Highlights of the Crick's response to the COVID-19 pandemic

Through support from the Crick leadership and collaborative working across diverse areas of expertise and STP laboratories, the Crick and its two clinical partners, University College London Hospital (UCLH) and Health Services Laboratories, were able to set up and run a clinically approved RT-PCR assay within three weeks. By August 2020, >100,000 healthcare worker staff and patient samples from nine NHS hospital sites across North London had been tested, protecting patients from pre-symptomatic or asymptomatic transmission from frontline healthcare workers. This pipeline has transformed clinical care across North London hospitals, allowing the creation of COVID-protected hubs for patient care. As of March 2021, in excess of 400,000 swabs have been analysed, and there is broad ethics approval to go back to positive patients for serology and to track vaccine responses, escape mutants and viral evolution in the Crick/UCLH Biomedical Research Centre Legacy Cohort.

The operation involved in excess of 150 Crick staff working in the influenza, TB and malaria groups and many other labs, the advanced sequencing and high throughput STPs, pathology reporting, IT and computing, health and safety, and quality control and quality assurance. Crick scientists repurposed their skills to deliver a clinically validated assay, qualified for clinical use, at a time when there was a global shortage of RNA extraction kits, RT-PCR kits and numerous other materials vital to the pipeline.

There are many examples of the Crick COVID-19 Consortium’s agility, flexibility and innovative approaches. An RNA extraction method previously optimised in the Skoglund group for ancient nucleic acid work was repurposed for use in the testing pipeline; Gamblin and the BRF team developed a SARS-CoV-2 sample reception and viral inactivation laboratory from an empty room in the basement in under a month; the health and safety team supervised and led the most critical phase of the pipeline — the viral inactivation step — in the category 3 facilities at the Crick; and management teams and IT and computing staff came together to work seven days a week to set up a booking and test barcoding hub, and to interface the Crick reporting system with that of the NHS to deliver a 24-hour-turnaround reporting system.

Apart from the obvious clinical and societal benefits of this initiative, there are currently a number of publications resulting from the Crick COVID-19 Consortium’s work. Two are of particular note. Aitken et al (2020) describes how the Crick repurposed itself to set up the rapid testing pipeline, and provides a link to freely available protocols and key documents for other labs seeking to emulate it. The paper is important in that it is a template for the future, describing how excess testing capacity could be rapidly generated from a network of academic research laboratories in times of national emergency.

The evidence of health care worker infection presented in Houlihan et al (2020) contributed to a change in UK government policy. In late March and early April, at the peak of the first wave of the pandemic, samples from two hundred frontline staff at UCLH showed that 44% had evidence of COVID-19 infection, more than twice the rate in the general population in London. Importantly, 38% of staff testing positive by PCR were asymptomatic, and for those that did develop symptoms, the average onset time was four days after a positive test. Of those testing negative by all methods, 13% tested positive within one month of follow-up. This evidence of high rates of infection, of which a significant proportion were asymptomatic, led the government to mandate regular testing for all health care staff, rather than continuing its previous policy of testing only those with symptoms.

In January 2020, before the Crick was able to readily access clinical samples in the UK, Crick group leader Markus Ralser, who holds a joint appointment at the Charité-Universitätsmedizin Berlin, collaborated with his German clinical colleagues to identify a 27-strong biomarker panel in blood samples taken in Berlin from critically ill patients, using the Ralser lab’s advanced high throughput mass spectrometry platform (Messner et al, 2020; discussed in Case Study 9).

Similar work began shortly afterwards at the Crick, in a collaboration co-led by Crick group leader Adrian Hayday and his KCL colleague, intensive care consultant Manu Shankar-Hari (Laing et al, 2020). In a study called COVID-IP (COVID immunophenotyping), the team analysed blood samples from 63 highly heterogeneous patients with SARS-CoV-2 who were treated at Guy’s and St Thomas’ Hospitals, London, comparing them against 55 healthy individuals and 10 patients with non-COVID respiratory tract infections.

To increase opportunities for future global meta-analysis, the group picked simple high-throughput procedures and readily observable target molecules for their study. Immune cell composition was measured using eight multiparameter flow cytometry panels, and further assays were set up to monitor expression of 22 cytokines, and to detect antibodies against the SARS-CoV-2 nucleocapsid, spike and receptor binding domain, and against autoantigens.

Using these methods, a core COVID-19 peripheral blood immune signature was detected, which included discrete changes in B and myelomonocytic cell composition, profoundly altered T cell phenotypes,
selective cytokine/chemokine upregulation and SARS-CoV-2 specific antibodies. Some traits in the signature could be linked with other diseases and vaccine responses in terms of immunoprotection and immunopathology, and others, including basophil and plasmacytoid
dendritic cell depletion, correlated strongly with disease severity.

Most significantly, a third set of traits, not seen in the other respiratory illnesses included in the study, included a triad of IP-10, IL-6, and IL-10 expression which indicated how the disease would progress: patients with COVID-19 who had higher levels of these molecules when first admitted to hospital went on to become more severely ill. By extending the study to an additional 285 patients, the group found that IP-10 levels in particular seemed to be a more robust predictor of disease severity than many other commonly used indicators. IP-10 levels are also high in SARS and MERS, and conceivably, chemokine dysregulation may be a core component of pathogenic coronavirus infections; interestingly, three chemokine receptor genes lie within a region associated with severe COVID-19 respiratory disease susceptibility.

Hayday and his team are now working with public health statisticians in Oxford with two goals in mind: to see if their IP-10 results hold true in further cohorts, and to determine how IP-10 measurement might realistically be integrated into current standard-of-care procedures. Should IP-10 be a suitable prognostic marker, tests for accurately measuring IP-10 levels in the clinical setting are already available for immediate implementation, contingent on official recommendation.

Although no human had encountered SARS-CoV-2 prior to 2019, almost all have been repeatedly infected with the four seasonal human coronavirus (HCoV) types that are endemic in the population. Despite using different host receptors for cellular entry, all HCoVs express a spike protein on their surface comprising two subunits: S1 varies between different viruses and contains the receptor binding domain (RBD), responsible for docking with the respective host receptor, and S2 mediates viral and host cell membrane fusion, and cell entry. Regions of the S2 subunit are highly conserved among seasonal HCoVs, zoonotic coronaviruses, including SARS-CoV-2, and other animal coronaviruses.

Using a sensitive flow cytometric assay to detect serum antibodies, Kassiotis’s graduate student Kevin Ng and his colleagues found that individuals who had never encountered SARS-CoV-2 nevertheless had antibodies that specifically recognised the S2 subunit. These cross-reactive antibodies were class-switched to mature IgG isotypes, suggesting that the B cells producing them had been induced by a previous immune response against a different HCoV. The team also demonstrated that such cross-reactive antibodies exhibited neutralising activity against infectious SARS-CoV-2, despite not targeting the RBD directly. Since then, monoclonal antibodies specific for the SARS-CoV-2 S2 subunit and able to neutralise infectious virus have been isolated by several groups.

Infection rates of seasonal HCoVs are higher in children and adolescents, and cross-reactive antibodies able to recognise the SARS-CoV-2 S2 epitopes were far more prevalent in these groups than in adults. However, the presence of such cross-reactive antibodies is unlikely to be sufficient to prevent SARS-CoV-2 infection completely; antibodies that work against multiple coronaviruses exist in only a few adults at very low levels, and children, who do have more of them, are clearly not protected from catching different varieties of common cold virus. Nevertheless, the presence of such cross-reactive antibodies and memory B cells may modify the course or severity of COVID-19, even if infection is not prevented. Large studies are required to address whether pre-existing immunity from HCoVs can explain the characteristic age or geographical pattern of COVID-19 severity.

The identification of these cross-reactive antibodies opens the possibility of a vaccine that would work against all current and possibly future coronaviruses. This is something the team are currently investigating.

During infection with any respiratory virus, disease severity is linked to lung epithelial destruction, which contributes to acute respiratory distress syndrome, pneumonia and increased susceptibility to bacterial superinfections. Restoring damaged epithelium is therefore paramount to maintain lung function and barrier protection, but it was unknown how IFN responses might affect lung epithelial repair. In August 2020, Andreas Wack and colleagues showed in a mouse model that both Type I and Type III IFNs can aggravate respiratory viral infection if present late during infection when epithelial repair sets in (Major et al, 2020).

Using a combination of mouse influenza models and primary human and mouse airway epithelial cell cultures, PhD student Jack Major and colleagues showed that all three IFN subtypes reduced lung epithelial cell proliferation during influenza recovery, with IFN-β and IFN-λ having the strongest effects; growth inhibition, manifested by necrosis and apoptosis, was only seen in actively dividing cells, suggesting that progression through the cell cycle was failing. However, only endogenous IFN-λ compromised subsequent repair, likely because increased IFN-λ production during infection combined with greater induction of anti-proliferative pathways. This latter was primarily due to p53 being significantly up-regulated by IFN-λ in repairing lung epithelial cells. 

Interestingly, IFN-λ treatment early during influenza virus infection is known to be protective in mice, offering anti-viral protection without the proinflammatory responses associated with IFNα/β, so there may be a role for it in treating SARS-CoV-2 if timing and duration of treatment are carefully monitored. Optimal protection could be achieved by strong induction of IFN-stimulated genes early during infection to curb viral replication followed by timely down-regulation of IFN responses to allow efficient lung epithelial repair.

In a series of important papers, Gamblin’s postdocs Antoni Wrobel and Donald Benton co-led projects in which they described the structure of the SARS-CoV-2 spike (S) protein and compared it to that of its closest relative, a bat coronavirus (Wrobel et al, 2020), and then generated high-resolution cryo-EM structures showing the process by which SARS-CoV-2 S protein unfurls and docks with its receptor, human ACE2 (Benton et al, 2020). Most recently, they have demonstrated that at least part of the enhanced infectivity of SARS-CoV-2 variants carrying the D614G mutation is likely due to a more open spike structure which is primed for binding to ACE2, and perhaps also for membrane fusion (Benton et al, 2021). This in-depth understanding of spike structure not only provides a direct insight into the mechanism of infection but also provides invaluable information for the design of anti-spike therapeutics.

Infection by SARS-CoV-2 is initiated by virus binding to ACE2 cell surface receptors, followed by fusion of the virus and cell membranes to release the virus genome into the cell. As described above, both receptor binding and viral fusion are mediated by the viral spike (S) glycoprotein, which is post-translationally cleaved by furin into S1 and S2 components that remain associated after cleavage. When the SARS-CoV-2 spike protein is compared to that of the bat virus RaTG13, its closest known relative, this furin cleavage site is missing. This is of significance, as similar furin cleavage sites have been found in related coronaviruses, including MERS-CoV, and their acquisition is associated with increased pathogenicity in other viruses such as influenza virus.

Previous structural studies of the spike glycoproteins of other coronaviruses have shown that the spike is a trimer, consisting of a central helical stalk — comprising three interacting S2 components — capped by the S1s. Each S1 component has two large domains, the N-terminal domain and the receptor-binding domain (RBD), with each associated with a smaller intermediate subdomain. Following purification and furin cleavage, Wrobel et al (2020) derived cryo-EM structures for the SARS-CoV-2 spike protein trimer, and found that they could distinguish three conformations: closed, intermediate and open. In the closed conformation, the RBDs were buried, with the ACE2 binding interfaces inaccessible inside the trimer, but in the intermediate form, one of the RBDs was now disordered; this disordered RBD was found rotated by ~60° in the open form so that the ACE2-interacting surface was fully exposed at the top of the assembly.

In furin-cleaved material, a much lower proportion of spike trimers were in the closed conformation, suggesting that as well as its requirement for membrane fusion, efficient cleavage might be selected to ensure there is a higher proportion of spike protein on the virus surface in an open conformation capable of receptor binding. Further evidence for this hypothesis came from a comparison of the cryo-EM structures of the SARS-CoV-2 and RaTG13 spike proteins. Mutations in the RBD-RBD interfaces of the S1 part of the trimer mean that the closed, uncleaved SARS-CoV-2 spike trimer is more stable than that of RaTG13, but once cleaved, the SARS-CoV-2 trimer has similar stability to its bat counterpart.

By using surface biolayer interferometry, the group also measured receptor binding of the SARS-CoV-2 and RaTG13 spikes. Their data showed that when compared to the RaTG13 spike protein, the changes present in SARS-CoV-2 allowed 1000- fold tighter binding to the ACE2 receptor, meaning that the RaTG13 spike would be unable to bind effectively. The RaTG13 virus is therefore unlikely to have ever infected humans directly; rather, the observations suggest the involvement of recombination between distinct coronavirus genomes in the generation of SARS-CoV-2.

Benton et al (2020) built on this work to generate a detailed picture of the binding of the SARS-CoV-2 spike trimer to the ACE2 receptor. Using cryo-EM, they were able to classify ten different molecular species on a spectrum from the unbound, closed spike trimer through the fully open ACE-2 bound trimer, to dissociated monomeric S1 bound to ACE2. Taken together, these structural data enable mechanistic suggestions for the early stages of SARS-CoV-2 infection of cells. 

As shown in Wrobel et al (2020), after furin cleavage between the S1 and S2 domains, the proportion of spike trimers in an open conformation with one erect RBD capable of binding ACE2 increases. Binding of the ACE2 receptor to an open RBD is incompatible with the RBD adopting a closed conformation, leading to several two-open-RBD conformations as well as the three-RBD-bound conformation. Successive RBD opening and ACE2 binding generates the fully open and ACE2-bound form in which the trimeric S1 ring remains bound to the core S2 trimer by limited contacts through the intermediate subdomains of S1. Once in this arrangement, the tops of the S2 helices are fully exposed, and the S trimer is primed for the helical rearrangements of S2 that are required for fusion of the viral and host cell membranes. By analogy with other coronaviruses, the fusion peptide is then projected towards the host cell membrane, a structural change that is driven by the cooperative displacement and dissociation of the three S1 monomers.

This work used the original version of the SARS-CoV-2 spike protein, but during 2020, a mutation, D614G, disseminated rapidly through the population to become the most abundant variant worldwide; this D614G mutation remains in all emerging variants including the highly infectious Kent and South African strains. To determine if there was a structural reason for the variant’s increased infectivity, Benton et al (2021) compared the cryo-EM structures of the G614-substituted spike protein with the D614 original. They found that unlike its progenitor, which is predominantly closed, the G614 spike protein adopts a predominantly open conformation. Therefore, G614 promotes opening of the spike, priming it for binding to ACE2 and possibly for its subsequent role in membrane fusion. This open conformation may be the reason for the current virus’s reported increased infectivity and its current predominance. 

While a worrying development, the G614 spike mutant may have an Achilles heel: during the first wave of infection, most people did not have antibodies to SARS-CoV-2, so the open-structured G614 spike protein could have been beneficial to the virus. However, as time goes on, more people will have antibodies as a result of previous infection or vaccination. As the mutant spike protein presents more surface area, it is more exposed to these antibodies, which could potentially be a disadvantage.