Interesting Preprints/Papers on Coronavirus

I’m starting this thread to gather commentary on interesting preprints discussing coronavirus. I’m going to start posting stuff in here with some basic commentary. Please feel free to chime in! Here’s one to start us off

First 12 patients with coronavirus disease 2019 (COVID-19) in the United States: This is a preprint that analyzes the outcomes for the first 12 US coronavirus patients. 3 of these patients were treated with Remdesivir. Here’s some analysis from Statnews about the results. The results are inconclusive about Remdesivir’s effectiveness, but there’s some interesting early clues to look at.


A human monoclonal antibody blocking SARS-CoV-2 infection:

Reports on a human monoclonal antibody that targets a communal epitope on SARS-CoV-2 and SARS-CoV. Coronavirus antibodies typically target the trimeric spike (S) glycoproteins on the viral surface. The spike proteins of SARS-CoV-2 and SARS-CoV are structurally very similar and both bind human angiotensis coverting enzyme 2 (ACE2).

Antibodies were originally identified using ELISA-(cross)reactivity assessed on antibody-containing supernatants of a collection of hybridomas from immunized transgenic H2L2 mice. An identified chimeric 47D11 antibody was recombinantly expressed as a human IgG1 antibody for further characterization.

The 47D11 antibody was found to potently inhibit infection of VeroE6 cells with SARS-S and SARS2-S pseudotyped VSV. Authentic infection of of VeroE6 cells with SARS-CoV and SARS-CoV2 was also neutralized, but with higher required IC50s.

The preprint claims this antibody will be useful developing antigen detection tests and serological assays targeting SARS-CoV2.

1 Like

Yeah, remdesivir seems unlikely to be a home run given the in-vitro data and limited in-vivo data. We can also assume that’s it’s been given to a lot of people in the US through compassionate use. If there was something big there we would have known by now.

I don’t know much about antibodies, but it seem like an interesting direction. I don’t have a feel for what timelines are like in large-molecule though.


The landscape of lung bronchoalveolar immune cells in COVID-19 revealed by single-cell RNA sequencing:

The authors of this work applied single cell RNA seq and single sell TCR seq to characterize lung bronchoalveolar lavage fluid (BALF) cells from 6 covid19 patients, 3 severe and 3 mild. The severe cases had monocyte-derived FCN1+ macrophages replacing FABP4+ alveolar macrophages (more on this later). Mild patients had highly expanded cloncal CD8+T cells.

scRNA-seq was performed using 10X genomics platform. 8 public scRNA-seq lunch samples were used as a control. Clustering analysis showed 36 different cell cluster types, with a higher proportion of T and NK cells in covid19 patient samples.

Recent studies have identified 3 different lung macrophage subsets by FCN1 (monocyte derived), SPP1 (pro-fibrotic), and FABP4 (alveolar macrophage). Clustering analysis showed a marked difference between the severe and mild patients, with severe patients having many more FCN1 dervied macrophages, and an almost complete loss of FABP4 macrophages.

Single cell TCR seq provides evidence that CD8+ T cells contain highly expanded clones, suggesting that CD8+ T cell response likely holds the key for successful viral control in covid19 patients.

The authors note that a limitation of their study is that it’s only based on 6 samples. This may mean that further observations are needed to confirm the conclusions of the current study. Computational analysis was performed using Seurat and pyScenic.

1 Like

Coronavirus data-sharing from Global Health Drug Discovery Institute with annotated preclinical studies and compound libraries for drug repurposing.

This page looks pretty interesting, has a nice collection of data from different preclinical studies (some annotated), previous clinical efforts for SARS/MERS, and compound libraries for drug repurposing. Could this be an interesting resource for cheminformatics or bioinformatics researchers? GHDDI also includes literature mining of previous coronavirus-related studies and their relationships.

They also shared the different approaches + datasets they used on their own models for drug repurposing:


Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

This is a published paper rather than a preprint, but a really important one. The authors of this work determined the structure of the trimeric spike glycoprotein of the virus using cryo-EM to get a 3.5 angstrom crystal structure of the virus in the prefusion conformation.

The structure of this protein is pretty similar to that for SARS-CoV, with a 3.8 angstrom RMSD difference over 959 C-\alpha atoms. The work also characterizes the dynamics of the interaction of SARS-CoV with angiotensin-converting-enzyme 2 (ACE2) via surface plasmon resonance. The binding affinity is 15 nM, about 10-20 fold higher than that for SAR-CoV. It’s possible this tight binding affinity helps explain the ease with which the virus can move from human to human. The work tested monoclonal antibodies against SARS-CoV against the current protein, but didn’t find any binding, despite the high similarity.


A serological assay to detect SARS-CoV-2 seroconversion in humans:

A serological assay can be used to detect SARS-CoV-2 antibodies. The development of a suitable serological assay for SARS-CoV-2 would allow for dynamic studies of the immune response to SARS-CoV-2, and can help determine the precise rate of infection in an affected area.

This preprint describes a serological ELISA assay developed using recombinant antigens derived from the SARS-CoV-2 spike protein. The developed assays can allow for screening and identification of COVID19 seroconverters as early as 3 days past symptom onset. The developed assay can be performed at BSL-2 (biosafety level 2) since it doesn’t involve live virus.

1 Like

A Trial of Lopinavir–Ritonavir in Adults Hospitalized with Severe Covid-19

Another important paper (not preprint). This is one of the first published clinical trials about Covid19 patients from a Chinese team. It unfortunately finds that a Lopinavir-Ritonavir treatment didn’t show any difference from standard care in an open-label trial with endpoint time to clinical improvement. There were 99 in the lopinavir-ritonavir arm and 100 in the standard. The median age of patients in the trial was 58.

Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an openlabel non-randomized clinical trial

This is the preprint that’s gotten a lot of attention with its marked results that hydroxycholoroquine and azithromycin combo could have very strong effect. Here’s the top level graph:


Derek Lowe’s blog has some excellent commentary. The main issue is that this is a small trial. Analysis on twitter has pointed out there may be some false positives in the hydroxychloroquine-only arm.

As a couple details from the paper, the trial lasted 14 days. The authors argue that the finding of the drop in the viral load could be very significant since recent work from China has shown that viral shedding can last between 20-37 days. If viral shedding can be controlled by treatment, this could have an impact on controlling spread of the virus. The authors note that even though azithromycin is more known as an antibiotic, there’s well established literature demonstrating its antiviral activity as well.

Structure, function and antigenicity of the SARS-CoV-2 spike glycoprotein

This preprint is from a little while ago, but is an important one so sharing here a little late. This work determined cryoEM structures of SARS-CoV-2 ectodomain trimer. This work also shows that SARS-CoV-2 uses ACE2 to enter cells and highlights similarities between SARS-CoV-2 and SARS-CoV.

The spike protein S comprises of two functional subunits. S_1 binds to the host cell receptor, and S_2 fuses the viral and cellular membranes. The S protein is the main target of neutralizing antibodies since it’s exposed on the surface and mediates entry into hosts. The SARS-CoV and SARS-CoV-2 spike proteins are highly similar, with a 4 amino acid insertion in SARS-CoV-2 which results in insertion of a furin cleavage site.

Structure of Mpro from COVID-19 virus and discovery of its inhibitors

The Covid19 protease is a key CoV enzyme that plays a pivotal role in mediating viral replication and transcription. This work uses computer aided drug design to find an inhibitor N3, then determines the crystal structure of the protease (Mpro) in complex with the compound. They used structure based virtual screening and high throughput assays to test 10,000 compounds for inhibition of Mpro. 6 of these compounds have IC50 values in the low micromolar range.

The N3 compound was designed to target proteases from SARS-CoV in previous work. A homology model was used to construct a model of the SARS-CoV-2 protease and docking was used to check whether N3 could dock against the SARS-CoV-2 protease. A crystal structure of N3 with the SARS-CoV-2 protease was constructed at 2.1 angstrom resolution.

The structure was then used to do in silico docking using Glide. A FRET assay was used to do the high throughput screen. The most potent compounds were tested to see if viral replication could be prevented in a cell based assay. A plaque reduction assay was performed as well.

SARS-CoV-2 vaccines: status report

This review article provides an overview of current therapeutics for Covid19, with a focus on vaccine development. Trials with remdesivir and protease inhibitors are currently ongoing in China. It’s not entirely clear what remdesivir’s mechanism of action is (it’s a nucleotide analog), but it may work by terminating RNA synthesis or by leading to incorporation mutagenesis. In addition a number of trials are being tried with HIV inhibitors such as lopinavir and ritonavir. Other treatment options being tried include dosing of human recombinant ACE2 and the antiviral arbidol. Human polyclonal IgG from transgenic cows was used for MERS-CoV and may also work here.

We know from SARS-CoV-1 that the spike protein is a good target for vaccine design. Several SARS-CoV-1 vaccines have been designed and tested in animal models. Vaccination with live virus resulted in lung damage in mice. Only a small number of SARS-CoV-1 vaccines made it to phase I trials before funding dired up. It’s possible that these SARS-CoV-1 vaccines might protect against SARS-CoV-2, but these aren’t currently available for use since they weren’t developed past phase I. There are some vaccines against MERS-CoV available, but these aren’t likely to protect against SARS-CoV-2.

An mRNA based vaccine (which expresses the target antigen in vivo) after injection of mRNA in lipiid nanoparticles recently started a phase I clinical trial. Other efforts in preclinical phases include recombinant protein based vaccines (focused on the spike protein), viral vector based vaccines (focused on the spike protein), DNA vaccines (focused on the spike protein), live attenuated vaccines, and inactivated virus vaccines. It’s not yet clear which of these is the best.

The paper tries to address the question of why it takes so long. Many of the technologies used for the vaccines are new and need to be tested for safety. This typically involves a test in an animal model of the virus. However, this is challenging since the virus doesn’t grow in wildtype mice and only induces mild disease in transgenic animals expressing human ACE2. In vitro neutralization assays can still be used to evaluate the serum from vaccinated animals even if the disease model isn’t accurate. Toxicity tests need to be performed (in rabbits typically). Usually this takes 3-6 months. Development of current Good Manufacturing Processes (cGMP) is also needed. All these processes need to be developed or amended for the SARS-CoV-2 vaccines. Safety and efficacy are evaluated in the usual Phase I, II, III trials.

Another challenge is that production capacity to produce sufficient vaccine might be challenging. For mature technologies such as inactivated or live attenuated vaccines, this already exists, but for novel technologies like mRNA, this may be challenging.

Finally, distributing the vaccine will take time. Vaccinating a large portion of the population may take weeks. More than one dose of the vaccine may be needed per person. This usually takes a prime-boost form, with immunity only achieved 1-2 weeks after the second. Given all these challenges, realistic time frames seem like 12-18 months.

The paper asks what longer term solutions might be. One might be to build up production capacity for respiratory viruses. More ambitious would be development of a broadly protective vaccine. This is being attempted for influenza but considerable work would be needed to develop this further.

The paper concludes by stating that in the past funds for vaccine development have only been available with looming pandemics. Given the seriousness of the current situation, it seems worth investing in better vaccine infrastructure. It also calls for developed emergency plans to be pulled out in case of future pandemics.

Therapeutic Strategies in an outbreak scenario to treat the novel coronavirus originating in Wuhan, China

This article review potential options to treat patients for coronavirus, including drug repurposing, neutralizing monoclonal antibodies, and oligonucleotide strategies targeting the viral genome. It then proposes a quick approach, the development of a biologic that blocks 2019-nCoV entry, using a soluble version of ACE2, fused to an immunoglobulin Fc domain. Treatment with ACE2-Fc could also replenish lowered ACE2 levels in the lungs during infection.

Screen Shot 2020-03-19 at 11.47.04 PM

An orally bioavailable broad-spectrum antiviral inhibits SARS-CoV-2 and multiple
2 endemic, epidemic and bat coronavirus

This works shows that the ribonucleoside analogue \beta-D-N^4 hydroxycitidine (NHC, EIDD-1931) has broad antiviral activity against SARS-CoV-2, SARS-CoV, and MERS-CoV. Mice infected with SARS-CoV and MERS-CoV improved pulmonary function, and reduced virus titer and body weight loss.

Antiviral assays were performed in continuous and primary human lung cultures. In a MERS-CoV assay in human epithelial cell line, NHC was potently antiviral with IC50 of .15\mu\textrm{M}. Similarly, NHC inhibited viral replication in Vero cells infected with SARS-CoV-2 with IC50 .3\mu\textrm{M}.

In vivo efficacy was tested with an orally bioavailable prodrug of NHC.

A Massively Parallel COVID-19 Diagnostic Assay for Simultaneous Testing of 19200 Patient Samples

This is a really interesting preprint in part because it’s released out on google drive! First time I’ve seen a google drive preprint.

This paper proposes a massively parallel workflow that uses next gen sequencing, PCR and reverse transcriptase (RT) to test up to 19200 patient samples per workflow. The authors computationally design 57600 high performance RT primers with unique barcode samples that link patient samples to cDNA products across 3 amplicons (N1 and N2 SARS-CoV-2 and RNAseP human control). The protocol instructions and primer sequences are released alongside the paper. The paper notes that next generation sequencing devices are now widely available at many institutions across the globe.

The paper comments that the key novelty of the technique is the unique labeling of patient samples at the RT step, the pooling of thousands of cDNA samples into a single PCR amplification step, the use of NGS in amplicon detection mode to measure thousands of variant cDNA levels, and rapid demultiplexing of NGS data.

Non repetitive RT primer sequences were designed by their algorithm, the non-repetitive parts calculator.

The convalescent sera option for containing COVID-19

This article makes a case that using human convalescent serum (isolated from the blood of recovered patients) could be a good option for the treatment and prevention of covid19.

The paper explains that passive antibody therapy involves the administration of antibodies against a given agent to a susceptible individual, while active vaccination requires the induction of an immune response that can take time to develop. Passive antibody therapy can provide immediate protection and has a long history going back to the 1890s. Evidence from SARS-CoV-1 shows that passive antibody therapy can work for coronaviruses.

As the number of cases continues to rise, there will also be more recovered patients who can provide convalescent serum. The paper makes the case that this could be used to protect high risk individuals (such as doctors on the frontlines). It might also be possible to treat individuals who are already sick, but the literature indicates this is less effective. Protection lasts for weeks to months depending on the individual.

There are some risks with passive antibody therapy. There’s always some risk with blood transfusions which would also hold here. There’s also the risk of antibody dependent enhancement of virus. It’s theoretically possible that antibodies for one SARS-CoV-2 strain could enhance the infection for another strain of the virus.

It’s worth noting that serological assays for SARS-CoV-2 will also be needed to measure viral neutralization in serum. Takeda is currently gearing up to produce antibodies from convalescent sera.

Viral dynamics in mild and severe cases of COVID-19

This article notes that patients who had higher viral load at time of admission seem to have more severe versions of the disease. The mean viral load of severe patients was 60 times that of mild cases. Mild cases had an early viral clearance, with 90% of patients testing negative on RT-PCR by day 10 and onwards. In contrast, all severe cases still tested RT-PCR positive on day 10. This observation matches similar observations made for SARS-CoV, suggesting that viral load might be a good estimator of disease severity.

Not a paper, but it seems relevant to this thread. Stanford is hosting a virtual conference about COVID-19 and AI on April 1.

1 Like

This blog has a lot of relevant preprints posted and updated. I’ve mainly been using Twitter to source preprints but I’ll start pulling from here as well. Anyone who’s interested in helping, please grab interesting preprints and write up brief summaries of them on this thread :slight_smile:

A SARS-CoV-2-Human Protein-Protein Interaction Map Reveals Drug Targets and Potential Drug-Repurposing

This is a large consortium paper by a giant author list from multiple groups at UCSF, University of Michigan Ann Arbor, Fred Hutchinson cancer research, Institut Pasteur, and UCSD.

The paper tries to build more understanding about the molecular details of SARS-CoV-2 infection. The effort cloned, tagged, and expressed 26 of the 29 viral proteins in SARS-CoV-2 in human cells. They then used affinity-purification mass spectrometry to identify the human proteins involved with these expressed viral proteins. This identified 332 high confidence SARS-CoV-2-human protein-protein-interactions. Of these, there are 66 druggable human proteins or host factors targeted by 69 existing FDA-approved drugs. The authors mention they are running live SARS-CoV-2 infection assays to evaluate these compounds.

Here’s a few details that caught my eye in the paper. The 30kB genome of the virus encodes up to 14 open reading frames. The 5’ Orf1a/Orf1ab encodes polyproteins which are auto-proteolytically processed into 16 non-structural proteins, forming the replicase/transcriptase process. The 3’ end of the genome has up to 13 ORFs, encoding 4 structural proteins including the spike protein, envelope protein, membrane protein, and nucleocapsid protein.

Mature nonstructural proteins were codon optimized and cloned into a mammalian expression vector. Protein expression plasmids were transfected into human HEK293T cells. Viral expression was checked with a western blot on input cell lysate.

The mass spec identified 332 PPIs. Gene ontology enrichment analysis was used to identify the major cell biological processes of the interacting proteins.