COVID-19 relapse associated with SARS-CoV-2 evasion from CD4+ T-cell recognition in an agammaglobulinemia patient

Summary A 25-year-old patient with a primary immunodeficiency lacking immunoglobulin production experienced a relapse after a 239-day period of persistent severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Viral genetic sequencing demonstrated that SARS-CoV-2 had evolved during the infection period, with at least five mutations associated with host cellular immune recognition. Among them, the T32I mutation in ORF3a was found to evade recognition by CD4+ T cells. The virus found after relapse showed an increased proliferative capacity in vitro. SARS-CoV-2 may have evolved to evade recognition by CD4+ T cells and increased in its proliferative capacity during the persistent infection, likely leading to relapse. These mutations may further affect viral clearance in hosts with similar types of human leukocyte antigens. The early elimination of SARS-CoV-2 in immunocompromised patients is therefore important not only to improve the condition of patients but also to prevent the emergence of mutants that threaten public health.

SARS-CoV-2 persistently infected a man with humoral immunodeficiency for 239 days COVID-19 relapsed and SARS-CoV-2 evolved during the infection period The virus isolated after relapse had an increased proliferative capacity in vitro SARS-CoV-2 may have evolved to evade recognition by CD4 + T-cells INTRODUCTION X-linked agammaglobulinemia (XLA) is a rare primary humoral immunodeficiency that was first reported by Bruton in 1952. In XLA, abnormalities in Bruton's tyrosine kinase prevent normal B-cell maturation, resulting in agammaglobulinemia. 1 The prognosis is generally good if globulin preparations are administered regularly. 1 On approximately day 235, the general condition of the patient worsened significantly. He presented chills, shivering, malaise, and uncontrollably high fever (approximately 40 C, treated with 1,800 mg of acetaminophen per day). Blood testing showed elevated levels of CRP and interleukin (IL)-6, and CT showed worsening pneumonia. The qRT-PCR test for SARS-CoV-2 on day 239 showed a decreased Ct value. The multiplex PCR testing of sputum samples for 19 respiratory tract infections (BioFire FilmArray Respiratory Panel 2.1) on the same day detected SARS-CoV-2 but no other viruses or bacteria causing respiratory tract infection (Table S2). The patient was readmitted to Osaka City General Hospital and treated with remdesivir from day 239. Remdesivir was administered at 200 mg/day on the first day and 100 mg/day from the second day onwards for 10 days in total, based on methods described in previous studies. 3,10 On the second day of remdesivir administration (day 240), the fever subsided, and the chills, shivering, and malaise disappeared. Blood testing showed rapid decreases in the levels of CRP and IL-6 ( Figure 1A, Table S3), and imaging tests showed the resolution of pneumonia ( Figure 1B). The qRT-PCR testing of sputum samples for SARS-CoV-2 showed elevated Ct values, and the patient tested negative for the virus on day 243 ( Figure 1A, Table S4). No SARS-CoV-2 was isolated from the sputum samples after the administration of remdesivir. The patient was discharged on day 248 after all of his symptoms had been resolved and the blood and imaging examination results had improved. No flare-ups have occurred for 9 months since his discharge, indicating that the treatment was successful. The patient was given intravenous immunoglobulin (IVIG) injections every month ( Figure 1A, Table S5) and was not vaccinated against SARS-CoV-2 during the entire course of this study. Tests for the anti-SARS-CoV-2 spike (S) protein in the plasma and IVIG lots administered to the patient during the study period yielded negative results ( Figure S1).

Viral genome and proliferative capacity analysis
Infectious viruses were isolated from cryopreserved specimens from a nasopharyngeal swab (NPS) on day 1, bronchoalveolar lavage fluid (BALF) and lung tissue on day 218, sputum on day 224, and fresh sputum on day 239 (Table S4). Whole-genome sequences were obtained from the five isolates and day-1 NPS specimen. A phylogenetic analysis showed that all of the viruses belonged to the clade B.1.1.284, which was a prevalent clade in Japan during the period in which the patient was infected. The isolated day-1 virus showed the same consensus sequence as the day-1 NPS specimen, and the other viruses formed a distinct cluster with the day-1 viruses ( Figure 2A, Table S6). It should be noted that most of the viruses in Japan between days 218-239, during which the mutated viruses in the patient were isolated, belonged to clade B.1.1.7. In contrast, the viruses isolated from the patient belonged to B.1.1.284, which nearly disappeared between days 218-239 in Japan. [15][16][17] These results demonstrate that the worsening of the condition of the patient after day 235 was due to a relapse of COVID-19, ruling out the possibility of reinfection with another virus. Nearly two-thirds of the mutations were non-synonymous (Table S7) iScience Article nucleotide substitutions were commonly observed in all of the other isolates ( Figure 2B). An additional two, one, three, and five amino acid changes were found in the isolates from the BALF and lung tissue from day 218 and those from days 224 and 239, respectively. Apart from the K3353R mutation in ORF1ab, these mutations have rarely been found in the CoV-GLUE database 18 (Table S7), indicating a unique evolution of SARS-CoV-2 in this patient. These mutations were unevenly distributed in the viral genome, and the  iScience Article average numbers of mutations per 1,000 nucleotides in these viruses exceeded 1 in the regions of E (6.67), M (3.00), and ORF7a (6.89), whereas those in ORF1a (0.454), ORF1b (0.278), S (0.720), and ORF3a (0.303) were less than 1. There were no mutations in the ORF6, ORF8, and N regions. The evolution rate of SARS-CoV-2 in this patient was higher (9.37 3 10 À4 /site/year) than that reported in general populations (3.95-6.67 3 10 À4 /site/year) ( Table 1). 19- 23 We calculated the SARS-CoV-2 evolution rate based on published sequence data from other immunocompromised patients 2,4,10,12,13,23,24 and confirmed that SARS-CoV-2 evolves faster in immunocompromised hosts than in the general population ( Table 1).
The proliferative capacities of the isolated viruses were evaluated using a focus-forming assay. Passage #1 viruses isolated from the NPS specimen on day 1, BALF and lung tissue on day 218, and sputum on days 224 and 239 were inoculated in VeroE6/TMPRSS2 cells, which were seeded in 24-well plates, at infection multiplicities of 0.01 and 0.001. The focus sizes of the viruses were compared 1 day after infection ( Figure 2C). The viruses isolated from the day-1 NPS specimen had small foci (mean size, 45.5 G 19.04 3 10 À3 mm 2 ), whereas those isolated from the sputum on day 239 had large foci (mean size, 102.8 G 78.61 3 10 À3 mm 2 ). Although the viruses isolated from the lung tissue on day 218 formed small foci (mean size, 34.2 G 27.26 3 10 À3 mm 2 ), those isolated from the BALF on day 218 formed both large (mean size, 101.3 G 55.72 3 10 À3 mm 2 ) and small foci (mean size, 20.0 G 9.82 3 10 À3 mm 2 ), suggesting that these isolates comprised a mixture of more than one viral clone. The viruses isolated from the sputum on day 224 formed medium-sized foci (mean size, 73.2 G 44.06 3 10 À3 mm 2 ). Although several factors, including the binding and destruction of the receptor of the viruses, can affect focus or plaque sizes, 25,26 it is generally believed that the sizes of foci are correlated with the proliferative capacities of viruses. These results suggest that SARS-CoV-2 evolved with an increased proliferative capacity in this patient.

The T32I mutation in ORF3a evades T-cell recognition
To profile the immune response against SARS-CoV-2 in the patient, the T-cell response against SARS-CoV-2 antigens was first examined. Peripheral blood mononuclear cells (PBMCs) were stimulated with peptide pools derived from proteins of the SARS-CoV-2 Wuhan-Hu-1 strain (Wu) in an ELISPOT assay ( Figure 3A). Interferon (IFN)-g-expressing cells were mainly observed when the PBMCs were stimulated with structural spike (S), membrane (M), nucleocapsid (N) proteins, accessory factors (ORF3a and ORF7a), and a non-structural protein (NSP3), suggesting that T-cell epitopes were mostly located in these proteins. These epitope iScience Article hotspots included almost half of the mutations detected in the patient, implying that they were selected under pressure from the host T-cell response.
To further investigate the pivotal mutations that aggravated the infection, the T-cell response to single epitopes covering the mutated regions was examined. PBMCs from the patient were expanded with peptides from Wu that covered the mutated regions ( Figure 3A) before they were re-stimulated with the same or corresponding mutant (mut) peptides (Table S8). The S1145, ORF1ab2973, and ORF3a25 Wu peptides induced greater IFN-g expression than their corresponding mutant peptides ( Figure 3B). In addition, a decreased expression of IFN-g induced by ORF1ab3061 mut and ORF7a105 T115I mut peptides was repeatedly observed in three experiments, indicating that mutations in these five regions were likely involved in the immune escape of the viral variants.
To further elucidate the characteristics of the T cells targeting these critical regions, T cells were sorted after being stimulated with S1145, ORF1ab2973, ORF3a25, or M121 Wu peptides, pooled, and analyzed using single-cell T-cell receptor (TCR)-and RNA-sequencing analyses ( Figure 4A). The TCR sequencing results showed that some of the clonotypes had higher frequencies after the in vitro stimulations within both the CD4 + and CD8 + clusters ( Figure 4B). Within the top-15 frequent T-cell clones identified in the single-cell analyses, 14 clones were located in clusters reflecting CD8 + CTL-like cells ( Figure 4C). The remaining clone-111 comprised CD4 + cells ( Figure 4C) and was predominantly from ORF3a25-stimulated cells ( Figure 4D), suggesting that this clone was preferentially expanded by the stimulation of ORF3a25. All of the top 15 frequent clones, except clone-111, were detected in the PBMCs of the patient before the in vitro stimulation ( Figure 4E), suggesting that clone-111 had a low frequency before the stimulation and became dominant afterward. Consistent with this, the ORF3a25 Wu peptide was found to activate the in vitro-expanded CD4 + T cells upon re-stimulation ( Figure 5A). To confirm the epitope of clone-111, TCR-a and-b of clone-111 were reconstituted in a TCR-deficient reporter cell line. The reporter cells were activated by the ORF3a25 Wu peptide in the presence of autologous antigen-presenting cells, but  iScience Article the T32I mutant peptide lost antigenicity ( Figure 5B). Of the human leukocyte antigen class II genes in the patient, DRB1*04:06 was identified as a restricting allele for ORF3a epitope recognition ( Figure 5C). These results indicate that the T32I mutation in ORF3a resulted in the evasion of T-cell recognition.
Clone-111, which was detected as 28 different barcoded cells in a CD4 + cluster ( Figure 4C), showed the gene signatures of highly activated cells ( Figure 5D). These cells also expressed genes related to heterogeneous functions, such as IFN-g, IL-21, PRF1, and IL-10 ( Figure 5D). Notably, clone-111 was located in the cluster with a high expression of T regulatory (Treg) signature genes ( Figure 5E), 27 although most of the clone-111 cells did not highly express FOXP3. These observations suggest the helper function, cytotoxicity, and regulatory activity of this clonotype during infection. 20, [27][28][29] The finding that SARS-CoV-2 escaped the CD4 + clonotype with multiple potentials in the absence of humoral immunity further suggests a previously unclear role of CD4 + cells in the immune protection against COVID-19.

DISCUSSION
We evaluated a patient with XLA in whom SARS-CoV-2 evolved uniquely for 239 days. During the whole infection period, 20 non-synonymous mutations were acquired by the virus, and many of these mutations were previously unreported (GISAID, https://www.gisaid.org). Among them, the ORF3a T32I mutation evaded recognition by CD4 + T cells with multi-functional characteristics, such as cytotoxic and regulatory function. Indeed, infection was controlled for 8 months when the original ORF3a epitope (ORF3a T32) was preserved, suggesting that T cells recognizing this epitope have a potent protective function even in the absence of humoral immunity. Consistently, when the mutation was acquired on day 239, the patient suffered severe deterioration with drastically increased levels of IL-6 and CRP and the sputum viral load. Thus, iScience Article this variant is considered to acquire escape capacity from protective T cells, which results in viral expansion and the subsequent exacerbation of pneumonia in the patient.
The patient was given immunoglobulin every month because no normal mature B cells were present in his peripheral blood. The globulin products administered during the observation period were manufactured from blood collected before the emergence of COVID-19 (Table S5). Therefore, no SARS-CoV-2-specific antibodies were present in the serum of the patient ( Figure S1). This case illustrates the pathophysiology of SARS-CoV-2 infection in the absence of specific humoral immunity.
Several precedents for SARS-CoV-2 evolution in immunocompromised patients with partially attenuated humoral responses have been described. 4,12,24,30-33 SARS-CoV-2 evolved faster in these patients (Table 1), and its mutations showed a certain degree of similarity to those in several variants of concern, such as B.  iScience Article have rarely been reported, except for the K3353R mutation in ORF1ab, possibly due to a near-complete lack of humoral immune responses.
In the absence of humoral immunity, cellular immunity should exert protective effects against pathogens. The 28 CD4 + T cells from clone-111 had the same TCRs but showed the characteristics of different cell types, such as T follicular helper (Tfh), CD4 CTL, and Treg cells (Figures 5D and 5E). 20, [27][28][29] These results suggest that this CD4 + T-cell clonotype may exhibit functional plasticity and balance the T-cell responses in the protective immunity against SARS-CoV-2 infection. It has been reported that epitope affinity may affect the fate of T-cell clonotypes. 34,35 However, in the case evaluated in our study, the development of heterogeneous T-cell types did not seem to be restricted by the epitope.
In our patient, CD4-mediated responses were likely to be dominant, because, in addition to ORF3a, four more mutations (ORF1ab F2979L in NSP4, ORF1ab L3829F in NSP6, K1149T in S, and T1155I in ORF7a) escaped from CD4 + T-cell recognition, whereas no mutations lost antigenicity to CD8 + T cells ( Figure S2). Several mutant peptides, such as M1, ORF1ab2509, and ORF1ab3061, activated more CD4 + T cells than the corresponding Wu peptides did. These mutations may have occurred due to viral fitness, which warrants further investigation. The reason for the dominance of CD4-mediated responses is unclear at present. Nevertheless, it is known that MHC class II deficiency, but not MHC class I deficiency, incapacitates the immune system during viral infection, 36 which suggests an indispensable role of CD4 + T cells in antiviral immunity. Therefore, the identification of de novo mutations that avoid T-cell recognition in the absence of protective antibodies may help to predict future ''T-escaping'' variants when broad neutralization antibodies prevail in developing vaccines. 37,38 The ORF3a protein is unique for the coronavirus subgenus Sarbecovirus, which includes SARS-CoV and SARS-CoV-2. ORF3a works as a viroporin, assisting the entry and release of the virus. 39 Because ORF3a has multiple roles in the life cycle of SARS-CoV-2 and the induction of host immune responses, it is hard to predict the influence of the T32I mutation in ORF3a. Nevertheless, given that we observed a deteriorated infection after the occurrence of this mutation, we suspect that this mutation does not disrupt or even benefit ORF3a function. The replicative speed of this mutant virus isolated on day 239 was higher than that of the day-1 virus without the ORF3a mutation.
As shown in the results in Figure 2C, the day-1 virus formed small foci, whereas the day-239 virus formed large foci. Like the day-1 virus, the day-218 lung tissue virus formed small foci, but the day-218-BALF virus formed both small and large foci. The day-239 and day-218 BALF viruses shared ORF1ab I6548V and S G1251R mutations, both of which were absent in the other viruses. These mutations were detected in approximately half of the day-218-BALF virus population and nearly all of the day-239 virus population (Figure 2B). These results suggest that ORF1ab I6548V and/or S G1251R mutations may have been involved in the increased proliferative ability of the virus. This is because the S protein plays an important role in viral entry into cells, 40 and ORF1ab 6548 is located at NSP15, which is an endoribonuclease important for viral replication. 41 The administration of remdesivir dramatically improved the symptoms, inflammatory markers, and imaging test results of the patient, and the shedding of infectious viral particles stopped immediately. Remdesivir induced strong antiviral activity in the patient, which was consistent with results from a previous report. SARS-CoV-2 may therefore be eliminated by remdesivir alone without SARS-CoV-2-specific antibodies. 42 In conclusion, SARS-CoV-2 evolved to evade recognition by CD4 + T cells and increased in its proliferative capacity during the persistent infection in our reported patient, which raises the possibility of the emergence of such variants in future even in immunocompetent patients, particularly when the escape from neutralizing antibodies is not sufficient for immune evasion. Thus, the control of chronic SARS-CoV-2 infection would also be beneficial for the prevention of virulent variants that evade such multi-functional T cells.
In particular, immunocompromised patients should be treated with available antiviral drugs to prevent the emergence of public health-threatening mutants. We recently reported that non-spike epitopes recognized by CD8 + T cells contribute to the protection. 43 Given the importance of multi-functional CD4 + T cells for protection in this study, the epitopes of such CD4 + T cells can be promising candidate antigens for the next-generation vaccines, because these epitopes are conserved in currently-reported VOCs. iScience Article Extensive investigation of variants and T cell responses is therefore a crucial measure for the sustainable control of SARS-CoV-2 worldwide.

Limitations of the study
Despite the importance of our findings, there were several limitations in this study. These included the number of cases, because we could not investigate other patients with XLA who were also infected with SARS-CoV-2 as such cases are rare. Nevertheless, we were able to show the accelerated evolution of SARS-CoV-2 in the case evaluated herein as well as in other reported immunocompromised cases. In addition, the clinical specimens, especially the PBMCs that were used for the T-cell analysis, were mainly collected after day 210, and we failed to obtain evidence for the immune evasion of some of the mutations that were identified. We were, however, able to identify ade novomutation that evaded the recognition of dominant T cells in the patient with XLA.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We are grateful to Dr. Ko Iida and Dr. Hidenori Nakagawa for their helpful discussions and comments. We are grateful to Dr. Charles Schutt for helpful discussion and editing this manuscript. We are grateful to Dr. Naoki Takahashi for providing technical assistance. We are profoundly grateful to the patient and his family for their kindness and generosity. This work was supported by funding from the Japan Agency for Medical Research and Development (JP20nf0101623 and JP20nk0101602). We gratefully acknowledge the contributions of the nurses and all of those who participated in caring for the patient. We also thank Editage (www.editage.jp) for editing this manuscript.

COVID-19 patient
The patient, a 25-year-old man, experienced repeated bacterial infections from approximately 8 months of age and was genetically diagnosed with X-linked agammaglobulinemia (XLA) due to an exon 6 skip mutation at 11 months. He has since been given immunoglobulin every month (Table S5) but no prophylactic antibiotics. He developed fever at the end of July 2020 and was diagnosed with COVID-19 based on a positive result from a quantitative real-time reverse transcription polymerase chain reaction (qRT-PCR) test for SARS-CoV-2 in early August.

Cell lines
VeroE6/TMPRSS2 cells were obtained from the JCRB cell bank (cat. JCRB1819). The cells were maintained at 37 C and with 5% CO 2 in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum, 1 mM L-glutamine, 100 units/mL penicillin, and 100 mg/mL streptomycin. Mouse T-cell hybridoma with a nuclear factor of activated T cells (NFAT)-green fluorescent protein (GFP) reporter gene 44 were maintained at 37 C and with 5% CO 2 in RPMI 1640 medium supplemented with 10% fetal bovine serum, penicillin, streptomycin, and 2-mercaptoethanol.

Sample preparation for PCR testing and viral infectivity studies
Fresh samples (obtained within 24 h of sampling) and those that had been cryopreserved at 80 C were obtained from a nasopharyngeal swab (NPS), bronchoalveolar lavage fluid, sputum and saliva, lung tissue obtained by a biopsy, and suction phlegm and used for PCR tests and viral infectivity studies. Before extracting the RNA, the sputum and saliva samples were diluted two-or four-fold using phosphate-buffered saline (PBS). A lung tissue sample was homogenized by adding a four-fold volume of PBS.

PCR test
A qRT-PCR was performed as previously described, 54  In-house ELISA of the anti-SARS-CoV-2 spike (S) protein antibody A 96-well flat-bottom microplate was coated with 0.1 mg/well of the S protein (Sino Biological) in 50 mL/well of carbonate-bicarbonate buffer (Sigma-Aldrich) and incubated at 4 C overnight. The plate was washed with 0.05% Tween 20 in PBS (PBS-T) and blocked with 300 mL/well of 25% BlockAce for 2hat 37 C. After washing the plate with PBS-T, 50 mL/well of diluted patient plasma (1:100), control sera, and diluted 10% ll OPEN ACCESS

Viral isolation
Viral isolation from the clinical specimens was performed as previously described. 54 In brief, VeroE6/ TMPRSS2 cells were seeded to 24-well plates (2 3 10 5 cells/well) 1 day before the experiment. The medium was removed, and 100 mL of the samples were added to each well. After 30 min, the samples were removed, and the cells were washed three times with serum-free MEM. The cells were then cultured in 1 mL of DMEM containing 2% fetal bovine serum, 200 units/mL of penicillin, 200 mg/mL of streptomycin, and 0.25 mg/mL of amphotericin B in an atmosphere with 5% CO 2 and at 37 C for up to 5 days. Cytopathic effects (CPEs) were macroscopically observed each day. When CPEs appeared, the culture supernatant was collected (passage #0). The isolated viruses were passaged once (passage #1) and titrated using focus-forming assays.

Viral sequencing and phylogenetic analysis
For viral sequencing, RNA was extracted from the culture supernatant of the isolated viruses (passage #0) using the Qiagen Viral RNA Mini kit (QIAGEN). cDNA synthesis was performed using the GenNext RamDAseq Single Cell Kit (TOYOBO). An Illumina library was constructed using the Nextera XT DNA Library Prep Kit (Illumina) and Nextera XT Index Kit v2 (Illumina). The resulting DNA libraries were sequenced using the NovaSeq6000 instrument (Illumina). The obtained reads were analyzed using Cutadapt to trim adapters; bwa to map the reads to the reference genome (NC_045512.2); Mutect2 to call mutations; snpEff to make annotations; and HaplotypeCaller and bcftools to make consensus sequences.
A phylogenetic tree was inferred from the alignment using the maximum likelihood approach in W-IQ-TREE. 51 The best-fit model was selected by ModelFinder, 52 and the ultrafast bootstrap 53 with 1,000 replicates was calculated. The viruses used for the phylogenetic analyses have been listed in Table S3. We used BLAST to obtain similar sequences to the virus from the patient using the NCBI database. We picked 17 sequences based on the highest identity score. As references, we added 8 sequences from various lineages. We then added 4 sequences from the same hospital and 16 sequences from the same prefecture iScience Article Bulk TCR sequencing and analysis A total of 3 x 10 5 fresh PBMCs were lysed in 200 mL of QIAzol (QIAGEN). Full-length cDNA was synthesized using SMARTer (Takara Bio) before variable regions of the TCR-a and-b genes were amplified. Each pair of reads was assigned a clonotype (defined as TR(A/B)V and TR(A/B)J genes and the complementarity-determining region 3) using MiXCR software. 70 Monocyte-derived dendritic cells CD14 + cells were isolated from PBMCs using human CD14 MicroBeads (Miltenyi Biotec) and cultured in RPMI 1640 medium supplemented with 5% human AB serum, 0.1 mM Non-Essential Amino Acid Solution (Gibco-BRL), 1 mM sodium pyruvate (Gibco-BRL), 10 ng/mL of human granulocyte-macrophage colonystimulating factor (GM-CSF) (Peprotech), and 10 ng/mL of IL-4 (Peprotech) for 13 days.

QUANTIFICATION AND STATISTICAL ANALYSIS
The cytokines were measured in duplicate. The data are shown as mean G standard deviation.