Successful treatment of COVID-19 with remdesivir in the absence of humoral immunity

The response to the coronavirus disease 2019 (COVID-19) pandemic has been hampered by lack of an effective severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antiviral therapy. Here we report the successful use of remdesivir in a patient with COVID-19 and the prototypic genetic antibody deciency X-linked agammaglobulinaemia (XLA). Despite evidence of complement activation and a robust T cell response, the patient developed persistent SARS-CoV-2 pneumonitis, without progressing to multi-organ involvement. His unusual clinical course identies a key role for SARS-CoV-2 antibodies in both viral clearance and progression to severe disease. In the absence of these confounders, we took an experimental medicine approach to examine the in vivoutility of remdesivir. Over two independent courses of treatment, we observed a dramatic, temporally correlated clinical and virological response, leading to clinical resolution and viral clearance, with no evidence of acquired drug resistance. We therefore provide unambiguous evidence for the antiviral ecacy of remdesivir in vivo, and its potential benet in selected patients.


Introduction
The prodrug nucleoside analogue remdesivir is a broad-spectrum antagonist of viral RNA-dependent RNA polymerase (RdRp) enzymes, leading to inhibition of SARS-CoV-2 replication in vitro 1, 2, 3 and pre-clinical bene t in a macaque model of COVID-19. 4 Two recent RCTs have tested the e cacy of remdesivir in patients. The rst was underpowered, and failed to show clinical bene t. 5 Preliminary data from the second showed a statistically signi cant reduction in illness duration, and a trend to reduced mortality. 6 No convincing evidence of virological e cacy was reported in either study.
Whilst RCTs provide the gold-standard for evaluation of the e cacy of new therapeutic interventions, comparing the average responses of patients in heterogeneous treatment and control groups may mask the potential bene ts for individual patients. Evaluation of therapeutics for COVID-19 is particularly complicated by the highly variable clinical course. Furthermore, as well as mediating clearance of SARS-CoV-2, the immune response may also contribute to severe COVID-19 pathology, independent of viral replication.
It is therefore unclear whether the limited response to remdesivir observed in RCTs re ects inadequate in vivo antiviral activity, or the need for concurrent immunomodulation. To minimise heterogeneity attributable to the immune response, we therefore take a reductionist, experimental medicine approach to evaluate the e cacy of remdesivir for treatment of COVID-19 in vivo, by studying a rare patient in whom the contribution of humoral (antibody-dependent) immunity to viral clearance and immunopathology is controlled genetically by the primary immunode ciency XLA.

Results
Uncomplicated persistent COVID-19 pneumonitis in a patient with XLA XLA is caused by mutations in the gene encoding Bruton's tyrosine kinase (BTK), leading to an absence of mature B lymphocytes and immunoglobulins (antibodies). The subject of this study is a 31 year old man with XLA, whose past medical history is summarised in the Methods. His illness began with fever, cough, nausea and vomiting. On day 19, a nasopharyngeal/throat swab was positive for SARS-CoV-2 RNA, and he commenced treatment with hydroxychloroquine and azithromycin (Supplementary Figure 1).
His symptoms persisted, and a repeat nasopharyngeal/throat swab on day 28 remained positive for SARS-CoV-2 RNA.
The patient was admitted to hospital on day 30 because of worsening dyspnea, and supplemental oxygen was commenced (Fig. 1a). Blood tests showed a rising CRP and mild lymphopaenia, elevated IL-6, IL-10, TNF-a and IFN-g, but normal D-Dimer and troponin, with no evidence of signi cant coagulopathy, renal or liver dysfunction ( Fig. 1 and Tables [1][2][3][4]. Blood and sputum cultures for bacteria remained negative throughout the course of the illness, procalcitonin was repeatedly normal, and two trials of broad-spectrum intravenous antibiotics had no improved effect (Supplementary Figure 1). A sputum sample remained positive for SARS-CoV-2 RNA, and a CT chest scan showed widespread patchy groundglass opacity in the lower lobes, consistent with COVID-19 pneumonitis (Fig. 1b).
Immunocompetent adults with severe COVID-19 typically exhibit a monophasic acute illness, with hospital admission and progressive respiratory failure 7-10 days after symptom onset 7,8 . In contrast, our patient exhibited a very unusual pattern of SARS-CoV-2 infection, with persistent fever and pneumonitis for >30 days, but without progression to acute respiratory distress syndrome (ARDS) or multi-organ involvement. This relatively stable baseline allowed the detailed assessment of clinical, virological and immune responses during two independent challenges with remdesivir.
Clinical response to remdesivir On day 34, hydroxychloroquine and azithromycin were discontinued, and the patient commenced a 10 day course of remdesivir. His fever and dyspnea improved within 36 hours of the rst dose, nausea and vomiting ceased, and rising oxygen saturation allowed discontinuation of supplemental oxygen. This dramatic clinical response was accompanied by a progressive decrease in CRP, a rise in total lymphocyte count, and an improvement in ground-glass opaci cation on repeat CT chest ( Fig. 1a-b). The patient was therefore discharged on day 43. 7 days after discharge, his fever, dyspnea and nausea returned. He was readmitted to hospital on day 54, and supplemental oxygen was commenced. A further sputum sample was positive for SARS-CoV-2 RNA, and a CT pulmonary angiogram showed evidence of ongoing pneumonitis ( Fig. 1a-b). Consistent with a recrudescence of COVID-19, the patient's CRP increased, and his lymphocyte count fell. On day 61, he therefore began treatment with a further 10 day course of remdesivir. Once again, his symptoms rapidly improved, his fever and requirement for supplemental oxygen resolved, and his CRP and lymphocyte count normalised (Fig. 1a).
The patient therefore exhibited a dramatic clinical response, tightly correlated with the administration of remdesivir, over two independent challenges. Because of his underlying immunode ciency, protracted illness, and relapse after the rst course of remdesivir, he was further treated with 2 units of convalescent plasma on days 69 and 70, to provide secondary prophylaxis against SARS-CoV-2 infection (Fig. 1a). He was discharged 3 days later, and has remained apyrexial and asymptomatic over a further 28 days of follow-up.
Virological response to remdesivir SARS-CoV-2 RNA was identi ed on nasopharyngeal/throat swabs early in the patient's illness, but became undetectable on these samples from day 36 (Fig. 1a). Because of his underlying bronchiectasis, the patient habitually expectorates small volumes of sputum. SARS-CoV-2 RNA was readily detectable in these samples until day 64, 4 days into his second remdesivir course, allowing non-invasive monitoring of his lower respiratory tract. Samples from blood, urine, faeces and a rectal swab were all negative.
Strikingly, levels of SARS-CoV-2 RNA in sputum fell progressively during the patient's rst course of remdesivir, corresponding with his clinical response (Fig. 1a). Nonetheless, SARS-CoV-2 RNA remained detectable at low levels. Upon cessation of remdesivir treatment, levels of SARS-CoV-2 RNA increased again in parallel with the recrudescence of symptoms. The effect of the second course of remdesivir was even more rapid and complete, with SARS-CoV-2 RNA becoming undetectable after 4 days (Fig. 1a). The patient therefore exhibited a dramatic virological response, tightly correlated with both the administration of remdesivir and resolution of his symptoms.
To con rm that the recurrent detection of SARS-CoV-2 RNA re ected viral persistence, rather than reinfection, isolates were sequenced regularly over the course of the patient's illness ( Fig. 1a and Supplementary Figure 2 and Table 5). All isolates belonged to the B2.6 lineage, a very uncommon lineage globally with only 47 isolates in total from the UK, making reinfection extremely unlikely. Compared with the Wuhan-Hu-1 reference sequence, 17 single nucleotide polymorphisms (SNPs) were present in at least one isolate (Supplementary Figure 2). Amongst these, the only SNP in the remdesivir target gene nsp12, encoding the RdRp, was a synonymous variant (C14805T), present at equivalent abundance across all isolates. Despite the combination of underlying immunode ciency, prolonged viral shedding and repetitive treatment, there was therefore no evidence of acquired remdesivir resistance.
Contribution of antibodies to clearance of SARS-CoV-2 XLA is the prototypic genetic disorder of the humoral immune system, and the most obvious explanation for our patient's failure to clear his infection spontaneously is his lack of antibodies to SARS-CoV-2. Patients with XLA are known to be at risk from persistent RNA viral infections, particularly chronic enteroviral meningoencephalitis 9 . This risk is mitigated by immunoglobulin replacement therapy, including antibodies to enteroviruses. 10 Conversely, available pooled immunoglobulin preparations antedate the COVID-19 pandemic, and lack speci c SARS-CoV-2 antibodies.
We therefore assessed the humoral immune response to SARS-CoV-2 in our patient. As expected, despite maintenance of regular immunoglobulin replacement, antibodies to SARS-CoV-2 spike and nucleocapsid proteins were undetectable by immunoassay prior to treatment with convalescent plasma (Fig. 2a-b), and no neutralisation activity was observed against lentiviral particles pseudotyped with SARS-CoV-2 spike protein (Fig. 2c). Following administration of convalescent plasma, antibody levels and neutralisation activity rose commensurately (Fig. 2a-c).
Together with CD8+ T cells, antibodies are known to contribute to the control of other RNA viruses, including HIV 11 , Hepatitis C 12 and LCMV infection of mice 13,14 . Two recent reports of patients with XLA and COVID-19 15, 16 did not describe persistent clinical disease. This suggests that, similar to the LCMV model, CD8+ T cell immunity can sometimes compensate for humoral de ciency in the control SARS-CoV-2. We therefore assessed the antigen-speci c CD8+ T cell response to SARS-CoV-2 in our patient.
Immunodominant CD8+ T cell epitopes of SARS-CoV-2 17 and other coronaviruses 18,19 are contained within the spike protein. We rst measured the quality of the CD8+ T cell response by ow cytometry for antigen-stimulated effector protein expression 20 (Fig. 3a). The frequency of spike-speci c CD8+ T cells within the circulation was comparable to age-matched health care workers with acute COVID-19, trending upwards over the course of infection. Increasing polyfunctional effector capacity was evident from day 62 (Fig. 3b), particularly the appearance of TNF-a and IFN-g producing CD8+ T cells (Supplementary Figure 3).
We further assessed the coverage of the patient's antigen-speci c response by measuring CD8+ T cell proliferation to peptide pools covering the S1, S2, M, N, ORF3, ORF6, ORF8 and ORF7 SARS-CoV-2 proteins, following viral clearance. At days 77 and 83, the patient had striking responses to all viral peptides relative to age-matched COVID-19 infected HCWs ( Fig. 3c-d), likely re ecting the prolonged duration of antigen exposure. Taken together, these data con rm the presence of a robust CD8+ T cell response to SARS-CoV-2. Whilst insu cient to resolve the infection spontaneously, this likely contributed to the clearance of virus during the second course of remdesivir.

Role of antibodies in immunopathogenesis of COVID-19
As well as their role in the anti-viral response, antibodies have the potential to cause immune pathology 21 . In a retrospective study of patients with SARS-CoV, the development of neutralising antibodies early in infection correlated with worse disease outcome 22 , and high antibody titres are associated with severe disease in COVID-19 23 . Our patient's lack of SARS-CoV-2 antibodies may therefore explain why, despite his persistent infection, he did not progress to ARDS or multi-organ involvement.
Antigen-antibody complexes are able to activate the classical complement pathway, with release of the anaphylatoxins C3a and C5a and formation of the multi-subunit terminal complement complex (TCC).
C3-de ciency reduces lung pathology in a mouse model of SARS-CoV 24 , complement activation is a feature of COVID-19 25 , and deposition of TCC is observed in the microvasculature of patients with severe disease 26 .
We therefore assessed the level of complement activation in our patient. Despite the lack of disease progression, levels of C3a C3c, C5a and TCC were markedly elevated, comparable to levels seen in COVID-19 patients admitted to the intensive care unit (Fig. 4a-d). Whilst complement activation correlates with and may be required for the development of ARDS and/or multi-organ failure, complement activation alone is therefore insu cient for disease progression, and the classical pathway is not required for complement activation in COVID-19.

Discussion
Taken together, our observations con rm that remdesivir is a potent antiviral agent for treating SARS-CoV-2 infection in vivo. It is striking that previous reports from immunocompetent patients have generally not demonstrated the profound clinical and virological responses observed in this study. This is likely because, amongst the hospitalised patients included in RCTs 5, 6 or compassionate use programmes 27 to date, progression or resolution of disease has been determined not by the level of SARS-CoV-2 replication, but by the evolution of immune pathology. Whilst remdesivir was given chronologically late in our patient's disease, he had not developed ARDS or multi-organ involvement. The utility of remdesivir in immunocompetent patients may therefore be maximised by early treatment in the most at risk, or by combination with targeted immunomodulatory therapies in the most sick.
The persistent disease observed in this patient strongly suggests that antibodies contribute to the control of SARS-CoV-2. Conversely, the lack of disease progression suggests that they also contribute to immune pathology in COVID-19. Aside from activating complement, antibodies may interact with Fc receptors on immune cells to drive macrophage activation and in ammatory cytokine production 28 . Strategies which target these interactions include plasmapheresis, or blockade of Fc receptors by saturating doses of intravenous immunoglobulin (IVIG). The potential for both bene cial and deleterious effects of SARS-CoV-2 antibodies revealed by our study suggests that these interventions, like the administration of convalescent plasma, should be tailored to individuals or subgroups of patients with distinct clinical characteristics.

Oversight
The study was approved by the East of England -Cambridge South national institutional ethics review board (17/EE/0025). The patient provided written informed consent. Additional healthy controls, patients and healthcare workers with COVID-19 were enrolled to the NIHR BioResource Centre Cambridge (17/EE/0025) and the Oxford Gastrointestinal Illness Biobank (16/YH/0247).

Patient characteristics
The subject of this study is a 31 year old man, born in the UK to parents of Pakistani origin, and diagnosed with XLA at the age of 12. He has a family history of XLA affecting his brother, 4 maternal uncles and a male maternal cousin. Following investigation for recurrent chest infections, he was found to lack mature B cells and circulating immunoglobulins, and con rmed to have inherited the familial c. 1430delT mutation in the gene encoding BTK. Because of prior incomplete adherence to replacement immunoglobulin therapy, he developed middle lobe bronchiectasis.

Remdesivir treatment
The rst course of remdesivir was administered as part of a Gilead SIMPLE study (NCT04292899), the second course was provided as part of the Gilead Expanded Access Program (NCT04323761). In each case, the patient received an initial dose of 200mg IV, followed by 9 daily doses of 100mg IV (10 days total).

Convalescent plasma
Convalescent plasma was collected from individuals with previous, laboratory-con rmed SARS-CoV2infection at least 28 days after resolution of symptoms using established infrastructure and standard UK donor selection guidelines, as previously described 29 . Signed consent was obtained from each donor at the time of donation using NHS Blood and Transplant-approved consent forms, in accordance with Blood Safety and Quality Regulations enforced by the Medicines & Healthcare products Regulatory Agency.
In brief, a total of at least 540ml of plasma (containing <1x10 6 leucocytes per component) was collected via plasmapheresis from each donor, divided into two units, rapidly frozen and stored at -25°C. Donor blood samples were tested for SARS-CoV-2 RNA (Public Health England) and antibodies (EUROIMMUN (IgG) assay, PerkinElmer, London, UK). A signal to cut-off (S/CO) ratio of 9.1 in the EUROIMMUN assay was previously shown to identify donations with a neutralising antibody titre of >/=1:100 in a SARS-CoV-2 (isolate England/2020) microneutralisation assay with a speci city of 100%. 29 Both donations used in this study were collected in May 2020 and con rmed to be negative for SARS-CoV-2 RNA. EUROIMMUN S/CO ratios were 37.171 ( rst unit, 290ml, administered on day 69) and 7.271 (second unit, 283ml, administered on day 70), respectively. For the second unit (lower S/CO ratio), the neutralising antibody titre was con rmed to be >/=1:100 using the SARS-CoV-2 microneutralisation assay. 29 Plasma was defrosted using a 37˚C waterbath, and transfused within 4 hours of defrosting.

Patient sampling
Upper respiratory tract samples (nasopharyngeal/throat swabs) were collected in viral transport medium according to Public Health England (PHE) guidelines. Lower respiratory samples (sputum) were collected in universal containers and extracted following mucolysis with Mucolyse PL.701 sputum liquefying agent (Pro-Lab Diagnostics). To minimize RNA degradation, the time between sample collection and processing did not exceed 2 hr.
Peripheral blood mononuclear cells (PBMCs) were isolated using Leucosep tubes (Greiner Bio-One) with Histopaque-1077 (Sigma) by centrifugation at 800g for 15 mins at room temperature. PBMCs at the interface were collected and rinsed twice with autoMACS running buffer (Miltenyi Biotech). All samples were processed within 4 hr of venepuncture. Samples were then resuspended in freezing medium (FBS 90% supplemented with DMSO 10%), frozen overnight at -80°C in a Mr Frosty cell freezing containger (Nalgene) then transferred to liquid nitrogen for long term storage.

SARS-CoV-2 molecular testing
Sample testing was carried out using an in-house real time uniplex RT-PCR diagnostic assay for the detection of SARS-CoV-2 in the PHE Clinical Microbiology and Public Health Laboratory at Addenbrooke's Hospital, Cambridge. This assay targets a 222 base-pair region of the SARS-CoV-2 nsp12 gene (encoding RdRp), and has been validated for clinical use.
In brief, nucleic acid extraction was undertaken using the NUCLISENS easyMAG platform (Biomerieux, Marcy L-Etoile), in accordance with the manufacturer's instructions. Nucleic acids were extracted from 500μL sample, with a dilution of MS2 bacteriophage (4600pfu per extraction) added pre-extraction to act as an internal extraction and inhibition control. Molecular grade water was used as a negative control. Positive control material, BetaCoV/England/02/2020, was obtained from PHE Colindale (essentially, puri ed viral RNA diluted to give a cycle threshold (CT) value of 26-28).
The RdRp gene was detected using primers ATGGGTTGGGATTATCCTAAATGTGA and AGCAGTTGTGGCATCTCCTGATGAG with a FAM-labelled MGB RdRp Probe ATGCTTAGAATTATGGCCTCAC. The internal extraction control was detected using the MS2 forward primer TGGCACTACCCCTCTCCGTATTCACG, the MS2 reverse primer GTACGGGCGACCCCACGATGAC and a ROX-BHQ2 labelled MS2 probe CACATCGATAGATCAAGGTGCCTACAAGC. Ampli cation reactions and detection of PCR products were performed using the RotorgeneTM PCR instrument. A typical reaction contained 400 nM of forward and reverse primers for the RdRp gene and 200nM of the MS2 internal control forward and reverse primer pair, along with 120nM of the RdRp and MS2 probes. The cycle conditions were as follows: 25°C 2 mins, 50°C 15 mins, 95°C 2 mins followed by 45 cycles of 95°C and 60°C. Samples that generated a CT value ≤36, de ned as 0.01 uorescence units as per the RotorgeneTM manufacturer's instructions, were considered positive (roughly equivalent to 8 genome copies, the lower limit of detection for the assay). Where indicated, additional samples were tested in the clinical laboratory at the Royal London Hospital, London.

Viral sequencing and bioinformatics
Samples were analysed by Nanopore sequencing as part of the COG-UK sequencing project, following the ARTICnetwork V3 protocol (https://dx.doi.org/10.17504/protocols.io.bbmuik6w), and assembled using the ARTICnetwork assembly pipeline (https://artic.network/ncov-2019/ncov2019-bioinformaticssop.html). The accession numbers of the sequences included in this study are available in Table 5. Median genome depth of coverage was 1,346x. Consensus FASTA sequences were analysed after QC ltering, de-duplication and matching with metadata. Variants were initially assessed using the ARTICnetwork assembly pipeline VCF output les, with a SNP being called when >50.1% frequency at a single base. These were then independently con rmed using iVar analysis (https://andersenlab.github.io/ivar/html), set to a quality threshold of Q20 and a minimum frequency of 10%, with a SNP being called when >50.1% of the reads at a particular position differed from the reference sequence, Wuhan-Hu-1 (GenBank Accession MN908947.3). The location of each SNP was examined to identify mutations in the nsp12 gene, encoding RdRp (the target of remdesivir treatment). For the kinetic assessment, proportions of each variant as a fraction of all reads at their corresponding locations (minimum depth 20x) were plotted as a proportional stacked area chart using ggplot2 and RStudio.

Serological Assessment
Serological reactivity to SARS-CoV-2 spike and nucleocapsid proteins Recombinant SARS-CoV-2 nucleocapsid and spike proteins were covalently coupled to distinctive carboxylated bead sets (Luminex; Netherlands) to form a multiplex assay. For protein coupling, beads were rst activated with 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride (Thermo Fisher Scienti c) in the presence of N-hydroxysuccinimide (Thermo Fisher Scienti c), according to the manufacturer's instructions, to form amine-reactive intermediates. The activated bead sets were incubated with the corresponding proteins at a concentration of 50ug/ml in the reaction mixture for 3 hr at room temperature on a rotator. Beads were washed and stored in a blocking buffer (10mM PBS, 1% BSA, 0.05% NaN3).
Coupled bead sets were incubated with patient or control sera at a dilution of 1/100 for 1 hr in 96-well lter plates (MultiScreenHTS; Millipore) at room temperature in the dark on a horizontal shaker. Fluids were aspirated with a vacuum manifold and beads were washed three times with 10mM PBS/0.05% Tween-20. Beads were incubated for 30 mins with a PE-labeled anti-human IgG-Fc antibody (Leinco/Biotrend), washed as described above, and resuspended in 100μl PBS/Tween-20. They were then analyzed on a Luminex analyzer (Luminex / R&D Systems) using Exponent Software V31. Speci c binding was reported as mean uorescence intensities (MFI). Stored sera collected in the diagnostic immunology laboratory prior to November 2019 were used as healthy controls. Sera collected from patients with PCR-con rmed COVID-19 were used as positive controls.
Virus production: pseudotyped lentiviral stocks were generated through the co-transfection of HEK293T cells with a lentiviral expression vector plus the packaging plasmid pCMVΔR8.91 and either pMD.G (VSV-G) or pCG1-SARS-CoV-2 D19 spike plasmid, using TransIT-293 transfection reagent (Mirus) according to the manufacturer's recommendations. Viral supernatants were harvested 48 hr post-transfection and cell debris was removed with a 0.45-um lter prior to use or freezing.
Neutralisation assay: pseudotyed pHRSIN-re y luc-Puro R lentivirus supernatants were pre-incubated with diluted serum or plasma samples at 37°C for 60 mins, before addition of 1.25x10 4 HEK293T cells to each well. Transduction e ciencies were quanti ed 48 hr post-transduction by measuring re y luciferase activity in cell lysates using Bright-Gloä luciferase assay system (Promega) and read in a CLARIOstar luminometer (BMG). Data were plotted using GraphPad Prism 8 (non-linear regression, (Inhibitor) vs.
response --Variable slope, four parameters) and virus neutralising antibody titre was obtained by determining the dilution factor which gave a 50% reduction in re y luciferase levels, compared to the no serum control wells.

Complement assays
The complement components C3a, C3c and C5a were measured in plasma samples from patients with COVID-19 using solid-phase ELISA assays (Hycult Biotech) according to the manufacturer's instructions.
TCC was measured using an in-house ELISA with the anti-TCC neo-epitope antibody, ae1 as capture.

Data Availability
Viral sequence data is available from COVID-19 Genomics UK and the GSAID. The accession numbers are detailed in Table 5     Representative ow cytometry dot plots (c) and % proliferating CD8+ T cells (d) after stimulation ± peptide pools covering the indicated SARS-CoV-2 proteins. Patient samples are compared with 2 HCWs with PCR-con rmed COVID-19 at indicated time points.