CONCLUSIONS
Here, we have created a stoichiometric biomass function for the
COVID-19-causing SARS-CoV-2 virus and incorporated this into a human
lung cell genome scale metabolic model. The viral biomass function
highlights the key building blocks required to synthesize a SARS-CoV-2
virus and its simulation within the human metabolic model enables
predicting optimal flux distributions in the host for sustaining either
SARS-CoV-2 reproduction or host maintenance. We used the latter
capability to identify reaction perturbations that can inhibit
SARS-CoV-2 reproduction in general or selectively, without inhibiting
the host metabolic maintenance. The identified reactions primarily fall
onto glycolysis and oxidative phosphorylation pathways, and their
connections to amino acid biosynthesis pathways. The latter finding is
in line with the additional observation we made here, that the relative
stoichiometries of specific amino acids differ in SARS-CoV-2 biomass vs.
host cell maintenance estimated using human lung cell protein expression
data. Together, these results highlight the possibility of targeting
host metabolism for inhibition of SARS-CoV-2 reproduction in human cells
in general and in human lung cells specifically.
The predictions presented here are based crucially on the structure of
the metabolic model as well as the two key assumptions of the flux
balance analysis (FBA); the assumptions of metabolic steady state and
the optimality of metabolic fluxes towards a specific metabolic
function. On the case of the model structure, the RECON2.2 model we used
here presents the most-comprehensive and up-to-date curated human genome
scale model (Thiele 2013; Swainston 2016). This model contains
confidence levels for most included reactions and associated gene
information, which can be improved by future studies updating the model
structure, an area of active development for genome-scale models
(Chindelevitch 2015). In the case of the key assumptions of FBA
(Schuster 2008; Raman 2009), these are expected not to affect
qualitative predictions on which metabolic reactions might be required
to carry flux for a given metabolic process or how specific
perturbations might impact such processes. For example, FBA-based
approaches have been successful in predicting and explaining
experimental observations on gene deletion and environmental
perturbations in both microbial (Ibarra 2002) and eukaryotic systems
(Papp 2004; Gille 2010). In summary, the presented predictions of our
FBA-based approach should be taken as starting points for experimental
studies and to be improved in a systems biology theory-experiment cycle.
The virus life cycle consists of environmental circulation, infection
and subsequent host cell attachment and entry, reproduction within the
host cell, and exit for a new round of infection. The presented approach
focuses solely on the reproduction within the host cell and the
metabolic aspects of that. While this is a limited focus, reproduction
in the host cell is a crucial and essential aspect of the virus life
cycle. The importance of this stage is highlighted in several studies,
which demonstrate that viruses tend to re-program host metabolism for
increased viral production (El-Bacha 2004; Munger 2006; Munger 2008; Yu
2011) or encode enzymes that can participate in host metabolic functions
(Maynard 2010a). These findings show that metabolic basis of host-virus
interaction is crucial for the success of viruses and suggests that such
interaction could be under significant evolutionary selection. Emergent
viruses, such as SARS-CoV-2, are argued to not be well-adapted to their
new host and undergo rapid evolution dictated by host-determined factors
(Simmonds 2019). It has been highlighted, for example, that there is a
codon usage bias in virus genomes that possibly evolve in time to align
with their host (Wong 2010). The presented approach suggests that there
might be a similar adaptation of viruses to their host metabolism. We
argue that differences in metabolic requirements of a virus vs. its host
could be a ‘physiological mismatch’ that contributes to this
evolutionary dynamic. Before metabolic adaptations happen, however,
inhibition of the host metabolism might be a possible strategy to
selectively inhibit reproduction of emergent viruses in new hosts.
Specific host-based metabolic perturbations have already been shown
experimentally to be effective against viruses (Munger 2008; Carinhas
2010), while general perturbation of end-points of nucleotide
biosynthesis through nucleoside analogs underpins the mode of action of
several existing antiviral drugs. The predictions listed here present
possible new antiviral targets that are primarily within central carbon
metabolism, and in particular in glycolysis and oxidative
phosphorylation. There are already several drugs that are shown to
interact with the predicted enzymes in this study, opening up the
possibility of experimentally testing the presented predictions usingin vitro assays and cell cultures.
In the development of host-based metabolic strategies to inhibit
viruses, metabolic modelling, as presented here, can play a useful role.
In particular, our approach can be applied relatively rapidly to any
host-virus pair both for existing and emerging viruses, and allow
generating experimentally testable hypotheses for virus inhibition. This
approach can be applied as long as structural and genomic information
can be converted into an estimation of biomass composition for the virus
and a suitably detailed metabolic model for the host can be constructed.
The former process can be enhanced by further databases of viral
structural and genomic information, while the latter process would
benefit from extending human-focussed efforts such as the human
metabolic atlas database (Robinson 2020) to cover also cell lines of
common animal hosts.