The discovery may help explain why some COVID-19 patients can become severely ill despite having high levels of antibodies against the virus.
In their research, published in Science Advances today (April 22, 2021), teams from the Francis Crick Institute, in collaboration with researchers at Imperial College London, Kings College London and UCL (University College London), found that biliverdin and bilirubin, natural molecules present in the body, can suppress the binding of antibodies to the coronavirus spike.
As vaccines are rolled out globally, understanding immunity to SARS-CoV-2 and also how the virus evades antibodies is critically important. However, there are still many unknowns.
The ability of the immune system to control the infection and the quality of the antibody response are highly variable, and not well correlated, between individuals.
The Crick researchers were involved in the development of tests that see if a person has been exposed to the virus. The scientists discovered that the SARS-CoV-2 spike protein strongly binds to biliverdin, a molecule which was giving these proteins an unusual green colouration.
Working with teams at Imperial College London, UCL and Kings College London, they found that this natural molecule reduced antibody binding to the spike. They used blood sera and antibodies from people who were previously infected with SARS-CoV-2 and found that biliverdin could suppress the binding of human antibodies to the spike by as much as 30-50%, with some antibodies becoming ineffective at neutralizing the virus.
Such a significant impact was completely unexpected, as biliverdin only binds to a very small patch on the virus’ surface. To find out the mechanism at work, the team at the Crick used cryo-electron microscopy and X-ray crystallography to look in detail at the interactions between the spike, the antibodies and biliverdin.
They found that biliverdin attaches to the spike N-terminal domain and stabilizes it so that the spike is not able to open up and expose parts of its structure.
This means that some antibodies are not able to access their target sites and so cannot bind to and neutralize the virus.
Annachiara Rosa, first author and postdoctoral training fellow in the Chromatin structure and mobile DNA Laboratory at the Crick, says: “When SARS-CoV-2 infects a patient’s lungs it damages blood vessels and causes a rise in the number immune cells. Both of these effects may contribute to increasing the levels of biliverdin and bilirubin in the surrounding tissues.
And with more of these molecules available, the virus has more opportunity to hide from certain antibodies. This is a really striking process, as the virus may be benefiting from a side effect of the damage it has already caused.”
Peter Cherepanov, author and a group leader of the Chromatin structure and mobile DNA Laboratory at the Crick, says: “In the first months of the pandemic, we were extremely busy churning out viral antigens for SARS-CoV-2 tests. It was a race, as these tests were urgently needed.
When we finally found the time to study our green proteins, we expected a mundane answer. Instead, we were astonished to discover a new trick the virus uses to avoid antibody recognition. This is a result of a collaborative effort of several amazing teams working at the Crick and three partner universities, led purely by scientific curiosity.”
The researchers will continue this work from various angles, including measuring the levels of biliverdin and other haem metabolites in patients with COVID-19 and also exploring if it is possible to hijack the binding site used by biliverdin to potentially find new ways to target the virus.
Trimeric coronaviral spike glycoproteins form prominent features on viral particles that are responsible for the attachment to a receptor on the host cell and, ultimately, fusion of the viral and cellular membranes (1, 2). Encoded by a single viral gene, the mature spike glycoprotein comprises two subunits, S1 and S2, which mediate binding to the receptor and facilitate fusion, respectively. The recognition of the betacoronavirus SARS-CoV-2 host receptor, the cellular membrane protein angiotensin-converting enzyme 2, maps to the S1 C-terminal domain (referred to as the receptor binding domain, RBD) (3–5), while the function of the N-terminal domain (NTD) remains enigmatic. Both domains can be targeted by potent neutralizing antibodies that arise in infected individuals. The majority of characterized neutralizing antibodies bind the RBD, while minimal structural information exists about neutralizing epitopes on the NTD (6–10). The immune properties of the spike glycoprotein underpin ongoing SARS-CoV-2 vaccine development efforts (11).
In the course of our activities to support the development of serology for SARS-CoV-2, we produced a range of recombinant coronaviral spike antigens by expression in human cell lines (Fig. S1A). Surprisingly, preparations of SARS-CoV-2 trimeric spike and S1 carried a distinctive green hue, with prominent peaks at ~390 and 670 nm in their light absorbance spectra (Fig. S1A–B). These unusual features were also evident in the spectrum of S1 from the 2003 SARS-CoV-1 isolate, but not those from the seasonal human coronaviruses NL63 and OC43 (Fig. S1B–C).
The property was confined within the spike NTD and absent in isolated RBD (Fig. S1B). The spectra of the SARS-CoV spike constructs were consistent with biliverdin (Fig. S1B), a product of haem metabolism responsible for the coloration of bruises and green jaundice. We isolated the pigment from denatured SARS-CoV-2 S1 and confirmed the presence of biliverdin IXα by mass spectrometry (Fig. S2). Biliverdin is produced at the first step of haem detoxification by oxygenases and is then reduced to bilirubin, the final product of tetrapyrrole catabolism in humans (12).
We measured tetrapyrrole binding to immobilised SARS-CoV-2 S1 using surface plasmon resonance (SPR) and estimated the dissociation constant (Kd) for the interaction with biliverdin and bilirubin at 9.8 ±1.3 nM and 720 ±250 nM, respectively (Fig. S3A–B, Table S1). SARS-CoV-1 S1 likewise displayed nanomolar affinity for biliverdin (Fig. S3I, Table S1). The spike bound haem considerably more weakly, with Kd of 7.0 ±1.2 μM, while no interaction was observed with protoporphyrin IX (Fig. S3C–D, Table S1).
Next, we imaged single particles of the trimeric SARS-CoV-2 spike ectodomain (3, 13) in the presence of excess biliverdin using cryo-electron microscopy. Image processing resulted in the 3D reconstruction of closed (3RBDs-down) and partially open (1RBD-up conformation) states of the spike at 3.35 and 3.50 Å resolution, respectively (Fig. 1A, Fig. S4, Table S2). Close inspection of the cryo-EM maps revealed features interpretable as a biliverdin molecule buried within a deep cleft on one side of each of the NTD domains (Fig. 1A, Fig. S5A).
To define the structural basis for the interaction more precisely, we co-crystallised the isolated NTD with biliverdin and determined the structure at 1.8 Å resolution (Fig. 1B, Fig. S5B, Table S3). The metabolite fits snugly into the cleft with the pyrrole rings B and C buried inside and propionate groups appended to rings A and D projecting toward the outside. The pocket is lined by hydrophobic residues (Ile101, Trp104, Ile119, Val126, Met177, Phe192, Phe194, Ile203, and Leu226), which form van der Waals interactions with the ligand. Biliverdin packs against His207, which projects its Nε2 atom towards pyrrolic amines, approaching three of them at ~3.6 Å. Pyrroles A and B are involved in a π-π stacking with side chain of Arg190, which is stabilised by hydrogen bonding with Asn99.
The binding of biliverdin largely buries the side chain of Asn121, which makes a hydrogen bond with the lactam group of pyrrole D. In agreement with the extensive interactions observed in the crystal structure, the melting point of isolated NTD increased by over 8°C in the presence of biliverdin (Fig. S3K). Unidentified entities at the tetrapyrrole binding site were observed in published SARS-CoV-2 spike reconstructions (1, 4, 13–17), presumably obtained with partial occupancy by the metabolite; in some cryo-EM maps, the bound biliverdin molecule is resolved remarkably well (Fig. S6) (14–17).
Structures of SARS-CoV-2 spike-biliverdin (A,B) and spike-P008_056 Fab (C) complexes.
(A) Cryo-EM 3D reconstructions of trimeric SARS-CoV-2 spike ectodomain in 3RBD-down (left) and 1RBD-up (right) conformations determined under saturation with biliverdin. Spike protomers are color-coded. Biliverdin and glycans are shown in green and grey, respectively. (B) Details of the biliverdin binding pocket in the crystal structure. SARS-CoV-2 NTD is shown as cartoons with selected amino acid residues and biliverdin in sticks. Carbon atoms of the protein chain, sugars (NAG), and biliverdin are in purple, grey and green, respectively; the remaining atoms are coloured as follows: oxygen, red; nitrogen, blue; and sulphur, yellow. Dark grey dashes are hydrogen bonds.
The presence of a histidine residue in the biliverdin binding pocket (Fig. 1B) suggested that the interaction may be pH-dependent. In agreement with this hypothesis, the Kd of the S1-biliverdin interaction increased to 250 ±100 μM at pH 5.0 (Fig. S3E,H and Table S1), and purification under acidic conditions greatly reduced the biliverdin content of recombinant SARS-CoV-2 S1 (Fig. S1D).
Substitutions of spike residues closely involved in ligand binding (H207A, R190K and N121Q) diminished pigmentation of purified recombinant protein (Fig. S1E). The biliverdin binding affinity of SARS-CoV-2 S1 was reduced by two and three orders of magnitude by the R190K and N121Q amino acid substitutions, respectively (Fig. S3F–H, Table S1). The latter ablated the interaction with bilirubin, confirming that the tetrapyrroles share the binding site on the spike (Fig. S3H). SARS-CoV-2 spikes carrying these mutations supported infection of Vero and Huh7 cells by a pseudotyped retroviral vector, indicating that biliverdin binding is not required for viral entry under tissue culture conditions (Fig. S7).
Because biliverdin binding conceals a deep hydrophobic cleft on the NTD (Fig. 1B), we suspected that it may mask or modify the antigenic properties of the viral spike. To test this hypothesis, we measured the reactivity of sera from 17 SARS-CoV-2-infected and convalescent individuals with full-length WT and N121Q SARS-CoV-2 spike using a flow cytometry-based assay (18). Remarkably, addition of 10 μM biliverdin reduced binding of patient IgGs to WT spike, supressing the reactivity of some of the immune sera by as much as 50% (Fig. 2).
By contrast, antibody binding to N121Q SARS-CoV-2 spike was not affected by addition of biliverdin (Fig. 2). Binding of IgM and IgA antibodies, which are present at lower titres in these patients (18), was more variably affected (Fig. S8B–D). In a separate experiment, we tested 91 clinical serum samples in an IgG capture enzyme-linked immunosorbent assay (ELISA). SARS-CoV-2 specific antibodies were detected using WT or N121Q S1 conjugated to horseradish peroxidase. Biliverdin-depleted WT S1 was significantly more reactive than the same protein supplemented with excess metabolite (Fig. S9). By contrast, addition of biliverdin did not dampen detection of the SARS-CoV-2 antibodies with N121Q S1 (Fig. S9).
Biliverdin strongly downmodulates the reactivity of SARS-CoV-2 spike with antibodies present in immune sera.
Left: Mean fluorescence intensity (MFI) of IgG staining of HEK293T cells expressing full-length WT or N121Q SARS-CoV-2 spike by individual patient sera in the absence or the presence of 10 μM biliverdin. Each symbol represents an individual patient (n=17) and coloured dotted lines represent the linear regression for each spike variant. The inset shows posterior probability density plots of values for pairwise contrasts (±biliverdin) for the WT and N121Q spikes. Black dots indicate the median of the distribution, thick and thin line ranges correspond to the 85% and 95% highest density interval, respectively; the dotted vertical line indicates a zero difference. Right: Changes in MFI caused by the addition of 10 μM biliverdin, as percent of staining without biliverdin, for serum for IgM and IgA antibodies. Each pair of connected symbols represents an individual patient.
It is remarkable that a small molecule with a footprint of 370 Å2, corresponding to only ~0.9% of solvent-exposed surface (per spike monomer, Fig. 1A), competes with a considerable fraction of the spike-specific serum antibody population (Fig. 2). These results prompted us to evaluate a panel of human antibodies cloned from B cells of SARS-CoV-2 convalescent individuals.
We used 38 IgGs reported in a recent study (10), as well as a panel of 15 novel SARS-CoV-2 S1-specific IgGs obtained from individuals with asymptomatic infection, or mild/severe disease undergoing characterization in one of our laboratories (Graham et al., in preparation). We tested these monoclonal antibodies for binding to recombinant WT and N121Q S1 by ELISA. Remarkably, 9 of 53 (17%) IgGs lost binding to WT, but not to N121Q S1, in the presence of 10 μM biliverdin (Fig. 3A,,C,C, Fig. S10A). Furthermore, addition of biliverdin strongly suppressed binding of these antibodies to full-length WT but not N121Q SARS-CoV-2 spike expressed on the surface of transfected HEK293T cells (Fig. 3B,,D).D). By contrast, reactivity of both RBD-specific controls (COVA1–18 and COVA1–12), was not affected by the metabolite (Fig. 3A–D).
Biliverdin decreases binding to SARS-CoV-2 spike by a group of human monoclonal IgGs.
(A) Antibodies were titrated 6-fold and assayed by direct ELISA for binding to recombinant S1 biliverdin-depleted by purification under acidic conditions (−biliverdin), same protein but supplemented with biliverdin (+biliverdin) or N121Q S1. Area under the curve (AUC) is shown for IgG that were sensitive to biliverdin and two unaffected control IgGs. AUC values are colour-coded as per the key; fold change compared to WT protein are reported. (B) Biliverdin-sensitive IgGs were titrated 10-fold and incubated with 293T cells expressing full-length WT or N121Q SARS-CoV-2 spike with or without 10 μM biliverdin. Binding was detected using an anti-IgG antibody and reduction in binding in the presence of biliverdin is shown as % MFI reduction and colour-coded as a heatmap of the quartile values. (C) ELISA titration curves for four neutralizing IgG including the biliverdin-insensitive control COVA1–18. (D) Relative MFI dose-dependent curves for four neutralizing IgG including the biliverdin insensitive control COVA1–18. Relative MFI calculated by normalising to the MFI of the biliverdin-insensitive COVA1–18 at the highest concentration against spike. (E) IgG indicated above each graph were titrated 5-fold against SARS-CoV-2 spike pseudotype, in the presence and absence of 10 μM biliverdin, and a version of spike encoding the mutation N121Q. COVA1–18 was used as a biliverdin-insensitive control IgG. (F) Neutralization of SARS-CoV-2 (England 02/2020/407073) by IgGs was measured in the absence and presence of 10 μM biliverdin in Vero-E6 cells. P003_027 was used as a biliverdin-insensitive control IgG.
Two of the biliverdin-sensitive monoclonal IgGs, COVA1–22, COVA2–17, were previously reported to efficiently neutralize SARS-CoV-2 pseudotyped retrovirus (10). Addition of biliverdin suppressed neutralization of the pseudotype carrying WT but not N121Q spike (Fig. 3E). The mutation had a differential effect on neutralization showing a decrease in potency for COVA1–22 and no effect for P008_056 or COVA2–17.
As expected, addition of biliverdin had no effect on neutralization by COVA1–18. Neutralization of replication-competent SARS-CoV-2 by P008_056, COVA1–22 or COVA2–17 was substantially decreased by addition of biliverdin (Fig. 3F). Intriguingly, while P008_056 seemed somewhat less potent in the pseudotype assay (Fig. 3E), it efficiently neutralized live SARS-CoV-2 virus, achieving 50% and >90% inhibition at concentrations of 0.03 and 1.56 μg/ml, respectively, in a biliverdin-sensitive manner (Fig. 3F, Fig. S10B).
To establish the structural basis for SARS-CoV-2 neutralization by a biliverdin-sensitive antibody, we imaged single particles of the viral spike in complex with P008_056 antigen binding fragment (Fab). Cryo-EM image processing resulted in reconstruction of structures with one, two and three Fab moieties bound per trimeric viral glycoprotein (Fig. S10A–B), and the best map was obtained for the complex containing a single Fab (Fig. 4A, Fig. S11C–D).
The reconstruction with a local resolution of ~4 Å at the NTD-Fab interface allowed for unambiguous tracing of the protein backbone and revealed positions of key amino acid side chains (Fig. 4A, Fig. S5C). P008_056 binds the spike at the side of the NTD β-sandwich fold, which undergoes profound conformational rearrangements (Movie S1). Access to the epitope is gated by a solvent-exposed loop composed of predominantly hydrophilic residues (“gate”, SARS-CoV-2 spike residues 174–188; Fig. 1).
To allow P008_056 binding, the loop swings out of the way, with a backbone displacement in the middle of the loop of ~15 Å (Fig. 4B). The gating mechanism is accompanied by insertion of Phe175 and Met177, which are located in the beginning of the loop, into the hydrophobic pocket vacated by biliverdin (Fig. 4B, Fig. S5C).
Thus, when bound, the metabolite appears to act as a wedge that restricts gate opening. Antibody binding is additionally complemented by an upward movement of a β-hairpin (“lip”, SARS-CoV-2 residues 143–155), which overlays a cluster of aromatic residues (Fig. 1, Fig. 4B). The marked gain in thermal stability upon biliverdin binding (Fig. S3K) is consistent with resistance to an antibody that requires major conformational remodeling of the NTD for binding (Movie S1).
Cryo-EM structure of the spike-Fab complex.
(A) Reconstruction obtained with multibody refinement in Relion (left) and a zoom on the spike-Fab interface in the structure obtained by consensus refinement (Fig. S11d). (B) Refined model of the spike-Fab complex shown as cartoon, with selected amino acid side chains in sticks and indicated. Carbon atoms of the gate and lip NTD elements that relocate to allow Fab binding (arrows), are shown in black. Fab heavy (HV) and light (LV) chains are shown in blue and beige, respectively.
It is well-established that viruses employ extensive glycosylation of their envelopes to shield antibody epitopes from recognition by humoral immunity (19–21). Here, we identified and structurally characterised a novel class of a neutralizing epitope, present on SARS-CoV-2 spike, which is differentially exposed through recruitment of a metabolite. In contrast to glycosylation, co-opting a metabolite may allow conditional unmasking, for example under acidic conditions within the endosomal compartment. Biliverdin levels in plasma of healthy individuals (0.9–6.9 μM) and more so under pathological conditions (>50 μM) (22) greatly exceed the Kd of its interaction with the spike (~10 nM) and are therefore sufficient to affect SARS-CoV-2 antigenic properties and neutralization.
Although SARS-CoV-2 spike bound bilirubin with lower affinity (Fig. S3), this final product of haem catabolism accumulates at higher levels in vivo (22). Moreover, elevated bilirubin levels correlate with the symptoms and mortality among COVID-19 patients (23–26). Therefore, the tetrapyrroles likely share a role in SARS-CoV-2 immune evasion. Severe COVID-19 symptoms and death are associated with neutrophil infiltration in pulmonary capillaries and alveolar space (27).
Indeed, nasopharyngeal swabs of COVID-19 patients are enriched in neutrophil myeloperoxidase (28), a highly abundant haem-containing protein responsible for coloration of mucus (29). Alongside extensive vascular damage, these symptoms provide rich source of haem catabolites, which may contribute to the inability to control the infection in severe cases. Mutations within the SARS-CoV-2 spike NTD are associated with viral escape from antibody immunity (30, 31) and have been observed in circulating viral strains (32, 33). Our results demonstrate a remarkable structural plasticity of the NTD and highlight the importance of this domain for antibody immunity against SARS-CoV-2.
reference link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7852234/
More information: SARS-CoV-2 can recruit a haem metabolite to evade antibody immunity, Science Advances (2021). DOI: 10.1126/sciadv.abg7607 , advances.sciencemag.org/conten … 04/22/sciadv.abg7607
