Towards a broad-spectrum antiviral, the myristoyltransferase inhibitor IMP-1088 suppresses viral replication – the Yellow fever NS5 is myristoylated

Although a potent Yellow fever vaccine is available since 1937, up to 200.000 severe cases are reported per year, which indicates that virus vaccines require additional support by antiviral therapies. Direct-acting antiviral drugs against severe and widespread diseases, such as DENV and Yellow fever infections with more than millions of diagnosed diseases per year, are still not available. Since antivirals’ development against neglected diseases is uneconomical, a broadspectrum antiviral compound would be of public benefit. Here, we show that IMP-1088, a recently published myristoyltransferase-1/2 inhibitor suppressing Rhino- and Polioviruses, inhibits replication of HIV-1, Yellow fever virus, Dengue virus, Vaccinia virus, CMV, and human Herpesvirus 8 in the low nanomolar range, indicating that IMP-1088 has broad-range activity against different pathogenic virus families. The inhibition relies on virally encoded myristoylation signals since Zika, Chikungunya, and Enterovirus 71 are not affected by IMP-1088. Furthermore, we show that the Yellow fever NS5 protein is myristoylated and IMP-1088 treatment of Dengue and Yellow fever infected cells leads to a re-localisation of the viral NS5 proteins. Author Summary Treatment of viral diseases requires the development of tailored drugs specific to inhibit certain virus families. This specificity results in missing treatment options for important human pathogens such as Yellow fever and Dengue virus infection since the development is laborious and costly. Substances acting on various virus families could solve this problem. Here, we describe that IMP-1088, an inhibitor of the cellular myristoyltransferase, inhibits HIV-1, Dengue virus, Yellow fever viruses, Vaccinia virus, and Herpesviruses at low concentrations, which do not affect cell proliferation. Viruses without predicated myristoylation sites, such as Zika viruses, were not inhibited by IMP-1088. Since no experimental evidence was provided that Yellow fever virus proteins are myristoylated, we analysed the post-translational modification of Yellow fever NS5 protein. We determined the subcellular localisation to understand the mechanism of the IMP-1088 mediated suppression and could show that both the Dengue and the Yellow fever NS5 proteins are re-localised by IMP-1088 treatment.


Abstract:
Although a potent Yellow fever vaccine is available since 1937, up to 200.000 severe cases are reported per year, which indicates that virus vaccines require additional support by antiviral therapies. Direct-acting antiviral drugs against severe and widespread diseases, such as DENV and Yellow fever infections with more than millions of diagnosed diseases per year, are still not available. Since antivirals' development against neglected diseases is uneconomical, a broadspectrum antiviral compound would be of public benefit. Here, we show that IMP-1088, Thus, effective antiviral therapy is still urgently needed.
Most antiviral drugs used to treat viral infections target viral proteins by inhibiting either virus entry or viral enzymes, such as polymerases, integrase or protease. These treatment strategies can result in rapid viral adaptation leading to the development of viral resistance. It was recently published that IMP-1088, an inhibitor of the cellular myristoyltransferase, will suppress Polio-and Rhinoviruses' replication (7). Interestingly, myristoylated viral proteins are involved in the replication of several human pathogenic viruses at different stages of the viral life-cycle. While Vaccina L1R and A16L proteins of the fusion complex were shown to be myristoylated, myristoylation of the HIV-1 Gag protein is essential for assembly. The requirement of post-transcriptional myristoylation for the replication of RNA and DNA viruses offers the rare possibility to use a single drug active against different and unrelated viruses.
Furthermore, since this inhibitor acts on a cellular enzyme, viral adaption seems to be more unlikely.
Furthermore, we show that the YF NS5 protein is myristoylated and that the recently described substance suppresses the YF and DENV. We provide evidence that this effect is virus-specific since the replication of other viruses, such as Zika, Chikungunya or Rubella viruses, are not affected by IMP-1088.

Yellow fever NS5 is myristoylated
The plus-stranded RNA viruses like DENV, Poliovirus, and YF encode a single open reading frame, which is structured in the coding regions of the structural proteins and the coding regions of the non-structural (NS) proteins. In Polioviruses, the start codon used to translate this precursor protein is followed by a glycine residue, which is myristoylated and finally leads to a modified VP4 protein. The DENV and YF precursor proteins lack this myristoylation site.
Cellular proteases and the viral protease are responsible for cleaving the viral precursor polyprotein into functional proteins. The viral precursor protein processing is essential for viral infectivity and results in functional subunits with predicted myristoylation sites in the NS5 protein (8). Thus, IMP-1088 might inhibit the viral replication of other viruses too.
It has recently been suggested that the NS5 proteins of DENV and YF could be myristoylated, but neither the NS5 myristoylation nor its influences have been determined so far. Thus, we analysed whether the amino terminus of the YF proteins contains a potential myristoylation site using bioinformatics (9), and the N-terminal amino acid sequence of NS5 was predicted representing myristoylated signals with an ExPASy score of 0.935. This led to the hypothesis that the Yellow fever precursor protein could be cleaved by viral and cellular proteases, and subsequently, the NS5 protein could be myristoylated. HEK293T cells were transfected with a codon-optimised NS5-expression plasmid (pCDNA-3.4-YFsynNS5-FLAG), encoding a FLAGepitope at the carboxyl terminus to demonstrate that NS5 is myristoylated. First, the NS5 expression was visualised with an anti-FLAG-antiserum by Western blotting. Next, the transfected cells were lysed, and NS5 was affinity purified with M2-Flag-Resin. The eluted protein was visualised on the coomassie stained PAGE and excised. The protein was digested with either Elastase, Thermolysin, Trypsin or Trypsin+10%Acetonitril to obtain overlapping peptides, which were subjected to mass spectrometry. We found that the glycine residue at the NS5 cleavage site is myristoylated or acetylated ( Figure 2). These results indicated that YF might be sensible to IMP-1088 as well.

IMP-1088 suppresses YF replication
Before analysing the effects of IMP-1088 on the YF replication, the influence of the compound on cellular growth was determined. Vero cells were seeded in optical 96 well plates. On the next day, cell numbers per individual well were determined. Afterwards, cells were incubated with 500nM, 100nM, or 10nM of IMP-1088 for two days, and again all cells per well were counted. We did not observe any influence of IMP-1088 on cellular growth. Next, Vero cells were incubated with either IMP-1088 or DMSO as control and subsequently infected with the YF. Cellular supernatants were collected two days after infection, viral RNAs were isolated, and RTqPCR determined genome copy numbers with primers and probes described before (10). The efficiency of the RTqPCR was determined with RNA dilutions rows. Viral copy numbers were quantified with a synthetic standard of known concentration. IMP-1088 decreased the number of viral genome copies at 100nM and 10nM by two or one orders of magnitude, respectively (Figure 3), indicating that IMP-1088 is a potent inhibitor of YF replication. The EC50 was determined with 11.3 ± 1.9 nM.

IMP-1088 changes the subcellular localisation of YF NS5
Since myristoylation of NS5 might influence membrane targeting and, thereby, the subcellular localisation of the protein, microscopic analyses with an NS5-specific antibody were performed. Vero cells were seeded on microscopic slides, infected with YF, and fixed with paraformaldehyde two days after infection. Staining with the NS5 specific antibody revealed that YF NS5 was localised in the cytoplasm and nucleus in infected Vero cells. In contrast, NS5 in YF infected cells treated with 100nM IMP-1088 was localised almost exclusively in the nucleus ( Figure 3). This indicates that the cytoplasmic localisation of the YF NS5 protein is myristoylation dependent. In order to show that no other viral protein is required for the IMP-1088 dependent re-localisation, a FLAG-tagged codon-optimised NS5 was synthesised.
HepG2 cells were seeded on microscopic slides and transfected with pcDNA3.4-synNS5-FLAG. Cells were fixed and permeabilised two days after transfection. The subcellular localisation was visualised with a rabbit-anti-FLAG antibody ( Figure 3). Again, the NS5 protein was localised in the cytoplasm and nucleus. Treatment with IMP-1088 resulted in relocalisation of the NS5 protein to the nucleus showing that no other viral protein is required for the change in subcellular localisation. These results provide evidence that the NS5 localisation is dependent on the myristoylation. Furthermore, the non-myristoylated form of NS5 seems to be localised almost exclusively in the nucleus.

IMP-1088 suppresses Dengue viruses specifically and leads to re-localisation of NS5
Since YF responded so well to IMP-1088, we sought to analyse other plus-stranded RNA viruses, such as Zika and Dengue. In similar experiments, Vero cells were incubated with IMP-1088 and infected with Dengue virus serotype 2 (DENV2). After 3 days, viral RNAs were isolated and quantified by RTqPCR. 100nM IMP-1088 suppressed viral loads about one order of magnitude, while 10nM of IMP-1088 reduced viral genome amounts to 45%. This indicated that DENV2 is susceptible to IMP-1088 ( Figure 4). In addition, the EC50 values for DENV2 (0.15 ± 0.14 nM) were calculated.
Previously it was shown that the subcellular localisation of the DENV2-NS5 protein is controlled by a short stretch of amino acids at the C-terminus and that NS5 shuttles between the cytoplasm and the nucleus. This prompted us to analyse whether the IMP-1088 would
In summary, we provide evidence that Vaccinia virus, mCMV and HHV-8 need cellular myristoyltransferase for optimal replication in cell culture. However, the effect of IMP-1088 is less pronounced compared to the plus-stranded RNA viruses.
In general, usually, antiviral therapies are virus-specific and target viral enzymes. This sets viruses on an evolutionary pressure and leads to the selection of viral escape mutants. The resistances associated mutants have been extensively studied for HIV-1, where in some cases, more than 20% of the PR encoding amino acid residues were exchanged during antiviral therapy. Targeting of the cellular myristoyltransferase 1 and 2 would block in case of HIV two essential steps, Gag assembly at the plasma membrane and Nef function. It is unlikely how the virus could adapt since the functions of two critical protein are affected. In this regard, IMP-1088 is different from previously used antivirals which inhibited single viral proteins.
Furthermore, the inclusion of IMP-1088 in combination therapies would provide an additional treatment option for the patient all therapeutic options exhausted and for HHV-8 seropositive HIV-1 infected patients.
The Flaviviral NS5 protein is highly conserved and multifunction. Here, we presented evidence that YF NS5 is myristoylated and that myristoylation is required for viral infectivity of Dengue and Yellow fever viruses. This is supported by the IMP-1088 resistance of Zika viruses, which lack any potential myristoylation site in its sequence. This result indicates that the inhibition of myristoylation of viral DENV and YF NS5 proteins is responsible for the antiviral effect of IMP-1088. Similar results were obtained in previous experiments with myristoyl transferase 1knockdown cells showing that NMT activity is required for DENV replication (16). The negative impact on other RNA viruses provides evidence that the observed effect is virus-specific, not related to the inhibition of general cellular factors.
Flaviviruses replicate in the cytoplasm. Thus, it is likely that viral proteins are localised there.
However, previous reports and we described localisation of the YF NS5 protein in both the nucleus and the cytoplasm (17). The inhibition of the myristoylation with IMP-1088 resulted in almost exclusive nuclear localisation, indicating that the cytoplasmic YF NS5 might interact with cellular membranes. In contrast, the NS5 protein of the DENV Serotype 2 encodes functional nucleolar localisation signals and is almost exclusively localised in the nucleus (18,19). Treating cells with IMP-1088 led to the nuclear exclusion of DENV-2 NS5 protein. The nucleolar localisation contributes to the dysregulation of the host-translation and splicing, which enhances virus replication, and importin inhibitors led to reduced virus replication (20,21). This could explain the observed reduction of DENV-2 titers in IMP-1088 treated cells.
In summary, inhibition of the cellular myristoyltransferase can specifically and efficiently suppress the replication of RNA and DNA viruses. After two days, the cells were fixed with methanol/acetone (1:1), extensively washed with PBS and ß-Galactosidase activity was visualised with X-Gal substrate. The blue-stained infected cells were counted, and viral titers were calculated. Viral RNA isolation and RTqPCR: Viral RNAs were isolated with the viral NA isolation kit (Roche, Germany) or with the MagNa Pure 24 NA isolation device according to the manufacturer's instructions (Roche). Genomes quantification of YF was done as described before (10). All other viruses were quantified with LightMix Assays (Roche). The RNA Master Probe kit was used for amplification as described by the manufacturer (Roche). PCR-Setup was performed with the Liquid Handling Station (BRAND, Germany) in triplicate assays. All PCRs were performed using either the LightCycler96 or 480II (Roche). Quantifications were performed with the respective cycler software. EC50 values were calculated using GraphPad Prism.

Viruses
Immunofluorescence and Transfection: Vero cells were seeded on coverslips, incubated with IMP-1088 and infected with DENV2 or YF. Cells were fixed with 5% PFA and permeabilised with 0.25% Triton X in PBS. Unspecific binding of the antibodies was blocked with 3% bovine serum albumin. Viral NS5 proteins were visualised with an NS5 specific antiserum (Biozol, Germany) and a Cy3 coupled donkey anti-rabbit IgG antibody (Dianova, Germany). HepG2 cells were transfected with the pcDNA3.4-synNS5-FLAG. The localisation of the codon-optimised NS5 protein was detected with a rabbit-anti-FLAG antibody (Sigma, Germany).