Viral afterlife: SARS-CoV-2 as a reservoir of immunomimetic peptides that reassemble into proinflammatory supramolecular complexes

Significance At present, there are no criteria to evaluate whether a coronavirus can cause pandemics with severe inflammation or just common colds. We provide a possible answer by considering the virus not only as an infectious agent but as a reservoir of replicated peptide motifs that are not themselves pathogen associated molecular patterns (PAMPs) that specifically bind to pattern recognition receptors but are nevertheless capable of drastic immune amplification via self-assembly with PAMPs. We show evidence that viral peptide fragments from SARS-CoV-2 but not harmless coronavirus homologs can “reassemble” with dsRNA into a form of proinflammatory nanocrystalline condensed matter, resulting in cooperative, multivalent immune recognition and grossly amplified inflammatory responses.

To determine whether any SARS-CoV-2 proteins and protein fragments are found in host biofluids and thereby exposed to diverse cell types, and to assess whether such fragments are potentially harmful from their overlap with xenoAMPs, we apply liquid chromatography-tandem mass spectrometry (LC-MS/MS to analyze protein fragments in tracheal aspirate samples from 29 critically ill COVID-19 patients with hypoxic respiratory failure.(P1-P29) We have chosen to use trypsin as the protease used to enable the mass spectrometry measurements, The reason is that trypsin cleavage sites are the same as those from trypsin-like serine proteases already present in the host innate immune system.The general expectation is that some of the highest scoring AMP-like fragments will bind strongly to nucleic acids and lipids, since such binding is related to their antimicrobial mechanism of action.Consequently, the highest scoring sequences may not be observable in isolation.Furthermore, we do not want to bias our analysis of the proteome by pre-selecting the peptides to be detected in tracheal aspirate from ICU patients.Therefore, we have performed bottom-up proteomics analysis rather than bias the proteomics toward the detection of specific peptides "predicted" to be AMP-like by machine learning.Finally, our analysis is performed in data dependent acquisition (DDA) mode: during the chromatographic elution only the most abundant peptides were fragmented to obtain sequence information, which implies that only the most abundant peptides that remain in isolation are detected.We investigate whether any xenoAMPs or fragments of xenoAMPs can be observed under these conditions, and in how many patients.As a comparison, we perform a similar study for LL-37, a host AMP expected to be produced during neutrophilia, one of the inflammatory consequences in COVID-19.
Consistent with previous observations from gargle lavage fluid and nasopharyngeal swabs, SARS-CoV-2 peptide fragments are observed: 7 peptides from the spike protein (sequence coverage (SC) of 7.38%), 3 peptides from the membrane protein (SC of 16.6%), 12 peptides from the nucleocapsid protein (SC of 42.95%).(A) One of the most abundant fragments, 948-964 (LQDVVNQNAQALNTLVK), is a peptide fragment originated from spike protein heptad repeat 1(HR1), which adopts an alpha-helical conformation with a large hydrophobic face on one side and hydrophilic face on the other side.(B-E) Using SAXS, we show that this fragment can remodel membranes the same way that host AMPs do, via generation of negative Gaussian curvature topologically necessary for pore formation.The fragment induces a Pn3m cubic phase with a lattice constant of 17.7nm with a peptide/lipid molar ratio of 1/72 (data not shown).As a calibrating readout, fragments of LL-37, a host AMP generated during the inflammatory process, are found in 20 of the 29 patient samples.In comparison, SARS-CoV-2 peptide fragments are found in 28 of the 29 patient samples, with some of the fragments having high enough scores to be xenoAMPs themselves.(B, C) A comparison of peptide-dsRNA complexes formed using xenoAMP(ORF1ab) only, form using host LL-37 only, and formed using a 1:3 mixture of xenoAMP(ORF1ab) and LL-37.While all complexes exhibit the pro-inflammatory poly(I:C) spacing around 3.3-3.7nm,the mixed xenoAMP-LL-37 complex is the most ordered with additional higher order peaks, with a sharper linewidth and larger domain size.

Fig
Fig. S2.| Sequence alignment of six human coronaviruses.The SARS-CoV-2 ORF1ab protein, spike protein, and membrane protein are aligned with the corresponding proteins from five human coronaviruses, with one highly pathogenic one (SARS-CoV) and four less pathogenic ones (HCoV-HKU1, HCoV-OC43, HCoV-229E and HCoV-NL63).The phylogenetic tree shows the evolutionary relationship between these 6 human coronaviruses.The whole alignments are zoomed in to where xenoAMP(ORF1ab), xenoAMP(S) and xenoAMP(M) are derived.Color is assigned to each residue using the ClustalX color scheme.Each amino acid position is scored between 0-11 to show the degree of conservation.11(+) means fully conserved.10(*) means the properties are conserved.0(-) means not conserved.σ denotes the AMPness.A sequence with large, positive σ is AMP-like sequence, while a sequence with large, negative σ score does not mimic AMP.The number of basic amino acids (Lysine(K), arginine(R) and histidine(H)) and acidic aa (aspartic acid(D) and glutamic acid(E)) are shown in the chart.

Fig. S3 |
Fig. S3 | Mass spectrometry measurement of tracheal aspirate samples from critical COVID-19 patients.To determine whether any SARS-CoV-2 proteins and protein fragments are found in host biofluids and thereby exposed to diverse cell types, and to assess whether such fragments are potentially harmful from their overlap with xenoAMPs, we apply liquid chromatography-tandem mass spectrometry (LC-MS/MS to analyze protein fragments in tracheal aspirate samples from 29 critically ill COVID-19 patients with hypoxic respiratory failure.(P1-P29) We have chosen to use trypsin as the protease used to enable the mass spectrometry measurements, The reason is that trypsin cleavage sites are the same as those from trypsin-like serine proteases already present in the host innate immune system.The general expectation is that some of the highest scoring AMP-like fragments will bind strongly to nucleic acids and lipids, since such binding is related to their antimicrobial mechanism of action.Consequently, the highest scoring sequences may not be observable in isolation.Furthermore, we do not want to bias our analysis of the proteome by pre-selecting the peptides to be detected in tracheal aspirate from ICU patients.Therefore, we have performed bottom-up proteomics analysis rather than bias the proteomics toward the detection of specific peptides "predicted" to be AMP-like by machine learning.Finally, our analysis is performed in data dependent acquisition (DDA) mode: during the chromatographic elution only the most abundant peptides were fragmented to obtain sequence information, which implies that only the most abundant peptides that remain in isolation are detected.We investigate whether any xenoAMPs or fragments of xenoAMPs can be observed under these conditions, and in how many patients.As a comparison, we perform a similar study for LL-37, a host AMP expected to be produced during neutrophilia, one of the inflammatory consequences in COVID-19.

Fig. S4 |
Fig. S4 | Assembly of pro-inflammatory SARS-CoV-2 xenoAMP-poly(I:C) complexes under non-ideal conditions: (A) Variation of peptide length from non-ideal cleavage: Complexes from incubating xenoAMP(M) (RGHLRIAGHHLGRCDIKDLPKEIT, σ=1.19) with poly(I:C) and complexes from the truncated version of xenoAMP(M) (GHLRIAGHHLGRCDIKDLPK, σ=1.32) with poly(I:C), both at the isoelectric peptide to nucleic acid stoichiometry, are measured with SAXS.Strong pro-inflammatory structural signatures are seen in both cases regardless of truncation, with the former having a dsRNA spacing at 3.68nm, and the latter at 3.64nm.(B,C) A comparison of peptide-dsRNA complexes formed using xenoAMP(ORF1ab) only, form using host LL-37 only, and formed using a 1:3 mixture of xenoAMP(ORF1ab) and LL-37.While all complexes exhibit the pro-inflammatory poly(I:C) spacing around 3.3-3.7nm,the mixed xenoAMP-LL-37 complex is the most ordered with additional higher order peaks, with a sharper linewidth and larger domain size.