Toxin-like peptides in plasma, urine and faecal samples from COVID-19 patients

Background: SARS-CoV-2 that causes COVID-19 disease and led to the pandemic currently affecting the world has been broadly investigated. Different studies have been performed to understand the infection mechanism, and the involved human genes, transcripts and proteins. In parallel, numerous clinical extra-pulmonary manifestations co-occurring with COVID-19 disease have been reported and evidence of their severity and persistence is increasing. Whether these manifestations are linked to other disorders co-occurring with SARS-CoV-2 infection, is under discussion. In this work, we report the identification of toxin-like peptides in COVID-19 patients by application of the Liquid Chromatography Surface-Activated Chemical Ionization – Cloud Ion Mobility Mass Spectrometry. Methods: Plasma, urine and faecal samples from COVID-19 patients and control individuals were analysed to study peptidomic toxins’ profiles. Protein precipitation preparation procedure was used for plasma, to remove high molecular weight proteins and efficiently solubilize the peptide fraction; in the case of faeces and urine, direct peptide solubilization was employed. Results: Toxin-like peptides, almost identical to toxic components of venoms from animals, like conotoxins, phospholipases, phosphodiesterases, zinc metal proteinases, and bradykinins, were identified in samples from COVID-19 patients, but not in control samples. Conclusions: The presence of toxin-like peptides could potentially be connected to SARS-CoV-2 infection. Their presence suggests a possible association between COVID-19 disease and the release in the body of (oligo-)peptides almost identical to toxic components of venoms from animals. Their involvement in a large set of heterogeneous extra-pulmonary COVID-19 clinical manifestations, like neurological ones, cannot be excluded. Although the presence of each individual symptom is not selective of the disease, their combination might be related to COVID-19 by the coexistence of the panel of the here detected toxin-like peptides. The presence of these peptides opens new scenarios on the aetiology of the COVID-19 clinical symptoms observed up to now, including neurological manifestations.


Introduction
Numerous clinical extra-pulmonary manifestations co-occurring with COVID-19 disease have been reported (e.g. neurological, haemorrhagic, and thrombotic) and evidence of their severity and persistence is increasing. Gupta et al. reviewed the extrapulmonary organ-specific pathophysiology of patients with COVID-19, 'to aid clinicians and scientists in recognizing and monitoring the spectrum of manifestations, and in developing research priorities and therapeutic strategies for all organ systems involved' 1 . Liotta et al. characterized the incidence of neurological manifestations in a cohort of hospitalised patients with confirmed COVID-19: the most frequent were myalgia, headache, encephalopathy, dizziness, dysgeusia, and anosmia; encephalopathy was found to be 'associated with increased morbidity and mortality, independent of respiratory disease severity' 2 . Whether these manifestations are linked to disorders co-occurring with SARS-CoV-2 infection is under discussion, including their concomitant occurrence, which could be strongly related COVID-19 disease. Frontera et al., by conducting a prospective, multi-centre, observational study of hospitalised adults with laboratory-confirmed SARS-CoV-2 infection, concluded that 'neurologic disorders were detected in 13.5% of COVID-19 patients during the study timeframe. Many of these neurologic disorders occur commonly among patients with critical illness. Encephalitis, meningitis or myelitis referable to SARS-CoV-2 infection did not occur, though post-infectious Guillain-Barre syndrome was identified. Overall, neurologic disorders in the context of SARS-CoV-2 infection confer a higher risk of in-hospital mortality and reduced likelihood of discharge home' 3 .
Studies on the use of mass spectrometry in COVID-19 context focus on the search for augmented human inflammatory molecules to be used as biomarkers to assess the severity status of COVID-19 (see for example the work 4 of Messner and colleagues). Different studies report the use of proteomic approaches to characterise SARS-CoV-2 proteins 5-7 . Other studies highlight challenges in their use due to the need of enriching the protein fraction to be analysed for maximizing the technology sensitivity 8 .
Liquid Chromatography Surface-Activated Chemical Ionization -Cloud Ion Mobility Mass Spectrometry (LC-SACI-CIMS) is reported as a high sensitivity mass spectrometry technique able to maximize the peptide signal intensity 9-12 . We used LC-SACI-CIMS to reveal the presence of metabolites that could explain the clinical descriptions of neurological, coagulation and inflammatory symptoms, and here we present the results of our analyses. We found toxin-like peptides in plasma, urine, and faecal samples from COVID-19 patients, but not in control samples. As our findings do not correspond with current thinking of the aetiology related to the observed clinical manifestations in COVID-19 patients, we feel their immediate sharing with the scientific community is critical.

Rationale
Liquid Chromatography-Surface Activated Chemical Ionization -Cloud Ion Mobility Mass Spectrometry (LC-SACI-CIMS) exhibits a high selectivity in peptide detection thanks to its ability to selectively isolate peptide ions through an in-source ion mobility (IM) effect. In fact, it allows a selective regulation of the potential difference between the low voltage of the SACI surface (47 V) and the entrance lens (-50 / -600 V), and a selective focalization on solvent ion cloud containing species at low or high m/z ratio. By switching the entrance voltage lens between -50 and -600 V during the analysis, it is possible to separate the low m/z from the high m/z potential signal, to avoid ion trap saturation, and to maximize the number of detected compounds. The mass spectra chemical noise is also strongly reduced due to the lower amounts of solvent cluster ions that are produced in low voltage ionization conditions. Thus, the peptide detection efficiency is strongly increased by the IM selectivity and lower chemical noise with respect to the classical high voltage ionization approaches. Thanks to the specificity of the SACI-CIMS technology in focalizing the solvent ion clouds containing the high m/z (oligo-)peptide species, it was possible to increase the detection efficiency.
In the use of LC-SACI-CIMS, the following strategies have been adopted: • To reduce the presence of contamination as much as possible and to avoid the formation of acetonitrile polymers occurring in acid conditions (as reported by Eizo et al. 13 ), formic acid was not added to the CH 3 CN chromatographic phase.
• To separate low from high m/z solvent ion clusters by reducing the ion trap saturation, the space/charge effect, and by increasing the detected compounds recovery, LC-SACI-CIMS entrance lens voltage was switched between -50 and -600 V every 10 ms during the analysis.
• To enhance the SACI ionization efficiency, NH 4 HCO 3 was added to the samples. As reported in the literature 14,15 , the peptide ionization efficiency (and consequently the sensitivity) is enhanced in SACI conditions when ionic salts are present in the sample, due to peptide ion specific coordination.
• To decrease the total run time, a shot gun chromatographic gradient was used to desalt the sample.
• To avoid sample molecular profile alteration, and to evaluate the potential biological activities of the circulating species, no enzymatic digestion was applied to samples.
• To normalize the m/z signal intensity, 5 µL of standard ESI tune mix (Agilent, USA) were added to each sample extract.

Cohort
Samples used in the present study: plasma samples collected from 20 COVID-19 patients from different cities of Italy and from 10 control individuals (i.e. negative to SARS-CoV-2 tests and not affected by cancer or autoimmune diseases); urine samples collected from two additional COVID-19 patients and from two control individuals; stool samples from three COVID-19 patients and from three control individuals. The human biological samples used in the experimentation were collected and used with the expressed free and informed written consent, of the person from whom the material was taken, according to current legislation. The study received approval from "Comitato Etico Campania Sud" (n.36/2021, request submitted on 06-05-2020). Apart from positivity to SARS-CoV-2, no additional information (i.e. age, sex, blood serotype, severity of the disease, time of the collection, fatality, etc.) was provided.

Sample preparation
Plasma. Each plasma sample was treated as follows: 5 µL of CH 3 CN were added to 50 µL of plasma and vortexed for one minute. The procedure was repeated 10 times. Then the sample was centrifuged at 1,500 g for 10 minutes and two 100 µL aliquots of supernatant were dried and resuspended in 70 µL of NH 4 HCO 3 50 mmol. The solution was analysed by LC-SACI-CIMS (see Rationale).
Urine. Each urine sample was treated as follows: an equivalent volume of bi-distilled water was added, followed by centrifugation at 1,500 g for 10 minutes. 100 µL were dried and resuspended in 70 µL of NH 4 HCO 3 50 mmol. The sample was analysed by LC-SACI-CIMS (see Rationale).
Stool. Each stool sample was treated as described by Cristoni et al. 11 and analysed by LC-SACI-CIMS (see Rationale).

Liquid chromatography
The Ultimate 3000 LC (by ThermoFisher) was used to achieve separation of analytes for each sample prior to mass spectrometry (MS) analysis. A reversed phase Kinetex C-18 LC column (50 × 2.1 mm; particle size, 5 µm; pore size, 100 Å, by Phenomenex, USA) was used. The eluent flow was 0.25 mL/min and the injection volume was 15 µL. The mobile phases were: The elution gradient was: 2% (v/v) of B between 0 and 2 min; 2 to 30% between 2 and 7 min; 30 to 80% between 7 and 9 min; 80% between 9 and 12 min; 80-2% between 12 and 12.1 min. The column was rebalanced with 2% of B between 12.1 and 17 min.

Mass spectrometry
All samples were analysed for the presence of proteins with potential toxic effect by using the LC-SACI-CIMS as already described in the literature 9-12 . Samples were analysed with an ORBITRAP mass spectrometer (Breme, Germany) coupled to a surface-activated chemical ionization (SACI) source and operated in positive ion mode.
The surface voltage was 47 V and the entrance lens was switched between -50 and -600 V each 10 ms. Auxiliary gas: 2 L / min; Nebulizer gas: 80 psi; Temperature: 40 °C. Full scan spectra were acquired in the 40-3,500 m/z range for non-targeted metabolomics/proteomics analyses to detect analytes. The same m/z range was used for both discovery and selective biomarker identification, and to standardize (primarily in terms of scan rate) the instrument. The software used for data elaboration is SANIST, a modified version of the Global Proteome Machine (GPM, https://www.thegpm.org/GPM/), implanted as described in 9-12. SANIST output files are available as supplementary material 16 (see section Data availability).
SANIST software here used is freely available, upon email request to CranioMed group (dir.brogna@craniomed.it).
Mass spectrometry on samples was performed with collision-induced dissociation using data dependent scan and helium as the collision gas. The ion trap was applied to isolate and fragment the precursor ions (windows of isolation, ± 0.3 m/z; collision energy, 30% of its maximum value, which was 5V peak to peak), and the ORBITRAP mass analyser was used to obtain fragments with an extremely accurate m/z ratio (resolution 15,000; m/z error <10 ppm).

Data elaboration
Detected high m/z peptides were used to identify toxins thanks due to the selectivity given by their long chain.
The complete UniprotKB set of manually reviewed venom proteins and toxins (UniprotKB, Animal toxin annotation project. https://www.uniprot.org/program/Toxins, Accessed October 4, 2020), mixed with a subset of non-venom proteins and toxins from UniprotKB database 17 was used as reference protein dataset in order to give statistical significance to the results. The information reported in Table 1 has been retrieved from the UniprotKB database and from the NCBI Taxonomy database 20 , after confirmation by BLAST sequence comparison analysis 18 .
SANIST was set to perform the database search considering all potential protein points and post-translational modifications, and to consider proton rearrangements. No enzyme cutting rules were specified, but all the protein subsequence combinations were considered. Database search calculation was performed by means of General Processing Graphic Processing Units (GPGPU).
The MS data are available on the ZENODO platform 16 (see section Data availability).

Results and discussion
The presence of (oligo-)peptides characterised as toxic components of animal venoms was observed in plasma and urine samples from SARS-CoV-2 infected patients and never in plasma, urine and faecal samples from control individuals. Examples of SACI-CIMS chromatograms are reported in Figure 1 and Figure 2 (panels a and b), showing the spectra acquired by means of the LC-SACI-CIMS technology. Figure 2c and d show the spectra obtained using ESI extracted at the same retention time. SACI-CIMS give rise to higher signal intensities probably due to the low ion trap saturation.
Several (oligo-)peptides (between 70 and 115, depending on the analysed sample) matched to different animal venom proteins and toxins like conotoxins, phospholipases A2, metalloproteinases (86% of assignments have a -log(e) higher than 25). An overview of 36 proteins covered by the toxin-like peptides found is reported in Table 1; details of -log(e) and false discovery rates are reported in Table 2. Examples of mass spectra peptide characterization together with the peptide ion fragmentation pathways are shown in Figure 3a. All the MS/MS signal were assigned to the different N-terminal y,z (blue and purple colour) and c-terminal b,c (red and yellow colour) fragmentation series (see Figure 3b for fragmentation series details). In the defined SACI-CIMS conditions, doubly charged m/z ion of medium-high molecular weight peptide species are produced, allowing high identification accuracy, in line with what is already described in the literature that high identification statistical rates are achieved analysing peptide doubly charged species with medium high molecular weight. Different fragmentation anomalies with proton rearrangements have also been detected and considered in phase of data elaboration. Only mass spectra exhibiting a statistical -log(e) score higher that 10 and a false discovery rate lower than 0.05 were considered for the identification (see Figure 3c). False discovery rate and statistical score were estimated by means of reverse sequence approach.
Some of the toxin-like peptides found mapped on the same reference protein (UniprotKB: D2DGD8), are reported in Figure 4: these peptides were found in the five plasma samples and in the three faecal samples.
The types of toxic-like peptides found resemble known conotoxins, phospholipases A2, metalloproteinases, prothrombin activators, coagulation factors, usually present in animal venoms, which are known to have high specificity and affinity towards human ion channels, receptors, and transporters of the nervous system, like the nicotinic acetylcholine receptor. Cheng et al. 21 reported the discovery of a superantigen-like motif in the S1 Spike protein, as well as two other neurotoxin-like motifs that have peptide similarities to neurotoxins from Ophiophagus (cobra) and Bungarus genera. They conclude that neurotoxin-like motifs are present in SARS-CoV-2 protein products, acting as neurotoxin-like peptides. We checked in the full set of peptides we got (here we report only 36 examples), and we identified, in plasma and faecal samples, toxin-like peptides mapping on kappa 1a-bungarotoxin, Kappa 1b-bungarotoxin from Malayan krait, kappa-2-bungarotoxin and alpha-bungarotoxin from many-banded krait (Uniprot Accession Numbers Q8AY56, Q8AY55, P15816, and P60615, respectively), which were reported by Cheng and colleagues. Furthermore, we looked at the amino acid changes currently reported in GISAID data 22 , analysed by CoV-GLUE-Viz (update 15/09/2021) 23 , and occurring in the Y 674 QTQTNSPRRAR 685 motif identified by these authors as homologous to neurotoxin motifs of animal venom proteins. We observed the existence of amino acid variations which makes this motif even more similar to the neurotoxin motifs of animal venom proteins (like variations Q677S and T676A observed in sequences assigned to PANGO Lineage B.1.596). Experiments to assess neurotoxicity of these peptides and of spike protein on 3D neuronal/glial model ("neurospheres") obtained from human induced Pluripotent Stem Derived Neural Stem Cells (iPS-NSCs) are currently ongoing.
What follows is our attempt to elaborate a potential relation between their presence and extra-pulmonary COVID-19 symptomatology.

Conotoxins
Conotoxins are neurotoxic peptides isolated from the venom of marine (genus Conus) cone snails. In their mature form, they consist of 10 to 30 amino acid residues, with often one or more disulphide bonds, which are used to classify them in structural classes (µ-conotoxins, ω-conotoxins, and α-conotoxins are the major classes). The mechanism of action of conotoxins is not yet fully understood 24 . Studies have found that they are able to modulate the activity of several receptors, including ion channels, nicotinic acetylcholine receptors The presence of conotoxin peptides might explain the occurrence of many symptoms (like hyposmia, hypogeusia and the signs typical of Guillain-Barre syndrome) observed in some COVID-19 patients. Their presence can alter normal functioning of ion channels, nicotinic acetylcholine receptors and of acetylcholine levels.

Phospholipases A2
Phospholipases A2 (PLA 2 , E.C. 3.1.1.4) hydrolyse phospholipids and lead to release of lysophosphatidic acid and arachidonic acid 29 . Arachidonic acid is a major precursor of many pro-inflammatory mediators like leukotriene, thromboxane and prostaglandin; as a consequence, abnormal presence of active PLA 2 can induce severe inflammation 30 . In animal venoms, PLA 2 act as neurotoxic proteins: they hydrolyse membrane phospholipids of the motor nerve terminal, and the plasma membrane of skeletal muscle, thus triggering a severe inflammatory degenerative response, which in turn leads to degeneration of the nerve terminal and skeletal muscle 29 . The drug dexamethasone can inhibit prostaglandins synthesis and leukotriene formation 31 . As dexamethasone is still the only therapeutic shown to be effective against the novel coronavirus in patients 32 with severe symptoms, it can be that the positive effect of this drug on COVID-19 patients is also due to the reduction of the here identified PLA 2 -like peptides.

Metalloproteinases
The last example of identified toxin-like peptides is those recognised as metalloproteinases present in animal venoms, zincdependent enzymes of varying molecular weight having multidomain organization. These toxic enzymes cause haemorrhage,    Table 1) almost identical short stretches (up to 10 amino acids) of these peptides in potential coding regions of many bacterial and viral sequences, but no long potential coding frame entirely covering any of them was found. Consequently, at the time of writing we have not yet identified the "genetic source" of these peptides, which could be: • The SARS-CoV-2 RNA genome with its protein reading set, as proposed by Brogna 35 , who reported the identification in SARS-CoV-2 RNA of many regions encoding for oligopeptides (four-five amino acids long) identical to neurotoxin peptides typical of animal venoms. Only mass spectra exhibiting a statistical -log(e) score higher that 10 and a false discovery rate lower than 0.05 were considered for the identification (reported in panel c). False discovery rate and statistical score were estimated by means of reverse sequence approach.
• The SARS-CoV-2 genome directly read by bacteria, assuming that the SARS-CoV-2 genome, or parts thereof, is capable of replicating with a possible 'bacteriophage-like' mode of action, as previously described 36 .
• Genomes of bacteria, which, as a reaction to the presence of the virus, secrete these peptides. This could happen by using still not well known and debated mechanisms, like alternative reading due to rRNA sequence heterogeneity (as described in 37,38), or the involvement of small bacterial ncRNA (sRNAs), known to be key players of gene regulation under conditions like stress response, quorum sensing, or virulence (in this context, in 1984 Coleman et al. reported the micF non-coding RNA as a functional bacterial sRNA 39 ).
• A combination of the above e.g. the 'toxin' genetic code is present in the bacteria and expression may be triggered by SARS-CoV-2, acting like temperate bacteriophages, which are known to interact with bacteria so that they express (or not) certain genes, as described by Carey et al. 40 .
A detailed 3D structural similarity analysis between the toxinlike peptides found and reference proteins has not yet been conducted. Accordingly, at the time of writing, we can only speculate that these toxin-like peptides are involved in the clinical extra-pulmonary manifestations in symptomatic COVID-19 patients. According to our knowledge, these toxin-like peptides have never been searched in animals considered reservoirs of SARS-CoVs.

Conclusions
The presence of (oligo-)peptides almost identical to toxic components of venoms from animals has been observed. Data and results reported here suggest an association between COVID-19 disease and the release in the body of these, and raise a series of questions: • Are these findings in line with what was proposed by Tizabi et al. 41 , i.e. a potential therapeutic role for nicotine, nicotinic agonists, or positive allosteric modulators of nicotinic cholinergic receptors in COVID-19?
• If induced by SARS-CoV-2, can the production of toxin-like peptides be involved in the neurological disorders and injuries observed in hospitalized COVID-19 patients?
• If induced by SARS-CoV-2, can the production of toxin-like peptides influence complex diseases apparently triggered or enhanced by COVID-19, like e.g. Guillain-Barré Syndrome 42 or Parkinson's disease 43 ?
• Are toxin-like peptides associated with SARS-CoV-2 infection or to other viral infections or, more in general, is their presence related to sickness condition?
• Are our findings supporting the suggestion made by the iVAMP Consortium 44 on the relationships between animal venom glands and microorganisms' microenvironments?
We consider that the immediate sharing of these results can contribute to the untangling of the multifaceted set of clinical manifestations in symptomatic COVID-19 patients, and to the further understanding of the mechanisms involved.

Consent
The human biological samples used in the experimentation were collected and used with the expressed free and informed written consent of the person from whom the material was taken, according to current legislation.

Nasopharyngeal Swabs: A Proof-of-Concept Focused on a 3 Min Mass
Open Peer Review

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Partly We will address all of them, and wait for those of other reviewers, in order to provide a fully revised version of the manuscript.
Best regards, Mauro Petrillo, on behalf of the authors. You are perfectly right: it can't be excluded that the findings aren't specific to COVID and that might be common to other conditions.
And in fact, one of the questions of the Conclusions section of the manuscript is "Are toxin-like peptides associated with SARS-CoV-2 infection or to other viral infections or, more in general, is their presence related to sickness condition?" The aim of our manuscript is to immediately share these observations with the scientific community as they are (together with a series of other observations which we have recently reported in https://doi.org/10.12688/f1000research.52540.3) quite unexpected, at least to us.
Thanks again for your time and interest. I am happy to further discuss, also privately.

Best regards, Mauro Petrillo
Competing Interests: None Reader Comment 19 Jul 2021 Ernesto de Bernardis, ASP SR, Italy I don't understand why the Authors' hypotheses about the origin of these peptides don't include the host response during severe inflammation or ARDS, or during a pharmacological therapy similar to those given to COVID patients.
The paper compares peptides from COVID patients with those from healthy controls, but doesn't