key: cord-1049913-eaejj26k
authors: Mehta, Ravi; Chekmeneva, Elena; Jackson, Heather; Sands, Caroline; Mills, Ewurabena; Arancon, Dominique; Li, Ho Kwong; Arkell, Paul; Rawson, Timothy M.; Hammond, Robert; Amran, Maisarah; Haber, Anna; Cooke, Graham; Noursadeghi, Mahdad; Kaforou, Myrsini; Lewis, Matthew R.; Takats, Zoltan; Sriskandan, Shiranee
title: Antiviral metabolite 3'-Deoxy-3',4'-didehydro-cytidine is detectable in serum and identifies acute viral infections including COVID-19
date: 2022-01-31
journal: Med (N Y)
DOI: 10.1016/j.medj.2022.01.009
sha: c16dbc06f7dd998b2a81cac15c95fc0b1343efb1
doc_id: 1049913
cord_uid: eaejj26k

Background There is a critical need for rapid viral infection diagnostics to enable prompt case identification in pandemic settings and support targeted antimicrobial prescribing. Methods Using untargeted high-resolution liquid chromatography coupled with mass spectrometry, we compared the admission serum metabolome of emergency department patients with viral infections including COVID-19, bacterial infections, inflammatory conditions, and healthy controls. Sera from an independent cohort of emergency department patients admitted with viral or bacterial infections underwent profiling to validate findings. Associations between whole-blood gene expression and the identified metabolite of interest were examined. Findings. 3'-Deoxy-3',4'-didehydro-cytidine (ddhC), a free base of the only known human antiviral small molecule ddhC-triphosphate (ddhCTP), was detected for the first time in serum. When comparing 60 viral to 101 non-viral cases in the discovery cohort, ddhC was the most differentially abundant metabolite, generating an area under the receiver operating characteristic curve (AUC) of 0.954 (95% CI: 0.923-0.986). In the validation cohort, ddhC was again the most significantly differentially abundant metabolite when comparing 40 viral to 40 bacterial cases, generating an AUC of 0.81 (95% CI 0.708-0.915). Transcripts of viperin and CMPK2, enzymes responsible for ddhCTP synthesis, were amongst the five genes most highly correlated to ddhC abundance. Conclusions The antiviral precursor molecule ddhC is detectable in serum and an accurate marker for acute viral infection. Interferon-inducible genes viperin and CMPK2 are implicated in ddhC production in vivo. These findings highlight a future diagnostic role for ddhC in viral diagnosis, pandemic preparedness, and acute infection management.

Early differentiation of acute infectious aetiologies is now a priority in diagnostic innovation.

Conventional methods relying on pathogen identification through culture, polymerase chain reaction or antigen detection are time-consuming and/or insensitive, leading to diagnostic delays that result in inappropriate antimicrobial prescription and infection transmission. [1] [2] [3] There is therefore renewed interest in novel biomarkers of infection classes that can better guide therapeutic and infection control decisions in real-time.

Metabolomics technologies for large-scale characterisation of low-molecular-weight metabolites have the potential to aid discovery of novel biomarkers of infectious diseases.

Liquid chromatography coupled with mass spectrometry (LC-MS) and nuclear magnetic resonance (NMR) spectroscopy stand out among the most commonly employed techniques in the field. The use of mass spectrometry has already revolutionised modern microbiology by enabling rapid detection of bacterial species from cultured colonies. 4 Despite its growing impact on biomedical research, metabolic profiling of biofluids has produced candidate biomarkers in only a small number of infectious states. One study identified a two-metabolite serum signature differentiating infected from non-infected patients within a systemic inflammatory response syndrome cohort. 5 Metabolomic interrogation of cerebrospinal fluid from patients with meningitis was able to differentiate between M. tuberculosis and other infectious causes. 6 Wang et al. examined the lipidome of 40 patients in a paediatric cohort prior to the COVID-19 pandemic and identified a 3-lipid signature that discriminated bacterial from viral infection, although wider metabolomic changes were not reported. 7 A number of more recent studies report metabolic differences J o u r n a l P r e -p r o o f between patients with and without SARS-CoV-2 infection, [8] [9] [10] but comparator groups did not include bacterial infections.

We investigated the serum metabolome of adult patients presenting to two UK emergency departments with a range of suspected infection syndromes, including COVID-19, to derive and cross-validate novel biomarkers for viral and bacterial infections. We used point-of-admission samples to replicate the timepoint where a discovered biomarker would be used clinically. To ensure diagnostic certainty, we adopted a case-control approach with laboratory-proven infections. Our sampling included unwell, non-infected cases to ensure that any biomarkers identified accounted for cases of inflammatory conditions unrelated to infection. 11 J o u r n a l P r e -p r o o f

In the discovery cohort, serum from 232 patients and 13 healthy controls underwent metabolomic profiling with LC-MS (Figure 1 ), producing the hydrophilic interaction liquid chromatography (HILIC+) and reversed-phase chromatography (RPC+/-) datasets. All discovery cohort patients were admitted via one of two emergency departments at Imperial College Infection study . Owing to insufficient sample volume or data quality, four samples were excluded from the discovery primary analysis in both lipidomics assays, and five from the HILIC+ assay. 80 of the 112 COVID-19 samples were not transferred to a -80°C freezer within five days of collection and so this sub-group was also excluded from the discovery Analysis of the discovery cohort HILIC+ dataset identified several significantly differentially abundant (SDA) features with a median absolute log2 fold-change of >4 and p-value <0·01 when comparing viral cases (pre-COVID-19 viral and COVID-19) versus all other groups, and viral versus bacterial cases (Gram-positive and Gram-negative bacteraemia) (Figures 2A & 2B) .

The top SDA discriminator was the feature 248·0647 m/z at 1·96 minutes, which showed a 36fold change in the median intensity in viral cases compared to all other groups (adjusted pvalue <1x10 -18 ). This metabolite was identified as 3'-Deoxy-3',4'-didehydro-cytidine (ddhC), a free base of the ribonucleotide ddhC-triphosphate (ddhCTP) recently reported to have antiviral properties in vitro. 12 Using all samples from the discovery primary analysis cohort, ddhC returned an area under the receiver operating characteristic curve (AUC) of 0·954 (95% CI 0·923-0·986; sensitivity 88·1%, specificity 91·7%) in discriminating viral infections from all other groups, and 0·944 (95% CI 0·905-0·983; sensitivity 89·8%, specificity 86·7%) in discriminating viral from bacterial infections ( Figure 2C ). When we included the sub-group of samples that spent more than five days outside a -80°C freezer, similar results were achieved with AUCs of 0·966 and 0·959 respectively (Supplementary Figure 2A) .

In the discovery primary analysis cohort, ddhC demonstrated a higher relative intensity among patients with viral infections compared to other groups ( Figures 2D & 2E) . Similar results were achieved when including samples that spent more than five days outside a -80°C freezer (Supplementary Figures 2B & 2C) . There was no correlation between ddhC and age, and no significant difference in the median ddhC intensity between sex and ethnicity subgroups (Supplementary Figures 2D, 2E & 2F ). There was a low correlation between admission serum creatinine and ddhC (correlation coefficient 0.555, p-value <1x10 -12 ), but there was no significant difference in the median serum creatinine between viral and non-viral groups 

To assess and cross-validate the discriminatory performance of single markers distinguishing infection groups in the data, we utilised the FS-PLS method. For each comparison (viral versus other, viral versus bacterial, bacterial versus other) we present the discriminating feature that was selected most frequently out of 100 different training:test FS-PLS runs and the median and IQR of the test AUCs generated (Supplementary Table 3 J o u r n a l P r e -p r o o f ddhC performs better than white cell count, lymphocyte count & C-reactive protein (CRP) as a biomarker for viral infections.

We compared the ability of ddhC to differentiate viral infection from other groups to alternative biomarkers such as white cell count, lymphocyte count, and CRP, which were taken as part of routine admission clinical laboratory tests. We used the discovery primary analysis HILIC+ cohort (n=161) and excluded healthy controls (n=13), for whom there were no routine laboratory test data (n=148 in total). All patients had a white cell count and lymphocyte count recorded, and 122/148 patients had a CRP. Routine admission clinical laboratory tests performed poorly compared to ddhC ( Figure 2F ).

Sera from a separate cohort of 80 patients from the UCLH arm of the BioAID study with confirmed viral (n=40) and bacterial infections (n=40) underwent untargeted metabolomic profiling using the HILIC+ assay. PCA did not show clustering of samples by age or sex and the corresponding eigencor plots did not show correlation above 0·3 (Supplementary Figures 3A, 3B & 3C). Using the same empirical thresholds as the discovery analysis, no SDA discriminators were found. Adjusting the median absolute log2 fold-change threshold to 2, the top SDA discriminator was the feature 248·0648 m/z at 1·93 minutes, which is the same ion [M+Na] + of ddhC identified in the discovery cohort and showed a 6·7-fold change in the median intensity in the viral compared to bacterial group ( Figure 3A , adjusted p-value <1x10 -3 ). ddhC returned an AUC of 0·811 (95% CI 0·708-0·915, sensitivity 72·5%, specificity 92·5%) in discriminating viral To assess the role of ddhC as a prognostic indicator, we performed a post-hoc exploratory analysis of the ddhC response in all viral infections (pre-COVID-19 viral and COVID- 19) categorised by outcome severity. We used the total discovery patient cohort (including samples that were transferred to -80°C storage after 5 days, n=138) to maximise power to differentiate between categories. The median relative ddhC intensity was 35,525 in mild disease (admission duration 0-2 days), increasing to 71,569 in moderate (admission duration 3- Table 4 . Two of the five genes are directly implicated in ddhCTP metabolism -RSAD2 (viperin), aided by CMPK2, mediates ddhCTP production during viral infection. 12 The correlation coefficient for viperin expression and ddhC intensity was 0·748 (p-value <1x10 -22 ); viperin was more highly expressed in patients with viral infections (Figure 4 ). Data for CMPK2

showed the same trends (Supplementary Figure 4 ).

The ability to rapidly differentiate infection syndromes is an urgent requirement, underlined by the ongoing COVID-19 pandemic and the growing threat of antimicrobial resistance. We capitalised on the sensitivity of high-resolution liquid chromatography coupled with mass spectrometry to discover that 3'-Deoxy-3',4'-didehydro-cytidine (ddhC), a free base of the antiviral molecule ddhC-triphosphate (ddhCTP), was detectable in patient serum. In a discovery cohort, ddhC was found to have a 36-fold higher median intensity in patients with viral infections, including COVID-19, compared to those with bacterial infections, non-infected inflammatory states, and healthy controls, corresponding to an AUC of 0·954, sensitivity of 88·1% and specificity of 91·7%. It outperformed white cell count, lymphocyte count, and CRP as a viral biomarker (AUCs of 0·688, 0·545 and 0·585, respectively). In an independent validation cohort, ddhC was again the most significantly differentially abundant metabolite when comparing patients with viral versus bacterial infections, generating an AUC of 0·811.

ddhCTP has recently been shown to be the first and, to the best of our knowledge, only small molecule produced by humans that is capable of directly inhibiting viral replication machinery. 12 Gizzi et al. showed that the enzyme viperin (virus inhibitory protein, endoplasmic reticulum-associated, interferon-inducible), aided by the genomically adjacent enzyme cytidylate monophosphate kinase 2 (CMPK2), catalyses the conversion of CTP to ddhCTP, which acts as a chain terminator for multiple viral RNA-dependent RNA polymerases (RdRPs). 12 Synthetic ddhC traversed the plasma membrane of Vero and HEK293T cells, suggesting a mechanism for how ddhC might eventually reach the serum in detectable quantity. ddhC has also been detected in prokaryotic cells; Escherichia coli production of ddhCTP after viperin homolog expression was associated with T7 phage RdRP suppression, suggesting a role for J o u r n a l P r e -p r o o f ddhCTP in bacterial immunity to viruses. 13 To our knowledge, ddhC has hitherto not been identified in human or other mammalian serum, nor associated with COVID-19. We showed that this antiviral molecule was a sensitive and specific serum biomarker for a range of viral infections, including COVID-19, in clinical samples of patients presenting to hospital. In a subset of patients for whom RNA-Seq data was available, we showed that viperin and CMPK2 expression was also increased in patients with viral infections. Furthermore, of more than 18,000 genes, their expression was amongst the top five most highly correlated with ddhC intensity (correlation coefficients 0·75 and 0·76 respectively), providing a plausible mechanism by which ddhC may be produced during viral infection. facilitating further investigation of its use. 15 Seifert et al. showed that exogenous ddhCTP can inhibit SARS-CoV-2 polymerase activity in Huh7-hACE2 cells, although it did not decrease SARS-CoV-2 N-protein immunofluorescence. 16 Future work will ascertain whether exogenous and/or endogenous ddhCTP can inhibit SARS-CoV-2 viral replication in other cell lines, and its effect on other viruses. Here we show that ddhC is produced naturally in vivo in response to viral infections, at levels that are detectable in the circulation, increasing the likelihood of an acceptable safety profile of ddhCTP as a therapeutic.

Our study demonstrates a number of strengths. We deliberately included both healthy and unhealthy non-infected controls, reducing the likelihood of selecting biomarkers confounded by inflammation unrelated to infection. We used stringent inclusion criteria for infected patients, excluding those where the timing of clinical presentation or PCR/culture result might have affected infection status at the point of sample acquisition. We used admission-day samples taken prior to any intervention, the timepoint where a diagnostic test would be most useful.

In conclusion, using high-fidelity metabolic profiling of serum from patients attending hospital, we found that the antiviral molecule ddhC is present in human serum during viral infection and represents an accurate biomarker for a wide range of viral infections, including COVID-19.

These findings pave the way for a universal blood test to rapidly identify acute viral infections, which could play a key role in both pandemic preparedness and routine acute infection management.

Our study should be viewed in the context of its limitations. Firstly, to ensure diagnostic certainty, we only included bacterial infections associated with bacteraemia, and did not include fungal and protozoal infections. We plan to assess the performance of ddhC in new patient cohorts that include a wider range of infection syndromes. Secondly, we allowed a period of up to five days from serum sample acquisition to -80°C storage for the discovery cohort, which may have introduced variability in potential metabolite degradation at 4°C during the 0-5 day period. However, ddhC performed similarly in COVID-19 samples that had not been frozen within 5 days, which suggests it is likely to be a robust marker that is not highly susceptible to temperature-associated degradation and suitable for 'real-life'

biochemical analytics, where samples may require time to reach laboratories prior to testing.

The serum samples for both arms of the discovery cohort were collected in the same way from the clinical diagnostic laboratory, albeit that patients were prospectively recruited to BioAID, while retrospectively recruited to Microbial Products in Infection. Thirdly, our severity analysis was exploratory without consideration of potential confounders and incorporated hospitalisation duration, which can be affected by factors other than severity. Fourthly, we did not have access to the patients' pre-admission medication history, which could potentially affect the serum metabolic profile, but we do not suspect this would significantly differ between viral and non-viral groups. Fifthly, our cohort represents patients unwell enough to seek hospital attention -further work will be required to assess the role of ddhC in less unwell patients presenting to primary care and determine whether it is detectable in minimally invasive samples such as urine or saliva. 

The authors declare no competing interests. Healthy controls not included as WCC, lymphocyte count, and CRP not available. Methods S1 -Supplementary items related to STAR Methods metabolite identification Table 1 Targeted feature extraction related to STAR Methods metabolite identification 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Professor Shiranee Sriskandan (s.sriskandan@imperial.ac.uk).

This study did not generate new unique reagents.  Any additional information required to reanalyse the data reported in this paper is available from the lead contact upon request.

Adult patients presenting to the emergency department were recruited into two separate cohorts: the discovery cohort and a post-hoc external validation cohort.

In the discovery cohort, patient serum samples were obtained from two parallel studies at Imperial College Healthcare NHS Trust (ICHNT), the Imperial arm of the cross-site Bioresource 

Serum samples from patients in one of the following six categories were used in the discovery cohort: culture-confirmed Gram-positive bacteraemia, culture-confirmed Gram-negative bacteraemia, PCR-confirmed viral infection prior to detection of SARS-CoV-2 in the UK (January 2020), PCR-confirmed COVID-19, non-infected patients, and healthy controls.

N=24 samples in each of two comparator groups were required to achieve a power of >90% to identify an AUC of at least 0·8, at a significance level of 0·01. Thus to enable all comparisons, accounting for potential sample exclusion (e.g. assay failure, poor data quality), we used n=30 samples in each clinical group apart from COVID-19, where we included all available samples (n=112) to facilitate exploration of severity differences in this cohort. Infection categories were assigned using electronic diagnostic pathology data pertaining to admission only and confirmed by a clinician. Non-infected patients were identified from the database where there J o u r n a l P r e -p r o o f was no positive microbial diagnostic test and no infection-related ICD-10 diagnostic code from BioAID admission. Sera from n=13 healthy controls were available from a subcollection of the ICHT BRC Tissue Bank.

To facilitate multi-omic comparison, serum samples from BioAID patients were prioritised if whole blood RNA-Sequencing (RNA-Seq) had already been undertaken as part of an earlier study, where samples had been selected from the BioAID database using a random number generator. 18 Sera from additional BioAID patients were selected randomly from within individual infection groups using a random number generator in Excel.

Bacteraemic patients were excluded if the isolated bacterium was deemed a contaminant, or if the blood culture was taken >24 hours prior to/post the admission serum sample. COVID-19 patients were excluded if their positive PCR test was taken >10 days prior to admission, or >2 days post admission, to avoid non-COVID-19 related admissions and hospital-acquired COVID-19, respectively. Microbiologically confirmed co-infections across different infection classes were excluded.

In the validation cohort, we compared two patient groups: bacteraemic and viral. Based on discovery data, to achieve a power of >90% to identify an AUC of 0.75, at a significance level of 0.01, we required 36 patients in each group. Accounting for potential sample exclusion, we used n=40 samples in each group, selected from the UCLH BioAID database using a random number generator in Excel. For all patients, the same inclusion and exclusion criteria were applied as in the discovery cohort.

Serum samples from 245 patients in the discovery cohort and 80 patients in the validation cohort were analysed using ultra-performance LC-MS following previously described analytical and quality control (QC) procedures. 19 

Serum samples were prepared as previously described. 20 Briefly, for each assay, samples were analysed in a randomised order demonstrating no correlation with study design variables, precluding any confounding effect of analysis order. To facilitate quality assessment and preprocessing, a pooled quality control (QC) sample was prepared by combining equal parts of each study sample and analysed periodically among study sample analyses. In addition, for assessment of analyte response, 21 a series of QC sample dilutions was created (10 x 100%, 5 x 80%, 3 x 60%, 3 x 40%, 5 x 20%, 10 x 10%, 10 x 1%) and analysed at the start and end of each set of sample analyses. Targeted extraction of the four associated features, each absent from the XCMS profiling dataset owing to non-detection, was performed using peakPantheR software and preprocessed as described above for the XCMS profiling dataset. 25 

For the validation cohort, the identification of ddhC was based on the same m/z and similar retention time to that observed in the discovery data.

We examined the interaction between whole blood gene expression and the feature of interest identified in the discovery cohort. Gene expression data were obtained from RNA-Seq of Imperial BioAID patient RNA samples, performed prior to this study in two cohorts. Full details for the first patient cohort (recruited pre-COVID-19 pandemic) have been described previously. 18 For the second patient cohort (recruited during the COVID-19 pandemic), whole blood was collected in the same way as the first cohort. 18 Material was quantified using RiboGreen (Invitrogen) on the FLUOstar OPTIMA plate reader (BMG Labtech) and the size profile and integrity analysed on the 2200 TapeStation (Agilent, RNA ScreenTape). Input material was normalised and strand specific library preparation was completed using NEBNext® Ultra™ II mRNA kit (NEB) and NEB rRNA/globin depletion probes following manufacturer's instructions. Libraries were on a Tetrad (Bio-Rad) using in-house unique dual indexing primers (based on Lamble et al). 29 Individual libraries were normalised using Qubit and pooled together. The pooled library was diluted to ~10 nM for storage and denatured and further diluted prior to loading on the sequencer. Paired end sequencing was performed The Wellcome Centre for Human Genetics in Oxford UK using a Novaseq6000 platform at 150 paired end configuration, generating a raw read count of 30 million reads per sample. The RNA-Seq analysis pipeline consisted of quality control using FastQC, 30 MultiQC 31 and annotations modified with BEDTools, 32 alignment and read counting using STAR, 33 SAMtools, 34 FeatureCounts 35 and version 89 ensembl GCh38 genome and annotation. 36 J o u r n a l P r e -p r o o f Genes completely missing in either of the RNA-Seq cohorts were removed, in addition to ribosomal genes. The two RNA-Seq cohorts were merged and the batch effects between the two cohorts, in addition to the plate effects within the first cohort, were removed by combat_seq. 37 The raw counts were normalised using DESeq2. 38 In patients for whom both metabolic and transcriptomic data were available, we assessed the correlation between log2transformed feature intensities of a metabolite of interest and log2-transformed expression of associated genes using Pearson correlation coefficients. We restricted further analysis to the five genes most highly correlated to the metabolite of interest.

Data analysis was performed using R. 39 Power calculations were performed using the pROC package. 40 Unit-variance scaled principal component analysis (PCA) and eigencor plots were performed to identify the major sources of variation in the datasets, using the PCAtools package. 41 For PCA, features where >98·5% of samples returned an intensity of zero were excluded (n=3/1572 in discovery HILIC+ dataset, nil in both discovery lipidomics datasets, n=1/1194 in validation HILIC+ dataset).

In the discovery cohort, we compared all viral cases (COVID-19 and pre-COVID-19) versus others, all bacterial cases (Gram-positive and Gram-negative bacteraemia) versus others, and all viral versus all bacterial cases. In the validation cohort, we compared all viral cases versus all bacterial cases. In each comparison, we assessed the fold-change between the infection groups' median intensities for each feature. P-values were generated using the two-sided Wilcoxon test and were adjusted using the Benjamini-Hochberg procedure. 42 Volcano plots were generated comparing median log2fold-change and -log10 p-values.

In order to cross-validate findings in the discovery cohort, we used the variable selection method, forward selection-partial least squares (FS-PLS). 43 FS-PLS has been described in detail elsewhere. 7, 44 Briefly, it is a forward-selection method that selects variables most strongly associated with the groups of interest. It can be used to select a multi-feature signature composed of non-correlated variables, but in this study the 'max' parameter was set to one to evaluate the performance with only one feature. Feature intensities were log2 transformed. A p-value threshold of 0·01 was used, which determined the selection of a variable or termination. 100 runs of FS-PLS were applied to the dataset for every comparison, each time with a different training:test split at a ratio of 70:30. In each FS-PLS run, the feature identified on the training set was tested on the test set, and its performance was assessed using the AUC generated. For the feature that was selected in the most FS-PLS runs out of 100, the median and interquartile range (IQR) of the respective test AUCs were generated.

To assess the diagnostic utility of a feature of interest in the discovery cohort and compare it to the traditional biomarkers C-reactive protein (CRP), white cell count, and lymphocyte count (procalcitonin levels were not routinely available), as well as examine its use in the validation cohort, AUCs were generated using the pROC package. 40 The Youden's J statistic was used to determine thresholds for sensitivity and specificity. 45 In an exploratory post-hoc analysis, to investigate the relationship between the intensity of a feature of interest and illness severity in viral infections in the discovery cohort, we developed a three-point severity scale that differentiated between mild, moderate and severe illness in both COVID-19 and other viral illnesses. We incorporated duration of hospital admission, re-J o u r n a l P r e -p r o o f admission to hospital, admission to the intensive care unit (ICU) and death in this scale.

Severity group 1 (mild) included patients admitted to hospital for 0-2 days, group 2 (moderate) included patients admitted for 3-8 days, and group 3 (severe) included patients admitted for >8 days and those who were admitted to ICU or died at any point during admission. Readmission to hospital within 5 days of discharge was counted as the same admission. P-values were generated using the Kruskal-Wallis test. 

Considerations for diagnostic COVID-19 tests

False negative rate of COVID-19 PCR testing: a discordant testing analysis

Appropriateness of antibiotic prescribing in the Emergency Department

MALDI-TOF mass spectrometry: an emerging technology for microbial identification and diagnosis

Metabolite Profiles in Sepsis: Developing Prognostic Tools Based on the Type of Infection

)H nuclear magnetic resonance-based metabolic profiling of cerebrospinal fluid to identify metabolic features and markers for tuberculosis meningitis

Plasma lipid profiles discriminate bacterial from viral infection in febrile children

Large-Scale Multiomic Analysis of COVID-19 Severity

Proteomic and Metabolomic Characterization of COVID-19 Patient Sera

Integrative Modeling of Quantitative Plasma Lipoprotein, Metabolic, and Amino Acid Data Reveals a Multiorgan Pathological Signature of SARS-CoV-2 Infection

Overlap in serum metabolic profiles between non-related diseases: Implications for LC-MS metabolomics biomarker discovery

A naturally occurring antiviral ribonucleotide encoded by the human genome

Prokaryotic viperins produce diverse antiviral molecules

RNA-Dependent RNA Polymerase as a Target for COVID-19 Drug Discovery

Chemical Synthesis of the Antiviral Nucleotide Analogue ddhCTP

Inhibition of SARS-CoV-2 polymerase by nucleotide analogs from a single-molecule perspective

Cohort study protocol: Bioresource in Adult Infectious Diseases (BioAID)

Discovery and validation of a 3-gene transcriptional signature to distinguish COVID-19 and other viral infections from bacterial sepsis in adults; a case-control then observational cohort study

Development and Application of Ultra-Performance Liquid Chromatography-TOF MS for Precision Large Scale Urinary Metabolic Phenotyping

The effects of kisspeptin on beta-cell function, serum metabolites and appetite in humans

Representing the Metabolome with High Fidelity: Range and Response as Quality Control Factors in LC-MS-Based Global Profiling

XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification

peakPantheR, an R package for largescale targeted extraction and integration of annotated metabolic features in LC-MS profiling datasets

METLIN: A Technology Platform for Identifying Knowns and Unknowns

Improved workflows for high throughput library preparation using the transposome-based Nextera system

MultiQC: summarize analysis results for multiple tools and samples in a single report

BEDTools: a flexible suite of utilities for comparing genomic features

STAR: ultrafast universal RNA-seq aligner

The Sequence Alignment/Map format and SAMtools

featureCounts: an efficient general purpose program for assigning sequence reads to genomic features

ComBat-seq: batch effect adjustment for RNA-seq count data

Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2

R: A language and environment for statistical computing (R Foundation for Statistical Computing

pROC: an open-source package for R and S+ to analyze and compare ROC curves

PCAtools: Everything Principal Components Analysis

Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing

Lachlancoin/fspls: Minimal TB Biomarkers

Identification of Reduced Host

The authors acknowledge the NHS and NIHR CRN clinical research teams, diagnostic laboratory leads, and data managers whose work supports BioAID. The authors also acknowledge Ash

 Global metabolomic profiling of serum from infection patients presenting to ED  Single metabolite differentiates viral from bacterial infections and unwell controls  Identified as ddhC -a free base of only known human antiviral small molecule  ddhC has key potential role as rapid diagnostic for viral infections eTOC blurb Mehta et al. perform untargeted metabolic profiling of serum samples from patients presenting to the emergency department with different infections. The antiviral small molecule 3'-Deoxy-3',4'-didehydro-cytidine (ddhC) was detectable in human serum and accurately differentiated viral infections from other infection syndromes in both a discovery and validation patient cohort.

Researchers at the Imperial College London looked at all the small molecules in blood samples from a cohort of adult patients admitted to hospital with different types of infection, including COVID-19, other viruses, and bacteria. They found that the molecule 'ddhC' was produced in excess by patients with viral infections and was rarely present in those without. This result was replicated in a separate validation patient cohort. ddhC was previously shown to have antiviral properties in laboratory cells but has not to date been detected in living humans. Using ddhC to rapidly diagnose viral infections could prove beneficial in early recognition and containment of future pandemics and help to reduce unnecessary antibiotic use.J o u r n a l P r e -p r o o f

The table highlights the reagents, genetically modified organisms and strains, cell lines, software, instrumentation, and source data essential to reproduce results presented in the manuscript. Depending on the nature of the study, this may include standard laboratory materials (i.e., food chow for metabolism studies, support material for catalysis studies), but the table is not meant to be a comprehensive list of all materials and resources used (e.g., essential chemicals such as standard solvents, SDS, sucrose, or standard culture media do not need to be listed in the table). Items in the table must also be reported in the method details section within the context of their use. To maximize readability, the number of oligonucleotides and RNA sequences that may be listed in the table is restricted to no more than 10 each. If there are more than 10 oligonucleotides or RNA sequences to report, please provide this information as a supplementary document and reference the file (e.g., See Table S1 for XX) in the key resources table.

Please report the information as follows: REAGENT or RESOURCE: Provide full descriptive name of the item so that it can be identified and linked with its description in the manuscript (e.g., provide version number for software, host source for antibody, strain name  SOURCE: Report the company, manufacturer, or individual that provided the item or where the item can be obtained (e.g., stock center or repository). For materials distributed by Addgene, please cite the article describing the plasmid and include "Addgene" as part of the identifier. If an item is from another lab, please include the name of the principal investigator and a citation if it has been previously published. If the material is being reported for the first time in the current paper, please indicate as "this paper." For software, please provide the company name if it is commercially available or cite the paper in which it has been initially described. IDENTIFIER: Include catalog numbers (entered in the column as "Cat#" followed by the number, e.g., Cat#3879S). Where available, please include unique entities such as RRIDs, Model Organism Database numbers, accession numbers, and PDB, CAS, or CCDC IDs. For antibodies, if applicable and available, please also include the lot number or clone identity. For software or data resources, please include the URL where the resource can be downloaded. Please ensure accuracy of the identifiers, as they are essential for generation of hyperlinks to external sources when available. Please see the Elsevier list of data repositories with automated bidirectional linking for details. When listing more than one identifier for the same item, use semicolons to separate them (e.g., Cat#3879S; RRID: AB_2255011). If an identifier is not available, please enter "N/A" in the column. o A NOTE ABOUT RRIDs: We highly recommend using RRIDs as the identifier (in particular for antibodies and organisms but also for software tools and databases). For more details on how to obtain or generate an RRID for existing or newly generated resources, please visit the RII or search for RRIDs.Please use the empty table that follows to organize the information in the sections defined by the subheading, skipping sections not relevant to your study. Please do not add subheadings. To add a row, place the cursor at the end of the row above where you would like to add the row, just outside the right border of the