig. 2.

Effect of MPA on the virus life cycle. (A) Time of addition experiments were performed using full-length SARS-CoV-2 (25,000 PFU). A549-A/T cells were seeded at 8 × 10⁴ cells/mL and treated either with the vehicle control (EtOH) or MPA (2.5 µg/mL) for time periods indicated by the black arrows. Twenty-four hours post infection the supernatant was collected, and infectious progeny production was assessed by an end-point dilution assay to determine TCID₅₀/mL. The long dashed line indicates viral loads calculated for the UTC. Cell viability was monitored by an MTT assay (Right). Statistical significance was estimated by one-way ANOVA with Dunnett’s multiple comparison (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001, ****P ≤ 0.0001). All experiments were performed in triplicates (mean ± SD). Dots indicate technical replicates. (B) A549-A/T cells were seeded at 8 × 10⁴ cells/mL and incubated until attached. One-third of the cells were treated with 25 µg/mL MPA 10 h before infection (preMPA). All cells were infected with SARS-CoV-2 (25,000 PFU) for 1 h and subsequently washed thrice with 1× PBS. The cells were then left untreated or treated with indicated concentrations of MPA. At the indicated time points the supernatant was collected and the cells were subjected to two freeze-and-thaw cycles. Extracellular and intracellular viral titers were determined by an end-point dilution assay. (C) Representative fluorescence images of infected A549-A/T cells stained for the nucleocapsid. All experiments were performed in triplicates (mean ± SD). ns—not significant. TCID₅₀—50% tissue culture infectious dose. p.i. = post infection.

A time-course analysis and subsequent determination of intra- and extracellular viral titers (Fig. 2B) revealed the formation of infectious viral particles within A549-A/T cells takes approximately 6 h post infection (hpi) and was delayed by an additional hour before becoming detectable in the supernatant. When MPA treatment is initiated simultaneously with infection, there is no discernible difference in viral loads until the 10-h time point. We also pretreated the cells 10 h before infection, mimicking immunosuppressive treatment before infection. In this case, we observed a delay of infectious particle production by another 3 h for both intra- and extracellular virus (Fig. 2B). Furthermore, immunofluorescence staining confirmed a reduction of SARS-CoV-2 N-protein positive A549-A/T cells upon MPA treatment (Fig. 2C), suggesting that a reduced number of cells establish productive infection upon MPA treatment. Together, these results demonstrate that MPA affects postentry stages of the SARS-CoV-2 replication cycle, thereby reducing the amount of infectious progeny produced over time.

Similar to RBV, MPA inhibits IMPDH, an essential enzyme in cellular metabolism that regulates the production of guanine nucleotides necessary for DNA and RNA synthesis (63). Specifically, IMPDH facilitates the conversion of inosine monophosphate (IMP) to xanthosine monophosphate (XMP), a critical step in the de novo biosynthesis of guanosine nucleotides. Consequently, MPA-mediated inhibition of IMPDH leads to reduced levels of GTP (63). However, studies have shown that this effect can be reversed by introducing exogenous guanosine (G) or guanosine monophosphate (GMP) during RBV treatment (20, 65). To explore this mechanism in the context of infected cells, we treated A659-A/T cells with a single dose of MPA while simultaneously adding different concentrations of G and/or GMP (Fig. 3). Increasing concentrations of G and/or GMP resulted in restored viral titers, thus counteracting the antiviral effects of MPA (Fig. 3 C and F), without affecting cell viability (Fig. 3 A and D). Of note, individual addition of either GMP, G or GMP and G together did not affect virus titers (Fig. 3 B and E). Addition of the monophosphates AMP, CMP, UMP, or TMP during MPA treatment did not lead to a rescue of viral infectivity (SI Appendix, Fig. S4). These findings suggest that MPA’s broad antiviral effect is likely mediated by the depletion of cellular GTP and can be reversed by the exogenous substitution of G and/or GMP, but not other monophosphates.

Fig. 3.

The antiviral effect of MPA is reversible by ectopic substitution of guanosine (G) and/or GMP. Experiments were performed on Huh7 cells for 229E (A–C) and HEp-2 cells for RSV (D–E). Cells were seeded at 8 × 10⁴ cells/mL and incubated until fully attached. (A and D) Afterward the cells were treated with GMP and/or G (100 µM to 0.78 µM) and cell viability was assessed 24 h post treatment. Additionally, cells were infected with HCoV-229E and RSV (MOI 0.1) for 1 h. To remove the inoculum the cells were washed thrice with 1× PBS. Afterward the cells were treated with (B and E) solely GMP and/or G (100 µM) or with (C and F) GMP and/or G (100 µM to 0.78 µM) in combination with 2.5 µg/mL MPA. The supernatant was harvested 24 h post infection. Infectious viral titers were determined by an end-point dilution assay (TCID₅₀/mL). Statistical significance was estimated by one-way ANOVA with Dunnett’s multiple comparison (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001, ****P ≤ 0.0001). All experiments were performed in triplicates (mean ± SD). ns—not significant. TCID₅₀—50% tissue culture infectious dose. UTC—untreated control. GMP—guanosine monophosphate. G—guanosine. MPA—mycophenolic acid.

We hypothesized that combinational treatment of MPA with direct acing antivirals could enhance the antiviral effect, resulting in complete virus inhibition. Currently approved treatment options for COVID-19 include the direct acting antivirals Nirmatrelvir und Ritonavir (Paxlovid®, Pax), as well as RBV (66). In line with previous reports, single treatment of either compound significantly inhibited SARS-CoV-2 replication (SI Appendix, Fig. S5A) without any effect on cell viability (SI Appendix, Fig. S5B). Strikingly, combination treatments with MPA significantly enhanced the antiviral activity, confirming additive effects and superior activity compared to single treatment (SI Appendix, Fig. S5 C and D). These observations underline the potential superiority of combinational treatment approaches targeting both viral and host processes.

SARS-CoV-2 Can Rapidly Adapt to the Selective Pressure of Mycophenolic Acid.

Direct-acting antivirals including RBV or Remdesivir have the potential to increase viral mutagenesis, thereby driving replication beyond the error threshold, leading to viral extinction, or facilitating the emergence of variants with enhanced viral fitness (46–49). Medications directed at cellular proteins typically pose a lower risk of promoting antiviral resistance. However, in the case of MPA, the identified mode of action could still play a role in viral mutagenesis. To investigate this, SARS-CoV-2 was serially passaged in A549-A/T cells under different treatment regimens encompassing MPA (M); MPA and Tacrolimus (MT); or MPA, Tacrolimus, and Prednisolone (MTP). Viral titers were monitored throughout the passaging experiments (Fig. 4A). Consistent with our initial findings (Fig. 1 and SI Appendix, Fig. S1), viral titers decreased upon MPA treatment compared to virus propagated on untreated cells (UTC) until passage 5. However, within 8 passages, viral titers recovered to comparable levels as the UTC, pointing toward adaptation to the treatment. No further changes in viral titers were observed during the remaining passaging time (up to 20 passages, p20). The different combinational treatment regimens displayed no discernible differences. Alongside recovered infectivity, phenotypic changes resulting in a significantly increased plaque size, that was not observed in the UTC, became apparent (Fig. 4 B and C). These findings indicate viral adaptation, potentially involving changes in replication fitness and/or resistance to treatment.

Fig. 4.

SARS-CoV-2 passaging under the presence of MPA in physiological concentrations. A549-A/T cells were seeded in 75 cm² flasks and treated with solely MPA (2.5 µg/mL; M), MPA, and Tacrolimus (6 ng/mL; MT), MPA, Tacrolimus and Prednisolone (20 ng/mL; MTP) or left untreated (UTC). One hour later the cells were inoculated with 200,000 PFU for 1 h. Hereafter, the inoculum was removed and replaced with fresh MPA, Tacrolimus and/or Prednisolone-containing DMEM. The cells were then incubated for 24 h and considered as the first passage of virus (p1). For the following passages, pretreated A549-A/T cells were inoculated with supernatant from the previous passage for 1 h. Hereafter the inoculum was replaced by fresh 5% FCS-containing DMEM with the applicable treatment. (A) For every passage infectious progeny production was determined as PFU/mL. (B) Cell viability was assessed 24 h post treatment. (C) Phenotypic changes of plaques were documented and (D) quantified using ImageJ. Statistical significance was estimated by one-way ANOVA with Dunnett’s multiple comparison (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001, ****P ≤ 0.0001). PFU—plaque forming units.

To exclude a potential gain-of-function, passaged viruses were tested for their potential to evade neutralizing antibodies and antiviral treatment approaches. Although passaged viruses exposed to MPA exhibited reduced sensitivity (SI Appendix, Fig. S6 A and B), their responsiveness to clinically relevant drugs such as Molnupiravir, Paxlovid® or Remdesivir and neutralization by neutralizing antibodies remained unchanged (Fig. 5). Notably, extracellular infectivity was found to be enhanced, particularly for viruses passaged in the presence of MPA, compared to the virus passaged in the UTC cells (SI Appendix, Fig. S6C). This suggests altered viral fitness noticeable through accelerated replication kinetics and the efficient release of infectious viral particles from A549-A/T cells. Given the increased pressure upon depletion of cellular GTP pools, we hypothesized that virus–host interactions may be modulated as viruses adapt toward differential cellular conditions. Therefore, the ratio of virus–host reads and cellular transcriptional responses to infection in UTC and MPA-treated cells were investigated by RNA-sequencing (RNA-seq) (67) (Fig. 6). In UTC cells, virus read numbers generally remained stable at ~20% of cellular reads across all passages. In contrast, in MPA treated cells, early passage virus-mapped reads (p1, p3, and p5) were demonstrably lower (<10%), followed by a massive increase to over 70% of total cellular reads at later passages (p9, p15, and p20) (Fig. 6A). No obvious dysregulation of viral cofactors or lung markers was observed (Fig. 6B). Based on these data, comparison of global transcriptional responses to infection in UTC cells (n = 6) were made separately to cells infected with either early passage virus where replication was suppressed, (n = 3) or late passage virus where replication was enhanced (n = 3) (Fig. 6 C–F). In early virus passage infected cells, differentially expressed gene (DEG) and Gene Ontology (GO) analysis unveiled a dysregulation of cellular genes (Fig. 6 C and D), associated with distinct biological processes (Fig. 6 D–F). Specifically, MPA-treated cells showed enhanced protein modification processes as well as response to stress, NFκB signaling, and biosynthetic and metabolic processes (Fig. 6 E and F). In contrast, late passage virus-infected cells showed minimal transcriptional differences when compared to UTC (Fig. 6 C–F). Together, these data highlight the host genes and cellular processes associated with MPA suppression of SARS-CoV-2. Additionally, the rapid loss of MPA associated inhibition, evidenced by the massive increase of viral replication observed with late passage virus, are indicative of the emergence of breakthrough mutations, conferring altered viral fitness restoring the ability of the virus to replicate efficiently in the context of cellular metabolic changes induced by MPA.

Fig. 5.

Sensitivity of passaged SARS-CoV-2 toward clinically relevant drugs. A549-A/T cells were seeded at 8 × 10⁴ cells/mL and incubated until fully attached. The cells were infected with SARS-CoV-2 (25,000 PFU) for 1 h. The inoculum was removed and the cells were washed thrice with 1× PBS. Hereafter, the cells were treated with the indicated concentrations of Molnupiravir (Mol), Paxlovid® (Pax) and Remdesivir (Rem). Twenty-four hours post infection (A) cell viability was assessed by MTT assay and (B and C) the supernatant was collected for subsequent determination of infectious viral titers performing an end-point dilution assay. (D) A neutralization assay was performed to calculate the neutralization efficacy of the WHO standard against the passaged viruses. The 50% neutralization dose (ND₅₀) was calculated by a linear regression model.

Fig. 6.

Transcriptional profiling of MPA treated cells. SARS-CoV-2 was serially passaged in naïve A549-A/T cells either in combination with MPA, or in UTC cells. Subsequently, total RNA was isolated from infected cells from selected passage (p1, p3, p5, p9, p15, and p20) and subjected to next generation sequencing. (A) Reads were mapped to both the human (Hg38) and SARS-CoV-2 genome. Based on the virus mapped read numbers for MPA treated cells, passages were divided into early cell passages (p1, p3, and p5) and later cell passages (p9, p15, and p20) for subsequent analyses of host cell gene dysregulation. (B) The abundance of SARS-CoV-2 entry factor mRNAs, as well as lung, hepatocyte, and brain markers were determined. (C) The cellular response to virus infection is displayed as volcano plots. gray—n.s. and low L2FC, dark green—n.s. and L2FC > 1, light blue—P ≤ 0.05 and low L2FC, light red—P ≤ 0.05 and L2FC > 1. (D) Differentially expressed genes (RPKM > 1, FDR P < 0.05) and enriched gene ontology categories (FDR P < 0.05) were determined. (E) Selected GO categories colored according to the mean L2FC. Circle size indicates gene ratio. No outline—n.s.; blue outline—P ≤ 0.05. (F) Genes of interest belonging to enriched gene ontology categories. UTC—untreated control. Early passages—p1, p3, and p5. Late passages—p9, p15, and p20. DEG—differentially expressed gene. GO—gene ontology. L2FC- log₂ fold change.

Mutational Changes during Viral Adaptation.

The recovery of infectious viral titers of SARS-CoV-2 when passaged in MPA-treated A549-A/T cells, coupled with evident phenotypic changes in plaques, strongly suggests the emergence of variants potentially carrying mutations that alter viral fitness and/or reduce sensitivity to MPA. To investigate this further, viral RNA from each passage (p0–20) was sequenced. The SARS-CoV-2 genome coverage was comparable across different passages of the virus cultured in UTC cells, while the coverage increased in later passages for the virus continuously exposed to MPA (Fig. 7A). Despite the selective pressure from MPA, nucleotide usage across the SARS-CoV-2 genome remained consistent throughout the passaging process (SI Appendix, Fig. S7 A and B). This suggests that while MPA’s antiviral activity is mediated by cellular GTP depletion, this metabolic stress did not prompt the virus to alter its nucleotide composition as an adaptive response. When examining polymorphic sites, the vast majority of sites contained variants at substantially less than 1% frequencies, and no enrichment of mutation frequencies at synonymous sites was observed, indicating no obvious mutagenic effects of MPA. We found that only a minority of SNVs occurred at frequencies >1% of the viral population, with only a small proportion of these SNVs becoming fixed in the population and leading to a consensus sequence change (SI Appendix, Fig. S7C). In total, seven mutations were detected with a frequency exceeding 50% in at least one passage (Fig. 7B). Among these mutations, two were also observed in the untreated control (UTC), indicating likely adaptations to cell culture conditions. Notably, three mutations (S P812R, ORF3 Q185H, and E S6L) became dominant and were correlated with “breakthrough” viral load increases from continued exposure to MPA, providing direct evidence of SARS-CoV-2 adaptation to MPA treatment (Fig. 7C). These mutations became dominant in both intracellular and extracellular viruses. To study the impact of the MPA-related mutations, S P812R, ORF3 Q185H, and E S6L were introduced into the wild-type (WT) SARS-CoV-2 genome (with D614G) through reverse genetics, generating three single mutants and a triple mutant harboring all three mutations. Interestingly, the triple variant demonstrated increased production of infectious virus, in contrast to individual mutants, which exhibited comparable titers to WT virus (Fig. 8A). In line with this finding, time kinetics indicate enhanced replication of the triple variant in comparison to the WT virus and all three single mutants by immunofluorescence staining for the SARS-CoV-2 Nucleoprotein (NP) in relation to the total cell number (Fig. 8B). Furthermore, bright field microscopy and analysis of plaque size suggested enhanced cell death for the triple mutant (Fig. 8 C and D). However, no significant increase in IC₅₀ values was observed for any of the variants. The mutations in the E and ORF3 proteins instead increased susceptibility to MPA (Fig. 8E) which is presumably offset by the massive increase in replication and virus production. In hAECs, the triple variant showed no clear replication advantage compared to the WT strain; however, the single variants appeared to modestly reduce infectious progeny production, a deficit that seems to be compensated for in the triple variant (SI Appendix, Fig. S8). In conclusion, these findings suggest that MPA-selected mutations confer altered viral fitness rather than MPA resistance.

Fig. 7.

Analysis of SARS-CoV-2 variant emergence. SARS-CoV-2 was passaged during exposure to different drug combinations. Subsequently, viral RNA was isolated from the supernatant and A549-A/T cells of selected passage and subjected to RNA-seq. (A) SARS-CoV-2 genome coverage for intracellular virus is plotted in correlation to the genome position. (B) SNV frequencies were determined for extracellular virus and (C) variants that exceeded a frequency of 50% in at least one passage are depicted in the heatmaps. The virus used for the initial inoculation was set as a reference. ORF—open reading frame. SNV—single nucleotide variant. E—envelope. S—spike.

Fig. 8.

Characterization of MPA-selected variants. (A) A549-A/T cells were seeded at 8 × 10⁴ cells/mL and infected with SARS-CoV-2 variants (WT, S P812R, ORF3 Q185H, E S6L, and the combined triple variant; 25,000 PFU) for 1 h. After 24 h the supernatant was collected and infectious viral titers were determined by plaque assay. (B) Similarly, A549-A/T cells were infected with the different variants and the number of nucleocapsid expressing cell was quantified by immunofluorescence staining at 8 and 24 h p.i. (C) Virus induced cell death (cytopathic effects) was monitored. Representative images were taken using bright field microscopy. (D) Additionally, representative images of the plaque assay performed for (A) are displayed in a gray scale image. (E) Plaque size was quantified using ImageJ. (F) Sensitivity toward MPA was evaluated by dose dependency assay. WT—wild-type. NP—nucleocapsid protein. PFU—plaque forming units. UTC—untreated control. MPA—mycophenolic acid. IC₅₀—50% inhibitory concentration.

Discussion

Immunosuppression, whether induced by disease or medical treatment, places individuals at an increased risk for severe infections. This vulnerability stems from the compromised state of their immune systems, leaving them less equipped to mount an effective defense against invading pathogens. Numerous studies indicate that immunocompromised individuals, even after vaccination, tend to experience more severe disease upon viral infection (30–33). In the context of the recent SARS-CoV-2 pandemic, it has been reported that approximately 40% of hospitalized individuals with SARS-CoV-2 breakthrough infections after vaccination are immunocompromised (31). Furthermore, the use of antiproliferative agents has previously been associated with a poor humoral immune response (68). Next to potentially more severe disease outcomes, immunosuppression has been intensively discussed concerning viral adaptation and the emergence of novel variants (35–37). Especially in the context of viral infections, and specifically due to the ongoing circulation of SARS-CoV-2, it is thus crucial to enhance our understanding on how immunosuppression impacts viral infection. In line with previous observations, we could demonstrate that the immunosuppressive drug MPA broadly reduces virus infection in vitro. Mechanistically, we provide compelling evidence demonstrating that MPA influences virus replication by depleting intracellular GTP pools. This is substantiated by the dose-dependent restoration of viral titers through the ectopic substitution of guanosine and/or GMP. Conversely, viral titers could not be recovered for the alternative monophosphates, suggesting a direct link between GMP availability and MPA treatment efficacy. Coupled with the observation that individuals undergoing immunosuppression frequently experience prolonged viral replication (35, 40–43), MPA-induced immunosuppression may thus serve as a potential source for the emergence of new variants. Chronic viral infections as caused by HCV (69–71) or HEV (39), but also SARS-CoV-2 (34), have already demonstrated an increased risk of quasispecies development, characterized by heterogeneous viral swarms or mutant clouds. These viral populations, composed of closely related but nonidentical genomes typical for RNA viruses (72, 73), play a crucial role in the adaptation of viruses to environmental changes and selective pressures imposed by the host’s immune system (74). We could demonstrate that increased exposure to MPA treatment led to rapid adaptation of SARS-CoV-2, which, next to viral titer recovery, was accompanied by phenotypic changes resulting in increased plaque sizes. We found that the majority of SNVs occurred at low frequencies and only a very small proportion of these SNVs became fixed in the population, suggesting that most mutations were transient and did not contribute to long-term genomic adaptation. We observed a rapid jump in virus production and intracellular viral RNA levels under MPA-treatment, which coincided with the parallel emergence of three nonsynonymous mutations. We confirmed these mutations confer enhanced virus replication fitness and not MPA resistance. On the virus side, no obvious mutagenic effects or base exchange biases were observed in viral genomes mediated by MPA treatment, outside of the detected mutations. On the host side, MPA-mediated transcriptional dysregulation affected a range of cellular processes. Thus, the exact mechanism underlying MPA-driven mutational emergence remains elusive. However, in line with the evolutionary dynamics of rapidly evolving RNA virus populations, we propose reduced viral population sizes observed under MPA treatment can facilitate the more rapid fixation of mutations in the population. In these restricted populations, adaptive mutations which confer a replication advantage will rapidly become dominant in the absence of immune pressure. Along these lines, SARS-CoV-2 Alpha VOC was hypothesized to have emerged in an immunosuppressed suppressed patient, where chronic infections due to impaired humoral and cellular responses results in greater chances of accumulating fitness-enhancing mutations, and subsequent onward transmission (75).

Specifically, deep sequencing revealed the presence of three commonly arising mutations during adaptation, present in the Spike protein (S P812R), the accessory protein ORF3 (ORF3 Q185H) and the Envelope protein (E S6L), respectively. Interestingly, only the combination of all three mutations increased viral fitness in comparison to the WT virus, whereas the single variants did not affect viral fitness suggesting that they may have complementary roles. This hypothesis is supported by the observation that these mutations, when arising independently as in the UTC, do not affect viral fitness. The linked emergence of these three mutations in the sequencing analysis further suggests a compensatory relationship, indicating they likely function collectively rather than independently. Importantly, susceptibility to neutralizing antibodies as well as approved antiviral treatment regimens remained comparable to the WT. This holds particular significance in light of the growing number of viral variants linked to immune escape and/or diverse pathogenesis (75).

In general, propagation in cell culture correlates with a significant increase in substitutions in the genomic region coding for the Spike protein, which could mediate enhanced infectivity (76, 77). Here, we identified S P812R to be of significant interest. Although reported as a cell culture adaptation conferring a selective phenotypic advantage (78), S P812R had also been reported to occur in vivo (79). An S P812S mutation has been observed in association with upper airway specific evolution in an immunocompromised individual with chronic SARS-CoV-2 infection suggesting a mutational hotspot that also contributes to evasion from neutralizing antibodies through structural rearrangements (80). In fact, S P812 is located close to the S2’ cleavage site. Upon amino acid substitution from proline to arginine the sequence changes from PSKR to RSKR, which corresponds to creation of an additional furin cleavage motif (81). Introducing a second cleavage site could potentially increase membrane fusion and subsequently infectivity thereby compensating for suboptimal conditions (82, 83). Interestingly, the proline at position 812 is also present in SARS-CoV whereas an arginine is found in MERS-CoV (77). Next to the Spike protein, ORF3 has been identified as a driver of pathogenicity affecting virulence, infectivity, ion channel activity, morphogenesis, and virus release (84, 85), as well as a potent interferon antagonist (86). Here, we identified a substitution within ORF3 at position 185 (ORF3 Q185H). Notably, this mutation located in the transmembrane domain of ORF3a has been frequently detected in waste water samples in New York City (87) and is one of the characteristic mutations found in lineage B.1.258.17, which previously circulated in Slovenia, Switzerland, Germany, Sweden, and Austria (88–90). It has also been linked with disease progression (84). Controversially, ORF3 Q185H was additionally found to substantially decrease the stability of the protein (91), adding ambiguity toward its proposed function. The amino acid substitution in the envelope (E S6L) has been reported to occur in vivo (92) while others have reported emergence of this variant during in vitro passaging experiments (36, 93). The envelope protein is the smallest of the major structural proteins, and associated with viral assembly, budding, envelope formation, and pathogenesis. Furthermore, the assembly to pentameric channels directly affects virus replication (94, 95). Currently, there is no evidence indicating specific functional or phenotypic consequences associated with this mutation. However, since all reported mutations have also been observed in vivo, albeit separately and never within the same genome, we still believe that each likely holds biological relevance.

Comparable findings regarding the appearance of viral variants with altered viral fitness or characteristic mutations similar to those displayed by VOCs were documented during the passaging of SARS-CoV-2 in the presence of low doses of Remdesivir (50) and similarly with poliovirus exposed to RBV (46, 49). Our findings also suggest that mutations observed in circulating variants can emerge through in vitro passaging. Consequently, this passaging approach holds significant potential as a valuable tool for assessing the development of viral variants under certain conditions. The emergence of new viral variants presents a risk to the general population by affecting various aspects of virus–host biology such as transmission dynamics, public health measures, vaccine efficacy, antiviral treatment, disease severity, and pathogenesis, among others. Therefore, it is crucial to thoroughly investigate potential drivers of mutagenesis and understand the impact of specific mutations. While immunosuppression has always been a major concern in pathogen-induced diseases, the identification of novel variants raise questions about potential consequences for the long-term management of viral infections in immunocompromised individuals. This is particularly noteworthy not only because SARS-CoV-2 continues to circulate in humans and wild animal reservoirs, posing a constant threat of new variant emergence, but also considering the growing numbers of immunocompromised individuals. Understanding these dynamics becomes crucial in developing more effective and sustainable strategies for treating infections in vulnerable populations. The insights gleaned from our study have the potential to be extrapolated to other virus infections and can contribute to evaluating the pandemic potential of emerging variants in distinct populations.

Methods

Cell Culture.

VeroE6, Vero E6/TMPRSS2 [Provided by M. Takeda, NIBSC/CFAR, Cat. No. 100978 (96)], Huh7, MDCK II, HEp-2, and Vero76 cells were cultured in Dulbeco’s Modified Eagle’s Medium (DMEM) supplemented with 10% (v/v) fetal calf serum (FCS), 1% (v/v) nonessential amino acids (NEAA), 100 IU/mL penicillin, 100 μg/mL streptomycin, and 2 mM L-glutamine. VeroE6/TMPRSS2 were selected with 1 mg/mL Geneticin (G418). ACE2 and TMPRSS2 overexpressing A549 (A549-A/T) cells were cultured in DMEM with 5% (v/v) FCS, 1% (v/v) NEAA, 100 IU/mL penicillin, 100 μg/mL streptomycin, and 2 mM L-glutamine. The cells were additionally selected with Blasticidin (10 µg/mL) and Puromycin (0.5 µg/mL). HepG2/C3A cells were cultured in Eagle’s minimum essential medium (MEM) supplemented with 10% (v/v) ultralow IgG FCS (Gibco, Cat. Nr. 16250-078, Lot 1939770), 2 mM L-glutamine, 100 μg/mL gentamicin, 1 mM sodium pyruvate and 1% (v/v) NEAAs. HepG2/C3A cells were further grown on rat collagen-coated (SERVA Electrophoresis GmbH) cell culture dishes. Human airway epithelial cells (hAEC) were obtained from explanted lungs (Hannover medical school) or from tracheal tissue of healthy donors of lung transplants (Department Clinic for Thoracic and Cardiovascular Surgery, University Hospital Essen). The local ethics committee of the Hannover medical school (3346/2016) and the Medical Faculty of the University Duisburg-Essen as well as the Westdeutsche Biobank Essen (19-8717-BO, 20-WBE-102) approved the use, collection, and storage of lung transplant tissue after informed written consent that was obtained from all patients or legal representatives involved in the study prior to tissue donation. General selection criteria for lung transplant donors are listed in the Eurotransplant guidelines (https://www.eurotransplant.org/wp-content/uploads/2022/02/H1-Introduction-July-28-2016.pdf). Isolated and subsequently differentiated hAECs [as described recently (97, 98)] were cultured in S/D medium [1:1 mixture of DMEM and BEpiCM-b (ScienCell) supplemented with 100 U/mL penicillin and 100 µg/mL streptomycin, 12.5 mL/L of 1 M HEPES, 1× Bronchial Epithelial Cell Growth Supplement (ScienCell) and 5 mM EC-23]. Fully differentiated hAECs were washed with Hank’s balanced salt solution (HBSS) apically before infection.

Ethics.

The experiments involving the adaptation of SARS-CoV-2 to MPA were approved by the local commission for ethics in security-relevant research (KEF).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

RNA-seq data is deposited in the NCBI GEO database (GSE295532) (67). All other data are included in the manuscript and/or SI Appendix.